Application of Anomaly Detection Methods in the Housing and Utility Infrastructure Data

Efficient and timely fault detection is a significant problem due to the intensifying use of modern technological solutions in machine condition monitoring. This work is carried out as part of a project that is aimed at development of software solutions for a housing and utility condition monitoring system. An experimental setup was designed and assembled for the study of basic housing infrastructure elements operating modes. The setup includes electric pumps, power transformers, ventilation and air conditioning systems (HVAC), heaters and electric boilers. Every element is equipped with various sensors. Sensor readings were gathered, processed and analyzed. This dataset was used to fit statistical and probabilistic models such as linear regression and Hidden Markov model in order to classify regular and faulty operating modes of equipment. Nine classes of equipment malfunction were modeled, these models are intended to be used as a theoretical basis for the design of industrial housing and utility condition monitoring systems.