The principle of a multivariate sensor system for controlling of waste water streams is described. The sensor system bases on an array of several simple sensor elements for example the pH-value, conductivity, turbidity or temperature. The sensor elements are arranged single or multiple in the sensor array. An image of forbidden or allowed waste water states arrises by means of methods of pattern recognition. In recent publication the fuzzy clustering and the use of neural network are discussed. The aggressive conditions of a waste water duct can change the sensitivity of sensor elements. Therefore the recognition and characterising of trends are necessary by means of the method of fuzzy logic. A comparison of the mentioned sensor system with analytical methods for the waste water problem as CSB, TOC or methode of specific off-line methods is discussed.