Distributed/boundary model predictive control for transport-reaction systems (crystal growth processing)

最先进的半导体加工，制药，石化和化学工艺工厂的操作改进在于优化和控制器合成领域，可以满足更严格的环境法规，提供所需的产品规格，并确保在能源成本增加和全球竞争时的利润。在所有这些过程中，目标是设计一个“最佳”控制器，即模型预测控制器（MPC），它应该：1)为过程工厂调节提供最优解决方案；2)包括自然存在的限制（例如，对用于过程调节的可用能量的限制——打开阀门加热或冷却过程，由于产品规范规定的性能要求、安全或危害考虑而产生的限制）；3)考虑致动器/传感器的移动和扫描能力，以优于标准的固定致动器/传感器配置。本研究计划的理论和工业应用相关目标是探索三个相关的例子，它们占目前半导体晶体加工行业中95%的工艺：1) Czochralski晶体生长过程，2)Bridgman晶体生长过程，3)熔化区精炼过程，并在上述每个过程中探索主导输运现象，并设计基于模型的相关模型预测控制器，以改进当前工业应用的控制器。该计划的预期结果是提供一个最终的最优控制器，可以处理高成本的晶体生长处理，同时节省资本成本，降低化学成本，提高产品质量，并提供更有效的电力使用。除其他好处外，该计划将通过博士和硕士学生的积极参与，培养能够掌握半导体晶体生长加工控制的高素质人才，并将吸引对工艺改进感兴趣的半导体行业当前成员的注意。

The state-of-the-art semiconductor processing, pharmaceutical, petrochemical and chemical process plants operational improvement lies within the realm of optimization and controller synthesis which can meet tighter environmental regulations, provide desired product specifications and ensure profit at the time of increasing energy costs and worldwide competition. In all these processes, the goal is to design a ``best'' possible controller, a model predictive controller (MPC), which should: 1) provide an optimal solution for the process plant regulation; 2) include naturally present constraints (e.g., limitations on available energy used for process regulation-valves opening to heat up or cool down the process, limitations due to the performance requirements imposed by the product specifications, safety or hazard considerations); 3) consider the moving and scanning capability of an actuator/sensor so as to outperform standard fixed actuator/sensor configurations. The theoretical and industrial application related objectives of this research program are to explore three relevant examples which account for 95% of the processes present in the current semiconductor crystal processing industry: 1) Czochralski crystal growth process, 2) Bridgman crystal growth process, 3) melted zone refining process, and within each of the above processes to explore dominant transport-phenomena and to design relevant model based model predictive controllers which should improve upon the current industrially applied controllers. The expected outcome of this program is to provide an ultimate optimal controller that can handle high cost crystal growth processing, and at the same time result in capital cost savings, chemical cost reduction, improved quality of products, and provide for more efficient use of power. Amongst other benefits, the program will provide for the training of highly qualified personnel by active involvement of PhD and MS students capable of mastering the control of semiconductor crystal growth processing, and it will draw attention of current members of the semiconductor industry interested in the process improvements.