Development Of ANFIS Algorithm To Predict The Sheet Metal Cut Quality Of Carbon Dioxide Laser
Loading...
Date
2011-07
Authors
Chong, Zhian Syn
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
The major trend in laser beam cutting industry is to produce a good cutting quality, which involve with cutting geometry (kerf width), cutting surface quality (surface roughness), mechanical properties (hardness) and metallurgical characteristics (dross inclusion) of the end product. This trend has necessitated the development of predictive models in order to achieve and improve high quality and productivity of laser cutting process. Thus, an empirical comparative study has been carried out using the application of artificial intelligence (AI) approach, namely artificial neural networks (ANNs), fuzzy logic (FL) and adaptive network based fuzzy inference system (ANFIS) to predict the effect of carbon dioxide laser cutting quality based on laser cutting parameters onto 1 mm thickness of Incoloy® alloy 800. All the model developments were implemented on MATLAB toolbox. Experiments were performed to collect data for training and validation purposes, and a set of extra experimental data were used for the verification purpose in order to find out the best AI model architecture for prediction. Based on the results of the study, despite all the three AI approaches gave promising results in term of mean absolute percentage error (MAPE) during training and validation phase, but ANFIS with grid partition technique (ANFIS-GRID) was selected based on the least MAPE during testing phase in the final selection of prediction with the values of 3.30% for kerf width, 12.41% for surface roughness, 2.15% for hardness and 12% for dross inclusion. On the other hand, the prediction accuracy by the finalized four ANFIS models have yielded up to 87% and above proving the prediction stability. Results obtained reveal that the reliability and good predictability of ANFIS model outperforms the ANN and FL model for the laser cutting prediction in terms of training performance and prediction accuracies.
Description
Keywords
Anfis algorithm to predict the sheet metal , cut quality of carbon dioxide laser