|
| Randomized balancing factor C algorithm and its application research |
| Received:September 13, 2023 Revised:November 08, 2023 |
| View Full Text View/Add Comment Download reader |
| DOI：10.7511/jslx20230912002 |
| KeyWord:data-driven computational mechanics balance factor randomized value selection algorithm convergence accuracy |
| Author | Institution |
| 冯凡丁 |
武汉大学 土木建筑工程学院, 武汉 |
| 姜清辉 |
武汉大学 土木建筑工程学院, 武汉 |
|
| Hits: 58 |
| Download times: 30 |
| Abstract: |
| The Randomized Balancing Factor C Algorithm was proposed to address the issue of inappropriate selection of the balancing factor C,which can lead to local optima or low convergence accuracy in data-driven computational mechanics.In this algorithm,the balancing factor C was no longer a predefined constant value,but was determined based on clustering the strains or stresses in the database and calculating the tangent stiffness of each cluster.The minimum and maximum tangent stiffness values within each cluster were used as the range of C values,and a random balancing factor was employed in each iteration.The effectiveness of the randomized balancing factor C algorithm was evaluated through a comparative analysis in the context of nonlinear elastic truss problems,as well as linear and nonlinear plane problems.Iteration paths of representative points and strain-stress relative errors were compared under different predefined C values and the randomized algorithm.The results indicate that,under the same database capacity,the randomized balancing factor C algorithm achieves better convergence accuracy in the various elastic problems. |
|
|
|