产销一体化模式下CSPS系统的优化控制

发布时间：2018-07-03 08:33

本文选题：传送带给料加工站 + 产销一体化模式　；参考：《合肥工业大学》2014年硕士论文

【摘要】：现实世界的一些生产制造业中,通常存在一类配有恒速传送带的生产中心,工件随机到达生产中心进行必要的加工。通常,为了提高生产线的柔性生产,加工站配置一个有限容量的缓冲库用于临时存储工件,这类源于福特生产线的生产模型叫做传送带给料加工站(Conveyor-Serviced Production Station,简称CSPS). CSPS可以被视为生产自动化过程中的抽象模型,例如机器人生产线。作为一类智能生产模型,CSPS被广泛用在现实生产中。因此研究该类系统的最优控制问题,具有重要的现实意义,也是工业工程中的一个重要领域。现今,由于在供应链管理(SCM)时代经济发展的影响,实时的生产可能会极大地受到影响了市场,因此本文主要研究产销一体化模式下传送带给料加工站系统的优化控制,该类系统被视为一个连接到销售中心的生产中心,加工后的成品流入销售中心的成品库,本系统的特点是随机的工件到达和顾客到达,随机的加工时间和有限的缓冲库和成品库容量。所以类似这类产销一体化模式下CSPS系统的控制策略受生产和销售水平的共同影响。本文主要目标是随机控制问题的建模和提供解决方法以得到平均和折扣准则下的最优前视控制策略。这部分首先为产销一体化模式下CSPS系统建立详细的以缓冲库和成品库剩余量为联合状态的半Markov决策过程,在系统参数精确已知的情况下可以通过数值迭代和策略迭代求解该问题。为了避免理论优化算法中“维数灾”和“建模难”的问题,文中给出基于模拟退火的Q-学习算法以推导近似解。最后仿真结果表明,通过我们建立的模型和给出的优化算法,在缓冲库和成品库容量设计合理的情况下,系统可以得到最优或者次优前视控制策略。现实生产中,生产加工站的服务率是决定系统运行效率的关键之一,因此对服务率进行合适地控制可以提高系统的生产率,减少等待时间和周期时间,从而使系统在一定条件下达到最优。因此本文研究服务率可变的产销一体化模式下CSPS系统的优化控制,主要目标是对随机优化控制问题进行建模和提供解决方案。首先,以缓冲库和成品库剩余容量为联合状态,以站点前视距离和工件服务率为行动变量,将该优化控制问题建模为半Markov决策过程。因此,可用策略迭代和基于模拟退火思想的Q-学习算法等方法进行求解,获得系统在平均准则或折扣准则下的最优前视距离和服务率。
[Abstract]:In some manufacturing industries in the real world, there is usually a kind of production center with constant speed conveyor belt. In order to improve the flexible production of the production line, the processing station is equipped with a limited capacity buffer library for temporary storage of the workpiece. This kind of production model derived from the Ford production line is called the Conveyor-Serviced production Station (CSPS). CSPS can be regarded as an abstract model in the process of production automation, such as a robot production line. As a kind of intelligent production model, CSPS is widely used in practical production. Therefore, it is of great practical significance to study the optimal control problem of this kind of systems, and it is also an important field in industrial engineering. Nowadays, because of the influence of the economic development in the era of supply chain management (SCM), the real-time production may be greatly affected by the market. Therefore, this paper mainly studies the optimization control of the belt feed processing station system under the integrated mode of production and marketing. This kind of system is regarded as a production center connected to the sales center, the finished product after processing flows into the finished product store of the sales center, this system is characterized by the random arrival of the work piece and the arrival of the customer. Random processing time and limited buffer and finished product storage capacity. Therefore, the control strategy of CSPS system is influenced by the production and sales level under the integrated mode of production and marketing. The main goal of this paper is to model the stochastic control problem and provide a solution to obtain the optimal forward view control strategy under the average and discount criteria. In this part, a semi-Markov decision-making process with buffer and finished product surplus as joint state is established for CSPS system under the integrated mode of production and marketing. This problem can be solved by numerical and policy iterations when the system parameters are known accurately. In order to avoid the problems of "dimension disaster" and "difficult modeling" in the theoretical optimization algorithm, a Q- learning algorithm based on simulated annealing is presented to derive the approximate solution. Finally, the simulation results show that the optimal or sub-optimal forward view control strategy can be obtained under the reasonable design of the capacity of the buffer library and the finished product warehouse through the proposed model and the given optimization algorithm. In real production, the service rate of the production and processing station is one of the key factors that determine the efficiency of the system operation. Therefore, proper control of the service rate can improve the productivity of the system and reduce the waiting time and the cycle time. Thus, the system can be optimized under certain conditions. Therefore, this paper studies the optimal control of CSPS system under the integrated mode of production and marketing with variable service rate. The main goal of this paper is to model the stochastic optimal control problem and provide a solution. Firstly, the optimal control problem is modeled as a semi-Markov decision process with the residual capacity of the buffer library and the finished product bank as the joint state, the forward distance of the site and the service rate of the workpiece as the action variables. Therefore, the optimal forward distance and service rate of the system under the average criterion or discount criterion can be obtained by means of strategy iteration and Q-learning algorithm based on simulated annealing.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TB49;TP273

【参考文献】