|
| Database establishment of filling bodies and strength prediction model based on GA-SVM |
|
Received:May 13, 2024
Revised:June 19, 2024
Accepted:July 02, 2024
Published Online:March 24, 2025
|
| View Full Text View/Add Comment Download reader |
| DOI： |
| KeyWord:Unconfined;compressive strength, Filler;Database, Intensity;prediction, GA-SVM;model |
| Author | Institution |
| lvfengbin |
Beijing University of Technology |
| Zhang Zhihong |
Beijing University of Technology |
| Guo Lijie |
BGRIMN Technology Group, Beijing |
| liuguagnsheng |
BGRIMN Technology Group |
|
| Hits: 1206 |
| Download times: 603 |
| Abstract: |
| The filling mining method demonstrates significant advantages in enhancing ore recovery, preventing surface subsidence, and achieving environmentally friendly mine construction. The strength characteristic of backfill plays a crucial role in determining the effectiveness of the filling process. Evaluating the compressive strength of backfill through experiments is both time-consuming and costly; however, a deep learning-based strength prediction model can effectively address this issue. This study establishes a database comprising 2338 sample data based on experimental results. Four prediction models (BP, PSO-BP, SVM, and GA-SVM) are compared and evaluated using root-mean-square error, average absolute error, and determination coefficient as evaluation criteria. The findings indicate that the GA-SVM model yields the most accurate predictions for backfill strength and can serve as an effective tool for such predictions. Furthermore, the random forest algorithm is employed to assess the importance of 13 factors influencing backfill strength; notably, curing age emerges as the most critical factor. This research offers valuable insights into predicting the compressive strength of mine backfill. |
| Close |
|
|
|