After four successful editions, the EluciDATA Lab of Sirris is organising a fifth edition of its seminar on fleet-based analytics. The seminar focuses on the opportunities and challenges of optimising the operation and maintenance of a fleet of machines by means of data analytics and AI. Every year, it provides the participants with an overview of the current status in this domain, via keynote presentations by lead companies and state-of-the-art research results illustrated with industrial use cases in different sectors.
While many companies that operate a fleet of machines continuously monitor its operation, only few fully exploit the enormous amount of collected data to its full potential. Usually this data is merely used for operational dashboarding or by maintenance and service personnel for reactive maintenance. Some lead user companies have recently shifted to intelligently exploit this data for proactive maintenance, but even in these cases the analysis is typically performed at individual machine level. This leaves a huge potential of untapped exploitation opportunities at fleet level.
By comparing the behaviour of a single machine with respect to the fleet in which it is operating, more detailed and more accurate knowledge can be extracted. This offers several new possibilities to improve the operations or optimise the maintenance, including:
- identifying malfunctioning or underperforming assets
- dealing with lost data connections and missing data
- benchmarking assets against one another
- identifying remaining useful life of assets
- modelling new assets that lack historical data
In each case, the fleet context provides additional insights that can help you to improve the operational efficiency of your assets.
This seminar will give you an update of the current state of the art in this domain, through keynote presentations by lead companies and research results illustrated by means of industrial use cases in different sectors.
13:30: Introduction and setting the context: the opportunities and challenges of fleet-based analytics
13:45: Industrial keynote AW Europe: GAN-based Anomaly Detection for PCBA images (Arnaud Bougaham)
Abstract: Industry 4.0 and recent deep learning progress make it possible to solve problems that traditional methods could not. This is the case for anomaly detection that received a particular attention from the machine learning community, and resulted in a use of generative adversarial networks (GANs). In this talk, an industrial use case is considered, using intermediate patches in WGANs for the inference step, to make the anomaly detection possible on full size Printed Circuit Board Assembly (PCBA) images. This technique can be used to support or replace actual industrial image processing algorithms and, to avoid a waste of time and manual labour for industries, when verifying false positives.
14:15: Using fleet-based AI for assisting asset configuration (Sirris)
Abstract: A particular challenge when connecting assets deployed in the field to a cloud-based data gathering platform is onboarding, i.e., configuring the data ingestion process for that asset and the correct configuration of its digital twin so that the data can be interpreted correctly. Onboarding can be a time-consuming and error-prone process due to a lack of standardization, e.g. different assets produced by various manufacturers, that use different communication protocols, different data formats, ad-hoc configurations of on-premise data logging systems, etc. In this presentation, we will present how this process can be supported using fleet-based AI techniques that recommend configurations automatically.
14:40: Industrial innovation pitch: Industrial equipment & cloud services: l'embarras du choix. (Jurgen Devlieghere, VP Technology Digital, Flint Group Digital)
Abstract: With cloud and IoT services overwhelming any observer, the first steps in developing cloud services for industrial equipment are daunting.
Hundreds of possible applications on the one side and multiple platforms with each hundreds of supporting services on the other side make it difficult to connect the dots. Nevertheless, some insights can help to make choices, both in what to do and in how to do it.
15:15: Advanced discretisation and visualisation techniques for performance profiling of fleets (Sirris)
Abstract: Many industrial assets as wind turbines are instrumented with monitoring systems generating a vast amount of time series data. Making sense of such data is not always trivial due to the high frequency and level of detail, which might obscure interesting long-term trends and patterns. Clever segmentation and subsequent discretisation, allows to convert the time series into a representation, which might be much more suitable for the application of advanced machine learning and visualization approaches. In this presentation, we illustrate how a novel circular binning approach allows to generate insightful visualisations. Next, prototypical fleet-wide models are extracted via non-negative matrix factorisation, enabling to detect anomalies and perform production forecasts.
15:40: Industrial innovation pitch: Training, deployment and usage of AI models across a fleet of similar assets: finding the right mix of no-code, low-code and coded methods (Jan Verhasselt, Managing Director, Yazzoom)
16:00: Context-sensitive benchmarking of industrial assets (Sirris)
Abstract: In many domains, an asset’s performance is benchmarked against theoretical specifications that are based on a limited number of operational contexts and asset setups . This makes it difficult to identify whether an asset is performing optimally given an actual operational context or to assess which type of asset would perform best in a specific context . When data on a fleet of assets is available, the behaviour of an asset can be benchmarked against the behaviour of other assets in order to assess the actual performance in real-life operational conditions. In this presentation, we will illustrate how this can be achieved, present the challenges that need to be overcome and discuss the solutions we have realized.
340 EUR (excl. BTW)
Companies located in Flanders can participate with a discount of 70% at the price of 100 EUR, thanks to the support of VLAIO through the Industry Partnership initiative.
Our general terms and conditions
Any cancellation has to be made by e-mail (email@example.com). Cancellations made before the 3 business days preceding the event are free of charge. After this deadline, 50% of the participation fee will be charged (incl. VAT). In case of cancellation the day itself, the full amount of the registration will be due. In case of 'no-show' the full amount of the registration fee will be due too. Replacement by a colleague is always possible if notified in advance by e-mail to firstname.lastname@example.org.