SCALE-UP OF CHEMICAL PROCESS MODELS:RE-CALIBRATING FUNDAMENTAL MODELS USING DATA MINING

摘要

To understanding a chemical kinetic reaction scheme, arnlarge number of experiments are often performed in arn'Lab' or small-scale reactor. The experimental results canrnbe represented by a fundamental mathematical model,rntypically a series of ordinary differential equationsrn(ODEs) that are solved numerically. When the realrnprocess is 'scaled up' to a much larger 'Plant reactor,rnsome of the underlying model's process parameters canrnchange to new, unknown values. This paper proposes arnsimple method to re-calibrate the fundamental modelrnbased on lab data to accurately represent the plantrnprocess. The method involves running the lab scalernfundamental model at a large number of design points,rnrecording the simulation predictions and derivatives withrnrespect to the model inputs. Only a few experiments needrnto be run in the plant, and those data are combined withrnthe simulations in a specially trained support vectorrnmachine (SVM) that simultaneously adjusts its parametersrnto fit both sets of data. The resulting empirical modelrngives an accurate representation of the plant scale process.