@inproceedings{01KAVA7Y78ZTV0EHZSJ5GJMPZK,
  abstract     = {{This study investigates data-driven fault detection and health prognostics for residential multi-split air-conditioning systems using large-scale operational sensor data. Split-system air conditioners, widely deployed in residential buildings, provide thermal comfort by transferring heat between indoor and outdoor units. Their performance degrades over time due to component wear, environmental conditions, and irregular usage patterns. Traditional maintenance strategies rely on reactive interventions or scheduled servicing, which can lead to avoidable energy consumption, refrigerant leakage, reduced comfort, and premature component failure. Recent advances in sensor availability and machine learning provide an opportunity to transition toward predictive maintenance approaches that detect faults early and estimate remaining useful life (RUL).
The core objective of this research is to monitor the condition of split air-conditioning equipment by constructing health indicators from high-frequency operational telemetry and developing time-series models that characterize degradation. Using sensor-rich data—including temperature, humidity, current, fan speed, inverter parameters, and heat-exchanger measurements—we derive health indicators that capture changes in thermal, compressor, and airflow behavior. These indicators enable segmentation of the component life cycle into meaningful degradation stages. When clear deterioration patterns emerge, RUL forecasting is performed using a predefined failure threshold.}},
  author       = {{De Rocker, Yaël and Bogaert, Matthias and Van den Poel, Dirk}},
  booktitle    = {{56th Annual Conference of the Decision Sciences Institute, Abstracts}},
  language     = {{eng}},
  location     = {{Orlando, USA}},
  title        = {{Leveraging sensor data for fault detection in split-system air conditioners}},
  year         = {{2025}},
}