Causal Artificial Intelligence–Driven Approach for HVAC Preventive Maintenance Explanation

Abstract

Preventive maintenance of heating, ventilation, and air conditioning (HVAC) systems aims to monitor and detect early-stage failures, thereby enhancing thermal comfort satisfaction and reducing energy consumption. However, it is a complex and uncertain factor, and technicians must understand the root cause of the problem to interpret prefailure events accurately. This study addresses this concern using a causal machine learning model, which encodes HVAC system behavior, indoor ambiance, and the outdoor environment based on random variables, with a particular focus on air conditioning units. It models their causal representation concerning prefailures using the structural causal model. It employs d-separation and d-connection to justify cause-and-effect relationships with the model and applies the expectation maximization algorithm to fit model parameters given observational data. The causal model is verified using a do-operator to rationalize the sound explanations (d-separated and d-connected) and employing odds ratio and confidential interval to prove their statistical strength. The model achieves high causal significance, with odds ratios of the submodel lying between 0.3-0.9 and 1.4-6.0 (exclude 1.0). The results indicate that the model and its explanations can encode human-like interpretation and achieve high causal significance aligned with real-world prefailure events. This refers to a reasoning process that is similar to that of humans, where information is included or excluded based on explanations of cause and effect. This confirms that the contribution of this study is suitable for causal-based decision-making systems for HVAC preventive maintenance.

Keywords

• Causal inference
• causal artificial intelligence
• Internet of Things
• building engineering
• thermal comfort