Prediction of Mechanical Properties of Cotton Fibers by a BP Neural Network Model Optimized by Genetic Algorithm
| Autor: | Junyang Wang, Limin Zhang, Xiang Liu, Jinchan Zhang, Wanxin Wang, Hong Xu |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2024 |
| Předmět: | |
| Zdroj: | Journal of Natural Fibers, Vol 21, Iss 1 (2024) |
| Druh dokumentu: | article |
| ISSN: | 15440478 1544-046X 1544-0478 |
| DOI: | 10.1080/15440478.2024.2389161 |
| Popis: | In this experiment, a general purpose BP neural network (BP) based on genetic algorithm (GA) was developed for predicting fiber properties. The experiment is based on the breaking force of cotton fibers, and the controlled variable method is used for sampling test to collect 878 datasets containing four eigenvalues. The first 850 items of this dataset were then utilized to train the designed BP, and the remaining 28 items were evaluated for error. Next, the model is parameterized using a genetic algorithm to reduce the overall network size, thus optimizing the fit. Finally, the improved model was evaluated using the same dataset. The results were obtained: the MAPE was reduced from 10.94% to 3.7869%, the MAE was reduced from 0.39586 to 0.14584, and the MSE was reduced from 0.32161 to 0.05201. The results show that this GA-BP has better results for nonlinear fitting, and it can make a better correspondence to the outliers in the dataset, and also produces a smaller error in the fitting results, the Overall, the method is effective. |
| Databáze: | Directory of Open Access Journals |
| Externí odkaz: |
|