基于MLP神经网络的海上发射平台主尺度优化研究  

Research on principal dimensions of launch platform based on MLP neural network

作　　者：卢志童 王宝来 王超 王棋 LU Zhitong;WANG Baolai;WANG Chao;WANG Qi(Yantai Research Institute of Harbin Engineering University,Yantai 264000,China)

机构地区：[1]烟台哈尔滨工程大学研究生院,山东烟台264000

出　　处：《舰船科学技术》2025年第6期94-100,共7页Ship Science and Technology

基　　金：山东省海上航天装备技术创新中心(鲁东大学)开放基金项目(MAETIC202202);山东省重点研发计划(2020CXGC010701);烟台先进材料与绿色制造山东省实验室开放课题基(AMGM2022A01);烟台哈尔滨工程大学研究院校企合作基金(210F0401004)。

摘　　要：海上发射平台是确保海上发射安全的重要一环,利用多层感知机(MLP)神经网络对海上发射平台进行优化。参考海上发射的总体性布置设计,建立了火箭在发射船上刚性连接段的弯矩模型,计算平台的不同浪向角下的运动响应确定0°浪向角作为最优浪向角;选取6个平台主尺度参数作为优化变量,通过MLP神经网络模型对母型平台进行优化设计,得到的优化平台火箭弯矩相比母型平台减少了21.2%;随后对部分优异的样本点进行建模和数值仿真。结果表明,MLP神经网络模型误差在1%左右,验证了模型准确性,为发射平台的设计优化提供一种新方法,同时为海上发射安全性提供一种预测手段。Marine launch platform is an important part to ensure the safety of Marine launch.This paper uses the multilayer perceptron(MLP)neural network to optimize the Marine launch platform.According to the general layout design of the launch at sea,the bending moment model of the rigid connection section of the rocket on the launch ship is established,and the 0°wave direction Angle is determined as the optimal wave direction Angle by calculating the offshore platforms respond to motion at different wave angles.Six main scale parameters of the platform were selected as optimization variables,and MLP neural network model was used to optimize the parent platform.The bending moment of the optimized platform was reduced by 21.2%compared with that of the parent platform.Then,some excellent sample points are modeled and simulated.The results show that the error of MLP neural network model is about 1%,which verifies the accuracy of the model.It provides a new method for the design optimization of launch platform and a prediction method for the safety of launch at sea.

关 键 词：MLP神经网络 全析因设计法 火箭弯矩 半潜式平台优化 

分 类 号：U661.2[交通运输工程—船舶及航道工程]