Fault diagnosis system for a robot manipulator through neuro fuzzy approach

The adoption of an efficient online Fault Detection and Isolation (FDI) tool is becoming utmost importance for robots, especially for those operating in remote or hazardous environments, where a high degree of safety and self-diagnostics capabilities are required. This saves time and cost in repairing the robot. A number of researchers have proposed Fault Diagnosis (FD) architectures for robotic manipulators using the model-based analytical and redundancy approach. One of the main issues in the design of fault detection system is to model the rigid link robotic manipulators with modelling uncertainties. In this paper, a new approach neuro fuzzy-based FDI for robot manipulators is discussed. A learning architecture, with neural network as online approximates the off-nominal system behaviour, which is used for monitoring the robotic system for the faults. This generates the residual by comparing the actual output from robot. Fuzzy inference system is applied to identify and isolate the faults which provide the adaptive threshold under the varying conditions. The concepts discussed were validated through simulation study using a Scorbot-ER 5 plus manipulator to illustrate the ability of the neuro fuzzy-based FD scheme to detect and isolate fault. Although a robotic manipulator is used to illustrate the effectiveness of this approach, it can also be applied to the non-linear systems.