Fault Detection in Supermarket Refrigeration Systems Using Machine Learning

Sathiyaseelan, Gowsikan ; Martin Garcia, Antonio Javier

4. term

Control and Automation, Master

2021

2021-06-03

109 pages

Supermarket refrigeration systems are responsible for maintaining a constant temperature to preserve the quality of the goods. Failure in these appliances can lead to economical losses and food spoilage. However, manually monitoring the systems is inconvenient and expensive. The objective of this thesis is therefore to use fault detection for determining the status of the systems and predict the failure before changes occur in the system. To solve this problem, a Support Vector Machine classifier (SVM) has been developed, using data from a simulation, to predict the state of the system in real-time. Prior to the classification, the data has been preprocessed by reducing its dimensions to facilitate the work of the classifier using Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). Several tests have been carried out to verify the performance of the developed methods under different initial conditions and faults in the system. It can be concluded from the results that LDA is the superior method for dimensionality reduction in this application and SVM is a feasible solution for automatic fault detection having close or equal to 100 % accuracy in most tests.

Refrigeration ; Principal Component Analysis ; Linear Discriminant Analysis ; Support Vector Machine ; Machine Learning ; Classification

Download
View record in AAU Student Projects