An expert system for diagnosing operational transients in a nuclear power plant is discussed. Hypothesis and test is used as the problem-solving strategy with hypotheses generated by an expert system that monitors the plant for patterns of data symptomatic of known failure modes. Fuzzy logic is employed as the inferencing mechanism with two complementary implication schemes to handle scenarios involving competing failures. Hypothesis testing is performed by simulating the behavior of faulted components using numerical models. A filter has been developed for systematically adjusting key model parameters in an attempt to obtain agreement between simulations and actual plant data. Pattern recognition is employed as a decision analysis technique for choosing among several hypotheses based on simulation results. An artificial intelligence framework based on a critical functions approach is used to deal with the complexity of a nuclear plant. A prototype system for diagnosing transients in the reactor coolant system of a pressurized water reactor has been developed to test the algorithms described here. Results are presented for the diagnosis of data from the Three Mile Island Unit 2 loss-of-feedwater/small-break loss-of-coolant accident.