The Use of Inverse Formulation in Design and Control of Transient Radiant Systems

This study considers the design and control of a radiant furnace by the use of an inverse formulation. The furnace considered is one that is used to heat an object from an initial temperature to a final steady state while the temperature across the heated object is kept spatially uniform. The design problem considered is a transient boundary condition estimation problem. The necessary heater distributions can be calculated by inverse design but when such a system is built, some discrepancies will result between the design based on the numerical model and what the physical system achieves. There are many reasons for differences, and the design and control engineer should correct for the discrepancies at some level. A control algorithm is developed in this study that makes use of the information provided by the inverse design and data from the physical system to train a neural net, which is then used to control the system to overcome the errors in the predicted operation.