Abstract: To predict the dilution rate of 15-5PH alloy powder laser cladding layer and improve the performance of the cladding layer,a 15-5PH laser cladding layer dilution rate prediction method based on Particle Swarm Optimization(PSO) optimization Support Vector Regression(SVR) was proposed.15-5PH was used as cladding material and 45 steel was used as substrate in laser cladding experiment.The SVR model between the process parameters and dilution rate of 15-5PH cladding layer was established based on the experimental results.The kernel function of SVR model was optimized and PSO was used to optimize the parameters of SVR model.The results showed that PSO-SVR model had the best predicted performance when the Gaussian kernel function was selected;the PSO-SVR model had higher accuracy in predicting the dilution rate of 15-5PH cladding by comparing with SVR model and Back Propagation Neural Network model.The coefficient of determination of the model was 0.9647,the mean square error was 0.0003,and the mean relative error was 3.6%.

Key words: laser cladding, 15-5PH allog powder, dilution rate, support vector regression, particle swarm optimization algorithm

摘要： 为了预测15-5PH合金粉末激光熔覆层稀释率进而改善熔覆层性能,提出一种基于粒子群算法(PSO)优化支持向量回归(SVR)的15-5PH激光熔覆层稀释率预测方法。以15-5PH为熔覆材料,45钢为基体进行激光熔覆实验;基于实验结果建立了工艺参数与15-5PH熔覆层稀释率间的SVR模型;优选SVR模型的核函数并运用PSO优化SVR模型的参数。结果表明:PSO-SVR模型选择高斯核函数时预测性能最好;与SVR模型和BP神经网络模型的预测结果相比,PSO-SVR模型对15-5PH熔覆层稀释率的预测结果更准确,模型的决定系数为0.9647,均方误差为0.0003,平均相对误差为3.6%。

关键词: 激光熔覆, 15-5PH合金粉末, 稀释率, 支持向量回归, 粒子群算法