As an application of ART 2 neural networks, computer aided monitoring of pump efficiency is successfully examined for an industrial waste-liquid treatment process with measured data of valve openness and liquid flow rates. By running the neural networks in parallel, we confirm that accuracy to detect system changes is good, and the adjustment of classifier parameters is relatively easy. Investigating the resulting classes carefully, frequency of each class is correlated with pump efficiency. The relative amount of variables are also related to the classes.