S4: Diagnostic reasoning

Model based reasoning to support diagnostics in complex systems

As systems become increasingly complex, diagnosing system failures and performance issues becomes a true challenge for engineers. Too often, solving problems requires extensive involvement of multiple R&D experts. Diagnostic reasoning gives engineers the tools they need to make decisions about system behaviour and performance without having to know all the intrinsic system details. In addition to presenting a state-of-the-art overview and a look into the future, this session presents industrial cases in which data, modelling, and reasoning are brought together to solve diagnostic challenges.

Moderation: Emile van Gerwen, ESI (TNO)

Chair: Hans Onvlee, ASML

In his introduction Hans Onvlee will present a state-of-the-art overview on diagnostic reasoning and give a look into the future.

Jeroen Voeten, ESI (TNO)

Diagnosing timing bottlenecks in large-scale component-based software

We  introduce  a  new  measurement-based approach  to  diagnose  timing  bottlenecks  of  existing  large-scale  component-based  software.  The approach is based on Timed  Message  Sequence  Charts (TMSCs) that

  • capture the  run-time execution of component-based software  systems  in a concise and intuitive way

  • are amenable to formal timing analysis, and

  • are automatically inferred from execution traces. We demonstrate the  effectiveness  of  our  approach  by automatically  computing  critical  paths  in  the  software  of  an ASML lithography  scanner  to  identify  timing  bottlenecks.

Ton van Velzen, IBM

AI for Manufacturing

Artificial Intelligence applications in manufacturing range from improving reliability of equipment to improving operational effectiveness and providing quality insights. Using a simple framework this talk will discuss some practical examples of machine learning to predict failures and optimize equipment maintenance and process effectiveness. It will be shown that for a large class of algorithms explainability and transparency of machine learning based recommendations is an essential part of the solution. A discussion of a knowledge representation application will show how diagnostic reasoning can be assisted by artificial intelligence.

download pdf

go to S5 Acceptable AI