基于BP神经网络与L-M算法的潜艇声纳自噪声预报

Submarine Sonar Self-Noise Forecast Based on BP Neural Network and Levenberg-Marquart Algorithm

L-M(L evenberg-M arquart)算法与BP(B ack-P ropagation)神经网络相结合,使神经网络在多样本、大变量输入的情况下,具有更快的收敛速度和更高的逼近精度。将BP神经网络与L-M算法相结合应用于潜艇声纳自噪声预报;分析了影响潜艇声纳自噪声的各种声源参数;利用潜艇声纳实测数据进行网络训练,训练好的神经网络可以对潜艇声纳自噪声进行精确预报。

If L-M （Levenberg-Marquart） algorithm is combined with BP（Back-Propagation） neural network, faster speed of convergence and higher learning accuracy of BP neural network can be acquired as multiple parameters and large patterns are inputted. This paper combines L-M algorithm with BP neural network to forecast submarine sonar self-noise. All kinds of parameters that have function to submarine sonar self-noise have been analyzed. Actual data are utilized to train BP neural network and then the trained BP neural network can be used to accurately forecast submarine sonar self-noise.