Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
559143 | Mechanical Systems and Signal Processing | 2016 | 15 Pages |
•Introduced a distributed sensor fault detection framework for non-linearity faults.•Illustrated nonlinear sensor fault detection is equivalent to the LER problem.•Presented a low complexity algorithm suitable for embedding in wireless sensors.•Evaluated the performance of the algorithm extensively through simulations.
Wireless sensors operating in harsh environments have the potential to be error-prone. This paper presents a distributive model-based diagnosis algorithm that identifies nonlinear sensor faults. The diagnosis algorithm has advantages over existing fault diagnosis methods such as centralized model-based and distributive model-free methods. An algorithm is presented for detecting common non-linearity faults without using reference sensors. The study introduces a model-based fault diagnosis framework that is implemented within a pair of wireless sensors. The detection of sensor nonlinearities is shown to be equivalent to solving the largest empty rectangle (LER) problem, given a set of features extracted from an analysis of sensor outputs. A low-complexity algorithm that gives an approximate solution to the LER problem is proposed for embedment in resource constrained wireless sensors. By solving the LER problem, sensors corrupted by non-linearity faults can be isolated and identified. Extensive analysis evaluates the performance of the proposed algorithm through simulation.