APPLICATION OF MACHINE LEARNING FOR MACHINE MONITORING AND DIAGNOSIS

Authors:

• David A. Maluf, Stanford University, Stanford, maluf@db.stanford.edu
• Laeeque K. Daneshmend, Queens University, Canada, lkd1@post.queensu.ca

Abstract:

This paper reports on research to evaluate the application of artificial neural networks to pump condition monitoring. Based on historical velocity vibration measurements, artificial neural networks were developed to assess pump condition. Pumps, in general, behave like most rotating equipment and evolve or deteriorate as a function of a single variable, namely time. Such evolution forms the basis of the current work, where the monitoring of the vibration measurements at different times establishes a pattern in that evolution. Therefore, given adequate data of the initial life, optimum behavior, and final stages of machine failure of the pump, the relative pump health can be determined.

Keywords:

Machine Monitoring, Machine Learning, artificial neural network, rotating equipment.