Extended Lessons Learned and Evaluations Integrated ...€¦ · 1.1.2 Challenges for big data...

Project ID 604674 FITMAN – Future Internet Technologies for MANufacturing

30/09/2015 Integrated Deliverable D12.5 – D13.5 – D14.5

D12.5 – D13.5 – D14.5

Extended Lessons Learned and Evaluations

Integrated Deliverable

Document Owner: Thomas Fischer (DITF)

Contributors: Eva Coscia, Silvia Crippa, Jacopo Cassina (Holonix), Toni Ventura (Datapixel), Aitor

Romero (Datapixel), Karl Hribernik, Marco Franke (BIBA), Silke Balzert, Jan Sutter

(DFKI), Konrad Pfleiderer (DITF), Ioan Toma, Benjamin Hiltpolt (STI), Sonja.Pajkovska-

Goceva (COMPlus), Gash Bullar (TANET), June Sola (Innovalia), Mirla Ferreira

(CONSULGAL), Marek Eichler (VW), Javier Martinez (AIDIMA), Roberto Sanguini

(AW), Nenad Stojanovic (FZI)

Dissemination: CONFIDENTIAL

Contributing to: WP12-13-14 Extended lessons learned and evaluations (final)

Date: 14/11/2015

Revision: 1.1



FITMAN Consortium Dissemination: CONFIDENTIAL 2/93

VERSION HISTORY

Version Date notes and comments

0.1 24/09/2015 DOCUMENT STRUCTURE AND TABLE OF CONTENT,

FIRST INPUT 14.5

0.2 05/10/2015 FIRST CONTRIBUTIONS FOR 12.5

0.3 13/10/2015 UPDATED CONTRIBUTIONS FOR 12.5

0.4 20/10/2015 CONTRIBUTION 13.5

0.5 26/10/2015 INPUT KPI Part 1

0.6 05/11/2015 INPUT KPI Part 2

1.0 06/11/2015 FINALISATION

1.1 14/11/2015 FEEDBACK FROM PEER REVIEW INCLUDED




Table of Contents

EXECUTIVE SUMMARY 5

ACRONYMS AND ABBREVIATIONS 6

LIST OF FIGURES 7

1 DELIVERABLE 12.5 8

1.1 WHIRLPOOL USE CASE 8 1.1.1 Introduction 8 1.1.2 Challenges for big data analytics for process improvement 9 1.1.3 Our approach for big data clustering 10 1.1.4 Use Case: Whirlpool washing machine testing 11 1.1.5 Results 14 1.1.6 Evaluation of the Experimentation 19 1.1.7 Visualization of Deviation Maps 21 1.1.8 Lessons Learnt 22 1.1.9 Conclusion 23

1.2 TRW USE CASE 23 1.2.1 Introduction 23 1.2.2 Integration architecture 24 1.2.3 Visualization examples 25 1.2.4 Lessons learnt 26

1.3 REFERENCES 26

2 DELIVERABLE 13.5 27

2.1 INTRODUCTION AND DOCUMENT SCOPE 27 2.2 EVALUATION METHODOLOGY 27 2.3 METHODOLOGY MAIN STEPS 27

2.3.1 Feedback collection channels 27 2.3.2 Interview structure 27 2.3.3 Interviews and reporting 28

2.4 IMPROVEMENTS AND EXTENSIONS RELATED TO END USER FEEDBACKS 28 2.4.1 CONSULGAL TRIAL 28 2.4.2 AIDIMA TRIAL 33 2.4.3 VW TRIAL 34 2.4.4 WHIRLPOOL TRIAL (Digital factory) 34 2.4.5 TRW TRIAL (Digital Factory) 35

2.5 INTERVIEWS WITH END USERS 36 2.5.1 Interviews in CONSULGAL 36 2.5.2 Interviews in AIDIMA 38 2.5.3 Interviews in VW 41 2.5.4 Interviews in WHIRLPOOL 43 2.5.5 Interviews in TRW 45

2.6 LESSONS LEARNT 46 2.6.1 Lessons Learnt in AIDIMA 46 2.6.2 Lessons Learnt in CONSULGAL 49 2.6.3 Lessons Learnt in VW 50 2.6.4 Lessons Learnt in TRW and WHIRLPOOL 51

2.7 CONCLUSIONS 53

3 DELIVERABLE D14.5 54

3.1 INTRODUCTION 54 3.2 GENERATION AND TRANSFORMATION OF VIRTUALIZED ASSETS (GETOVA) 54

3.2.1 Short overview of GeToVA SE 54 3.2.2 Experiments and Results 56 3.2.3 Lessons learned 61




3.3 ADVANCED MANAGEMENT OF VIRTUALIZED ASSETS (MOVA) 61 3.3.1 Short overview of MoVA SE 61

3.4 EXPERIMENTS AND RESULTS 61 3.4.1 Data Modelling 61 3.4.2 Importing 62 3.4.3 Cluster Search 67 3.4.4 Integrating MoVA with SME Cluster 67 3.4.5 Lessons learned 70

3.5 INTERVIEW WITH TANET 71 3.6 CONCLUSIONS 72

4 UPDATE ON KPIS IN THE TRIALS 73

4.1 AGUSTAWESTLAND 73 4.2 AIDIMA 78 4.3 VOLKSWAGEN 80

4.3.1 MR Update Cost 81 4.3.2 MR Update Time 81 4.3.3 Average lead time to access experts knowledge 82 4.3.4 Evaluation Accuracy 82 4.3.5 Inquiry Respond Time 82 4.3.6 Inquiry Respond Cost 83

4.4 CONSULGAL 83 4.5 TRW 86

4.5.1 Trial Results and Progress 86 4.5.2 TRW KPIs Analysis 87 4.5.3 Consolidated Trial Experience 89

4.6 TANET 90 4.6.1 Overview 90 4.6.2 General Comments about KPI’s 91

4.7 COMPLUS 92 4.7.1 Network Transparency For More Efficient Supplier Search 92 4.7.2 Transparency And Consistency Of ITs And Documents 92




Executive Summary

This joint deliverable D12.5 D13.5 and D14.5 documents the final results of the evaluation of

work in Tasks 12.3, 13.3 and 14.3. These Tasks focus on involving the FITMAN Trials in the

technical evaluation of the customized platforms and solutions that have been implemented in

WP4+WP12; WP5+WP13 and WP6+WP14. In this sense, this joint deliverable is the summa

of all the experiences gained in our 10 industrial Trials. Trials (end users and their IT

partners) have been asked to experiment with the trial integrated systems and to report

essentially about their correctness (the solution behaves as expected), their completeness (the

business processes are supported as expected and all the key functionalities have been

implemented) and their quality (regarding performance, security and user friendliness). The

result of this evaluation has been analysed to extract lessons learnt at both trail and single

component point of view and therefore derive possible improvements for them.

Interviews have been conducted with the IT specialists and end users, to capture their

evaluation of the provided solutions and to know from them in particular if critical

functionalities were missing and how difficult has been for them to start using the software.

Overall, all comments have been largely positive and encouraging, with the confirmation

from the end users that the provided solution fully meets the expectations and supports the

selected business processes. Moreover, trials foresee the possibility to continue using the

solutions after the closure of FITMAN. This will be reported in the exploitation report of

WP9.

An update of the KPI analysis of some of the trials, based on the results of the latest 6 Months

of experimentation concludes this deliverable.

As a consequence of the evaluations performed in this phase, in Smart Factory the

components for data clustering had to be extended in order to deal with huge amount of data

as in the WHIRLPOOL trial. The components for visualization which could help validating

the results needed to be adapted accordingly.

In Digital Factory, some performance requirements asked for optimisation of the SEs

involved. One example is a virtual meeting with a high number of concurrently active

widgets. Some of the SEs, such as the 3DWV, might require some time to complete the

computation/transformation tasks and thus are slower in responding. Therefore, some

extensions and adaptations have been made to solve these issues.

In Virtual Factory, the cluster modelling for the automatic suggestion of suitable clusters was

a real challenge. The ontology modelling and search algorithms needed to be extended

accordingly. In addition, the import and export of data proved to be a challenge which was

met by optimised interfaces and APIs.

The intense experimentation activity in all the trials throughout the full lifetime of FITMAN

highlights the enormous interest of the industrial partners in the project.




Acronyms and Abbreviations

3D PCAP 3D Point Cloud Analysis Processing

BP Business Process

BS Business Scenario

C3DWV Collaborative 3D Web Viewer

DCC Digital Content Creation

CAD Computer Aided Design

DF Digital Factory

GE Generic Enabler

GUI Graphic User Interface

MR Machine Repository

SE Specific Enabler

SEMed Semantic Mediator

SF Smart Factory

TSC Trial Specific Component

VF Virtual Factory

VW Volkswagen




List of Figures

Figure 1: Correlation between power and rotation speed ......................................................... 12 Figure 2: Power parameter of multiple tests ............................................................................. 12 Figure 3: Comparison of two signals – red signal is shifted even more ................................... 13 Figure 4: Generated medoids .................................................................................................... 14

Figure 5: First cluster ................................................................................................................ 15 Figure 6: Second cluster ........................................................................................................... 15 Figure 7: First anomaly ............................................................................................................ 16 Figure 8: Second anomaly ........................................................................................................ 16 Figure 9: Third anomaly ........................................................................................................... 17

Figure 10: First cluster (large dataset) ...................................................................................... 17

Figure 11: Second cluster (large dataset) ................................................................................. 18

Figure 12: Third cluster (large dataset) .................................................................................... 18 Figure 13: Fourth cluster (large dataset)................................................................................... 19 Figure 14: Bigger dataset makes validation/visualization of analysis more complex .............. 21 Figure 15: Web App for the Visualization of Deviation Maps ................................................ 22

Figure 16: Class diagram for extension of TRW use case ....................................................... 24 Figure 17: Final architecture .................................................................................................... 25 Figure 18: Examples of two employee positions during work ................................................. 26

Figure 19: the original unique Dam Zone view ........................................................................ 30 Figure 20 The view Dam zones shows the list of the concrete compositions used in a slected

zone of the dam......................................................................................................................... 31 Figure 21: The new "Concrete Operations" view ..................................................................... 32 Figure 22: The NPAB details ................................................................................................... 33

Figure 23. Whirlpool part used for the trial. ............................................................................. 34

Figure 24: Diagram of the structure of the numerical dataset .................................................. 35 Figure 25. TRW spindle used in the trial.................................................................................. 35 Figure 26. TRW spindle used in the trial.................................................................................. 47

Figure 27: GeToVA Architecture ............................................................................................. 55 Figure 28: GeToVA integrated with MoVA in TANET trial ................................................... 56

Figure 29: TANET cluster generated by GeToVA .................................................................. 57 Figure 30: Individual Profile in JSON extracted by GeToVA from LinkedIn ......................... 58 Figure 31: Individual Profile in JSON extracted by GeToVA from LinkedIn ......................... 59

Figure 32: GeToVA in COMPlus trial ..................................................................................... 60 Figure 33 MoVA: Startscreen and model ................................................................................. 62

Figure 34: MoVA: Add new supplier ....................................................................................... 64 Figure 34: MoVA: Imported suppliers ..................................................................................... 64

Figure 38: MoVA: Restful API Import suppliers ..................................................................... 67 Figure 35: MoVA: Data of MoVA in SMECluster .................................................................. 68 Figure 36: MoVA: Restful API Result of Level 0 search ........................................................ 69 Figure 37: MoVA: Restful API Result of Level 1 search ........................................................ 69 Figure 38: MoVA: Restful API Result of Level 2 search ........................................................ 69

Figure 39: MoVA: MoVA Cluster Search Result in SMECluster ........................................... 70




1 Deliverable 12.5

This deliverable summarizes the results and the lessons learnt in WP12, focusing on the two

trials Whirlpool and TRW. Other trials such as Piacenza had already completed their

experiments.

1.1 Whirlpool Use Case

One of the most important challenges in manufacturing is the continuous process

improvement that requires new insights about the behavior/quality control of processes in

order to understand the optimization/improvement potential. This deliverable elaborates on

usage of big data-driven clustering for an efficient discovering of real-time anomalies in the

processes. Our approach extends traditional clustering algorithms (like k-Means) with

methods for better understanding the nature of clusters and provides a very efficient big data

realization. We argue that this approach paves the way for a new generation of quality

management tools based on big data analytics that will extend traditional statistical process

control and empower Lean Six Sigma through big data processing. The proposed approach

has been applied for improving process control in Whirlpool (washing machine tests, factory

in Italy) and we present the most important finding from the evaluation study. We note here

that the results of this deliverable will be published IEEE BigData 2015, Special Session -

From Data to Insight: Big Data and Analytics for Smart Manufacturing Systems [8].

1.1.1 Introduction

Due to dynamically changing business environment and especially permanently increasing

competition, one of the most important challenges in manufacturing nowadays is the

continuous process improvement (CPI) - defined as an ongoing activity aimed at

improving processes, products and services through sustainable changes over a period of time.

Most CPI strategies incorporate the Lean Six Sigma1 principle, which is a combination of

techniques and tools from both the Six Sigma Methodologies and the Lean Enterprise2. The

Six Sigma methodology is based on the concept that a "process variation” can be reduced

using statistical tools. The goal of Lean is to identify and eliminate non-essential and non-

value added steps in a business process in order to streamline production, improve quality and

gain customer loyalty.

Lean Six Sigma practitioners have been improving processes for years through statistical

analysis of process data in order to identify the critical parameters and the variables that have

the most impact on the performance of a value stream and to control their variations.

However, the main constraint is the complexity of the statistical calculations that should be

applied on the (large) datasets. A well-known example is the multivariate statistical analysis.

Standard quality control charting techniques (e.g., Shewhart charts, X-bar and R charts, etc.)

are applicable only to single variable and cannot be applied to modern production processes

with hundreds of important variables that need to be monitored. Indeed, the diversity of

process measurement technologies from conventional process sensors to images, videos, and

indirect measurement technologies has compounded the variety, volume, and complexity of

process data3. For example, it is typical in a modern FAB (semiconductor manufacturing) that

1 https://en.wikipedia.org/wiki/Lean_Six_Sigma

2 https://en.wikipedia.org/wiki/Lean_enterprise

3 http://wenku.baidu.com/view/59f0c1bd84254b35effd346f.html?re=view




over 50,000 statistical process control charts are monitored to control the quality of over 300

manufacturing steps in the fabrication of the chip [1].

Moreover, generated data can be extremely big: e.g. in-process monitoring in additive

manufacturing (3D printing) produces 100MB – 1GB data, whereas in-process geometry

inspection generates 1- 10GB data volume per part4.

Although process operations are rich in data, without effective analytical tools and efficient

computing technology to derive information from data, it is often the case that data is

compressed and archived for record keeping and only retrieved for use in emergency analysis

after the fact rather than being used in a routine manner in the decision-making process.

In this deliverable we report the testing and experimentation of a novel big data approach for

continuous process improvement that exploits above mentioned advantages for enabling

better understanding of a (dynamic) nature of a process and boosting innovations.

We argue that by performing Big data analytics on the past process data we can model what is

(statistically analyzed) usual/normal for a selected period and check the variations from that

model in the real-time (as Six Sigma requires). Additionally, these data-driven models can

support the root-cause analysis that should provide insights about what can be eliminated as a

waste in the process (as Lean requires). However, due to the above mentioned variety and

volume of data, the analytics must be a) robust – dealing with differences efficiently and b)

scalable - realized in an extremely parallel way.

The proposed approach has been applied for improving process control in Whirlpool factory

in Italy based on washing machine tests. In this deliverable we present the most important

finding from the evaluation study.

This section is organized in the following way: we start with description of the challenges for

big data analytics and then we continue with presenting our approach for big data clustering.

We provide details about the case study and lessons learnt from the trial. Finally, we

summarize the results.

1.1.2 Challenges for big data analytics for process improvement

In the nutshell of improving a process is the understanding of the nature of the process – what

is its normal/usual behavior? However, the pace of change is continuously increasing and

introduces new computational challenges for continuous process improvement. There are two

main issues that challenge traditional Lean Six Sigma approach for continuous improvement:

the number of parameters that can be measured in a process and corresponding size of

data to be analysed is exploding (note that it is strongly influenced by the supply-chain

networks, which expands the space of interest dramatically) and

process variations cannot be checked against predefined (expert) rules – the dynamics

of the process context requires the dynamicity in rules to be applied.

Therefore the detection of variations is not anymore the question of optimizing formulas from

statistics, but rather the challenge for defining what is “normal/usual” in the dynamically

changing business environment. This is where Big Data comes to the game:

by being inherently data-driven, big data processing of manufacturing data is able to

generate valid models of process behaviour.

by being very scalable, big data processing is able to work in high-dimensional spaces

of interests with low latency.

4 Sigma Labs, In-process Quality Assurance, Industrial 3D Printing Conference




by using unsupervised learning, big data processing can continuously improve/adapt

the performances of the underlying task.

In this deliverable we present how variations in a manufacturing process can be detected

using unsupervised data analytics, namely big data clustering.

We define the following three requirements that should be satisfied in a big data approach for

detecting anomalies

R1: Precision: how to define what is similar (metrics) – what is usual?

R2: Interpretation: how to understand why something is not similar – why it is

unusual?

R3: Scalability: how to ensure that by using as much as possible data, the results of

processing will be calculated as fast as possible?

1.1.3 Our approach for big data clustering

Clustering algorithms tend to identify groups of similar objects and produce partitions for a

given dataset. A number of clustering algorithms exist, from partitioning algorithms such as

K-means5, over hierarchical algorithms that form a tree of clusters (dendogram) by

performing clustering on different levels, to density based algorithms such as DBSCAN [2]

that group objects based on the neighborhood of each object. All of these algorithms have

their own purpose, advantages and weaknesses, so a great caution is needed while choosing

the appropriate clustering method.

K-medoids6 is a partitioning algorithm, similar to K-means, that uses medoids to represent

clusters. Unlike the K-means algorithm where centroids are used to represent clusters, in case

of K-medoids, medoid is one of the objects from the dataset that is the best representative of

the cluster. PAM (Partitioning Around Medoids) is the most common realization of this

algorithm. The basic steps of K-medoids algorithm are initialization, assignment of objects to

closest medoid and new medoid selection for each of the clusters. Medoid selection is the

most expensive procedure of the algorithm. FAMES is a medoid selection algorithm that tries

to overcome this problem.

K-medoid algorithms try to find optimal medoids in the dataset, while finding a single

medoids requires O(n2) distance calculations. This makes this algorithm practically unusable

for bigger datasets. FAMES (FAst MEdoid Selection) [3] represents an improvement of K-

medoids algorithm by offering a fast selection of good representatives.

Our solution represents a combination of scalable K-means|| [4] (K-means parallel)

initialization and K-medoids like algorithm that relies on FAMES for medoid selection. K-

means|| represents an improvement of K-means++ [5] algorithm. The major downside of the

K-means++ is its inherent sequential nature, which makes it difficult to use in case of big data

because it requires k passes over the data to find good initial set of centers. On the other hand,

K-means|| is able to find good initial medoids in n iterations, where n is usually much smaller

than k. Another good property of this initialization approach is that it can be easily distributed

therefore achieving greater speed than K-means++ and can be used in the case of much bigger

datasets.

5 https://en.wikipedia.org/wiki/K-means_clustering

6 https://en.wikipedia.org/wiki/K-medoids




1.1.4 Use Case: Whirlpool washing machine testing

The problem setup for washing machine tests for Whirlpool use case is as follows:

Washing machines functional tests are provided by Whirlpool;

Functional test is performed on every washing machine that was assembled and it is

used to examine if machine functions properly;

Various parameters are measured, such as power, speed and water inlet;

The size of the data is very large (too large to be processed on a single machine);

The goal is to detect anomalies – washing machines that behaved strangely during

functional test;

By detecting anomalies during functional tests quality of production can be increased;

Data is provided in a compressed form, so the first step is to decompress it;

Based on analysis a solution for detecting anomalies automatically should be

implemented.

The goal is to find a way to define normal parameter values and implement a solution that

will be able to identify unusual patterns in functional tests. Dataset provided by Whirlpool

contains three parameters:

Power;

Speed;

Total water inlet.

Initial analysis of the dataset was performed in order to find correlations between parameters,

as presented in Figure 1. The goal of our analysis is to describe normal behavior and discover

anomalies as behavior that does not conform to the defined model. To do that, we are using

clustering approach described in the previous section. We elaborate shortly on the arguments

to use that approach.

First of all, we have defined the process of detecting anomalies as follows:

Cluster the data using some clustering algorithm that will not only produce clusters,

but will also produce cluster representatives;

Cluster representatives and cluster variances form the model;

Each new measurement is compared to the existing model and the dissimilarity from

the model determines its status as anomalous or normal.

Most of clustering algorithms that respect these conditions belong to the group of partitioning

algorithms. Before we describe the concrete algorithm that we use from that group, it is

necessary to take another look at the dataset, this time considering multiple tests at the same

time.

Power parameter values of multiple tests are presented on Figure 2. We can notice that even

though all the graphics present the value of the same parameter, they may have very different

shapes. This has a huge impact on our analysis. There is a need for robust measure that is able

to cluster time series based on the shape of the series.




Figure 1: Correlation between power and rotation speed

Figure 2: Power parameter of multiple tests

Another question occurs due to the latest condition, clustering time series based on shape.

How should the cluster representative look like for a cluster that contains series of different

shape? Some of the algorithms that produce cluster representatives give some kind of a mean

or an average of all the elements in the cluster as a representative. But that would not be an

acceptable solution in our case for the following reason – different tests may be performed

under different conditions. By conditions we mostly mean different timings. In case of one

test, centrifuge can be started part of a second earlier than in the other, or can have slightly

greater duration. That part of a second makes mean of series impossible to use as a

representative. Because of that, we turn to another group of clustering algorithms, algorithms

that select one of the objects from the cluster as a representative of the cluster. That object is

the one that is the most similar to all the other objects in its cluster, and it’s usually called a

medoid. This is the reason we selected k-Medoids as clustering algorithm.




First of all, we consider initial medoids selection. This question has more relevance than it

may seem. Quality of final clustering heavily depends on initial medoid selection due to

phenomenon of local minimum. One option would be to select k objects randomly from the

dataset, but that can result in poor clustering. That is why we use a different kind of

initialization in our implementation. The idea is to select objects that are placed somewhere in

the core of existing clusters as initial medoids. In this way, only a small number of iteration is

needed to produce final clusters, and its role is just to refine initial clustering. This makes our

algorithm faster and more precise than in case of random initialization.

Second, we consider the question of objects similarity. To perform clustering it is necessary to

define a measure that will describe how similar objects are. Similarity measure is often

compared to distance measure, since the objects are observed in N-dimensional space.

Maximal similarity between two objects can have a value of 1, which means that distance

between them is minimal, that is, equal to 0. There are different kinds of distance measures,

such as Euclidean, Manhattan or cosine, but we will see that they are not very useful in our

case. The reason for that is that we are comparing time series that can be shifted in time, or

skewed, and distance-measures like Euclidean do not tolerate this. That is why we approached

another kind of distance measures that considers shapes of two signals. The measure is called

Dynamic Time Warping (DTW) [4] and it is able to find the optimal alignment between two

signals. To show this we will observe two very similar time series, presented on Figure 3. It

should be noticed that one signal is actually a modification of another created by shifting the

first signal by a certain time interval. The figure represents a comparison of Euclidean, cosine

and DTW distances of the two signals.

It can be noticed that Euclidean distance is very large, and so is the cosine distance. DTW, on

the other hand, gives distance equal to zero. This implies that DTW is immune to the

phenomenon of shift, no matter how big it is. This could be very useful in our case, since

different conditions are used while performing functional tests. We could interpret this shift in

the following way – counter clockwise rotation was started later in the case of second signal

than in the case of the first signal, so every next step in the test (for example, centrifuge being

started) also gets shifted. But this doesn’t mean that something is wrong with the other test.

The values are normal; they are just shifted in time.

Figure 3: Comparison of two signals – red signal is shifted even more




1.1.5 Results

In this section we present the main conclusion from the performance test.

We have initially considered Power parameter and took a small sample to validate our

methods. We used a variant of K-medoids that we have mentioned and got five clusters, while

three clusters where clusters singletons. Cluster singleton is a cluster that contains only one

test. We consider these clusters anomalous, since the number of objects they contain is very

small, so they differ from the rest of the dataset. Medoids that were produced are presented on

Figure 4, while clusters are presented from Figure 5 to Figure 9.

By observing the shape of signals contained in the sample we can conclude that there really

are two groups of signals. At the same time we can notice that signals belonging to clusters

singletons have different shape than signals that exist in non-singleton clusters. We must

emphasize that our primary goal is not to find anomalies while clustering, since this can be a

long-term operation, but to generate a model that can be used in real time to detect anomalies.

Even so, we may detect suspicious tests in the dataset, like in our example, so the best

solution is to report them, and remove them from the dataset, for safety reasons.

The initial sample was good for validation of the approach and the implementation, but

afterwards all the tests provided by Whirlpool were analyzed. The dataset currently provided

contains about 15.000 functional tests, but a much larger amount of data is expected (amount

that demands a Hadoop cluster for processing – learning what represents normal behavior).

Figure 4: Generated medoids




Figure 5: First cluster

Figure 6: Second cluster




Figure 7: First anomaly

Figure 8: Second anomaly




Figure 9: Third anomaly

Clustering results for the whole dataset are now presented from Figure 10 to Figure 13.

Again, there are clusters of normal behaviors and there is a cluster of potential anomalies

(cluster presented on Figure 11).

Based on results we conclude that our algorithm represents a good solution for the problem of

detecting anomalies in functional tests provided by Whirlpool. We even got a confirmation

that we were able to detect tests that represent a problem that really exists in one of Whirlpool

facilities (Figure 10 represents an example of such tests), which they are aware of. We

developed a solution that is able to compare test series based on their shape, and to cluster

tests based on it. We also used that similarity and clusters being produced to determine which

tests look unusual comparing to normal examples found in the dataset.

Figure 10: First cluster (large dataset)




Figure 11: Second cluster (large dataset)

Figure 12: Third cluster (large dataset)




Figure 13: Fourth cluster (large dataset)

1.1.6 Evaluation of the Experimentation

In this section we summarize the lessons learnt from the trial.

1. What were the problems in using Engineering infrastructure for big data

experiment?

- Access to machines in Engineering infrastructure

o There was only one machine with public (static) IP address

o We were able to connect to that machine only from machines in our office

using SSH (there was a whitelist for server access)

o There were some problems with the specific port given by Engineering for

access

o Number of machines that could connect to machine with public access was

limited

o There were a lot of ssh connections “hanging”, even if have closed the

connection (which led to messages such as “ssh_exchange_identification:

Connection closed by remote host” and “ssh_exchange_identification: read:

Connection reset by peer”, which made it unable to connect to machines)

- Missing support for the components in Hadoop ecosystem

o There was lack of support (knowledge) for setting up (using) Oozie on

Engineering machines

o A lot of difficulties occurred due to insufficient permissions for starting-up

Hadoop jobs from Oozie coordinator

o Number of workflows limitation as a result of small cluster size

o Non-default YARN settings (port) led to more problems




2. What was the main difference in the results (from the analytics point of view)

between the first and second experiment in Whirlpool

- More data that needed to be cleaned

o The dataset provided by Whirlpool was in a compressed form and after

decompressing it has been noticed that it contains some irregular records. For

example:

There were some records in which all the parameters had zero value all

the time

Values for a parameter are pipe-separated “|”, which is fine, but records

could be found that contain long sequences of such pipes without any

actual values in between. For example sequences such as “||||||” exist

Exceptions were thrown during decompression, such as

“java.io.IOException: incorrect data check”

o This requested data cleaning before the actual analysis

o In the first phase analysis was performed on a small sample

o In the second phase analysis was performed on the whole dataset

o The number of functional tests (records) significantly increased (the size of the

subset was 20, the size of the full dataset was about 15.000)

o With the dataset size increase the number of “dirty” records increased as well

o That means that cleaning the big dataset is a “big job” itself

- More difficult visualization/interpretation of results

o The greatest part of the analysis was based on clustering

o Clustering is an unsupervised machine learning method

o Clustering as an unsupervised method has its strengths (the dataset doesn’t

have to be labelled, for first)

o As an unsupervised method clustering has a lot of challenges (clustering

algorithm, distance measure, number of clusters, quality of clusters…)

o One of the biggest challenges is validation of the analysis

o With bigger dataset validation becomes even a greater challenge

o With the small subset it was easier to perform validation

o Visualization (which could help validating the results) is much more complex

in the case of the bigger dataset (following figure illustrates this)

o A lot of effort (and creativity) needs to be put so the orientation could be found

in the “big mess” that comes with large amount of high dimensional data




Figure 14: Bigger dataset makes validation/visualization of analysis more complex

1.1.7 Visualization of Deviation Maps

Additionally to the results presented so far, a Web application for the visualization of

deviation maps for the Whirlpool use case was developed which can be deployed in any Web

server. The idea of a deviation map is to visualize the differences (i.e. deviations) of a

physical part from the CAD model which was defined for its design. For this purpose, usually

a point cloud is produced for the physical part. In the context of FITMAN this point cloud is

produced by a high accuracy laser scanner. The point cloud is then compared with the CAD

model which results in the deviation map. The scanning and computation of the deviation is

not part of the Web application. The Web app purely displays the deviation maps which were

already computed. The easiest way to access the Web application is to go to

http://xml3d.org/xml3d/scenes/magnifi/.

http://xml3d.org/xml3d/scenes/magnifi/




Figure 15: Web App for the Visualization of Deviation Maps

Figure 15 displays a screen shot from this Web app for displaying deviation maps. The

deviation maps are in the first place provided as binary data which is, however, easy to

interpret. It consists of a list of triangles all of which have a deviation value assigned. For the

visualization of the deviation map in a browser on the basis of XML3D, it was necessary to

convert the data into XML3D. For this the pure model information (i.e. the triangles) were

extracted. The deviation information is given as a separate vector where the position of the

deviation values in the vector need to correspond with position of the respective triangle in

the mesh of the model to which the deviation information belongs.

Additionally to the Web application a DyVisual Web client was developed for the FiVES

synchronization server. The advantage of using the DyVisual Web client is that the deviation

map can be investigated cooperatively, i.e. several users located at geographically distributed

site can view the model simultaneously while all changes of the view in which the deviation

map is visualized is synchronized among all clients which are connected to the same

DyVisual server. The Web client provides a restful service interface where rest services for

uploading a new deviation map and for adding or modification of the deviation information

for a given model.

The idea of the synchronization is that a group of experts which might be located at different

sites can cooperatively investigate a given deviation map. The assumption is that only one

expert is manipulating the synchronized view. We assume that the experts have an audio

connection for their discussion which they can use to agree on who is allowed to modify the

view.

1.1.8 Lessons Learnt

From the experiences with the FITMAN trials one can conclude that DyVisual is a powerful

tool for the visualization of dynamic 3D content in the context of the World Wide Web. The

main issue to solve for an applications is to acquire the data for the 3D models in the first

place. The data representation format of this data is of course an issue. DyVisual supports the




most important standards for representing 3D models and with this a large number of

applications. The pure visualization of the model in a browser is straightforward. However,

when the model should be dynamically modified or even animated, deeper knowledge of

XML3D, FiVES and 3D data in general is necessary. For applications like the visualization of

deviation maps special shaders might be necessary which also requires special knowledge of

XML3D. However, the generic enablers which form the basis for DyVisual, i.e.

XML3D/Xflow and FiVES are well supported and documentation can be found at

http://catalogue.fiware.org/enablers/3d-ui-xml3d and https://github.com/fives-team/fives.

1.1.9 Conclusion

The clustering of big data for the real-time discovery of deviations can be achieved by an

extension of traditional clustering algorithms. Such new approaches will extend traditional

quality management tools and will thus empower current frameworks such as Lean Six

Sigma.

The proposed approach has been applied for improving process control for washing machine

tests in Whirlpool factory in Italy and the results are very promising. We have started a large-

scale case study for the presented washing machine functional tests that should prove the

feasibility of the approach for production environment.

1.2 TRW Use Case

1.2.1 Introduction

The task was to integrate the detection movement and the joint positions it is reporting with

our DyCEP, whose results will be visualized using DyVisual.

The first step was to create a NGSI10-to-RabbitMQ mapping web service to import the

movement events into DyCEP. DyCEP is designed as a reactive component, reacting on

events published in real time, so the preferred (in Storm almost mandatory) way of input is a

broker/queue. NGSI10 is a request/response protocol built on top of HTTP and it is not quite

suited for communication including subscribers. However, since NGSI10 is mandatory, we

created a bridge accepting NGSI10 XML-based messages and transforming them into an

internal format, published onto a local instance of RabbitMQ. The service is written in Scala

using PlayFramework and we called it the Collection Service.

The Storm topology within DyCEP is subscribed to the broker and executes its algorithms.

The output of the DyCEP generally should include the last angle and risk per distinct joint for

the visualization to have appropriate input. We needed to slightly change the operation of

DyCEP since it initially did more of a statistical overview of the worker movement in the past

five minutes and the risk distribution. We had to change the pattern to make a view of only

the last movements and exclude the statistics and to create appropriate adaptation logic for the

DyVisual input.

DyCEP also has to map the joint movement from a numerical sensor id notation to explicit

joint label suitable for DyVisual. This is done by specification provided by Innovalia.

The DyVisual has a REST endpoint for updating the avatar position, with a configuration

string specifying all the joints position in 3d space. The missing coordinates have default

values. Configuration string is shown in the example.

http://catalogue.fiware.org/enablers/3d-ui-xml3d




A set of classes is created which are used for generating the configuration string.

Figure 16: Class diagram for extension of TRW use case

1.2.2 Integration architecture

The following figure shows the architecture after the integration. Notice the change with the

output format of the CEP and the destination of that output, as well as the path which is used

to visualize the position and risk level.

{

"avatarID":"0",

"aniName":"dummy.bvh",

"configString":"update,Spine,y,50,RightForeArm,y,100"

}




Figure 17: Final architecture

Kinect (2) detects worker (1) joint angles. After some adaptation to the NGSI10 interfaces the

sensor ids (where a sensor id corresponds to a joint) and joint angles are reported to the

collection web service, explained earlier. The service adapts the messages to a pub/sub

interface (the broker) and the CEP (5) consumes those messages executing the algorithms.

When a result is ready, which means a set of coordinated joint positions and calculated risks,

a message for the DyVisual API is created. In the previous architecture the message was

different; it was an overview of the risks for each joint in the last 5 minutes, while the new

one is a message with the latest angle position paired with the risk. The API call is placed to

the DyVisual backend, which is hosted on some web server. It could be on the same web

server as the collection web service, like on the image, or might not be. The backend updates

the visualization, presumably viewed from a web browser on a computer. On the image, the

computer viewing the avatar and the computer which has the Kinect connected are the same,

however that doesn’t have to be the case. The avatar can be viewed from multiple computers

in general.

It should be noted that the avatar position API call and the markers (the colored orbs) are set

in multiple calls. The avatar can be set in one call, while each orb is set by a separate. Before

updating any of the markers, all of them must be deleted since they stack up on joints.

1.2.3 Visualization examples

On these two images shown in Figure 17 we can see examples of two employee positions

during work. The avatar itself represents the body position, while the orbs around the joints

show the risk level of that joint. Red being the highest, risk level 3, and the green is the

lowest, risk level 1. If there is no orb than the joint is visualized in its default position, and

there is no movement detected.




Figure 18: Examples of two employee positions during work

1.2.4 Lessons learnt

In this section we summarize the lessons learnt from this trial.

1. Real-time processing of the signals from Kinect can be important for different

working situation

2. The signal from Kinect is rather complex and requires a substantial pre-processing in

order to get the data in the proper form

3. Big data opens new possibilities for process optimization, based on data collected in

all phases

4. Big data analytics enables powerful observing/sensing and reacting if needed

5. The rules for quality control are defined manually (and are not conceptually sound)

1.3 References

[1] Qin SJ, Cherry G, Good R, Wang J, Harrison CA. Semiconductor anufacturing process control and monitoring: a Fab-wide framework. J Process Control 2006;16: 179–91.

[2] http://www.dbs.ifi.lmu.de/Publikationen/Papers/KDD-96.final.frame.pdf

[3] Adriano Arantes Paterlini, Mario A. Nascimento, Caetano Traina Jr., Using Pivots to Speed-Up k-Medoids Clustering, JOURNAL OF INFORMATION AND DATA MANAGEMENT Vol 2, No 2 (2011)

[4] http://theory.stanford.edu/~sergei/papers/vldb12-kmpar.pdf

[5] https://www.math.uwaterloo.ca/~cswamy/papers/kmeansfnl.pdf

[6] http://static.googleusercontent.com/media/research.google.com/en//archive/mapreduce-osdi04.pdf

[7] Black, A. W.; P. Taylor: Automatically clustering similar units for unit selection in speech synthesis. In: Proc. Eurospeech ’97

[8] Stojanovic N., Dinic M., Stojanovic L., Big Data Process Analytics for Continuous Process Improvement in Manufacturing, to appear in IEEE BigData 2015, Special Session - From Data to Insight: Big Data and Analytics for Smart Manufacturing Systems, Oct 29 – Nov 01, 2015.

http://www.dbs.ifi.lmu.de/Publikationen/Papers/KDD-96.final.frame.pdf

http://theory.stanford.edu/~sergei/papers/vldb12-kmpar.pdf

https://www.math.uwaterloo.ca/~cswamy/papers/kmeansfnl.pdf




2 Deliverable 13.5

2.1 Introduction and Document Scope

The main objective of task T13.3 is to analyse and evaluate the experimentation provided by

T13.2 and derive from that the bottlenecks and opportunities for further expansion of the trials

and for improvements of the single SEs and PCs

This document is the update of the D13.4 deliverable, reporting final improvements operated

on the results delivered in T13.2 and the outcomes of the interviews with the end users,

conducted after their experimentation of the final versions of the integrated prototypes.

2.2 Evaluation Methodology

The first evaluation round (reported in D13.3) collected feedback about single components.

This feedback was provided by mail and phone. The second round was done via dedicated

experiments with end users.

Since the end users conducted an experimentation of the overall integrated solution for their

trial, with possibly an incomplete visibility of the exact role played by each single component

in the solution, it has been decided to first collect feedback and comment on the overall

solution and then, whenever possible, to try to make the evaluation report more specific for

the single components. If that was not possible, either the mediation of the IT partner has been

asked, or the IT partners of WP13 analysed the provided feedback to identify bottlenecks,

gaps and suggested improvements that can be reported to the SEs and PCs.

2.3 Methodology main steps

2.3.1 Feedback collection channels

Some channels and tools have been used to collect feedback directly from end users:

Emails, web meetings, excel file, and, at the end of the evaluation phase, finally direct

interviews with the end users. This feedback is reported here.

In addition to that, other feedback data have been mediated by the IT partners from WP5, who

were in charge of the trial platform and acted as interface with the end users for the technical

and integration issues: IPK has been responsible for the integration of the WP13 prototype in

the VW trial and UNINOVA did the same for the CONSULGAL one.

2.3.2 Interview structure

The overall structure of interview is the following:

Evaluation of the integrated solution

o Questions for the IT partner in charge of the Trial platform: here the questions

are focused on installation and deployment issues, overall performances etc..

o Questions for the end user: here questions are about the coverage of the

selected business processes, the coverage of requirements and level of

completeness of the provide set of functionalities

Evaluation of the documentation and provided support: here the questions are on the

completeness and easiness of use of the SEs documentation available on the catalogue

and on the received support.




Evaluation of specific SEs and PCs: in this part, questions about the completeness of

the solution, the adopted approach, missing functionalities, and

replicability/adaptability to other processes were asked.

2.3.3 Interviews and reporting

The end users have been asked to dedicate a 1-hour slot to conduct the interviews through

web conference channels.

The Trials experimenting with the SE (ADIMA, VW, and CONSULGAL, TRW and

WHIRLPOOL) received the interview structure in advance and had the possibility to get

prepared for the interview that has been conducted by the IT partners (one IT partner has been

appointed as responsible for each interview) using a web meeting tool.

The results of the interview have been shared within the WP13 partners, so that each IT

provider analysed the answers to identify the lessons learnt not only ad trail level, but also at

SE/PC level.

2.4 Improvements and extensions related to end user feedbacks

This section reports the minor modifications and improvements on the SEs, PCs and overall

integration solutions, developed for the trials during the experimentation phase to meet the

feedbacks received from the end users during the initial validation activities. We highlight

here the latest feedback and do not include feedback already processed earlier in the project,

for example provided by Augusta Westland.

2.4.1 CONSULGAL TRIAL

2.4.1.1 SEMed improvements

The configuration of SEMed was customized and deployed for the iLike views provided

before CONSULGAL suggested some re-organization of the information visualized. During

the specification phase additional attributes for the concrete operations view, changes in the

attribute mappings and rearrangement of attributes for the concrete operations view were

necessary. In consequence, the configuration of SEMed has been adapted, deployed and a

new set of SPARQL queries has been provided to enable the new content of the information

flows. The information flows which are mentioned in D13.3 are still consistent.

2.4.1.2 ILike improvements

During the experimentation with the provided version of iLike, CONSULGAL provided

clarifications on the semantic of some data that have been transferred from the Trial platform

to the SEMed and from there visualized in the iLike interface; moreover CONSULGAL

suggested some re-organisation of the information visualized in the iLike interface, to better

support the selected business process.

Such modifications are summarized in the table below:




Visualisation Provided information Improvements/modifications

“Last concrete operations” offering a synthetic view of

the concreting operations

executed in the last period,

including links to visualize

the information of the

extracted samples; the

requested modifications

where;

This view was not existing

before: CONSULGAL

required it to better support

the activities of the

inspectors, that need to

search the system to retrieve

complete description of

concreting operations

“NPAB view” information on the concrete

composition, position in the

dam and approval date

Some data have been re-

named to reflect their

meaning; consistency check

with other views

implemented

“Dam Zone” providing access to the

information of each zone of

the Dam

This view was already

available in the previous

version; some corrections on

the structure of the visualized

data have been implemented

The first working version, proposed to the trial, offered a simple list of the concrete

operations filtered by the dam zones (shows the list of the concrete compositions used in a

selected zone of the dam). The details of the concrete composition and the samples were

contained in modal panel. This kind of view was developed in order to offer the simplest

interaction to the user but, after a first evaluation with CONSULGAL, the following weak

points have been identified:

the filters are limited at the dam zone while others important parameter are ignored

the dam zones must be identified by 4 coordinates instead of 3

the concrete operation was not identified uniquely

the details of document NPAB was not available from the views




Figure 19: the original unique Dam Zone view

In order to address the identified problems and improve the navigation, the system has been

modified splitting the view (see following Figure 20) in two different views: “Concrete

Operations” and “Dam Zones”. Basically the set of displayed information are almost the same

but they are organized in two different ways to meet the needs of the user in different time of

the dam building.

The Dam zones view

The view “Dam Zones” is useful in the BP1: “Identification of concrete class and concrete

composition process”. This view allows the user to focus on the concrete class element, and

also to analyse the results of the testing on samples collected from the concrete operation that

use the selected concrete class. This view is useful in the long time because it allows

understanding the behaviour of a concrete during the time, but also it provides a summary of

the state of the dam area.




Figure 20 The view Dam zones shows the list of the concrete compositions used in a selected zone of the

dam

The Concrete Operations view

The BP4 “Slump tests results for each concreting operations” requested to create a view with

a collection of the details of each concreting operation. From a concrete operation element, it

is now possible to retrieve the details of NPAB (the formal document approving the

concreting operation) and of the concrete composition; moreover for each concreting

operation it is possible to access the list of the samples taken during the single concreting

activities.




Figure 21: The new "Concrete Operations" view




Figure 22: The NPAB details

These examples show how the SEs involved have been optimised throughout the trials the

meet the needs of CONSULGAL.

2.4.2 AIDIMA TRIAL

2.4.2.1 C3DWV improvements

End users required some support to correctly transform the 3D models and make them

visualized in the C3DWV.

2.4.2.2 Virtual Obeya improvements

An updated version of the VO has been provided to AIDIMA, including templates that can be

used to easily create Obeyas and set up collaboration sessions.

Moreover, from some interactions with AIDIMA; it emerged that having the possibility to

share in the VO documents is critical during the design collaboration meetings. Whereas the

creation of widgets to share Google Docs is supported by the VO through a set of pre-defined

templates, AIDIMA and its associates cannot use this kind of documents, but rather share

documents using OneDrive. Therefore, Holonix and AIDIMA worked together to find a

solution that could be easy for OneDrive documents too.




2.4.3 VW TRIAL

SEMed and 3DWV have been updated slightly throughout the trial.

2.4.4 WHIRLPOOL TRIAL (Digital factory)

The Whirlpool trial in the Digital Factory is focused on the dimensional quality control for

plastic parts produced in a Whirlpool manufacturing environment. In this trial the objective is

to use the 3DScan SE and the 3D Point Cloud Analysis Processing PC as a whole system to

produce dimensional data that will be used as an input by the Dynamic CEP SE to analyse

statistically the information obtained. The samples analysed can be also visualised

dynamically in the DyVisual SE, as well as stored and visualised using the 3DScan SE

storage and visualisation components.

The scanned part and the CAD model should share the same reference system to be

compared, so the colour mapping with the deviations can be calculated. This requirement,

together with the need of sharing and exchanging data together with the Dynamic CEP SE has

forced to make adjustments in the alignment mechanism of the part and the CAD model. As

the analysed part (see Figure 23) has revolution symmetry it is important that all the parts

belonging to the working sample are oriented in the same sense.

Figure 23. Whirlpool part used for the trial.

To use that revolution axis to install the reference system it has been necessary to develop a

small software customisation to accomplish with this requirement and make a specific

alignment. This is mainly to the fact that the part used in the trial has special geometrical and

functional properties, as it is revolution and rotating part.

The adjustment performed, from a conceptual perspective, consists of a mesh that is

calculated in the whole surface of the Whirlpool helix and divided in different regions with

triangle forms (see following figure). So the numerical dataset created after scanning the helix

is composed by a list of triangles (expressed with numerical coordinates, x, y, z, of the three

points that compose the triangle). Each triangle also contains a deviation that informs about

the difference between the scanned part and the CAD model (expressed in millimetres). In

some cases the deviation is zero, this usually means that the system has not been able to

calculate the deviation of this particular triangle. All the parts of the sample contain exactly

the same triangles, so each part can be easily compared with another one.




Figure 24: Diagram of the structure of the numerical dataset

2.4.5 TRW TRIAL (Digital Factory)

The TRW trial is clearly oriented to the dimensional quality control of components for the

automotive industry produced by TRW in one of its manufacturing plants. In this trial the

objective is to use the 3D Scan and the 3DPAP as a relevant tool to identify parts (Figure 25

shows the part used in the trial) that do not meet the dimensional specifications and to

advance towards the Zero Defect Factory paradigm, where dimensional defective parts are

identified through the means of a control system.

The system deployed in TRW (3DScan + 3DPAP) is very close to the standard applications of

the technology of Datapixel. In any case, some specific customisations were necessary to

adapt the technology to the trial. The technology was installed and configured in TRW’s IT

infrastructure. The system has been deployed in TRW manufacturing plant and this requires

always some specific customisations in terms of configuration to adapt it to the IT

environment and to the particular production line.

An important aspect that has been agreed between Datapixel and TRW is the selection of the

part used during the trial. Finally the part selected is an endless spindle.

Figure 25. TRW spindle used in the trial.

Deviation (mm) D

D’ D’’




Another important issue that Datapixel and TRW have agreed is which type of measurements

should be done and which type of defects should be detected. The clear definition of these

aspects has led to a reduction of the necessary steps for the usage.

2.5 Interviews with end users

These interviews summarise the final feedback at the end of the trials and after the

improvements made throughout the trials.

2.5.1 Interviews in CONSULGAL

2.5.1.1 Interview execution

The interview with CONSULGAL has been conducted on GoToMeeting and attended by

Paulo Rodrigues (CONSULGAL) and Sudeep Ghimire (Uninova)

During the interview, Mr. Rodrigues answered the questions about the completeness and

correctness of overall integrated prototype and of iLike, whereas the questions about SEMed

could be answered by UNINOVA, as the partner in charge of connecting the Trial Platform

with this SE.

2.5.1.2 Interview Results

2.5.1.2.1 Evaluation of the overall integrated prototype

As an IT integrator, UNINOVA considered the solution has no defects. The classification was

chosen according to the correct handling of the output of the web services of the

CONSULGAL Trial Platform. The easiness of the of application was classified at Level 2,

according to the fact that application requires fair amount of work for integration, specifically

caused by the constraints over the data model at the source. The resulting efficiency was

classified by UNIVOVA as expected for such solution and is “…acceptable for the scenario

we were working with…”. The reliability was classified by UNINOVA as high. A weakness

is the effort for configuration of the information flow which was classified as medium.

UNINOVA mentioned that the configuration of the system is time consuming and required

clear understanding of the documentation; it was worth it due to the value created.

CONSULGAL confirmed that the structure and the format of the information visualized in

the iLike UI corresponds to what the inspectors needs to know while performing their

activities.

Thus the solution is judged correct and complete.

Whereas the Trial platform developed in WP5 by Uninova covers the business processes BP1,

BP2 and BP3 of the CONSULGAL use case, WP13 solution covers the BP4 (Slump tests

results for each concreting operation) , requiring to compile, connect and make available all

the information collected during the other BPs.

The improvement in terms of efficiency provided by the solution is highly valuable, as

presently the information on concrete operations a slump tests must be searched across

different systems, the search covers and files and paper documents: with the Trial platform

and the solution from WP13, inspectors and other stakeholders can immediately retrieve this

information, even in mobility.




The impact in terms of time reduction has not been precisely quantified, but it is extremely

high.

2.5.1.2.2 Evaluation of the documentation

The documentation provided on the FI-WARE catalogue supporting the usage of the SEMed

has not been used by CONSULGAL, as end user.

2.5.1.2.3 Evaluation of the single components

SEMed

The integration and application of SEMed were driven through the integrators namely BIBA,

HOLONIX and UNINOVA.

UNINOVA defines the role of SEMed in the Trial’s Business Scenario as the mediator for

data between CONSULGAL’s Trial platform and iLike. So, the data being produced and

stored by the CONSULGAL’s Trial platform are provided to iLike via SEMed to leverage the

functionalities provided by iLike which was an added value for the business scenarios of

CONSULGAL. UNINOVA mentioned that the requirements according to the data integration

in the business processes within the scope of FITMAN for CONSULGAL were fulfilled. In

addition, the overall supported data integration approach is flexible enough for data

integration via web services. SEMed maintains loose coupling between data producers and

consumers thus providing a cleaner data integration approach. On the basis of these

experiences UNINOVA declared “…SEMed approach is quite independent of the domain of

business scenarios. So, we strongly believe that SEMed is applicable for other data integration

challenges beyond the scope of FITMAN…”. HOLONIX mentioned that the existing

flexibility for data integration is given, but the creation of queries for solution provider should

be simplified. The application of SEMed in other business processes could be possible by

UNINOVA within the next two years. UNINOVA also mentioned that the application of open

source in a production environment would be possible in the case that the support would be at

the same level that they got within the scope of FITMAN. In cases where open source is not

applicable for production environments, UNIVOVA could imagine to implement an own

solution. The idea would be to extend the SEMed solution with functionalities that can arise

from new business requirements. This is a path that UNINOVA believes they could follow.

UNINOVA mentioned finally that the integration of SEMed required some efforts to get a

stable solution, but it was worth the effort due to the value created by integrating SEMed in

the trial solution and that “…We hope to use SEMed in future projects to deal with data

integration problems…”.

iLike

The iLike interface has been evaluated very positively: all the information are visualised as

required and the requests of modifications provided after the preliminary evaluation phase

have been implemented.

A remarkable comment is that the solution is considered extremely intuitive and easy to be

adopted, with almost no need of specific trainer for the professionals who will use it.

As for the applicability to other sectors, Mr. Rodrigues said that it can definitely be used to

support the supervision of concreting operations for other type of projects (i.e. hotels or road

constructions) as the information required to verify the quality of the concrete are exactly the

same one.

Moreover, the usage of the solution by other stakeholders, such as contractors or responsible

of the testing operation, is straightforward, as they need to know exactly the same data that in

FITMAN are provided to the inspectors. As the main Trial Platform supports workspaces for




different categories of stakeholders, from these workplaces the users could be redirect to iLike

to visualise the information that has been entered into the system and that are of interest for

all of them.

Some improvements are suggested for the feature, to make the solution more complete and

therefore interesting for the market:

From the main platform of WP5, users should have the possibility to directly access,

through a button or similar, the iLike interface for the visualisation of the information

entered in the system

The guide number should be used as the unique identifier for the concreting operations

Filters and also drop down box should be available in the zones

Finally the solution should be able to manage several projects and thus to visualise

information about more than one dam.

Finally, a strong recommendation is to always verify the correctness of data transfer between

the main trial platform and the iLike interface, as the visualised information are very critical

and end users should fully trust the system.

2.5.2 Interviews in AIDIMA


The interview with AIDIMA has been conducted on GoToMeeting and attended by Maria

Josè Nunez and Fernando Gigante.

During the interview, all the questions of the first three sections have been analysed and

answered. The written answers have been provided on the day after the web meeting.



AIDIMA provided an overall very good evaluation of the platform as it has been delivered at

the end of the project.

It provided comments and feedbacks both from the IT integrator (here reporting also the

comments from UPV) and end user point of view (here representing the associated SMEs

from the furniture sector).

As an IT integrator and end user, the solution is considered correct; no defects remained after

some interactions with the IT developers to remove problems encountered initially in

configuring correctly the VM containing the integrated prototypes. Some adaptations have

been performed to successfully apply it to the existing environment, merely to solve

configuration issues.

The overall efficiency and reliability of the solution is considered high, considering that the

evaluation has been conducted on a Virtual Machine with quite important requirements. In

general, performances highly depend on the number and nature of widgets that are

simultaneously embedded into an Obeya rooms. Some of the SEs, such as the 3DWV, might

require some time to complete the computation/transformation tasks and thus are slower in

responding.

Configuration required some effort and time in the first evaluation (for the installation of the

first VM) but has been consistently improved for the configuration of the final version, and




now AIDIMA says that the documentation provided was useful and adequate to perform the

installation and configuration of the tools. .

The business processes to be supported in the AIDIMA trial were:

1. Understanding customer requirements: Analyse external and internal information

sources such as social information analysis, trends, sales information, exchange of

ideas, etc.

2. Project management and functional design brief: Creation of a design brief

involving design specifications, environmental aspects, CAD files, manufacturing

issues, quality levels, price, etc., and exchanging opinions about.

3. Iterative sketch development and technical design rollout. Preparation,

presentation and selection of sketches according to design brief. Once approved,

technical design is generated involving cost calculation, BOM, revision and

validation of technical design until its final version is achieved.

All these business processes are fulfilled by the solution and no critical functionality is

missing in the system.

However, AIDIMA suggested some possible improvements for the future (see below) and

recommended to improve the documentation and training material.

The solutions are easy to understand and to learn by AIDIMA technical people, as end users,

with the provided manuals.

For SMEs, the learning processes could be long and presently requires external support, thus

it is suggested to improve the available documentation for the final users: the information in

the FI-WARE catalogues are good for IT developers and system integrators, but the final end

users (especially designers) need videos and tutorial to learn how to work with the integrated

solution-by examples.

As a suggestion, in the case of VO, tooltips in some actions could be included to ease the use

of the tool (i.e.: the “edit Obeya” control box).

The C3DWV is a bit more complex for end users so it provides more controls. Furthermore,

SEMed requires specific training.

Concerning the efficiency of the solution, it is difficult to make a benchmarking with similar

solutions already in use, as SMEs and designers are nowadays collaborating by exchanging

fields via cloud-based solutions, while the collaborative meeting are usually organized in

skype.

Thus, the collaborative platform offered by WP13 is a completely new approach that

rationalize and improve dramatically the as is situation. Moreover, the end users perceive as

of very useful not only the synchronous collaboration support offered by the virtual Obeya,

but also the possibility to access ad different times to the information sources and tools

embedded in the Obeyas room: that is particularly useful when people in the design team are

distributed geographically and might have difficulties in meeting together at the same time,

due to different time zones.

A very positive feedback on the interest of SMEs for the VO and for the integrated solution

provided by WP13 has been collected by AIDIMA during a workshop organized on 29th

September with associated SMEs. Some of them expressed their interest in attending a

training event organized by AIDIMA, to better learn how to use the solution.





The documentation provided in the FI-WARE catalogue supporting the SEs (SEMed and

3DWV) is considered of high quality and complete enough for the IT developers and

integrators. As mentioned, for the final end users it is expected, to introduce them to the usage

of the overall integrated platform and to give them a flavour of the possible application of the

SEs.

It is suggested that the documentation provided should include some FAQs in order to give

some clues about how to solve the most common issues when installing, configuring or using

the SEs. This could be considered a dynamic registry which can be built according to the user

experience.

2.5.2.2.3 Evaluation of the single SEs and PCs

SEMed For AIDIMA, SEMed provides significant flexibility in terms of connection of data sources

and performing specific queries.

The graphical interface provided by SEMed for the configuration of the data source is user

friendly. However, mapping and onto-related files need to be created by hand.

No missing features have been found at the moment. Further tests are useful to identify

specific missing functions.

In principle only Materializa (one feature of SEMed) was found to be useful to be integrated

in the collaborative environment but other kind of data sources such as ERP-related could be

made accessible via any tool by using SEMed.

In terms of applicability of the SE, it is commented that SEMed could be used to integrate

semantics in the learning platform used by AIDIMA to offer training services to its associates.

iLike

Designers benefit from the information provided by iLike through the prototype widgets.

Thus iLike is used as a backend tool that stores information and feed the prototype widgets; it

also communicates with the 3DWV and with Materializa, through the SEMed. Thus the iLike

interface is considered appropriate to this end.

As a remark, AIDIMA reports that the data model presented by iLike is not as flexible as

AIDIMA would like. In principle there are only 3 categorisation levels but in furniture

product data categorisation is not enough, thus additional levels should be introduced.

In terms of applicability to other processes, AIDIMA can be successfully applicable to other

domains in addition to the furniture one. Considering the scope of the scenario proposed,

iLike can be also useful to model data about production processes although this is not

considered for the proposed scenario

Virtual Obeya (VO)

The collaborative meetings approach provided by VO fulfils the requirement of having a

collaborative space supporting iterative sketch development and technical design rollout,

preparation, presentation and selection of sketches according to design brief, cost calculation

of approved design, BOM, revision and validation of technical design until its final version is

achieved. This approach is applicable to any other challenge: its success will depend on the

implementation/selection of appropriate widgets embedded in the Obeyas. AIDIMA suggests

that it can be used to support training/learning processes related to manufacturing and other

disciplines.




The overall evaluation of the VO is high and AIDIMA can imagine running collaborative

meeting using the VO tool.

Some future improvements (beyond the expectations for the project) are suggested

The sharing of documents is directly implemented in VO for Google Docs; something similar

should be provided also for OneDrive, as presently it requires a tricky method to be achieved.

The visibility of widgets in the Obeya could be associated to roles, so that only selected

categories of users can see and use some widgets in the Obeyas.

Tracking of completed sessions could be included (i.e.: session log, modified widgets, logged

users, session length). The system could send notifications to all the users involved in the

session to inform about changes.

C3DWV

AIDIMA reports that the remote visualization and collaborative approach provided by

C3DWV supports the selected business processes and requirements: indeed the 3D

visualization increases the efficiency of the arranged meetings, mainly those in which

designers are involved as 3D models can be directly evaluated during the remote meeting

sessions.

2.5.3 Interviews in VW


The interview with VW has been conducted on GoToMeeting and attended by Marek Eichler

as end user, Frank-Walter Jäkel und Jan Torka as integrators and Marco Franke as

interviewer.

The questionnaire was sent to the end users and integrators before the interview appointment.

The filled out questionnaire was provided before the interview by IPK/VW.

During the interview, an update of the answers were discusses and minor changes created.



VW and IPK provided a good evaluation of the platform as it has been delivered during the

project.

The interview provided comments and feedbacks both from the IT integrator (IPK) and end

user point of view (VW).

As an IT integrator and end user, the solution is considered as relatively minor defects. The

classification was chosen according to the missing full automation of the workflow and the

varying performance in the solution. The ease of application was classified by IPK between

the range: “Applicable with significant amount of work” and “Applicable with some

adaptions”. IPK mentioned that the integration of all components occurred in work, which

required the support of the SE/PC owners. Faster response times and a more detailed

documentation could be possible approaches, which would improve the ease for the next time.

The overall efficiency was considered as expected for such a solution. In contrary, the

reliability varies significantly according to its performance. The usage of the whole solution




in a collaborative environment requires a higher performance. The latency of specific

components varies too much, which are described in detail in the SE specific sections.

The effort for configuration and integration of all components were classified as high. IPK

mentioned that they haven’t used the documentation of the FITMAN catalogue. The

configuration of all components the configuration requires the involvement of the solution

providers. Finally, the instantiation and integration was possible and the solution’s efficiency

was good as expected.

The business processes are fully supported. VW explained that the requirements according to

the web based availability of the Machinery Repository (MR) with a GUI to all relevant

persons and the aggregating and abstraction of data from different sources are fulfilled in a

good way. From the perspective of VW, the usability of the solutions differs between the

contained components. The solution provides functionalities to create, submit and evaluate

inquiries and to manage the MR content. Most of these functionalities are easy to understand

and to learn. Only the data aggregation workflow with manual XML export from the PLM

system is a bit more complex. In this case, VW mentioned also the missing fully automated

data extraction process. BIBA mentioned that this is possible but would include a multistage

extraction process and more extended capabilities in SEMed. The performance of C3DWV

was mentioned as fact for the hard usage of WP13 solution. VW mentioned that the overall

solution improved the business processes according to its grouping of all important services

and data. Furthermore manual tasks were (semi) automated, which reduce the effort. The

efficiency of business processes could be more improved through adding the fully automated

tasks for the data integration, the intelligent support for the evaluation functionality in VO and

the full-JT support to import 3D models without the necessity of manual conversion tasks.

The provided documentation was not sufficient to solve all integration issues of VW and IPK.

Instead, VW and IPK take advantage of the support by BIBA, DFKI, and Holonix. In so

doing, a couple of telephone conferences and physical meetings were hold. The direct contact

to the developers of the solution provider was necessary to solve integration issues, such as

for example the transferring process of OPC to IPK cloud of C3DWV and a non-terminating

SEMed instance were issues. All these issues could be solved by DFKI and BIBA.

2.5.3.2.2 Evaluation of C3DWV

IPK and VW define the role of the C3DWV as visualization of the machinery to provide the

engineer a better impression of the machinery, which corresponds to the planned role of

C3DWV. The satisfied requirement was to show 3D models from the PLM system based on

JT-files. This is currently only possible via manual mapping of the JT-file to XML3D. Apart

of the functional requirements, VW mentioned that the performance when used in IPK cloud

is not acceptable. The overall visualization approach for the business processes increases the

overall impression of the machinery, but does not affect the performance or efficiency

directly. Furthermore, VW mentioned that especially the collaborative part of the C3DWV

could be used for meetings and feedback in a global frame (e.g. planning discussions with

China) and therefore could be used in other business processes. To gain a further aid in the

business processes VW mentioned that a full JT-support is required. The usage of C3DWV in

future project is possible: “…Generally yes, due to the big number of 3d data during the

product development process. But JT support must be provided…”.

2.5.3.2.3 Evaluation of SEMed




IPK and VW define the role of the SEMed as the data extraction from “PLM XML export

file” into MR database, which corresponds to the planed role of the SE. The satisfied

requirement was to extract and generalize data from PLM into MR planning database. In

particular, the extraction of a single station’s information from PLM to MR is available.

Extraction methods, which required arithmetic capabilities, were not provided by SEMed but

by TSC (LogoLayout Extractor). This missing feature was mentioned by VW to enable a fully

automated data extraction process. The overall supported data integration approach slightly

improved the flexibility for connecting specific parts of data sources. IPK/VW mentioned also

that SEMed is also applicable for another integration challenge in their domain. The

evaluation of SEMed for another domain within the next two years depends on the future

evolution of the SEMed. One example would be the parameterisation by an ontology derived

from the end user needs. The creation of the ontology and mapping during the configuration

should be simplified and more end user driven. SEMed is an open source tool and VW

mentioned the application of open source in the production environment is general possible if

the security, trust and the maintenance aspects are ensured. In other cases, IPK/VW said the

normal way would be to create an own data integration solution on basis of the results of

SEMed.

2.5.3.2.4 Evaluation of VO

IPK and VW define the role of VO for the support of a professional user interface for the

business processes, which corresponds to the planed role of the PC. The satisfied

requirements were the support of PHP (Hypertext Preprocessor), providing of different user

roles and the display different widgets. The synchronization between widgets was a

requirement, which was not satisfied, but this missing functionality was known by IPK/VW

since the first introduction. VW mentioned that collaboration approach provided by the VO is

applicable for other remote cooperation challenges in the domain of DF. The combination of

VO and C3DWV is applicable for another business process of company/business

ecosystems/associates. To support business processes better, extended security support, such

as smartcards, would be necessary. IPK/VW mentioned that another evaluation within the

next year is foreseen but in general, they could imagine running collaborative meetings using

the VO.

2.5.4 Interviews in WHIRLPOOL


The interview has been answered by Pierluigi Petrali, Operations Excellence Manager,

belonging to the Manufacturing R&D division of Whirlpool during the last week of

September 2015 and by Mauro Isaja as ENG (IT provider) representative.



Whirlpool has provided, in general, a good evaluation of the integrated platform delivered

during the project, considering that the solution fulfils all the trial requirements. The objective

of the platform solution is to provide a 3D measure of a microwave fan.




ENG, as the IT provider, considers that there are no major defects in the platform and that the

correctness degree is highly satisfactory. From an applicability point of view, ENG considers

that the platform could be applicable to the user environments with a little amount of work

and some extra actions. By the way, the solution has proved efficient, being capable to

provide the appropriate performance with very reasonable resources consumption. Regarding

the openness, the platform shows a degree of interoperability maturity that can be defined as

Baseline Unified Approach (International Standard exist) in the case of 3DPCAP private

component and a level of Open Unified Approach (No international Standard exists) in the

case of the 3DScan SE, which means that there is a strong possibility to interact with other

systems. In terms of openness the solution, always according to Innovalia, shows a no

barriers, allowing developers to view and study the requirements and implement them as they

wish, in the case of the 3DPCAP private component. A high degree is considered of this

parameter, in the case of the 3DScan SE, allowing to consult with the use cases about their

needs and contribute to the source repository, designing documents and bug reports.

Whirlpool, considers that the platform solution is reliable enough to be used and keep a

specified level of service when used in the factory environment and settings. According to

sustainability, both, Whirlpool and ENG, consider that the software solution is easy to

maintain and modify.

The Business Processes are fully supported. Whirlpool evaluates that the four defined

requirements

access to point clouds stored

upload and retrieve the point clouds to the storage unit,

visualization of the point cloud stored

visualization of the 3D results through a colour mapping representation)

are covered by the solution proposed. Whirlpool in this sense considers that in case of the

visualisation of the point cloud stored and the 3D visualisation by means of a colour mapping,

the solution could cover a slight variation of the business processes. For the other two

business processes (access to the point cloud stored and upload and retrieve of the point

clouds stored) the solution could cover medium variations.


Whirlpool took advantage of DATAPIXEL’s configuration and did therefore not use the

provided administration documentation. In any case, some meetings were held, by telephone

and physically, to adapt the solution to the type of samples Whirlpool would use during the

trial. All these issues were solved together by Whirlpool and Datapixel, but implemented and

configured in the platform by Datapixel.

On the other hand Whirlpool used and found useful the user documentation.

Evaluation of 3DScan and 3DPCAPWhirlpool considers that 3DScan could be useful to solve

other industrial problems raised in its manufacturing environment. In this sense, Whirlpool is

fully open to use open source solutions in its production environment and could perfectly

adopt 3DScan or similar solutions in future projects. However Whirlpool suggests that

3DScan should include openness to other hardware solutions to be part of its IT infrastructure.

Whirlpool states that they could perfectly conceive a dimensional quality control system

based on 3DScan.

Whirlpool has used 3DPCAP PC always as an integrated subsystem of the platform solution.

So the evaluation of 3DScan can be assumed for 3DPCAP.




2.5.5 Interviews in TRW


The interview has been answered by Ignacio Arcona, IT Director of TRW production plant in

Pamplona and by June Sola from Innovalia (IT provider). The interview was answered during

the last week of September 2015.



TRW has provided a positive evaluation of the integrated platform delivered during the

project, considering that the solution fulfils all the defined trial requirements. The role of the

solution in the trial is to scan an endless spindle manufactured by TRW, provide a 3D point

cloud and analyse it in order to produce 3D results, including measurements and deviations.

Innovalia, as the IT provider, considers that the solution has a high degree of correctness; no

defects are detected in its specification and implementation. From an application perspective,

Innovalia states that the solution is applicable to its production and IT environment with a bit

of work in configuration. In terms of efficiency, the solution shows a high performance in

relation to the amount of resources used. By the way, Innovalia considers that the solution

provided has an Open Unified Approach (No International Standard exists) related to

interoperability maturity, namely the capability of the software to interact with other systems

in the case of the 3DScan SE. On the other hand, the 3DPCAP private component presents in

this aspect a Baseline Unified Approach (International Standard Exists). However, Innovalia

evaluates a high degree of openness with respect to the 3DScan SE and a low degree in the

3DPCAP private component, defining this level as the possibility of developers to view and

study the requirements and implement them as they wish.

In relation to reliability, TRW has considered for the solution a high degree of the software to

maintain a specified level of performance when used in the factory environment and settings.

Finally, both, TRW and Innovalia, consider that the software solution has a high degree of

sustainability, e.g. that the software solution is easy to maintain and modify.

The business processes are fully supported. TRW evaluates the defined business processes are

fulfilled. BP5: The manufactured parts are correctly digitalised through 3D scanning

technologies, is mainly done through 3DPCAP. Once this is performed the point clouds are

conveniently stored by the 3DScan SE and can be uploaded and retrieved from the storage

unit. 3DScan also satisfies the need to visualise the point clouds stored and to visualise the 3D

results by means of a colour mapping. 3DScan finally has been proved to be useful to

evaluate if a part contains dimensional defects or not. This is mainly due to the fact the

3DPCAP is able to measure the surfaces obtained and to perform the dimensional analysis

defined.


TRW has used the documentation provided, both the administration documentation and the

user documentation. According to the administration documentation TRW considers that the

documents provided clearly indicate the necessary steps to install and configure the SE and to




use it, in the case of the user documents. In this sense, TRW states to be satisfied with both,

the administration and the user documentation.

2.5.5.2.3 Evaluation of 3DScan and 3DPCAP

TRW considers that both, 3DScan and 3DPCAP, could be used to solve other industrial

problems, as the use of 3D Point Clouds is not only devoted to the automotive sector.

Therefore these technologies integrated or by separate, could be used in any industrial sector

in which parts are manufactured and need a high level of dimensional quality. In this sense,

these technologies could be used by TRW in future projects in the coming years. Regarding

open source solutions, TRW states that the automotive industry is a very traditional sector

where confidentiality of the data managed is an essential issue. In any case, if open source

solutions offer a huge stress on security issues, there is no reason why this technology cannot

be adopted in TRW’s production environment. However TRW suggests that 3DScan should

include access control and other security functionalities so it can be part of its IT

infrastructure. Concerning dimensional quality controls, TRW, as an automotive supplier, is

deeply concerned on dimensional quality so any system that can help to improve it will be

welcomed.

2.6 Lessons Learnt

Finally, the present report concludes with lessons learnt from the validation activities

conducted so far. Whereas the initial lessons learnt included in D13.3 resulted mostly from

the validation of the T13.2 results conducted within the WP13, by the IT partners, these

lessons learnt results from the interactions and the feedback collected directly from the Trail

owners, after their experimentations.

2.6.1 Lessons Learnt in AIDIMA

From the interactions with AIDIMA carried on during the evaluation period, several elements

have been extracted to create an analysis of strengths, opportunities but also bottlenecks and

further improvements for both the integrated solution developed in T13.2, as well as for the

SEs an PCs provided by the WP13 partners.

2.6.1.1 Strengths and Opportunities

From the end user perspective, the solution has been judged as mature and complete w.r.t. the

expected coverage of the business processes and the provided functionality: presently, it

offers a completely new way of working for the team of designers and furniture SMEs in a

cooperative way and no major functionality is missing. The SMEs that have been presented

the solution demonstrated interest and AIDIMA is evaluating how to continue the

experimentation with some of them, after the end of the project.

The final version is easy to be integrated in the trial environment and does not require specific

configuration effort.

As a demonstration of the interest proved by AIDIMA in experimenting with the solution,

below are reported some pictures of the collaborative environments created by AIDIMA in

the Virtual Obeya, as examples to be demonstrated to the associated SMEs:




Figure 26. TRW spindle used in the trial.

2.6.1.1.1 SEs/PCs specific benefits

Virtual Obeya The Virtual Obeya offers a new approach for the collaboration and it has been successfully

valuated by AIDIMA; it considers of great value the possibility of accessing the contents of

the Obeya both on synchronous and asynchronous modalities, thus overcoming the limitations

of web conference tools, currently used by companies to host remote collaboration sessions,

where it is possible to share documents or tool interfaces, but just once a time, with just one

partner having control of it and with no possibility of accessing the tools and the results of the

collaborative session.

AIDIMA expressed the interest to continue the experimentation of the tool, with the objective

of making it available for the associates, to conduct virtual meetings with them and possibly

explore the possibility to use it support the training activities.

iLike The widgets of iLike provide a simplified interface to interact with the underlying platform,

and the integration with the 3D viewer allows to easily visualizing within the VO the

prototype data that have been uploaded on iLike.

C3DWV

The C3DWV, integrated within the VO, offers a valuable service to designers that

collaboratively work on a project and through this SE can visualize the 3D models they are

ideating, making them visible to all the participants to the meeting. Moreover, in the provide

solution, the 3D models are directly linked to the technical data of the prototype, thanks to the

integration with iLike.

SEMed The SEMed has been positively evaluated by AIDIMA from the point of view of IT provider

and integrator, since it provides the integration of Materializa data within the widgets for the

prototype creation. Whereas further analysis and experimentation would be necessary to grasp

all the potentialities of the tool, the availability of a web interface for the configuration proved

to be helpful for the IT integrators




2.6.1.2 Bottlenecks and improvements

AIDIMA has identified some potentialities to further improve the solution and thus increase

the adoptability by the final end users.

Some of these improvements are valid for the overall solution, some other ones ae SEs/PCs

specific.

Overall solution

There are essentially two main barriers that could prevent the full adoption of the solution.

The first one is the easiness to learn how to work with the platform as a whole and how to use

the single elements. The preparation of documentation and training material, in the form of

videos, online help and tutorials is strongly recommended.

Also, the simplification of the installation procedures and the formalisation of the suggestion

provided to AIDIMA by email, phone calls and written instructions during the evaluation

would be highly beneficial.

Furthermore, the responsiveness of the collaboration environment significantly reduced when

more than 5-6 widgets are embedded in the same Obeya. However, it is recognised that this is

an extreme situation, and in any case other tools currently used for collaborative meeting are

offering more severe limitation, as the web meeting tools usually allows sharing one screen

and cannot offer all participants the possibility to interact with the visualized document or

tool.

SEMed

Even if ADIMA has not experimented with the configuration of SEMed directly, as the

configuration for Materializa has been provided by BIBA, it is recommended to make it as

intuitive as possible and to reduce the steps to be performed manually.

Virtual Obeya

After having expressed its satisfaction for the tool, AIDIMA suggested some directions for

further improvement of the tool, to improve its usability and flexibility. Below are the initial

suggestions:

Simplify the mechanisms for sharing documentation, extending the mechanism of

document template creation already available for Google Docs

Introduce a mechanism for role-based control of the visibility of widgets in the

Obeyas

Clarify and, if necessary, improve, the level of security of the applications: companies

would like to know how secure is to share information within the Obeya; it is

recommended to guarantee the highest possible level of security, prevented

unauthorized access to these information.

A mechanism for recording the activities performed by participants during a virtual

meeting, in particular as for the modifications operated in the shared information and

documents, would be highly beneficial.

C3DWV

AIDIMA recommends improving the uploading mechanism of files, removing the problems

with COLLADA models that sometimes appear and, possibly, to reduce the computational

time that makes the application a bit slow (at least when embedded in the VO). Moreover,




since 3D models cannot be changed directly in the viewer, an annotation service, able to

collect the comments and request of modifications of the user would be highly beneficial.

2.6.2 Lessons Learnt in CONSULGAL

During the evaluation period CONSULGAL conducted several very exhaustive testing

sessions, that complemented the testing performed by the IT partners: this has been very

important as the end user was able to detect some bugs and inconsistencies of data that only

someone knowing the semantic of the data could detect and that guided the completing of the

integration between the Trial platform and the WP13 solution.

From these testing reports and from the above reported interview, BIBA and Holonix

extracted very useful information to create an analysis of strengths, opportunities but also

bottlenecks and further improvements for both the integrated solution developed in T13.2, as

well as for the SEs an PCs provided by the WP13 partners.


Overall solution

From the end user perspective, the solution has been judged as, literally, “nearly perfect” as it

fully covers the business process BP4 and offers access to all the necessary information, in the

way suggested by CONSULGAL.

SEs/PCs specific benefits

SEMed

The role of SEMed as data integration solution achieved good results and its applicability for

other business processes were assigned by UNINOVA in its role as integrator. Furthermore,

SEMed’s integration capabilities were recognized as “…quite independent of the domain of

business scenarios,,,”. Both assignments increase the chance of application of SEMed in near

future projects. ..

iLike

Its usage is highly intuitive and no specific training is required to start using the solution. This

is due to the fact that the information is visualized in the way inspectors need them.

Moreover, the iLike interface is of interest for other stakeholders in addition to the inspectors

(contractors, testing labs) that could access exactly the same information, with no need of

implementing filtering mechanisms to hide/display data depending on the user role.

Similarly, the same solution, with almost no major modification, could be used to support the

concreting control process for other projects.


No specific bottlenecks are reported by the end user, and also the performances in terms of

responsiveness and reliability of the solution are positive.

CONSULGAL has identified some potentialities for further improvements that could improve

the adoptability by the final end users.

Such improvements are related to the overall solution, with no distinction between SEMed

and iLike components.




Improvements of the overall solution

Basically, it is envisaged to use the same solution for accessing information related to

different projects and constructions that are supervised by CONSULGAL or similar

organisations, as inspectors normally control more than one concreting process and on

different projects.

A seamless integration between the Trial platform interface and the iLike one is missing and

might cause confusion or trouble to the end user to access two different systems: this can be

easily solved by adding an icon or a tab in the trial platform interface to redirect the user to

the visualization of data in iLike.

In addition, searches should be as intuitive and quick as possible and here having dropdown

boxes and auto completion of words could be useful.

From the point of view of BIBA and Holonix, an important lessons learnt is the importance of

acquiring a deep understanding of the meaning of the technical data managed by

organisations such as CONSULGAL, so to be able to test the correctness and coherence of the

data acquired from external systems and visualized in iLike before making the system

available to the end users.

In addition to that, the involvement of the end users in all the phase of the development for

early validation is extremely important to be sure that the way data are visualized or searched

into the system is fully responding to the end user requirements and the interface is highly

usable and intuitive for the final consumers of the data.

2.6.3 Lessons Learnt in VW

Information has been extracted to create an analysis of strengths, opportunities but also

bottlenecks and further improvements for both the integrated solution developed in T13.2, as

well as for the SEs an PCs provided by the WP13 partners


Overall solution From the end user perspective, the fulfilled requirement has been judged as “…fulfilled in a

good way…” In particular the provided functionality “…web based availability of the MR

with a GUI to all relevant persons and the aggregating and abstraction of data from different

sources…” covers the addressed business processes. The impact of the solution was

summarized as “… The solution is more efficient due its grouping of all important services

and data. Furthermore manual tasks were (semi) automated….” In addition the overall

solution, each of the single solutions was suggested to be applicable to other business

processes/ domains in the DF.

SEs/PCs specific benefits

C3DWV The C3DWV offers a valuable service to visualization of the machinery to provide the

engineer a better impression of this machinery. Through the web-based access, the

visualization increases the overall impression of the machinery. In particular, the

collaborative part of the C3DWV could be used for meetings and feedback in a global frame.




SEMed

The SEMed offers a valuable service to data extraction and abstraction from PLM into MR

planning database. SEMed was suggested by IPK/VW to be suitable for data integration

challenges in other business processes and domains.

Virtual Obeya

The VO offers a professional user interface for a web based collaborative environment which

offers in combination with C3DWV a valuable service. IPK/VW mentioned that the

combination is applicable for other business processes and business ecosystems in the domain

of DF.


Overall solution

IPK/VW mentioned that the performance is varying between the SEs/PCs in the solution and

results as a common outcome. It was estimated as too low. The performance of the

visualization is an important bottleneck, which should be prioritized and solved. Apart of the

performance some functionality should be extended to increase the semi-automated processes

to a fully automated which in particular address the 3D models and the PLM data extraction

processes.

C3DWV The non-functional bottleneck of this SE is the performance within the proposed cloud

infrastructure. A detailed investigation of reasons and corresponding adaption would improve

the applicability in daily usage significantly. Apart of the performance, the semi-automated

transformation of a JT model into a XML3D model to a fully automated transformation would

also increase the applicability for the spontaneous usage of C3DWV in daily usage.

SEMed The functional bottleneck of this SE is to provide additional functionality to enable not only a

semi-automated but also fully automated data integration. In particular, an arithmetic function

is required to count amounts of information in data source and to be capable of offer an auto

increment function for keys in the information forwarding processes. This extension would

enable the fully automated data integration approach so far. Apart of the functional

capabilities of SEMed the creation process of the configuration must be simplified and driven

towards the perspective of the end user. In particular, this focuses on the creation and

maintenance of the information models that are the basis for SEMed.

Virtual Obeya

The functional bottleneck of this PC is to provide the capability to synchronize the data

between the widgets.

2.6.4 Lessons Learnt in TRW and WHIRLPOOL

The feedback from TRW and WHIRLPOOL was submitted in a slightly different structure

which was maintained in order to remain authentic.

http://www.dict.cc/englisch-deutsch/prioritised.html




Good results:

The trials performed within FITMAN have shown advantages and positive results with

respect to the technology implemented. The following paragraphs point out the main positive

aspects of the technology in both, the TRW and the Whirlpool Trials.

Easiness to configure the solution:

Both trials have shown that the solution does not present great difficulty to be configured. It’s

true that a first configuration could require some specific know-how, but this can be easily

learned by the IT personnel of the final user and be deployed for future configurations or

maintenance. It is believed that this is a great advantage in comparison to other existing

solutions that may require much more know-how to configure the solution.

Expected acceptance:

The solution presented has been accepted by the end users of both trials, mainly due to the

added value that the information obtained can offer, combined with the fact that it is easy to

use and to adopt. From this point of view, the solution has been accepted in terms of usability

and adoptability, as it does not present a significant learning curve for the quality experts of

the final user. As well, it does not present a great challenge for the IT department in terms of

infrastructure and deployment. These two issues (added value and easiness to adopt and use)

facilitate its acceptance in the end user organisations.

Added value:

The results obtained by the deployed solution have an important added value for the final

user, as they deliver dimensional information about the production explaining if the analysed

components accomplish with the defined specifications. This can be one of the key processes

to advance to a Zero Defects Factory paradigm, where the whole production can be controlled

and defective parts identified easily in the production plant, as the TRW trial shows. The

Whirlpool trial, on the other hand, shows that the production can be controlled, from a

dimensional point of view, attending to trends that could help to anticipate and predict future

deviations.

Open issues and suggestions for further improvements:

The implementation of the technology in the two trials has left some open issues or unsolved

problems. These issues can be used to do some suggestions for further improvements that

should finally enhance the solution for next implementations in industry applications.

Integration:

In the Whirlpool trial the integration of 3DScan and 3DPAP with the statistical analysis

system (Dynamic CEP) shows the importance of standardization of data formats. This

standardization includes 3D Point Clouds and CAD models. In both cases it would be equally

essential to define and use standardized models in the market. Currently there are no

standardized models available in the market and this has been clearly a weak point that in the

near future that could be easily improved by defining common standards. In the Whirlpool

trial, to integrate both systems, the developers have finally agreed to use ASCII format for the

3D Point Clouds and stl for the CAD models. The colour mapping has been shared using a

specific de facto standard used for this kind of deviation maps.




Traceability of the samples:

Both trials, especially the Whirlpool one, have shown that it would be very interesting to

connect the solution with a reliable traceability system. This could easily improve the global

quality system, as it would allow to have more control over the whole production and to

deploy hypothetical action plans, once a quality problem has been detected.

2.7 Conclusions

In T13.3, end users have been asked to experiment and evaluate the solutions that WP13

partners developed in T13.2 to support the business processes of the trials and implement the

requirements defined in T13.1.

These evaluations have been conducted by providing the solutions to the end users for testing,

either as VMs to be installed in the Trail environment (as for AIDIMA and VW), or as web-

accessible solutions, integrated with the main platform (as in the case of CONSULGAL).

In TRW the solutions has been implemented in the trial environment and in the case of

Whirlpool the analysed parts have been analysed in a platform were data transfer between its

components has been performed virtually.

Useful feedback has been collected through different channels: during the evaluation, end

users and IT partners asked for clarifications, additional support and reported bugs by email,

phone and web calls; this feedback has been very useful to identify bugs and minor

improvements implemented by the end of the project.

After the end of the evaluation, end users have been interviewed by the WP13 partners; the

structure of the interview covered both an overall evaluation of the integrated prototype and

more SE and PC specific questions that end users could (partially) answer or have been

answered by the IT partners responsible for the integration with the main trial platform.

The evaluation outcomes are, overall, highly positive and in all the trials the end users

confirmed the correctness and completeness of the solutions, which covered all the selected

business processes. The end users also suggested improvements for further developments of

the solutions and to better meet the expectations of the market and improve the adoptability of

the SEs and PCs.

During the evaluation and after an analysis of the feedback, useful lessons learnt have been

elaborated and will guide the future improvements of the SEs/PCs and potentially will support

the usage of the SEs in the FI-PPP.




3 Deliverable D14.5

3.1 Introduction

The objective of this document is to report on the lessons learnt and evaluation results of the

FITMAN Specific Enablers (SE) generated in the context of FITMAN Task T14.3

specifically focusing extended virtual platform.

The APR trial did focus on other components than MoVA and GeToVa and the

experimentation did already end a bit earlier, therefore this trial is not reported here.

Being an accompaniment document to the technical prototype there is one dedicated chapter

for each of the SE implemented; each of which includes:

- general information

- experiments done in the context of FITMAN Trials and their results

- lessons learnt

The conclusions provide an overview about how the SEs are implemented

3.2 Generation and Transformation of Virtualized Assets (GeToVa)

3.2.1 Short overview of GeToVA SE

Extracting knowledge from multiple data sources, representing it in a meaningful, structured way,

as well as clustering, visualization and transformation into various formats in order to support

interoperability is one large requirement manufacturing enterprises usually have. The data sources

can vary from webpages, e-mails, text documents, spreadsheets to news articles, collaborative

posts, and patents. The FITMAN Specific Enabler for Generation and Transformation of

Virtualized Assets is aiming at providing a state-of-the-art Information Extraction-driven

semantic tool for (semi-)automatic Virtualized intangible Assets in order to heavily reduce

manual data entry for the population of the FITMAN-CAM Specific Enabler. The GeToVA

Specific Enabler provides the following core functionalities:

1. Extraction of Virtualized Assets information from real-world semi-structured

enterprise and network resource;

2. Generation of semantic representation of Virtualized intangible Assets according to

ontological models;

3. Clustering of Virtualized intangible Assets enabling better search of such assets

4. Multi-format ontology transformation between various formats, mapping and

exchanging Future Internet (FI) data e.g. USDL

The GeToVA Specific Enabler is provided as a set of RESTFul services being implemented on

top of the FITMAN baseline VF Platform. The GeToVA services APIs have been designed as

fully compatible with FITMAN Platform components, namely the Data.SemanticsSupport for the

GeToVA multi-formation ontology transformation and the Apps.Repository for registration of the

assets generated by GeToVA. During the last months of the project, we have refined and extended

the GeToVA architecture taking into account the shortcomings discovered during the

experimentations in the trials. The high-level, final architectural of FITMAN-GeToVA is

depicted in the following figure:




Figure 27: GeToVA Architecture

FITMAN-GeToVA Specific Enabler includes seven components which are briefly recapped

below (a detailed description is available in D14.1, D14.2 and D14.3).

GeToVA includes several components:

1. Extraction - responsible for extracting information from real-world semi-structured

enterprise and network resource. It relies on the GATE system and includes as well

support for tagging and annotations. The Tagging and Annotations allow the user to define

annotations that can reused to automatically spot properties within semi-structured data.

Extraction is done either automatically (e.g. from LinkedIn profiles) or semi-automatically

using GATE Support.

2. Transformation - responsible for transforming the Base RDF formatted generated by the

FormatHandler into various formats, according to various ontologies. In includes

subcomponents such as the Europass Format Handler which is used to manage

structured data i.e. XML according to Europass Format and the Converted which is able

to transform between semantic formats (JSON-LD, XML, RDF) using SPARQL

Constructs. An Ontology Manager is used to create RDF data that is valid to the used

ontologies within our platform.

3. Clustering - which provides clustering of Virtualized intangible Assets

4. Search - provide Full Text search among our data

5. Database and Search Engine – for storing the raw and processed information

6. RESTful API - exposed in a unified RESTful API all the functionalities / components

mentioned above.

7. Dashboard – build on top of the RESTful API providing an intuitive user interface to

consume GeToVA functionality.




3.2.2 Experiments and Results

GeToVA has been deployed and used in the context of two FITMAN Virtual Factory Trials,

namely TANET and ComPlus as follows.

3.2.2.1 TANET

In the context of the TANET trial GeToVA is used to import suppliers and tenders from

unstructured and semi-structured sources (e.g. The Welsh Automotive Forum and

Sell2Wales). The integration and usage of GeToVA SE in the TANET trial is illustrated in the

following figure.

Figure 28: GeToVA integrated with MoVA in TANET trial

Being integrated in the TANET trial, GeToVA provides the following functionalities. Given a

set of suppliers and tenders GeToVA is able to semi-automatically extract information about

these companies from unstructured and semi-structure data sources such as raw documents

and web sites. The information extraction is performed using the knowledge Extractor

GeToVA component and then represented internally in GeToVA as RDF using the Ontology

Manager component. Information is also transformed in other formats using the Converter

component. The information in RDF is than imported into our sister SE, MoVA which offers

additional functionalities for the SME Cluster in TANET (see section on MoVA for more

details).

GeToVA was also used to cluster, i.e. create groups of suppliers with similar profiles. A

cluster created by the GeToVA Clustering component using TANET trial data is shown in the

following figure.




Figure 29: TANET cluster generated by GeToVA

For the TANET trial we generated 143 tender opportunities from the Sell2Wales-website. An

example of a generated virtual asset is listed below.

_:g2157029680 <http://fitman.sti2.at/company/hasLegalName> "CastAlum\n" .

_:g2157029680 <http://fitman.sti2.at/company/hasDescription> "Diecast and

machined alumnium components, Design for manufacture\n" .

_:g2157029680 <http://fitman.sti2.at/company/hasWebsite>

"www.Castalum.com\n" .

_:g2157029680 <http://fitman.sti2.at/company/hasLegalAddress> "Buttington

Cross Enterprise Park\nWelshpool\n, Powys, SY21 8SL\n" .

_:g2157029680 <http://fitman.sti2.at/company/hasHQAddress> "Powys, SY21

8SL\n" .

Benjamins-MacBook-Air:ditf benjaminhiltpolt$ cat 'Magor Designs

.rdf'

_:g2157293240 <http://fitman.sti2.at/company/hasLegalName> "Magor

Designs\n" .

_:g2157293240 <http://fitman.sti2.at/company/hasDescription> "Design

Engineering, Precision Engineering\n" .


"www.magordesigns.co.uk\n" .

_:g2157293240 <http://fitman.sti2.at/company/hasLegalAddress> "Neath Vale

Business Park\nResolven\n, Neath, SA11 4SR\n" .

_:g2157293240 <http://fitman.sti2.at/company/hasHQAddress> "Neath, SA11

4SR\n" .

In addition we have generated 180 assets out of companies description in unstructured format.

An example is given below.

_:g2181915040 <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>

<http://fitman.sti2.at/company/Company> .






_:g2181254880 <http://fitman.sti2.at/company/name> "Rumm Ltd" .

_:g2181254880 <http://fitman.sti2.at/company/hasLocality> " Mid Glamorgan"

.

_:g2181254880 <http://fitman.sti2.at/company/country> "Wales" .

_:g2181254880 <http://fitman.sti2.at/company/postalCode> " CF82 7EH" .

_:g2181254880 <http://fitman.sti2.at/company/hasWebsite> "

http://www.rumm.co.uk " .

_:g2181254880 <http://fitman.sti2.at/company/hasMail> "[email protected]" .

_:g2181254880 <http://fitman.sti2.at/company/locatedInRegion> " Ystrad

Mynach" .

_:g2181254880 <http://fitman.sti2.at/company/hasStreetAddress> "Tredomen

Gateway Centre Tredomen Business Park" .

_:g2181254880 <http://fitman.sti2.at/company/category> "Engineering and

Technical Development Services" .

_:g2181254880 <http://fitman.sti2.at/company/category> "Other Support

Organisations" .

_:g2181254880 <http://fitman.sti2.at/company/category> "Support for

management, productivity, accreditation and IT" .

Since the previous release the Extractor component was further developed to support the

needs of the TANET Trial. The extraction from LinkedIn individuals and companies profiles

is available. It extracts information automatically and makes it available as JSON. For

example extracting the https://at.linkedin.com/in/ioantoma LinkedIn profile will result in the

generation of the following JSON data:

Figure 30: Individual Profile in JSON extracted by GeToVA from LinkedIn

For the TANET trial we further increased our functionality by providing fully automated

LinkedIn profile page extraction. This is done by scraping LinkedIn profiles and store the

information as JSON. The JSON is available on our platform and accessible via our REST-

API. To access this functionality a simple Web frontend is available at

http://fitman.sti2.at/tanet_linkedins. The frontend allows the user to provide URLs to public

profiles, which are then scraped. It further allows to show, edit and remove scraped profiles

https://at.linkedin.com/in/ioantoma

http://fitman.sti2.at/tanet_linkedins






The profiles for the TANET trials are available as REST endpoint at

http://fitman.sti2.at/tanet_linkedins. To extract the profiles we used a scraper based on the

opensource software available at https://github.com/yatish27/linkedin-scraper which is only

able to crawl public LinkedIn profiles. The REST-API allows the following methods:

• Base URL: http://fitman.sti2.at/tanet_linkedin

• GET /tanet_linkedins/:id to retrieve a specific LinkedIn profile (or all if no id is

provided)

• PATCH /tanet_linkedins/:id to update a LinkedIn profile

• PUT /tanet_linkedins/:id to manually add a LinkedIn profile

• DELETE /tanet_linkedins/:id to delete a LinkedIn profile

Further the endpoint: http://fitman.sti2.at/scrape_linkedin triggers the scraping of a provided

URL. One needs to do a GET on http://fitman.sti2.at/scrape_linkedin using as parameter the

LinkedIn URL of the form https://at.linkedin.com/LINKEDIN_PROFILE. Figure 31 shows

the extraction of a LinkedIn profile from the GeToVA dashboard.

Figure 31: Individual Profile in JSON extracted by GeToVA from LinkedIn

3.2.2.2 COMPlus

In the context of the COMPlus trial GeToVA is used for enrichment of the knowledge base

use by the network manager. By having a richer knowledge base, the network transparency is

improved and becomes easier for her/him to be aware of all possible choices of business

partners and chose the most appropriate ones for their business network. The integration and

usage of GeToVA SE in the COMPlus trial is illustrated in the following figure.



https://github.com/yatish27/linkedin-scraper

https://github.com/yatish27/linkedin-scraper

http://fitman.sti2.at/tanet_linkedin

http://fitman.sti2.at/scrape_linkedin








Figure 32: GeToVA in COMPlus trial

Being integrated in the COMPlus trial, GeToVA provides the following functionalities. Given

a set of company profiles GeToVA is able to semi-automatically extract information about

these companies from unstructured and semi-structure data sources such as raw documents

and web sites. The information extracted includes the company name, type, location, web site

address, industry branch, etc. Such information is extracted using the Knowledge Extractor

GeToVA component and then represented internally in GeToVA as RDF using the Ontology

Manager component. In this way we can generate structured, semantic representations of the

companies profiles which will be used by COMPlus to enrich the knowledge base In addition

the information is transformed in other formats using the Converter component. The

information in RDF is than imported in the COMPlus ontological based where it can be

queried and reasoned upon for the COMPlus improved network transparency case.

In total we have processed a total of 76 LED company profiles and generated virtual assets

from them using GeToVA functionality. An example of COMPlus LED company information

generated by GeToVA is provided below.



_:g2173020640 <http://fitman.sti2.at/company/name> "spectral" .

_:g2173020640 <http://fitman.sti2.at/company/hasLocality> "Freiburg" .

_:g2173020640 <http://fitman.sti2.at/company/country> " Germany\n" .

_:g2173020640 <http://fitman.sti2.at/company/postalCode> "79111" .


"http://www.spectral-online.de" .

_:g2173020640 <http://fitman.sti2.at/company/hasMail> "info@spectral-

online.de" .

_:g2173020640 <http://fitman.sti2.at/company/locatedInRegion> "

Germany\n" .

_:g2173020640 <http://fitman.sti2.at/company/hasStreetAddress> "

Bötzinger Straße 31\n" .

_:g2173020640 <http://fitman.sti2.at/company/produces> " Arbeits- und

Leseleuchten " .

_:g2173020640 <http://fitman.sti2.at/company/produces> " LED-Strahler " .




_:g2173020640 <http://fitman.sti2.at/company/produces> " Lichtböden / -

decken " .

For the COMPlus trial we also made available the companies profiles as JSON. They are

available at http://fitman.sti2.at/complus.

3.2.3 Lessons learnt

During the second and last iteration of trial experiments we have faced a coupled of

challenges. We have imported data from different sources such as LinkedIn profiles, web

pages and text documents. Due to the high diversity of information sources from where we

want to extract knowledge, the Extraction component requires fine tuning and adaptation.

However once the right setup is done, extraction works with high accuracy. As experiences in

the TANET and COMPlus trials, GeToVA SE and its functionalities can be very easily used.

The RESTful API we provide covers the needs of the trials and MoVA, the other Virtual SE

developed in WP14. The dashboard also fully supports the users in consuming GeToVA

functionalities.

3.3 Advanced Management of Virtualized Assets (MoVA)

3.3.1 Short overview of MoVA SE

MoVA has been deployed and used in the context TANET FITMAN Virtual Factory Trial.

The architecture has already been described in previous reports. Thanks to the MoVA

flexibility it could be applied without any problem.

3.4 Experiments and Results

3.4.1 Data Modelling

In the context of the TANET trial MoVA is used to identify new clusters (groupings of

SMEs) responding to a new tender opportunity. The search needs to assure that the cluster fits

in terms of competence and tender requirements. Therefore the following structure was

modelled in MoVA.

http://fitman.sti2.at/complus




Figure 33 MoVA: Start screen and model

The model of the cluster and the lessons learned have already described in earlier reports

(D14.3). Throughout the trial all partners acknowledged the flexibility of MoVA.

3.4.2 Importing

Using the MoVA Backend and Plugin API the import was implemented. This step was crucial

for the acceptance and the success of the trial as the import saves a lot of time (one KPI).

The Import Routines are implemented using the MoVA Plugin API and MoVA Backend API.

The plugin are integrated in the repository by creating a folder for the plugin in the

repositories plugin directory. In the file register_application_components.php the following

code activates the import code.

$plugInPath = '../plugins/tanet/';

{ # Additional Functions




$subPlugInPath = $plugInPath.'code/';

{ #

$r-

>register_JavaScriptFile($subPlugInPath.'importDomainsSuppliers.js');

$r->register_JavaScriptFile($subPlugInPath.'importTenders.js'); $r-

>register_JavaScriptFile($subPlugInPath.'importSuppliers.js');

}

}

The special import code for the three import functions is integrated in three JavaScript Files

which defines the menu structure for calling the import which is running on the server. Each

of the three import function has its own file with its own server side code.

function importSuppliers() {

[…]

}

function importTenders() {

[…]

}

function importDomainsAndSuppliers() {

[…]

}

For importing the suppliers the JSON content from GeToVa URL

http://fitman.sti2.at/tanet.json is read and parsed into an array. For each new content element

there is created a new supplier in MoVA. This can be done very easy by using the MoVA

Backend API. The values of the supplier can also be set using the MoVA API.

#Create new object

$O_Suppliers = $OT_Suppliers->addObject();

[…]

$O_Suppliers->setAttributeValue_noCheck( $OA_UUID_Suppliers_Supplier_name,

"value_text", $companyName, true );

if (property_exists( $importedAttributes, "c:hasDescription" )) {

$O_Suppliers->setAttributeValue_noCheck(

$OA_UUID_Suppliers_Description, "value_text", trim($importedAttributes-

>{"c:hasDescription"}), true );

}

$addressParts = array();

if (property_exists( $importedAttributes, "c:hasLegalAddress" )) {

#ignore hasHQAddress as the data is the same as

hasLegalAddress, but with less information


$OA_UUID_Suppliers_Address, "value_text", trim($importedAttributes-

>{"c:hasLegalAddress"}), true );

$addressParts[] = $importedAttributes->{"c:hasLegalAddress"};

}

if (property_exists( $importedAttributes, "c:country" )) {

http://fitman.sti2.at/tanet.json





$OA_UUID_Suppliers_Country, "value_text", trim($importedAttributes-

>{"c:country"}), true );

}

else if (property_exists( $importedAttributes, "c:locatedInRegion" ))

{


$OA_UUID_Suppliers_Country, "value_text", trim($importedAttributes-

>{"c:locatedInRegion"}), true );

}

if (property_exists( $importedAttributes, "c:hasMail" )) {


$OA_UUID_Suppliers_Email, "value_text", trim($importedAttributes-

>{"c:hasMail"}), true );

}

if (property_exists( $importedAttributes, "c:hasWebsite" )) {


$OA_UUID_Suppliers_Website, "value_text", trim($importedAttributes-

>{"c:hasWebsite"}), true );

}

Figure 34: MoVA: Add new supplier

Figure 35: MoVA: Imported suppliers

For importing domain ontology the import function parses the Excel saved as CSV-File and

does the import. The plugin code used the MoVA Backend which is full object orientated.

Besides importing the domain ontology the suppliers related in the Excel are also imported, if

they are not yet in the system. For example they can also be imported by the




importSuppliers() function. Also the relations between domain entry and suppliers are set,

even if the suppliers are already in the system. Furthermore the import routine sets some

default values and does some mapping of data.

function importDomainsAndSuppliers() {

[…]

# read import file

$inputArray = readCSVinArray(IMPORT_DOMAIN_SUPPLIERS_SOURCE);

$companyInformation = array_shift($inputArray);

$qualityInformation = array_shift($inputArray);

$costsInformation = array_shift($inputArray);

$timeInformation = array_shift($inputArray);

$suppliers = array();

# generate suppliers in mova

for ($i=10; $i<count($companyInformation); $i++) {

#prepare information

$companyName = $companyInformation[$i];

$companyQuality = $qualityInformation[$i];

$companyCosts = $costsInformation[$i];

$companyTime = $timeInformation[$i];

{ # default values

if (empty($companyQuality)) {

$companyQuality = 0.25;

}

if (empty($companyTime)) {

$companyTime = 0.25;

}

if (empty($companyCosts)) {

$companyCosts = 0.25;

}

}

{ # generate supplier if not existing

[…]

$searchAttributes = array();

$searchAttributes['searchType'] = "must";

$searchAttributes['search_text'] = $companyName;

$Os_Suppliers_AllreadyInSystem = $OT_Suppliers-

>retrieveBy_attributeValues( array($OA_UUID_Suppliers_Supplier_name =>

$searchAttributes) );

if (count($Os_Suppliers_AllreadyInSystem)) {

[…]

$O_Suppliers = array_shift (

$Os_Suppliers_AllreadyInSystem );

$O_Suppliers = $O_Suppliers['object'];

$countUpdatesSuppliers++;

} else {

#Create new object

$O_Suppliers = $OT_Suppliers->addObject();




#Setting name


$OA_UUID_Suppliers_Supplier_name, "value_text", $companyName, true );

$countInsertsSuppliers++;

}

}

{ # setting quality, cost, time information


$OA_UUID_Suppliers_Time, "value_listKey", strval(floatval($companyTime)),

true );

[…]

}

{ # link supllier to facilitator

[…]

}

{ # update supplier

$O_Suppliers->update();

$suppliers[$i] = $O_Suppliers;

}

}

{ # generate domains in mova

foreach ($inputArray as $inputLine) {

set_time_limit(10);

{ # preparing domain entry

[…]

}

{ # generate domain if not existing

$searchAttributes['search_text'] = $domainName;

$O_Domain_Entity_AllreadyInSystem = $OT_Domain_Entity-

>retrieveBy_attributeValues( array($OA_UUID_Domain_Entity_Domain_Keyword =>

$searchAttributes) );

if (count($O_Domain_Entity_AllreadyInSystem)) {

$O_Domain_Entity = array_shift (

$O_Domain_Entity_AllreadyInSystem );

$O_Domain_Entity = $O_Domain_Entity['object'];

$countUpdatesDomains++;

} else {

#Create new object

$O_Domain_Entity = $OT_Domain_Entity->addObject();

#Setting name

$O_Domain_Entity->setAttributeValue_noCheck(

$OA_UUID_Domain_Entity_Domain_Keyword, "value_text", $domainName, true );

$countInsertsDomains++;

}

{ # setting additional values

$O_Domain_Entity->setAttributeValue_noCheck(

$OA_UUID_Domain_Entity_only_for_structuring, "value_listKey",

strval($domainOnlyForStructuring), true );

}

{ # link domain to domain - hierarchy

$currentDomain[$currentHierarchyLevel] =

$O_Domain_Entity;

if ($currentHierarchyLevel > 0) {




{ #lookup top level domain

$O_HigherLevel_Domain =

$currentDomain[$currentHierarchyLevel-1];

}

{ # check if relation allready exists

[…]

}

{ #Linking Domain to subdomain - if not

allready linked

[…]

}

{ #Setting attributes on relation

[…]

}

}

}

{ # link domain to supplier

[…]

}

{ #update domain

$O_Domain_Entity->update();

}

}

}

}

{ # output

$output = array();

$output['updatesSuppliers'] = $countUpdatesSuppliers;

$output['insertsSuppliers'] = $countInsertsSuppliers;

$output['updatesDomains'] = $countUpdatesDomains;

$output['insertsDomains'] = $countInsertsDomains;

return $output;

}

}

Figure 36: MoVA: Restful API Import suppliers

3.4.3 Cluster Search

Using the MoVA Backend API and Plugin API the cluster search was implemented. The

Cluster Search is started navigating to Requirements. Several Requirements are already

imported. The article which should be produced is set under Domain Keyword. Again, this

was already described in D 14.3.

3.4.4 Integrating MoVA with SME Cluster

Using the MoVA Restful API SME Cluster has access to the data stored in MoVA. The

Restful API offers a lot of simple functions, which can be combined to complex function. For

example for requesting all suppliers there are needed many requests: Asking for get all object

types, then get all objects of the found object type with the name Suppliers. After receiving a

list of all objects from type Suppliers for each supplier it’s necessary to do a request to receive

the attributes of the object type.




Figure 37: MoVA: Data of MoVA in SMECluster

Furthermore it’s possible to call the Restful API functions of the plugins.

The plugin code of the TANET plugin provides a cluster search. This cluster search can not

only be used by using MoVA directly but also by calling the Restful API function from

external system. This is done by the SMECluster Website to display the result of cluster

search directly in the website.

If calling the cluster search function there is returned a complex json array, which is parsed by

SME Cluster to do the output. Each level needs its own request:

{"suppliersWithBenchmark":{"domainName":"Foam

Manufacturing","subdomains":[{"consistsOf":{"calc":20,"value":

20,"unit":"%"},"domain":{"domainName":"Seating

Foam","subdomains":[]}},{"consistsOf":{"calc":20,"value":20,"u

nit":"%"},"domain":{"domainName":"Trim


nit":"%"},"domain":{"domainName":"Moulded

Foams","suppliers":[{"supplier":{"O_v_UUID":"7faa3052-1a57-

11e5-b0e9-02004e435049","name":"[…]

Ltd"},"benchmark":{"value":25,"unit":"%","parts":{"time":0.125

,"cost":0,"quality":0.125},"partsAsString":"Time: 0.25 \/

Cost: 0.25 \/ Quality:

0.25"}},{"supplier":{"O_v_UUID":"7d2d4ed9-1a57-11e5-b0e9-




02004e435049","name":"[…]"},"benchmark":{"value":25,"unit":"%"

,"parts":{"time":0.125,"cost":0,"quality":0.125},"partsAsStrin

g":"Time: 0.25 \/ Cost: 0.25 \/ Quality:

0.25"}}]}},{"consistsOf":{"calc":20,"value":20,"unit":"%"},"do

main":{"domainName":"Open Cellular

Foams","subdomains":[]}},{"consistsOf":{"calc":20,"value":20,"

unit":"%"},"domain":{"domainName":"Polyurethane

Foams","subdomains":[]}}],"clusters":false}}

Figure 38: MoVA: Restful API Result of Level 0 search



20,"unit":"%"},"domain":{"domainName":"Moulded

Foams","suppliers":[{"supplier":{"O_v_UUID":"7faa3052-1a57-

11e5-b0e9-02004e435049","name":"A2B Plastics

Ltd"},"benchmark":{"value":25,"unit":"%","parts":{"time":0.125

,"cost":0,"quality":0.125},"partsAsString":"Time: 0.25 \/


0.25"}},{"supplier":{"O_v_UUID":"7s2d4ed9-1a57-11e5-b0e9-

02004e435049","name":"Applied Component Technologies

(ACT)"},"benchmark":{"value":25,"unit":"%","parts":{"time":0.1

25,"cost":0,"quality":0.125},"partsAsString":"Time: 0.25 \/


0.25"}}]}},{"consistsOf":{"calc":20,"value":20,"unit":"%"},"do

main":{"domainName":"Open Cellular




unit":"%"},"domain":{"domainName":"Seating



Foam","subdomains":[]}}],"clusters":false}}




20,"unit":"%"},"domain":{"domainName":"Moulded


unit":"%"},"domain":{"domainName":"Open Cellular




unit":"%"},"domain":{"domainName":"Seating



Foam","subdomains":[]}}],"clusters":false}}





Figure 41: MoVA: MoVA Cluster Search Result in SMECluster

3.4.5 Lessons learned

Before you can start the implementing the import routine you’ve to get familiar with the

MoVA Plugin and Backend API. If you’ve done this orientation you can implement the

import very easy.

The data provided by GeToVa for the suppliers is very good structured. So it was very easy to

specify the import. Also the data for the Assets and Domain Entities seems to be good. The

import could be done very easy. The GeToVa fields are mapped to the MoVA structure. By

inspecting the assets and domain entities there are was found a lot of unstructured data, which

GeToVa itself couldn’t filtered out, because the source data itself contains this wrong

unstructured data. Therefore the imported data has to be inspected and validated by hand.

The structure of the Excel file is very good, as it is easy to parse by the import routine. The

development of the plugin was very easy. The mapping to the suppliers, which are imported

from GeToVa wasn’t always possible, as the suppliers were listed with different (full and

shorted) name of the companies. Due to this, some companies were create two time, although

they are indeed the same.

As only a few requirements contain a link to a Domain Entity a cluster search can only started

for this small amount of requirements. The cluster search is working good and results suitable

data. As the Domain Entity model is still not fully built, the cluster search is only working for

level 0 and 1 at the moment. A facilitator has a lot of experiences which suppliers work well

together and which do not.




The next steps were to clean up the Domain Entities from the wrong data and to build up the

full domain hierarchy. After that the cluster search can be evaluated again to test the results.

MoVA stores actual clusters responding to a requirement. The facilitator can evaluate their

overall performance [0%-100%] and their performance in terms of time, quality and cost.

This evaluation is considered when suggesting new clusters.

Using the MoVA Restful API the SME Cluster website has access to the data stored in

MoVA. The presentation is done very nice.

The MoVA SE proves to be very useful for the TANET trial as it combines hard facts

(domain keywords, sub-domains) with the human way of working (searching for either high

quality or quick delivery). MoVA is able to handle such fuzzy values and to integrate them

into the technically exact model.

MoVA requires apparently a little experience in modelling complex information systems. The

GUI fully supports the system and shows step by step what is happening. Hence, it supports

the user in gaining this experience.

3.5 Interview with TANET

The following interview was conducted towards the end of the trial phase with the general

manager of TANET. The other trial, COMPlus, stated its experience already earlier in the

project.

Question: Could you please briefly describe the challenge your Trial was facing before the

implementation of the Specific Enabler of the open call?

Answer: We wanted to enhance the services of our SMECluster services by giving the cluster

managers tools ad hand that would improve the quality of the services why reducing time.

Q: Describe your experience in implementing MoVA and GeToVa, please.

A: The feedback I received from the Technical implementers at Control 2K was that it was

relatively easy to implement both SE’s into the SMECluster Platform. The support provided

was excellent.

Q: How well did the Specific Enablers support the business logic of your trial?

A: They have enabled the SME Platform to be extended to connect to the Automotive Forum

data and pull out the relevant suppliers from the association and allow Tim Williams who is

the Chief Exec of the operation to create clusters of companies who are able to tender for

business.

Q: How easy or difficult was the technical integration into your system?

A: I believe it was easier than the previous GE’s and SE’s that we worked with. Please

consult our Engineers or look at other deliverables for more detailed information.

Q: What are the key functions and benefits of MoVA and GeToVa? Could you fully exploit

them in your trial?




A: www.smecluster.com is live so you can see them operational in the platform. They are

helping the platform to provide much needed data in a format friendly to use. Especially the

clustering, the import and the advanced cluster search are beneficial. We are now also able to

integrate subjective evaluations and rankings of the cluster managers.

Q: Did the Specific Enablers open up your Trial by offering new opportunities and functions

you did not plan to exploit in the beginning?

A: In a way yes. We did need the added features but the initial GE’s and SE’s did not provide

the required functionality.

Q: Which additional features would you like to see developed for MoVA and GeToVa in the

near future?

A: Better integration with LinkedIn and tenders to be pulled from the Welsh Tendering portal

http://www.sell2wales.gov.uk/

3.6 Conclusions

In this deliverable we have reported on the final round of experiments, results and lessons

learned from using the two SEs developed as part T14.1 namely:

- advanced Management of Virtualized Assets (MoVA) aiming to support Virtual

Factories (VF) in intuitively generating, composing, and transforming virtual

representations of in-/tangible assets (VAaaS) within Manufacturing Ecosystems.

Design and implementation of an intuitive user-centric graphical interface for dynamic

discovery and flexible composition of Virtualized in-/tangible Assets (as a Service)

targeting at team building applications as well as advances in production networks;

- Generation and Transformation of Virtualized Assets (GeToVA) aiming to

support Virtual Factories (VF) in semi-automatic generation and clustering of

Virtualized intangible Assets (VAaaS) from real-world semi-structured enterprise and

network resources. GeToVa enables as well multi-format ontology transformation

between various representations of Virtualized in-/tangible Assets.

The Specific Enablers have been incorporated into the FITMAN Architecture and have been

used in the FITMAN pilots, especially in the Virtual Factories (WP6) that created the

specification for the open calls to which these SEs fulfil.

Both SEs contributed successfully and significantly to the success of the two trials where they

have been incorporated. Valuable insights to further streamline and optimise the SEs have

been gained.

http://www.smecluster.com/

http://www.sell2wales.gov.uk/




4 Update on KPIs in the Trials

In this section an update on the KPIs in the trials is provided.

4.1 AgustaWestland

AgustaWestland has performed the experimentation of SMART and DIGITAL trials. At the

end of this phase AW has been able to produce significant lesson learnt referring to the

collected results.

Trial results

Following is the list of gathered BIs, due to confidentiality purposes all Performance

Indicators were unified to percentage.

Trial - Scenario - BPI Progress_1 Progress_2 Progress_3

3-AGUSTAWESTLAND 70,4% 100,0% 100,0%

1- SUPPORT FOR MANAGEMENT OF DOCUMENTATION AND

REPORT CREATION 55,6% 100,0% 100,0%

ANDR_1-AVERAGE NUMBER DISCREPANCY REDUCTION 100,0% 100,0% 103,70%

RAT_1-REDUCTION OF AVERAGE TIME 11,1% 100,0% 100,0%

2- SUPPORT FOR MONITORING AND MANAGEMENT OF

TOOL TRACKING 100,0% 100,0% 100,0%

TDTM_2-TAILORED DATA FOR TRAINING MATERIALS 100,0% 100,0% 100,0%

Digital Trial

As a general comment the digital trails confirmed the initial expectations and the desiderata

target have been achieved.

With respect to the BI defined at the beginning of the project, a further BI was measured

relevant to a significant reduction of number of discrepancies between the analysed data from

different sources and complied by different persons, it’s linked especially to possible human

error of transcription inside the DB of Quality Production.

As regards the BPI “reduction of the average time”, the great improvement measured at

month 18 is due to various reasons:

at the beginning some GEs has been found not yet mature and not very stable.

consolidating of the application and fixing of the main issues that were present in the

first version tested at 12 months, in particular the connecting to the corporate

repositories;

tuning of the virtual machine where the application was running in order to manage

multiple services;

bugs fixing.

At month 27 further significant improvements were not measured even if some additional

functionaries were implements in the system such as possibility to filter the data sources or




the display of image of the components (if available). These new features, although

appreciated by the users, have not brought substantial improvements on the average time to

collect data.

Additional improvements could be obtained by increasing the number of sources dates

interrogated by the system.

As regard the BPI “Average Number Discrepancy Reduction”, the expected result was

already achieved at month 12. The application has replaced the “past and copy” of data from

company archives to the Db Quality Production handmade by users. Furthermore we must

consider that this DPI is time independent research data that is measured instead by the

previous PBI.

The result was confirmed at 18 months, while a slight improvement was measured at 27

months when an automatic consistency check between data sources was implemented. This

new feature checks if there are mismatches between the data from different sources (e.g.: a

different part number for the same component provided by different vendors) and in this case

it selects the data contained in the source that has been defined as master (in example the

IETP is the master for the part number). The other new features (filters and image of

components) have not affected the BPI.

General speaking, times saving as well as reduction of discrepancies can contribute to

enhancing the efficiency of the activities performed by the Production Quality department that

is committed in the preparation of all documentation required and necessary for the delivery

of the helicopter to the customer, for example to have a real helicopter picture of “as built” vs.

“as design”. This is very important for the post-delivery activities (spare parts procurement,

updating of technical publications, training, etc.) that constitute a major portion of the AW

business.

An interesting indication has emerged during the trial in particular saving time during

documents searching, the possibility to use the system to support the instructors of AWTA




(AW Training Academy) in preparation of the courses for pilots and maintenance technicians.

The availability of the information in a shorter time makes more efficient the activities of the

AW training organisation; the instructors, saving time in data collection, can focus on training

aspects and on preparation of training materials (training manual, presentation, multimedia,

etc.) used during type rating courses for pilots and technicians that are purchased by the

customer.

Purpose: The system searches the data linked to a specific helicopter through queries in 4

different sources and compiles the relative cell inside a Db of Quality Production

Business Performance Indicators:

PI1 Reduction of average time

PI2 Average number discrepancy reduction

Idl Trial Scen

ario PI_Description

PI_name AS IS (middle indicative value)

TO BE * (desiderata)

Ex-pect-ed Target

Comments PI_ Class

UM

3

AW

Digital Trial Case

Reduction of average time to make data in a digital format after/before the implementation during the period

PI1 Reduction of average time

50÷65 min

(for each person, monthly) (this is an indicative value, it’s a medium value per person)

5÷10 min

(for each person, monthly)

- 45÷55%

Target is a percentage of reduction of time to search, copy or write data from different sources inside the Database used by Quality Production for the Logbook data compilation

Lead Time in minutes (LT)

% reduction

3

AW

Digital Trial Case

Reduction of average data discrepancies after/before the implementation during the period (for each helicopter)

PI2 Average number discrepancy reduction

30÷40 (number

of discrepancy per each helicopter)

5÷10 (number of discrepancy per each helicopter)

- 25÷30%

Reduction of number of discrepancies between the analysed data from different sources and complied by different persons, it’s linked especially to possible human error of transcription inside the Database of Quality Production for the Logbook data compilation

Number

% reduction




* Data regarding tool utilization are partially simulated as errors are not statistically frequent

enough in comparison to the pilot dynamics

Smart Trial

For the smart trial the BPI is qualitative in terms of availabily of documents reporting the

events which can occur to the tools (tool tracking) used during assembly of the helicopters at

the FAL or during maintenance activities of the helicopters at a service center.

Selected a period of time to be monitored, the following information are provided:

the last date in which the event happens;

the typology of event (for example not correct position of tool in the smart tools box at

the end of activity);

the tool connected / associated to the logged event;

the number of time that the event linked to the selected tool happens (this is linked to

the entire chronology, starting from the first use of the toolbox and referring tools).

These documents are used by stakeholder involved in FOD (Foreign Object Debris)

prevention to better define and tailor training courses to be provided to technicians as

prescribed by regulations for all workers employed in the aviation industry (for example

taking into consideration events that occur more frequently, tools most forgotten, etc.)

The aim is to keep high the awareness on safety that, in all its forms, is essential for the

avoidance of risk to helicopter’s users (passengers and crew).

The expected result was already achieved at month 12 and it was confirmed at month 18 and

month 27. Evaluation of BPI was made not only based on the availability of documents, but

also on their usefulness as confirmed by some senior instructor who appreciated the type of

information inside.




In particular at month 27 additional features were implements in the system such as a more

rational organization of the collected information that have been grouped by source (smart

tool box) allowing to better understand the areas where the events occurred, or the possibility

to display the image of the tools (if available). The users really appreciated these new features

because provide them further information.

It is important to point out that additional benefits could be obtained using the results of the

trial as input of TELL ME (Technology Enhanced Learning Living Lab for Manufacturing

Environments) platform developed within the an integrated project in technology Enhanced

Learning for Manufacturing workplaces of the future. The aim of this project is to support the

Blue Collar Worker at the workplace providing specific training by using the latest

technologies and insights.

The TELL ME scenario developed for aeronautical sector refers to technicians that, at service

stations, carry out maintenance activity on AW helicopters. Also in this case FOD prevention

is one of the main topics and it was faced with an innovative approach based on Precision

Teaching methodology.

Purpose: The system produces period reports relative to Tools events recorded. The reports contain

data useful for the preparation of further and future tailored Training Material linked to Tools FOD

Prevention.

Business Performance Indicators: PI6_ more tailored data for training materials linked to results of

new tracking tools methodology

Id Tri

al Scenario

PI_Description

PI_name AS IS (middle indicative value)

TO BE * (desiderata)

Expected Target

Comments

PI_ Class

UM

3

AW

Smart Trial Case

Report of tools use tracking useful for the future preparation of further dedicate and tailored training material linked to Tools FOD prevention

PI6_ more tailored data for training materials linked to results of new tracking tools methodology

N/A

Not fully achieved

Qualitative value though a binary indicator: yes / no

After the implementation it will be possible to know if the system and the referring will be able to furnish data useful for the raw training material ( yes) or not (no)

Qualitative Indicator

Qualitative value

Document




* Data regarding tool utilization are partially simulated as errors are not statistically frequent

enough in comparison to the pilot dynamics

Consolidated Trial Experience

The aeronautical manufacturing is a high-technology industry characterized by complex

processes that are also regulated by national and international bodies. This is meant to reach

and ensure the highest possible level of safety. The implementation of FITMAN solution took

into consideration these aspects in order to avoid any interference that could compromise the

safety of the flight and the airworthiness is why a testing environment has been set up as an

exact replica of the real environment.

In the same way AW guidelines and policies oblige that all data has to be treated with high

confidentiality and should not be accessible by public in any case. As a consequence a cloud

solution has been excluded preferring the use of the internal network.

Some problems were faced during implementation due do the software quality of FIWARE

GEs available in the catalogue that is still not mature and bug free.

The use of different operating systems (UNIX and MICROSOFT) has led to some problem of

interfacing.

The Future Internet technology such as that experienced in FITMAN trial 3 is in line with the

AgustaWestland strategy towards customers who are asking for more – more safety

assurance, more quality on delivery and more reliability and availability in service.

4.2 Aidima

Furniture Trends Forecasting for Product Development / UC1

Type Indicator

s

Descriptio

n

Unit Curren

t value

Future

expecte

d value

TO-

BE

Value

s

1, 2, 3

Impac

t level

Comments

Efficiency Search

time

process per

source

Reduction

of searching

time

(working

hours

saving) per

source when

browsing

electronic

sources,

identifying

weak

signals and

classifying

them.

Workin

g hours

8 hours

per

source

approx.

6 hours

per

source.

6

6

5.1

Resear

ch

Productivit

y

Sources Increase of

number of

electronic

sources

analyzed by

trends

experts due

Number 20 +40/year 25

25

60

Resear

ch

Software

allows to

analyze as

many sources

as analysts

input into the

system.




to FITMAN

automated

solutions

Number of

sources can be

incremented

as needed

without limit

Productivit

y

Weak

signals

Increase of

number of

weak

signals

identified

due to

FITMAN

automated

solutions

Number 200

approx.

400/year 220

220

462

Resear

ch

Weak signals

are index

cards with any

score

Productivit

y

Index

cards

Increase of

number of

index cards

due to

FITMAN

automated

solutions.

Number 100

approx.

300/year 150

150

286

Resear

ch

Index cards

are weak

signals with a

score higher

than 3 stars

and that are

printed out.

The new

rating system

has a lot of

potential and

can increment

the number of

index cards

dramatically

Opinion Mining in Furniture Products / UC2

Type Indicators Descriptio

n

Unit Curre

nt

value

Future

expecte

d value

TO-

BE

Value

s

1, 2, 3

Impa

ct

level

Comment

s

Efficien

cy

Complaint

s

resolution

time

process.

Time

saving

when

addressing

customer

complaints

or negative

opinions.

Days >1 <1 1

1

0,2

Comp

any

Complain

resolution can

be reduced

dramatically

since the

answers can be

carried out via

Facebook or

Twitter. It all

depends on the

time dedicated

on a regular

working day Marketin

g

Opinion

retrieval

Number of

identified

Percenta

ge

0% -

10%

100% 30

30

Comp

any




electronic

customer

opinions

about the

firm or its

products,

services and

brands

75

Social Identificatio

n of non-

reported

dissatisfacti

on

Increase of

cases of

non-

reported

customer

online

dissatisfacti

on related to

product

and/or

service. Not

directly

reported to

the

company

Percenta

ge

0% 100% of

online

comment

s. On

specified

sources

20

20

100

Custo

mers

Social Opinion

leaders

Identificatio

n of opinion

leaders

amongst

customers

(i.e.

bloggers,

etc.). Not

professional

.

Number 0 Up to 5 1

1

7

Custo

mers

4.3 Volkswagen

KPI Name AS-

IS

value

TOBE

1

TOBE

2

TOBE

3

Target

value

Progress

1

Progress

2

Progress

3

BP: Management of the Machinery Repository

MR

Update

Cost

100 85 75 60 50 30,00% 50,00% 80,00%

MR

Update

Time

100 80 70 50 46 37,04% 55,56% 92,59%

BP: Inquiry Service

Average




lead time

to access

experts

knowledge

100 60 55 31 29 56,34% 63,38%

97,18%

Evaluation

Accuracy

100 90 85 65 50 20,00% 30,00% 70,00%

Inquiry

Respond

Time

100 95 90 83 80 25,00% 50,00% 85,00%

Inquiry

Respond

Cost

100 95 92 91 90 50,00% 80,00% 90,00%

4.3.1 MR Update Cost

Comment:

This KPI indicates the cost of updating or adding a machinery module inside the MR. This

task is done by the responsible engineer and the value is calculated by the effort in hours and

the hourly wage of the engineer.

Trend:

The numbers in the table above show a continuous reduction of the update cost over the

project period and the implementation/evolution of the trial system (final achievement 80%).

Reasons:

This reduction is achieved by using the FITMAN trial system. Due to its web-based services

the engineer can easily and fast accesses the MR to enter machinery data. One major

improvement is the semi-automated extraction of machinery data from the PLM system. This

function aggregates and abstracts the detailed and unsorted data from an XML file and stores

it into the MR. By this process the manual effort is reduced a lot.

4.3.2 MR Update Time

Comment:

This KPI indicates the time of updating or adding a machinery module inside the MR and to

make it public (accessible to all engineers). This task is done by the responsible engineer and

the value is the time in hours.

Trend:

The numbers in the table above show a continuous reduction of the update time over the

project period and the implementation/evolution of the trial system (final achievement

92,59%).

Reasons:

The engineer can easily and fast accesses the MR to enter machinery data. After the

successful updating/adding of machinery data, this data is instantly accessible by other




engineers and no further manual distribution is needed. This results in a much shorter update

time.

4.3.3 Average lead time to access experts knowledge

Comment: This KPI indicates the time to get in contact with relevant experts in production planning and

to receive information about production topics.

Trend:

The numbers in the table above show a continuous reduction of the lead time to access experts

knowledge over the project period and the implementation/evolution of the trial system (final

achievement 97,18%).

Reasons:

Every user can get in direct contact with engineers who are responsible for different assembly

sections by using the FITMAN system web services. The user has only to choose the product

or assembly section and the inquiry will be forwarded to the responsible engineer. Due to this

no manual effort for searching the responsible person or department is needed and the process

is fastened.

4.3.4 Evaluation Accuracy

Comment:

This KPI indicates the accuracy of cost estimations during the production system planning

phase and is based on estimated and real cost. Unfortunately only a long term measurement

could provide reliable data for this KPI, which was not possible inside the frame of FITMAN

with respect to the implementation date of the system. To deal with this issue older car

projects were evaluated with the FITMAN trial system. The shown KPI values are based on

these results and by using a pessimistic approach. But even an achievement of 70% (equal to

an improved accuracy by 35%) is very good.

Trend:

The numbers in the table above show a continuous improvement of accuracy over the project

period and the implementation/evolution of the trial system (final achievement 70%).

Reasons:

This reduction is achieved by using the FITMAN trial system. The cost evaluation was

improved due to the aggregated and abstracted machinery data in the MR.

4.3.5 Inquiry Respond Time

Comment:




This KPI indicates the time which is needed to receive, analyse and evaluate an inquiry and to

create a report.

Trend:

The numbers in the table above show a continuous reduction of time for the inquiry respond

over the project period and the implementation/evolution of the trial system (final

achievement 85%).

Reasons:

Thanks to the web services the engineer can easily and fast accesses the system to view and

analyse this inquiry. The evaluation of the inquiry is supported by the aggregated and

abstracted machinery data in the MR, which reduces the manual effort for data aggregation.

After finishing the evaluation an online report is created and sent back to the requester. This

report is instantly available on his/her computer or mobile device.

4.3.6 Inquiry Respond Cost

Comment:

This KPI indicates the costs which are spent for the evaluation of production related inquiries.

Its value is calculated by the effort in hours and the hourly wage of involved persons.

Trend:

The numbers in the table above show a continuous reduction of cost for the inquiry respond

over the project period and the implementation/evolution of the trial system (final

achievement 90%).

Reasons:

The engineer can easily and fast accesses the system to view and analyse an incoming inquiry.

The evaluation of the inquiry is supported by the aggregated and abstracted machinery data in

the MR, which reduces the manual effort for data aggregation and evaluation. The main

amount of the evaluation is still based on the engineer’s experience.

4.4 Consulgal

Performance indicators are measures that describe how well a program is achieving its

objectives. Following, we describe what the data show for each of the indicators measured

until M27:

PI1: Ratio: Average lead time to access the information relating to concrete

characteristics and concreting plan after/before the DV/AV implementation during the

concrete control process.

AS-IS: 4 hours. Target value: 98% of reduction.

This performance indicator is to provide information on the time saved by the elimination of

waiting time in the process.




What the data show?

The values obtained represent a 99.96% of reduction in time. This value is very close

to the expected value for this PI (98% of reduction in time.)

PI2: Ratio: Average number of pages used in the test results recording, archival,

after/before the DV/AV implementation during one concrete operation.

AS-IS: 5 pages. Target value: reduce by 40%

This performance indicator provides information about the average number of pages used for

recording the test results during one concrete operation.

What the data show?

In the simulations made we did not print information but this will not be more than 2

pages per concreting operation. This represents 60% of reduction.

PI3: Average lead time needed to perform and record the test results after/before the

DV/AV implementation during one concrete operation.

AS-IS: 27.5 minutes. Target value: 30% of reduction.

This performance indicator is to provide information about the time saved due to automation

of the process.

The values obtained represented a 75.39% of reduction in time. This value definitely exceeds

the expected value for this this PI.

The application has removed the registration time in Excel, and additionally it reduced the

other times beyond our expectation, because it is not necessary anymore to register several

times the data that identify the concrete operation. This later aspect has not been considered in

our AS IS analysis and, for that reason, the value of this performance indicator significantly

exceeded our initial estimate.

PI4: Ratio: Average lead time needed to analyse the test results after/before the DV/AV

implementation during one concrete operation.

AS-IS: 39 days. Target value: 98% of reduction

This performance indicator is to provide information about the time we hope to save by the

elimination of waiting time in the process.

What the data show?

The time to analyse the test results is minutes. We consider the values 0 due to the

magnitude scale. This represents 100% of reduction.

PI5: Ratio: Time for data exchange between stakeholders after/before the DV/AV

implementation during the concrete control process.

AS-IS: 8 hours. Target value: 98% of reduction.




This performance indicator is to provide information about the time expected to be saved by

improving the actual exchange information between stakeholders.

What the data show?

We made measurements in all the Business Scenarios. The values obtained represent a

99.97% of reduction in time. This value is very close to the expected value for this PI

(98% of reduction in time.)

PI6: Ratio: Average cost needed to perform and record the test result after/before the

DV/AV implementation during one concrete operation.

AS-IS: 2.04€. Target value: reduce by 30%.

This performance indicator is to provide information about the average cost of human

resources involved in the process.

The values obtained represented a 74.01% of reduction in time. This value definitely exceeds

the expected value for this this PI. The values obtained in this PI exceeded the value of the AS

IS, due to the values obtained in the PI3.

PI7: Ratio: Average cost needed to analyze the test result after/before the DV/AV

implementation during one concrete operation.

AS-IS: 1.41€. Target value: reduce by 65%

This performance indicator is to provide information about the average cost of human

resources involved in the process.

What the data show?

The values obtained represent a 67.37% of reduction in cost. This value exceeds the

expected value for this PI.

In general, we can say that the improvements that may be obtained from using the application,

in the concrete control process, have exceeded our expectations. Nevertheless, there is room

for significant improvements on what the user interfaces are concerned, the functionality of

some of the features, the versatility allowed and user-friendliness.

Table with the values of BPIs (ASIS, TOBE 1, 2, 3 , Target value)

7 CONSULGAL

1

IDENTIFICATION OF CONCRETE CHARACTERISTICS AND CONCRETING PLAN T

OB

E

1

TO

BE

2

TO

BE

3

Tar

get




EXCH.TIME TIME FOR DATA EXCHANGE 28800 5,15 6,35 5,12 576

Values are in seconds. (8 hours = 28800 seconds) Target reduction in percentage is 98%

LT Char.&Plan

AVERAGE LT TO ACCESS INFORMATION 14400 7,5 5,39 4,73 288

Values are in seconds. (4 hours = 14400 seconds). Target reduction in percentage is 98%

2 SAMPLES COLLECTION AND TESTING

COST RES.

AVERAGE COST TO PERFORM AND RECORD RESULT 2,04 0,55 0,47 0,57 1,43

Values are in € Target reduction in percentage is 30%

EXCH.TIME TIME FOR DATA EXCHANGE 28800 8,2 5,1 4,89 576

Values are in seconds ( 8 hours = 28800 seconds) Target reduction in percentage is 98%

LT RES.

AVERAGE LT TO PERFORM AND RECORD RESULTS 1650 424 358 436 1155

Values are in seconds (27.5 minutes = 1650 seconds) Target reduction in percentage is 30%

NUM.PAG. AVERAGE NUMBER OF PAGES 5 2 2 2 3

Values are in number of pages. Target reduction in percentage is 40%

3 TEST RESULTS TREATMENT AND EVALUATION

COST AN.RES. AVERAGE COST TO ANALYZE RESULT 1,41 0,52 0,41 0,45 0,49

Values are in € Target reduction in percentage is 65%

EXCH.TIME TIME FOR DATA EXCHANGE 28800 8 10,5 8,9 576

Values are in seconds Target reduction in percentage is 98%

LT AN.RES. AVERAGE LT TO ANALYZE RESULTS 39 0 0 0 0,78

Values are in days Target reduction in percentage is 98%

4.5 TRW

TRW has performed the experimentation of SMART trial. At the end of this phase TRW has

been able to produce significant lesson learnt referring to the collected results.

4.5.1 Trial Results and Progress

The table in the next page summarises the expected target, the real values measured in the

TRW trial during the whole project and the progress of the measured KPIs..

TRW trial will use percentages of improvement and decrease of the business performance

indicator as measuring unit, avoiding the usage of absolute values. The main reason for this

choice is the misuse that external users can do with current data of TRW, getting them out of




context and creating non-desirable image for a worldwide leader branch in the automotive

sector. Due to this unfortunate and possible situation, TRW will use percentages comparing

current and future values of each indicator.

Additionally, the most important target of TRW due to business performance indicator is to

not only assess the impact of the FITMAN system instantiation, but also report and

communicate this impact in the manufacturing and production activities thanks to FI

technologies deployment. In order to reach these objectives of assessment and

communication, percentage values of TRW indicators are as useful as absolute values, since

they are able to reflect the evolution of the business processes in the factory.

PI Name of the PI Expected Target TO

BE 1 TO

BE 2 TO

BE 3 Progress 1 Pro. 2 Pro. 3

TRW Trial 83,32% 123% 178,44%

BS1 - RISK MODELLING 78,3% 116,7% 184,7%

BS1PI 1

Number of standards and regulations (added) in the repository after/before the DV/AV implementation during a period

Increase of 5% Good

Increase of 7% Very good

Increase of 15% Excellent

4 6 10 80,0% 120,0% 200,0%

BS1PI 2/ BS2PI 1

Number of accidents and incidents in the factory after / before the DV/AV implementation during a period

Reduction of 10% Good

Reduction of 15% Very good

Reduction of 20% Excellent

9 13 17 90,0% 130,0% 170,0%

BS1PI 3

Number of risks that has been defined using the new system after / before the DV/AV implementation during a period

Increase of 30% Good



25 40 50 83,3% 133,3% 166,7%

BS1PI 4

Number of preventive actions using the new systems after /before the DV/AV implementation during a period




18 30 50 60,0% 100,0% 166,7%

BS1PI 5

Number of human errors in the design of prevention strategy planning after /before the DV/AV implementation during a period

Reduction of 10% Good

Reduction of 20% Very good

Reduction of 30% Excellent

- 10 22 100,0% 220,0%

BS2 - RISK DETECTION AND INFORMATION 88,3% 130,1% 172,2%

BS2PI 2

Number of deployed monitoring systems after / before the DV/AV implementation during a period




50 70 95 90,9% 127,3% 172,7%

BS2PI 3

Number of risk detections, alarms and warnings set up after / before the DV/AV implementation during a period




60 80 95 92,3% 123,1% 146,2%

BS2PI 4

Number of training sessions regarding safety after /before the DV/AV implementation during a period




20 35 50 80,0% 140,0% 200,0%

4.5.2 TRW KPIs Analysis

The main reason for the high progress in the TRW trial (over 100%) is due to the inexistence

of customised and effective tools and systems for the optimisation of the preventive planning

design.




Furthermore, these KPIs demonstrate that the use of the FITMAN system in the TRW

production line allow a successful prevention of the injuries and illnesses that later can

provoke important musculoskeletal disorders, enhancing the health and safety of the

workers.

BS1PI 1: Number of standards and regulations (added) in the repository after/before the

DV/AV implementation during a period

This performance indicator aims to measure the time invested and the reduction of

inefficiencies (time) in the broad application of current regulations and standards.

TRW is currently using REBA, NIOSH and OCRA standards, which are the most important

ones. With the new system, the time invested in the full application of these standards and the

range of information controlled (parameters controlled) has been optimised, not changing the

costs.

BS1PI 2/ BS2PI 1: Number of accidents and incidents in the factory after / before the


This is a key performance indicator, which ensures that the system is able to reduce the

number of injured workers and reduce the lost days in the production line.

The TRW trial has achieved a reduction of 17% in the accidents and incidents with the use of

the FITMAN system. As a result, the rates of injured workers with musculoskeletal disorders

have been significantly reduced, decreasing the number of lost days, with the important

savings that this supposes.

BS1PI 3: Number of risks that has been defined using the new system after / before the


The system allows setting up risks that can happen in the factory, specifying concrete

parameters and thresholds to detect them. The type of risks that can be found in the

production line are defined by important organisms, so these cannot be modified. But thanks

to the FITMAN system, several risk have been deeply configured and customised, which as a

result provides a better prevention and detection.

BS1PI 4: Number of preventive actions using the new systems after /before the DV/AV

implementation during a period

The system allows setting up preventive actions, linked to the risks detected. The increase of

50% of design of preventive actions is directly related to the previous KPI. However, more

important than the quantity is the quality of the preventive actions, and thanks to the FITMAN

system more accurate and customised actions are possible, getting support from the latest

technologies.

BS1PI 5: Number of human errors in the design of prevention strategy planning after

/before the DV/AV implementation during a period

This performance indicator is focused on checking that the human errors are reduced, which

is one of the main problem of current systems. Nowadays the prevention technicians are the

ones in charge of the risk detection and preventive action definition. Even if they have huge

experience, some errors might appear. In the TRW trial these errors have been decreased to

22%.




BS2PI 2: Number of deployed monitoring systems after / before the DV/AV


Nowadays there are hardly any IT systems supporting the prevention activities. In the TRW

trial of FITMAN new IT equipment and infrastructures have been deployed in the selected

production line, increasing in a 95% the usage of technology for this area.

BS2PI 3: Number of risk detections, alarms and warnings set up after / before the


This is a key performance indicator, since it determines the effectiveness of the systems in the

risk detection and preventive actions deployment. Due to the FITMAN system

implementation in the TRW factory, the number of risks detected and alerts send to the

workers has increased in a 95%, which has directly contribute in the enhancement of the

workers well-being. Additionally, new tools and mechanisms for performing the prevention

actions have been set up.

BS2PI 4: Number of training sessions regarding safety after /before the DV/AV


The objective of this performance indicator is to probe the increase in the awareness of the

importance of H&S adoption in the TRW factory. Thanks to the information gathered and

analysed in the TRW trial, 50% increase in the training sessions has been performed.

Therefore, the workers have more knowledge on their postures and behaviours, which in the

end becomes into less injures and health problems.

4.5.3 Consolidated Trial Experience

To implement the FITMAN platform in the industry sector some points have to be

considered:

- Make sure all goals/objectives of the planned usage and all needed functionalities are

clarified.

- Check the needed functionalities with the provided platform components whether they

fit or not. Maybe additional components like SEs or TSCs are required.

- Make sure your infrastructure complies with the component’s requirements

(Hardware, Software, OS).

- Clarify existing guidelines and policies and check if they interfere with the FITMAN

platform components.

- Develop an implementation roadmap and experimentation plan. Maybe it is advisable

to test each component separately and to implement the components step-by-step.

- If problems or questions are occurring, contact the owner or developer of the

component. They can provide help or needed adaptions.

More concretely, and due to concrete aspects of the TRW trial, some other important factors

has to be taken into account. These differential aspects are the location of the trial in the shop

floor (in the production line) and the importance of workers safety and security in the trial.

The first important activity to be achieved is the calibration of the sensors deployed in the

shop floor. The point is that depending on different aspects such as the light, vibrations,

location, etc. the results provided by the sensors cannot be reliable. Therefore, some tests and




calibrations have to be performed in the concrete location where the devices will be deployed

in the shop floor, to ensure their accuracy.

The second advice is related to the selection of the personnel involved in the trial. The

TRW trial monitors and processes information about the production line workers, adding

some difficulties to the technical implementation. Therefore, the selection of the personnel

participating in the trial has to be done very carefully, since those people will not only provide

their data for the benefit of the trial, but also will give some feedback and opinions about

functionalities, usability and other important aspects of the solution. The TRW solution will

be mainly used by the blue collar workers and prevention technicians, so they should be the

main source for developing an intuitive and easy-to-use solution based on the FI

technologies.

4.6 TANet

4.6.1 Overview

TANET

PI_Desc PI_Name AS-IS TO-BE1 TO-BE2 TO-BE3 Target Comments UM

IMPORT OF TENDER OPPORTUNITIES

FAC.NUM.

NUMBER OF ACTIVE

FACILITATORS 1 2 2 3 3

Current Value and Target are in number of active facilitators Added Welsh

Automotive Forum as facilitator into use case. Discussing inclusion with third

partner. M18 - still in talks with third partner, dependent on open call

integration ability to import member data. Third facilitator tentatively

onboard, based on GetOva import of suppliers.

number of active

facilitators Productivity (P)

SERV.PR.NUM.

NUMBER OF REGISTERED

SERVICES PROVIDERS 23 23 71 101 115

Current Value and Target are in number of registered service providers

SMECluster is not yet advertising for new service providers. This is a long-

term goal increase. M18 - improved by using GeToVA, available at:

http://fitman.sti2.at/companies - GeToVA pulled in third facilliators

suppliers.

number of registered

service providers Productivity (P)

TEND. TENDERS ACCRUED MONTHLY 3 3 12 18 20

Current Value and Target are in numbers of tenders No automated process

for acquiring tenders exists yet. M21 planned completion using open call

components. M18 - tender entry by facilitator using SMECluster platform.

Tend 3 - tenders entered by facillitator still, Getova SE was aimed more at

import of suppliers. number of tenders Productivity (P)

IMPROVEMENT OF FACILITATOR ROLE

CLUST. END-TO-END CLUSTERING 6 5 2 2 2

Current Value and Target are in hours Decrease in time due to use of CAM as

data store. SCAPP implementation expected to significantly reduce time by

providing negotiation tools. Open call components will also reduce time

through import of tender opportunities. M18 - SCAPP negotiation rooms has

hugely simplified the process of negotiation between facilitator and supplier

members -open call components were not used to automate tendering so no

changes to the end-to-end clustering time. hours Lead Time (LT)

TEND.AUT.

AUTOMATED TENDER INPUT

TIME 30 30 6 2 1

Current Value and Targets are in minutes Open call components will be used

to automate import of tenders - completion planned for M21. M18 - Majority

of data storage and representation supported by CAM, reducing input time;

input time will be further reduced using OCSEs - time was reduced further by

the addition of Mova which gives a better interface to the ontology thus

allows annotating in a better manner. Getova was originally to be used for

fully automating tendering however this was dropped as the focus became

further toward the import of suppliers. minutes Lead Time (LT)




4.6.2 General Comments about KPI’s

4.6.2.1 Import of Tender Opportunities:

These all relate to improving the throughput of business opportunities by implementing the

GE’s, SE’s and TSE’s.

Number of Active Facilitators:

It was always envisaged that SMECluster as a tender platform would have been made

available to other Facilitator based organisations. The testing was conducted primarily with

the Welsh Automotive Forum who have always required such a platform and the additional 2

businesses interested were Aerospace Forum and Industry Wales.

Number of Registered Service Providers:

This was based on the number of providers that could offer services via the SME portal and

whilst expressions of interest increased through-out the trial, the figures reflect the likely

uptake in the local region.

Tenders Accrued Monthly:

This KPI is a calculated value based on connecting to sites such as www.sell2wales.co.uk and

scraping information about current tenders from the relevant areas. Also the importing of

suppliers was introduced to SMECluster via the newer GE’s such as MoVA and GeToVA.

4.6.2.2 IMPROVEMENT OF FACILITATOR ROLE:

This KPI was to measure the effectiveness of the new tools coming out of FITMAN to

improve the efficiency of the facilitators.

End to End Clustering:

This KPI was to measure the overall reduction of “processing” time for facilitators through

negotiating tools provided by FITMAN.

Automated Tender Input Time:

The numbers connections to pull opportunities from tender sites such as

http://www.tendersdirect.co.uk/ and www.sell2wales.co.uk increased the speed of processing

TANET 8

1 IMPORT OF TENDER OPPORTUNITIES AS

-IS

ToB

e1

ToB

e2

ToB

e3

Targ

et

Pro

gre

ss1

Pro

gre

ss2

Pro

gre

ss3

Comments

FAC.NUM. NUMBER OF ACTIVE FACILITATORS 1 2 2 3 3 50.00% 50.00% 100.00%

Current Value and Target are in number of active facilitators Added Welsh

Automotive Forum as facilitator into use case. Discussing inclusion with

third partner. M18 - still in talks with third partner, dependent on open

call integration ability.

SERV.PR.NUM.

NUMBER OF REGISTERED SERVICES

PROVIDERS 23 23 71 120 115 0.00% 52.17% 105.43%

Current Value and Target are in number of registered service providers

SMECluster is not yet advertising for new service providers. This is a long-

term goal increase. M18 - improved by using GeToVA, available at:

http://fitman.sti2.at/companies.

TEND. TENDERS ACCRUED MONTHLY 3 3 12 12 20 0.00% 52.94% 52.94%

Current Value and Target are in numbers of tenders No automated

process for acquiring tenders exists yet. M21 planned completion using

open call components. M18 - tender entry by facilitator using SMECluster

platform.

2 IMPROVEMENT OF FACILITATOR ROLE

CLUST. END-TO-END CLUSTERING 6 5 2 2 2 25.00% 100.00% 100.00%

Current Value and Target are in hours Decrease in time due to use of CAM

as data store. SCAPP implementation expected to significantly reduce time

by providing negotiation tools. Open call components will also reduce time

through import of tender opportunities.

TEND.AUT. AUTOMATED TENDER INPUT TIME 30 30 6 2,5 1 0.00% 82.76% 94.83%

http://www.sell2wales.co.uk/

http://www.tendersdirect.co.uk/

http://www.sell2wales.co.uk/




tenders compared to the original manual method. Further work is required to actually make

these opportunities to available to facilities.

4.7 COMPlus

4.7.1 Network Transparency For More Efficient Supplier Search

Type Indicators Description Unit Cur-rent

value

Future

expected

value

TO-BE

Values

1, 2, 3

Comments

Lead Time Time used

for

configuratio

n and data

entry

Reduction of

configuration

and searching

time due to the

configuration

of the supply

network.

% Reduc-

tion of

35%

Reduction

of 85%

5

20

35

This indicator

shows the level of

support to the

configuration

process of supply

network.

This KPI

improves during

the use of the

solution. As the

number of

included entities

into the

knowledge base

increase, the

maturity of the

system improves.

Produc-

tivity Level of

transparency

Improvement

of level of

transparency

of the supply

network

% Improve

ment of

50%

Improvem

ent of 80%

0

20

50

This indicator

shows the level of

achieved

transparency

within the supply

networks. With

the continuous use

of the solution,

the knowledge

base is being

enriched and

hence the level of

transparency if the

network

improves.

Lead Time Reduction of

the time

needed for

searching of

a supplier

Reduction of

time needed to

search for an

existing or a

new supplier

% Improve

ment of

35%

Reduction

of time to

search for

a supplier

of 80%

5

15

35

This indicator

shows the level of

decrease of time

needed to search

for a new or

existing supplier

within the

network

4.7.2 Transparency And Consistency Of ITs And Documents




Type Indicators Description Unit Current

value

Future

expecte

d value

TO-BE

Values

1, 2, 3

Comments

Quality Reduction

of Mistake

and Errors

This

indicator

show the

ratio of the

reduction of

mistakes and

errors during

the

configuration

of the supply

network

% 35 80 5

15

35

The

improvement of

this ratio

improves with

the enrichment

of the supplier

knowledge base

and the maturity

of the solution.

Produc-

tivity

Standardise

d IT

Landscape

This

indicator

shows the

ration of the

standardised

IT

Landscape

% 30 55 15

25

30

This ratio

improves with

the number of

shared best

practices and

maturity of the

solution.

Extended Lessons Learned and Evaluations Integrated ...€¦ · 1.1.2 Challenges for big data...

Documents

Transcript of Extended Lessons Learned and Evaluations Integrated ...€¦ · 1.1.2 Challenges for big data...