|
Comparison of artificial neural network and response surface methodology in predicting the tensile strength and optimization of 3D printed objects |
|---|---|
| รหัสดีโอไอ | |
| Creator | 1. Pichai Janmanee 2. Pongpun Ratchapakdee |
| Title | Comparison of artificial neural network and response surface methodology in predicting the tensile strength and optimization of 3D printed objects |
| Publisher | Faculty of Engineering, Khon Kaen University |
| Publication Year | 2567 |
| Journal Title | Engineering and Applied Science Research |
| Journal Vol. | 51 |
| Journal No. | 6 |
| Page no. | 704-715 |
| Keyword | Fused deposition modelling, Prediction, Response surface methodology, Artificial neural network, Tensile strength |
| URL Website | https://ph01.tci-thaijo.org/index.php/easr/index |
| Website title | Engineering and Applied Science Research |
| ISSN | 2539-6161 |
| Abstract | This study investigates the geometric properties of 3D-printed objects using fused deposition modelling. It focuses on optimising parameters for printing in terms of their significant impact on both the quality and cost of printed objects. To enhance the quality of 3D-printed objects, it is crucial to predict the geometric properties in advance. The development of an artificial neural network model (ANN) is employed to predict the tensile strength properties of polylactic acid material during experiments. The model considers three variables: printing temperatures at three levels (190°C, 210°C, and 230°C), printing speeds at three levels (30 mm/s, 50 mm/s, and 70 mm/s), and material infill density at three levels (40%, 60%, and 80%). Tensile strength testing was conducted, and the predictive performance of ANN models was compared with mathematical models derived from the application of response surface methodology (RSM). The goal was to determine suitable printing conditions. Tensile strength testing revealed that printing temperature, printing speed, and infill density significantly impact tensile strength. The ANN configuration consists of a 3-input layer with 3 neurons, a hidden layer with 14 neurons, and an output layer with 1 neuron, denoted as 3-14-1. The model exhibited a testing decision-making accuracy of 0.938. The average error for the ANN model was 0.307, lower than the average error from the full factorial model, which was 1.392. The optimized printing conditions for maximum tensile strength were found to be a print temperature (X1) of 230C, a feed rate (X2) of 30 mm/sec, and an infill density (X3) of 80%, resulting in a tensile strength of 43.107 MPa. The mathematical model derived from RSM demonstrated efficacy in predicting and controlling the quality of printed objects, aiming to reduce production costs and enhance efficiency in future 3D printing processes. |
