Adaptive, hybrid modeling and optimization for design and control of startup processes

加拿大（和全球）几乎所有的生产过程都经历过瞬态：由于定期维护或不可预见的事件而导致的停机和启动。目前的启动实践依赖于食谱或一系列预先决定的行动顺序。这些政策虽然可行，但远非最佳，通过将建模、优化和反馈控制原则应用于这些过程，可以取得重大的环境和财务收益。虽然反馈和控制的概念已被普遍采用的连续操作，他们还没有被广泛用于处理启动的目的，由于两个主要原因。第一个是开发可用于优化和控制的良好模型的困难，另一个是在这种情况下自然出现的优化问题的独特性质（由于最小化例如启动时间的要求）。拟议的研究将集中在这两个方面，并提出自适应数据驱动的混合建模，优化和控制方法，并与工业合作伙伴（普莱克斯公司）合作启动氢气和空气分离装置。由于这些瞬态操作在加拿大所有生产过程中无处不在，研究结果将有利于加拿大所有制造业。

Almost all production processes in Canada (and worldwide) go through transients: shutdowns and startups resulting from scheduled maintenance or unforeseen events. Current startup practices rely on recipes or a series of pre-decided sequence of actions. Such policies, while feasible, are far from optimal, and significant environmental and financial gains can be made by applying modeling, optimization and feedback control principles to these processes. While the notions of feedback and control have been universally adopted for continuous operation, they have not been widely utilized for the purpose of handling startups due to two primary reasons. The first is the difficulty in developing good models that could be then used for optimization and control, and the other being the unique nature of optimization problems that naturally arise in this context (due to the requirement of minimizing for instance the startup time). The proposed research will focus on both these aspects and present adaptive data driven hybrid modeling, optimization and control approaches with applications to startup of a hydrogen and air separations unit in collaboration with the industrial partner (Praxair Inc). Owing to the ubiquitous presence of these transient operations in all productions processes across Canada, the research results will benefit all manufacturing industries in Canada.