Machine learning application for refrigeration showcase fault discrimination

Open refrigeration showcases are commonly utilized equipment in super markets and convenience stores to maintain the temperature and quality of foods and drinks. Often set in a broader refrigeration arrangement, where a number of showcases and outdoor condensers are connected, it is a system that remains susceptible to fault events, which can lead to financial losses for stores. Therefore, faults and early abnormal behaviors that can lead to future problems should be identified. To classify events as in-control of faulty, samples or patterns for both types of events are needed, however, it is often the case in practical industrial applications where only the in-control type of data is available. This paper assesses the applicability of the machine learning approaches for supervised and unsupervised learning in the task of identifying unusual behavior in real showcase data.