基于自适应交互多模型算法的传感器故障诊断

Failure Detection and Diagnosis of Sensor Using Adaptive Interacting Multiple Model Algorithm

在标准的交互多模型算法中,模型切换概率和模型噪声方差都假定为先验信息,但是系统的模型信息一般隐含在当前的量测信息中,因此本文提出一种在线辨识模型切换概率和噪声方差的自适应交互多模型算法,在建立传感器的全局故障和局部故障模型的基础上,对自主水下航行器多个传感器的局部故障与全局故障进行了诊断。在仿真过程中发现,与一般的多模型自适应估计方法(MMAE)相比,此方法算法简单,能够更加准确可靠地诊断故障。

In the IMM algorithm, the Markov transition probabilities and the variances of model-conditional process noise are priori information. In fact, the information is hidden in the current measurement, thus this paper presents the fault detection and isolation method using AIMM algorithm to identify above parameters. Based on the sensor failure models of AUV control system, a simulation example of an AUV control system is given for illustration. Compared with standard multiple model method, it is more accurate and easy to realize.