【作者】 马元驰；

【导师】 孙国正；

【作者基本信息】 武汉理工大学 ， 机械设计及理论， 2004， 硕士

【摘要】 在我国，混凝土泵车的设计与制造时间不长，一般生产36M扬程以下的产品，42M以上的大多是进口组装，特别是结构系统。36M以下的产品还存在振动较大，有时还出现臂架结构疲劳断裂等问题。由于臂架对长臂混凝土泵车的性能有重大影响，为了尽快提高我国混凝土泵车的设计制造水平，受企业委托，本文以国内新近开始设计生产的42m混凝土泵车臂架系统为研究对象，将BP神经网络、有限元(模态)分析、演化策略有机结合，对混凝土泵车臂架系统的结构优化设计进行了研究。 本文利用能处理模糊、含有噪音数据和具有很强非线性映射功能的BP神经网络进行结构动态特性的预测输出。利用BP神经网络建立起臂架结构设计参数与固有频率的非线性映射关系，即建立了基于BP网络的结构分析器。BP网络训练样本由有限元模型模态分析所得。本文还对BP神经网络的理论和实现方法进行了较深入的研究。针对BP神经网络的缺点，研究了一种动态自适应调整学习参数的改进型BP算法。本文利用进化算法调用BP网络训练结果进行结构设计参数优化，并开发了混凝土泵车臂架优化设计软件。通过优化，臂架结构主要设计参数有不同程度改善，达到了优化目的。 本文共分为7章。第1章为绪论，综述了国内外人工神经网络、模态分析、有限元、进化算法(包括演化策略)研究的历史、发展和现状，以及国内混凝土泵车的产业发展状况，阐述了本课题的提出、目的和意义。第2章系统叙述了人工神经网络特别时BP神经网络的理论、结构、主要缺点和改进措施，提出了一种改进型BP算法。第3章概述了进化算法特别是进化策略的理论及其实现。第4章建立了混凝土泵车臂架结构的力学和有限元模型，并进行了实验验证。第5章介绍了基于BP网络的臂架结构和固有特性分析器的构建。第6章建立了臂架结构模态分析的优化数学模型，介绍了优化设计的实现。第7章为全文研究工作的总结，提出了今后进一步研究的发展方向。更多还原

【Abstract】 In our country, concrete pump design and manufacture of car time long, produce 36M lift following products generally, 42M a the above-mentioned one mostly to import and assemble, especially the structure system. 36M following products have vibration to be relatively heavy also , appear arm shelf structure tired question of rupturing etc. sometimes. Because arm shelf have great influence on performance of the long arm concrete pump car, for improve of our country concrete pump designing and manufacturing level of car as soon as possible, entrust by enterprise, with begin 42m concrete pump car arm shelf system produced to design as the research object recently at home this text, analyse BP neural network, finite element, hereditary organic to combine algorithm, carry on research to concrete pump car arm shelf structure optimization design of system.This text can deal with to utilize fuzzily , contain noise data and have very strong to shine upon BP neural network of function carry on structure prediction of dynamic characteristic export while being non-linear. Utilize BP neural network set up arm shelf structure design parameter shine upon the relation with nonlinearity of natural frequency, namely has set up the structure analysor based on BP network. BP network train sample mode analyse the income by finite element model. This text carry on deeper studying to BP neural theory and implementation method of network also. To BP neural shortcoming of network , study one dynamic self-adaptation is it study improvement type BP algorithm of parameter to adjust. This text is it evolve algorithm transfer BP network train result carry on structure design parameter optimized to utilize, and has developed the concrete pump train arm shelf optimization design software. Through optimize, arm shelf structure main design parameter improve in various degree, have achieved the goal of optimizing.This text is divided into 7 chapters altogether. Chapter one the introduction, survey artificial neural network , modal analysis , finite element both at home and abroad , evolve algorithm history , development and current situation ofstudying, and domestic concrete pump industry state of development of car, explain copies of proposition, purpose and meaning of subject. Chapter two system BP neural theory , structure , main shortcoming of network improve measure when narrating artificial neural networking very, have put forward a kind of improvement type BP algorithm. Chapter three sum up evolve algorithm especially evolve the theory of tactics and realize. Set up concrete pump car arm shelf mechanics and finite element model of structure, and has carried on the experiment to verify. Introduce based on BP network arm shelf structure and inherent characteristic construction of analysor. Set up arm shelf structure optimization mathematics model that mode analyses, have introduced the realization of optimization design. Chapter seven full text summary of research work, propose developing direction that will study further in the future.更多还原