摘要
为提高齿轮箱故障诊断性能,建立了以齿轮箱振动信号的时频域特征为输入,以齿轮箱的主要故障形式为输出的神经网络。采用粒子群优化算法代替反向传播算法来训练神经网络的权重和阈值,利用训练后的神经网络对齿轮箱进行了故障诊断,并比较了基于粒子群优化算法与BP算法的诊断结果。结论是基于粒子群优化算法神经网络具有较好训练性能,收敛速度快,迭代步数少,诊断精度高,具有良好的故障识别率。
In order to improve the performance of gearbox fault diagnosis, the neural network (NN) was established in this paper, based on the feature of vibration signal in time domain and frequency-domain of gearbox used as input vector, while its main fault types used as output vector of NN. The particle swarm optimization ( PSO )algorithm was used to train the weights and the thresholds of NN instead of back propagation (BP) algorithm. The NN trained by PSO was applied to gearbox fault diagnosis. The diagnostic results between PSO and BP algorithm were compared. The conclusion is that NN based on PSO has better training performance, faster convergence rate , minimum iterations ,higher accuracy and an good identification probability of diagnosis.
出处
《振动．测试与诊断》
EI
CSCD
2006年第2期133-137,共5页
Journal of Vibration,Measurement & Diagnosis
基金
国家自然科学基金资助项目(编号:50575214)
关键词
齿轮箱
故障诊断
神经网络
粒子群优化
gearbox fault diagnosis neural network particle swarm optimization (PSO)
