The accurate diagnosis of accidents in a nuclear power plant has fundamental importance for decision making necessary to mitigate their consequences for the power plant as well as for the general public, on the basis of emergency planning. Two main characteristics should be achieved in this kind of diagnostics, namely, real-time features and adaptive capacity. The first characteristic gives the operators the possibility of predicting degraded operations and monitoring critical safety functions related to that specific situation. The second one allows the system to be able to deal with accidents that were not incorporated in the training sample set, in which case the operators are unprepared because they were not trained to face an event that they did not observe even in simulator training. The Three Mile Island accident is a classic one to demonstrate that these kinds of events are possible. Several methodologies have been tried to match those characteristics. While the first one is achieved through the permanent evolution of new faster processors, the second one can only be achieved through the simulation of human cognitive processes, which show higher adaptive behavior. Our model utilizes a neural network, fuzzy sets, and a genetic algorithm to simulate that behavior. We have used a neural network activated by an additive model and trained with an unsupervised competitive training law. Once trained with three accidents (steam generator tube rupture, blackout, and loss-of-coolant accident), a synaptic matrix was obtained, in which the elements represent the interchanging weights between neurons in the concatenated input / output space and the competitive neurons that fight to encode the input-output vector. This kind of competition establishes a statistical classification of the state variables, changing with time, clustering them in centroids labeling the kind of accident for which variables are being sampled. Thus, the accident identification is done in real time with the synaptic matrix. However, the centroids are located in the same time value, in view of the fact that the neural network algorithm treats the variable time as an independent one. Therefore, a genetic algorithm is applied to a fuzzy system formed by the partition of the variables space with fuzzy sets determined by the neural network centroids, in order to estimate the optimal positions in the time variable where the fuzzy system centroids must be located. As a consequence, the diagnostic can be done in representative regions of each accident with maximum confidence.