基于低复杂度最大空闲矩形的非线性传感器故障诊断方法

针对现存的很多传感器故障诊断方法假设前提多以及复杂度高的问题,提出一种分布式诊断方法来识别无线传感器网络(WSN)中的非线性故障。首先,对局部传感器的输出值进行分析,得到一系列特征值；然后,在交叉误差函数的基础上,将传感器非线性故障诊断等效为最大空闲矩形(LER)问题。并使用提出的低复杂度最大空闲矩形算法予以解决；最后,通过定义一个阈值来诊断有故障的传感器,且不需要使用参考传感器就可以检测一般非线性故障。仿真实验使用了双音谐波信号激励和白噪声信号激励,比较了双线性和指数非线性两种情况下的性能。相比集中式故障诊断方法,提出的算法节省了大量数据传输功率,且获得了非线性模型正常区域边界的准确值。相比最优LER算法,提出的低复杂度LER算法检测性能与之相似,但复杂度更低。

Method of nonlinear sensor fault diagnosis based on low computational largest empty rectangle

As the multiple assumptions of many existing sensor fault diagnosis methods and the high computation, a distributed diagnosis method is proposed to identify nonlinear faults in wireless sensor networks (WSN). Firstly, the output value of the local sensors is analyzed, and a series of eigenvalues are obtained. Then, based on the cross error function, the nonlinear fault diagnosis of the sensor is equivalent to the problem of largest empty rectangle (LER). And low computational largest empty rectangle algorithm is adapted to solute the problem. Finally, sensor fault diagnosis is through the definition of a threshold, and the proposed method does not need reference sensors to detect general nonlinear fault. The two-tone signal excitation and white noise excitation are used in the simulation experiments, and bilinear and exponential non-linearity are considered. Compared with the centralized fault diagnosis methods, proposed method not only saves a large amount of data transmission power, but also obtains the accurate value of the normal region boundary of the nonlinear model. Compared with the optimal LER algorithm, the performance of the proposed low computational LER is similar, but the complexity is lower.