Enhancing Chemical Process Simulation through a GPU-Optimized Framework: Implementation and Validation of Equation-Oriented Methods using CUDA

• Published in: Computer Aided Chemical Engineering Vol. 53; pp. 3271 - 3276
• Main Authors: Yang, Shaoyi, Tian, Xufei, Qu, Shifeng, Duan, Zhaoyang, Yang, Minglei, Qian, Feng, Du, Wenli
• Format: Book Chapter
• Language: English
• Published: 2024
• Subjects: CUDA; Equation-Oriented; GPU; nonlinear solver; CUDA; GPU; Equation-Oriented; nonlinear solver
• ISBN: 9780443288241; 0443288240
• ISSN: 1570-7946
• DOI: 10.1016/B978-0-443-28824-1.50546-9

In the realm of chemical engineering, the simulation of complex process systems frequently entails solving interconnected equations on a large scale. Traditional methods are difficult to handle higher-dimensional computing, particularly hose necessitating increased real-time processing capabilities. This study aims to significantly enhance the computational efficiency of equation-oriented (EO) methods by leveraging Graphics Processing Units (GPUs), primarily utilizing the CUDA programming paradigm exploit the parallel processing capabilities of GPUs. We present a novel GPU-based framework tailored for chemical process simulations. This framework segments the thermodynamic computations and other complex tasks within the EO method, using GPU's massive parallel capabilities to simultaneously update each thermodynamic parameter in batches, significantly enhancing the overall efficiency of parameter computation. This work first demonstrates the tasks split method within the EO structure, followed by the utilization of CUDA graph on GPUs for parallel computation of thermodynamic parameters after the breakdown of tasks. Extensive result analyses are provided, validating that using a consumer-grade GPU achieve nearly a 100-fold enhancement in performance, while preserving identical accuracy.