Design And Fault Diagnosis Of Multi Sensor Redundant System For Mobile Robot

With the rapid development of relevant technologies,mobile robots have been widely used in the military,industrial and other fields.In order to enhance the reliability of sensors in mobile robot,the paper designed a redundant system consisting of six single degree of freedom gyroscopes,and the fault detection of the sensors in the system was studied by analyzing two kinds of traditional detection algorithms.To overcome the limitations of the algorithms,the improved methods were proposed,the experimental platform was built to verify the validity of algorithms.Firstly,we designed an inertial sensor redundant system of mobile system based on the sensor redundancy allocation problem which included sensors geometry structure design,circuit design of single axis Z gyroscope module,and embedded sensors data processing circuit.For convenience,we also designed a liquid crystal display module to show the faults detection effect.Secondly,in order to effectively detect the fault,we analyzed the fault mechanism and mathematical model of inertial sensor,and based on this,we analyzed traditional principal component analysis(PCA)fault detection method in redundant sensors system,To solve the low accuracy existing in the fault detection with non-Gauss sensors noise by using traditional PCA method,the PCA-ICA fault detection method was studied.The algorithm firstly used the independent component analysis to remove the non-Gauss noise and then used the principal component analysis(PCA).The effectiveness of two methods was verified by setting the non-Gauss noise and deviation fault in the simulation.Thirdly,there is a strong correlation between the sensors noise which effect the fault detection by using the traditional PCA method.To overcome this,we used the Mahalanobis distance applied in the fault detection.And on this basis,the improved Mahalanobis distance method was put forward to solve the problem that iterative weighted least squares estimation residual effects the real-time performance.This algorithm can save the time of residual estimation and the parity residuals obtained from the data preprocessed by the parity vector can be used for fault detection.Furthermore,to solve the unsteadiness in the calculation of the Mahalanobis distance method,the robust Mahalanobis distance aiming to detect the outliers and fault was studied.Finally,the experimental platform was built to verify the proposed algorithm.The experimental results show that,the accuracy of the Mahalanobis distance is higher than the traditional PCA method,the improved Mahalanobis distance method could improve the real-time efficiency in the fault diagnosis and the effectiveness using robust Mahalanobis distance is better than the traditional one in the detection of outliers and fault.