基于LMBP神经网络的声诱饵对抗鱼雷效能评估

Evaluation of Anti-Torpedo Operational Effectiveness for Acoustic Decoy Based on LMBP Neural Network

根据实际情况,补充调整建立了声诱饵对抗鱼雷效能评估指标体系。结合专家调查法和仿真试验,采用基于LM快速算法的BP神经网络(LMBP)综合评估法对声诱饵对抗鱼雷效能进行评估,并对LM算法和常用的Traingdx算法进行了对比分析,证明LM算法误差更小、训练速度更快。最后,通过实例证明将BP神经网络评估方法应用于水声对抗效能评估切实可行。该方法能综合考虑专家经验和试验数据等主客观因素,最大限度减少单纯主观或客观赋权带来的误差,通过神经网络的"自学习"得出明确的综合评估值,具有较强的实用性和通用性。

To better evaluate anti-torpedo operational effectiveness for various acoustic decoys some effectiveness evaluation factors are presented and analyzed according to the actual .condition. Combining Delphi and simulation experiment, back propagation （BP） artificial neural network based on Levenberg-Marquardt （LM） algorithm is applied to evaluate anti-torpedo operational effectiveness for acoustic decoys. Traingdx algorithm and LM algorithm are compared and analyzed, and the latter shows faster training speed and smaller errors. By comparing the counterwork effects of two kinds of acoustic decoys, the evaluation method based on BP neural network is proved to be more feasible.