Just as it is important for us to monitor our own body in order to maintain good health, we should have an accurate understanding of the state of engineering systems in order to prevent failures. In this project, entitled "Interpretable Condition Monitoring for Complex Engineering Systems", we investigate machine learning methods to build a condition monitoring framework and apply it to real engineering systems such as artificial satellites and unmanned aerial vehicles.
An important research question of the project is how to make the condition monitoring methods interpretable for human operators. The lack of interpretability of machine learning models often hinders the effective integration of learning-based methods into real engineering systems. This is particularly evident in condition monitoring applications, where human operators should manually investigate alarms generated by condition monitoring for reasons such as system safety. We address this issue through the idea of grey-box (hybrid) modeling, where data-driven, machine learning models and theory-driven, expert models are combined. Expert models of engineering systems are interpretable by design but often cannot accurately predict the state of the system in real-world situations. Machine learning models can adapt to real-world situations based on data but lack interpretability. We take the best of the two regimes to build data-driven condition monitoring methods that are reasonably interpretable.
The project is funded by the Swiss National Science Foundation under the Strategic Japanese-Swiss Science and Technology Program. It is an international collaborative project with a counterpart in Japan; we are working with the Artificial Intelligence Lab of the Research Center for Advanced Science and Technology, the University of Tokyo.
Relevant publication from the DMML group:
Naoya Takeishi, Alexandros Kalousis: Physics-Integrated Variational Autoencoders for Robust and Interpretable Generative Modeling. Advances in Neural Information Processing Systems 34, 2021.
Naoya Takeishi, Alexandros Kalousis: Deep Grey-Box Modeling With Adaptive Data-Driven Models Toward Trustworthy Estimation of Theory-Driven Models. arXiv:2210.13103, to appear in AISTATS 2023.
Read more
A vast amount of Earth Observation (EO) data is produced daily and made available through online services and repositories. Contemporary and historical data can be retrieved and used to power existing applications, to foster innovation and finally improve the EU citizens’ lives. However, an undersized audience follows this activity, leaving huge volumes of valuable information unexploited. EO4EU aims to provide innovative tools, methodologies and approaches that would assist a wide spectrum of users, from domain experts and professionals to simple citizens to benefit from EO data. EO4EU strives to deliver dynamic data mapping and labelling based on AI adding FAIRness to the system and data. EO4EU introduces an ecosystem for holistic management of EO data, bridging the gap among domain experts and end users, and bringing in the foreground technological advances to address the market straightness towards a wider usage of EO data. EO4EU envisages to boost the Earth Observation data market, providing a digestible data information modeling for a wide range of EO data, through dynamic data annotation and a state-of-the-art serverless processing by leveraging important European Cloud & HPC infrastructures.
The role of our group within EO4EU is linked with the Machine Learning related tasks of the project, which can be summarized in two main directions:
The study and application of Self-Supervised learning models which will help us exploit the vast volume of unlabelled EO data, in order to minimize the annotation effort required in downstream supervised tasks. In this way, new use cases will be enabled to efficiently train classifiers, with a significantly reduced required budget of annotated/ labelled data. This would facilitate the greater public and non-experts interested institutions and individual to access and efficiently use these data. To achieve this, we explore the latest advancements in Self-Supervised Contrastive Learning.
The development of tailored models of Learned Compression, suitable to the EO ecosystem. Based on previous expertise in the development of models of Learned Compression, we will develop the models that will increase the compression efficiency comparing to standard, ... Read more
LiBertaS aims to liberate the access of Location Based Services (LBS) to the highly accurate fingerprinting methods in business practice, by greatly downscaling the data volume needed for their operation. Location Based Services (LBS) have recently undergone a tremendous increase in popularity. Despite the proven superiority of the fingerprinting methods in terms of localization accuracy, their requirement for an extensive data collection and the related costs, proves to be the main barrier preventing their market adoption.
LiBertaS plans to tackle this issue following a multifaceted approach. LiBertaS will operate as a synergistic power that will study the optimal combination of previous findings, linking pieces in a puzzle, to construct a first-of-its-kind comprehensive framework. In LiBertaS, we will study, evolve and combine 3 main research directions in a synergistic manner. We will explore the individual and combined capabilities of three innovative concepts: (i) estimate selection based on the Dynamic Accuracy Estimation, (ii) Proximity-based fingerprint augmentation, and (iii) Generative Modeling. The work plan of LiBertaS aims to bring the relevant research line of our work full circle.
LiBertaS will combine our previous contributions, which were mainly produced within the Eratosthenes project. More particularly, LiBertaS builds on top of these works:
G. G. Anagnostopoulos and A. Kalousis, “ProxyFAUG: Proximity-based Fingerprint Augmentation”, in 2021 International Conference on Indoor Positioning and Indoor Navigation (IPIN), Nov. 2021, pp. 1–7. doi: 10.1109/IPIN51156.2021.9662590.
G. G. Anagnostopoulos and A. Kalousis, “Analysing the Data-Driven Approach of Dynamically Estimating Positioning Accuracy”, in ICC 2021 - IEEE International Conference on Communications, Jun. 2021, pp. 1–7. doi: 10.1109/ICC42927.2021.9500369.
G. G. Anagnostopoulos and A. Kalousis, “Can I Trust This Location Estimate? Reproducibly Benchmarking the Methods of Dynamic Accuracy Estimation of Localization”, Sensors , vol. 22, no. 3, p. 1088, Jan. 2022, doi: 10.3390/s22031088.
Dr. Grigorios Anagnostopoulos, research associate of the dmml group, is the grantee of the LiBertaS project. LiBertaS is funded under the Small Projects funding scheme of the Halser Foundation, who is gratefully acknowledged by the grantee.
Related previous projects:
Eratosthenes , funded under the Spark SNF funding scheme
Orbiloc, an ... Read more
Eratosthenes is a research project funded under the Spark funding scheme of the Swiss National Science Foundation (SNSF). The aim of the Spark is to fund postdoctoral researchers to implement "projects that show unconventional thinking and introduce a unique approach". The relevant criteria for the award of Spark grants are the originality/novelty of the idea, the unconventionality of the proposed research project, the scientific quality of the project and the potential for significant impact. Dr. Grigorios Anagnostopoulos, postdoctoral researcher of the dmml group, is the grantee of the Eratosthenes project.
The ambition of the Eratosthenes project is to integrate the recent advancements of the field of generative modelling within the application domains of indoor and outdoor positioning. More specifically, the aim is to mitigate the barrier of the data collection requirement of fingerprinting localization methods through generative modelling.
SNSF webpage of the project: http://p3.snf.ch/project-195964
Related past project: http://dmml.ch/orbiloc/
Related Blog post: http://dmml.ch/outdoor-positioning-for-the-iot-world-2/
Abstract
The proliferation of Location Based Services through smart devices has supported a variety of services. Unlike outdoor positioning, where systems such as the GPS or Galileo are considered a de facto standard, indoor positioning has presented systems utilizing a big variety of technologies and positioning methods. The main categorical division of these methods is set between ranging and fingerprinting methods. Moreover, the recent emergence of the usage of Internet-of-Things (IoT) technologies and their usage in creating Low Power Wide Area Networks (LPWAN) has shaped a new landscape in the field of outdoor localization. The low-power devices used in LPWANs cannot afford the battery consumption of a Global Navigation Satellite System (GNSS) chip set, such as the GPS. Therefore, an alternative approach is needed in order to localize these low power devices. Similarly, to the indoor positioning, ranging and fingerprinting methods are two main competing approaches for LPWAN positioning as well. Fingerprinting methods generally offer a higher accuracy of localization but suffer from the disadvantage of requiring a tedious and costly data collection phase. Since fingerprinting is a purely data-driven approach, the way data are collected, the volume of the collected data and the way they ... Read more
IAI is an Innosuisse project, funded under the joint call in which partners from South Korea and Switzerland are invited to collaborate. At the Swiss side of the consortium, our team collaborates with ABB, a pioneering technology leader with a vivid interest and a profound knowledge in predictive maintenance. The South Korean side of the consortium is composed by three partners: The SHRM Lab from the Seoul National University contributes its considerable experience in Prognostic and Health Management, the SMLD team of the Dongguk University contributes with its physical modelling and data cleansing experience and, lastly, OnePredict brings its expertise in predictive maintenance.
The predictive maintenance field has not yet benefited from the recent advances in deep learning. This is mainly because deep learning is rather data hungry, typically needing tens of thousands of training examples. Real-life maintenance data are composed of large quantities of samples referring to normal operating conditions with only very few samples of abnormal or faulty behaviour. In addition, the real-life samples are specific to the particular operating conditions under which they were collected. A change in the operating environment can have an important effect on the behaviour of the equipment and therefore the monitoring data.
In IAI, we build on the recent advances in deep learning and develop methods which are tailored to the particularities of predictive maintenance. To do so, we study regularization techniques to bring in domain knowledge into the modelling and data augmentation relying either on physical models and simulators or on machine learning conditional generative models to complement the scarce real-life data. The tools that will be developed in this project aim to push the state of the art in the field, by offering significant improvements to the predictive performance, resulting in significant savings in maintenance and operational costs.
Read more