锻造模具的随机疲劳损伤分析

塑性应变控制的低周疲劳失效是锻造模具的主要失效形式，由于受到多种因素的影响，其损伤过程呈现随机性。应用有限元与BP神经网络，建立了一种基于损伤累积理论的锻造模具的随机疲劳损伤分析模型。首先用有限元对锻造过程中模具内的场变量进行分析，由计算结果找到模具的危险部位，认为危险部位失效时的寿命即为模具寿命，并计算出确定性损伤；再考虑模具材料及实际工况的随机性，应用BP神经网络对模具的损伤进行模拟；根据损伤的累积效应，得出考虑随机因素作用下的模具疲劳寿命。以锻造齿轮模作为分析对象，得到不同工况下模具疲劳寿命的频数分布，其分布规律基本服从Weiubull分布。另外，还分析了打击速度对模具寿命的均值和离差的影响及可靠性随使用次数的变化。该 模型可用来对模具进行寿命预测和可靠性分析。

STOCHASTIC FATIGUE DAMAGE ANALYSIS OF THE FORGING DIE

Low cycle fatigue controlled by plastic strain is a main failure form of the forging die. Because affected by many factors, the damage process shows a randomicity. In this paper a stochastic fatigue damage analysis model is put forward based on the damage accumulation theory, in which the finite element method (FEM) and BP neural network are adopted. The distribution of stress, strain and temperature in the die during forging is analyzed with FEM. According to the results the location liable to failure can be found, the fatigue life of which is regarded as that of the forging die. After getting the definite damage, a three layered BP neural network is trained to simulate the damage process taking the variation of the die material's fatigue properties and the load cases into account. Based on the damage accumulation effect, the fatigue lives under different load cases can be gotten. As an example, a gear forging die is analyzed with this model. The distribution of failure frequency and the change of reliability with stroke times are gotten. The distribution law is approximate Weibull distribution. Next the effect of load cases on the distribution features, the mean value and the variation, is analyzed. Based on the results the reliability or the failure probability of the die can be derived, which gives a direction in reliability design and analysis of the forging die. The model can be used for life prediction and reliability analysis of the forging die.