Machine learning classification for pump fault and failure detection

This project aims to develop an automated ability capable of detecting faults with pumps. This is referred as “Automated Fault Detection and Diagnosis” (AFDD). Equipment performance begins to worsen throughout time due to various reasons, where these reasons are referred to as “faults”. Generally, there is an understanding of the various faults and causes for equipment failure, but the challenge arises in development of a tool capable of accurately and automatically detecting these issues. An AFDD tool would receive sensor data from a pump and use various algorithms to 1) “Detect” a fault and 2) “Diagnose” the exact fault. The expected outcome of this project is an algorithm that can read data from installed equipment and use it to determine whether or not the equipment is (a) operating correctly, (b) operating with a problem (fault), or (c) in danger of failing, and provide both the building owners and Armstrong service team with the necessary information to fix it.

Faculty Supervisor:

Jennifer McArthur

Student:

Rony Shohet

Partner:

Armstrong Fluid Technology

Discipline:

Architecture and design

Sector:

Manufacturing

University:

Ryerson University