半导体晶圆制造系统的瓶颈管理及调度优化研究

发布时间：2018-08-18 16:56

【摘要】：我国的半导体芯片行业随着电子信息技术的飞速发展得到了快速的崛起。在半导体芯片生产过程中最重要和最复杂的制造阶段当属半导体晶圆制造系统,它具有生产周期长、生产工艺繁杂、生产环境不确定以及生产方式反复重入加工的显著特点,并且是多重入、规模大、多产品、工艺复杂制造系统的鲜明典范,同时也是现如今世界上存在的最复杂的制造系统。因此,对半导体芯片生产厂商来说,如何有效的进行半导体晶圆制造系统的瓶颈管理与调度优化来满足快速变化的市场需求是取得行业竞争优势的关键所在。半导体晶圆制造系统的瓶颈管理与调度优化问题关系着企业实现缩短加工周期、提高设备生产率和增加晶圆产品产出量的目标实现。瓶颈管理可以分为瓶颈识别与瓶颈漂移预测两方面,它是企业迫切要解决的问题之一。同样,调度优化问题是根据半导体晶圆制造系统的有限资源进行合理配置与安排,目的是企业能满足客户需求,从而使自身赢得市场份额、获得发展与壮大。本文以半导体晶圆制造系统为研究对象,基于调控生产过程的思想,对半导体晶圆制造系统的瓶颈管理与调度优化问题做出了研究。首先,瓶颈管理中的瓶颈识别问题主要是对瓶颈识别方法探究。瓶颈概念的非统一性,使得瓶颈识别方法的研究呈现出多样化特点。本文从设备与生产系统的角度提出的综合瓶颈度的判定方法。设备角度方面给出了设备固有瓶颈度的识别方法分析,它是在前人研究的基础上对识别方法进行改进。生产系统角度是应用复杂网络理论的知识对半导体晶圆制造系统进行抽象,从而给出了复杂网络设备瓶颈度的识别方法。综合瓶颈度是将两者通过一定的转化提出的一种瓶颈识别的方法。其次,瓶颈管理中的瓶颈漂移预测问题主要是对模型建立和解析的探究。通过利用半导体晶圆制造系统日常采集的各种设备数据,采用基于支持向量机方法,对系统内漂移的瓶颈展开预测。同时,为了解决支持向量机中某些参数拟定的任意性,提出了采用粒子群算法对其进行改进优化,从而保证预测的准确可靠。最后,半导体晶圆制造系统的调度优化问题主要是基于充分利用系统瓶颈设备的思想,提出了综合投料控制策略和分层工件调度策略,从而保证晶圆产品在生产过程中使系统的多项性能达到最优。
[Abstract]:With the rapid development of electronic information technology, semiconductor chip industry in China has been rising rapidly. The most important and complex manufacturing stage in semiconductor chip production is semiconductor wafer manufacturing system, which is characterized by long production cycle, complicated production process, uncertain production environment and repeated re-processing of production mode. It is a vivid example of multi-reentry, large scale, multi-product and complex manufacturing system, and it is also the most complex manufacturing system in the world. Therefore, for semiconductor chip manufacturers, how to effectively manage and optimize the bottleneck management and scheduling of semiconductor wafer manufacturing system to meet the rapidly changing market demand is the key to obtain the competitive advantage of the industry. The bottleneck management and scheduling optimization of semiconductor wafer manufacturing system is related to the realization of the goal of shortening the processing cycle, increasing the equipment productivity and increasing the output of wafer products. Bottleneck management can be divided into two aspects: bottleneck identification and bottleneck drift prediction. Similarly, scheduling optimization problem is based on the limited resources of semiconductor wafer manufacturing system reasonable allocation and arrangement, the purpose is that the enterprise can meet the needs of customers, so that it can win market share and gain development and growth. Taking semiconductor wafer manufacturing system as the research object, the bottleneck management and scheduling optimization of semiconductor wafer manufacturing system is studied based on the idea of regulating the production process. First of all, the bottleneck identification problem in bottleneck management is mainly to explore the method of bottleneck identification. Because of the non-unity of the bottleneck concept, the research of the bottleneck identification method presents a variety of characteristics. This paper presents a method for judging the degree of comprehensive bottleneck from the point of view of equipment and production system. In the aspect of equipment angle, the identification method of the inherent bottleneck degree of equipment is analyzed, which is an improvement on the basis of previous research. The angle of production system is to abstract semiconductor wafer manufacturing system by using the knowledge of complex network theory, and the method of identifying bottleneck degree of complex network equipment is given. Comprehensive bottleneck degree is a method to identify bottleneck by transforming them to a certain extent. Secondly, the prediction of bottleneck drift in bottleneck management is mainly about modeling and analysis. Based on the data of semiconductor wafer manufacturing system and support vector machine (SVM), the bottleneck of drift in the system is predicted. At the same time, in order to solve the arbitrariness of some parameters in support vector machine, particle swarm optimization (PSO) algorithm is proposed to improve and optimize it, so as to ensure the accuracy and reliability of prediction. Finally, the scheduling optimization problem of semiconductor wafer manufacturing system is mainly based on the idea of making full use of the bottleneck equipment of the system. A comprehensive feeding control strategy and a layered workpiece scheduling strategy are proposed. In order to ensure that the wafer products in the production process to achieve the optimal performance of the system.
【学位授予单位】：天津理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN305

【相似文献】