【作者】 陈铁梅；

【导师】 胡跃明；

【作者基本信息】 华南理工大学 ， 控制理论与控制工程， 2012， 博士

【摘要】 表面贴装技术是将电子元器件组装在印刷电路板（PCB）上的一种精密组装制造技术。元器件贴装是表面贴装技术中工作量大且耗时的工序。完成该工序的核心设备是贴片机，其技术难度大，精度要求高，价格昂贵，是整条生产线的瓶颈。因此本论文对拱架式贴片机优化运行的理论和方法进行研究，从而缩短贴装时间，降低能耗和成本，提高电子制造生产效率。贴片机的运行优化问题主要涉及两个优化子问题：喂料器分配优化和元器件贴装顺序优化。这两个问题高度关联，都属于NP难的组合优化问题，求解难度大。因此本论文依据从简单到复杂的研究思路，在研究各子问题的基础上再进行贴装过程的整体优化研究。针对这些问题，论文从建立优化模型和构建有效的求解算法出发，进行了如下几个方面的研究：1、针对在元器件贴装顺序已知的前提下，研究贴片机的喂料器在喂料槽上分配优化问题。建立了以贴片头在喂料槽移动距离最小化为优化目标的喂料器分配优化模型。设计了一种基于蚁群算法与遗传算法的混合算法对贴片机的喂料器分配问题进行求解。混合算法在每次迭代中首先进行蚁群搜索，再引入遗传算法的交叉和变异算子产生多个不同的新解，并用于更新蚂蚁搜索得到的结果。混合算法利用了遗传算法的全局搜索能力来弥补蚁群算法容易收敛到局部最优解的弱点。实验数据表明该种算法能有效求解出贴片机喂料器分配问题的近优解,与单一的遗传算法相比，平均效率提高了4.48%。2、针对喂料器的位置确定的条件下，研究贴片机的元器件贴装顺序优化问题。建立了新的元器件贴装顺序的优化模型。针对该问题的路径寻优特点，把混合蛙跳算法与蚁群算法相融合对问题进行求解。在蚂蚁搜索结果的基础上，再利用混合蛙跳算法的局部深度搜索和全局信息交换策略对解进行进一步改进，从而有助于克服蚁群算法容易陷入局部最优的弱点。此外，在算法中提出了适应于贴片机实际贴装情况的分段启发函数、分段信息素以及信息素的组更新策略等多种改进方法。实验结果表明，算法具有较好的求解精度和全局搜索能力，与单一的混合蛙跳算法和蚁群算法相比，平均效率分别提高了7.89%和3.79%。3、研究了贴装过程的整体优化问题。分析了该问题的难度及其必要性，建立了符合实际贴装过程的集成优化模型。在问题求解过程中，基于分解与协调的思想，在算法每次迭代中采用禁忌算法和改进的混合蛙跳算法分别对喂料器分配优化和元器件贴装顺序进行优化，在求解一个子问题时，均采用另一个子问题的当前最好解作为优化前提，从而实现两个子问题的协调求解。实验结果表明，该算法能获得较好的贴片机贴装优化解，与混合遗传算法相比，平均效率提高了11.99%。4、为在更快速的时间内获得比上面算法更好的贴片机贴装过程优化解，进一步基于禁忌搜索算法研究贴片过程的整体优化。提出了一种带扰动和变异因子的改进禁忌搜索算法求解贴片机的整体优化。算法在传统的禁忌搜索算法的基础上，以长期记忆为基础的多元化扰动策略和块变异算子来扩大贴片机贴装顺序优化搜索空间，并结合局部下降搜索策略优化喂料器分配，最终实现贴片机贴装整体优化。实验数据表明，带扰动和变异因子的禁忌算法能快速有效的获得较好的贴片机贴装优化解。与之前所提出的整体优化算法相比，在求解质量和求解速度上有较大的优越性。最后，本文在总结现有研究成果的基础上，对表面贴装生产优化问题的未来研究作了展望。本课题来源于国家自然科学基金重点项目“面向精密电子贴装生产线的关键视觉检测与优化控制问题”(资助号：60835001)、国家自然科学青年基金项目“表面贴装中的生产优化与调度方法研究”(资助号：60804053)、广东省教育部产学研重大专项“高端全自动表面贴装成套装备研发及产业化”(资助号：2009A090100027)以及教育部博士点基金项目“电子贴装中的优化方法研究”(资助号：200805611065)资助。更多还原

【Abstract】 Surface mounting technology (SMT) is one of the precise assembly manufacturingtechnologies, in which components are placed one the surface of Printed Circuit Boards(PCBs). Surface mounting is a time-comsuming process and has a large workload.Therefore, the surface mounting machine (SMM) is a key equipment to complete the process.It includes high technical difficulties, has high precision and is very expensive. So, themachine is the bottleneck machine in the whole assembly line. In this paper, theoptimization theory and method to the mounting process of surface mounting machines withover-head gantry are considered. It is helpful to minimize the assembley time, reduce theproduction cost and improve the production efficiency of the electronic manufacturing.The mounting process optimization problem for the SMM involves two highlyinterrelated sub-problms, one is feeder assignment optimization problem and the other iscomponent mounting sequence optimaization. The two sub-problems are well-knownNP-hard and are difficult to work out. Therefore, according to the general research method,from simple to complex, subproblems are firstly focosed on and then the whole optimizationproblem is considered. For these problems, around building optimization models anddeveloping their efficient sovling algorithms, the following research work has been done:1. The feeder allocation opitimization is considered given that the components mountingsequence is known. An optimization model is presented with the objective is to minimize thetravel distance of the header equipped on the SMM. A new hybrid algorithm of ant-colonyoptimization algorithm (ACO) and genetic algorithm (GA) is proposed to solve the problem.In the algorithm, the ACO is firstly used to search a good solution in each iteration, then thecrossover operator and mutation operator of GA are used to obtain some new solutions whichare then used to update the solutions for the ACO. The algorithm takes avdvange of the GAwhich has wide global search capacity and makes up of ACO which is easily trapped in thelocal optium. The results show the algorithm could obtain satisfied near-optimal solutions tothe feed allocation optimization and the solutions make an average improvement4.48%onthose obtained by the single genetic algorithm.2. The component mounting sequence opitimization is considerd given that the feeders have been assigned to slots. A new optimization model is established for the problem.According to the path optimization characteristic of the problem, a new hybrid algorithm ofACO and shuffled frog-leaping algorithm (SFLA) is proposed to solve it. After the ACOobtain it results at each iteration, a local area deep-search and a global information exchangeprocedure of the SFLA are adapted to futher improve the obtained solution. It is helpful toovercome the weakness of being easily trapped into local optimum of the ACO. Furthermore,a few search strategies such as segmented heuristic function, segmented pheromone andgrouping pheromone update stragegy, suitable to the actual mounting situation, are proposedin the algorithm. The results show the algorithm has a global search capability,and thesolutions obtained make an improvement7.89%averagely on those by the single shuffledfrog-leaping algorithm and3.79%by the single ant colony algorithmon.3. The whole mounting process optimization including the two subproblems decribed inprevious is investigated. Firstly, the difficulties and necessity of the whole optimizationproblem are analysed, and then an optimization model is established according to thecharacteristics of the mounting process of the machine. In the problem solving approach,based on the idea of decomposition and cooperation, tabu search and improved shaffledfrog-leaping algorithm are adopted to solve the feeder assignment subproblem and thecomponent mounting sequence subproblem respectively at each iteration of the algorithm.When solving each subproblem, the best solution found so far for the other subproblem isused. By this means, the two subproblems could be cooperatively solved. Experimentalresults show that it could obtain satisfied near-optimal solutions to the whole optimizationproblem, and the solutions make an improvement11.99%on those obtained by the hybridgenetic algorithm reported in literature.4. The whole mounting process optimization and a tabu search algorithm is futherinvestegated in order to obtain better solutions within shorter time compared to the previousalgorithm proposed. A modified tabu search algorithm with diversification perturbationoperator and mutation operator is developed to solve the whole mounting processoptimizaiton problem. In order to achieve the whole optimization of problem, the algorithm isbased on the traditional tabu search algorithm, adopts a diversification perturbation procedurebased on long-term frequency information and a mutation operator to expand the search place, and a local descent search strategy is embedded into the algorithm to optimize the feederassignment subproblem. Experimental results show that the proposed algorithm could obtainsatisfied near-optimal solutions to problem in a short length of time. Compared to thealgorithm presented previously, it makes an improvement on both solution quality andcomputation speed.In the end, the whole research in the dissertation is summarized and the futureinvetigation on optimization problem in surface mount manufacturing is presented.The dissertation is supported by the State Key Program of National Natural ScienceFoundation of China “Research on the Key Technology of Vision Detection and OptimalControl Oriented to the Precise Electronic Assem bly Lines”(Grant No.60835001), theNational Natural Science Foundation of China “Research on Production Optimization andScheduling Method in Surface Mounting Lines”(Grant No.60804053), the Key ResearchCooperation Project of Guangdong Province and the Ministry of Education “Research anddevelopment and industrialization of high-end automatic SMT equipments”(Grant No.2009A090100027) and the National Research Foundation for the Doctoral Program of HigherEducation of China “Research on Optimization in Electronic assembly”(Grant No.200805611065).更多还原