A Phase-Divided MSPC Approach for Wing Strain Data Characterization and Anomaly Identification.

The effective characterization and anomaly identification of flight strain data are critical to aircraft structural health monitoring. To address the challenges of large-volume wing strain data and the limitations of univariate methods in capturing synergistic variations among multiple features, this paper proposes a phase-divided multivariate statistical process control (MSPC) approach for wing strain data characterization and anomaly identification. Focusing on the rear spar flange of a specific aircraft wing, independent multivariate control baselines are established for the takeoff, cruise, and landing phases. A six-dimensional feature vector is constructed using peak value, root mean square, standard deviation, peak-to-peak value, kurtosis, and skewness to quantitatively characterize the strain data of the wing structure across different flight phases. After feature standardization, a Hotelling's multivariate statistical control model is developed to identify anomalies in structural strain states. Based on 30 normal sorties, the upper control limit is determined as UCL = 26.58 at a 99% confidence level. Anomaly diagnosis is performed on six test sorties using