(570ae) Integrated Simulation and Optimization Platform for System Design and Operation

Rui Huang

United Technologies Research Center

East Hartford, CT 06108

Computer-aided system simulation has been increasingly important in industrial environments particularly for the smart manufacturing in the digital age. System modeling and simulation can contribute insights to engineers with more information about the process and the consequences of the design and control decisions. There are many commercial modeling and simulation packages on the market, such as Aspen Suite for chemical processes simulation, Dymola (such as Aspen for the simulation of chemical processes or the general-purpose object-oriented modeling tool Dymola) for computational fluid simulation, etc. In addition, many companies and institutes have custom-built in-house simulation tools tailored to particular processes. 

On the other hand, more and more system design and operation tasks involve solving optimization problems. Although significant efforts are devoted in system modeling and simulation in the simulation platforms, optimization engineers usually find it hard to extract the system equations to perform equation-oriented optimization tasks. They are often left without choices to run exhaustive search or simulation-based optimization such as genetic algorithms.

In this presentation, an integrated framework for system modeling, simulation and optimization is described. This framework allows engineers to model the systems by compiling different components as the state-of-the-art practice used in Aspen and Dymola. Then a set of equations of the system is assembled in the background, and derivative information is calculated based on finite differences or automatic differentiation. In addition, different simulation solvers and optimization solvers are integrated to the backend of the platform. As a result, engineers are able to simulate the system by solving the set of system equations or optimize the system with user-defined objectives constrained by the set of system equations. A vapor compression cycle is presented as an example.

The benefit of the framework is to integrate the system modeling and simulation with the optimization. Given the system is properly modeled, systematic equation-based optimization can be performed without significant additional efforts.