【作者】 凌琳；

【导师】 刘明周；

【作者基本信息】 合肥工业大学 ， 工业工程， 2014， 博士

【摘要】 随着经济和生产力的发展，市场需求日趋多样化和个性化，使得生产模式由传统的单一品种（或少品种）大批量逐渐转变为多品种小批量，生产作业车间内存在的各种不确定性因素对物料流影响随之增大，给车间内物料流控制带了严峻的挑战。推拉结合的TOC控制模式虽然能够通过对关键瓶颈的控制，提高物料流与生产节奏的同步性；然而，传统的TOC控制模式无法适用于由于不确定性因素而导致车间瓶颈频繁漂移的生产环境。为此，本文提出并研究了适应于不确定环境下存在瓶颈漂移现象的作业车间物料流控制方法。首先，对目前物料流控制相关邻域的研究现状进行了综述，分析了现代制造模式下作业车间环境的特点和问题，指出了现代制造环境下物料流控制方式与传统TOC控制模式的区别，将识别瓶颈和预测瓶颈漂移趋势作为物料流控制的前提。接着，围绕着不确定环境下物料流控制的核心问题，从漂移瓶颈辨识、物料投放、物料流路径以及物料流稳定性等方面进行了深入研究，具体包括：1)在明确了作业车间物料流内涵以及物料流瓶颈漂移定义的基础上，提出了导致物料流瓶颈漂移的作业车间不确定性因素变量的数字化描述方法；利用马尔科夫链、神经网络算法分析了不确定因素间耦合关系；建立物料流瓶颈综合指数体系包括瓶颈指数、瓶颈漂移指数和瓶颈敏感指数，以实现对不确定环境下作业车间内漂移瓶颈的识别和预测。2)根据作业车间内各制造单元的瓶颈状态、漂移特点以及主次瓶颈个数，提出了基于漂移瓶颈物料投放控制方法，包括对初始物料的准入控制以及制造单元间生产(转运)批量和提前期的控制；针对物料准入控制，提出了作业车间内瓶颈单元在制品上限、瓶颈间在制品上限和瓶颈上游单元在制品目标值作为准入依据；针对制造单元间的期量控制，采用物料流瓶颈综合指数表征制造单元内在制品流动节奏，建立以最小化非增值加工时间和相邻制造单元物料流瓶颈程度差异为优化目标的制造单元间期量设定模型，以保证各制造单元间物料流节奏的同步性，并利用粒子群算法求解该模型。3)在计及漂移瓶颈对物料配送路径选择影响的基础上，采用瓶颈指数和瓶颈漂移指数表征实时变化的制造单元物料配送优先级，对路径选择过程中违反此优先级的行为设置惩罚成本，提出以最小化包括车辆运输成本和违反优先级的惩罚成本在内的总配送成本为优化目标，建立时变的物料配送路径优化模型；为保证运输车辆所载物料全额配送，避免非必要负载以及由此造成的非必要配送子路径，令该模型允许运输车辆非满载和物料拆分配送，以提高物料配送效率降低配送成本；并结合模型特点将贪婪策略融入遗传算法对优化模型求解。4)对物料流鲁棒性和稳定性内涵进行了区分，以单机车间为特例建立不确定加工时间环境下的物料流稳定性控制模型并提出两种启发式算法对该NP-hard问题进行求解；在此基础上，将单机问题扩展至多制造单元作业车间，提出了两种多制造单元物料流稳定性控制策略(瓶颈单元控制策略和瓶颈上游单元控制策略)，并通过实例比较和分析了两种策略的适用范围，表明运用该稳定性控制策略能够抑制作业车间物料流瓶颈的漂移、降低物料流控制的难度。更多还原

【Abstract】 With the development of economy and productivity, the market and customer demand becomesdiversification and individuation gradually, which makes the production mode change fromsingle-item big-lot to multi-item small-lot. The influence of various uncertain factors in job shop onthe material flow increases as the production mode changes, which leads to severe challenges for thecontrol of material flow. Although TOC control mode can enhance the synchronism betweenmaterial flow and production pace by controlling the key bottleneck, traditional TOC mode can notadapt the uncertain production environment with bottleneck shifting. Therefore, this dissertationproposes and researches the control method of the material flow under uncertainty with thephenomenon of the shifting bottleneck.Firstly, the literature review of the related research field about material flow control issurveyed. The characteristics of production environment under modern production mode is analyzed,then the difference between the material flow control method under modern production mode andtraditional TOC control method is indicated. It shows that the identification and prediction of theshifting bottleneck is the premise of the material flow control.Then, focusing on the core problem of matrial flow control under uncertainty, the identificationand prediction of the shifting bottleneck, the matrial release, the route of matrial flow and thestability of the material flow are studied. The content is as follows:1) Based on the concept of the material flow and the definition of the shifting bottleneck in jobshop, the digitalization method of the uncertain factor which cause the bottleneck shift is proposed.Using Markov chain and neural network to analyze the coupling relationship among the uncertainfactors, then establish the integrated bottleneck measurement system including bottleneck index,bottleneck shifting index and the bottleneck sensitivity index, in order to identify and prodict theshifting bottleneck in job shop under uncertainty.2) According to the bottleneck status of each manufacturing unit, bottleneck shifting trend andthe number of the bottlenecks, a control method of material release based on the shifting bottleneckis proposed, which includes the control of the raw material input and the control of the transfer lotand lead time. Aming at the control of raw material input, WIP upbound of the bottleneck unit, theWIP upbound between bottleneck units and the WIP target value of upstream unit is presented as thecriterion of the material input. Aming at the control of the transfer lot and lead time, adopting thebottleneck index and the bottleneck shifting index to represent the pace of WIP flow, establish thesetting model of the transfer lot and lead time of which the objectives are minimizing thenon-value-added processing time and minimizing the difference of the bottleneck degree betweenthe adjacent manufacturing units, and use the particle swarm heuristic to solve this model.3) Based on considering the influence of the shifting bottleneck on the optimization of materialdistribution route, use the bottleneck index and the bottleneck shifting index to indicate the priorityof material distribution of each manufacturing unit, and provide the punishment cost if violate thepriority during the distribution route optimization. Hence, the objective of the established model oftime-varing material distribution route optimization is minimizing the total distribution costincluding the transportation cost and the punishment costdue to the violation of distribution priority.In order to guarantee that the material in the vehicle is distributed totally before the vehicle comesback to the distribution center and to avoid unnecessary distribution sub-route, suppose the condition that the vehicle haven’t to be loaded fully and material can be distributed to one manufacturing moretimes, which can enhance the distribution productivity and decrease the distribution cost further.Then, a greedy-based genetic algorithm is presented to solve the proposed optimization model.4) On the basis of differentiating the concept of the stability and roubustness of the materialflow, take a single machine shop as a special case to build the staility control model of the matrialflow under uncertain processing time and propose two heuristics to solve this NP-hade problem.Then generate the model and the result from a special case (single machine job shop) to the generalcase (multi-machine job shop) and present two control policy for multi-machine job shop—thecontrol policy of the bottleneck unit and the control policy of the upstream unit of the bottleneck.Further, according to the case study, analyze and compare the adaptation condition of two controlpolicies and the results shows that the two control policies can help to hedge the bottleneck shift andreduce the difficulty of the matrial flow control.更多还原