水下航行器声隐身性能快速评估研究

针对传统水下航行器声隐身性能评估方法计算时间长、实时性不强的缺点,将评估由数值计算问题变为基于多传感器信息融合的模式识别问题来解决。通过将RBF神经网络模型和LVQ神经网络模型相结合,提出了基于组合神经网络的水下航行器声隐身性能快速评估模型。利用加速度传感器测得壳体表面振动信息,抽取分析频段内每个频带的功率作为特征向量,通过组合神经网络模型进行识别分类,快速评估出航行器当前的声隐身状态。并利用水下双层加肋圆柱壳体模拟航行器舱段缩比模型,进行了水下声学试验,验证了方法的实时性和有效性。该评估方法计算速度快、评估正确率较高、通用性较强,可很好地应用于各类水下结构的声学状态评估。

Research on Fast Assessment of Acoustic Stealth Situation for Underwater Vehicles

Traditional algorithm on underwater vehicle radiated noise calculation could not satisfy real-time request in practical use due to its long calculating time. Aimed at this disadvantage, the problem can be converted into a pattern recognition problem based on multisensor data fusion, and a new method of fast evaluation acoustic stealth condition of underwater vehicle has been researched. Using the measured acceleration of the cylindrical structure shell, a rapid method of acoustic stealth situation assessment of underwater vehicle based on RBF and LVQ combination neural-network was realized. And the power of each frequency band in analysis band was extracted as the feature vectors. An iron underwater cylindrical structure was simulated as underwater vehicle cabin. A underwater acoustic experiment was made to prove the high accuracy and short calculating time of the method. So this method can be well used for acoustic stealth situation assessment of most kinds of underwater structure.