Method of predicting elevator failure based on bp neural network optimized by pso

Abstract

The invention discloses a method for predicting elevator faults based on a PSO-optimized BP neural network. The initial inertia weight value of the existing PSO particle swarm optimization algorithm is not large enough, and the inertia weight value decreases not fast enough in the later stage. The invention improves the standard PSO algorithm, uses the global optimal value obtained by the improved PSO algorithm as the weight and bias of BP, and obtains the BP neural network optimized by the improved PSO algorithm. The improved PSO algorithm of the present invention has a larger inertia weight than the standard PSO algorithm in the initial stage of iteration, so that the ability of particles to converge on the global optimal value is improved, and has a strong global search ability; the approximate range of the optimal value is searched in the later stage of iteration When , the improved PSO algorithm has a smaller inertia weight than the standard PSO algorithm, so that the improved PSO algorithm shows more diversity of particle swarms than the standard PSO algorithm, and completes the fine search.

Description

technical field [0001] The invention relates to the technical field of intelligent elevator safety monitoring, in particular to a method for predicting elevator faults based on a PSO-optimized BP neural network. Background technique [0002] In today's rapidly developing technological society, fault prediction has begun to develop towards automation and intelligence, and many hybrid intelligent algorithms have also emerged as a result. PSO particle swarm optimization algorithm is an intelligent search algorithm, which can quickly find the optimal solution of the problem, but PSO particle swarm optimization algorithm itself is easy to fall into local optimum. The inertia weight ω of the particle swarm optimization algorithm decreases linearly. The initial inertia weight value is large, the convergence speed is fast, and the rough search is conducive to the global search, which meets the requirements of the particle exploration ability in the early stage; while the inertia wei...

AI Technical Summary

Problems solved by technology

First, if the particle swarm finds a better point in the early stage, it is hoped that the particle will quickly converge to the global optimal point, but the value of the initial inertia weight is not large enough, and the convergence speed of the particle needs to be improved; second, in the later stage of the iteration, because the inertia weight The value decreases linearly, not fast enough to make fine-grained searches less capable

Images