Traffic Pattern Recognition Of Elevator Systems Based On Fuzzy Neural Networks

With the developing of high buildings, passengers have more needs to elevator services, but one elevator couldn't satisfy all needs of passengers. In order to cut down waiting time, reduce energy using, we need put many elevators together and control them together in reason, and this optimization dispatch systems to all the elevators name elevator group control systems (EGCS). Traffic flow is an important factor of EGCS. so the search of it is very important, and a new development of EGCS is it can recognize the change of traffic flow. Traffic flow is expressed by the number of passengers, the period of passengers appearing, as well as the positions of passengers. Traffic flow expresses a traffic status of EGCS. It is possible and needed to use intelligent control methods to elevator group for it randomicity. nonlinearity. and difficulty to set up an exact mathematics mode.Fuzzy logic has been used to elevator group control systems successfully, for it can get rid of the tie of exact mathematics modes, summarize and use the men's manipulation and control experiences, imitate the logic reasoning and decision-making of men's brain, be good robust, deal good with systems' diversity, randomicity, and nonlinearity. But it has not a learning ability, and it all bases on a lot of control rules that were made by experts. Neural networks (NN) is a dynamic nonlinearity system, which bases on some topology structures. It has strong abilitiesof learning, toleration and extending. And the main disadvantage of NN is the difficulty to certain its structure, needing a lot of training stylebooks, long training periods, and can not provide a certain frame which is used to express network knowledge. Considering the virtues and flaws of both, we band the two methods together, take in the virtues and get ride of the flaws, and get fuzzy neural networks (FNN). FNN has an expression frame, one side it could provide a comprehensible mode structure, which provides expression and reasoning, and another side it has the abilities of getting knowledge and learning.A traffic pattern recognition method of elevator group control systems based on fuzzy neural networks is presented in this paper. The recognition works by two FNN and two-step. The first network recognizes the proportions of up-peak traffic pattern, down-peak traffic pattern, idle traffic pattern, and inter-floor traffic pattern by all the passenger, the passenger who get in the hall and the passenger who get out the hall, when all this happen to a unit time. And the second network recognizes balance inter-floor traffic pattern, two-way traffic pattern, and four-way traffic pattern, by the passenger of the biggest floor and the passenger of the second biggest floor, when the proportion of inter-floor traffic pattern is bigger. Stylebooks are made by expert knowledge, and a two-step hybrid method is adopted to train the two fuzzy neural networks. Testing results proves the efficiency of the two FNN that can recognize all kinds of traffic patterns, the efficiency of this method that can instruct the group controller to optimize dispatch strategy and improve the service of the elevator group.