GPU-Based Parallel Calculation Method for 1Molecular Weight Distribution of Batch Free Radical Polymerization

Abstract The rigorous model for calculating the time evolution of the molecular weight distribution (MWD) in batch free radical polymerization processes consists of a very large number of differential and algebraic equations (DAEs). A sequential variable decoupling (SVD) method has been proposed to enable the calculation in a sequential mode. With this SVD method, the calculation of MWD can be achieved for any reasonably large chain length number. In this paper, we aim to further accelerate the calculation of MWD with a parallel computing method. The algorithm analysis proves that the proposed method has the best efficiency in parallel computing. The NVIDIA Graphic Processing Unit (GPU) is applied for the parallel computing with the advantages of high parallel performance for float computing and low energy cost. The Compute Unified Device Architecture (CUDA) provided by NVIDIA is used for the programming. Running on a Tesla C 1060 GPU, the application shows that the parallel computing implementation speeds up the calculation of the MWD model by a factor of 7.3. Good agreement with the sequential result is also guaranteed.