单轨约束条件下多机器人柔性制造单元的建模与调度方法研究

发布时间：2018-05-06 05:00

本文选题：柔性制造 + 多机器人　；参考：《广东工业大学》2016年博士论文

【摘要】：随着我国智能制造战略的实施,机器人柔性制造单元被越来越多的制造企业所采用。针对模具这类单件小批定制生产问题,机器人柔性制造单元需要进行实时调度以适应外部环境变化以及模具零件类型、加工工艺和工时等随订单变化的需求。目前柔性制造单元的理论和方法面临以下挑战,模型的结构和规模随着工件、机床或机器人的变化而变化导致建模繁琐复杂；大量工时短、交货急的工件所需要的柔性搬运对单轨上多移动机器人的争夺混乱；给出初始生产计划后,缺少含死锁控制和纠正计划功能的仿真系统进行评估和反馈等方面。本文在国家自然科学基金(51105082)的资助下研究单轨约束条件下多移动机器人柔性制造单元的建模和调度问题。结合模具制造的特点,提出随机令牌智能变迁(STITPN)模型理论,探讨单轨上多机器人运动冲突消解问题,实现包含死锁控制策略和纠正计划功能的状态推演仿真算法,创造多维空间侦察机制及算法优化生产性能,单轨多机器人柔性制造单元决策支持系统的实施,有助于提高制造单元效率。本文的主要工作包含以下几个方面：1、分析模具零部件生产的特点,抽象出单轨约束条件下多机器人柔性制造单元的问题,针对主要的不确定因素进行分析,考虑了工件离散随机到达、工序加工时间随机分布、机床可用和休整的随机性等因素,最后提出研究的框架。2、针对传统Petri网模型难以表达模型结构和规模随工件、机床或机器人的变化而变化的问题,提出随机令牌智能变迁petri网(STITPN)理论。首先,将工件、机床和机器人及其组合建模为令牌,接着根据情景演化路径建模库所和变迁,然后建立系统状态推演机制,Petri网中令牌携带随机实时信息并可以停留在变迁中,变迁根据令牌的实时信息进行触发。基于STITPN理论的制造单元模型,以后只需不断输入生产任务及配置参数即可推演得到符合约束条件的调度方案。本文用实例演示建模和状态推演过程,输入经典案例数据进行计算验证模型理论的有效性。3、在柔性制造单元的STITPN模型中,多机器人令牌需要主动决定去向,然而在单轨约束条件下的多机器人柔性搬运存在冲突问题。为此建立多机器人的运动方程模型、冲突判据模型和消解策略。当驱动事件发生时实时规划机器人运动方程,通过联立多个方程求解来判断是否有冲突。提出冲突消解策略,即机器人的实时规划路径有冲突时需废弃等待下一驱动事件,多个机器人同时启动但有冲突时按工件优先级选择执行,对冲突路径中的静止机器人规划运动并循环检测冲突等策略。最后,用实例计算和案例应用验证方法的有效性。4、由于初步的总体生产计划普遍只考虑工件及机床,没有考虑机器人的搬运,需要柔性制造单元的STITPN模型进行推演仿真。在资源高度共享中存在死锁阻塞问题,计划也可能需要纠正。针对该实际问题,提出变迁的智能函数、死锁控制策略和状态推演等算法。在变迁的智能函数中,通过检测令牌的可移动方向判断是否满足触发条件,然后按优先级移动令牌；为了避免死锁,工件令牌先预测资源占用情况再决定是否进入搬运变迁；根据确定好工序选择的机床和各工序的处理次序,在模型运行中根据期望时间触发事件,使能变迁并触发,由状态推演路径得到完善后的机床加工计划和机器人搬运规划。5、在生产计划优化方面,由于工序选择机床和工序先后排列存在组合爆炸性问题,本文采用元启发式智能算法,由侦察队伍评估和搜索目标所启发,提出多维空间侦察机制及优化算法。在评估目标方面,将多维空间点解码为车间计划,输入柔性制造单元的STITPN模型中进行推演得到的生产性能为适应度值。在快速收敛方面,在各士官所代表的优化点的各个方向进行随机扰动寻求最优值。在避免过早落入局部最优值,扩大搜索范围方面,求出多个信息交互指挥官之间的中心点,在代表指挥官、中心、上限和下限的四个点之间形成新侦察点。本文把经典案例数据作为生产任务进行优化计算,并与典型的算法DE(差分进化算法),ElitistES(进化策略),gGA(遗传算法)和PSO(粒子群优化算法)等进行比较,本文提出的算法在平均最优值方面比DE和PSO好；在最优值标准差方面比ElitistES,gGA和PSO好,说明本文提出的侦察算法是很有竞争力的。6、分析了企业应用背景和制造单元动态调度策略,设计了系统的功能模块和数据库,开发了一套单轨多机器人柔性制造单元决策支持系统,用于合作企业辅助生产车间活动,应用效果良好。
[Abstract]:With the implementation of China's intelligent manufacturing strategy, more and more manufacturing enterprises are adopted by the flexible manufacturing unit of robot. Aiming at the problem of small batch production of single piece of mold, the robot flexible manufacturing unit needs real-time scheduling to adapt to the change of the external environment and the type of the die zero, the processing technology and the working hours and so on with the order change. At present, the theory and methods of FMS are facing the following challenges. The structure and size of the model changes with the changes of the workpiece, machine tool or robot, which leads to complicated and complicated modeling; the flexible handling required by a large number of workhours and quick delivery of the workpiece is scramble for the chaos of the multi shift robot on the monorail; the initial production meter is given. After delimit, the simulation system with the function of deadlock control and correction plan is lack of evaluation and feedback. Under the support of the National Natural Science Foundation (51105082), this paper studies the modeling and scheduling problem of the flexible manufacturing unit of a multi mobile robot under the condition of monorail constraints. ITPN) model theory, discuss the problem of multi robot movement conflict resolution on Monorail, realize the state inference simulation algorithm including the deadlock control strategy and the correction plan function, create multi-dimensional space reconnaissance mechanism and algorithm to optimize the production performance. The implementation of the decision support system for the flexible manufacturing unit of monorail robot is helpful to improve the manufacturing unit. The main work of this paper includes the following aspects: 1, analyzing the characteristics of the production of mould parts and parts, abstracting the problem of the multi robot flexible manufacturing unit under the condition of monorail constraint, analyzing the main uncertain factors, considering the discrete random arrival of the workpiece, the random distribution of the processing time of the working sequence, the available and the rest of the machine tool. In the end, the frame.2 of the research is put forward. In view of the problem that the traditional Petri net model is difficult to express the model structure and the scale change with the workpiece, machine tool or robot, the stochastic token intelligent change Petri net (STITPN) theory is proposed. First, the workpiece, machine tool and robot and their combinations are modeled as token, and then the scene is performed. The transformation of the path modeling library and the change, and then establish the system state inference mechanism, the token carrying random real-time information in the Petri network can stay in the change, the change is triggered according to the token's real-time information. Based on the STITPN theory, the manufacturing unit model can be deduced to be conformed only by the continuous input of the production task and configuration parameters. In this paper, an example is used to demonstrate the modeling and state inference process, and the validity of the model theory is.3. In the STITPN model of FMS, the multi robot token needs to decide the direction actively, but there is a conflict in the flexible handling of the multi robot under the condition of the monorail. To solve this problem, the motion equation model of multi robot, the conflict criterion model and the resolution strategy are established. When the driving event occurs, the motion equations of the robot are planned in real time and the conflict is determined by solving the multiple equations. The conflict resolution strategy is proposed, that is, the next driver is abandoned and waiting to be abandoned when the robot's real-time routing path is conflict. At the same time, a number of robots are started at the same time but are chosen to execute according to the priority of the workpiece in conflict, plan the motion of the static robot in the conflict path and circulate the conflict detection strategy. Finally, the validity of the method is.4 with example calculation and case application. Because the initial overall production plan is generally only considered the workpiece and the machine tool, there is no consideration. The handling of the robot requires the STITPN model of FMS to carry out the simulation. There is a deadlock blocking problem in the high sharing of resources. The plan may also need to be corrected. For the actual problem, an intelligent function, a deadlock control strategy and a state inference algorithm are proposed. In the changing intelligent function, the detection token is detected. The moving direction determines whether the trigger condition is satisfied and then the token is moved by priority; in order to avoid the deadlock, the work token first predicts the occupancy of the resource and determines whether it enters the transfer change; according to the processing order of the machine tool and the various processes selected for the selection of the good procedure, the event is triggered in the model running according to the expected time, so that the change can be changed and Triggered by the state deduction path, the machine tool processing plan and robot handling planning.5 are improved. In the aspect of production planning optimization, because of the problem of combinatorial explosions in the process selection of machine tools and processes, this paper uses the meta heuristic intelligent algorithm, which is inspired by the evaluation of the reconnaissance team and the search target, and puts forward the multi-dimensional space reconnaissance. The multi-dimensional space point is decoded into a workshop plan, and the production performance obtained in the STITPN model of the flexible manufacturing unit is considered as the fitness value. In the fast convergence, the optimal value is sought in the direction of the optimization points represented by the officers in each direction. In terms of local optimal value and enlarging the range of search, the center points between multiple information interactive commanders are found, and new reconnaissance points are formed between four points representing commanders, centers, upper limits and lower limits. This paper optimizes the classic case data as a production task, and the typical algorithm DE (differential evolution algorithm), ElitistES (Evolution). Strategy), compared with gGA (genetic algorithm) and PSO (particle swarm optimization algorithm), the algorithm proposed in this paper is better than DE and PSO in the average optimal value. It is better than ElitistES, gGA and PSO on the standard deviation of the optimal value. It shows that the reconnaissance algorithm proposed in this paper is a competitive.6, and the enterprise application background and the dynamic scheduling policy of the manufacturing unit are analyzed. The function module and database of the system are designed, and a set of decision support system for MP flexible manufacturing unit is developed, which is used in the cooperative enterprise to assist the production workshop, and the application effect is good.

【学位授予单位】：广东工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP242;TH165

【相似文献】