Carl Schweiger

Industrial application of nonlinear model predictive control technology for fuel ethanol fermentation process

There are currently 134 ethanol biorefineries in the United States with a production capacity of nearly 7.2 billion gallons per year, with an additional 6.2 billion gals per year capacity under the construction [1]. Approximately two thirds of these are dry-mill production facilities.

Extrapolating gain-constrained neural networks - effective modeling for nonlinear control

Nonlinear model predictive control (NLMPC) is now a widely accepted control technology in many industrial applications. Since the quality of the model of a physical non-linear process plays a critical role in the successful development, deployment, and maintenance of a NLMPC application, the mathematical representation of such models has been the subject of significant research in both academia and industry. In this paper, extrapolating gain-constrained neural networks (EGCN) is described as a key component of a NLMPC technology that has been in use in more than 100 industrial applications over the past 7 years. Simulation results are presented which compare EGCN models to traditional neural network training methods as well as to the recently proposed bounded-derivative network (BDN). These results highlight the critical advantages of EGCN in nonlinear process modeling for optimization and control applications and underscore the effectiveness of EGCN models in providing guarantees on global gain-bounds without compromising accurate representation of available process data.

Plant-wide optimization of an ethanol plant using parametric hybrid models

Ethanol plants are highly integrated systems consisting of many different processing units. As a result, the optimal operation of the ethanol plant can only be achieved if the plant model properly captures the operation of the individual units and the integrated nature of the plant components. The main challenge in plant-wide optimization, therefore, lies in the need for a compromise between accuracy and computational efficiency of (a) the constituent models of the plant, and (b) the manner by which these models are integrated. This challenge is further complicated by the fact that these models must capture true operating conditions and constraints of the plant in real-time for the optimization solution to have any chance of being implementable. This paper introduces parametric hybrid modeling as a framework for achieving a workable compromise between model complexity and computational efficiency. We represent process units as parameterized shortcut models with parameters that are empirically modeled based on actual plant data. We demonstrate the viability of our approach via a simulation study in which the parametric hybrid model of an actual ethanol plant is used to determine the optimal operation set points for the ethanol plant under different economic conditions.

Development of PUNDA (Parametric Universal Nonlinear Dynamics Approximator) Models for Self-Validating Knowledge-Guided Modelling of Nonlinear Processes in Particle Accelerators \& Industry

The difficult problems being tackled in the accelerator community are those that are nonlinear, substantially unmodeled, and vary over time. Such problems are ideal candidates for model-based optimization and control if representative models of the problem can be developed that capture the necessary mathematical relations and remain valid throughout the operation region of the system, and through variations in system dynamics. The goal of this proposal is to develop the methodology and the algorithms for building high-fidelity mathematical representations of complex nonlinear systems via constrained training of combined first-principles and neural network models.