Research on Fault Diagnosis Modeling Methods Driven by Route QAR Data

Diagnosing flight route faults based on QAR data is crucial for preventing flight accidents and reducing maintenance costs. This paper presents a method for diagnosing flight route faults using a Long Short-Term Memory Network and Decision tree algorithm (LSTM-DT). We used feature optimization algorithms to process the original QAR data, including rolling window operation and principal component analysis. The optimized features were then used to establish an LSTM anomaly detection model to identify abnormal points of the faulty route. We established a decision tree fault diagnosis model using these detected outliers to determine the corresponding fault types. When tested with real airline data, the fault diagnosis accuracy of this method reached 99.45%, confirming its effectiveness.