摘 要

世界经济全球化以及我国“海上丝绸之路”，一带一路的国家战略发展，进出口贸易也越来越多，我国作为生产制造的大国，离不开海洋运输，未来对船舶的数量需求会变大，对技术含量会变高。而在船舶制造过程中，船舶轴系校中不仅影响船舶制造工期。因此，轴系校中质量的研究在船舶领域扮演者重要角色。大量的相关研究工作者致力于将轴系校中中轴承位移直观的计算出来。本文将基于神经网络，建立神经网络模型来表示轴系特性，解决船舶建造时轴系校中耗费时间长，校中质量不好的问题，并且可以为智能化船舶建造过程打下基础。

本文针对上述问题主要做了以下研究工作：

（1）构建神经网络模型。通过对轴系神经网络模型的初步训练和分析，选择最适合轴系特性的激活函数、学习函数、训练函数和数据归一化函数。

（2）寻找最适合轴系的神经网络参数。通过改变神经网络模型的几个参数，得到当隐含层神经元为5，学习率为0.008，目标误差值为1E-5时，对轴系的拟合效果最好。

（3）设计应变片法测量轴系负荷的方法。从器材的选用、操作安装的工艺以及计算过程及原理来分析应变片测轴系受力的原理。

关键词：神经网络；应变片法；轴系校中

Abstract

With world economic globalization and “maritime Silk Road” developing，the import and export become more than before.China，as one of the bigest manufacturing country，cannot develop without ocean transportation. Also the demand for ships will bincrease and the technology content of ship will increase.In the process of shipbuilding，the shafting alignment not only affects the building time ，but also affect the safe and steady.Therefore the shafting play a important role.There are many researchers that devote to the shafting alignment calculation.in this paper，we established a neural network model that simplify calculation process.

the following research work has been done to solve the above problems:

1. Construct neural network model. The neural network model is trained by using the data of shafting characteristics， and the neural network model which accords with shafting characteristics is obtained.

(2)Search for the most suitable parameters of neural network. By changing several parameters of neural network model， the fitting effect of neural network model on shafting under different parameters is obtained and analyzed.

(3) Design strain gauge method to measure the load of shafting. From the selection of equipment， operation and installation process， calculation process and principle， strain gauge is used to measure the stress of shafting.

Key Word: Neural Networks;Strain gauge ; Shafting Alignment

目 录

第1章 绪论.....................................................................................................................................1

1.1 研究的目的和意义................................................................................................................1

1.2 国内外研究现状分析............................................................................................................1

1.3 论文的主要工作...................................................................................................................2

1.3.1 研究内容........................................................................................................................2

1.3.2 研究方法........................................................................................................................2

1.3.3 论文的章节安排.............................................................................................................3

第2章 神经网络模型......................................................................................................................4

2.1 BP神经网络的原理................................................................................................................4

2.1.1 BP神经网络.....................................................................................................................4

2.1.2 GA-BP神经网络...............................................................................................................6

2.2 BP神经网络函数....................................................................................................................8

2.2.1 激活函数........................................................................................................................8

2.2.2 学习和训练函数...........................................................................................................11

2.2.3 数据归一化函数...........................................................................................................12

2.2.4 BP神经网络建立...........................................................................................................14

2.3遗传算法优化.......................................................................................................................14

2.3.1基因编码及适应度函数................................................................................................14

2.3.2工具箱及函数................................................................................................................15

2.4 本章小结.............................................................................................................................16

第3章不同参数对神经网络精度的影响.....................................................................................17

3.1 样本选取原则.....................................................................................................................17

3.2 样本处理.............................................................................................................................17

3.2.1 训练集的选取…….........................................................................................................18

3.2.3 仿真误差计算…........….................................................................................................18

3.3 实验结果…….....................................................................................................................…19

3.3.1 初始权值与阈值……....................................................................................................19

3.3.2 隐含层节点数…...........................................................................................................19