Early fault classification in dynamic systems using Case-Based Reasoning

In this paper we introduce a system for early classification of several fault modes in a continuous process. Early fault classification is basic in supervision and diagnosis systems, since a fault could arise at any time, and the system must identify the fault as soon as possible. We present a computational framework to deal with the problem of early fault classification using Case-Based Reasoning. This work illustrates different techniques for case retrieval and reuse that have been applied at different times of fault evolution. The technique has been tested for a set of fourteen fault classes simulated in a laboratory plant.