Collaborative Proposal: Feedback Control Theory, Computation, and Design for Scheduling and Blending

协作提案：用于调度和混合的反馈控制理论、计算和设计

基本信息

批准号：
2026980
负责人：
Christos Maravelias
金额：
$ 27.65万
依托单位：
Princeton University
依托单位国家：
美国
项目类别：
Standard Grant
财政年份：
2020
资助国家：
美国
起止时间：
2020-10-01 至 2024-09-30
项目状态：
已结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=2026980&HistoricalAwards=false
关键词：
Collaborative Proposal Feedback Control Theory

项目摘要

The objectives of this project are to develop new theory, design methods, and computational algorithms to improve two essential chemical manufacturing operations: (i) chemical production scheduling; and (ii) raw material and final product blending. New theory is needed to establish the level of performance that can be achieved using automatic feedback and rescheduling as process measurements become available and when large process disturbances occur, such as equipment breakdowns and scheduled task delays. Computationally efficient algorithms are required to ensure the calculations can be carried out in real time; because these fast solutions may be suboptimal, a means of assuring the performance guarantees of the optimal, but slower solution, must be developed. Finally, because of the wide variety of scheduling problems that exist in the chemical processing industries, a corresponding range of optimization methods must be investigated to achieve required performance goals under process uncertainties and disturbances. While this research will target applications in both traditional and new classes of chemical production scheduling and material blending operations, the modeling, design, and solution methods developed in this research will be sufficiently general to be applied to scheduling problems arising in any manufacturing facility having production targets and constraints on materials, workflows, and inventories. A significant innovation of the proposed approach is to enable automatic rescheduling with minimal disruption on the arrival of new measurement information. This automated use of corrective feedback is absent in almost all manufacturing scheduling approaches in use today, and so this work will provide a transformative opportunity for improved business performance across many industrial sectors. The intuitive notion of online, repeated optimization of a model-based forecast as a means of designing an automatic feedback control system has now taken hold in most advanced control technologies applied in the chemical process industries as well as many other industrial sectors such as robotic motion control, flight and land vehicle guidance control, etc. The intellectual merit of the proposed research is to advance the state of the art in designing such systems and linking the design parameters to the performance and robustness properties of the closed-loop operating systems. The target applications in this proposal are characterized by discrete decisions (scheduling) and nonlinear models (blending). Designing the objective function and constraints, and demonstrating the performance under significant model uncertainty for this challenging class of applications will enhance both the underlying fundamental control theory as well as the application of these technologies to complex industrial manufacturing facilities. In batch scheduling, the assumption that all events (both decisions and disturbances) take place at an integer multiple of the sample time is often inaccurate. Therefore, a state estimation method tailored to batch scheduling that can automatically infer the state of the process from the available measurements, regardless of when an event occurs, will be developed. Finally, in the area of raw material and product blending, we face the problem of mixed-integer nonlinear programming (MINLP) models that must be solved repeatedly in real time. To develop reliable online operational capabilities for this challenging class of problems, better solution methods are required. Efforts will focus on solution methods that exploit a known, nearly feasible/optimal solution because, in the context of real-time operations, such a solution is typically available. Moreover, unlike previously proposed solution approaches, this research program will build upon tightening and reformulation methods that have been developed for MILP scheduling models.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

该项目的目标是开发新的理论、设计方法和计算算法，以改进两个基本的化学品制造操作：（i）化学品生产调度； (ii) 原材料和最终产品混合。需要新的理论来确定当过程测量可用时以及当发生大的过程扰动（例如设备故障和计划任务延迟）时，使用自动反馈和重新调度可以实现的性能水平。需要计算高效的算法来保证计算能够实时进行；由于这些快速解决方案可能不是最优的，因此必须开发一种方法来确保最佳但较慢的解决方案的性能保证。最后，由于化学加工行业中存在各种各样的调度问题，必须研究相应的优化方法，以在过程不确定性和干扰下实现所需的性能目标。虽然本研究将针对传统和新型化学生产调度和材料混合操作中的应用，但本研究中开发的建模、设计和解决方法将足够通用，可应用于任何具有生产能力的制造工厂中出现的调度问题。材料、工作流程和库存的目标和限制。所提出的方法的一个重大创新是能够在新测量信息到达时以最小的干扰实现自动重新调度。当今使用的几乎所有制造调度方法中都没有这种自动使用纠正反馈的方法，因此这项工作将为提高许多工业部门的业务绩效提供变革性机会。基于模型的预测的在线重复优化作为设计自动反馈控制系统的一种手段，这一直观概念现已在化学加工工业以及许多其他工业领域（例如机器人运动）中应用的最先进的控制技术中占据主导地位。所提出的研究的智力价值是推进设计此类系统并将设计参数与闭环操作系统的性能和稳健性联系起来的最先进技术。该提案中的目标应用的特点是离散决策（调度）和非线性模型（混合）。设计目标函数和约束，并展示此类具有挑战性的应用在显着模型不确定性下的性能，将增强底层的基本控制理论以及这些技术在复杂工业制造设施中的应用。在批量调度中，所有事件（决策和干扰）都发生在采样时间的整数倍的假设通常是不准确的。因此，将开发一种适合批量调度的状态估计方法，无论事件何时发生，该方法都可以根据可用的测量自动推断过程的状态。最后，在原材料和产品混合领域，我们面临着必须实时重复求解的混合整数非线性规划（MINLP）模型问题。为了针对此类具有挑战性的问题开发可靠的在线操作能力，需要更好的解决方法。工作重点将集中在利用已知的、几乎可行/最优的解决方案的解决方法上，因为在实时操作的背景下，这样的解决方案通常是可用的。此外，与之前提出的解决方案不同，该研究计划将建立在为 MILP 调度模型开发的紧缩和重新制定方法的基础上。该奖项反映了 NSF 的法定使命，并通过使用基金会的智力价值和更广泛的影响进行评估，被认为值得支持审查标准。