基于PSO-BP复合网络的掘进机截割部故障智能诊断
Intelligent fault diagnosis of cutting part of tunnel boring machine based on PSO-BP composite network
张世丽1
作者信息
- 1. 潞安化工集团常村煤矿,山西长治 046000
- 折叠
摘要
针对井下掘进机故障诊断频发,传统诊断方法和BP神经网络诊断周期长的情况,以常村煤矿矿用EBZ-160TY型掘进机为背景,提出基于PSO-BP神经网络模型的掘进机截割部智能诊断模型.该模型能够弥补BP神经网络模型收敛周期长、局部最优搜索差的缺点,实现模型的快速收敛和故障准确预测.通过设置PSO-BP神经网络模型参数、样本数据训练,同时经过数据测试,确定PSO-BP神经网络模型预测结果故障预测率为100％,而BP神经网络的预测精度为80％,且在同时间下,PSO-BP神经网络较BP神经网络预测精度更高.在同精度下,PSO-BP神经网络模型收敛速度更快,在精度为1 ×10-5时,PSO-BP神经网络模型仅需7步,BP神经网络平均需要198.5步.综合测试结果说明,PSO-BP神经网络模型能够较快实现掘进机故障的预测,且达到较高的预测精度,为掘进机故障诊断提供依据.
Abstract
In view of the frequent fault diagnosis of underground tunnel boring machine(TBM)and the long diagnosis period of traditional diagnosis methods and BP neural network,taking EBZ-160TY TBM in Changcun Coal Mine as the back-ground,an intelligent diagnosis model of TBM cutting part based on PSO-BP neural network model is proposed,which can o-vercome the long convergence period of BP neural network model and realize the fast convergence of the model and the ac-curate prediction of fault.By setting PSO-BP neural network model parameters,sample data training,and data testing at the same time,it is determined that the fault prediction rate of the PSO-BP neural network model is 100％,while the prediction accuracy of BP neural network is 80％.Moreover,at the same time,the prediction accuracy of PSO-BP neural network is higher than that of BP neural network.Under the same accuracy,the convergence speed of PSO-BP neural network model is faster.When the accuracy is 1 × 10-5,the PSO-BP neural network model only needs 7 steps,while the average BP neural network needs 198.5 steps.The comprehensive test results show that the PSO-BP neural network model can realize the fas-ter prediction of fault for TBM with higher accuracy,which provides a basis for fault diagnosis of TBM.
关键词
掘进机截割部/PSO-BP神经网络模型/故障智能诊断/数据样本/收敛速度/预测精度Key words
TBM cutting part/PSO-BP neural network model/intelligent fault diagnosis/data sample/convergence rate/pre-diction accuracy引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据