应用BP神经网络和均值一次二阶矩法的连杆频率可靠性分析

Frequency Reliability Analysis of Connecting Rod Using BP Neural Network and Mean First Order Second Moment Method

为研究机械零部件尺寸参数及材料性能参数对机械零部件频率失效的影响,对NC450型单缸发动机连杆进行了研究,应用多平台集成设计对其进行频率可靠性分析。首先,建立ANSYS有限元模型进行模态分析,得到了连杆前四阶固有频率值及振型;其次,通过ANSYS与Isight集成设计模态联合仿真及DOE样本实验,获得用于频率可靠性分析的样本数据,并经DOE样本分析确定了连杆的最低阶固有频率的敏感设计变量;然后,由MATLAB编制BP神经网络程序进行回归分析,拟合连杆最低阶固有频率和敏感设计变量之间的函数关系,建立了基于共振失效原理的频率失效功能函数;最后,结合均值一次二阶矩法有效分析了连杆在稳定工况下的可靠性,为机械零部件的可靠性分析及设计提供新的方法。

In order to study the influence of mechanical component size parameters and material performance parameters on the frequency failure of mechanical components,the NC450 single-cylinder engine connecting rod is taken as the research object,and the multi-platform integrated design is applied to analyze the frequency reliability.Firstly,the finite element model for ANSYS analysis is established for modal analysis,and the first four natural frequency values and vibration modes of the connecting rod are obtained.Secondly,through the joint simulation of ANSYS and Isight integrated design mode and DOE sample experiment,the sample data for frequency reliability analysis are obtained,and the sensitive design variables of the lowest natural frequency of the connecting rod are determined by DOE sample analysis.Then,the BP neural network program is compiled by MATLAB for regression analysis,and the functional relationship between the lowest natural frequency of the connecting rod and the sensitive design variables is fitted,and the frequency failure function based on the resonance failure principle is established.Finally,the reliability of the connecting rod under stable working conditions is effectively analyzed by the mean first-order second-moment method,which provides a new method for reliability analysis and design of mechanical parts.