摘要
反鱼雷对抗效果是水声对抗装备效能评估的核心和难点,运用神经网络进行效果评价具有一定的优势,但由于单纯采用专家评分的方法,使得训练样本集的可靠性不高,降低了网络权值的可信度。为此分析了影响干扰器对抗效果优劣的几类指标,结合仿真试验技术,提出了运用改进算法的BP网络来进行效果评价的方法,合理地构建了由专家评分及仿真试验的指标值共同组成的混合训练样本集,提高了训练样本的充分性和可信性,最后通过对两型干扰器的对抗效果评价比较,验证了方法的有效性。
Anti - torpedo underwater acoustic countermeasure effectiveness is the key and knot to the effectiveness evaluation of acoustic warfare equipments. The application of neural network in the effectiveness evaluation bears certain advantage. But because of the simplex employment of expert evaluation method, the reliability is not high in training sample collections. Therefore, the credibility of network values is reduced. The article analyzes several indexes affecting acoustic countermeasure effectiveness of noise jammers： Combined with simulative experiment, a new effectiveness evaluation method with improved algorithm back -propagation neural network（ BP NN）is proposed. And a rationally combined training sample collection consisting of both expert evaluation and index values gained by simulating test is constructed. As a result, the reliability and credibility of training samples are enhanced. Finally, through the comparison of evaluations between two types of noise jammers, the effectiveness of the method is proved.
出处
《计算机仿真》
CSCD
2008年第10期16-19,81,共5页
Computer Simulation
关键词
反鱼雷
对抗效果
效能评估
神经网络
Anti - torpedo
Countermeasure effectiveness
Effectiveness evaluation
Neural network
