A fully distributed voting strategy for AHU fault detection and diagnosis based on a decentralized structure

Air handling units (AHUs) are essential for the regulation and circulation of indoor air in modern buildings and for providing a comfortable environment. However, the performance of AHUs can be easily affected by various sensor faults during long-term operation. To precisely detect and diagnose AHU faults, many researchers have applied model-based and data-driven approaches based on existing centralized architectures. However, such methods require comprehensive knowledge of physical constraints or sufficient labeled data and thus are impractical in real scenarios. Inspired by the merits of emerging decentralized systems, we implement insect intelligent building (I 2 B), a fully decentralized architecture, for intelligent buildings to improve building operation performance. In addition, to ensure the AHU efficiency, we optimize the process of AHU fault detection and diagnosis and propose a fully distributed voting strategy using only adjacent computing process nodes (CPNs) in a decentralized architecture. Extensive experiments were conducted to analyze the performance of the proposed method in different real scenarios. Based on a comparison with related state-of-the-art methods, the experimental results highlight the superiority of our method, which provides high accuracy, plug-in capability and strong adaptability.