Multi-fidelity Data Fusion for Electromagnetic Field Prediction of Electromagnetic Railgun Field

High-fidelity (HF) numerical simulations play a crucial role in realizing the fine design of armature and track structures for electromagnetic railguns, but are limited by the high cost. In contrast, low-fidelity (LF) numerical simulation serves as an important alternative paradigm that also reflects the flow of electromagnetic fields, but is less accurate compared to experiment. By fusing high-fidelity and LF electromagnetic field data, accurate prediction of electromagnetic field can be realized at a low computational cost. In this paper, a multi-fidelity convolutional long- and short-term memory neural network, in which both HF and LF data are introduced into the loss function with appropriate weighting factors to balance the overall accuracy, is developed for predicting the electromagnetic field distribution in the orbit of an electromagnetic railgun. The results show that the proposed method can accurately predict the electromagnetic fields distribution in an electromagnetic railgun track using an appropriate amount of LF data and a small amount of HF data. The performance of the proposed method in this paper is superior compared to the convolutional long- and short-term memory neural network method that only works from either high or low fidelity.