Smart Maintenance

21
Smart Maintenance: Concept Development of a Cognitive Integrated Vehicle Health Management System Ali Baghchehsara Research & Development VDev Systems and Services The Royal Aeronautical Society London, United Kingdom Sep 4, 2017

Transcript of Smart Maintenance

Smart Maintenance:

Concept Development of a

Cognitive Integrated Vehicle Health Management System

Ali BaghchehsaraResearch & DevelopmentVDev Systems and Services

The Royal Aeronautical SocietyLondon, United Kingdom

Sep 4, 2017

OUTLINE

Introduction

Introfrom Architecture,

Technology, to Operation

DevelopmentsInterpretation of

Results in the concept& Resolution

Results

Use CaseApproach

Key Facts about Approaching the IVHM

Concept

Test Setup, Execution and

Analysis

Of a Vehicle

Health

To Manage

Different Systems

Integrationof

Key Point:

Managing Maintenance based on Health (Status and Trend) of the Vehicle

IVHM?

MAP

Developed a flowchart Expanded the flowchart to testing process

Examined some test cases

Phase 2

Literature Research, Communication Architecture

Identified the key functions of IVHM, “Prognostic” Technically possible / not possible?

Phase 1

Method verified via a real use case on an A320 with VDev

DLRK 2016 Extended my method for other A/C systems

Phase 3

Start: Joining DLR

Design & Implementation of IVHM in process

Closing

Big Picture

Status Quo of Maintenance

• A, B, C, D checks (or Task Checks)

• Conditional Based Maintenance

Objectives

• Predicting Failure

• Preventing Failure

Results

• Higher Safety in Operation

• Costs Reduction

Own Development based on HUMS, 1991 - updated in 2016

Reactive Maintenance

Preventive Maintenance

Prognostic

Current Objective

Percent

Key Facts

83%

17%

5

AOG

UF

Fixing the components before they cause AOGs.

Observing health status and enabling communication between systems and humans to turn unexpected failures into expected maintenance actions.

Toward Zero AOG

Unexpected Failures

Enabling Communication

Monitoring the Health Status

74%

26%

6

Key Facts from the Unexpected to the Expected

Predicting …

Years, Decades or Century?

Complex Systems do not need to be onboardSystems, sensors and software are

reliable TODAY

Today with Internet of Things, Cloud Services are available onboard

Systems are able to learn and develop Diagnostic Models

Input Data, Event

Utilizing Today’s Building Blocks

Systems are able to communicate with systems and humans

The system would self-develop the Knowledge Base (Cognitive Computing Development)

Maintenance Management based on

Failure Prediction

68%

32%

7

Key Facts

1 ObserveThe system would see the the happenings and evidences as we does . I n sys tems i t means observing inputs and outputs.

3 EvaluateExtending the evaluat ion to understand the interpretation and evaluating which hypothesis are (more likely) right or wrong.

4 DecideChoosing set of options that seems best and acting accordingly based on the decision trees, gained knowledge, and existing models.

2 InterpretLearning, exploring, and analyzing the meanings of what we are seeing. To generate hypothesis and interpret their meaning.

Decide

Observe

Evaluate

Interpret

Cognitive System68%

32%

8

Key Facts

The ADE

Health Manager

Data Process

A/C DATA

Cloud Server, But?

Maint. Exec.

LogisticMaint.

Support

Maintainer

Verified Algorithm

Failure Prediction

58%

42%9

Airline

Baghchehsara, 2016

Algorithm Verification

Cognitive

Scripting

Maint. Exec. Maint.

Events

Autonomous Development Architecture

Generate Indication & Predict ADMM

50% 50%10

1 2 4

Concept Concept development and c u s t o m i z a t i o n ,

Personal izat ion of the

cognitive computings in a

w a y t h a t fi t s t o t h e stakeholder’s settings.

Development Model Developments:

- Test cases

- Flight Simulators

- Fleet data extension of

models to selected fleet

(Beta)

Service S e r v i c e o f I V H M i n

operation of the fleet:

- Delivery Into Service

- Operative Life

3

“Lifecycle of the designated IVHM concept demands a Transition phase between regular maintenance in today’s operation, to the point where IVHM is mature and learnt enough to take over the maintenance management. “

Transition This phase is continuous

self-learning, validation and

verification of the models created, in ops and keep

matching the maintenance

records.

Transition to Operation

Life Cycle of IVHM

- Extraction of Knowledge

(End of Life of an Aircraft)

Use Case - A Feasibility Study

Test Case: Ram Air Outlet Flap

Failure Forecast in ATA 21 (Air-conditioning system)

The case hypothesis is to find an indication of Failure in advance.

Executing test needed a measurement system but no new sensors.

Failure model creation followed steps a cognitive system does.

Test Set up Actuation Data Collection

Airbus AFM, 2009

Steps:

43%

57%11

RamAirIn/Outlet

Actuators

MeasurementDevice

ACSystemControl

TapdatafromModularTerminalBlock

ACPanel(Cockpit)

Computer(TestResult)

Measurement System and Computer

Modular Terminal Block in Avionic Bay

Test Set Up

Air-conditiong System Control

41%

59%12

Bradley, 2015 Own Illustration, VDev Systems and Services, 2015 Doering, 2013

Test Setup

Test Result

Observation & Interpretation

Healthy Operation Profile (day 2)

Malfunction Profile (day 5)Test Operation Profile (day 3)

Unhealthy Operation Profile (day 4)

Measures of Spread

36%

64%

13

Reference: Own Work

Analysis of the Results

Evaluation - Failure Prediction in Reality

0

25

50

75

100

Day 2 Day 3 Day 4 Day 5

Upcoming Failu

re

31%

69%

14

Reference: Own Work

Observed Range of the CM data from in day 2 is: 52 (day 3: 57)

Range of the CM data in day 4 is: 75

Range of CM data in beginning of day 5 is: 96

Automatization of Developments

Observe Interpret Evaluate Decide & Act Develop/ Optimize

Geijtenbeek, 2013

Observe: A/C Data and Maintenance Records.

Interpret: Using the data, indications has been created.

Evaluate: a right failure prediction model.

Decide accordingly:

Machine Learning Feasibility

21%

79%

15

ATOM PROTECTION MODEL OF THE AUTONOMOUS SECURITY BOT (ASB)

• H. Butz, A, Baghchehsara

Adaptive Security

Target Localizer

Threat Identifier

Classical Scan &

Protection

Data Base

Function

NoEffect Minor Major HazardousCatastrophic

Frequent

Probable

Remote

ExtremelyRemote

ExtremelyImprobable

SafetyImpact

Likelihoo

d

BasedonED–202acceptabilityRMUnacceptable

ASBSteps*:1-Identifiestheattack(UnexpectedChange,unacceptablenumber)2-SafetyimpactishazardousandlikelihoodisFrequent(e.g.Takeoffdataiscracked)3-ProperCountermeasureIdentification(e.g.Pilotisavailabletore-confirmthedata)4-TakeAction(e.g.asksthepilotstowriteandconfirmvalues)

ASB CONNECTS SAFETY AND SECURITY

Baghchehsara,Butz,2017

WHERE MY INVESTIGATIONS STANDS NOW?

• What have I done ? • I designed a technical concept for IVHM and then proved it via a use

case. In this case, I observed the system health status, and created a prediction for failure with the same Configuration (No New Sensor)

• What it brings to operation ? • Incident: Berlin A332 at Berlin on Oct 1st 2016, rejected takeoff due to

air conditioning problem • IVHM (if implemented) could have prevented this problem,

considering the use case feasibility

• Where we are standing ? • Investigation point: Implementation • Security Safety Robot (Autonomous Security Bot) is implemented in

the IVHM system for ensuring safe operation. (e.g. concerning Cloud Based Transfer)

GroundControlStation-duringalabtest

Orangebox(Functions:DataTransfer&ASB)totheMaintenance&OPSSupportCenter(VDevSystemsandServices,2017)

Baghchehsara,Butz,2017

Cloud Service Protected by ASB

CONCLUSION

• It sounds feasible to further investigate use of cognitive computing to automatically create models instead of a human doing so. (Automated Environment)

• Health detection of Air-conditioning system’s flap actuator of A320 was possible. (without re-configuration (Causing highest costs and delays compared to other systems -no monitoring possible in Ops)

REFERENCES

HUMS, 1991 - Updated in 2016Airbus HUMS, D. (n.d.). Health and Usage Monitoring System. TN EV52-505/91. - Updated in a testimony with Lufthansa Technik AG in Frankfurt, Nov 2016

IBM Watson, 2014: IBM Watson, How it Works?, 2014 Retrieved on Feb 2016. Video Link: https://www.youtube.com/watch?v=_Xcmh1LQB9I

Baghchehsara, 2016 Ali Baghchehsara, Amine Boughalem, Henning Butz, Automated Technology Development of Integration of Vehicle Health Management System United States Patent No. 62451844, Granted on Feb 2017

Baghchehsara, 2016 Ali Baghchehsara, Development of Integrated Vehicle Health Management System (IVHM), with a focuss on Architecture, Technology and Operation. Master Thesis, Bremen University of applied sciences, Germany, Bremen,.

Baghchehsara, 2016 Ali Baghchehsara, Identification and Interpretation of Integrated Vehicle Health Management (IVHM) Generic Architecture and a Case Study. DLR Congress . DLRK 2016.

Airbus AFM, 2009 Airbus AFM. (2009). MASTER CONFIGURATION DEVIATION LIST RAM AIR INLET FLAP ILLUSTRATION. Airbusworld. Retrieved from MASTER CONFIGURATION DEVIATION LIST: http://www.avsoft.com/scrm/courses/B734-C16/AircraftGeneral/AircraftGeneral-29.jpg

Eric Bradley, 2015Air-conditioning Panel, Overhead Scan - Air Conditioning Panel in A320 Cockpit, drawing by: Eric Bradley, 2015, retrieved in Jan 2017 from: http://mcdu.equicom.net/blog20150716g.php

Doering, 2013Mathias Doering, Image taken in 2013, Retrieved from VDev Systems and Services, 2016

VDev Systems and Services , 2015Jan Robbe, Measurement System, Image taken in 2014, Image Courtesy: VDev Systems and Services, 2016

Geijtenbeek, 2013 Thomas Geijtenbeek, Michiel van de Panne, Flexible Muscle-Based Locomotion for Bipedal Creatures. Retrieved on Feb 2016 Link: https://www.youtube.com/watch?v=pgaEE27nsQw

Baghchehsara,Butz,2017AliBaghchehsara,HenningButz,AutonomousSecurityBot-4069AirworthinessReviewReportonAug2017

… Story Continues

THANK YOUQ & A

Ali BaghchehsaraResearch and Development [email protected]+49 (0) 421 5976 290