GOALI: Optimal Design and Operation of Reliable Process Systems

空气分离装置将大气分离成主要气体成分：氮气、氧气、氩气和其他含量较低的气体。空气分离厂也是许多重要行业的主要来源供应商，如医院、钢铁厂和化学精炼厂。空分厂面临的一个主要问题是设备故障可能导致产品交付中断。为了提高这些工厂和其他相关化学过程的可靠性，目前的技术水平是使用模拟工具来评估设备冗余、额外库存和预防性维护方面的不同设计方案。考虑到大量的设计选择，显然需要开发基于优化的系统工具来解决这些问题。此外，如果最终目标是设计可持续的过程系统，那么研究与其他目标的相互作用是必不可少的，例如最小成本、最小环境影响和最大安全性。一个学术研究小组和一个工业伙伴普莱克斯之间的拟议合作，旨在解决如何将可靠性纳入过程系统的最佳设计和操作的问题，这是一个在文献中很少受到关注的问题。该研究项目将为参与的研究生提供普莱克斯暑期实习机会。这个项目的基本方法和结果正在通过卡内基梅隆大学高级过程决策中心的网站传播给石油、化学和工程/软件公司。本科生参与了这项研究项目，研究团队正在开展针对高中生的基于工程的推广活动。这个研究项目背后的概念是基于一个通用的多目标优化框架，其中成本最小化和可用性最大化（定义为工厂正常运行时间的百分比）是主要目标。基于混合整数非线性规划的可用性目标最大化数学建模是本研究项目的重点，因为可靠性是本项目主要关注的问题。首先解决的是依赖于基于概率的方法的模型，这些模型主要旨在确定工艺设备的最佳冗余水平。这些模型对于过程系统的综合阶段是有用的。接下来，将考虑基于马尔可夫过程的模型的发展。这些模型更通用，更适合于详细的设计和改造，因为它们处理与时间相关的概率、库存和流程流（尽管它们在计算上很昂贵，因为它们明确地处理系统的离散状态）。还考虑了这些模型对维护计划和备件库存的扩展，目标包括环境影响、安全性、灵活性和弹性。提议的多目标优化框架将考虑其他目标，如安全性、环境影响、灵活性和弹性。

An air separation plant separates atmospheric air into its primary gas components: nitrogen, oxygen, argon and other less abundant gases. Air separation plants are also primary source providers for a number of important industries, such as hospitals, steel mills, and chemical refineries. A major issue that air separation plants face is the potential for disruption in the deliveries of their products due to equipment breakdown. The current state of the art for increasing reliability of these plants and other related chemical processes is to use simulation tools to assess different design alternatives in terms of equipment redundancy, additional inventory and preventive maintenance. Given the very large number of design alternatives, there is a clear need for developing systematic tools based on optimization that can address these problems. Furthermore, the study of interactions with other objectives, such as minimum cost, minimum environmental impact, and maximum safety, is essential if the ultimate goal is to design sustainable process systems. The proposed collaboration between an academic research group and an industrial partner, Praxair, aims to address the problem of how to incorporate reliability in the optimal design and operation of process systems, a problem that has received little attention in the literature. This research project will involve summer internships at Praxair for participating graduate students. The basic methodologies and findings from this project are being disseminated to petroleum, chemical and engineering/software companies through the website of the Center for Advanced Process Decision-making at Carnegie Mellon. Undergraduate students are involved in this research project, and the research team is pursuing engineering-based outreach activities targeted towards high school students.The concept behind this research project is based on a general multi-objective optimization framework in which the minimization of cost and maximization of availability, defined as the percentage of plant uptime, are included as major objectives. This research project focuses on mathematical modeling based on mixed-integer nonlinear programming for the maximization of the availability objective, since reliability is the major concern in this project. Models that rely on a probability based approach and are mostly aimed at determining the optimal level of redundancy of process equipment are being addressed first. These models are useful for the synthesis stage of a process system. Next, the development of models based on Markov processes are being considered. These models are more general and well suited for detailed designs and retrofits, because they handle time dependent probabilities and inventories and process flows (although they are computationally expensive, since they work explicitly with discrete states of the system). The extension of these models for maintenance scheduling and spare parts inventory is also considered, with objectives such as environmental impact, safety, flexibility and resiliency. The proposed multi-objective optimization framework will allow consideration of other objectives such as safety, environmental impact, flexibility and resiliency.