修正指数遗忘RLS算法及其在故障诊断上的应用

Self-Adjustable Exponential Forgetting Recursive Least Square Algorithm for Fault Diagnosis

汽车电子系统在线故障诊断的算法执行效率、动态跟踪速度和稳态估计精度是检测突变故障参数的关键指标。针对基于参数估计故障诊断中常用的递推最小二乘RLS算法存在的典型问题,提出了一种在线修正遗忘因子的方法。理论分析和仿真结果均表明,修正后的方法能有效解决一般递推算法的“数据饱和”问题,与通常的遗忘算法和滑动数据窗法相比,表现出了明显的优越性。为进一步的实车故障诊断提供了更加有效的理论根据。

To detect abrupt parameter effectively, rapidly and accurately is quite crucial for online fault diagnosis based on parameter estimarion methods. An ameliorative adaptive algorithm named SAEFRLS （Self- Adjustable Exponential Forgetting Reeursive Least Square） algorithm is suggested here to avoid problems that traditional RLS （Recursive l.east Square） algorithm can not solve. In the application of fault diagnosis, SAEFRLS algorithm is not only capable of solving ＂ data saturation＂ problem, but also attesting to be of significant advantage o ver traditional Exponential Forgetting RLS and Sliding- Window BI.S algorithm, which provides theoretical basis to real-time fault diagnosis more effectively, rapidly and accurately.