Pattern Recognition Techniques Applied to Diagnostics

The application of Bayes' Procedures to effect statistically optimum diagnostic decisions is described. The thesis is advanced that pattern recognition techniques can be used to reduce processing complexity and thus permit implementation of Bayes' Procedures in practical field portable equipment.

This thesis is under evaluation in a test program in which journal bearing and wrist pin defects are detected from vibration measurements. Processing of vibration data by a computer programmed Pattern Recognition Technique resulted in design of “predetection” filters which can simplify the equipment required to detect the faults in real time.