【作者】 刘鹏；

【导师】 杜正春； 谭璜；

【作者基本信息】 上海交通大学 ， 工业工程， 2013， 硕士

【摘要】 全球半导体产业日趋成熟，尤其是日、美、西欧等先进国家和地区已经进入稳定的半导体增长时期，较之以前人们把目光更多的集中在产品技术的更新换代上，如何不断缩短半导体产品的设计周期，实现更多个性化功能，并将产品快速投入市场以满足客户需求已成为各大半导体制造商竞相追逐的目标。因此产能规划在整个半导体营运过程中起着极为重要的作用。产能规划问题长期以来一直是半导体制造业中最重要和最困难的问题之一。因为其要求在复杂的制造流程中准确预测在制品的走势，同时还必须考虑各设备的实时制造能力等约束条件。合理的产能规划能够避免千万美金的资本支出，降低运营成本、提高设备记忆人员的利用率，从而提高车间的生产效率，满足终端客户的需求。本文重点解决的是半导体制造测试单元的产能规划问题。建立设备静态产能模型，通过优化设备维护周期，提高了SORT设备探针移动速度，从而提高了设备的生产能力9%；通过增加设备运行中的平行作业的重叠时间来改善设备利用率；在不影响质量的前提下，兼顾工程试验，质量，设备能力和设备产能，合理的安排设备的停机进行预防性维护和工程实验的时间。通过上述综合措施，对设备产能进行最大化的改善，实施结果表明SORT设备产能在生产线满负荷运行状态下从159.4片/周，提高到228.2片/周，产能提升了43.1%。基于离散型仿真技术理论，建立生产系统在制品预测仿真系统WFS（Work-In-Process Forecast System）。该系统根据设备实时状态而进行动态预测在制品走势，较之最传统的按照前段投片量预测流入SORT的在制品数量，以及静态STCM（短期产能模型），动态在制品预测在准确性上得到了大大的提高。通过对比我们可以发现，WFS对于在制品走势预测的动态预测偏差一般能够控制在15%以内，准确性相比其他方法得到了大大的提升，通过更准确地在制品走势预测建立动态模拟，计算测试单元的产能需求，最后通过产能改善，找出各种产能需求下的最佳解决方案。论文研究与实践为半导体制造公司的生产线产能规划提供了有效地理论支持和实践样本。企业已经开始应用本文基本思想，开始着手研发更高级的供应链管理系统。更多还原

【Abstract】 The semiconductor industry is getting more and more mature. Instead of only focusingon the product R&D and manufacturing technology upgrade, the industry has recognized theimportance of shortening the cycle time from product design, Customization and diversity,and how to rapidly respond to the market requirement. The manufacturing capacity planningis accordingly one of the key factors to achieve that goal.The capacity planning is one of the most important and challenging areas in thesemiconductor fabrication operations because of the complicated manufacturing process andthe dynamic operation environment. Work-in-product (WIP) profile is changing all the timebecause of the constraint of the equipment capacity are dynamically shifted based on toolstatus. An appropriate capacity planning will bring the right tools to the production line in theright time and million dollars saving as the consequence. It will also enable the cost reductionby reducing the consumption of the chemical/parts and enhance the productivity byoptimizing the resource such as tool set and manpower. Eventually it will help the companyto meet the customer’s satisfaction.This article will focus on the methodology of capacity planning for a SORT unit in thesemiconductor manufacturing factory. We setup a static capacity model considering theprocess flow and the equipment unit capacity. By optimizing the equipment maintenancecycle, improving the transfer speed of the sort prober, the equipment productivity is increasedby9%; we also maximize the overlap time between each individual operation to enhance theequipment utilization; We comprehensively consider the factor of quality, technologydevelopment, equipment capability and capacity, make the appropriate arrangement forequipment maintenance and engineering experiments. By doing all the improvement above,the sort equipment capacity was significantly increased from159.4wafers per week to228.2wafers per week, the improvement rate is43.1%during the factory full load period.Based upon the discrete type of system simulation technique, we programmed a dynamic WIP Forecast System (WFS). This system is to real-time simulating the WIP profile byconsidering the equipment current status. The factory used to take front-end wafer startvolume as the only input to calculate the Sort incoming WIP,(so called static Short-TermCapacity Model, STCM). WPS is more advanced in terms of the accuracy of the WIP forecast.The vibration can be easily controlled within15%. The dynamic system enables the pre-activecapacity optimization actions in order to maximize the output.The principles and methodology discussed in this dissertation have provided strongtheoretic support and practical examples. Certain enterprises have initiated implementationon these fundamentals, and started investment on the R&D of more advanced supply chainmanagement.更多还原