摘要
机械结构在长期的存储中,疲劳裂纹的变化受到多种环境因素的影响,针对目前疲劳裂纹预测准确率低的问题,提出一种基于最小二乘支持向量机方法(LSSVM)来预测机械结构的疲劳裂纹长度,通过改进的粒子群优化算法对LSSVM进行参数优化。改进的粒子群参数优化算法采用二次型惯性权重递减策略,使粒子群优化算法的优化过程更接近实际的非线性和高复杂过程。经仿真实验验证,结果表明,基于改进的粒子群参数优化的最小二乘支持向量机(PSO_LSSVM)对于机械结构的疲劳裂纹长度预测优于传统方法,收敛速度快,预测准确。
For mechanical structure in long-term storage, fatigue crack changes were influenced by a variety of environmental factors. Fatigue crack prediction accuracy was low. This paper proposed a method based on least squares support vector machine method to predict the fatigue crack length of mechanical structure. By improved particle swarm optimization algorithm optimized the LSSVM parameters. Improved particle swarm optimization algorithm used quadratic decreasing inertia weight strategy, made PSO algorithm optimization process closer to actual high nonlinear and complex process. It has been tested by simulation experiments,the results show that mechanical structure fatigue crack length prediction is better than traditional methods, h has fast convergence speed, high accuracy.
出处
《计算机应用研究》
CSCD
北大核心
2013年第12期3597-3599,3609,共4页
Application Research of Computers
基金
国家自然科学基金中国工程物理研究院联合基金资助项目(NSAF:11176027)
关键词
粒子群参数优化
最小二乘支持向量机
疲劳裂纹
二次型惯性权重递减策略
particle swarm parameter optimization ( PPSO )
least squares support vector machine ( LSSVM )
fatigue crack
quadratic decreasing weight