摘要
目的为了解决掘进机截割头故障振动信号信息量大且难以充分提取和现有的掘进机截割头故障诊断模型参数量多且计算量大,训练需要耗费大量时间,部署和使用对硬件设备的性能要求严苛的问题。方法提出一种小体积和运行速度快的轻量化网络模型时空特征融合(STFF)-ShuffleNet用于掘进机截割头故障诊断研究。ShuffleNet在模型进行时空特征提取时引入深度可分离卷积提高模型的运算效率,再对融合后的时空特征做通道混洗和分组卷积,既解决了通道分组的问题又提高了诊断精度。通过悬臂式掘进机截割头实验台模拟不同故障条件,将采集到的振动信号输入模型中进行测试。结果实验表明,STFF-ShuffleNet模型对掘进机截割头的诊断与其他传统模型相比参数量少,收敛速度快,运行效率更高,更容易满足掘进机截割头故障诊断高实时性要求,准确率达到99.3%。结论STFF-ShuffleNet模型对提高掘进机截割头故障诊断效率和降低硬件要求提供了一种新的解决思路。
Objective To address the issues that the vibration signal of the roadheader cutting head contains extensive information and is challenging to extract comprehensively,and that the existing fault diagnosis model for the roadheader cutting head has a considerable number of parameters and a significant amount of computation,requiring a considerable amount of time for training,as well as demanding high-performance hardware equipment for deployment and utilization.Methods A lightweight network model featuring small volume and rapid running time,namely space-time feature fusion(STFF)-ShuffleNet,was proposed for the fault diagnosis of the roadheader cutting head.ShuffleNet introduced depth-separable convolution to enhance the operational efficiency of the model when extracting spatio-temporal features,and subsequently conducted channel mixing and grouped convolution for the merged spatio-temporal features,not only resolving the problem of channel grouping but also improving the diagnostic accuracy.By simulating diverse fault conditions on the cutting head test bench of the cantilever roadheader,the collected vibration signals were input into the model for testing.Results The experimental results indicate that the STFF-ShuffleNet model ha fewer parameters,a faster convergence speed,and higher operational efficiency compared to other traditional models,and it is more likely to fulfill the requirement of high real-time fault diagnosis of the roadheader cutting head,achieving an accuracy rate of 99.3%.Conclusion The STFF-ShuffleNet model offers a novel approach to enhance the fault diagnosis efficiency of the roadheader cutting head and reduce the hardware requirements.
作者
马天兵
陈旭升
童玮
李长鹏
许吉禅
MA Tianbing;CHEN Xusheng;TONG Wei;LI Changpeng;XU Jichan(State Key Laboratory of Mining Response and Disaster Prevention and Control in Deep Coal Mines,Anhui University of Science and Technology,Huainan Anhui 232001,China;School of Mechanical and Electrical Engineering,Anhui University of Science and Technology,Huainan Anhui 232001,China)
出处
《安徽理工大学学报（自然科学版）》
CAS
2024年第4期1-10,共10页
Journal of Anhui University of Science and Technology:Natural Science
基金
国家重点实验室基金资助项目(SKLMRDPC20ZZ01)
安徽省重点研究与开发计划基金资助项目(202104a07020005)
安徽省智能矿山技术与装备工程实验室开放基金资助项目(AIMTEEL202202)
安徽省高校自然科学研究基金资助项目(2023AH051196)
安徽高校协同创新基金资助项目(GXXT-2022-019)。
关键词
掘进机截割头
故障诊断
时空融合
轻量化网络
卷积神经网络
roadheader cutting head
fault diagnosis
space-time fusion
lightweight network
convolutional neural network
作者简介
马天兵(1981-),男,安徽合肥人,教授,博士,研究方向:振动测试与故障诊断。
