Multi-fault identification of a hydraulic lifting system via partitioning of unstructured residuals

Abstract

The structured residual monitoring approach attributed to binary fault signatures is most common for fault isolation in model-based fault detection and isolation. When residuals are unstructured due to lesser number of sensors, their corresponding parameters values are estimated numerically through optimization. Thereafter, the estimated values of the parameters are compared with their corresponding nominal values for fault isolation. This article presents a state of the art fault localization scheme by partitioning of a typical array of unstructured residuals where a number of parameters commonly affected the unstructured as well as structured residuals. A reduced set of parameters that are necessary to be estimated by optimization are found and are estimated through least-square optimization for fault isolation. The problem space (dimension of the objective function) in optimization thus gets reduced and optimization performance shows improvement. A high-pressure hydraulic circuit used in the dumper and several other lifting systems is used as a simulation example for the proposed approach.