RELIABILITY AND SENSITIVITY ANALYSIS OF A CONDITION MONITORING TECHNIQUE

Abstract

A condition monitoring technique based on vibration measurements and evaluation by means of neural networks has previously been investigated. Preliminary results were promising. In this work, the reliability and the sensitivity of the method are discussed. This study has used a simulation approach to study the sensitivity. An experimentally verified simulation model which yields the internal forces in the pump has been used in combination with experimentally found transfer functions to find the vibration on the surface of the pump housing. Different designs of the valve plate have been studied to simulate an increased internal leakage in the pump. For the evaluation of simulation or measurement results, neural networks are used.