基于BP神经网络遗传算法水轮机修复专用机器人结构优化设计
|
摘要
结构优化设计自60年代以来,随着计算机技术、数学、生物学等学科的进步而得到了迅速发展。优化技术由最优化理论、计算机技术及工程技术相结合而形成的一种现代设计方法与技术。在解决复杂的工程设计问题时,优化技术的应用可以较快地实现设计方案的最佳化,提高设计效率。本文将神经网络和遗传算法两者相结合,成功的实现了将改进型BP神经网络和遗传算法相结合应用到机器人结构优化设计中。
首先系统地分析了优化理论的发展和优化设计应用的现状,探讨神经网络的发展及其在结构优化设计过程中的应用,总结了传统的优化算法和现代优化算法的优缺点,并论证了基于遗传算法的神经网络应用的可行性和优越性。
本文以Ansys Workbech软件为工具,针对水轮机修复专用机器人在实际工作情况几个代表性姿态,在满足各种约束条件下进行结构静力分析,得到水轮机修复专用机器人工作末端的位移和整体应力分布情况,为基于BP神经网络遗传算法的机器人结构优化的网络训练、学习及检测网络性能提供代表性数据样本。
通过对BP网络模型结构的数据准备、隐层节点数确定、网络结构设计等进行分析和新的探讨,针对其缺点,提出了改进办法,即通过引入动量常量提高网络收敛速度;引入遗传算法避免了出现局部最小值,提出基于人工神经网络遗传算法的算法。该算法综合了遗传算法的全局性和改进型神经网络的快速收敛等特点,可克服遗传算法最终进化至最优解较慢和神经网络易陷入局部解的缺陷,具有较好的全局性和收敛速度。利用在Ansys Workbech获取的数据样本,验证了该方法的可行性和有效性。根据该算法对网络的拓扑结构进行了新设计,并依据MATLAB工具箱通过程序设计,真正意义上实现将BP神经网络、有限元分析、遗传算法有机结合,对水轮机修复专用机器人进行了结构优化设计。
Since 1960's,with the fact that development of discipline such as computer technology, Optimization technology have got prompt development Optimization theory, computer technology and engineering technology have combined closely, forming optimum technology, which is a modern design method, and technology. When resolving the complicated engineering design problem, the application optimizing a technology can realize the design plan optimization quicker, improve efficiency and quality of design. Have come true successfully owing to improving robot structural optimization design of integration of type BP neural networks with genetic algorithm in this paper.
At first, introduction analyzes the development of optimization theory and actuality of optimum design application carefully. Analysing process of development BP neural networks and introducing that Applytion of Structural optimization designed detailedly by the summary. Have analysed the shortcoming and merit of tradition optimization algorithm and modern excellent optimization algorithm. Analyze feasibility and superiority of genetic algorithm based on ANN.
The main body of the paper take Ansys Workbench software as implement. Choosed several typical posture in actual job condition of Special-purpose Robot for Hydraulic Turbine Repair.Under satisfying various constraint condition, Come true the analysis of the static and dynamics of robot. It is produced that the robot stress distribution cloud chart and the work terminal biggest displacement, which provide the data sample for network function training,studying and checking function based on BP neural networks and genetic arithmetic structural optimization.
Analyze and research BP network's data prepares , hide layer of node numbers , network structure design.Be aimed at to whose shortcoming , bring forward improvement way. Be that improves network convergence speed by the constant momentum, Lead into genetic algorithm having avoided appearing part minimum value. Bring forward GA-BP algorithm for opitmum design.In the algorithm, Characteristics such as the overall property having owed genetic arithmeti and the neural networks.GA provides global initial solutions, from which ANN obtains the finial solutions.Thus,the defects of slow convergence with GA and easily falling into local lutions with ANN can be overcome. Have the fairly good overall situation and convergence speed. Making use of data sample gaining in Ansys Workbech to study finally, verify feasibility and validity of that method. According to that algorithm, construct a new method about net topological frame structure and join according to MATLAB toolbox passes programming. The structural optimization realizing BP neural networks , finite element method analysis , genetic algorithm being organically combined , being in progress to the Special-purpose Robot for Hydraulic Turbine Repair structural optimization on significance designs that really.
首先系统地分析了优化理论的发展和优化设计应用的现状,探讨神经网络的发展及其在结构优化设计过程中的应用,总结了传统的优化算法和现代优化算法的优缺点,并论证了基于遗传算法的神经网络应用的可行性和优越性。
本文以Ansys Workbech软件为工具,针对水轮机修复专用机器人在实际工作情况几个代表性姿态,在满足各种约束条件下进行结构静力分析,得到水轮机修复专用机器人工作末端的位移和整体应力分布情况,为基于BP神经网络遗传算法的机器人结构优化的网络训练、学习及检测网络性能提供代表性数据样本。
通过对BP网络模型结构的数据准备、隐层节点数确定、网络结构设计等进行分析和新的探讨,针对其缺点,提出了改进办法,即通过引入动量常量提高网络收敛速度;引入遗传算法避免了出现局部最小值,提出基于人工神经网络遗传算法的算法。该算法综合了遗传算法的全局性和改进型神经网络的快速收敛等特点,可克服遗传算法最终进化至最优解较慢和神经网络易陷入局部解的缺陷,具有较好的全局性和收敛速度。利用在Ansys Workbech获取的数据样本,验证了该方法的可行性和有效性。根据该算法对网络的拓扑结构进行了新设计,并依据MATLAB工具箱通过程序设计,真正意义上实现将BP神经网络、有限元分析、遗传算法有机结合,对水轮机修复专用机器人进行了结构优化设计。
Since 1960's,with the fact that development of discipline such as computer technology, Optimization technology have got prompt development Optimization theory, computer technology and engineering technology have combined closely, forming optimum technology, which is a modern design method, and technology. When resolving the complicated engineering design problem, the application optimizing a technology can realize the design plan optimization quicker, improve efficiency and quality of design. Have come true successfully owing to improving robot structural optimization design of integration of type BP neural networks with genetic algorithm in this paper.
At first, introduction analyzes the development of optimization theory and actuality of optimum design application carefully. Analysing process of development BP neural networks and introducing that Applytion of Structural optimization designed detailedly by the summary. Have analysed the shortcoming and merit of tradition optimization algorithm and modern excellent optimization algorithm. Analyze feasibility and superiority of genetic algorithm based on ANN.
The main body of the paper take Ansys Workbench software as implement. Choosed several typical posture in actual job condition of Special-purpose Robot for Hydraulic Turbine Repair.Under satisfying various constraint condition, Come true the analysis of the static and dynamics of robot. It is produced that the robot stress distribution cloud chart and the work terminal biggest displacement, which provide the data sample for network function training,studying and checking function based on BP neural networks and genetic arithmetic structural optimization.
Analyze and research BP network's data prepares , hide layer of node numbers , network structure design.Be aimed at to whose shortcoming , bring forward improvement way. Be that improves network convergence speed by the constant momentum, Lead into genetic algorithm having avoided appearing part minimum value. Bring forward GA-BP algorithm for opitmum design.In the algorithm, Characteristics such as the overall property having owed genetic arithmeti and the neural networks.GA provides global initial solutions, from which ANN obtains the finial solutions.Thus,the defects of slow convergence with GA and easily falling into local lutions with ANN can be overcome. Have the fairly good overall situation and convergence speed. Making use of data sample gaining in Ansys Workbech to study finally, verify feasibility and validity of that method. According to that algorithm, construct a new method about net topological frame structure and join according to MATLAB toolbox passes programming. The structural optimization realizing BP neural networks , finite element method analysis , genetic algorithm being organically combined , being in progress to the Special-purpose Robot for Hydraulic Turbine Repair structural optimization on significance designs that really.
引文
[1]王凤歧等.现代设计方法[M],天津大学出版社,2005
[2]梁尚明,殷国富.现代设计方法[M],化工工业出版社,2005
[3]王英杰,大型机械系统复杂构件结构优化设计方法研究:[燕山大学博士论文]:秦皇岛:燕山大学机械工程学院,2001
[4]陈科,谢守振,赵韩.基于Bol t zmann机人工神经网络的机构优化设计[J].农业机械学报,2004
[5]Charnes A,Cooper W W.Chance constrained programming.Management Science 1959,6(1):73-79.
[6]Hilton,H.H,Feigen,M.Minimum weight analysis based on structural reliability.Journal of Aerospace Science,1960,27(9):641-651.
[7]P.Pederson.Optimal Joint Positions for Space Trusses.J.Struct.Div.,ASCE,1999,(ST10):2459-2476
[8]O.C.Zienkiewicz,J.S.Campbell.Shape Optimization and Sequential Linear Programming.In:Optimal Structural Design,Edited by R.H.Gallagher and O.C.Zienkiewicz,John Wiley,New York,1973R.H.GallaghO.C.Zienkiewicz,John Wiley,New York,1973
[9]Cz I.N.Rozvany,F.Moses.Automated Design of Trusses for Optimum Geometry J.Struct.Div.,ASCE,1998(ST3):671-690
[10]王希诚,钱令希.多层次联合的结构优化设计[J].计算结构力学及其应用,1988,5(4):69-75
[11]程耿东.可靠性最大的结构优化设计[J].计算力学学报,1984,12(4):27-131.
[12]冯元生.结构余度理论及其对结构设计的影响[J].农业机械学报,1987,6(4):56-59
[13]宋天霞.复杂结构优化选型的计算方法[J].固体力学学报,1990,11(3):235-245
[14]杜轩,陈柏鸿,彭立焱,钟毅芳.机械产品的全性能优化建模及求解[J].机械设计与制造工程.2000.29(4):20-23
[15]刘仁云.基于计算智能技术的结构可靠性优化设计研究:[吉林大学博士论文]:吉林:吉林大学固体力学,2006
[16]Shahab Mohaghegh,Andrei Popa,Sam Ameri,Design optimum frac jobs using virtual intelligence techniques,Computers &Geosciences 26(2000)927-939
[17]Steven Walczak,Narciso Cerpa,Heuristic principles for the design of artificial neural networks,Information and Software Technology 41(1999)107-117
[18]Glodberg D E.,Richardson J.Genetic Algorithm with Sharing for Multimodel Function Optimization Genetic Algorithms and Their Applications Proceedings of the Second In ten rational Conference on Genetic Algorithm,1987.41-49;
[19]Lin C-Y Hajela P.Genetic algorithmsin optimization problems with discrete andintegerdesign variables[J].Eng.Opt.,1992.19(4):309-32
[20]GenM,Kwan W K,Yamazak 1G.Project scheduling using hybrid genetic algorithm withfuzzy logic controller in SCM environment[J].Tsinghua Science and Technology,2003,19(4):19-29)
[21]Li Quanyong.Optimization design of antenna structure by genetic algorithm[A]Proceedings of the CJK-OSMI[C].XID-IAN UNIVERSITY PRESS,1999:188-191
[22]Han Yingshi,Guo Pengfei.A Hybrid Genetic Algorithm for Structural Optimization with Discrete Vari-ables,Optimization of Structural and Mechanical Systems[C].Xidian
[23]王洪元 史图栋.人工神经网络技术及其应用[M],北京:中国石化出版,2002
[24]焦李成.神经网络计算[M],西安:西安电子科技大学出版社,1993
[25]王洪元 史国栋.人工神经网络技术及其应用[M],北京:中国石化出版,2002
[26]焦李成.神经网络计算[M],西安:西安电子科技大学出版社,1993
[27]杨建刚.人工神经网络实用教程[M],杭州:浙江大学出版社,2001
[28]郭代仪,廖小云,雷文宇.神经网络及其在机械工程中的应用[M],重庆:重庆大学出版社,1998
[29]靳蕃等.神经网络与神经计算机原理·应用[M],四川:西南交通大学出版社,1991。
[30]姜绍飞,贾连光,于军.基于神经网络的精架结构优化设计[J].东北大学学报(自然科学版),2003.24(2):166-169
[31]许素强,夏人伟.杆架结构拓扑优化与遗传算法[J].计算机结构力学及其应用,1994.11(4):436-446
[32]白广忱,黄洪钟.机械系统可靠性的多目标模糊优化设计[J].机械设计,1998
[33]董小兰,吴卫,李晓慧.BP神经网络及其在压铸模具优化设计中的应用[J].模具技术,2006
[34]王伟,赵美英,赵锋.基于人工神经网络的结构布局优化设计[J].机械设计,2006
[35]V.Srinivas,K.Raman janeyulu,An integrated approach for optimum design of bridge decksusing genetic algorithms and artificial neural networks
[36]Shahab Mohaghegh,Andrei Popa,SamAmeri,Design optimum frac jobs using virtual intelligence techniques,Computers &Geosciences 26(2000)927-939
[37]雷铁安.基于人工神经网络(ANN)的结构学习及应用:[北京交通大学硕士学位论文].北京:北京交通大学车辆工程,2005
[38]李雄飞,李军编.数据挖掘与知识发现[M],北京:高等教育出版社,2003
[39]张文修,梁怡编著.遗传算法的数学基础[M],西安:西安交通大学出版社,2000
[40]刘远.基于遗传算法的大型起重船拔杆式臂架优化设计研究:[天津大学硕士学位论文].天津:天津大学船舶与海洋结构物设计制造,2004
[41]谢桶,陈英俊.遗传算法的改进策略及其在桥梁抗灌优化设计中的应用效果[J].工程力学,2000
[42]张明辉,王尚锦.遗传算法在结构形状优化中的应用明[J].机械科学与技术,2001
[43]唐文艳.结构优化中的遗传算法研究和应用田[J].大连:大连理工大学,2002
[44]陈凯.混凝土泵车臂架结构的智能优化设计[J].武汉:武汉理工大学,2003
[45]闵惜琳,刘国华.人工神经网络结合遗传算法在建模和优化中的应用[J].计算机用研究,2002
[46]李烁,徐元铭.基于神经网络响应面的复合材料结构优化设计[J].复合材料学报,2005
[47]马光文,王黎.遗传算法在桁架结构优化设计中的应用[J].工程力学,1998
[48]杨建国,李蓓智,项前.基于免疫遗传计算的零件多目标优化[J].工程图学学报,2003
[49]国际水轮机修复专用机器人展暨学术研讨会论文集全部论文.中国哈尔滨,2000
[50]杨萍,李鹤岐等.水轮机修复专用机器人发展现状及应用展[J].机器人技术及应用,2001
[51]黄平,朱文坚.机械设计基础[M],广州:华南理工大学出版社,2003
[52]章成器.优化设计方法在工程机械中的应用[M],上海:同济大学出版社,1993
[53]孙靖民.机械优化设计[M],北京:机械工业出版社,2004
[54]谢维信.工程模糊数学方法[M],西安:西安电子科技大学出版社,1991
[55]周延美,蓝悦明.机械零件与系统优化设计建模及应用[M],北京:化学工业出版社,2004
[56]姚祖康著.水泥混凝土路面设计理论和方法[M],北京:人民交通出版社,2003
[57]陈屹,谢华.现代设计方法及其应用[M],北京:国防工业出版社,2004
[58]王富耻,张朝晖.ANSYS 10.0有限元分析理论与工程应用[M],北京:电子工业出版社,2006
[59]王庆五,左叻,胡仁喜.ANSYG 10.0机械设计高级应用实例[M],北京:机械工业出版社,2006
[60]小飒工作室.最新经典ANSYS及WORKBENCH教程[M],北京:电子工业大学出版社,2004
[61]安世亚太.ANSYS WORKBENCH中文培训资料
[62]高隽.人工神经网络原理及仿真实例[M].北京:机械工业出版社,2003
[63]李国勇.智能控制及其MATLAB实现[M],北京:电子工业大学出版社,2005
[64]谢庆生等.机械工程中的神经网络方法[M],北京:机械工业出版社,2003
[65]姜绍飞.基于神经网络的结构优化与损伤检测[M],北京:科学出版社,2002
[66]陈允平 王旭蕊 韩宝亮.人工神经网络原理及其应用[M],北京:中国电力出版社,2002
[67]李敏强,寇纪淞,林丹等.遗传算法的基本理论与应用[M].科学出版社.2003
[68]王小平,曹立明.遗传算法—理论、应用与软件实现[M].西安交通大学出版社,2003
[69]雷英杰,张善文,李续武等.MATLAB遗传算法工具箱及应用[M].西安电子科技大学出版社,2005
[2]梁尚明,殷国富.现代设计方法[M],化工工业出版社,2005
[3]王英杰,大型机械系统复杂构件结构优化设计方法研究:[燕山大学博士论文]:秦皇岛:燕山大学机械工程学院,2001
[4]陈科,谢守振,赵韩.基于Bol t zmann机人工神经网络的机构优化设计[J].农业机械学报,2004
[5]Charnes A,Cooper W W.Chance constrained programming.Management Science 1959,6(1):73-79.
[6]Hilton,H.H,Feigen,M.Minimum weight analysis based on structural reliability.Journal of Aerospace Science,1960,27(9):641-651.
[7]P.Pederson.Optimal Joint Positions for Space Trusses.J.Struct.Div.,ASCE,1999,(ST10):2459-2476
[8]O.C.Zienkiewicz,J.S.Campbell.Shape Optimization and Sequential Linear Programming.In:Optimal Structural Design,Edited by R.H.Gallagher and O.C.Zienkiewicz,John Wiley,New York,1973R.H.GallaghO.C.Zienkiewicz,John Wiley,New York,1973
[9]Cz I.N.Rozvany,F.Moses.Automated Design of Trusses for Optimum Geometry J.Struct.Div.,ASCE,1998(ST3):671-690
[10]王希诚,钱令希.多层次联合的结构优化设计[J].计算结构力学及其应用,1988,5(4):69-75
[11]程耿东.可靠性最大的结构优化设计[J].计算力学学报,1984,12(4):27-131.
[12]冯元生.结构余度理论及其对结构设计的影响[J].农业机械学报,1987,6(4):56-59
[13]宋天霞.复杂结构优化选型的计算方法[J].固体力学学报,1990,11(3):235-245
[14]杜轩,陈柏鸿,彭立焱,钟毅芳.机械产品的全性能优化建模及求解[J].机械设计与制造工程.2000.29(4):20-23
[15]刘仁云.基于计算智能技术的结构可靠性优化设计研究:[吉林大学博士论文]:吉林:吉林大学固体力学,2006
[16]Shahab Mohaghegh,Andrei Popa,Sam Ameri,Design optimum frac jobs using virtual intelligence techniques,Computers &Geosciences 26(2000)927-939
[17]Steven Walczak,Narciso Cerpa,Heuristic principles for the design of artificial neural networks,Information and Software Technology 41(1999)107-117
[18]Glodberg D E.,Richardson J.Genetic Algorithm with Sharing for Multimodel Function Optimization Genetic Algorithms and Their Applications Proceedings of the Second In ten rational Conference on Genetic Algorithm,1987.41-49;
[19]Lin C-Y Hajela P.Genetic algorithmsin optimization problems with discrete andintegerdesign variables[J].Eng.Opt.,1992.19(4):309-32
[20]GenM,Kwan W K,Yamazak 1G.Project scheduling using hybrid genetic algorithm withfuzzy logic controller in SCM environment[J].Tsinghua Science and Technology,2003,19(4):19-29)
[21]Li Quanyong.Optimization design of antenna structure by genetic algorithm[A]Proceedings of the CJK-OSMI[C].XID-IAN UNIVERSITY PRESS,1999:188-191
[22]Han Yingshi,Guo Pengfei.A Hybrid Genetic Algorithm for Structural Optimization with Discrete Vari-ables,Optimization of Structural and Mechanical Systems[C].Xidian
[23]王洪元 史图栋.人工神经网络技术及其应用[M],北京:中国石化出版,2002
[24]焦李成.神经网络计算[M],西安:西安电子科技大学出版社,1993
[25]王洪元 史国栋.人工神经网络技术及其应用[M],北京:中国石化出版,2002
[26]焦李成.神经网络计算[M],西安:西安电子科技大学出版社,1993
[27]杨建刚.人工神经网络实用教程[M],杭州:浙江大学出版社,2001
[28]郭代仪,廖小云,雷文宇.神经网络及其在机械工程中的应用[M],重庆:重庆大学出版社,1998
[29]靳蕃等.神经网络与神经计算机原理·应用[M],四川:西南交通大学出版社,1991。
[30]姜绍飞,贾连光,于军.基于神经网络的精架结构优化设计[J].东北大学学报(自然科学版),2003.24(2):166-169
[31]许素强,夏人伟.杆架结构拓扑优化与遗传算法[J].计算机结构力学及其应用,1994.11(4):436-446
[32]白广忱,黄洪钟.机械系统可靠性的多目标模糊优化设计[J].机械设计,1998
[33]董小兰,吴卫,李晓慧.BP神经网络及其在压铸模具优化设计中的应用[J].模具技术,2006
[34]王伟,赵美英,赵锋.基于人工神经网络的结构布局优化设计[J].机械设计,2006
[35]V.Srinivas,K.Raman janeyulu,An integrated approach for optimum design of bridge decksusing genetic algorithms and artificial neural networks
[36]Shahab Mohaghegh,Andrei Popa,SamAmeri,Design optimum frac jobs using virtual intelligence techniques,Computers &Geosciences 26(2000)927-939
[37]雷铁安.基于人工神经网络(ANN)的结构学习及应用:[北京交通大学硕士学位论文].北京:北京交通大学车辆工程,2005
[38]李雄飞,李军编.数据挖掘与知识发现[M],北京:高等教育出版社,2003
[39]张文修,梁怡编著.遗传算法的数学基础[M],西安:西安交通大学出版社,2000
[40]刘远.基于遗传算法的大型起重船拔杆式臂架优化设计研究:[天津大学硕士学位论文].天津:天津大学船舶与海洋结构物设计制造,2004
[41]谢桶,陈英俊.遗传算法的改进策略及其在桥梁抗灌优化设计中的应用效果[J].工程力学,2000
[42]张明辉,王尚锦.遗传算法在结构形状优化中的应用明[J].机械科学与技术,2001
[43]唐文艳.结构优化中的遗传算法研究和应用田[J].大连:大连理工大学,2002
[44]陈凯.混凝土泵车臂架结构的智能优化设计[J].武汉:武汉理工大学,2003
[45]闵惜琳,刘国华.人工神经网络结合遗传算法在建模和优化中的应用[J].计算机用研究,2002
[46]李烁,徐元铭.基于神经网络响应面的复合材料结构优化设计[J].复合材料学报,2005
[47]马光文,王黎.遗传算法在桁架结构优化设计中的应用[J].工程力学,1998
[48]杨建国,李蓓智,项前.基于免疫遗传计算的零件多目标优化[J].工程图学学报,2003
[49]国际水轮机修复专用机器人展暨学术研讨会论文集全部论文.中国哈尔滨,2000
[50]杨萍,李鹤岐等.水轮机修复专用机器人发展现状及应用展[J].机器人技术及应用,2001
[51]黄平,朱文坚.机械设计基础[M],广州:华南理工大学出版社,2003
[52]章成器.优化设计方法在工程机械中的应用[M],上海:同济大学出版社,1993
[53]孙靖民.机械优化设计[M],北京:机械工业出版社,2004
[54]谢维信.工程模糊数学方法[M],西安:西安电子科技大学出版社,1991
[55]周延美,蓝悦明.机械零件与系统优化设计建模及应用[M],北京:化学工业出版社,2004
[56]姚祖康著.水泥混凝土路面设计理论和方法[M],北京:人民交通出版社,2003
[57]陈屹,谢华.现代设计方法及其应用[M],北京:国防工业出版社,2004
[58]王富耻,张朝晖.ANSYS 10.0有限元分析理论与工程应用[M],北京:电子工业出版社,2006
[59]王庆五,左叻,胡仁喜.ANSYG 10.0机械设计高级应用实例[M],北京:机械工业出版社,2006
[60]小飒工作室.最新经典ANSYS及WORKBENCH教程[M],北京:电子工业大学出版社,2004
[61]安世亚太.ANSYS WORKBENCH中文培训资料
[62]高隽.人工神经网络原理及仿真实例[M].北京:机械工业出版社,2003
[63]李国勇.智能控制及其MATLAB实现[M],北京:电子工业大学出版社,2005
[64]谢庆生等.机械工程中的神经网络方法[M],北京:机械工业出版社,2003
[65]姜绍飞.基于神经网络的结构优化与损伤检测[M],北京:科学出版社,2002
[66]陈允平 王旭蕊 韩宝亮.人工神经网络原理及其应用[M],北京:中国电力出版社,2002
[67]李敏强,寇纪淞,林丹等.遗传算法的基本理论与应用[M].科学出版社.2003
[68]王小平,曹立明.遗传算法—理论、应用与软件实现[M].西安交通大学出版社,2003
[69]雷英杰,张善文,李续武等.MATLAB遗传算法工具箱及应用[M].西安电子科技大学出版社,2005