Improved identification algorithm for B-spline modelling of output probability density functions

An improved modelling technique has been developed for the B-spline approximation of the output probability density functions (PDFs). By reformulating the B-spline links between the weights vector and the output PDFs, the new algorithm decomposes the previous scanning identification into two separate operations, therefore reduces the dimension of the least square calculation and improves its efficiency. A PDF control strategy is then adopted to ensure that the output PDF will follow a desired PDF. Simulation studies have been performed for the molecular weight distribution (MWD) control of a styrene polymerization process.