基于代理模型的螺旋输送器特征值反求

螺旋输送器的动态特性设计可归结于特征值反问题的求解.针对结构参数到结构响应之间的非线性映射关系,通过一种基于神经网络代理模型的优化策略,采用正交试验设计在设计空间中选择初始样本点,构造神经网络代理模型,神经网络结合遗传算法求解,利用神经网络的非线性拟合能力和遗传算法的非线性寻优能力,引入训练后的BP神经网络预测结果作为个体适应度值,获得全局最优值及对应输入值.解决了遗传算法能获全局最优解与有限元大量结构重分析之间的矛盾,是结构反问题的一种有效求解策略.     

基于GA-BP神经网络的PDC刀头钎焊残余应力优化分析

针对PDC硬质合金基底与45钢焊后残余应力对刀头钎焊强度的影响,在对钎焊残余应力进行有限元分析的基础上,利用MATLAB平台,将BP神经网络与遗传算法应用于钎焊参数优化设计。首先设计正交试验样本,对样本进行有限元模拟获得最大钎焊残余拉应力,并作为基于遗传算法优化的BP神经网络的导师信号;利用遗传算法,把减小最大残余拉应力作为目标,得到钎焊最优参数;最后通过数值仿真验证,计算数值模拟结果与遗传算法预测结果相吻合。     

基于数值模拟的B柱加强板成形工艺优化

在对使用拼焊板拉深成形的汽车B柱加强板进行数值模拟的基础上,采用BP神经网络建立工艺参数与成形质量之间的非线性映射关系,通过多目标遗传算法NSGA-II获得最优成形工艺参数.研究结果表明:神经网络结合多目标遗传算法可以获得最优成形工艺参数,可较好地解决B柱加强板的成形问题.     

粗糙集-遗传算法-神经网络集成分类器及其在转子故障诊断中的应用研究

针对转子故障诊断问题,在综合粗糙集理论、遗传算法及神经网络学习算法各自优点的基础上,提出了一种新的粗糙集-遗传算法-神经网络（RS-GA-NN）集成分类器模型。在该模型中,利用粗糙集理论的离散和约简算法实现对样本数据的特征选取;利用神经网络实现样本特征向量与故障之间的非线性映射;利用遗传算法实现对神经网络的结构优化以使神经网络的泛化能力达到最优。利用转子故障实验台模拟了不平衡、不对中、碰摩及油膜涡动4种故障的127个样本,构建了多故障识别的RS-GA-NN集成分类器,进行了转子故障的智能诊断实验,获得了很好的效果。     

基于神经网络和遗传算法的集装箱龙门起重机结构系统优化研究

论文综合利用有限元方法、正交试验法、人工神经网络以及遗传算法对龙门起重机结构系统进行优化研究。首先利用有限元模型对结构进行灵敏度分析,确定对结构系统特性敏感的设计变量作为神经网络的输入变量。然后利用正交试验法确定神经网络训练样本,并利用有限元模型计算出样本数据,建立人工神经网络模型。最后利用遗传算法对所建立的神经网络模型寻优。     

密封条结构参数优化设计方法

为了对轿车车门密封条结构参数进行优化设计，采用遗传算法和神经网络相结合的策略，首先利用神经网络建立密封条结构设计参数与压缩负荷、应力等的非线性全局映射关系，获得求解结构优化问题所需的目标函数，然后用遗传算法进行优胜劣汰的寻优搜索运算，求出最优解。优化结果表明，椭圆形结构在壁厚为1．5mm、高度为20mm时，压缩负荷和应力能达到目标函数要求。压缩负荷和应力的优化结果与理论计算值的误差分别为7．4％、9％，因此，利用神经网络和遗传算法进行结构参数优化的方法是可行的。     

基于遗传算法和神经网络算法的吊车结构优化设计与实现

论文综合利用BP神经网络、遗传算法有限元法以及正交试验法对吊车结构系统进行优化研究。利用遗传算法和BP神经网络建立复杂结构系统动态优化的计算模型,该模型可代替系统原来的有限元模型。首先对吊车起重机结构系统进行模态分析及谐响应动力学分析,找出对结构动态特性影响最大的模态频率,再利用灵敏度分析,确定对动态特性较敏感的设计变量作为神经网络的输入变量,并利用正交试验法确定神经网络训练样本,用有限元模型计算出样本点数据,建立反映结构振动特性的人工神经网络模型,最后利用遗传算法对所建立的神经网络模型寻优,得到使结构动态性能最优的设计参数。     

应用BP网络和遗传算法的快速分析与优化

将BP神经网络和遗传算法相结合运用于快速设计的结构优化问题。利用已有的实例对神经网络进行训练达到要求精度,使神经网络能够代替有限元分析达到快速分析的目的。用神经网络计算遗传算法的适应度,对结构进行优化。把BP神经网络和遗传算法相结合进行分析和优化符合快速设计的思想。此方法运用于某拱梁的快速优化设计中效果良好。     

基于GS理论和神经网络遗传算法函数寻优的板料成形优化

基于GS理论和神经网络遗传算法函数寻优法,利用非线性有限元分析软件Dynaform,对非标准方形盒成形过程参数寻优。借助正交试验法,初步获取不同组合下的减薄率数值;基于GS理论,对获取的数据进行分析,找出影响减薄率的两个主要因素即摩擦因素和冲压速度;利用拉丁超立方抽样对选出的两个主要因素进行抽样;基于神经网络遗传算法函数寻优模型,摩擦因数和冲压速度作为输入,最大减薄率作为输出,用输入输出数据训练BP神经网络。最后,用遗传算法寻优把训练后的BP神经网络预测结果作为个体适应度值,找到函数全局最优解和对应输入值。对比优化前后的数值模拟结果可知,优化后的冲压参数可以有效提高板料成形性能。     

神经网络与遗传算法在拉延筋参数反求中的应用

以某车型前地板角支撑板的拉延工序为例,讨论BP神经网络技术与遗传算法在拉延筋几何参数反求中的综合应用问题,建立能描述反映成形效果的三个参数与半圆形拉延筋几何参数之间非线性映射关系的神经网络模型,并运用遗传算法对神经网络结构进行了优化。提出逐次局部密化样本点的样本点设计方法。该方法有助于加快神经网络的设计进程,提高神经网络的模拟精度。当训练样本数据可通过有限元法自动获得时,使用该方案则更为便利。     

柔性制造系统中机器装载计划研究

研究了FMS生产环境下的机器装载问题。机器装载问题涉及将工件工序及其所需刀具合理地安排到加工机器上，在满足某些约束的前提下，使给定的性能指标得到优化。描述了问题的整数规划模型，同时提出了基于遗传算法的解决方案。最后以实例检验了算法的有效性，提出的方案能适用于较大规模的机器装载问题。     

并联杆系机床工作空间逆向问题研究

研究了基于ｓｔｅｗ ａｒｔ平台机构的并联杆系机床工作空间的逆向问题,即怎样由预先给定的工作空间来设计机床的有关结构尺寸。     

Genetic Algorithm Modelling and Solution of Inspection Path Planning on a Coordinate Measuring Machine (CMM)

A multiple component inspection path planning problem (MCIPP) can be formulated as an optimisation problem, referred to as a non-deterministic polynomial complete problem (NP). An MCIPP consists of testing points, which will be visited by a CMM probe only once, and dummy points which are set to avoid collision and may be visited by a CMM probe more than once. This paper considers the application of genetic algorithms (GAs) acting as optimisers for optimal inspection path planning systems. The paper explores the techniques used in the GA optimal inspection path planning system. The paper also discusses the comparison of integer programming models and genetic models.     

A hybrid assembly sequence planning approach based on discrete particle swarm optimization and evolutionary direction operation

Assembly sequence planning (ASP) has always been an important part of the product development process, and ASP problem can usually be understood as to determine the sequence of assembly. A good assembly sequence can reduce the time and cost of the manufacturing process. In view of the local convergence problem with basic discrete particle swarm optimization (DPSO) in ASP, this paper presents a hybrid algorithm to solve ASP problem. First, a chosen strategy of global optimal particle in DPSO is introduced, and then an improved discrete particle swarm optimization (IDPSO) is proposed for solving ASP problems. Through an example study, the results show that the IDPSO algorithm can obtain the global optimum efficiently, but it converges slowly compared with the basic DPSO. Subsequently, a modified evolutionary direction operator (MEDO) is used to accelerate the convergence rate of IDPSO. The results of the case study show that the new hybrid algorithm MEDO-IDPSO is more efficient for solving ASP problems, with excellent global convergence properties and fast convergence rate.     

Web上支持产品设计的灵活的实例抽取技术

CBD设计支持系统采用 Web上基于实例推理与基于规则推理相结合的方法。由于知识获取瓶颈的存在 ,文中提出 ,修改过程应当由人借助于机器推理的结果来完成 ,而不是直接由机器来完成。提出的灵活的实例模板技术 ,解决了系统实例结构的可进化 ;提出的用户视图技术 ,较好地解决了实例的客户化问题。并讨论了有助于实例抽取的实例形式化问题     

Concurrent process/inspection planning for a customized manufacturing system based on genetic algorithm

Producing products with multiple quality characteristics is always one of the concerns for an advanced manufacturing system. To assure product quality, finite manufacturing resources (i.e., process workstations and inspection stations) could be available and employed. The manufacturing resource allocation problem then occurs, therefore, process planning and inspection planning should be performed. Both of these are traditionally regarded as individual tasks and conducted separately. Actually, these two tasks are related. Greater performance of an advanced manufacturing system can be achieved if process planning and inspection planning can be performed concurrently to manage the limited manufacturing resources. Since the product variety in batch production or job-shop production will be increased for satisfying the changing requirements of various customers, the specified tolerance of each quality characteristic will vary from time to time. Except for finite manufacturing resource constraints, the manufacturing capability, inspection capability, and tolerance specified by customer requirement are also considered for a customized manufacturing system in this research. Then, the unit cost model is constructed to represent the overall performance of an advanced manufacturing system by considering both internal and external costs. Process planning and inspection planning can then be concurrently solved by practically reflecting the customer requirements. Since determining the optimal manufacturing resource allocation plan seems to be impractical as the problem size becomes quite large, in this research, genetic algorithm is successfully applied with the realistic unit cost embedded. The performance of genetic algorithm is measured in comparison with the enumeration method that generates the optimal solution. The result shows that a near-optimal manufacturing resource allocation plan can be determined efficiently for meeting the changing requirement of customers as the problem size becomes quite large.     

基于蚁群算法的喷涂机器人喷枪路径规划

根据复杂曲面的几何特性,对曲面进行分片处理,并建立喷枪路径的评价函数,在每一片上进行喷涂机器人喷枪路径的规划。为了使喷涂机器人喷涂作业的时间最短,将整个曲面上的喷枪路径优化组合问题看成乡村邮递员问题(ORPP),并利用改进的蚁群算法求解ORPP问题。算法中通过在进化过程中动态调整蚂蚁的个体信息,激励蚁群搜索可行空间并发现最优解集。最后以一个复杂曲面工件为喷涂对象进行喷涂实验,结果证明了改进的蚁群算法比一般性随机算法效果更佳。     

基于混合类电磁机制算法的机械臂逆运动学解

机械臂的逆运动学问题可转化为等效的最小化优化问题,并采用数值优化方法求解。提出一种基于类电磁机制(Electromagnetism-like mechanism,EM)和模式搜索(Pattern search,PS)数值求解机械臂逆运动学问题的混合类电磁机制算法(Hybrid electromagnetism-like mechanism,HEM)。该方法利用修改的类电磁机制(Revised electromagnetism-like mechanism,REM)与模式搜索方法各自特性相互融合平衡算法对解空间的全局探索和局部开发能力,基准函数测试结果表明该算法改善了全局搜索性能及寻优解的可靠性。在此基础上以6自由度TX60型号史陶比尔工业机械臂为例,考虑以机械臂末端位姿误差构建适应度函数并采用上混合类电磁机制算法数值求解,仿真结果表明,与其他方法比较,该方法用较少的适应度值计算次数下就能搜索到期望精度的寻优解,并且能搜索到所有可能关节角逆解的潜在能力,验证了该方法的有效性。     

Detection of Landmines from Acoustic Images Based on Cepstral Coefficients

This paper introduces a cepstral approach for the automatic detection of landmines from acoustic images. This approach is based on treating the problem of landmine detection as a pattern recognition problem. Cepstral features are extracted from a group of landmine images which are transformed first to 1-D signals by lexicographic ordering. Mel frequency cepstral coefficients (MFCCs) and polynomial shaping coefficients are extracted from these 1-D signals to form a database of features, which can be used to train a neural network with the landmine features. The landmine detection can be performed by extracting features from any new image with the same method used in the training phase. These features are tested with the neural network to decide whether a landmine exists or not. The different domains are tested and compared for efficient feature extraction from the lexicographically ordered 1-D signals. Experimental results show the success of the proposed cepstral approach for landmine detection at low as well as high signal to noise ratios. Results also show that the discrete cosine transform is the most appropriate domain for feature extraction.     

基于粒子群遗传优化算法的多机器人任务分配研究

在粒子群优化算法中,引入遗传算法中的克隆算子和变异算子,提出了粒子群遗传优化算法,并将多机器人系统的任务分配问题转换为在多维解空间内寻找最优解的问题,利用粒子群遗传优化算法在此空间寻找最优解,以实现对多机器人任务的协调分配.算例仿真表明,粒子群遗传优化算法不但具有粒子群优化算法所具有的易于工程实现、计算效率高等优点,还克服了粒子群优化算法易早熟、粒子群整体收敛性差等缺点,能够解决多机器人任务分配问题.