Safe Adaptive Control

This application is a request for funding the cost of establishing and strengthening the preliminary consortium, and means for strategic enlargement of the consortium involved in the SADCO project. There is established an intentional agreement with Jernb aneverk for collaboration on the "SADCO project, and it is planned to be supervised by UIO. IFE provides interest in the projects related to the research activities in the OECD Halden Reactor Project. The objective of the SADCO project is to develop an ap proach to allow adaptive subsystems in a critical process controlling environment. The gain of solution would be that critical systems could incorporate sophisticated subsystems utilising e.g. adaptive neural networks or genetic algorithms tailored to foc us on performance goals in control applications and still ensure plant safety. Critical programmable systems must be verified for safety in operation. This project will look into the possibility of introducing state-of-the-art adaptive control in critic al control environments. The problem with adaptive controllers in such environments is that its behaviour change over time as it adapts and learns about its environment. This feature entails that one cannot with absolute certainty predict the outcome of t he controller at a given time with a given input, in other words the algorithm that the controller represents is not a deterministic structure. Since determinism is a basic condition in order to verify safety it is obvious that the use of adaptive control represents a great challenge. This project will investigate solutions to bypass or greatly reduce the V&V need on the adaptive controller by making it not critical. The overall system will be fault tolerant to any failure in the adaptive subsystem. In sh ort, the project goal is to increase performance with adaptive control and sustain safety with fault tolerance