Research On Gearbox Fault Diagnosis Method Based On Image Coding

With the rapid development of sensor technology and industrial Internet of Things,traditional fault diagnosis systems are facing a series of problems such as low diagnosis efficiency and poor diagnosis accuracy.Gearbox is the most widely used type of transmission components.Its working environment is diverse.Especially for planetary gearboxes with complex structure,continuous heavy load and harsh working conditions,the failure rate is relatively high.Therefore,the key components inside the gearbox are Timely and accurate fault diagnosis is extremely necessary.Deep learning has a wide range of applications in image recognition and speech processing due to its powerful feature extraction capabilities,but its application in the field of fault diagnosis is still in its infancy.Combining the deep learning algorithm with two new image coding methods,two image diagnosis network models are established to diagnose the planetary gears and rolling bearings with different fault forms,and the proposed models are fully verified by setting up a variety of comparative tests.,The results show that it has high diagnostic accuracy and strong generalization ability.(1)A fault diagnosis model of planetary gear based on gramin angle field-convolution neural network(GAF-CNN)was proposed.The vibration signal of planetary gear box was converted into two kinds of characteristic graphs,GADF and GASF,by using two different image coding methods of gramin angle field.Through the optimization of the relevant parameters of CNN algorithm,a higher diagnosis accuracy was obtained.By studying the influence of network parameters,different network layers and different diagnosis methods on the fault diagnosis model,the optimal combination of the models was obtained.The test and comparison analysis results show that GADF-CNN can provide higher recognition accuracy than GASF-CNN;compared with other intelligent algorithms,GADF-CNN has the best effect in planetary gear fault diagnosis,which shows that the fault diagnosis model has good feasibility and reliability in planetary gear fault classification.(2)Aiming at the problem of relevant information between data being not able to be fully utilized during one-dimensional signals haken as input of convolutional neural network(CNN).Using gamin angle difference field(GADF)to encode the collected vibration signal directly,and produce the corresponding GADF feature maps.Then,feature maps were input into the convolutional neural network(CNN)to adaptively complete the feature extraction and classification of rolling bearing with ten different faults.In order to verify the performance of the model,Case Western Reserve University,USA bearing data set was used to do bearing fault diagnosis and analysis,and a common neural algorithm was introduced to contrastively test the classification performance of different models.The results showed that,compared with other image coding methods and neural network,the proposed model still maintains good diagnostic performance during load variation and noise pollution.(3)A fault diagnosis model of rolling bearing based on nested scatter plot-convolution neural network(NSP-CNN)was proposed.Using different sensors as input,and simultaneous interpreting the different frequency bands with FFT,the NSP feature map containing fault information was generated.Then,the optimized CNN algorithm was used to realize the fault diagnosis of rolling bearings under different speed conditions.The results show that the accuracy of the test set of the model exceeded 99.30%.By set up other comparative experiments,the influence of sensor location,image size and different optimizers on the performance of the model was clarified.