Relational Model and Optimization of Process Parameters for Vertical Vibratory Finishing with Multi-layer Evolutionary Neural Network

ZHANG Liaoyuan; LI Wenhui; WEN Xuejie; ZHANG Yan; LI Xiuhong; WANG Haizhu; YANG Shengqiang

doi:10.16490/j.cnki.issn.1001-3660.2025.16.011

PDF(9877 KB)

Surface Technology ›› 2025, Vol. 54 ›› Issue (16) : 131-140. DOI: 10.16490/j.cnki.issn.1001-3660.2025.16.011

Relational Model and Optimization of Process Parameters for Vertical Vibratory Finishing with Multi-layer Evolutionary Neural Network

ZHANG Liaoyuan^1,2, LI Wenhui^2,3,*, WEN Xuejie^1,2, ZHANG Yan^1,2, LI Xiuhong^1,2, WANG Haizhu^1,2, YANG Shengqiang^1,2

Author information +

History +

Abstract

Vertical vibratory finishing technology exhibits superior surface finishing performance. However, the intricate relationships between its process parameters and resulting machining outcomes remain relatively vague. To construct a relational model of high-precision vertical vibratory finishing process parameters and optimize process parameters, the work aims to take TC4 titanium alloy specimens as the research object and delve into the relational model and optimization methods for process parameters.
Orthogonal experiments were conducted with vibration frequency (f), phase difference of eccentric blocks (α), mass of upper eccentric block (m_a) and mass of lower eccentric block (m_b) as process parameter variables and the reduction rate of surface roughness (ΔRa) as the evaluation index of machining effect. The degree of impact of various process parameters on ΔRa was obtained through variance analysis. With process parameters as inputs and ΔRa as outputs, an initial relational model for process parameters was constructed through mathematical regression and neural network methods. Among them, the training environment for the neural network was set up by MATLAB (R2021a), and the structure of the initial neural network was set as 4-6-1. The optimal structure of 4-5-3-1-1 was determined by iterating the number of hidden layer nodes during the neural network training process. Furthermore, the weights and biases of each neuron in the neural network were extracted and used as genes to be optimized. These genes were then input into the Genetic Algorithm (GA) for optimization. Subsequently, the optimized network weights and biases were updated to construct the relational model of the Multi-layer Evolutionary Neural Network (GA-MLP). This relational model was further coupled with the genetic algorithm to achieve process parameter optimization. According to the research, the prediction accuracy of the relational models for process parameters through mathematical regression and traditional neural networks stands at 75.6% and 76.4%. However, the prediction accuracy of the relational model based on multi-layer evolutionary neural networks can be improved to 96.6%, and the maximum error is reduced from 22.710 to 2.750. The optimized processing parameters are as follows: vibration frequency of 25 Hz, eccentric block phase difference of 98°, upper eccentric block mass of 1.55 kg, and lower eccentric block mass of 1.8 kg. With these process parameters, the surface roughness of the specimen can be reduced from 0.976 μm to 0.311 μm, achieving a surface roughness reduction rate of 68.12%. By observing the two-dimensional and three-dimensional morphology of the specimen surface, it can be seen that the specimen has obvious wear marks before processing, the horizontal surface fluctuates greatly, and the laser scanning track is clear. After machining under the optimal process parameters, the convex peaks and wear marks of the specimen are basically removed, and the surface morphology is improved to some extent.
The multi-layer evolutionary neural network proposed exhibits higher prediction accuracy compared to traditional mathematical regression and initial neural networks. The optimized process parameters can effectively reduce the surface roughness and enhance the reduction rate of the specimen. This study offers a novel approach for constructing and optimizing the relational model of process parameters for the vertical vibratory finishing.

Key words

vertical vibratory finishing / relational model for process parameters / neural network / genetic algorithm / parameter optimization

Cite this article

EndNote

Ris (Procite)

Bibtex

Download Citations

ZHANG Liaoyuan, LI Wenhui, WEN Xuejie, ZHANG Yan, LI Xiuhong, WANG Haizhu, YANG Shengqiang. Relational Model and Optimization of Process Parameters for Vertical Vibratory Finishing with Multi-layer Evolutionary Neural Network[J]. Surface Technology. 2025, 54(16): 131-140 https://doi.org/10.16490/j.cnki.issn.1001-3660.2025.16.011

References

[1] 杨胜强, 王秀枝, 李文辉. 振动式滚磨光整加工技术的研究现状及再发展[J]. 太原理工大学学报, 2017, 48(3): 385-392.
YANG S Q, WANG X Z, LI W H.Research Status and Further Development of Vibratory Finishing Technology[J]. Journal of Taiyuan University of Technology, 2017, 48(3): 385-392.
[2] MEDIRATTA R, AHLUWALIA K, YEO S H.State-of- the-Art on Vibratory Finishing in the Aviation Industry: An Industrial and Academic Perspective[J]. The International Journal of Advanced Manufacturing Technology, 2016, 85(1): 415-429.
[3] WONG B J, MAJUMDAR K, AHLUWALIA K, et al.Effects of High Frequency Vibratory Finishing of Aerospace Components[J]. Journal of Mechanical Science and Technology, 2019, 33(4): 1809-1815.
[4] 杨胜强, 李文辉, 陈红玲, 等. 表面光整加工理论与新技术[M]. 北京: 国防工业出版社, 2011: 60-67.
YANG S Q, LI W H, CHEN H L.Surface Finishing Theory and New Technology[M]. Beijing: National Defense Industry Press, 2011: 60-67.
[5] ZHANG C, LIU W W, WANG S W, et al.Dynamic Modeling and Trajectory Measurement on Vibratory Finishing[J]. The International Journal of Advanced Manufacturing Technology, 2020, 106(1): 253-263.
[6] WAN S, LIU Y C, WOON K S.A Simple General Process Model for Vibratory Finishing[J]. The International Journal of Advanced Manufacturing Technology, 2016, 86(9): 2393-2400.
[7] 邵新智. 离心式滚磨光整加工工艺实验研究及表面粗糙度预测模型的建立[D]. 太原: 太原理工大学, 2017: 50-56.
SHAO X Z.Experimental Study on Centrifugal Roller Finishing Process and Establishment of Surface Roughness Prediction Model[D]. Taiyuan: Taiyuan University of Technology, 2017: 50-56.
[8] SHI W Y, MA Y, YANG C M, et al.Evaluation of a Regression Prediction Model for Surface Roughness of Wood-Polyethylene Composite (Wpc)[J]. Surface Review and Letters, 2017, 24(Supp02): 1850033.
[9] ZHAO X D, ZHANG X J, CHENG B, et al.Co-Optimization of Magnetic Abrasive Finishing Behaviors of Zirconium Tube Surfaces with Fe-6.5 Wt% Si/SiC Abrasives Using BP Neural Network and Response Surface Methodology[J]. Materials Today Communications, 2024, 38: 107901.
[10] LIN Y C, WU K D, SHIH W C, et al.Prediction of Surface Roughness Based on Cutting Parameters and Machining Vibration in End Milling Using Regression Method and Artificial Neural Network[J]. Applied Sciences, 2020, 10(11): 3941.
[11] KANOVIC Z, VUKELIC D, SIMUNOVIC K, et al.The Modelling of Surface Roughness after the Ball Burnishing Process with a High-Stiffness Tool by Using Regression Analysis, Artificial Neural Networks, and Support Vector Regression[J]. Metals, 2022, 12(2): 320.
[12] MALI H S, MANNA A.Simulation of Surface Generated during Abrasive Flow Finishing of Al/SiCp-MMC Using Neural Networks[J]. The International Journal of Advanced Manufacturing Technology, 2012, 61(9): 1263-1268.
[13] QI J D, ZHANG D H, LI S, et al.Modeling and Prediction of Surface Roughness in Belt Polishing Based on Artificial Neural Network[J]. Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, 2018, 232(12): 2154-2163.
[14] 李淑, 覃娴萍, 翟晓童, 等. 神经网络结构自适应研究综述[J]. 模式识别与人工智能, 2023, 36(12): 1087-1103.
LI S, QIN X P, ZHAI X T, et al.A Review on Architecture Adaptation of Neural Networks[J]. Pattern Recognition and Artificial Intelligence, 2023, 36(12): 1087-1103.
[15] 张帅, 赵鑫, 彭祥玉, 等. 基于双隐含层BP神经网络的某金矿回收率预测研究[J]. 黄金科学技术, 2024, 32(1): 170-178.
ZHANG S, ZHAO X, PENG X Y, et al.Prediction of the Recovery Rate of a Gold Mine Based on Double Hidden Layer BP Neural Network[J]. Gold Science and Technology, 2024, 32(1): 170-178.
[16] JI F, CHEN C, ZHAO Y F, et al.Optimization of Copper Electroforming Process Parameters Based on Double Hidden Layer BP Neural Network[J]. Micromachines, 2021, 12(10): 1157.
[17] 徐良, 陈燕, 韩冰, 等. 基于进化神经网络的磁粒研磨表面粗糙度预测方法研究[J]. 表面技术, 2021, 50(12): 94-100.
XU L, CHEN Y, HAN B, et al.Research on Surface Roughness Prediction Method of Magnetic Abrasive Finishing Based on Evolutionary Neural Network[J]. Surface Technology, 2021, 50(12): 94-100.
[18] AHMAD S, SINGARI R M, MISHRA R S.Modelling and Optimisation of Magnetic Abrasive Finishing Process Based on a Non-Orthogonal Array with ANN-GA Approach[J]. Transactions of the IMF, 2020, 98(4): 186-198.
[19] 槐创锋, 黄涛, 贾雪艳. 结合神经网络与遗传算法的磨抛工艺参数优化[J]. 机械科学与技术, 2021, 40(7): 1025-1030.
HUAI C F, HUANG T, JIA X Y.Optimization of Processing Parameters in Grinding and Polishing Coupling Neural Networks with Genetic Algorithms[J]. Mechanical Science and Technology for Aerospace Engineering, 2021, 40(7): 1025-1030.
[20] 刘守河, 易建业, 谢晖. 基于改进BP算法的模具零件表面抛光质量预测研究[J]. 模具工业, 2023, 49(10): 9-19.
LIU S H, YI J Y, XIE H.Research on Quality Prediction of Surface Polishing on Die and Mould Parts Based on Improved BP Algorithm[J]. Die & Mould Industry, 2023, 49(10): 9-19.
[21] AHMAD S, SINGARI R M, MISHRA R S.Tri-Objective Constrained Optimization of Pulsating DC Sourced Magnetic Abrasive Finishing Process Parameters Using Artificial Neural Network and Genetic Algorithm[J]. Materials and Manufacturing Processes, 2021, 36(7): 843-857.
[22] 王鹏朝, 李文辉, 李秀红, 等. 滚抛磨块材质对SLM成型316L不锈钢表面滚磨光整效果影响的实验研究[J]. 中国科技论文, 2024, 19(8): 928-936.
WANG P Z, LI W H, LI X H, et al.An Experimental Study on the Effect of Material about Barrel Finishing Block on Surface Barrel Finishing of SLM Forming 316L Stainless Steel[J]. China Sciencepaper, 2024, 19(8): 928-936.
[23] 牛啸, 李秀红, 王嘉明, 等. 容器尺寸对叶片竖直强制振动光整加工效果的影响[J]. 中国表面工程, 2023, 36(5): 167-178.
NIU X, LI X H, WANG J M, et al.Influence of Container Size on Processing Effect of Vertical Forced Vibration Finishing for Blades[J]. China Surface Engineering, 2023, 36(5): 167-178.
[24] LU E H, GUO Y C, ZHU X L, et al.A New Grinding Surface Roughness Measurement Method Based on Image Quality Algorithm and BP Neural Network[J]. Surface Topography: Metrology and Properties, 2022, 10(4): 045024.
[25] SINGH B, MISRA J P.Surface Finish Analysis of Wire Electric Discharge Machined Specimens by RSM and ANN Modeling[J]. Measurement, 2019, 137: 225-237.
[26] RAJPUT A S, DAS M, KAPIL S.Optimization of Surface Roughness Parameters in Hybrid-Chemo Magnetorheological Finishing (HC-MRF) Process Using Response Surface Methodology and Genetic Algorithm[J]. Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, 2024, 238(8): 3273-3286.

Funding

The National Natural Science Foundation of China (51875389, 51975399, 52075362); Central Government Guided Local Development Foundation (YDZJSX2022B004, YDZJSX2022A020)