207 / 2025-06-14 11:13:54

DEWMA-IForest：Abnormal Detection Method for Subcomponents of Aircraft Air Conditioning System

Aircraft Air Conditioning System, Anomaly Detection, Exponentially Weighted Moving Average, Isolation Forest, Decision path analysis

The characteristics of aircraft air conditioning systems are complex operating conditions and high coupling of components, which makes it difficult for traditional aircraft maintenance modes to meet the dual requirements of safety and economy in modern aviation transportation. To address the above-mentioned problems, an abnormal detection method for sub-components of the aircraft air-conditioning system based on DEWMA-IForest is proposed in this paper. Firstly, the baseline data of the outlet temperature of the sub components of the aircraft air conditioning system is used as the data source. The parameter dynamic adjustment mechanism is used in the framework of the exponential weighted moving average algorithm to enhance the model's ability to capture the time-varying and changing trends of data features, and thus trend anomaly detection is achieved. Secondly, the isolation forest method is used to construct binary isolation trees to quantify the degree of anomaly of data points. Meanwhile, the local outlier factor algorithm and the one-class support vector machine algorithm are employed to conduct a secondary anomaly assessment on the outliers in the isolation forest, so that scatter anomaly detection is realized. Finally, an anomaly assessment mechanism based on decision path analysis is established. By obtaining the anomaly scores and decision path features of data points, the average isolation depth of each data point is calculated to quantify the degree of anomaly, and thus an interpretable and quantifiable index for outliers is established. The experimental results show that the dual-detection mechanism of trend anomaly and scatter anomaly is used by the proposed method, which significantly improves the transparency of anomaly criteria while maintaining high detection accuracy of abnormal performance states in the air conditioning subsystem. It can provide a scientific basis that combines accuracy and credibility for intelligent maintenance decision-making.