【作者】 陈凯；

【导师】 孙国正；

【作者基本信息】 武汉理工大学 ， 机械设计及理论， 2003， 硕士

【摘要】 我国目前混凝土泵车产业虽然有了很大发展，但同国外相比仍处于较低层次水准，42m以上长臂架泵车生产大多需要技术引进或进口关键部件。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、有限元分析、遗传算法有机结合，对混凝土泵车臂架系统的结构优化设计进行了研究。 针对臂架结构优化设计过程中需要反复进行有限元分析，费力费时，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神经网络进行结构分析的预测输出。利用BP神经网络建立起结构设计参数与变形、应力等的非线性映射关系，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运算得到。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：在每次学习过程中，进行二次动态学习参数的自适应调整。第一次是自适应调整学习率η和动量因子α。为了从整体上限制学习速率过大而出现计算溢出，对学习率还作了进一步的修正。第二是自适应调整允许均方误差e的值。本文将改进的BP网络算法用于结构分析器的训练。 本文利用遗传算法调用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8章。第一章为绪论，综述了国内外人工神经网络、有限元、遗传算法研究的历史、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。第二、三章系统叙述了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。第四章概述了遗传算法的理论及其实现。第五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。第六章介绍了基于BP网络的臂架结构分析器的构建。第七章建立了臂架结构系统的优化数学模型，介绍了优化设计的实现。第八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】 Current there is very big development in our country’s concrete pump vehicle industry. But compared with the abroad, it is still placed in a low level . The production of long boom concrete pump vehicles above 42m need to introduce the whole techniques or import key parts, so it is very necessary to improve the design level of concrete pump vehicle series with long booms and especially to the boom system. 42m concrete pumping vehicle has being produced recently in domestic. This paper researches into the frame optimizing design of boom system of the vehicle. The research combines BP neural networks, genetic algorithm and finite element.Because of the complexity of the booms’ system, we often calculate the results again and again when making use of finite element software. The efficiency and the model’s accuracy need further improving and the optimum results require further perfecting. The finite element model and computing models are established in the paper, and furthermore, the computing results answer the measuring data well. The intelligent algorithm of modified BP neural network is applied in the paper to design the nonlinear mapping relationship between the stress & the strain and the main structure variables. Subsequently, Simulation Anneal-Genetic Algorithm is used to optimize the data and gives a better result. According to the engineering and optimum needing, a kind of integrated optimum software is set up to design the booms’ system. The software not only is a good tool for the boom system’s designing, but also verifies the rationality and feasibility of these models. In a word, this paper achieves a series of useful theories and practical results.This article divides eight chapters altogether. Chapter one is introduction, which summaries the history, development and current situation of BP neural networks, genetic algorithm and finite element, talks about the domestic industry of concrete pump vehicle , then explains the proposition, purpose and meaning of this thesis. In chapter two to three, the theories, construction and main defects of BP neural networks are described systematically and a modified method, which can adjust study parameters dynamically and adaptively, is put forward. To the booms ’ system, the finite element model and computing model are built inchapter five. Subsequently, the rationality of the models is verified. Based on the improved BP neural networks algorithm, an analysis software tool used for the analysis is introduced in Chapter six. Chapter seven set up the optimizing model of the booms’ system, then discusses the realization of the optimum design. Chapter eight is a conclusion of the study work of the thesis and puts forward the direction of further research development in the future.更多还原