赵明轩,桑建兵,丛继坤,等.基于径向基函数神经网络重载发动机曲轴的可靠性分析[J].内燃机工程,2025,46(5):100-108.
基于径向基函数神经网络重载发动机曲轴的可靠性分析
Reliability Analysis of Crankshaft in Heavy-Duty Engine Based on Radial Basis Function Neural Networks
DOI:10.13949/j.cnki.nrjgc.2025.05.011
关键词:曲轴  可靠性分析  灰色关联度分析  径向基函数神经网络  粒子群优化算法
Key Words:crankshaft  reliability analysis  grey relational analysis(GRA)  radialbasis function neural network(RBFNN)  particle swarm optimization(PSO)
基金项目:国防科技重点实验室项目
作者单位E-mail
赵明轩* 河北工业大学 机械工程学院天津300401 202331205106@stu.hebut.edu.cn 
桑建兵* 河北工业大学 机械工程学院天津300401 sangjianbing@hebut.edu.cn 
丛继坤* 河北工业大学 工业技术研究院天津 300401 cjk@hebut.edu.cn 
钟星宇 河北工业大学 机械工程学院天津300401 202421201019@stu.hebut.edu.cn 
李长远 河北工业大学 机械工程学院天津300401 3129535573@qq.com 
摘要点击次数: 426
全文下载次数: 183
摘要:针对传统可靠性分析方法计算成本高且精度不高等问题,结合灰色关联度分析(grey relational analysis, GRA)、径向基函数神经网络(radial basis function neural network, RBFNN)及粒子群优化算法(particle swarm optimization, PSO),提出了一种针对重载发动机曲轴的可靠性分析方法。首先,根据曲轴的动力学分析和点火做功状态确定了其危险工况并利用有限元软件ANSYS Workbench建立了静力学计算模型。其次,结合曲轴的几何参数和总体结构确定了影响最大Mises应力的不确定性因素,并对其进行灰色关联度分析筛选出径向基函数神经网络的输入参数。最后,依据不确定性参数的分布情况使用最优拉丁超立方(optimal Latin hypercube sampling, OLHS)进行采样,根据第四强度理论确定曲轴的破坏准则后,引入粒子群优化算法,结合径向基函数神经网络和蒙特卡洛方法(RBFNN–Monte Carlo, RBFNN–MC)预测了曲轴的可靠度和失效概率。研究结果表明,RBFNN–MC方法与传统可靠性分析方法相比,在保证高精度的前提下具有更高的效率和更好的鲁棒性。
Abstract:In order to address the high computational cost and low accuracy issues of traditional reliability analysis methods, a reliability analysis method for heavy-duty engine crankshafts was proposed, combining grey relational analysis (GRA), radial basis function neural network (RBFNN) and particle swarm optimization (PSO). The dangerous working conditions of the crankshaft were identified based on dynamic analysis and firing work conditions, and a static calculation model was established using the finite element software ANSYS Workbench. The uncertainty factors affecting the maximum Mises stress were identified based on the crankshaft’s geometric parameters and overall structure, and GRA was performed to determine the input parameters for RBFNN. Finally, optimal Latin hypercube sampling (OLHS) was performed based on the distribution of uncertainty parameters. The failure criterion of the crankshaft was determined based on the fourth strength theory, and the reliability and failure probability of the crankshaft were predicted by the RBFNN–Monte Carlo (RBFNN-MC) method with the introduction of the PSO. Results show that compared with traditional reliability analysis methods, the RBFNN-MC method offers higher efficiency and better robustness while ensuring high accuracy.
查看全文  HTML   查看/发表评论