A Diagnostic Modeling Methodology for Dual Retrospective Cost Adaptive Control of Combustion

控制技术使机器人技术应用于制造业和自动驾驶技术应用于自动驾驶汽车成为可能。控制技术提高了生产率、效率和安全性。特别具有挑战性的应用程序需要能够适应系统及其环境中不可预测变化的计算机算法。该项目将使用自适应控制来改善在燃烧过程中燃烧燃料的发动机的性能。这些发动机在世界范围内用于为电网发电。具有挑战性的问题是，尽管电力需求发生了变化，但如何更有效地燃烧燃料并减少污染。该项目开发的技术将提高电网的运行和可靠性，同时减少温室气体和烟尘颗粒的排放。该项目将涉及来自多个工程学科的学生，并将增强在该技术领域工作的未来专业人员的多样性。该项目将通过开发回顾性成本自适应控制（RCAC）的诊断建模技术，促进对反馈控制理论和实践的认识和理解。RCAC所需的建模信息涉及特定特征的存在（例如右半平面零和非线性）以及必须知道这些特征的精度（右半平面零的位置和非线性的细节）。作为RCAC的扩展，自适应和闭环识别作为双RCAC （DRCAC）同时进行。该技术依赖于高效的双二次优化算法。如果使用反馈传感器识别DRCAC不能揭示基本的建模细节，那么将使用具有更深入诊断能力的非反馈传感器来探测系统，以获取校准降低保真度模型的数据。这些模型将被DRCAC用于在线分析和闭环仿真，并在必要时修改反馈传感和驱动策略。该项目的智力目标是更深入地理解双重控制和诊断方法的发展，以便在理论和实践中促进复杂系统的自适应控制。

Control technology makes it possible to use robotics for manufacturing and autopilots for autonomous vehicles. Control technology enhances productivity, efficiency, and safety. Applications that are especially challenging require computer algorithms that can adapt to unpredictable changes in the system and its environment. This project will use adaptive control to improve the performance of engines that burn fuel in combustion processes. These engines are used worldwide to generate energy for the electrical grid. The challenging problem is to burn the fuel more efficiently and reduce pollution despite changes in the demand for electricity. The technology developed under this project will enhance the operation and reliability of the electrical grid while reducing the emission of greenhouse gases and soot particles. The project will involve students from multiple engineering disciplines and will enhance the diversity of future professionals working in this area of technology. This project will advance knowledge and understanding in the theory and practice of feedback control by developing diagnostic modeling techniques for retrospective cost adaptive control (RCAC). The modeling information required by RCAC concerns the presence of specific features (such as right-half-plane zeros and nonlinearities) as well as the accuracy with which those features must be known (locations of the right-half-plane zeros and details of the nonlinearities). As an extension of RCAC, adaptation and closed-loop identification are performed concurrently as dual RCAC (DRCAC). This technique depends on efficient algorithms for biquadratic optimization. If identification with DRCAC using feedback sensors fails to reveal the essential modeling details, then non-feedback sensors with more in-depth diagnostic capability will be used to probe the system to obtain data for calibrating reduced-fidelity models. These models will be used by DRCAC for online analysis and closed-loop simulation, and, if necessary, the feedback sensing and actuation strategy will be modified. The intellectual objective of this project is a deeper understanding of dual control and the development of the diagnostic methodology in order to facilitate adaptive control of complex systems in theory and practice.