Plant water quality diagnosing control system

A neural network model comprising a stratified structure having at least one intermediate layer between an input layer and an output layer of water quality information from a sensor is used, and among water quality information in the past, a typical pattern of variable values at different time points is inputted. Then, a control variable value corresponding to the typical pattern is learned as a reference signal to the neural network model. A variable value for the aimed water quality control is determined by inputting the water quality pattern not yet learned to the learned neutral network model to make the neural network model learn, to operate a control system. With such a constitution, an optimal output signal corresponding to the input pattern is outputted depending on the learned level. Accordingly, if sufficiently learned neutral network is used, an optimal output control signal capable of controlling the water quality relative to the water quality pattern input signal of such a level that a man can not correspond can be obtained, even to a water quality pattern input signal at such a level as considered to be impossible for a man to correspond. (N.H.)