Sensorless Position Estimation in Electromagnetic Launchers Using Recurrent Neural Networks with Repeated k-Fold Cross-Validation

Abstract

Systems utilizing electromagnetic launchers transform energy from one state to another. In electromagnetic launchers, the launched object follows the magnetic field generated by one or more fixed coils to achieve motion. Knowing the position of the projectile within the coil is crucial for efficient launching. In traditional systems, the position of the projectile is detected using different sensors. In this study, a sensorless method is proposed using Recurrent Neural Networks (RNNs) to estimate the projectile’s position based on current and voltage signals. Various RNNs architectures, including SimpleRNN, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional GRU (Bi-GRU), are evaluated through repeated k-fold cross-validation. The GRU model achieved the highest accuracy with 94.88%, demonstrating the feasibility of sensorless position estimation in electromagnetic launchers. However, limitations such as the sensitivity of the model to noise in electrical measurements and the need for extensive data for training must be addressed. Despite these challenges, the proposed method offers a cost-effective and simplified alternative to sensor-based systems, particularly in environments where sensor integration is impractical. This approach has the potential to enhance the reliability and performance of electromagnetic launch systems.