Abstract
In this paper a signal processing approach for improving the performance of dynamic weighing systems by using a combination of Fuzzy Logic and Artificial Neural Networks is presented. The paper indicates that the technique of combining these two methods is a viable design solution for article weight estimation from stochastic signals. The Neuro-Fuzzy system adopted is an Adaptive Network based Fuzzy Inference System (ANFIS) [1]. This method is further enhanced by introducing a systematic approach in deciding the number and initial shape of the membership functions. The number and shape of membership functions were determined by means of fuzzy space clustering. This new Enhanced ANFIS (EANFIS) method was experimentally verified and the obtained results, when compared to the results of classical signal processing, showed a significant improvement in both the accuracy and the throughput rate of an industrial dynamic checkweigher.