Abstract: Elevator traffic demand pattern recognition is the prerequisite for effectively implementing the strategies of elevator group control system. In view of the characteristics of time-varying, nonlinear and uncertainty of elevator traffic demand, an elevator traffic pattern recognition method with FCM (Fuzzy C-means) clustering based fuzzy neural network is presented. The method introduces the fuzzy logic into the calculation and learning of BP neural network, and employs FCM clustering algorithm to cluster the original traffic demand to realize the fuzzy partition of the input space of fuzzy system to determine the initial value of network membership function and clustering center and to obtain the fuzzy rules, which improves the learning ability of neural network and makes the weighted coefficients of the membership function vary with different traffic patterns. The elevator traffic pattern is recognized by the parallel fuzzy reasoning of neural network. Simulation experiments show the validity of the presented method.

Key words: elevator traffic demand, pattern recognition, fuzzy neural network, FCM clustering algorithm, expert experience