Research Initiation: Nonlinear Predictive Process Control

化学制造过程的经济可行性取决于以某种规定的方式维护和调节系统。在保持产品规格的同时最大限度地减少能源或自然资源消耗的能力是通过使用过程控制系统实现的。该项目的目标是开发使用化工过程的非线性动态模型作为反馈控制基础的控制系统技术。将开发一种非线性参数估计程序来估计重要的过程参数，如效率系数或换热系数，以及过程扰动，如进料流组成。除了参数估计，还将使用基于非线性优化的程序来保持期望的输出变量轨迹，类似于已被证明对线性系统成功的技术。该过程模型由一组非线性微分方程组组成，利用有限元上的正交配置将其转化为代数方程。将序列二次规划用于求解具有代数方程等式约束和操纵变量和状态变量不等式约束的非线性优化问题。这种非线性控制策略将导致对重要的过程变量进行更严格的调节，从而实现化学过程的更优化操作。这项研究还将导致对线性预测控制技术的更全面的理解，因为许多建议研究的重要影响以前没有针对线性系统的情况进行处理。

The economic viability of a chemical manufacturing process is dependent on maintaining and regulating the system in some prescribed manner. The ability to maintain product specifications while minimizing energy or natural resource consumption is accomplished by the use of process control systems. The goal of this project is to develop control system techniques that use a nonlinear dynamic model of the chemical process as a basis for feedback control. A nonlinear parameter estimation procedure will be developed to estimate important process parameters such as effectiveness factors or heat transfer coefficients, as well as process disturbances such as feed stream composition. In addition to parameter estimation, a nonlinear optimization-based procedure will be used to maintain a desired output variable trajectory, similar to techniques that have been proven successful for linear systems. The process model, characterized by a set of nonlinear differential equations, will be transformed to algebraic equations using orthogonal collocation on finite elements. Sequential quadratic programming will be used to solve the nonlinear optimization subject to the algebraic equation equality constraints and manipulated and state variable inequality constraints. This nonlinear control strategy will result in tighter regulation of important process variables, allowing a more optimal operation of chemical processes. This research will also result in a more complete understanding of linear predictive control techniques, since many of the important effects proposed for study have not been previously addressed for the linear system case.