Scholarlink

<p>The nickel-based superalloy GH4169 maintains high strength at elevated temperatures but exhibits relatively poor thermal conductivity. Consequently, heat accumulates intensely during milling, resulting in rapid tool wear, work hardening, and deterioration of surface integrity. Therefore, developing an accurate model for predicting milling temperatures is essential for optimizing process parameters, improving machining quality, and enhancing production efficiency. In the work, the milling process of GH4169 was simulated through a finite element (FE) model developed in ABAQUS/Explicit, enabling the prediction of milling temperatures under varying milling parameters. The accuracy of the model was experimentally validated. To improve prediction efficiency, a novel backpropagation neural network optimized by the rime optimization algorithm (RIME-BP-NN) was developed based on the temperature data obtained from the FE simulations, using the cutting speed, feed per tooth, axial depth of cut, and radial depth of cut as input variables. The proposed model was compared with conventional NN models optimized using particle swarm optimization (PSO) and genetic algorithm (GA). The results demonstrated that the RIME-BP-NN model achieved excellent agreement with the actual values and outperformed the benchmark models across all evaluation metrics. Specifically, compared to the PSO-BP-NN model, the RIME-BP-NN model reduced the mean absolute deviation (MAD) by 20.21, the mean relative error (MRE) by 5.32%, and the mean squared error (MSE) by 883.62, while increasing the coefficient of determination (R²) by 0.027. Similarly, compared to the GA-BP-NN model, it reduced the MAD, MRE, and MSE by 16.67, 4.39%, and 654.54, respectively, and increased the R² by 0.020. Furthermore, tests on 9 sets of milling parameters confirmed the strong extrapolation performance of the model, with all prediction relative errors remaining below 14%, demonstrating its robustness and practical suitability for milling temperature prediction in industrial settings. Pearson correlation analysis indicated a positive correlation between milling temperature and the cutting speed, feed per tooth, radial depth of cut, and axial depth of cut. The cutting speed was identified as the most influential parameter, followed in descending order of significance by the feed per tooth, radial depth of cut, and axial depth of cut.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction of Milling Temperature for Nickel-based Superalloy GH4169 Using the RIME-BP-NN Model

  • 期刊论文

  • 发表日期：

    2025年12月22日

查看全文
  • Guangxu Zhu

摘要

The nickel-based superalloy GH4169 maintains high strength at elevated temperatures but exhibits relatively poor thermal conductivity. Consequently, heat accumulates intensely during milling, resulting in rapid tool wear, work hardening, and deterioration of surface integrity. Therefore, developing an accurate model for predicting milling temperatures is essential for optimizing process parameters, improving machining quality, and enhancing production efficiency. In the work, the milling process of GH4169 was simulated through a finite element (FE) model developed in ABAQUS/Explicit, enabling the prediction of milling temperatures under varying milling parameters. The accuracy of the model was experimentally validated. To improve prediction efficiency, a novel backpropagation neural network optimized by the rime optimization algorithm (RIME-BP-NN) was developed based on the temperature data obtained from the FE simulations, using the cutting speed, feed per tooth, axial depth of cut, and radial depth of cut as input variables. The proposed model was compared with conventional NN models optimized using particle swarm optimization (PSO) and genetic algorithm (GA). The results demonstrated that the RIME-BP-NN model achieved excellent agreement with the actual values and outperformed the benchmark models across all evaluation metrics. Specifically, compared to the PSO-BP-NN model, the RIME-BP-NN model reduced the mean absolute deviation (MAD) by 20.21, the mean relative error (MRE) by 5.32%, and the mean squared error (MSE) by 883.62, while increasing the coefficient of determination (R²) by 0.027. Similarly, compared to the GA-BP-NN model, it reduced the MAD, MRE, and MSE by 16.67, 4.39%, and 654.54, respectively, and increased the R² by 0.020. Furthermore, tests on 9 sets of milling parameters confirmed the strong extrapolation performance of the model, with all prediction relative errors remaining below 14%, demonstrating its robustness and practical suitability for milling temperature prediction in industrial settings. Pearson correlation analysis indicated a positive correlation between milling temperature and the cutting speed, feed per tooth, radial depth of cut, and axial depth of cut. The cutting speed was identified as the most influential parameter, followed in descending order of significance by the feed per tooth, radial depth of cut, and axial depth of cut.