A Wind Speed Forecasting Method Based on Machine Learning for Operating Unguided Submunitions from a Cluster Munition

Wind forecasting is essential when operating ballistic weapons, including an aircraft-mounted cluster munition. Simple linear wind models are commonly employed for wind forecasting. This paper presents a novel approach for predicting wind direction and speed at intermediate altitudes from the release point to the ground. It is based on ANN-based nonlinear regression function trained on wind historical data collected at observation stations where wind is measured at 10-min intervals using wind observation equipment. The dataset used for ANN training was obtained by processing and utilizing data downloaded from the Korea Meteorological Administration Open MET Data Portal. This dataset includes valuable information such as measurement locations, dates, times, altitudes, wind directions, and wind speeds for eleven observation stations in South Korea. Also, we propose a method to employ spatial interpolation techniques for wind speed prediction in areas without observation stations, thus developing a wind speed prediction model for the entire South Korean region. The spatial interpolation technique utilized the inverse distance weighted (IDW) method, which assigns greater weight to the actual values from nearby observation stations when performing interpolation. To evaluate the performance of the proposed WSP-ANN model, a comparative analysis was conducted with the predictions of the common launch acceptability region approach (CLARA)’s simple linear wind model proposed by the United States Air Force and SAE International. The results clearly demonstrated that the proposed WSP-ANN model outperforms the CLARA’s simple linear wind model in terms of wind prediction performance.