%0 Journal Article
%@ 19552513
%A Ting, H.Y.
%A Asmelash, M.
%A Azhari, A.
%A Alemu, T.
%A Saptaji, K.
%D 2022
%F scholars:17039
%I Springer-Verlag Italia s.r.l.
%J International Journal on Interactive Design and Manufacturing
%K Backpropagation; Design of experiments; Forecasting; Machining; Machining centers; Multilayer neural networks; Regression analysis; Support vector machines; Titanium alloys, Abrasive waterjet machining; Abrasive waterjets; Comparison between methods; Low pressures; Machining Process; Regression analysis models; Support vector regressions; Support vectors machine; Titanium (alloys); Waterjet machining, Surface roughness
%N 1
%P 281-289
%R 10.1007/s12008-021-00830-9
%T Prediction of surface roughness of titanium alloy in abrasive waterjet machining process
%U https://khub.utp.edu.my/scholars/17039/
%V 16
%X This work presents a comparison between methods for predicting surface roughness (Ra) of titanium alloy machined by low-pressure abrasive water jet machine developed in our laboratory. Artificial neural network (ANN), support vector machine (SVM) and regression analysis (RA) models were used to analyse the machining data. The work aims at the comparison and selection of the best prediction method based on the accuracy of the predicted surface roughness. An experiment was designed using full factorial experimental design. In the experiment, machining parameters of traverse speed (V), waterjet pressure (P) and standoff distance (h) were considered as model variables. The actual surface roughness values were collected based on the designed experiment. A feed forward back propagation neural network was used, structured as one input layer, one hidden layer with 5 hidden neurons and one output layer. Both ANN and SVM models were trained and analysis of variance was used. F-test was used to validate the RA model. The results showed that the proposed methods indicated an acceptable level of accuracy for predicting the surface roughness, however, ANN model had better accuracy than SVM and RA models as it produced lower relative errors between the predicted values and experimental results. © 2022, The Author(s), under exclusive licence to Springer-Verlag France SAS, part of Springer Nature.
%Z cited By 6