Predicting the Flexural & Hardness Properties of a 3D Printed specimen of different Polymer Materials using “ML&AI”
Sanman S*1,Lavakumar K S2, Nanda Kumar N3,Sudeep T*4, Vinay S R4, Girish V4, Rohith Gowda J4
1Associate Professor, Dept. of Mechanical Engineering, Acharya Institute of Technology, Bengaluru, Karnataka,India
2Assistant Professor, Dept. of Mechanical Engineering, Acharya Institute of Technology, Bengaluru, Karnataka, India
3Assistant Professor, Dept. of AI&ML, Acharya Institute of Technology, Bengaluru, Karnataka, India
4UG Students, Dept. of Mechanical Engineering, Acharya Institute of Technology, Bengaluru, Karnataka, India
Affiliations:
Department of Mechanical Engineering, Acharya Institute of Technology, Bengaluru, India
*sanman2289@gmail.com
*sudeepk480@gmail.com
Abstract
The current work is employing the application of ML/AI techniques to Predict the actual strength of 3D-printed parts of different polymer materials in comparison with experimental results. The test specimens are printed as per D780 ASTM standards mainly with arguably the most common FDM materials such as PLA, ABS, and PETG. Parameters relevant to printing of test specimen such as layer height, infill density, print direction, and temperature were adopted and 3D Printed specimen are tested for flexural and hardness properties which helps the ML models classify how different printing conditions affect the final mechanical strength.From the obtained results, it is observed that,flexural strength obtained in experimental test is 56.184MPa and for Maching learning models ANN predicted 43.657MPa,Random forest model predicted 49.483MPa and XGBoost model predicted 57.573MPa .and hardness in experimental test is 73N/mm2 and for Maching learning models ANN predicted 72.28N/mm2,Random forest model predicted 71.28N/mm2 and XGBoost model predicted 73.01N/mm2.
Finally it can be concluded that XGBoost model follows the experimental trend,for both flexural and hardness with accuracy of 98.53% and 97.25% hardness strength indicating superior predictive capability compared to ANN and Random Forest models.
Keywords
FDM 3D Printing, Machine Learning, Flexural and Hardness Properties, Predictive Modeling, Martials such as PLA,ABS and PETG
