A novel fault diagnosis and self-calibration method for air-handling units using Bayesian Inference and virtual sensing

The Air-handling unit (AHU) is a critical and energy-consuming component in heating, ventilation, and air conditioning (HVAC) systems. However, sensor and component faults influence the AHU operation and energy consumption. This study proposed a novel fault detection, diagnosis (FDD), and self-calibration method based on the Bayesian Inference coupling with virtual sensing to estimate various faults, including the sensor and component faults. The method uses a small amount of measurement data for estimation. Specifically, virtual sensing was employed to represent variables that were difficult to measure directly. The Markov Chain Monte Carlo (MCMC) algorithm was used to derive the statistical characteristics of fault levels. Subsequently, a detailed criterion was conducted based on the statistical characteristics. A series of fault scenarios in a typical AHU system were designed to comprehensively ver-ify the performance of the proposed method. Accordingly, the proposed method effectively recognized the system operating state and identified the fault positions. Following self-calibration, the decreased deviation rate was high up to 98.0% for most fault scenarios. It is confirmed that the proposed method exhibits effective performance even with a small number of datasets. The results highlight the potentials of the proposed method in fault detection, diagnosis and self-calibration of HVAC systems. (c) 2021 Elsevier B.V. All rights reserved.

Automated summary

This study introduced a novel fault detection, diagnosis, and self-calibration method based on Bayesian inference and virtual sensing to estimate various faults, including sensor and component faults. The method effectively recognized system operating state and identified fault positions, reducing deviation rate by up to 98.0% in most fault scenarios.