Investigation of surface roughness in turning unidirectional GFRP composites by using RS methodology and ANN
Investigation of surface roughness in turning unidirectional GFRP composites by using RS methodology and ANN
复制标题
DOI：
10.1007/s00170-005-0175-x
复制
发表时间：
2006-11-01
影响因子：
3.4
通讯作者：
Isik, Birhan
中科院分区：
文献类型：
--
作者：
Bagci, Eyup;Isik, Birhan
纤维增强塑料（FRP）是一种力学性能和热学性能截然不同的两相材料，在加工过程中基体与增强体之间存在着复杂的相互作用。表面质量和尺寸精度将在零件的使用寿命期间对零件产生很大影响，特别是在零件在使用寿命期间与其他元件或材料接触的情况下。因此，对它们的研究和定性极为重要，尤其是那些受到不利环境条件和与其他元素或材料接触的情况。因此，测量和表征表面特性是制造过程中最重要的方面之一。本文采用正交试验法，用金属陶瓷刀具对单向玻璃钢进行了切削试验。在测试期间，改变切割深度（a）、进给速率（f）、切割速度（Vc），而切割方向保持平行于纤维取向。采用统计学三水平全因子试验设计技术设计车削试验。建立了车削表面粗糙度的人工神经网络（ANN）和响应面（RS）模型。在预测模型的开发中，切削速度，切削深度和进给量的切削参数被认为是模型变量。预测模型所需的数据是通过进行一系列车削试验和测量表面粗糙度数据获得的。预测模型的结果和实验测量之间观察到良好的一致性。GFRP车削零件表面的ANN和RSM模型的精度和计算成本进行了比较。
Fibre reinforced plastics (FRP) contain two phases of materials with drastically distinguished mechanical and thermal properties, which brings in complicated interactions between the matrix and the reinforcement during machining. Surface quality and dimensional precision will greatly affect parts during their useful life especially in cases where the components will be in contact with other elements or materials during their useful life. Therefore, their study and characterisation is extremely important and, above all, those cases subjected to adverse environmental conditions and in contact with other elements or materials. Thus, measuring and characterising surface properties represent one of the most important aspects in manufacturing processes. In this paper, orthogonal cutting tests were carried out on unidirectional glassfibre reinforced plastics (GFRP), using cermet tools. During the tests, the depth of cut (a), feedrate (f), cutting speed (Vc) were varied, whereas the cutting direction was held parallel to the fibre orientation. Turning experiments were designed based on statistical three level full factorial experimental design technique. An artificial neural network (ANN) and response surface (RS) model were developed to predict surface roughness on the turned part surface. In the development of predictive models, cutting parameters of cutting speed, depth of cut and feed rate were considered as model variables. The required data for predictive models are obtained by conducting a series of turning test and measuring the surface roughness data. Good agreement is observed between the predictive models results and the experimental measurements. The ANN and RSM models for GFRPs turned part surfaces are compared with each other for accuracy and computational cost.