User Benefits
- Excellent results can be achieved with fewer trials with Bayesian optimization.
- When the AGS-V is used, tensile tests can be conveniently performed with excellent precision.
Introduction
3D printer technology is widely used in manufacturing and research and development to efficiently produce parts with complex shapes. However, its quality depends on various factors, such as the material used, the forming conditions, and equipment settings. Optimizing these factors is not easy. In particular, to improve performance indicators, such as forming accuracy, surface quality, and strength, it is necessary to adjust a large number of parameters, so traditional methods based on trial and error require large amounts of time and cost. In recent years, “Bayesian optimization” has been attracting attention as a solution. It is a statistical method that can efficiently search for optimum parameters with a limited number of trials. It is particularly effective for high dimensional, nonlinear problems. This article introduces an example in which Bayesian optimization was used on a model for manufacturing tensile test specimens using a resin 3D printer to obtain 3D printing conditions (explanatory variables) that optimized tensile strength and manufacturing times (objective variables).*
*1 The output conditions obtained in this test are optimal for dumbbell-
shaped test specimens, but they may not be optimal for other shapes.
Test Conditions
The equipment configuration is shown in Table 1. PETG-CF, which has excellent mechanical strength and high flexibility, was used in the filaments that were the raw materials for the test specimens. Fig. 1 shows the AGS-V precision universal testing machine; Fig. 2 shows the test setup; and Fig. 3 shows the results of a tensile test. The elongation to failure of the PETG-CF used in the filaments was low, and failure occurred in a brittle manner.
Bayesian Optimization and Procedures
With Bayesian optimization, the test results are input, and the next test conditions are output. This process was performed 6 times, and a total of 30 conditions were tried, so the optimum result obtained was taken to be the optimum solution. The Bayesian optimization conditions are shown in Table 2. In this case, the explanatory variables were set to “layer pitch,” “print speed,” and “filling density,” and the objective variables were set to “maximization of tensile strength” and “minimization of printing time.” Table 3 shows the nine conditions and their test results for the first iteration. The Bayesian optimization calculation was performed using a service provided by Mitsui Knowledge Industry Co., Ltd. (MKI-bayesopt).
Test Results
The Bayesian optimization trial results are shown in Table 4. For tensile strength, test specimen number 25 (red line) had the optimum conditions, and for printing time, test specimen number 14 (blue line) had the optimum conditions. Fig. 4 is a 3- dimensional chart that shows the correlation between the objective variable tensile strength and the explanatory variables. The orange part indicates high tensile strength, and the blue part indicates low tensile strength. It can be seen that filling density and print speed had a greater effect on strength than layer pitch. Higher filling density and slower print speed tended to produce higher strength. Fig. 5 is a 3-dimensional chart that shows the correlation between the objective variable printing time and the explanatory variables. The orange part indicates longer printing times, and the blue part indicatesshorter printing times. The larger the layer pitch, the faster the print speed, and the lower the filling density, the shorter the printing time.
English 
German 
Czech 
Slovak 
Bulgarian 
Romanian 