SlideShare a Scribd company logo
Deploying Machine Learning under Uncertainty in
Cyber-Physical Manufacturing System (CPMS)
Bang Xiang Yong
Supervisor: Dr. Alexandra Brintrup
Advisor: Prof Duncan McFarlane
• In the recent Made Smarter Review [1], the total value of AI/ML technologies to the United Kingdom industry
was estimated to be £198.7bn between 2017-2027.
• Classical statistic methods such as linear regression and multivariate regression were designed for data with
few dozen input variables and sample sizes which would be considered small by today’s standard. By contrast,
many ML algorithms are capable of analyzing high dimensional data [6].
• Classical statistics and ML are not mutually exclusive : classical statistics can be used to pre-process the data
before feeding into ML model such as Artificial Neural Network (ANN).
• As manufacturers begin to capture large volumes of data describing their process through Cyber Physical
Manufacturing Systems (CPMS), there is a need to derive value from the sensor data streams which are
complex, high dimensional and non-linear.
Introduction
• Manufacturing systems are characterised by heterogeneous distributed systems, dynamic and unpredictable
environments.
• A missing ingredient from existing approaches is the measurement of uncertainty of ML – current approaches
focus on
• maximising the accuracy of predictions
• executing them in the fastest way via scalable algorithms
• A first motivating question would be, "Can we still trust the predictions of ML, given that faults and
environment changes have occurred?".
• Without reasoning on the uncertainty of data-driven systems in dynamic environments, manufacturers are
exposed to risk in events – sensor failures, overconfident and erroneous predictions
Problem Statement
• Definition of Machine learning
• Data-driven approaches which build mathematical models based on sample data to make predictions or
decisions. [21]
• In my definition : an automated process of building a mathematical model using example data,
characterised by training and inference phases which relies on minimal prior knowledge of the underlying
relationships between variables in data.
• Methodologies such as CRISP-DM [24] have been developed to provide industrial standard frameworks for
encapsulating the process of data analytics, which have been adapted to various domains [25].
Research Background
• Definition of Cyber-Physical Manufacturing Systems:
• Rajkumar et al. [30] defined CPS as systems which bridge the cyber-world of computing and communications with the
physical world.
• Sanislav and Miclea [32] emphasised CPS through the input and feedback to and fro the cyber-physical environments,
management of distributed control, real-time executions, large spatial distribution and large scale control systems
within systems.
• My definition: Distributed and highly interconnected computer systems in manufacturing which are capable of obtaining
real-time data from the physical world through sensors and acting upon them.
Research Background
Research Background
Systematic literature review
• Systematic literature reviews are characterised by explicit
and reproducible steps with clear guidelines to reproduce.
• The methods of systematic review advocated by Pickering
and Byrne [39] were adopted.
• Modern manufacturing has focused less on studying ML methods compared to
other technologies as shown in a review by Liao et al. [41].
Systematic literature review
RQ1: What are the types of learning algorithms and technologies used in ML-CPMS?
RQ2: What are the types and properties of data in ML-CPMS?
RQ3: What are the characteristics and communication protocols of ML-CPMS framework?
RQ4: What are the application areas and use cases of ML-CPMS?
RQ5: What are the challenges and issues in ML-CPMS?
"What is the state of machine learning research in Cyber-Physical Manufacturing Systems?"
Main review question
Sub review questions
Systematic literature review
Keyword search at Scopus and Web of Science:
("cyber physical system*" OR "industrial internet") AND
("machine learning“ OR "knowledge discover*" OR "data
mine*" OR "artificial intelligence") AND "manufacturing"
Customised web application for systematic literature review
RQ1: What are the types of learning algorithms and technologies used in ML-CPMS?
RQ1: What are the types of learning algorithms and technologies used in ML-CPMS?
RQ2: What are the types and properties of data in ML-CPMS?
RQ3: What are the characteristics and communication protocols of ML-CPMS framework?
RQ4: What are the application areas and use cases of ML-CPMS?
RQ5: What are the challenges and issues in ML-CPMS?
There is not a uniform definition for uncertainty in ML-CPMS:
• Uncertainty of raw measurements [82]
• Economics and systems perspective [109]
• Decision-making [110]
• Events [111]
• Part quality [58]
• Model uncertainty [70, 112].
RQ5: What are the challenges and issues in ML-CPMS?
Uncertainty of machine learning in CPMS
Relevant scenarios for measuring uncertainty of ML were inferred:
• Overconfident predictions
• Concept drift
• Sensor faults and degradation
Uncertainty of ML - The level of confidence of the output given by a model.
Research Gaps
1. Lack of study addressing uncertainty holistically
2. Methods to quantify uncertainty of predictions.
3. Isolated study of scenarios relevant to uncertainty.
4. Acting upon quantified uncertainty.
Towards developing a framework for deploying machine learning in Cyber-
Physical Manufacturing Systems which quantifies uncertainty of prediction
and acts on it.
Research aim
Research Questions and Tasks
1. What are the characteristics of ML-CPMS and the uncertainty of ML-CPMS?
• Task 1.1 : Carry out systematic literature review to understand the status of MLCPMS
research by enumerating the types of learning algorithms, technologies,
challenges and issues which contribute to uncertainty.
• Task 1.2 : Study extant literature to identify the sources of ML uncertainty,
relevant scenarios and existing methods of dealing with these scenarios.
• Task 1.3 : Define a conceptual model of uncertainty in ML-CPMS based on Task 1.2.
2. How can a framework of ML-CPMS manage the uncertainties?
• Task 2.1 : Define components of framework and algorithms for uncertainty quantification
of ML predictions, sensors fault detection and concept drift detection.
• Task 2.2 : Identify ML algorithms which are capable of expressing uncertainty in
their predictions.
• Task 2.3 : Identify methods and algorithms which are able to reduce the uncertainty
of prediction.
• Task 2.4 : Combine outputs of Task 2.1, Task 2.2, Task 2.3 into a framework with
the ability to quantify uncertainty, detect the sources and subsequently reduce it.
Research Questions and Tasks
3. How effective is the proposed framework in handling uncertainty?
• Task 3.1 : Identify quantitative or qualitative performance metrics of proposed
framework.
• Task 3.2 : Develop a testbed to implement algorithms and ML-CPMS framework
on real industrial data-sets.
• Task 3.3 : Define and conduct laboratory experiments to evaluate the performance
of proposed framework.
• Task 3.4 : Deploy proposed framework in industrial case studies.
• Task 3.5 : Evaluate the outcomes of the deployment and compare with the
laboratory studies.
Research Scope Research Methodology
Relevant journals
1. International Journal of Production Research
2. Future Generation Computing Systems
3. Sensors
4. Computers in Industry
5. Journal of Manufacturing and Materials Processing
6. Journal of Machine Learning Research.
Dissemination
• Webinar was recorded to disseminate the developed agent-based system for handling uncertainty,
agentMET4FOF software package and is publicly available for viewing
• SOHOMA19 paper titled "Multi Agent System for Machine Learning Under
Uncertainty in Cyber Physical Manufacturing System"
Current publication
Preliminary Conceptual Framework
Preliminary Conceptual Framework
Preliminary Conceptual Framework
Exploratory Studies : 1. Agent-based system
“How can a software architecture for machine
learning be implemented which is modular and
flexible to investigate various use cases,
configurations, scenarios, algorithms in the
context of Cyber-Physical Manufacturing
System?”
Screenshot of
agentMET4FOF
web application
Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML
Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML
Classification of Hydraulic System condition using Bayesian Neural Network
Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML
Classification of Hydraulic System condition using Bayesian Neural Network
Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML
Regression for prognosis of electromechanical cylinder using Bayesian Neural Network
Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML
Regression for quality prediction of radial forge, AFRC
Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML
Regression for quality prediction of radial forge, AFRC
Exploratory study: 3. Deployment of agent-based system on embedded systems
Future works
1. Further develop, implement and analyse proposed architecture based on outline
performance criteria
2. Simulate behaviour of multi-agent system under conditions of uncertain scenarios
3. Analyse relationship between predictive uncertainty and uncertainty and how agents
can make decisions to manage uncertainty of ML models
Contributions
• Theoretical contribution - theoretical framework of ML-CPMS which quantifies, detects and acts
on uncertainty of prediction.
• Practical contribution - software package based on agent-based system which readily implements
the methods for quantifying, detecting and acts on uncertainty in ML-CPMS

More Related Content

What's hot

Predatory Publications and Software Tools for Identification
Predatory Publications and Software Tools for IdentificationPredatory Publications and Software Tools for Identification
Predatory Publications and Software Tools for Identification
Saptarshi Ghosh
 
Systematic Literature Reviews and Systematic Mapping Studies
Systematic Literature Reviews and Systematic Mapping StudiesSystematic Literature Reviews and Systematic Mapping Studies
Systematic Literature Reviews and Systematic Mapping Studies
alessio_ferrari
 
Literature Review
Literature ReviewLiterature Review
Literature Review
AIMS Education
 
Literature survey
Literature surveyLiterature survey
Literature survey
Anamika Banerjee
 
Word representations in vector space
Word representations in vector spaceWord representations in vector space
Word representations in vector space
Abdullah Khan Zehady
 
Machine learning clustering
Machine learning clusteringMachine learning clustering
Machine learning clustering
CosmoAIMS Bassett
 
Basics of research paper publishing
Basics of research paper publishingBasics of research paper publishing
Basics of research paper publishing
Azam Shams
 
Literature review and the PhD
Literature review and the PhDLiterature review and the PhD
Literature review and the PhD
Prof Simon Haslett
 
Fundamentals of literature review
Fundamentals of literature reviewFundamentals of literature review
Fundamentals of literature review
riveraleyann56
 
Research Process and Research Design.
Research Process and Research Design.Research Process and Research Design.
Research Process and Research Design.
Utkarsh Gupta
 
Presentation 1; research methods
Presentation 1; research methodsPresentation 1; research methods
Presentation 1; research methods
Syed Iqrar Hussain
 
Bibliometric study
Bibliometric studyBibliometric study
Thesis Proposal Presentation
Thesis Proposal PresentationThesis Proposal Presentation
Thesis Proposal Presentation
Turki Alshammary (CBAP, PMI-PBA, PMP)
 
Open Access Initiatives
Open Access Initiatives Open Access Initiatives
Open Access Initiatives
drningappaarabagonda
 
Sherpa Romeo
Sherpa RomeoSherpa Romeo
Sherpa RomeoMKH-QMUL
 
Philosophy and ethics
Philosophy and ethics Philosophy and ethics
Philosophy and ethics
jagannath Dange
 
Impact factor of Journal as per Journal citation report, SNIP, SJR, IPP, Cite...
Impact factor of Journal as per Journal citation report, SNIP, SJR, IPP, Cite...Impact factor of Journal as per Journal citation report, SNIP, SJR, IPP, Cite...
Impact factor of Journal as per Journal citation report, SNIP, SJR, IPP, Cite...
Omprakash saini saini
 
1 - Systematic Literature Reviews: introduction and methods
1 - Systematic Literature Reviews: introduction and methods1 - Systematic Literature Reviews: introduction and methods
1 - Systematic Literature Reviews: introduction and methods
Vittorio Scarano
 
How to publish in an isi journal حنان القرشي
How to publish in an isi journal  حنان القرشيHow to publish in an isi journal  حنان القرشي
How to publish in an isi journal حنان القرشي
vdsr_ksu
 
Impact factor of journals
Impact factor of journalsImpact factor of journals
Impact factor of journalsDr. Pinki Insan
 

What's hot (20)

Predatory Publications and Software Tools for Identification
Predatory Publications and Software Tools for IdentificationPredatory Publications and Software Tools for Identification
Predatory Publications and Software Tools for Identification
 
Systematic Literature Reviews and Systematic Mapping Studies
Systematic Literature Reviews and Systematic Mapping StudiesSystematic Literature Reviews and Systematic Mapping Studies
Systematic Literature Reviews and Systematic Mapping Studies
 
Literature Review
Literature ReviewLiterature Review
Literature Review
 
Literature survey
Literature surveyLiterature survey
Literature survey
 
Word representations in vector space
Word representations in vector spaceWord representations in vector space
Word representations in vector space
 
Machine learning clustering
Machine learning clusteringMachine learning clustering
Machine learning clustering
 
Basics of research paper publishing
Basics of research paper publishingBasics of research paper publishing
Basics of research paper publishing
 
Literature review and the PhD
Literature review and the PhDLiterature review and the PhD
Literature review and the PhD
 
Fundamentals of literature review
Fundamentals of literature reviewFundamentals of literature review
Fundamentals of literature review
 
Research Process and Research Design.
Research Process and Research Design.Research Process and Research Design.
Research Process and Research Design.
 
Presentation 1; research methods
Presentation 1; research methodsPresentation 1; research methods
Presentation 1; research methods
 
Bibliometric study
Bibliometric studyBibliometric study
Bibliometric study
 
Thesis Proposal Presentation
Thesis Proposal PresentationThesis Proposal Presentation
Thesis Proposal Presentation
 
Open Access Initiatives
Open Access Initiatives Open Access Initiatives
Open Access Initiatives
 
Sherpa Romeo
Sherpa RomeoSherpa Romeo
Sherpa Romeo
 
Philosophy and ethics
Philosophy and ethics Philosophy and ethics
Philosophy and ethics
 
Impact factor of Journal as per Journal citation report, SNIP, SJR, IPP, Cite...
Impact factor of Journal as per Journal citation report, SNIP, SJR, IPP, Cite...Impact factor of Journal as per Journal citation report, SNIP, SJR, IPP, Cite...
Impact factor of Journal as per Journal citation report, SNIP, SJR, IPP, Cite...
 
1 - Systematic Literature Reviews: introduction and methods
1 - Systematic Literature Reviews: introduction and methods1 - Systematic Literature Reviews: introduction and methods
1 - Systematic Literature Reviews: introduction and methods
 
How to publish in an isi journal حنان القرشي
How to publish in an isi journal  حنان القرشيHow to publish in an isi journal  حنان القرشي
How to publish in an isi journal حنان القرشي
 
Impact factor of journals
Impact factor of journalsImpact factor of journals
Impact factor of journals
 

Similar to First Year Report, PhD presentation

INCOSE IS 2019: AI and Systems Engineering
INCOSE IS 2019: AI and Systems EngineeringINCOSE IS 2019: AI and Systems Engineering
INCOSE IS 2019: AI and Systems Engineering
CARLOS III UNIVERSITY OF MADRID
 
AI4SE: Challenges and opportunities in the integration of Systems Engineering...
AI4SE: Challenges and opportunities in the integration of Systems Engineering...AI4SE: Challenges and opportunities in the integration of Systems Engineering...
AI4SE: Challenges and opportunities in the integration of Systems Engineering...
CARLOS III UNIVERSITY OF MADRID
 
International journal of engineering issues vol 2015 - no 2 - paper4
International journal of engineering issues   vol 2015 - no 2 - paper4International journal of engineering issues   vol 2015 - no 2 - paper4
International journal of engineering issues vol 2015 - no 2 - paper4
sophiabelthome
 
SESE 2021: Where Systems Engineering meets AI/ML
SESE 2021: Where Systems Engineering meets AI/MLSESE 2021: Where Systems Engineering meets AI/ML
SESE 2021: Where Systems Engineering meets AI/ML
CARLOS III UNIVERSITY OF MADRID
 
M.tech cse 10july13 (1)
M.tech cse  10july13 (1)M.tech cse  10july13 (1)
M.tech cse 10july13 (1)
vijay707070
 
MSIE-06-T-M2S1-Cyber-Physical-System-and-Data-Security.pptx
MSIE-06-T-M2S1-Cyber-Physical-System-and-Data-Security.pptxMSIE-06-T-M2S1-Cyber-Physical-System-and-Data-Security.pptx
MSIE-06-T-M2S1-Cyber-Physical-System-and-Data-Security.pptx
MahendraShukla27
 
ESTIMATING THE EFFORT OF MOBILE APPLICATION DEVELOPMENT
ESTIMATING THE EFFORT OF MOBILE APPLICATION DEVELOPMENTESTIMATING THE EFFORT OF MOBILE APPLICATION DEVELOPMENT
ESTIMATING THE EFFORT OF MOBILE APPLICATION DEVELOPMENT
csandit
 
Model-Simulation-and-Measurement-Based Systems Engineering of Power System Sy...
Model-Simulation-and-Measurement-Based Systems Engineering of Power System Sy...Model-Simulation-and-Measurement-Based Systems Engineering of Power System Sy...
Model-Simulation-and-Measurement-Based Systems Engineering of Power System Sy...
Luigi Vanfretti
 
2016-03-02 research seminar
2016-03-02 research seminar2016-03-02 research seminar
2016-03-02 research seminar
ifi8106tlu
 
ML-Approach-for-Telecom-Network-Operations-Management.pptx
ML-Approach-for-Telecom-Network-Operations-Management.pptxML-Approach-for-Telecom-Network-Operations-Management.pptx
ML-Approach-for-Telecom-Network-Operations-Management.pptx
ssuser818de4
 
Automated Feature Selection and Churn Prediction using Deep Learning Models
Automated Feature Selection and Churn Prediction using Deep Learning ModelsAutomated Feature Selection and Churn Prediction using Deep Learning Models
Automated Feature Selection and Churn Prediction using Deep Learning Models
IRJET Journal
 
Cybernetics in supply chain management
Cybernetics in supply chain managementCybernetics in supply chain management
Cybernetics in supply chain management
Luis Cabrera
 
ICMCSI 2023 PPT 1074.pptx
ICMCSI 2023 PPT 1074.pptxICMCSI 2023 PPT 1074.pptx
ICMCSI 2023 PPT 1074.pptx
ajagbesundayadeola
 
Machine Learning (ML) in Wireless Sensor Networks (WSNs)
Machine Learning (ML) in Wireless Sensor Networks (WSNs)Machine Learning (ML) in Wireless Sensor Networks (WSNs)
Machine Learning (ML) in Wireless Sensor Networks (WSNs)
mabualsh
 
Pathways Overview For Open House 19 Sep2010
Pathways Overview For Open House   19 Sep2010Pathways Overview For Open House   19 Sep2010
Pathways Overview For Open House 19 Sep2010
jmorriso
 
MODEL CHECKERS –TOOLS AND LANGUAGES FOR SYSTEM DESIGN- A SURVEY
MODEL CHECKERS –TOOLS AND LANGUAGES FOR SYSTEM DESIGN- A SURVEYMODEL CHECKERS –TOOLS AND LANGUAGES FOR SYSTEM DESIGN- A SURVEY
MODEL CHECKERS –TOOLS AND LANGUAGES FOR SYSTEM DESIGN- A SURVEY
csandit
 
Spring 2012 Seminar Catalogue
Spring 2012 Seminar CatalogueSpring 2012 Seminar Catalogue
Spring 2012 Seminar Cataloguealidade
 
Cloud-based control systems: a systematic literature review
Cloud-based control systems: a systematic literature reviewCloud-based control systems: a systematic literature review
Cloud-based control systems: a systematic literature review
International Journal of Reconfigurable and Embedded Systems
 
Ibm colloquium 070915_nyberg
Ibm colloquium 070915_nybergIbm colloquium 070915_nyberg
Ibm colloquium 070915_nyberg
diannepatricia
 
Accelerated Prototyping of Cyber Physical Systems in an Incubator Context
Accelerated Prototyping of Cyber Physical Systems in an Incubator ContextAccelerated Prototyping of Cyber Physical Systems in an Incubator Context
Accelerated Prototyping of Cyber Physical Systems in an Incubator Context
Sreyas Sriram
 

Similar to First Year Report, PhD presentation (20)

INCOSE IS 2019: AI and Systems Engineering
INCOSE IS 2019: AI and Systems EngineeringINCOSE IS 2019: AI and Systems Engineering
INCOSE IS 2019: AI and Systems Engineering
 
AI4SE: Challenges and opportunities in the integration of Systems Engineering...
AI4SE: Challenges and opportunities in the integration of Systems Engineering...AI4SE: Challenges and opportunities in the integration of Systems Engineering...
AI4SE: Challenges and opportunities in the integration of Systems Engineering...
 
International journal of engineering issues vol 2015 - no 2 - paper4
International journal of engineering issues   vol 2015 - no 2 - paper4International journal of engineering issues   vol 2015 - no 2 - paper4
International journal of engineering issues vol 2015 - no 2 - paper4
 
SESE 2021: Where Systems Engineering meets AI/ML
SESE 2021: Where Systems Engineering meets AI/MLSESE 2021: Where Systems Engineering meets AI/ML
SESE 2021: Where Systems Engineering meets AI/ML
 
M.tech cse 10july13 (1)
M.tech cse  10july13 (1)M.tech cse  10july13 (1)
M.tech cse 10july13 (1)
 
MSIE-06-T-M2S1-Cyber-Physical-System-and-Data-Security.pptx
MSIE-06-T-M2S1-Cyber-Physical-System-and-Data-Security.pptxMSIE-06-T-M2S1-Cyber-Physical-System-and-Data-Security.pptx
MSIE-06-T-M2S1-Cyber-Physical-System-and-Data-Security.pptx
 
ESTIMATING THE EFFORT OF MOBILE APPLICATION DEVELOPMENT
ESTIMATING THE EFFORT OF MOBILE APPLICATION DEVELOPMENTESTIMATING THE EFFORT OF MOBILE APPLICATION DEVELOPMENT
ESTIMATING THE EFFORT OF MOBILE APPLICATION DEVELOPMENT
 
Model-Simulation-and-Measurement-Based Systems Engineering of Power System Sy...
Model-Simulation-and-Measurement-Based Systems Engineering of Power System Sy...Model-Simulation-and-Measurement-Based Systems Engineering of Power System Sy...
Model-Simulation-and-Measurement-Based Systems Engineering of Power System Sy...
 
2016-03-02 research seminar
2016-03-02 research seminar2016-03-02 research seminar
2016-03-02 research seminar
 
ML-Approach-for-Telecom-Network-Operations-Management.pptx
ML-Approach-for-Telecom-Network-Operations-Management.pptxML-Approach-for-Telecom-Network-Operations-Management.pptx
ML-Approach-for-Telecom-Network-Operations-Management.pptx
 
Automated Feature Selection and Churn Prediction using Deep Learning Models
Automated Feature Selection and Churn Prediction using Deep Learning ModelsAutomated Feature Selection and Churn Prediction using Deep Learning Models
Automated Feature Selection and Churn Prediction using Deep Learning Models
 
Cybernetics in supply chain management
Cybernetics in supply chain managementCybernetics in supply chain management
Cybernetics in supply chain management
 
ICMCSI 2023 PPT 1074.pptx
ICMCSI 2023 PPT 1074.pptxICMCSI 2023 PPT 1074.pptx
ICMCSI 2023 PPT 1074.pptx
 
Machine Learning (ML) in Wireless Sensor Networks (WSNs)
Machine Learning (ML) in Wireless Sensor Networks (WSNs)Machine Learning (ML) in Wireless Sensor Networks (WSNs)
Machine Learning (ML) in Wireless Sensor Networks (WSNs)
 
Pathways Overview For Open House 19 Sep2010
Pathways Overview For Open House   19 Sep2010Pathways Overview For Open House   19 Sep2010
Pathways Overview For Open House 19 Sep2010
 
MODEL CHECKERS –TOOLS AND LANGUAGES FOR SYSTEM DESIGN- A SURVEY
MODEL CHECKERS –TOOLS AND LANGUAGES FOR SYSTEM DESIGN- A SURVEYMODEL CHECKERS –TOOLS AND LANGUAGES FOR SYSTEM DESIGN- A SURVEY
MODEL CHECKERS –TOOLS AND LANGUAGES FOR SYSTEM DESIGN- A SURVEY
 
Spring 2012 Seminar Catalogue
Spring 2012 Seminar CatalogueSpring 2012 Seminar Catalogue
Spring 2012 Seminar Catalogue
 
Cloud-based control systems: a systematic literature review
Cloud-based control systems: a systematic literature reviewCloud-based control systems: a systematic literature review
Cloud-based control systems: a systematic literature review
 
Ibm colloquium 070915_nyberg
Ibm colloquium 070915_nybergIbm colloquium 070915_nyberg
Ibm colloquium 070915_nyberg
 
Accelerated Prototyping of Cyber Physical Systems in an Incubator Context
Accelerated Prototyping of Cyber Physical Systems in an Incubator ContextAccelerated Prototyping of Cyber Physical Systems in an Incubator Context
Accelerated Prototyping of Cyber Physical Systems in an Incubator Context
 

More from Bang Xiang Yong

Bayesian Autoencoders (BAE) & Honest Thoughts on research
Bayesian Autoencoders (BAE) & Honest Thoughts on research Bayesian Autoencoders (BAE) & Honest Thoughts on research
Bayesian Autoencoders (BAE) & Honest Thoughts on research
Bang Xiang Yong
 
Bayesian Autoencoders for anomaly detection in industrial environments
Bayesian Autoencoders for anomaly detection in industrial environmentsBayesian Autoencoders for anomaly detection in industrial environments
Bayesian Autoencoders for anomaly detection in industrial environments
Bang Xiang Yong
 
Uncertainty Quantification with Unsupervised Deep learning and Multi Agent Sy...
Uncertainty Quantification with Unsupervised Deep learning and Multi Agent Sy...Uncertainty Quantification with Unsupervised Deep learning and Multi Agent Sy...
Uncertainty Quantification with Unsupervised Deep learning and Multi Agent Sy...
Bang Xiang Yong
 
Proposal for Linking Concept Drift and uncertainty of Machine learning
Proposal for Linking Concept Drift and uncertainty of Machine learningProposal for Linking Concept Drift and uncertainty of Machine learning
Proposal for Linking Concept Drift and uncertainty of Machine learning
Bang Xiang Yong
 
PhD First Year Conference (MAY 2019)
PhD First Year Conference (MAY 2019)PhD First Year Conference (MAY 2019)
PhD First Year Conference (MAY 2019)
Bang Xiang Yong
 
SOHOMA19 Conference slides
SOHOMA19 Conference slidesSOHOMA19 Conference slides
SOHOMA19 Conference slides
Bang Xiang Yong
 
Use cases - agentMET4FOF
Use cases - agentMET4FOFUse cases - agentMET4FOF
Use cases - agentMET4FOF
Bang Xiang Yong
 
Espriex 2017 slides
Espriex 2017 slidesEspriex 2017 slides
Espriex 2017 slides
Bang Xiang Yong
 

More from Bang Xiang Yong (8)

Bayesian Autoencoders (BAE) & Honest Thoughts on research
Bayesian Autoencoders (BAE) & Honest Thoughts on research Bayesian Autoencoders (BAE) & Honest Thoughts on research
Bayesian Autoencoders (BAE) & Honest Thoughts on research
 
Bayesian Autoencoders for anomaly detection in industrial environments
Bayesian Autoencoders for anomaly detection in industrial environmentsBayesian Autoencoders for anomaly detection in industrial environments
Bayesian Autoencoders for anomaly detection in industrial environments
 
Uncertainty Quantification with Unsupervised Deep learning and Multi Agent Sy...
Uncertainty Quantification with Unsupervised Deep learning and Multi Agent Sy...Uncertainty Quantification with Unsupervised Deep learning and Multi Agent Sy...
Uncertainty Quantification with Unsupervised Deep learning and Multi Agent Sy...
 
Proposal for Linking Concept Drift and uncertainty of Machine learning
Proposal for Linking Concept Drift and uncertainty of Machine learningProposal for Linking Concept Drift and uncertainty of Machine learning
Proposal for Linking Concept Drift and uncertainty of Machine learning
 
PhD First Year Conference (MAY 2019)
PhD First Year Conference (MAY 2019)PhD First Year Conference (MAY 2019)
PhD First Year Conference (MAY 2019)
 
SOHOMA19 Conference slides
SOHOMA19 Conference slidesSOHOMA19 Conference slides
SOHOMA19 Conference slides
 
Use cases - agentMET4FOF
Use cases - agentMET4FOFUse cases - agentMET4FOF
Use cases - agentMET4FOF
 
Espriex 2017 slides
Espriex 2017 slidesEspriex 2017 slides
Espriex 2017 slides
 

Recently uploaded

Criminal IP - Threat Hunting Webinar.pdf
Criminal IP - Threat Hunting Webinar.pdfCriminal IP - Threat Hunting Webinar.pdf
Criminal IP - Threat Hunting Webinar.pdf
Criminal IP
 
standardisation of garbhpala offhgfffghh
standardisation of garbhpala offhgfffghhstandardisation of garbhpala offhgfffghh
standardisation of garbhpala offhgfffghh
ArpitMalhotra16
 
一比一原版(QU毕业证)皇后大学毕业证成绩单
一比一原版(QU毕业证)皇后大学毕业证成绩单一比一原版(QU毕业证)皇后大学毕业证成绩单
一比一原版(QU毕业证)皇后大学毕业证成绩单
enxupq
 
哪里卖(usq毕业证书)南昆士兰大学毕业证研究生文凭证书托福证书原版一模一样
哪里卖(usq毕业证书)南昆士兰大学毕业证研究生文凭证书托福证书原版一模一样哪里卖(usq毕业证书)南昆士兰大学毕业证研究生文凭证书托福证书原版一模一样
哪里卖(usq毕业证书)南昆士兰大学毕业证研究生文凭证书托福证书原版一模一样
axoqas
 
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
ewymefz
 
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
yhkoc
 
1.Seydhcuxhxyxhccuuxuxyxyxmisolids 2019.pptx
1.Seydhcuxhxyxhccuuxuxyxyxmisolids 2019.pptx1.Seydhcuxhxyxhccuuxuxyxyxmisolids 2019.pptx
1.Seydhcuxhxyxhccuuxuxyxyxmisolids 2019.pptx
Tiktokethiodaily
 
FP Growth Algorithm and its Applications
FP Growth Algorithm and its ApplicationsFP Growth Algorithm and its Applications
FP Growth Algorithm and its Applications
MaleehaSheikh2
 
一比一原版(UMich毕业证)密歇根大学|安娜堡分校毕业证成绩单
一比一原版(UMich毕业证)密歇根大学|安娜堡分校毕业证成绩单一比一原版(UMich毕业证)密歇根大学|安娜堡分校毕业证成绩单
一比一原版(UMich毕业证)密歇根大学|安娜堡分校毕业证成绩单
ewymefz
 
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
ukgaet
 
Investigate & Recover / StarCompliance.io / Crypto_Crimes
Investigate & Recover / StarCompliance.io / Crypto_CrimesInvestigate & Recover / StarCompliance.io / Crypto_Crimes
Investigate & Recover / StarCompliance.io / Crypto_Crimes
StarCompliance.io
 
做(mqu毕业证书)麦考瑞大学毕业证硕士文凭证书学费发票原版一模一样
做(mqu毕业证书)麦考瑞大学毕业证硕士文凭证书学费发票原版一模一样做(mqu毕业证书)麦考瑞大学毕业证硕士文凭证书学费发票原版一模一样
做(mqu毕业证书)麦考瑞大学毕业证硕士文凭证书学费发票原版一模一样
axoqas
 
Malana- Gimlet Market Analysis (Portfolio 2)
Malana- Gimlet Market Analysis (Portfolio 2)Malana- Gimlet Market Analysis (Portfolio 2)
Malana- Gimlet Market Analysis (Portfolio 2)
TravisMalana
 
Predicting Product Ad Campaign Performance: A Data Analysis Project Presentation
Predicting Product Ad Campaign Performance: A Data Analysis Project PresentationPredicting Product Ad Campaign Performance: A Data Analysis Project Presentation
Predicting Product Ad Campaign Performance: A Data Analysis Project Presentation
Boston Institute of Analytics
 
Best best suvichar in gujarati english meaning of this sentence as Silk road ...
Best best suvichar in gujarati english meaning of this sentence as Silk road ...Best best suvichar in gujarati english meaning of this sentence as Silk road ...
Best best suvichar in gujarati english meaning of this sentence as Silk road ...
AbhimanyuSinha9
 
Empowering Data Analytics Ecosystem.pptx
Empowering Data Analytics Ecosystem.pptxEmpowering Data Analytics Ecosystem.pptx
Empowering Data Analytics Ecosystem.pptx
benishzehra469
 
Jpolillo Amazon PPC - Bid Optimization Sample
Jpolillo Amazon PPC - Bid Optimization SampleJpolillo Amazon PPC - Bid Optimization Sample
Jpolillo Amazon PPC - Bid Optimization Sample
James Polillo
 
Tabula.io Cheatsheet: automate your data workflows
Tabula.io Cheatsheet: automate your data workflowsTabula.io Cheatsheet: automate your data workflows
Tabula.io Cheatsheet: automate your data workflows
alex933524
 
一比一原版(YU毕业证)约克大学毕业证成绩单
一比一原版(YU毕业证)约克大学毕业证成绩单一比一原版(YU毕业证)约克大学毕业证成绩单
一比一原版(YU毕业证)约克大学毕业证成绩单
enxupq
 
Ch03-Managing the Object-Oriented Information Systems Project a.pdf
Ch03-Managing the Object-Oriented Information Systems Project a.pdfCh03-Managing the Object-Oriented Information Systems Project a.pdf
Ch03-Managing the Object-Oriented Information Systems Project a.pdf
haila53
 

Recently uploaded (20)

Criminal IP - Threat Hunting Webinar.pdf
Criminal IP - Threat Hunting Webinar.pdfCriminal IP - Threat Hunting Webinar.pdf
Criminal IP - Threat Hunting Webinar.pdf
 
standardisation of garbhpala offhgfffghh
standardisation of garbhpala offhgfffghhstandardisation of garbhpala offhgfffghh
standardisation of garbhpala offhgfffghh
 
一比一原版(QU毕业证)皇后大学毕业证成绩单
一比一原版(QU毕业证)皇后大学毕业证成绩单一比一原版(QU毕业证)皇后大学毕业证成绩单
一比一原版(QU毕业证)皇后大学毕业证成绩单
 
哪里卖(usq毕业证书)南昆士兰大学毕业证研究生文凭证书托福证书原版一模一样
哪里卖(usq毕业证书)南昆士兰大学毕业证研究生文凭证书托福证书原版一模一样哪里卖(usq毕业证书)南昆士兰大学毕业证研究生文凭证书托福证书原版一模一样
哪里卖(usq毕业证书)南昆士兰大学毕业证研究生文凭证书托福证书原版一模一样
 
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
 
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
 
1.Seydhcuxhxyxhccuuxuxyxyxmisolids 2019.pptx
1.Seydhcuxhxyxhccuuxuxyxyxmisolids 2019.pptx1.Seydhcuxhxyxhccuuxuxyxyxmisolids 2019.pptx
1.Seydhcuxhxyxhccuuxuxyxyxmisolids 2019.pptx
 
FP Growth Algorithm and its Applications
FP Growth Algorithm and its ApplicationsFP Growth Algorithm and its Applications
FP Growth Algorithm and its Applications
 
一比一原版(UMich毕业证)密歇根大学|安娜堡分校毕业证成绩单
一比一原版(UMich毕业证)密歇根大学|安娜堡分校毕业证成绩单一比一原版(UMich毕业证)密歇根大学|安娜堡分校毕业证成绩单
一比一原版(UMich毕业证)密歇根大学|安娜堡分校毕业证成绩单
 
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
 
Investigate & Recover / StarCompliance.io / Crypto_Crimes
Investigate & Recover / StarCompliance.io / Crypto_CrimesInvestigate & Recover / StarCompliance.io / Crypto_Crimes
Investigate & Recover / StarCompliance.io / Crypto_Crimes
 
做(mqu毕业证书)麦考瑞大学毕业证硕士文凭证书学费发票原版一模一样
做(mqu毕业证书)麦考瑞大学毕业证硕士文凭证书学费发票原版一模一样做(mqu毕业证书)麦考瑞大学毕业证硕士文凭证书学费发票原版一模一样
做(mqu毕业证书)麦考瑞大学毕业证硕士文凭证书学费发票原版一模一样
 
Malana- Gimlet Market Analysis (Portfolio 2)
Malana- Gimlet Market Analysis (Portfolio 2)Malana- Gimlet Market Analysis (Portfolio 2)
Malana- Gimlet Market Analysis (Portfolio 2)
 
Predicting Product Ad Campaign Performance: A Data Analysis Project Presentation
Predicting Product Ad Campaign Performance: A Data Analysis Project PresentationPredicting Product Ad Campaign Performance: A Data Analysis Project Presentation
Predicting Product Ad Campaign Performance: A Data Analysis Project Presentation
 
Best best suvichar in gujarati english meaning of this sentence as Silk road ...
Best best suvichar in gujarati english meaning of this sentence as Silk road ...Best best suvichar in gujarati english meaning of this sentence as Silk road ...
Best best suvichar in gujarati english meaning of this sentence as Silk road ...
 
Empowering Data Analytics Ecosystem.pptx
Empowering Data Analytics Ecosystem.pptxEmpowering Data Analytics Ecosystem.pptx
Empowering Data Analytics Ecosystem.pptx
 
Jpolillo Amazon PPC - Bid Optimization Sample
Jpolillo Amazon PPC - Bid Optimization SampleJpolillo Amazon PPC - Bid Optimization Sample
Jpolillo Amazon PPC - Bid Optimization Sample
 
Tabula.io Cheatsheet: automate your data workflows
Tabula.io Cheatsheet: automate your data workflowsTabula.io Cheatsheet: automate your data workflows
Tabula.io Cheatsheet: automate your data workflows
 
一比一原版(YU毕业证)约克大学毕业证成绩单
一比一原版(YU毕业证)约克大学毕业证成绩单一比一原版(YU毕业证)约克大学毕业证成绩单
一比一原版(YU毕业证)约克大学毕业证成绩单
 
Ch03-Managing the Object-Oriented Information Systems Project a.pdf
Ch03-Managing the Object-Oriented Information Systems Project a.pdfCh03-Managing the Object-Oriented Information Systems Project a.pdf
Ch03-Managing the Object-Oriented Information Systems Project a.pdf
 

First Year Report, PhD presentation

  • 1. Deploying Machine Learning under Uncertainty in Cyber-Physical Manufacturing System (CPMS) Bang Xiang Yong Supervisor: Dr. Alexandra Brintrup Advisor: Prof Duncan McFarlane
  • 2. • In the recent Made Smarter Review [1], the total value of AI/ML technologies to the United Kingdom industry was estimated to be £198.7bn between 2017-2027. • Classical statistic methods such as linear regression and multivariate regression were designed for data with few dozen input variables and sample sizes which would be considered small by today’s standard. By contrast, many ML algorithms are capable of analyzing high dimensional data [6]. • Classical statistics and ML are not mutually exclusive : classical statistics can be used to pre-process the data before feeding into ML model such as Artificial Neural Network (ANN). • As manufacturers begin to capture large volumes of data describing their process through Cyber Physical Manufacturing Systems (CPMS), there is a need to derive value from the sensor data streams which are complex, high dimensional and non-linear. Introduction
  • 3. • Manufacturing systems are characterised by heterogeneous distributed systems, dynamic and unpredictable environments. • A missing ingredient from existing approaches is the measurement of uncertainty of ML – current approaches focus on • maximising the accuracy of predictions • executing them in the fastest way via scalable algorithms • A first motivating question would be, "Can we still trust the predictions of ML, given that faults and environment changes have occurred?". • Without reasoning on the uncertainty of data-driven systems in dynamic environments, manufacturers are exposed to risk in events – sensor failures, overconfident and erroneous predictions Problem Statement
  • 4. • Definition of Machine learning • Data-driven approaches which build mathematical models based on sample data to make predictions or decisions. [21] • In my definition : an automated process of building a mathematical model using example data, characterised by training and inference phases which relies on minimal prior knowledge of the underlying relationships between variables in data. • Methodologies such as CRISP-DM [24] have been developed to provide industrial standard frameworks for encapsulating the process of data analytics, which have been adapted to various domains [25]. Research Background
  • 5.
  • 6. • Definition of Cyber-Physical Manufacturing Systems: • Rajkumar et al. [30] defined CPS as systems which bridge the cyber-world of computing and communications with the physical world. • Sanislav and Miclea [32] emphasised CPS through the input and feedback to and fro the cyber-physical environments, management of distributed control, real-time executions, large spatial distribution and large scale control systems within systems. • My definition: Distributed and highly interconnected computer systems in manufacturing which are capable of obtaining real-time data from the physical world through sensors and acting upon them. Research Background
  • 8. Systematic literature review • Systematic literature reviews are characterised by explicit and reproducible steps with clear guidelines to reproduce. • The methods of systematic review advocated by Pickering and Byrne [39] were adopted. • Modern manufacturing has focused less on studying ML methods compared to other technologies as shown in a review by Liao et al. [41].
  • 9. Systematic literature review RQ1: What are the types of learning algorithms and technologies used in ML-CPMS? RQ2: What are the types and properties of data in ML-CPMS? RQ3: What are the characteristics and communication protocols of ML-CPMS framework? RQ4: What are the application areas and use cases of ML-CPMS? RQ5: What are the challenges and issues in ML-CPMS? "What is the state of machine learning research in Cyber-Physical Manufacturing Systems?" Main review question Sub review questions
  • 10. Systematic literature review Keyword search at Scopus and Web of Science: ("cyber physical system*" OR "industrial internet") AND ("machine learning“ OR "knowledge discover*" OR "data mine*" OR "artificial intelligence") AND "manufacturing"
  • 11. Customised web application for systematic literature review
  • 12. RQ1: What are the types of learning algorithms and technologies used in ML-CPMS?
  • 13. RQ1: What are the types of learning algorithms and technologies used in ML-CPMS?
  • 14. RQ2: What are the types and properties of data in ML-CPMS?
  • 15. RQ3: What are the characteristics and communication protocols of ML-CPMS framework?
  • 16. RQ4: What are the application areas and use cases of ML-CPMS?
  • 17. RQ5: What are the challenges and issues in ML-CPMS? There is not a uniform definition for uncertainty in ML-CPMS: • Uncertainty of raw measurements [82] • Economics and systems perspective [109] • Decision-making [110] • Events [111] • Part quality [58] • Model uncertainty [70, 112].
  • 18. RQ5: What are the challenges and issues in ML-CPMS?
  • 19. Uncertainty of machine learning in CPMS Relevant scenarios for measuring uncertainty of ML were inferred: • Overconfident predictions • Concept drift • Sensor faults and degradation Uncertainty of ML - The level of confidence of the output given by a model.
  • 20. Research Gaps 1. Lack of study addressing uncertainty holistically 2. Methods to quantify uncertainty of predictions. 3. Isolated study of scenarios relevant to uncertainty. 4. Acting upon quantified uncertainty.
  • 21. Towards developing a framework for deploying machine learning in Cyber- Physical Manufacturing Systems which quantifies uncertainty of prediction and acts on it. Research aim
  • 22. Research Questions and Tasks 1. What are the characteristics of ML-CPMS and the uncertainty of ML-CPMS? • Task 1.1 : Carry out systematic literature review to understand the status of MLCPMS research by enumerating the types of learning algorithms, technologies, challenges and issues which contribute to uncertainty. • Task 1.2 : Study extant literature to identify the sources of ML uncertainty, relevant scenarios and existing methods of dealing with these scenarios. • Task 1.3 : Define a conceptual model of uncertainty in ML-CPMS based on Task 1.2. 2. How can a framework of ML-CPMS manage the uncertainties? • Task 2.1 : Define components of framework and algorithms for uncertainty quantification of ML predictions, sensors fault detection and concept drift detection. • Task 2.2 : Identify ML algorithms which are capable of expressing uncertainty in their predictions. • Task 2.3 : Identify methods and algorithms which are able to reduce the uncertainty of prediction. • Task 2.4 : Combine outputs of Task 2.1, Task 2.2, Task 2.3 into a framework with the ability to quantify uncertainty, detect the sources and subsequently reduce it.
  • 23. Research Questions and Tasks 3. How effective is the proposed framework in handling uncertainty? • Task 3.1 : Identify quantitative or qualitative performance metrics of proposed framework. • Task 3.2 : Develop a testbed to implement algorithms and ML-CPMS framework on real industrial data-sets. • Task 3.3 : Define and conduct laboratory experiments to evaluate the performance of proposed framework. • Task 3.4 : Deploy proposed framework in industrial case studies. • Task 3.5 : Evaluate the outcomes of the deployment and compare with the laboratory studies.
  • 24. Research Scope Research Methodology
  • 25. Relevant journals 1. International Journal of Production Research 2. Future Generation Computing Systems 3. Sensors 4. Computers in Industry 5. Journal of Manufacturing and Materials Processing 6. Journal of Machine Learning Research. Dissemination • Webinar was recorded to disseminate the developed agent-based system for handling uncertainty, agentMET4FOF software package and is publicly available for viewing • SOHOMA19 paper titled "Multi Agent System for Machine Learning Under Uncertainty in Cyber Physical Manufacturing System" Current publication
  • 29. Exploratory Studies : 1. Agent-based system “How can a software architecture for machine learning be implemented which is modular and flexible to investigate various use cases, configurations, scenarios, algorithms in the context of Cyber-Physical Manufacturing System?”
  • 30.
  • 32. Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML
  • 33. Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML Classification of Hydraulic System condition using Bayesian Neural Network
  • 34. Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML Classification of Hydraulic System condition using Bayesian Neural Network
  • 35. Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML Regression for prognosis of electromechanical cylinder using Bayesian Neural Network
  • 36. Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML Regression for quality prediction of radial forge, AFRC
  • 37. Exploratory Studies : 2. Relationship between uncertainty thresholds and accuracy of ML Regression for quality prediction of radial forge, AFRC
  • 38. Exploratory study: 3. Deployment of agent-based system on embedded systems
  • 39. Future works 1. Further develop, implement and analyse proposed architecture based on outline performance criteria 2. Simulate behaviour of multi-agent system under conditions of uncertain scenarios 3. Analyse relationship between predictive uncertainty and uncertainty and how agents can make decisions to manage uncertainty of ML models Contributions • Theoretical contribution - theoretical framework of ML-CPMS which quantifies, detects and acts on uncertainty of prediction. • Practical contribution - software package based on agent-based system which readily implements the methods for quantifying, detecting and acts on uncertainty in ML-CPMS