Development of an Expert System Focused on Improving the Supply Chain by Increasing the Availability of Equipment

Abstract

The maintenance area is a transcendental link within the supply chain since it is the unit in charge of maintaining the equipment destined to carry out various organizational tasks in perfect working order. The correct maintenance management through the supply chain allows one to fulfill a primary objective, to deliver the required quantity of goods on time at the minimum cost. Within the automotive industry, the supply chain aims to be as efficient as possible, so innovation and developing new technologies are opportunities for improvement. Among the various motors, the three-phase induction motor is the most used, and, like any industrial device, it requires diagnosis and maintenance to continue operating correctly. There are various methodologies to diagnose faults in electric motors; however, their capacity is limited since they can only analyze and detect one type of fault at a time. In the present work, the development of an expert system capable of recognizing three different fault patterns in electric motors is shown; the deterioration patterns involved in the study are severe damage or fatigue in the bearings and short circuits in the winding. The architecture of this system is composed of two artificial neural networks capable of classifying the types of faults in the electric motor, the first artificial neural network determines whether the motor is in good condition or not, and the second artificial neural network is activated when detecting an inconvenience in the motor to give rise to the fault diagnosis. The developed expert system was implemented in an industrial process and reached an average accuracy of 99%, which is why it is validated as a reliable method to diagnose the correct operation of electric motors and, therefore, improve the availability of said motors in addition to other teams with which they relate. The scientific contribution is based on the methodological proposal to achieve a correct diagnosis of the physical state of electric motors from artificial neural networks, which brings, consequently, an increase in manufacturing productivity and greater efficiency in the supply chain of organizations based on the accurate monitoring of the equipment.