Artificial Intelligence –A Driving Force in Industrial 4 · Jay Lee, HosseinDavari, Jaskaran...
Transcript of Artificial Intelligence –A Driving Force in Industrial 4 · Jay Lee, HosseinDavari, Jaskaran...
Artificial Intelligence – A Driving Force in Industrial 4.0Shaibal Barua, PhDResearcher, Artificial Intelligence and Intelligent Systems
29 May, 2020
● Artificial Intelligence – what’s the deal?
● Industrial Artificial Intelligence
● The applied AI workflow● Data cleaning and preparation● Data representation● AI problems and methods● Validation
● Use cases
● What’s next?
2
Outline
Part 1: Artificial Intelligence
3
“The ability to learn, understand and think in a logical way about things; the ability to do this well”
- Oxford dictionary
5
Intelligence
Capability to understand complex ideas, ability to reasoning, learning from experiences, adaptability to the environment, plan, problem solving …..
"Mai
nstr
eam
Sci
ence
on
Inte
llige
nce”
, 199
4, th
e W
all S
tree
tJo
urna
l
”Int
ellig
ence
: Kno
wns
and
U
nkno
wns
”,199
5, th
e Bo
ard
of
Scie
ntifi
c Af
fair
s of t
heAm
eric
an
Psyc
holo
gica
l Ass
ocia
tion
Artificial Intelligence (AI) is usually defined as the science of making computers do things that require intelligence when done by humans.
AI is the study of programmed systems that can simulate, to some extent, human activities
such as perceiving, thinking, learning and acting.
7
Artificial Intelligence
Fig: Turing test
Behavior by a machine that, if performed by a human being, would be called intelligent
8
Artificial Intelligence
Artificial Intelligence
Reasoning
Knowledge representationLearning
Planning Perception Robotics
Social Intelligence
Natural language
9
Three types of AI
Narrow AI General AI Superintelligence
• Singular task• Successfully realized
to date• Operate under a
narrow set of constraints and limitations
• Machine intelligence• Carry out any
cognitive function that a human can
• Knowledge transfer between domains
• Fujitsu’s ”K” supercomputer
• Hypothetical agent• Machines become
self-aware • Surpass the capacity
of human intelligence
● Sixth-fifth century BC● Aristotle layout the epistemological basis; introduces syllogistic logic● The Iliad – assorted automata from the workshops of Greek god
Hephaestus
● Late first century● Fable automata built by Heron of Alexander
● Fifteenth-sixteenth century● Mechanic clocks, Paracelus introduces a recipe for a humanculus, an
intelligent “little man”
● Eighteen century ● Philosophers try to formulate the laws of thought
● Nineteenth century● Literary artificial intelligences proliferation
● Twentieth century ● Alan Turin proposes an abstract of universal computing machine
11
How Old is the idea of AI?
Hoffman’s The SandmanGoethe’s FaustMary Shelley’s Frankenstein
12
AI: Past, Present and Future
Golden years: 1957-1974Symbolic AI, search algorithms, neural nets, industrial robots, etc.
Expert systems boom: 1980-1987Rule-based, logical systemsSelection of components based on customer requirements5th gen project (Japan)Neural networks, backprop.
Goals fulfilled: 1993-2011
Deep Blue (1997)Victory of the “neats” (2003)DARPA Grand Challenge (2005)AI untold successes in data mining, robotics, logistics, speech recognition, search engines
Deep learning, big data and general AI: 2011-presentAccess to large amounts of dataFaster computersDeep learning drives progress in image and video processing, text analysis, speech recognitionGoogle DeepMind defeats world champion in Go (2016)Widespread discussions around Strong AI:superhuman intelligence
The Turing test
"I propose to consider the question, 'Can machines think?’” (A. Turing, 1950)
An interrogator asks questions to an (unseen) person A. If A is replaced by an AI, can the interrogator detect this or not?
1 st AI Winter:
1974-1980
2 nd AI Winter:
1987-1993
2017 AlphaGo: Google’s AI beats world champion Ke
Jie. Notable for vast number of 2170 of possible positions
● Ethical Reasoning● Accountability, Responsibility,
Transparency
14
Responsible AI
Figure: Trolley problem dilemma
● Responsible AI concerned with the fact that decisions and actions taken by intelligent autonomous systems have consequences that can be seen as being of an ethical nature.
Figure: Interrelationship of the seven requirements: all are of equal importance, support each other, and should be implemented and evaluated throughout the AI system’s lifecycle
Source: EU Ethics Guidelines for Trustworthy AI, https://ec.europa.eu/futurium/en/ai-alliance-consultation/guidelines/1
Part 2: Industrial ArtificialIntelligence
16
1st Industrial Revolution1765
•Mechanical production•Industry instead of agriculture
as basis of economy•Water power•Steam engine
2nd Industrial Revolution1870
•Electricity, gas and oil•Combustion engine, steel industry, chemical
industry•Telegraph, telephone•Division of Labour (Taylorism), Mass
production (Ford)
3rd Industrial Revolution1969
•Nuclear energy•Electronics,
telecommunication, computers•Automation - PLCs, control
theory, PID regulators, etc.•Industrial robots
4th Industrial Revolution2011
The four industrial revolutions
17
•Connected machines•Complex human-machine interaction•Artificial intelligence
A systematic discipline, which focuses on developing, validating and deploying various machine learning algorithms for industrial applications with sustainable performance.
18
Industrial Artificial Intelligence
Jay Lee, Hossein Davari, Jaskaran Singh, Vibhor Pandhare, Industrial Artificial Intelligence for industry 4.0-based manufacturingsystems, Manufacturing Letters, Volume 18, 2018, Pages 20-23,
AI and Industry 4.0
19Adapted from: Jinjiang Wang, Yulin Ma, Laibin Zhang, Robert X. Gao, Dazhong Wu, Deep learning for smart manufacturing: Methods and applications, Journal of Manufacturing Systems, Volume 48, Part C, 2018, Pages 144-156,
AI/ML Enabled Advanced Analytics
Capture Products’ Condition,
environment and operation
Descriptive(What happened)
Examine the causes of reduced
product performance or
detect failure
Diagnostics(Why it happened)
Predict quality and patterns that signal impending events
Predictive(What will happen)
Identify measures to improve outcomes or
correct problems
Prescriptive(What action to take)
Product Company Manufacturer Supplier
Data AggregationSmart, connected products(Location, condition, use, etc.)
Enterprise(Service histories, warranty
status, etc.)
External(Price, weather, supplier
inventory, etc.)
Deep insights
Data Processing
Decision making and applications
Smart Connected Process
Knowledge
Pattern
Data
● Analytics technology (A),
● Big data technology (B),
● Cloud or Cyber technology (C),
● Domain knowhow (D) and
● Evidence (E)
20
Key elements in Industrial AI
21
Industrial AI
Figure: Comparison of Industrial AI with other learning systems
Jay Lee, Hossein Davari, Jaskaran Singh, Vibhor Pandhare, Industrial Artificial Intelligence for industry 4.0-based manufacturingsystems, Manufacturing Letters, Volume 18, 2018, Pages 20-23,
● Machine-to-machine interactions
● Machine-to-human interactions
● Data quality
● Cyber security
22
Challenges of Industrial AI
Part 3: The applied AI workflow
24
25
The Industrial AI stack
Deployment, maintenance and support
Validation
“Solving the problem”
Representation
Data collection and processing
Business UnderstandingBusiness Understanding
26
Planning, scheduling, etc.
Common in industrial problems everywhere:● How should we schedule a workforce?● How to order manufacturing steps in a product variant?● How to order individual manufacturing orders/items?● On what units should which maintenance be performed and
when?… etc.
Typically, a deep understanding of the business is needed.
27
The Industrial AI stack
Deployment, maintenance and support
Validation
“Solving the problem”
Representation
Data collection and processing
Business Understanding
Data from real applications is dirty:● Duplicates and missing data● Values with special meaning (ID 9999 means ”missing”)● Invalid data● Logically inconsistent data● Mystery data (railway cars which are 600 meters long)● Spiking data (temperature is 10e+10 for 1 millisecond)● Sensor drift, ”almost” values (0.6% really means 0.0%;
100.6% means 100%)● Multiple data files which are not in sync ● Misspellings● Different wordings
Data preparation and cleaning takes a long time!Validity threat: data cleaning removes realistic details
28
Data cleaning and preparation
29
Example
31
The Industrial AI stack
Deployment, maintenance and support
Validation
“Solving the problem”
Representation
Data collection and processing
Business Understanding
● Before a method is chosen, the representation should be considered● For machine learning – what should be the input?● E.g. vibration/noise analysis – representation in time/space or frequency
domain?● For planning, scheduling, simulation – what model
abstraction should be used?● E.g. microscopic model of robot movements, mesoscopic model of discrete
manufacturing steps, or macroscopic model of completion time distribution for product variants.
● In both cases, the representation of the problem can impact performance substantially.
● Finding the right representation requires in-depth understanding of the application!
● Stakeholders must agree to modeling assumptions!
32
Representation
33
The Industrial AI stack
Deployment, maintenance and support
Validation
“Solving the problem”
Representation
Data collection and processing
Business Understanding
● ML tasks are typically classified into following categories, depending on the nature of the learning "signal" or "feedback" available:
● Supervised learning – it uses inputs and their desired outputs• The program is “trained” on a pre-defined set of “training
examples”, which then facilitate its ability to reach an accurate conclusion when given new data.
● Unsupervised learning - no labels are given to the learning algorithm• The program is given a bunch of data and must find
patterns and relationships therein.
34
Types of ML
Some problems in SupervisedMachine Learning
35
Supervised Machine Learning
36
Problems in supervised ML
In classification, inputs are divided into two or more classes, and the learner must produce a model that assigns unseen inputs to one or more of these classes.
In regression, the outputs are continuous rather than discrete.
- An example of a regression problem would be the prediction of the length of a salmon as a function of its age and weight.
37
38
Machine Learning Algorithms
ML
Logic
Graphical
modelSupport vectors
Neu
ral
Net
wor
ks
Genetic
Programs
Accuracy
Squa
red
Erro
r
Fitness
Posterior
Probability
Margin
Inverse Deduction
Gra
dien
t Des
cent
Probabilistic Inference Genetic Search
Constrained Optim
ization
Conn
ectio
nist
Evolutionaries
BayesiansAnalogizers
Symbolists
REPR. EVAL.OPT.
• REPR: Representation• EVAL: Evaluation• OPT: Optimization
Some problems in UnsupervisedMachine Learning
39
Unsupervised Machine learning
40
Problems in unsupervised ML
In clustering, a set of inputs is to be divided into groups. - Unlike in classification, the groups are not known
beforehand, making this typically an unsupervised task.
Density estimation finds the distribution of inputs in some space.
Dimensionality reduction simplifies inputs by mapping them into a lower-dimensional space.
41
42
The Industrial AI stack
Deployment, maintenance and support
Validation
“Solving the problem”
Representation
Data collection and cleaning
Business Understanding
● During model building, random patterns in the sample areeasily found which are might not be present in the wholepopulation.
● To justify the performance of the built predictive model, thevalidation should be done with data points that were neverused while building the model.
● For a set of ML variants, optimize parameter selection (learn) on the training set.
● Find the ML variant which performs best on the cross-validation set (e.g. polynomial degree)
● Estimate the generalization error using the test set
43
Model Validation
● You’ve trained your model, now what?
● Overtraining – model doesn’t generalize to (perform well on) new data.
● “Validation” or evaluation is used to estimate the performance on new data, i.e. how the model would perform when actually used
● Validation results will always be too optimistic!
Overtraining
✕ Few data samples ✕ Complicated model ✕ Similar training, test and validation sets✕ Fine-tuning parameters ✕ Evaluating several models with the same validation set
Image by Chabacano / CC BY
● Training Set:● This set is used for training the predictive models.
● Validation Set:● Fixing the values of different parameters of the built model is donewith this set.
● Test Set:● Accuracy of the built model is determined using this set.
● In common practice, test set is made with larger part of thedata containing data points of all possible outcomes. The rest ofthe dataset is split into two sets for validation and testing.
● A widely used ratio of data splitting is 60:20:20 for training,validation and testing.
45
Data Splitting
46
K-fold Cross Validation
CV # 1
CV # 2
CV # 3
CV # 4
CV # 5
Training Set Validation Set
Original Data (n = 20)
Example: Dataset Splitting in 5 – fold Cross Validation.
Confusion Matrix
47
Performance Measures
Case 1
Accuracy = 85%
● Class 1 = 10● Class 2 =10
9 1
2 8
Case 2
Accuracy = 90%
● Class 1 = 9● Class 2 =1
9 0
1 0
● F1 score tells us how precise and robust a model is.
● It is the harmonic mean of Precision andRecall values
!1 = 2 11
%&'()*)+, +1
.'(/00
● When the False Negatives and False Positives are crucial
● When there are imbalanced classes
● Greater F1 Score indicates better performance for prediction models.
48
F1 ScoreTP FP
FN TN
49
Performance Measures
● Receiver Operating Characteristics(ROC)● ROC is the widely used metric for
validating binary classification models.
● Two basic terms for AUC:● Sensitivity: In other words, it is called True
Positive Rate (TPR). Sensitivity is calculated fromthe values of confusion matrix –
!"#$%&%'%&( )*+ = )*)* + ./
● Specificity: It is also termed as False PositiveRate (FPR). It is calculated with the formula –
!0"1%2%1%&( .*+ = )/)/ + .*
● Example: Consider a test of Covid-19● Test has 90% sensitivity that means the test will
correctly return a positive result for 90% of peoplewho have the disease. But will return a negative result— a false-negative — for 10% of the people who havethe disease and should have tested positive.
● What about specificity?
51
The Industrial AI stack
Deployment, maintenance and support
Validation
“Solving the problem”
Representation
Data collection and cleaning
Business Understanding
Out of scope of this lecture
52
The Industrial AI stack in reality
Deployment, maintenance and support
Validation
“Solving the problem”
Representation
Data collection and cleaning
Business Understanding
Often 80% of total effort
Foundations of value creation (20% of effort)
End-user value
Part 4: Use cases
53
Example: Machine Health Monitoring
54
ConventionalData-driven
MHMS
Physical-basedMHMS
Deep Learning based MHMS
Solution
Monitored Machine
Data Acquisition
Solution
Solution
Monitored Machine
Monitored Machine
Data Acquisition
Data Acquisition
Hand design physical model
Source: Rui Zhao, Ruqiang Yan, Zhenghua Chen, Kezhi Mao, Peng Wang, Robert X. Gao, Deep learning and its applications to machine health monitoring, Mechanical Systems and Signal Processing, Volume 115, 2019, Pages 213-237, ISSN 0888-3270
Example: Data Analytics in Industry 4.0
55
A Case Study in Power Transfer Unit
Project: AUTOMADProject Leader: Dr. Mobyen Uddin Ahmed, Docent Contact: [email protected]
Example: Monitoring and Quality Control
56
Prototype running at Volvo and Chalmers, cloud based solution,implemented by Ivan Tomašić MDH
Project Leader: Prof. Peter Funk, MDH Contact: [email protected]
Example: Pulp and paper
58Pr
oduc
tion
Rat
e
Qua
lityOpe
rati
ng c
ostDetection of Digester Faults
• Screen clogging.
• Hang ups and
• ChannellingDetection of Anomalies
• Sensor faults
• Something is wrong
Prediction
• Kappa value
Contact: [email protected] and [email protected]
● Big Data and Cloud Computing for Industrial Applications ● Study period 2020-11-09 - 2021-01-17
● Visit mdh.se/premium
60
Production engineering course autumn 2020
61
Interesting Reading
Thank you
63