ε机在曲轴疲劳裂纹扩展预测中的应用

Prediction for Crankshaft's Fatigue Crack Growth Using ε-machine

机械构件的疲劳裂纹的扩展行为往往表现出阶段性,疲劳失效作为一个系统耗散过程,在不同的扩展阶段内必然隐含着一些内部模式的演变,如果建立一种相对初始状态模式的异常测度对这些演变模式进行动态、实时的识别并进行预测,则可为预防破坏性疲劳失效事故的发生和科学合理地制定维修计划提供参考依据。本文对曲轴疲劳试验过程中谐振台架的振动加速度进行了测录,并采用基于复杂系统隐含模式发现的异常检测算法(ε机)对这一监测信号进行分析,计算了其在各个时间段中的异常度曲线。与曲轴疲劳裂纹扩展速率曲线进行对比,分析了ε机对裂纹扩展模式的预测能力。

Propagations of mechanical components＇ fatigue cracks are usually staged. As a process of system dissipation, in each stage of the crack propagation, a slightly shifting of pattern will occur. If the abnormal measurement, which is correspondent to the original state, derived from these features can identify shifting patterns and predict the coming behavior features in real time dynamically, the referential basis is to be provided for preventing destructive fatigue failures and establishing reasonable maintenance and repair schedules. As one of the anomaly detection algorithms based on hidden pattern, ε-machine was utilized to process the vibration acceleration data that were recorded in the fatigue test of engine crankshafts, in order to make identification and prediction for the fatigue crack growth behavior patterns. By comparison with the test results of crack growth rate, the prediction capability of the method for crack growth behavior pattern was discussed.