Symbolic Reasoning on Embedded Controllers for Industrial Diagnostics

Carlos Pachajoa, Felix Rempe, Marco Seravalli

The project team realised this honours project in close cooperation with the Siemens AG, Corporate Technology, Munich.

Symbolic reasoning presents great potentials for industrial production both in terms of flexibility and functionality. An abstract level of programming as well as efficient algorithms for handling queries and deducing information make implementations of logic formalisms highly interesting for the use in diagnosis systems in production sites. However, it is still open which types of logics are most feasible and which physical controllers are capable of running a reasoner. Those controllers need to be small, cheap and reliable as well as provide the required computational features to handle complex systems with many sensors in terms of response time, memory requirements and completeness of the result. This project gives first results about the feasibility of performing reasoning on embedded controllers for industrial diagnostics. For that purpose, promising reasoners and hardware have been selected and tests concerning memory requirements have been carried out.

memory increase

Illustration 1: Maximum use of memory as a function of the problem size for different multipliers increasing the number of compiled rules. Behaviour is linear with respect to the multiplier. The results were obtained on a Gumstix Overo Air.