理工大学学士学位论文PAGE\*MERGEFORMAT59摘要神经网络通过模拟大脑神经处理信息的方法，完成信息并行处理和非线性转换。由于神经网络具有强大的学习功能，可以比较轻松地实现非线性映射过程，并且具有大规模计算的能力。神经网络在自动化、计算机和人工智能领域都有着广泛的适用性，实际上也确实得到了大量的应用，解决了很多利用传统方法难以解决的问题。本文主要研究了神经网络机理、特性及模型，并利用神经网络工具箱针对齿轮箱开展故障诊断识别与研究。同时，结合常见的网络故障诊断方法和几种典型的神经网络模型，重点开展基于BP神经网络的故障诊断方法的建模与研究，通过已知数据训练BP神经网络模型，并利用Matlab实现BP神经网络算法的齿轮箱故障诊断，仿真结果表明模型能够快速收敛，并较好的对齿轮箱故障进行识别。关键词：故障诊断；神经网络；BP网络；MatlabAbstractNeuralnetworkmethodthroughthesimulationofthebrainprocessinformation,completetheparallelprocessingandnon-linearconversionofinformation.Becauseneuralnetworkhasstronglearningfunction,canmoreeasilytorealizenonlinearmappingprocess,andhastheabilityoflarge-scalecomputing.Neuralnetworkinthefieldofautomation,computerandartificialintelligencehavebroadapplicability,actuallydidgetalotofapplication,solvedmanyusingthetraditionalmethodisdifficulttosolvetheproblem.Inthispaper,westudythemechanism,characteristicsandneuralnetworkmodel,andusingneuralnetworktoolboxforgearboxfaultdiagnosisistoidentifyandstudy.Atthesametime,combinedwiththecommonnetworkfaultdiagnosismethodsandsometypicalneuralnetworkmodels,focusonthefaultdiagnosismethodbasedonBPneuralnetworkmodelingandresearch,throughtrainingtheBPneuralnetworkmodelofdataisknown,andusetheMatlabBPneuralnetworkalgorithmofgearboxfaultdiagnosis,simulationresultsshowthatthemodelcanfastconvergence,andbetterforgearboxfaultrecognition.Keywords:networkfaultdiagnosis;neuralnetwork;BPnetwork;Matlab目录TOC\o"1-4"\h\z\uHYPERLINK\l"_Toc359326128"1绪论PAGEREF_Toc359326128\h1HYPERLINK\l"_Toc359326129"1.1研究背景及意义PAGEREF_Toc359326129\h1HYPERLINK\l"_Toc359326130"1.2齿轮箱故障诊断研究现状PAGEREF_Toc359326130\h1HYPERLINK\l"_Toc359326131"1.3齿轮箱故障诊断的发展趋势PAGEREF_Toc359326131\h2HYPERLINK\l"_Toc359326132"1.4本文组织结构PAGEREF_Toc359326132\h3HYPERLINK\l"_Toc359326133"2神经网络特性分析与研究PAGEREF_Toc359326133\h4HYPERLINK\l"_Toc359326134"2.1神经网络概念及特性PAGEREF_Toc359326134\h4HYPERLINK\l"_Toc359326135"2.2神经网络发展史PAGEREF_To