基于人工神经网络方法的舰船参数灵敏度分析

舰船参数灵敏度分析不仅为判断各参数的重要性大小提供了依据,量化的灵敏度指标也是后续参数估计的前提。在舰船系统中,参数与舰船综合性能的显式函数关系不易得到,导致一阶灵敏度指标无法直接求取。本文确定了舰船的几个主要参数,建立起了舰船综合性能评估框架,利用人工神经网络的方法,用C语言编制了一套舰船综合性能的评估程序,将已有的数据、专家经验等量化后得到评判指标数值,形成样本,进行神经网络训练,然后通过训练完成的神经网络对各个参数进行了灵敏度分析。

Parametric sensitivity analysis of ships based on artificial neural network method

The parametric sensitivity analysis of ships can be used for ranking the importance of different pamameters,especially the quantitative sensitivity indices used in the following parameter estimation.In the systems of ships,however,the explicit functional relationship between parameters and main capability indices of ships is too complex to be derived,as a result,the first-order derivative sensitivity indices are unable to be calculated.In this paper,several parameters of ships are chosen,experts' advice considered,to establish an evaluating framework of ships' main capability by using artificial neural network.The input quantitative parameters are trained as samples in the artificial neural network program.After that, the parametric sensitivity analysis can be done by using the trained artificial neural network program.