| Hydraulic six degree of freedom parallel mechanisms have the advantages of high accuracy,high stiffness,versatility,and safety,and have become an important tool for modern industrial manufacturing and automated production.As the power system of a six degree of freedom parallel mechanism,the hydraulic system has the characteristics of high transmission efficiency,large load capacity,and fast response,etc.Once the hydraulic system fails,it will have a serious impact on the movement of the mechanism.Therefore,timely troubleshooting and solving hydraulic system faults is of great significance to ensure the normal operation and safety of the mechanism.Hydraulic system faults are characterized by high frequency,strong fault transmission and multiple types of faults,with complex and diverse fault forms.Artificial intelligence algorithms represented by convolutional neural networks overcome the shortcomings of traditional fault diagnosis based on artificial experience.Relying on the self-learning ability of the network,a faster diagnosis speed and a higher accuracy rate are obtained in the fault diagnosis of the hydraulic system.The Soft Max classification layer in convolutional neural networks can only perform probability distribution on feature data,and has poor classification effects for certain fault types.Aiming at the shortcomings of convolutional neural network,a dual channel fusion convolutional neural network(CNN)plus support vector machine(SVM)fault diagnosis algorithm for six degrees of freedom parallel mechanism hydraulic system is proposed.Select SVM with better classification effects to replace the Soft Max classification layer,further improving the accuracy of fault diagnosis results.The main research contents of this paper are as follows:Firstly,the composition and working principle of the hydraulic six degree of freedom parallel mechanism are analyzed,the kinematic inverse mathematical model of the hydraulic six-degree-of-freedom parallel mechanism is derived.Aiming at the characteristics of multi system collaborative work and strong coupling of each system of the hydraulic six degree of freedom parallel mechanism,a co-simulation model of the hydraulic six degree of freedom parallel mechanism was established using AMESim and MATLAB software.Completed the mechanism analysis of common faults in the hydraulic system,and conducted fault simulation for two types of faults,namely,hydraulic cylinder leakage and servo valve blockage.Through the simulation results,the impact of different faults on the hydraulic system was analyzed.Feature vectors such as hydraulic cylinder displacement,hydraulic cylinder rodless cavity pressure,and hydraulic cylinder rodless cavity flow rate in the hydraulic system are selected as subsequent fault diagnosis samples.Secondly,a one-dimensional convolutional neural network(1DCNN)fault diagnosis model and a two-dimensional convolutional neural network(2DCNN)fault diagnosis model are established.Before performing fault diagnosis,the selected fault diagnosis samples are normalized to eliminate amplitude differences between different signals.The processed samples are fed into the network model for fault classification.Aiming at the problem of incomplete feature extraction from 2DCNN and 1DCNN,a dual channel fusion CNN fault diagnosis model is proposed.Input fault samples into 1DCNN and 2DCNN for feature extraction,add a convergence layer to fuse the extracted feature vectors,and then classify them.Finally,the Soft Max classification layer in convolutional neural networks can effectively classify faults,but there are also some shortcomings.The performance of support vector machine in high-dimensional data processing ability and small sample data processing ability is better than that of Soft Max classification layer.Therefore,this paper chooses the support vector machine(SVM)to replace the Soft Max classification layer in the convolutional neural network to form a dualchannel fusion CNN-SVM fault diagnosis algorithm.The experimental results show that the dual-channel fusion CNN-SVM fault diagnosis model is significantly better than 1DCNN,2DCNN and dual-channel fusion CNN fault diagnosis models in terms of fault diagnosis accuracy. |
