Kubišová Milena, Pata Vladimír, Měřínská Dagmar, Škrobák Adam, Marcaník Miroslav
Faculty of Technology, Tomas Bata University in Zlín, Vavrečkova 275, 760 01 Zlín, Czech Republic.
Materials (Basel). 2021 May 17;14(10):2620. doi: 10.3390/ma14102620.
This work deals with investigative methods used for evaluation of the surface quality of selected metallic materials' cutting plane that was created by CO and fiber laser machining. The surface quality expressed by Rz and Ra roughness parameters is examined depending on the sample material and the machining technology. The next part deals with the use of neural networks in the evaluation of measured data. In the last part, the measured data were statistically evaluated. Based on the conclusions of this analysis, the possibilities of using neural networks to determine the material of a given sample while knowing the roughness parameters were evaluated. The main goal of the presented paper is to demonstrate a solution capable of finding characteristic roughness values for heterogeneous surfaces. These surfaces are common in scientific as well as technical practice, and measuring their quality is challenging. This difficulty lies mainly in the fact that it is not possible to express their quality by a single statistical parameter. Thus, this paper's main aim is to demonstrate solutions using the cluster analysis methods and the hidden layer, solving the problem of discriminant and dividing the heterogeneous surface into individual zones that have characteristic parameters.
这项工作涉及用于评估通过CO2和光纤激光加工产生的选定金属材料切割平面表面质量的研究方法。根据样品材料和加工技术，对由Rz和Ra粗糙度参数表示的表面质量进行了研究。下一部分讨论了神经网络在测量数据评估中的应用。在最后一部分，对测量数据进行了统计评估。基于该分析的结论，评估了在已知粗糙度参数的情况下使用神经网络确定给定样品材料的可能性。本文的主要目标是展示一种能够为异质表面找到特征粗糙度值的解决方案。这些表面在科学和技术实践中很常见，测量它们的质量具有挑战性。这种困难主要在于无法通过单个统计参数来表达它们的质量。因此，本文的主要目的是展示使用聚类分析方法和隐藏层的解决方案，解决判别问题并将异质表面划分为具有特征参数的各个区域。
