Predictive Modeling For Low Alloy Steel Mechanical Properties: A Comparison Of Machine Learning Algorithms And Parameter Optimization

Desmarita Leni; Lega  Putri Utami; Ruzita  Sumiati; Moh.  Camim; Sharif  Khan

doi:10.24036/ijimce.v1i1.7

Authors

Desmarita Leni Departement Mechanical Engineering, Universitas Muhammadiyah Sumatera Barat, INDONESIA
Lega Putri Utami Departement Mechanical Engineering, Universitas Andalas, INDONESIA
Ruzita Sumiati Departement Mechanical Engineering, Politeknik Negeri Padang, INDONESIA
Moh. Camim Departement Mechanical Engineering, Sekolah Tinggi Teknologi Warga Surakarta, INDONESIA
Sharif Khan AABCO Engineering, Machine Technology SDN BHD, MALAYSIA

DOI:

https://doi.org/10.24036/ijimce.v1i1.7

Keywords:

Modeling, machine learning, algorithms, yield strength, low alloy steel

Abstract

The development of machine learning in predicting the mechanical properties of alloy steel has become an important research subject in recent years. This is due to the ability of machine learning to extract complex patterns from large and intricate data, which can be used to understand the relationship between chemical composition, microstructure, and mechanical properties of alloy steel. This research aims to design a machine learning model to predict the mechanical properties of low alloy steel, such as Yield Strength (YS) and Ultimate Tensile Strength (UTS), based on the percentage composition of chemical elements in low alloy steel and the heat treatment applied. The machine learning model in this study consists of 10 input variables and 2 target variables. The research compares the performance of 3 machine learning algorithms, namely Decision Tree (DT), Random Forest (RF), and Artificial Neural Network (ANN). The research findings indicate that the ANN algorithm model performs best in predicting the mechanical properties of low alloy steel. This model has Mean Absolute Error (MAE) values of 16.5 and 19.593 for predicting YS and UTS, Root Mean Square Error (RMSE) values of 19.111 and 22.005, and coefficient of determination (R) values of 0.964 and 0.947 for YS and UTS respectively. The modeling uses the ANN algorithm with an 80% training data and 20% testing data split, and applies the K-Fold Cross Validation method with a value of K=5. The best parameters obtained are a learning rate of 0.001, momentum of 0.1, and a hidden layer neuron count of 9. These results indicate that ANN has great potential in addressing the complexity and variability in material data. The implications of these findings are that the implementation of ANN in manufacturing and material engineering industries can enhance the accuracy and efficiency in material strength prediction processes, which, in turn, can aid in designing and developing better and more durable products.

Downloads

Download data is not yet available.

References

D. Leni, F. Earnestly, R. Sumiati, A. Adriansyah, and Y. P. Kusuma, “Evaluasi sifat mekanik baja paduan rendah bedasarkan komposisi kimia dan suhu perlakuan panas menggunakan teknik exploratory data analysis (EDA),” Din. Tek. Mesin, vol. 13, no. 1, p. 74, Apr. 2023, doi: 10.29303/dtm.v13i1.624.

A. A. Morini, M. J. Ribeiro, and D. Hotza, “Early-stage materials selection based on embodied energy and carbon footprint,” Mater. Des., vol. 178, p. 107861, Sep. 2019, doi: 10.1016/j.matdes.2019.107861.

R. Šmak, J. Votava, and A. Polcar, “The Cooling Media Influence on Selected Mechanical Properties of Steel,” Acta Technol. Agric., vol. 23, no. 4, pp. 183–189, Dec. 2020, doi: 10.2478/ata-2020-0029.

N. M. Ismail, N. A. A. Khatif, M. A. K. A. Kecik, and M. A. H. Shaharudin, “The effect of heat treatment on the hardness and impact properties of medium carbon steel,” IOP Conf. Ser. Mater. Sci. Eng., vol. 114, p. 012108, Feb. 2016, doi: 10.1088/1757-899X/114/1/012108.

Y. Zhang, J. Yang, D. Xiao, D. Luo, C. Tuo, and H. Wu, “Effect of Quenching and Tempering on Mechanical Properties and Impact Fracture Behavior of Low-Carbon Low-Alloy Steel,” Metals, vol. 12, no. 7, p. 1087, Jun. 2022, doi: 10.3390/met12071087.

B. Hutchinson et al., “Microstructures and hardness of as-quenched martensites (0.1–0.5%C),” Acta Mater., vol. 59, no. 14, pp. 5845–5858, Aug. 2011, doi: 10.1016/j.actamat.2011.05.061.

N. S. Reddy, J. Krishnaiah, S.-G. Hong, and J. S. Lee, “Modeling medium carbon steels by using artificial neural networks,” Mater. Sci. Eng. A, vol. 508, no. 1–2, pp. 93–105, May 2009, doi: 10.1016/j.msea.2008.12.022.

L. Qiao, Z. Wang, and J. Zhu, “Application of improved GRNN model to predict interlamellar spacing and mechanical properties of hypereutectoid steel,” Mater. Sci. Eng. A, vol. 792, p. 139845, Aug. 2020, doi: 10.1016/j.msea.2020.139845.

N. Amiri, G. H. Farrahi, K. R. Kashyzadeh, and M. Chizari, “Applications of ultrasonic testing and machine learning methods to predict the static & fatigue behavior of spot-welded joints,” J. Manuf. Process., vol. 52, pp. 26–34, Apr. 2020, doi: 10.1016/j.jmapro.2020.01.047.

D. Leni, “Prediction Modeling of Low Alloy Steel Based on Chemical Composition and Heat Treatment Using Artificial Neural Network,” J. Polimesin, vol. 21, no. 5, pp. 54–61, Oct. 2023, doi: 10.30811/jpl.v21i5.3896.

D. Leni, Y. P. Kusuma, M. Muchlisinalahuddin, F. Earnerstly, R. Muharni, and R. Sumiati, “PERANCANGAN METODE MACHINE LEARNING BERBASIS WEB UNTUK PREDIKSI SIFAT MEKANIK ALUMINIUM,” J. Rekayasa Mesin, vol. 14, no. 2, pp. 611–626, Aug. 2023, doi: 10.21776/jrm.v14i2.1370.

A. Agrawal, P. D. Deshpande, A. Cecen, G. P. Basavarsu, A. N. Choudhary, and S. R. Kalidindi, “Exploration of data science techniques to predict fatigue strength of steel from composition and processing parameters,” Integrating Mater. Manuf. Innov., vol. 3, no. 1, pp. 90–108, Dec. 2014, doi: 10.1186/2193-9772-3-8.

Ade Usra Berli, Desmarita Leni, and Helga Yermadona, “Analisis Pengaruh Komposisi Kimia dan Suhu Perlakuan Panas Terhadap Sifat Mekanik Baja Tahan Karat untuk Aplikasi Konstruksi,” J. Surya Tek., vol. 10, no. 2, pp. 811–819, Dec. 2023, doi: 10.37859/jst.v10i2.6059.