Two-Dimensional Gust Identification for Large-Wingspan Elastic Aircraft Based on Deep Learning

Despite the development of many advanced gust detection devices, it remains a challenge to measure gusts directly during flight. The advent of gust identification technology provides an indirect method of measurement by identifying gusts from aircraft gust responses. Previous studies of gust identification predominantly adopt a one-dimensional gust hypothesis with small-wingspan aircraft as the research object, rarely considering aircraft elasticity. In this paper, two-dimensional gust identification for an elastic aircraft with a large wingspan is investigated. A deep learning model based on the long short-term memory neural network is developed, which uses multipoint acceleration responses of the aircraft to identify two-dimensional gust fields that vary along the wingspan, including discrete gusts generated by large-eddy simulation, continuous gusts calculated by the spatial function of the von Karman model, and mixed gusts formed by the superposition of both. Considering practical applications, the influence of sensor noise and model uncertainties on identification is analyzed. [Formula: see text]-medoids clustering is utilized to determine the optimal accelerometer layout, thereby reducing the number of sensors while guaranteeing identification accuracy.