This paper aims to address the common challenges of storage overhead and computational inefficiencies that arise in isogeometric topology optimization (ITO) when dealing with large-scale problems. To tackle these issues, the paper proposes a novel framework that combines a highly efficient data storage strategy with Graphics Processing Unit (GPU) accelerated optimization. By utilizing control point pairs and removing redundant matrix storage, the Isogeometric Compressed Sparse Row (IGA-CSR) technique effectively reduces storage requirements. Furthermore, the paper presents an order-ascending optimization strategy to avoid intensive calculations caused by large degrees of freedom in the early stage. What's more, the introduction of Graphics Processing Unit further improves the optimization process. Combining these methods, an efficient optimization framework is proposed, which allows efficient optimization even for problems that involve tens of millions of degrees of freedom via single NVIDIA GeForce RTX 3090 GPU with 24 GB. Validation through two 3D benchmark examples reveals that the IGA-CSR method shows the best performance comparing with existing methods in memory consumption. At the same time, it enhances computational efficiency about 65.4 % comparing with conventional second-order isogeometric topology optimization via GPU acceleration.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号