Auto-Recognition Of Shock Absorber Indicator Diagram Based On Support Vector Machine

As the key component for the vehicle suspension system, hydraulic shock absorbers need to be tested to check the performance. Nowadays the indicator diagrams are obtained from indicator machine, and judged by artificial way. The efficiency is low and depended on variety of uncertain factors. The support vector machine (SVM) technology is discussed in this paper. Auto-recognition function can be implemented for indicator machine.In this paper, the main work and achievements are done as follows:(1) The single cylinder hydraulic shock absorbers are tested on the indicator machine. Different typical defect types are discussed, unified pretreatment is made. Then invariant moments are selected as the feature extraction according to the diagrams characteristic. Sample library is established.(2) The statistical learning and support vector machine theory are discussed. The VC dimension, boundary of generation, structural risk minimization, linear and non-linear support vector machine, kernel functions and some multiple classification algorithms are included.(3) LibSVM toolbox is studied to realize the process of classification, mainly about data preparation, data training, modeling and identification. On8classes,534samples are tested about different normalization methods, choices of kernel function and optimization of parameters. Under [0,1] normalization, RBF kernel function and cross validation optimized parameters, above97%recognition rate is reached.(4) It is showed by the experiment that the recognition rate is vastly affected by the feature extraction, normalization method, type of kernel function and model parameters.The support vector machine learning theory is feasible to be used in the defective shock absorber indicator diagram auto-recognition. It is expected that indicator machine with this new auto-recognition function can be popularized in the field of shock absorber testing.