Detecção e diagnóstico remoto de falhas baseado em aprendizado de máquina para equipamentos de uso doméstico

Abstract

The conventional household equipment we use on a day-to-day basis is practically not monitored for real-time faults and defects. In this work, an approach will be presented to the application of machine learning with the use of selected data of the equipment when in operation, and of data references extracted from the datasheets. In order to analyze, compare and evaluate differences in data sets, the fault detection and diagnosis device was developed to classify the symptoms that may represent defects in the equipment in real time. The occurrences of these equipment failures are traditionally identified by the users themselves when the expected performance does not occur. With the use of a microprocessed board connected to the electronic sensors installed at strategic points in the equipment, the data comparison step is started, the data collected are transmitted to the server, which through the Machine Learning algorithm performs the tasks for identification of the Detected failures. Real-time monitoring of the behavior of electrical and physical magnitudes of conventional household equipment is aimed at monitoring functional behavior and informing the user of any faults using local or Internet resources.