Research Direct Algorithm Based On Artificial Neural Networks Elevator Guide Rails School

With the rapid development of the elevator guide rail industry,there is a greater demand for its accuracy,and straightness indicator is the most important factor.In recent years, automatic elevator rails alignment machine in our county also gradually replaced the previous manual straightening situation,but the problem is that straightening efficiency is not ideal.To solve this problem,some new ways need to be found to improve the alignment algorithm,thus improving the straightening efficiency of high precision guide rail.And in all factors that influence efficiency of high precision guide rail straightening, design of overvoltage is the most critical reason. So this paper raises calculation scheme of overvoltage based on neural netwok with the developed elevator guide rail straightening hardware platform.At first,the paper provides the elevator guide rail straightening system,including the overall scheme design,the concept of straightness error,the mechanical structure of rail alignment and straightening principle etc.Secondly this paper analyzes the basic knowledge of neural network including BP network model,BP learning algorithm and steps etc,lays the foundation for MFC programming realization of neural network algorithm.Then this paper describes the rail straightening algorithm from a unitary perspective,including the rail straightening process, the design of pressure point location and the design of single pressure point straighening process etc.This paper also introduces calculation method of overvoltage in detail from the local perspective,the focus is on the implementation method of training samples in MFC programming and application process of neural network in the straightening system.At last,debugging the straightening system based on neural network. When the guide rail initial value is smaller,the straightening efficiency is very high in the cases of different form of guide rail bending deformation.And Improved alignment algorithm is proposed according to larger initial value over5millimetre of guide rail.Meanwhile,the paper also puts forward the expansion of neural network application in view of the situation that the three indicators including whole long bending value and the left right500millimetre value are not qualified.Experiments show that the time of rail detection is reduced with the improved algorithm.