VIT-GADG: A Generative Domain-Generalized Framework for Chillers Fault Diagnosis Under Unseen Working Conditions

The extreme unbalance of training samples among different working conditions caused by complex and variable external environments makes the fault diagnosis of a chiller based on domain adaptation (DA) poor performance. Although recently emerging fault diagnosis methods based on domain generalization (DG) can learn domain-invariant knowledge from multiple source domains and can generalize to unseen target domains, these methods still rely on multiple similar source domain data and rarely consider how to enhance the ability to distinguish the joint distribution of features extracted from the source domain samples. To address these problems, a generative domain-generalized framework for chillers fault diagnosis, namely, vision transformer generative adversarial DG (VIT-GADG), is proposed. In VIT-GADG, a novel VIT domain generation network (VIT-DGN) is first designed to reduce DG’s dependence on multisource domain data by improving the diversity of the distribution of the source domain samples. Then, a new adversarial DG network called VIT conditional adversarial DG network (VIT-CADGN) is designed to extract domain-invariant knowledge from the source domain and latent domains that can be generalized to unseen target domains. Specifically, the VIT module can effectively extract the global statistical feature of input samples, which is conducive to the identification of joint distribution. Simultaneously, the collaborative conditional domain discrimination strategy is introduced to improve the distribution discrimination ability of the extracted global statistical features while simultaneously aligning its conditional distribution. In addition, a personalized adaptive weight strategy is proposed to improve the performance of VIT-CADGN. Finally, the comprehensive case study shows that VIT-GADG has a satisfactory ability to extract invariant features, which improves the diagnosis accuracy in the unseen target domain.

I. Introduction

A large amount of the energy used in a building is used for heating, ventilation, and air conditioning (HVAC), which is a sustainable use of total energy consumption. Besides, this share is steadily rising as a result of climate change. In the tropics, a major share of HVAC energy is used by chillers, with the compressor using a significant amount of power. The chiller failure will greatly increase the building’s energy consumption. Therefore, chiller fault diagnosis is an effective method to lower the utilization of energy in buildings. The classical fault diagnosis methods of chillers are divided into two main types: rule-based methods and model-based methods. These model-based and rule-based approaches heavily rely on the subjective prior knowledge of the chiller on the fault mechanism. With the increasing complexity of equipment, it is increasingly difficult for fault diagnosis methods based on mechanism analysis to achieve high performance. Recently, deep learning (DL), with its powerful data processing and feature extraction capabilities, has been widely put into the fault diagnosis of complex equipment and effectively enhanced the diagnostic performance [1], [2], [3], [4], [5], [6], [7], [8], [9], [10].