A neural model for transient identification in dynamic processes with “don't know” response

摘要: Abstract This work presents an approach for neural network based transient identification which allows either dynamic or a “don't know” response. The uses two “jump” multilayer networks (NN) trained with the backpropagation algorithm. is used because it useful to dealing very complex patterns, case of space state variables during some abnormal events. first one responsible identification. NN uses, as input, short set (in moving time window) recent measurements each variable avoiding necessity using starting other validate instantaneous (from net) through validation variable. net allowing system provide In order method, Nuclear Power Plant (NPP) problem comprising 15 postulated accidents, simulated pressurized water reactor (PWR), was proposed in process has been considered noisy data evaluate method robustness. Obtained results reveal ability both transients and correct Another important point studied this that shown be independent trigger signal indicates beginning transient, thus making robust relation limitation.