Expert system for classification of measured signal as a sequence of processes

Classification of the processes is frequently required task in technical diagnostics. If one has at one's disposal a graph of a data stream measured during the operation of a device, often only a human expert is able to recognize the processes that have occurred during the operation. The recognized sequence of processes must describe the measured data stream as exactly as possible but the device should also be able to perform these processes in sequence. The approach described in this paper uses two steps to handle the problem. In the first stage, technique called the Hidden Markov Models is used to scan the data stream for possible occurrences of the processes. A set of hypotheses is generated by this algorithm. During the second stage a rule-based expert system is used to verify these hypotheses, connect them together when it is possible, and verify the resulting sequences of processes. The system based on a blackboard architecture is used to control the second stage.