ISSN: 1304-7191 | E-ISSN: 1304-7205
Artificial neural network and multiple regression analysis for predicting abrasive water jet cutting of al 7068 aerospace alloy
1Department of Mechanical Engineering, Kastamonu University Kastamonu, Turkey
2Department of Motor Vehicles and Transportation Technologies, Sivas Technical Sciences Vocational School, Sivas Cumhuriyet University, Sivas, Turkey
Sigma J Eng Nat Sci 2024; 42(2): 516-528 DOI: 10.14744/sigma.2023.00102
Full Text PDF


This study aims to predict machinability and high performance optimum surface roughness (Ra) by developing multiple regression models and artificial neural network (ANN) model for abrasive waterjet cutting (AWJC) of Aluminum 7068 alloy. Important basic processing parameters such as pump pressure (3500-4000 Bar), nozzle distance (2-5 mm), abrasive flow rate (200-350 g/min), abrasive grain size (100-110, mesh), and nozzle traverse speed (240-300 mm/min) were selected in the study. To examine the effects of these parameters on Ra, 32 experiments were conducted using the L32 orthogonal array, and data was collected. Additionally, the most important factors and interactions affecting Ra were determined using multiple regression analysis and analysis of variance (ANOVA). The ANN model was designed to have multiple hidden layers using MATLAB. The model was trained and evaluated using experimental data, and its performance was measured using mean squared error (MSE) and mean absolute error (MAE). The model was optimized using hyper parameter tuning and cross-validation techniques. As a result, it was determined that the best R2 value of 95.65% from the multiple regression models created to estimate the surface roughness could be obtained from the linear regression model. While selecting the optimum process parameters for AWJC, it was determined that nozzle rotation speed, abrasive grain size and flow rate had the greatest effect by 35.5%, 25.4% and 21.9%, respectively. The optimized ANN model showed high accuracy in predicting Ra for different input parameter combinations. This study provides a reliable and efficient tool for predicting Ra in AWJC, which can contribute to improving process planning and control.