基于小波神经网络的设备故障诊断方法研究
|
摘要
神经网络以其固有的记忆能力、自学习能力以及强容错性为故障诊断问题提供了一个新方法。本文针对科学实验中广泛使用的平流泵的故障特点，深入研究了BP神经网络的故障诊断方法。
首先用小波包分析技术做信号处理。选取db3小波函数，用硬阈值小波包降噪的方法将信号降噪，然后进行小波包分解与重构，以提取信号的能量特征向量，并将得到的特征向量作为神经网络的输入。
本文采用具有一个隐含层的三层BP神经网络进行故障诊断，深入分析故障诊断的结果后发现：第一，网络容易陷入极小值而导致诊断失败；第二，网络的隐含层节点数难以确定。为了解决上述问题，本文研究设计了GA+BP算法。该方法是将遗传算法与神经网络相结合。首先，GA对BP神经网络做前期优化，确定出最佳网络结构及该结构对应的初始权值、阈值和网络的学习速率；然后，构造具有最佳结构和参数的神经网络来进行故障诊断。GA+BP算法的设计中，把每个染色体分解为连接基因和参数基因，对这两部分采取不同的遗传操作。连接基因采用二进制编码方法，参数基因采用实数编码方法；连接基因采用一点交叉方式和基本变异方式，参数基因中的权阈基因和速率基因各自采用算术交叉方式和非均匀变异方式。另外，交叉算子和变异算子都采用自适应的方法。
GA+BP神经网络与BP神经网络故障诊断的结果对比后可以看到：第一，GA+BP神经网络比BP神经网络的工作量少，且克服了陷入局部极小的缺点，有更好的训练性能；第二，GA+BP神经网络的故障诊断准确率高于BP神经网络。由此可见，GA+BP神经网络能够更好的进行平流泵的故障诊断工作。
Neural network offers a new method for fault diagnosis owing to its memory ability,self-learning ability and strongly fault tolerance. This paper makes research on the faultdiagnosis method of neural network deeply based on the fault characteristics of pump whichis widely used in experiment.
Wavelet packet analysis is used to do the signal processing. Wavelet db3 is chosen, andall signals are de-noised by hard threshold de-noising method. Then wavelet packetdecomposes and constructs the energy eigenvectors which are regarded as the inputeigenvectors of the neural network.
A three-layer BPNN is applied to do the fault diagnosis. The results of simulation showthat the network traps in local minimum easily, and both the number of hidden neurons andthe learning rate are difficult to decide either.
In order to solve these questions above, this paper designs GA+BP algorithm. In thisalgorithm, genetic algorithm is used to optimize the number of hidden neurons, the initialweights and thresholds, and the learning rate of BPNN first, and then fault diagnosis is doneby this neural network which has the optimum structure and parameters. In GA+BP neuralnetwork, each chromosome is divided into the connection genes and the parameter genes, anddifferent genetic operations are carried on two parts. Connection genes are binary type andparameter genes are real-valued. Mixed crossover and mutation operations are operated on theconnection genes and parameter genes separately. It means the connection genes adoptsingle-point crossover and simple mutation, and the parameter genes adopt arithmeticcrossover and non-uniform mutation. Both the crossover and mutation operators adoptself-adaptive method.
Comparing the simulation results of GA+BP neural network with BPNN, we know thatGA+BP neural network has less work but high training performance, and the local minimumis inexistent. In addition, the GA+BP neural network can diagnose the failure more correctlythan BPNN. In conclusion, GA+BP neural network can accomplish the pump fault diagnosismuch better.
首先用小波包分析技术做信号处理。选取db3小波函数，用硬阈值小波包降噪的方法将信号降噪，然后进行小波包分解与重构，以提取信号的能量特征向量，并将得到的特征向量作为神经网络的输入。
本文采用具有一个隐含层的三层BP神经网络进行故障诊断，深入分析故障诊断的结果后发现：第一，网络容易陷入极小值而导致诊断失败；第二，网络的隐含层节点数难以确定。为了解决上述问题，本文研究设计了GA+BP算法。该方法是将遗传算法与神经网络相结合。首先，GA对BP神经网络做前期优化，确定出最佳网络结构及该结构对应的初始权值、阈值和网络的学习速率；然后，构造具有最佳结构和参数的神经网络来进行故障诊断。GA+BP算法的设计中，把每个染色体分解为连接基因和参数基因，对这两部分采取不同的遗传操作。连接基因采用二进制编码方法，参数基因采用实数编码方法；连接基因采用一点交叉方式和基本变异方式，参数基因中的权阈基因和速率基因各自采用算术交叉方式和非均匀变异方式。另外，交叉算子和变异算子都采用自适应的方法。
GA+BP神经网络与BP神经网络故障诊断的结果对比后可以看到：第一，GA+BP神经网络比BP神经网络的工作量少，且克服了陷入局部极小的缺点，有更好的训练性能；第二，GA+BP神经网络的故障诊断准确率高于BP神经网络。由此可见，GA+BP神经网络能够更好的进行平流泵的故障诊断工作。
Neural network offers a new method for fault diagnosis owing to its memory ability,self-learning ability and strongly fault tolerance. This paper makes research on the faultdiagnosis method of neural network deeply based on the fault characteristics of pump whichis widely used in experiment.
Wavelet packet analysis is used to do the signal processing. Wavelet db3 is chosen, andall signals are de-noised by hard threshold de-noising method. Then wavelet packetdecomposes and constructs the energy eigenvectors which are regarded as the inputeigenvectors of the neural network.
A three-layer BPNN is applied to do the fault diagnosis. The results of simulation showthat the network traps in local minimum easily, and both the number of hidden neurons andthe learning rate are difficult to decide either.
In order to solve these questions above, this paper designs GA+BP algorithm. In thisalgorithm, genetic algorithm is used to optimize the number of hidden neurons, the initialweights and thresholds, and the learning rate of BPNN first, and then fault diagnosis is doneby this neural network which has the optimum structure and parameters. In GA+BP neuralnetwork, each chromosome is divided into the connection genes and the parameter genes, anddifferent genetic operations are carried on two parts. Connection genes are binary type andparameter genes are real-valued. Mixed crossover and mutation operations are operated on theconnection genes and parameter genes separately. It means the connection genes adoptsingle-point crossover and simple mutation, and the parameter genes adopt arithmeticcrossover and non-uniform mutation. Both the crossover and mutation operators adoptself-adaptive method.
Comparing the simulation results of GA+BP neural network with BPNN, we know thatGA+BP neural network has less work but high training performance, and the local minimumis inexistent. In addition, the GA+BP neural network can diagnose the failure more correctlythan BPNN. In conclusion, GA+BP neural network can accomplish the pump fault diagnosismuch better.
引文
[1]徐敏，黄昭毅.设备故障诊断手册[M].西安：西安交通大学出版社，1998：16-18
[2]黄文虎，夏松波，刘瑞岩.设备故障诊断原理、技术及应用[M].北京：科学出版，1996：4-5
[3]徐章隧，房立清.故障信息诊断原理及其应用[M].北京：国防工业出版社，2001：71-73
[4]谷云辉，刘亚斌.基于自动测试系统的故障诊断方法研究[J].微计算机信息，2005(21)：146-148
[5]周东华，叶银忠.现代故障诊断与容错控制[M].北京：清华大学出版社，2000：43-53
[6]朱平，黄文虎等.基于模型的传感器故障诊断技术的研究[J].传感技术学报，1999(1)：22-28
[7]徐涛,王祁.一种神经网络预测器在传感器故障诊断中的应用[J].传感技术学报，2005，18(2)：235-237
[8]张志涌，徐彦琴等. MATLAB教程[M].北京：北京航空航天大学出版社，2001：1-3
[9]魏海，沈兰荪.反对称双正交小波应用于多尺度边缘提取的研究[J].电子学报，2002，30(3)：313-316
[10]胡战虎.基于贝叶斯估计的多分辨图像滤波方法[J].电子学报，2002，30 (1)：66-68
[11]王卫卫，杨波，宋国乡.基于二进小波图像边缘的新相位水印算法[J].计算机学报，2002, 25(7)：767-771
[12]赵育良，赵友庚，李开端.基于小波变换的复杂航空图像的边缘提取[J].光电工程，2002，29(4)：57-60
[13]赵瑞珍，宋国乡，屈汉章.基于小波变换的汉语声调识别新方法[J].信号处理，2000,16(4)：57-361
[14]吴小培，冯焕清，周荷琴.基于小波变换的脑电瞬态信号检测明[J].数据采集与处理，2001，16(1)：86-89
[15]杜干，张群，张守宏.小波变换在分形信号参数估计中的应用[J].西南电子科技大学学报，2000，27(3)：281-284
[16] Turkheimer.F.E. A linear wavelet filter for parametric imaging with dynamic PET[J].IEEE Transactions on Medical Imaging, 2003, 22(3):289-301
[17] Barmada.S. Analysis of transmission lines with frequency-dependent parameters bywavelet-FFT method [J].IEEE Transactions on Magnetics, 2003, 3 9(3):1602-1605
[18] Menegaz.G. Three-dimensional encoding/two-dimensional decoding of medical data[J].IEEE Transactions on Medical Imaging, 2003, 22(3):424-440
[19] Mensah-Bonsu.C. Real-time digital processing of GPS measurements for transmissionengineering [J].IEEE Transactions on Power Delivery，2003，18 (1):177- 182
[20] Goumas.S.K. Classification of washing machines vibration signals using discrete waveletanalysis for feature extraction [J].IEEE Transactions on Instrumentation andMeasurement, 2002, 51(3):497-508
[21]赵红怡，武梦龙等．小波分析在突变信号检测中的应用[J].北方工业大学学报，2004，16(1)：21-24
[22]飞思科技产品研发中心，小波分析理论与MATLAB7实现[M].电子工业出版社，2005，109-120
[23]胡昌华，李国华等.基于MATLAB 6.X的系统分析与设计——小波分析[M].西安电子科技大学出版社，2004
[24]于润伟.MATLAB基础及应用[M].机械工业出版社，2003，10
[25]董长虹，高志，余啸海.小波分析工具箱原理与应用[M].国防工业出版社，2004
[26] Zhang Q H, Benveniste A. Wavelet networks [J]. IEEE Transactions on Neural Networks,1992, 3(6): 889-898
[27]田慕玲，王晓玲.电机故障诊断中的小波分析方法及小波基选取[J].煤矿机械，2007(5)：176-178
[28] Jafar Zarei，Javad Poshtan．Bearing fault detection using wavelet packet transform ofinduction motor stator current[J].Tribology International，2007,40(5):763-769
[29] N．G．Nikolaou，I．A．Antoniadis．Rolling element bearing fault diagnosis using waveletpackets[J].NDT&E International，2002；34：197-205
[30]徐涛，王祈.基于小波包神经网络的传感器故障诊断方法[J].传感技术学报，2006，19(4)：1060-1064
[31]飞思科技产品研发中心，神经网络理论与MATLAB7实现[M].电子工业出版社，2005，109-120.
[32] Abhinav Saxena, Ashraf Saad. Evolving an artificial neural network classifier forcondition monitoring of rotating mechanical systems [J]. Applied Soft Computing,7(2007): 441-454
[33]吴涛，许晓鸣，刘登流，张浙.基于改进算法的人工神经网络建模及其在干燥过程中的应用[J] .中国控制会议路论文集，1998，9，677-681
[34]李奇，李世华.一类神经网络智能PID控制算法的分析与改进[J].控制与决策，1998，13(4)，311-136
[35] Hong-Nan Li, Hao Yang. System identification of dynamic structure by the multi-branchBPNN[J]. Neurocomputing, 2007, V70: 835-841
[36] Howard Demuth&Mark Beale. Neural network toolbox for use with MATLAB[J]. TheMathworks, 2001:16-19
[37]阎平凡，张长水.人工神经网络与模拟进化计算[M].北京：清华大学出版社，2002：547-585
[38]席裕庚，柴天佑，浑为民.遗传算法综述[J].控制理论与应用，1996，13（6）：697-708
[39] B.Samanta. Artificial neural networks and genetic algorithms for gear fault detection [J].Mech. Syst. Signal Process, 2004, 18(5):1273-1282
[40]王宏刚，钱锋.基于遗传算法的前向神经网络结构优化[J]，2007，14（4）：387-190
[41] S.M. Jakubek, T.I. Strasser. Artificial neural networks for fault detection in large scaledata acquisition systems [J]. Engineering Applications of Artificial Intelligence,2004(17): 233-248
[42] Schaffer.J.D, Caruana.R.A, Eshelman.L.J. A study of control parameters affecting onlineperformance of genetic algorithms for function optimization [J]. Los Altos:MorganKaufmann Publishers, Inc,1989,51-60
[43] A.G.Olabia, G.Casalino, An ANN and Taguchi algorithms integrated approach to theoptimization of laser welding [J].Advances in Engineering Software, 37(2006): 643-648
[44] Dam M, Saraf D N. Design of neural networks using genetic algorithm for on-lineproperty estimation of data mining applications [J]. Computer and Chemical Engineering,2006, 30(4): 722-729
[45] LU Xin-lai, LIU Hu, WANG Gang-lin, WU Zhe. Helicopter sizing based on geneticalgorithm optimized neural network [J]. Chinese Journal of Aeronautics, 2006, 19 (3):212-218
[46] Shiwei Yu, Kejun Zhu, Fengqin Diao. A dynamic all parameters adaptive BP neuralnetworks model and its application on oil reservoir prediction [J].Applied Mathematicsand Computation, 195(2007): 66-75
[2]黄文虎，夏松波，刘瑞岩.设备故障诊断原理、技术及应用[M].北京：科学出版，1996：4-5
[3]徐章隧，房立清.故障信息诊断原理及其应用[M].北京：国防工业出版社，2001：71-73
[4]谷云辉，刘亚斌.基于自动测试系统的故障诊断方法研究[J].微计算机信息，2005(21)：146-148
[5]周东华，叶银忠.现代故障诊断与容错控制[M].北京：清华大学出版社，2000：43-53
[6]朱平，黄文虎等.基于模型的传感器故障诊断技术的研究[J].传感技术学报，1999(1)：22-28
[7]徐涛,王祁.一种神经网络预测器在传感器故障诊断中的应用[J].传感技术学报，2005，18(2)：235-237
[8]张志涌，徐彦琴等. MATLAB教程[M].北京：北京航空航天大学出版社，2001：1-3
[9]魏海，沈兰荪.反对称双正交小波应用于多尺度边缘提取的研究[J].电子学报，2002，30(3)：313-316
[10]胡战虎.基于贝叶斯估计的多分辨图像滤波方法[J].电子学报，2002，30 (1)：66-68
[11]王卫卫，杨波，宋国乡.基于二进小波图像边缘的新相位水印算法[J].计算机学报，2002, 25(7)：767-771
[12]赵育良，赵友庚，李开端.基于小波变换的复杂航空图像的边缘提取[J].光电工程，2002，29(4)：57-60
[13]赵瑞珍，宋国乡，屈汉章.基于小波变换的汉语声调识别新方法[J].信号处理，2000,16(4)：57-361
[14]吴小培，冯焕清，周荷琴.基于小波变换的脑电瞬态信号检测明[J].数据采集与处理，2001，16(1)：86-89
[15]杜干，张群，张守宏.小波变换在分形信号参数估计中的应用[J].西南电子科技大学学报，2000，27(3)：281-284
[16] Turkheimer.F.E. A linear wavelet filter for parametric imaging with dynamic PET[J].IEEE Transactions on Medical Imaging, 2003, 22(3):289-301
[17] Barmada.S. Analysis of transmission lines with frequency-dependent parameters bywavelet-FFT method [J].IEEE Transactions on Magnetics, 2003, 3 9(3):1602-1605
[18] Menegaz.G. Three-dimensional encoding/two-dimensional decoding of medical data[J].IEEE Transactions on Medical Imaging, 2003, 22(3):424-440
[19] Mensah-Bonsu.C. Real-time digital processing of GPS measurements for transmissionengineering [J].IEEE Transactions on Power Delivery，2003，18 (1):177- 182
[20] Goumas.S.K. Classification of washing machines vibration signals using discrete waveletanalysis for feature extraction [J].IEEE Transactions on Instrumentation andMeasurement, 2002, 51(3):497-508
[21]赵红怡，武梦龙等．小波分析在突变信号检测中的应用[J].北方工业大学学报，2004，16(1)：21-24
[22]飞思科技产品研发中心，小波分析理论与MATLAB7实现[M].电子工业出版社，2005，109-120
[23]胡昌华，李国华等.基于MATLAB 6.X的系统分析与设计——小波分析[M].西安电子科技大学出版社，2004
[24]于润伟.MATLAB基础及应用[M].机械工业出版社，2003，10
[25]董长虹，高志，余啸海.小波分析工具箱原理与应用[M].国防工业出版社，2004
[26] Zhang Q H, Benveniste A. Wavelet networks [J]. IEEE Transactions on Neural Networks,1992, 3(6): 889-898
[27]田慕玲，王晓玲.电机故障诊断中的小波分析方法及小波基选取[J].煤矿机械，2007(5)：176-178
[28] Jafar Zarei，Javad Poshtan．Bearing fault detection using wavelet packet transform ofinduction motor stator current[J].Tribology International，2007,40(5):763-769
[29] N．G．Nikolaou，I．A．Antoniadis．Rolling element bearing fault diagnosis using waveletpackets[J].NDT&E International，2002；34：197-205
[30]徐涛，王祈.基于小波包神经网络的传感器故障诊断方法[J].传感技术学报，2006，19(4)：1060-1064
[31]飞思科技产品研发中心，神经网络理论与MATLAB7实现[M].电子工业出版社，2005，109-120.
[32] Abhinav Saxena, Ashraf Saad. Evolving an artificial neural network classifier forcondition monitoring of rotating mechanical systems [J]. Applied Soft Computing,7(2007): 441-454
[33]吴涛，许晓鸣，刘登流，张浙.基于改进算法的人工神经网络建模及其在干燥过程中的应用[J] .中国控制会议路论文集，1998，9，677-681
[34]李奇，李世华.一类神经网络智能PID控制算法的分析与改进[J].控制与决策，1998，13(4)，311-136
[35] Hong-Nan Li, Hao Yang. System identification of dynamic structure by the multi-branchBPNN[J]. Neurocomputing, 2007, V70: 835-841
[36] Howard Demuth&Mark Beale. Neural network toolbox for use with MATLAB[J]. TheMathworks, 2001:16-19
[37]阎平凡，张长水.人工神经网络与模拟进化计算[M].北京：清华大学出版社，2002：547-585
[38]席裕庚，柴天佑，浑为民.遗传算法综述[J].控制理论与应用，1996，13（6）：697-708
[39] B.Samanta. Artificial neural networks and genetic algorithms for gear fault detection [J].Mech. Syst. Signal Process, 2004, 18(5):1273-1282
[40]王宏刚，钱锋.基于遗传算法的前向神经网络结构优化[J]，2007，14（4）：387-190
[41] S.M. Jakubek, T.I. Strasser. Artificial neural networks for fault detection in large scaledata acquisition systems [J]. Engineering Applications of Artificial Intelligence,2004(17): 233-248
[42] Schaffer.J.D, Caruana.R.A, Eshelman.L.J. A study of control parameters affecting onlineperformance of genetic algorithms for function optimization [J]. Los Altos:MorganKaufmann Publishers, Inc,1989,51-60
[43] A.G.Olabia, G.Casalino, An ANN and Taguchi algorithms integrated approach to theoptimization of laser welding [J].Advances in Engineering Software, 37(2006): 643-648
[44] Dam M, Saraf D N. Design of neural networks using genetic algorithm for on-lineproperty estimation of data mining applications [J]. Computer and Chemical Engineering,2006, 30(4): 722-729
[45] LU Xin-lai, LIU Hu, WANG Gang-lin, WU Zhe. Helicopter sizing based on geneticalgorithm optimized neural network [J]. Chinese Journal of Aeronautics, 2006, 19 (3):212-218
[46] Shiwei Yu, Kejun Zhu, Fengqin Diao. A dynamic all parameters adaptive BP neuralnetworks model and its application on oil reservoir prediction [J].Applied Mathematicsand Computation, 195(2007): 66-75