New and Improved Classifiers for Fault Diagnosis

Pattern classifiers are useful for many commercial and military applications such as target recognition, fault diagnostics, pattern recognition, etc. The main goal of this research is to improve the classification performance of existing classifiers. Our proposed architecture consists of two major parts. One is to improve the individual classification performance of each classifier. We propose to apply the latest development in classification theory known as Support Vector Machines (SVM) to improve the accuracy of classifiers. SVM has many attractive advantages. Recent applications include automatic target recognition, breast cancer prognosis, particle identification in high energy physics, etc. Another is to use a robust fusion algorithm called Dempster-Shafer theory to fuse the decisions from different classifiers so that the overall classification performance can be further improved. During Phase II we will further refine our techniques and take on more challenging problems in chemical vapor concentration estimation and mixture component estimation. We will build a real-time version of our classifier for the e-nose application. We will also work with Boeing engineers to prototype a solenoid valve health monitoring system. Besides, we will extend these applications to a number of additional shuttle components and subsystems, and/or other NASA systems of interest.