基于PSO-BP的齿轮箱故障检测机制研究  

Research on the Gear Box Fault Detection Based on PSO-BP Intelligent Computation

作　　者：蒋年华[1] 

机构地区：[1]广西农业职业技术学院电子信息工程系,南宁530007

出　　处：《计算机测量与控制》2013年第8期2083-2086,共4页Computer Measurement &Control

摘　　要：传统的齿轮箱故障检测主要依靠参数测量后人工对信号进行分析,没有形成自动检测的系统模型,由于相关参数复杂且数量较大,经常会出现漏诊与虚诊现象。借鉴计算机领域的人工智能算法,提出一种基于PSO-BP的齿轮箱故障检测方法,使用特种传感器对齿轮箱故障信号进行采集,构造以时域、频域信号为输入的BP神经网络,使用粒子群优化方法对网络权重系数与阈值进行优化矫正,将故障类型作为神经网络的输出,通过计算机中的模拟测试实验证明,经过优化后的神经网络模型可以有更好的局部优化性能,故障诊断的准确率(在精度一定的情况下进行实验,随着粒子数的不断上升)较优化前有20%～30%的提升,因此具有很强的实用价值。The traditional gear box fault detection relies mainly on the parameters measurement of signal after artificial analysis, forming no automatic detection system model, due to the correlation parameter complex and the amount is larger, often can appear misdiagnosis and virtual clinical phenomenon. Reference in the field of computer artificial intelligence algorithm, the paper proposes a kind of based on PSO -- BP gear box fault detection method, the use of special sensor on gear box fault signal acquisition of structure in time domain and frequency do- main signal as input of the BP neural network, using particle swarm optimization method for network weight coefficient and threshold optimi- zation correction, fault type as the output of neural network, through the computer simulation test experiment, through the optimized neural network model can have better local optimization performance, the fault diagnosis of the high accuracy（with about 20%--30% increasement） has great practical value.

关 键 词：粒子群优化 系数矫正 齿轮故障 信号特征提取 

分 类 号：TM61[电气工程—电力系统及自动化]