Abstract:To address the poor maintainability of traditional bridge finite element software caused by process-oriented development, and the difficulty in balancing object-oriented design with computational efficiency in high-performance computing scenarios, a software architecture that collaboratively integrates three paradigms of object-oriented, performance optimization and distributed computing is studied and implemented. Firstly, a highly modular object-oriented data structure is constructed based on C#. And then, a global matrix pool technique is innovatively introduced to manage memory, enabling the reuse and efficient scheduling of large objects such as stiffness matrices. Finally, a distributed solver is integrated to achieve parallel assembly and solution of large-scale matrices. Validation using a long-span prestressed bridge structure as a case study demonstrates that the architecture significantly enhances the maintainability and extensibility of the software. The matrix pool technique effectively reduces memory allocation times by over 90%, shortening total computation time by approximately 30~40%, overcoming the performance bottleneck of managed languages. And the distributed computing paradigm endows the software with large-scale solving capabilities beyond single-machine limits, which validates the effectiveness of the “multi-paradigm integration” architecture in both theory and practice, providing a feasible technical path for the development of modern CAE software that is independently controllable, easy to maintain and has excellent performance.
