Research On The Fatigue Life Prediction Method Based On Support Vector Machine

As the mechanical components to the large-scale,complicated and high temperature,high speed using environment the development direction,in some fields of high reliability and high safety requirements,components have high cost,small batch and complex failure modes and failure principle,which makes life prediction attract widespread attention at home and abroad and be the research technical problems.In this paper,through the analysis of previous fatigue life prediction model based on support vector machine(SVM)in engineering applications,it exists the shortage of the prediction precision is not stable,and not good at parameters optimization effect,etc,to propose hybrid kernel function of support vector machine based on particle swarm optimization of fatigue life prediction model.The main research content of this article includes the following several aspects:(1)Combining the theory of support vector machine,to research the performances of the radial basis kernel function and the polynomial kernel function respectively with debug parameter method.By comparing the model prediction results with the experimental data,to verify the selection of kernel function and its related parameters is a very crucial step to establish support vector machine model.(2)For the defects of parameter optimization effect and poor efficiency,to introduce the particle swarm intelligence algorithm,to set up support vector machine model,to study the optimization parameters of support vector machine by particle swarm algorithm.(3)Aiming at the limitation to the single kernel function,to combine advantages with the radial basis kernel function and polynomial kernel function,to propose a mixed kernel function,and to program hybrid kernel matrix operations directly,to research the overall and local of the mixed kernel function.(4)To set up the grid search method optimization hybrid kernel function support vector machine model,particle swarm optimization hybrid kernel function support vector machine model,particle swarm optimization radial basis kernel function support vector machine model and particle swarm optimization polynomial kernel function support vector machine model respectively,to Compare and analyse the fatigue test data with the model prediction results,Particle swarm optimization(pso)algorithm has obvious advantage of optimization effect,the efficiency of convergence fast and high precision,Hybrid kernel function of support vector machine is more superior than single kernel function in the global and local.In this paper,the support vector machine model based on particle swarm optimization of mixed kernel function has faster training speed,better prediction precision,and strong generalization.