arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.28160cs.CEcond-mat.mtrl-sci
收藏

一种高效的开源框架用于铣削加工中高保真的3D表面形貌及粗糙度预测

An efficient open-source framework for high-fidelity 3D surface topography and roughness prediction in milling

Centre Internacional de Mètodes Numèrics a l’Enginyeria (CIMNE)(国际数值工程方法中心);Universitat Politècnica de Catalunya (UPC)(加泰罗尼亚理工大学);Eurecat, Centre Tecnològic de Catalunya(埃雷卡特加泰罗尼亚技术中心);Grupo Sevilla Control(塞维利亚控制集团);Universitat de Vic-Universitat Central de Catalunya (UVic-UCC)(维克大学-加泰罗尼亚中央大学)

Hadi Bakhshan, Sima Farshbaf, Adrián Travieso-Disotuar, Luciano Mijaíl Villarreal, Fernando Rastellini Canela, Josep Maria Carbonell

AI总结:

本文提出一种高效的开源框架,利用数值方法生成高保真的3D表面形貌及粗糙度指标,通过优化算法提升预测精度和计算效率,实验验证显示预测误差可接受且计算速度提升达42.2倍。

摘要:

随着数据驱动方法在科学领域的兴起,其在制造领域的应用,特别是表面精密工程,日益受到关注。然而,由于实验生成大规模数据集的成本高且测量耗时,生成大型数据集往往不切实际。高保真的合成数据集如果能高效且准确地生成,则是一个可行的替代方案。为了解决这一挑战,本文提出了一种高效的框架,利用数值方法生成铣削加工中的准确3D表面形貌和粗糙度指标。首先,基于正向解法(FSM)开发了传统形貌预测模型。在此基础上,提出了一种优化的计算算法,以建立性能显著提升的高效FSM。该模型通过两个独立的实验结果集进行验证,评估了预测精度和计算效率。结果表明,预测误差可接受,平均计算速度提升达42.2倍。所提出的开源模型提供了一个可扩展的框架,用于大规模分析,能够生成大量数据用于数据驱动的替代模型。

英文摘要:

With the emergence of data-driven approaches in science, there is growing interest in their application to manufacturing, particularly in surface precision engineering. However, generating large datasets required for model training is often impractical experimentally due to high costs and the time-intensive nature of measurements. High-fidelity synthetic datasets offer a viable alternative if they can be generated both efficiently and accurately. To address this challenge, this paper presents an efficient framework for generating accurate 3D surface topographies and roughness indicators in milling operations using numerical methods. First, a conventional topography prediction model is developed based on the forward solution method (FSM). Building on this, an optimized computational algorithm is proposed to establish an efficient FSM with significantly improved performance. The model is validated against two independent sets of experimental results, assessing both prediction accuracy and computational efficiency. The results demonstrate acceptable prediction errors and an average computational speedup of 42.2x. The proposed open-source model provides a generalizable framework for large-scale analysis, enabling the generation of extensive datasets for data-driven surrogate modeling.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑