Towards selective laser sintering of objects with customized mechanical properties based on ANFIS predictions

• Published in: Journal of mechanical science and technology Vol. 34; no. 12; pp. 5075 - 5084
• Main Authors: Aldahash, Saleh A., Salman, Shaaban A., Gadelmoula, Abdelrasoul M.
• Format: Journal Article
• Language: English
• Published: Seoul Korean Society of Mechanical Engineers 01.12.2020; Springer Nature B.V; 대한기계학회
• Subjects: Adaptive systems; Cement; Compressive strength; Control; Dynamical Systems; Engineering; Industrial and Production Engineering; Iterative methods; Laser sintering; Manufacturing; Mechanical Engineering; Mechanical properties; Modulus of rupture in bending; Optimization; Original Article; Polyamide resins; Production costs; Rapid prototyping; Ultimate tensile strength; Vibration; 기계공학; Cement-filled PA12; Mechanical properties; Adaptive network-based fuzzy inference system (ANFIS); Selective laser sintering (SLS)
• ISSN: 1738-494X; 1976-3824
• DOI: 10.1007/s12206-020-1111-6

Recently, the adaptive network-based fuzzy inference system (ANFIS) has been used extensively in modeling of manufacturing processes to save both optimization time and manufacturing costs. ANFIS is a powerful iterative tool for optimizing non-linear and multivariable manufacturing operations. In the present study, ANFIS is used to predict the optimum manufacturing parameters in selective laser sintering (SLS) of cement-filled polyamide 12 (PA12) composite. For this purpose, a set of cement-filled PA12 test specimens is manufactured by SLS technique with 8 different values of laser power (4.5–8 Watt) and 8 different weight fractions of white cement (5 %–40 %). Mechanical characterization of cement-filled PA12 is carried out to evaluate the ultimate tensile strength (UTS), compressive strength, and flexural properties. The experimental data are then divided into two groups; one group for training the ANFIS model and the other group for checking the validity of the identified model. The built ANFIS model was validated experimentally and comparison with experimental results revealed mean relative errors of 2.92 %, 3.84 %, 4.75 %, and 3.31 % in the predictions of UTS, compressive strength, flexural modulus, and flexural yield strength, respectively.