Torpedo Hit Probability Prediction Method Based on Deep Neural Network

Abstract

In order to further improve the prediction ability of torpedo hit probability, a torpedo hit probability prediction method based on deep neural network(DNN) was proposed. Firstly, the situation characteristic information was extracted, and the desired situation space was set. In addition, the large sample data set of torpedo operation was constructed based on the Monte-Carlo method. On this basis, the Levenberg-Marquardt optimization algorithm was used to calculate the optimal gradient direction, which improved the computational efficiency of the algorithm. Finally, two typical operational application modes were given based on the model. Experimental results show that the proposed DNN-based prediction model has higher recognition accuracy than other typical intelligent algorithms, which verifies the effectiveness and superiority of the model.

Keywords

• torpedo
• deep neural network
• hit probability
• operational effectiveness