Innovative soft fault diagnosis method for dual-redundancy sensors

Abstract

Purpose – This paper aims to conduct soft fault diagnosis of dual-redundancy sensors. An innovative fault diagnosis method, which combines a tracking differentiator and a sequential probability ratio test, is proposed. Design/methodology/approach – First, two tracking differentiators are used to track and predict the two original signals, and determine their residuals. These residuals are used to calculate one quadratic residual. Then, a sequential probability ratio test is carried out on this quadratic residual to obtain log-likelihood ratio. A fault can be detected through comparing the log-likelihood ratio value with the threshold value. Finally, analyses of the difference in the residuals, which locates the fault, and of the difference in the original signals, which reveals the fault level and type, are completed successively. Findings – Results from experimentation show that this method can realise soft fault diagnosis for dual-redundancy sensors. Originality/value – The method proposed in the paper gives a new idea to study hybrid redundancy. The method provides a new application mode for tracking differentiators and sequential probability ratio test. The method can be used in robots, such as unmanned aerial vehicles and unmanned ground vehicles, to improve their fault tolerance. It can also be applied to the key parts of industrial production lines to decrease financial losses caused by sensor faults.