Methodology for Multi-parameter Self-validating Sensor Fault Diagnosis

This paper introduced the concept of multi-parameter self-validating sensor,including definition and function model. Fault diagnosis unit plays an important part in incoming the self-validating function model.The multi-parameter self-validating sensor fault diagnosis based on partial least square(PLS) and support vector machines(SVM) was researched.The method with PLS to get the principal components of the sensor's measuring data as feature matrix and SVM to classify the sensor's working status was proposed for sensor fault diagnosis.The PLS method was applied for feature extraction to get the feature matrix which denote all kinds of known working status of multi-parameter self-validating senso,and the feature matrix were encoded.The feature matrix as inputs and feature codes as outputs were proposed for training the SVM classifier to get the optimum parameters.In the fault diagnosis unit,it used the PLS method to acquire the on-line feature matrix of multi-parameter self-validating sensor.The on-line feature matrix was supplied to the trained SVM classifier as inputs to validate the working status of multi-parameter self-validating sensor.If the sensor is healthy,the validated outputs are given directly;otherwise the sensor should auto alarm and recover the outputs in a short period of time.Finally,the applicability and effectiveness of the multi-parameter self-validating sensor fault diagnosis based on partial least square(PLS) and support vector machines(SVM) are illustrated by the sensor's self-validating results.