Application of Artificial Intelligence Technology for Prompt Diagnosis of Cast Iron Mechanical Properties

Kinetic indentation is widely used to measure physical and mechanical properties of materials as one of the most universal methods for non-destructive testing. This paper uses the latest advances in artificial intelligence and capabilities of the Python programming language libraries allowing to carry out accurate measurements of cast iron hardness based on the data of the material’s micro-impact loading diagram. It has been shown that use of machine learning allows eliminating gross errors and reducing the error of indirect hardness evaluation in several times – down to 10 units according to Brinell HB. It has also been established that formation of additional features for training models (based on traditionally used characteristics: penetration depths, indenter movement speed and contact forces at certain points in time) has a positive effect on the accuracy of measurements, but amount of measurements should also be optimized. Feasibility of effective use of machine learning to evaluate hardness has been demonstrated by comparing of calculated hardness values with data obtained with standard testing methods. Advantage of the developed testing method is the fact that the developed algorithms can be used for prompt diagnostics of cast iron hardness using existing equipment. It is appropriate to extend the proposed approach for determination of other mechanical properties of cast iron: yield strength, strain hardening index, creep, relaxation, determined by indentation methods.

1. Roy E. Cast iron technology. Butterworth-Heinemann. 2014;252 p. DOI: 10.1016/B978-0-408-01512-7.50001-X

2. Ferro P. Cast Irons: Properties and Applications. Mdpi AG. 2020;150 p.

3. Bharadiya JP, Reji TK, Farhan A. Rise of Artificial Intelligence in Business and Industry. Journal of Engineering Research and Reports. 2023;3(25):85-103. DOI: https://doi.org/10.9734/jerr/2023/v25i3893

4. Shahhosseini M. [et al.]. Coupling machine learning and crop modeling improves crop yield prediction in the US Corn Belt. Sci Rep. 2021;(11):1606 p. DOI: 10.1038/s41598-020-80820-1

5. Xin D. [et al.]. Whither AutoML? Understanding the Role of Automation in Machine Learning Workflows. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems (CHI '21). 2021;(83):1-16. DOI: 10.1145/3411764.3445306

6. Lu J, Suresh S. Dynamic indentation for determining the strain rate sensitivity of metals. Mech Phys Solid. 2003;(51):11-12. DOI: 10.1016/j.jmps.2003.09.007Kren AP. [et al.]. Testing physical and mechanical properties of cast iron using ifmh-ch device. Casting and metallurgy. 2019;(3). (In Russ.).

7. Hassani M. [et al.]. Material hardness at strain rates beyond 106 s−1 via high velocity microparticle impact indentation. Scripta Mater. 2020;(177):198-202. DOI: 10.1016/j.scriptamat.2019.10.032

8. Kren A, Delendik M, Machikhin A. Non-destructive evaluation of metal plasticity using a single impact microindentation. International Journal of Impact Engineering. 2022;(162):104141 DOI: 10.1016/j.ijimpeng.2021.104141

9. Hackett BL. [et al.]. Advances in the measurement of hardness at high strain rates by nanoindentation. J Mater Res. 2023;5(38):1163-77. DOI: 10.1557/s43578-023-00921-1

10. Kren AP. [et al.]. Testing physical and mechanical properties of cast iron using ifmh-ch device. Casting and metallurgy. 2019;(3). (In Russ.).

11. Organek P, Gosowski B, Redecki M. Relationship between Brinell hardness and the strength of structural steels. Structures. 2024;59:105701. DOI: 10.1016/j.istruc.2023.105701

12. Antonov AV, Chepurko VA. Building non-parametric distribution density using censed information. Reliability. 2005;(2):3 p. (In Russ.).

13. Hoffman J. Categorical and Cross-Classified Data: McNemar's and Bowker's Tests, Kolmogorov-Smirnov Tests, Concordance. Basic Biostatistics for Medical and Biomedical Practitioners (Second Edition). 2019:233-247 pp. DOI: 10.1016/B978-0-12-817084-7.00015-2