Time Series Analysis in Compressor-Based Machines: A Survey

TL;DR

This survey systematically maps fault detection, fault prediction, forecasting, and change point detection applied to multivariate time series from compressor-based machines (refrigerators, HVAC, heat pumps, chillers). It synthesizes 44 works (2016–2023) identified via PRISMA on Scopus, comparing ML and DL approaches, datasets, and evaluation practices, and highlighting critical gaps such as scarce public real-data benchmarks and inconsistent cross-study comparisons. Key findings include the prevalent use of ASHRAE 1043-RP in FD studies, the growing but still uneven adoption of DL architectures (e.g., 1D-CNN, LSTM, DRNN), and the need for generalizable, interpretable models. The paper outlines promising directions, including multi-modal data fusion, physics-informed models (PINN), graph neural networks (GNN), foundation models, transfer learning, and the creation of open benchmarks to enable fair comparisons and accelerate practical deployment in industrial and residential settings.

Abstract

In both industrial and residential contexts, compressor-based machines, such as refrigerators, HVAC systems, heat pumps and chillers, are essential to fulfil production and consumers' needs. The diffusion of sensors and IoT connectivity supports the development of monitoring systems that can detect and predict faults, identify behavioural shifts and forecast the operational status of machines and their components. The focus of this paper is to survey the recent research on such tasks as FD, FP, Forecasting and CPD applied to multivariate time series characterizing the operations of compressor-based machines. These tasks play a critical role in improving the efficiency and longevity of machines by minimizing downtime and maintenance costs and improving the energy efficiency. Specifically, FD detects and diagnoses faults, FP predicts such occurrences, forecasting anticipates the future value of characteristic variables of machines and CPD identifies significant variations in the behaviour of the appliances, such as a change in the working regime. We identify and classify the approaches to the tasks mentioned above, compare the algorithms employed, highlight the gaps in the current status of the art and discuss the most promising future research directions in the field.

Figures (6)

• Figure 1: The temporal distribution of works across the 4 analyzed tasks. Fig. 1(a) shows the number of works by year and task. Fig.1(b) shows the percentage of works (with respect to each task) by year and task
• Figure 2: PRISMA flow diagram of the systematic review
• Figure 3: Visualization of pairwise comparisons of FD algorithms. The x-axis reports the dominant algorithms, while the y-axis reports the dominated algorithms. Each cell reports the number of times that a dominant algorithm is found to outperform a dominated algorithm in the surveyed literature
• Figure 4: Visualization of pairwise comparisons of FP algorithms. The x-axis reports the dominant algorithms, while the y-axis reports the dominated algorithms. Each cell reports the number of times that a dominant algorithm is found to outperform a dominated algorithm in the surveyed literature
• Figure 5: Visualization of pairwise comparisons of forecasting algorithms. The x-axis reports the dominant algorithms, while the y-axis reports the dominated algorithms. Each cell reports the number of times that a dominant algorithm is found to outperform a dominated algorithm in the surveyed literature