Learning from Data Streams
18 September 2017
IoT Streams forPredictive Maintenance
Third Workshop+ Tutorial
23 September 2022 (Online Event)
ECML/PKDD 2022 Workshop on IoT Streams for Predictive Maintenance
Maintenance is a critical issue in the industrial context for the prevention of high costs or injuries. Various industries are moving more and more toward digitalization and collecting “big data” to enable or improve the accuracy of their prediction. At the same time, the emerging technologies of Industry 4.0 empowered data production and exchange which lead to new concepts and methodologies exploitation of large datasets for maintenance. The intensive research effort in data-driven Predictive Maintenance (PdM) has been producing encouraging outcomes. Therefore, the main objective of this workshop is to raise awareness of research trends and promote interdisciplinary discussion in this field.
Data-driven predictive maintenance deals with big streaming data that include concept drift due to both changing external conditions, but also normal wear of the equipment. It requires combining multiple data sources, and the resulting datasets are often highly imbalanced. The knowledge about the systems is detailed but in many scenarios, there is a large diversity in both model configurations, as well as their usage, additionally complicated by low data quality and high uncertainty in the labels. In particular, many recent advancements in supervised and unsupervised machine learning, representation learning, anomaly detection, visual analytics and similar areas can be showcased in this domain. Therefore the overlap in research between machine learning and predictive maintenance continues to increase in recent years.
Maintenance is a crucial topic for industrial machines, medical equipment, energy systems, passengers transport vehicles and home appliances among others. Cost reduction, machine reliability, operation, safety and time reduction have been the main concerns of companies and organizations. Meanwhile, Industry 4.0 brought new opportunities of meaningful data collection and storage. Promising data-driven methodologies shrive for predictive maintenance becoming a strong alternative.
Motivation and focus
This workshop will be centred on questions such as when to perform a maintenance action? How to estimate components current and future status? How accurate are the existing methods? Which data should be used? What decisions tools should be developed for prognostic. How can Data Mining and Machine Learning (Artificial Intelligence in general) contribute to answering these questions?
Therefore, this event is an opportunity to bridge researchers and engineers to discuss the emerging topics and the key trends.
Aim and scope
Topics of interest for the workshop include, but are not limited to:
● Predictive and prescriptive maintenance
● Explainable AI for predictive maintenance
● Fault detection and diagnosis
● Fault isolation and identification
● Estimation of Remaining Useful Life of components, machines, etc.
● Forecasting of product and process quality
● Anomaly detection
● Early failure detection and analysis
● Automatic process optimisation
● Self-healing and self-correction
● Incremental, evolving (data-driven and hybrid) models for FDD and anomaly detection
● Self-adaptive time-series based models for prognostics and forecasting
● Adaptive signal processing techniques for FDD and forecasting
● Concept drift issues in dynamic predictive maintenance systems
● Active learning and Design of Experiment (DoE) in dynamic predictive maintenance
● Systems fault-tolerant control
● Industrial process monitoring and modelling
● Maintenance scheduling and on-demand maintenance planning
● Visual analytics and interactive machine learning
● Decision-making assistance and resource optimisation
● Planning under uncertainty
● Analysis of usage patterns
Real-world applications such as:
● Manufacturing systems
● Production processes and factories of the future
● Wind turbines (offshore/onshore/floating)
● Smart management of energy demand/response
● Energy and power systems and networks
● Transport systems
● Power generation and distribution systems
● Intrusion detection and cyber security
● Internet of Things,
● Next-Generation aerospace applications
● Big Data challenges in energy transition and digital transition
● Solar plant monitoring and management
● Healthcare equipment
● Distributed renewable energy management and integration into smart grids
● Smart cities
Submission and Review process
Regular and short papers presenting work completed or in progress are invited. Regular papers should not exceed 15 pages, while short papers are maximum 6 pages. Papers must be written in English and are to be submitted in PDF format online via the Easychair submission interface:
Each submission will be evaluated on the basis of relevance, significance of contribution, quality of presentation and technical quality by at least two members of the program committee. All accepted papers will be included in the workshop proceedings and will be publically available on the conference web site. At least one author of each accepted paper is required to attend the workshop to present.
Paper submission deadline: 1th of July 2022 (extended)
Paper acceptance notification: 30th of July 2022 (new date)
Paper camera-ready deadline: 10th of August 2022 (new date)
Program Committee members (to be confirmed)
Carlos Ferreira, LIAAD INESC Porto LA, ISEP, Portugal
Edwin Lughofer, Johannes Kepler University of Linz, Austria
Sylvie Charbonnier, Université Joseph Fourier-Grenoble, France
David Camacho Fernandez, Universidad Politecnica de Madrid, Spain
Bruno Sielly Jales Costa, IFRN, Natal, Brazil
Fernando Gomide, University of Campinas, Brazil
José A. Iglesias, Universidad Carlos III de Madrid, Spain
Anthony Fleury, Mines-Douai, Institut Mines-Télécom, France
Teng Teck Hou, Nanyang Technological University, Singapore
Plamen Angelov, Lancaster University, UK
Igor Skrjanc, University of Ljubljana, Slovenia
Indre Zliobaite, University of Helsinki, Finland
Elaine Faria, Univ. Uberlandia, Brazil
Mykola Pechenizkiy, TU Eindonvhen, Netherlands
Raquel Sebastião, Univ. Aveiro, Portugal
Anders Holst, RISE SICS, Sweden
Erik Frisk, Linköping University, Sweden
Enrique Alba, University of Málaga, Spain
Thorsteinn Rögnvaldsson, Halmstad University, Sweden
Andreas Theissler, University of Applied Sciences Aalen, Germany
Vivek Agarwal, Idaho National Laboratory, Idaho
Manuel Roveri, Politecnico di Milano, Italy
Yang Hu, Politecnico di Milano, italy
Szymon Bobek, Jagiellonian University, Cracow, Poland
Martin Atzmüller, Osnabrück University, Germany
Joao Gama, University of Porto, Porto, Portugal
Albert Bifet, Telecom-Paris, Paris, France, University of Waikato, New Zealand
Sławomir Nowaczyk, Halmstad University, Sweden
Carlos Ferreira, University of Porto, Porto, Portugal
Publicity and Social Media Chair
Tutorial: IoT Data Stream Mining in Practice
The challenge of deriving insights from the Internet of Things (IoT) has
been recognized as one of the most exciting and key opportunities for both academia
and industry. The advent of IoT applications is here: industry 4.0, connected indus-
try, industry automation, smart cities, smart grids, energy efficiency, etc. All this IoT
applications require advanced analysis of big data streams from sensors and small
devices, while addressing security and privacy concerns. This tutorial is a gentle
introduction to mining IoT big data streams. The first part introduces data stream
learners for several learning tasks including distributed algorithms. The second and third part
present some applications for predictive maintenance, prediction for renewable ener-
gies, and social network analysis for telecommunications data streams. The last part presents how to use Apache Spark Streaming for applying scalable machine learning on Big Data streams.
1.IoT Fundamentals and IoT Stream Mining Algorithms
– Predictive Learning
– Descriptive Learning
– Frequent Pattern mining
2. Case Study: Predictive Maintenance
– Problem Definition
– Change, Anomaly and Novelty Detection
– Failure Prediction and Detection
3. Case Study: Social Network Analysis
– Challenges in mining networked data,
– Online sampling
– Evolving centralities and communities
– Tracking the dynamics of evolving communities
4. Big Data Stream Mining using Spark Streaming
– Fundamental concepts
Heitor Murilo Gomes
with a focus on Explainable Predictive Maintenance (XPM)
Today, in many industries, various AI systems, often black-box ones, predict failures based on analysing sensor data. They discover symptoms of imminent issues by capturing anomalies and deviations from typical behaviour, often with impressive accuracy. However, PM is part of a broader context. The goal is to identify the most probable causes and act to solve the problem before it escalates. In complex systems knowing that something is wrong (i.e., to detect an anomaly) is not enough; the key is to understand its reasons and potential consequences, and to provide alternatives (solutions/advice) in order to mitigate those consequences. However, this often requires complex interactions among several actors in the industrial and decision-making processes. Doing it fully automatically is unrealistic, despite recent impressive progress in AI. For example, in wind power plants a repair requires synchronisation of inventory (e.g., availability of pieces to be replaced), logistics (e.g., finding a ship to reach the installation), personnel management (availability schedule), weather predictions (i.e., suitability of conditions to perform maintenance actions, in particular for high towers in offshore farms), and more. All these cannot be done by AI and require human expertise since the complete relevant context cannot be formalised in sufficient detail to allow automatic reasoning.
Recent developments in the field based on different Machine Learning and Artificial Intelligence methods are promising for fully- and semi-automated data-driven pattern recognition and knowledge creation enabled by IoT streams. Explanations of the models are necessary to create trust in prediction results for complex systems and non-stationary environments.
This tutorial aims to present current trends and promising research directions within the field of Machine Learning for Predictive Maintenance. In addition, we will present some of the state-of-the-art methodologies in Explainable AI (XAI) relevant for predictive maintenance problems. Further, we will provide hands-on examples of applying XAI in benchmark datasets. This year the focus will be on explainable predictive maintenance (XPM). We will also present a discussion about future challenges and open issues on this topic. Two case studies related to Predictive Maintenance challenges in the Metro operation and Steel factory will be presented, hands-on, during the tutorial.
Explainable Predictive Maintenance (XPM) focuses on creating methods that can explain the operation of AI systems within the PM domain. What makes the creation of the maintenance plans challenging is incorporating the PM AI output into the human decision-making process, and integrating it with human expertise. To make AI useful and trustworthy means putting fault predictions in a relevant context, and making them understandable for the humans involved. It requires explanations that are adapted to the roles and needs of different actors – for example, to lower-level engineers through connection to the blueprints of the technical installation; while
using other means for the managers who evaluate the costs of system downtime; or company lawyers assessing the possible liability in the case of safety-threatening failure.
The underlying theme of the XAI is building up the trust towards AI systems, but in order to maximise the real usefulness of PM, we need to go one step further. In particular, the primary novelty of the XPM is developing technological solutions that explicitly address four concrete reasons why the explanations are needed. Within these four rationales, the explanations of AI decisions will lead to the most significant improvement in the repair and maintenance actions taken by human experts. The first rationale is identifying (isolating and characterising) the fault. In complex industrial systems, AI is often able to identify deviations from normal behaviour, but due to a huge number of components with complex interactions, pinpointing the actual spot requires experts and their domain knowledge. The second is understanding fault consequences. The right maintenance depends on how the issue will evolve, what is Remaining Useful Life (RUL), and what can be the collateral damage or loss of productivity. These questions require considering the broader context, beyond the scope of the AI system. The third rationale, is supporting domain experts by helping human operators create the right repair plan, including optimising the system performances (e.g. uptime or safety) in the presence of the degradation. It also needs to strike the right trade-off between different quality criteria. Finally, the fourth is understanding the reasons why the fault has occurred and how to improve the system in the future. It can be related to incorrect usage, suboptimal design, as well as the monitoring process itself. For example, what optimisation of sensor types and placement will allow for earlier detection in the future, or what changes to the manufacturing process parameters will extend the lifetime of certain critical components.
Data-driven solution of predictive Maintenance supplemented with explainable tools is a very interesting application for the Machine Learning community. Especially since many XAI solutions are focusing on understanding ML models rather than communicating with the end-users. This tutorial will be interesting both for practitioners working in the area of Predictive Maintenance (who will get a chance to better understand new developments in Machine Learning as potential solutions) as well as for researchers developing new Machine Learning algorithms (presenting to them somewhat unique challenges present in the Predictive Maintenance setting and identify new promising directions of research and improvement).
Sepideh Pashami, Halmstad University, Sweden
Grzegorz J. Nalepa, Jagiellonian University, Krakow, Poland
Rita Paula Almeida Ribeiro, University of Porto, Porto, Portugal
Bruno Veloso, University of Porto, Porto, Portugal
Szymon Bobek, Jagiellonian University, Poland
Foundations of Trustworthy Machine Learning on Data Streams
Trustworthy AI is broadly used as an umbrella term encompassing multiple aspects of machine learning-based solutions, notably including robustness, reliability, safety, security, scalability, and interpretability. Some of these aspects are of higher importance for adopting mostly autonomous solutions, others - for applications involving human-in-the-loop in decision making. In this talk, I will use examples of both kinds of applications to illustrate a variety of research challenges and related trade-offs being addressed. I will focus then on the peculiarities of trustworthy AI in machine learning over evolving data streams.
Mykola Pechenizkiy is full professor, chair of Data Mining at the Department of Mathematics and Computer Science, TU Eindhoven. His main expertise and research interests are in predictive analytics on data evolving over time. He leads Trustworthy AI interdisciplinary research studying foundations of robustness, safety, trust, reliability, scalability, interpretability and explainability of AI and their applications in industry, healthcare and education. He works in close collaboration with industry on developing novel techniques for informed, accountable and transparent predictive analytics. He serves on the editorial boards and on the program committees of the leading data mining, machine learning and AI conferences and journals.
10:30 Tutorial Part 1
• Data-driven Predictive Maintenance in Industry 4.0: concepts and challenges (70 minutes)
(break 5 min)
• Overview of XAI methods and their connection to Predictive Maintenance (40 minutes)
(break 5 min)
12:30 Tutorial Part 2
• Use case 1 description: Steel factory (30 minutes
• Use case 2 description: Streaming data from Porto Metro (30 minutes)
14.30 Keynote: Mykola Pechenizkiy
15.30 Online Anomaly Explanation: A Case Study on Predictive Maintenance
Rita P. Ribeiro, Saulo Mastelini, Narjes Davari, Ehsan Aminian, Bruno Veloso and Joao Gama.
15.50 Fault forecasting using data-driven system modeling: a case study for Metro do Porto data set
Narjes Davari, Bruno Veloso, Rita P. Ribeiro and Joao Gama.
16.10 An online data-driven predictive maintenance approach for railway switches
Emanuel Tomé, Rita Ribeiro, Bruno Veloso and João Gama.
17.00 curr2vib: Modality Embedding Translation for Broken-Rotor Bar Detection
Amirhossein Berenji, Zahra Taghiyarrenani and Slawomir Nowaczyk.
17.20 Incorporating Physics-based Models into Data-Driven Approaches for Air Leak Detection in City Buses
Yuantao Fan, Hamid Sarmadi and Slawomir Nowaczyk.
17.35 Towards Geometry-Preserving Domain Adaptation for Fault Identification
Zahra Taghiyarrenani, Slawomir Nowaczyk, Sepideh Pashami and Mohamed-Rafik Bouguelia.
17.50 A systematic approach for tracking the evolution of XAI as a field of research
Samaneh Jamshidi, Slawomir Nowaczyk, Hadi Fanaee-T and Mahmoud Rahat.
18:05 Frequent Generalized Subgraph Mining via Graph Edit Distances, Richard Palme, Pascal Welke