Using LIBS to Gauge the Hardness of Steel Rails

A novel method utilizing laser-induced breakdown spectroscopy (LIBS), coupled to multivariate model leveraging machine learning techniques, has the capability to evaluate the hardness of steel rails for essential infrastructure projects, according to a recent study published in Spectrochimica Acta Part B: Atomic Spectroscopy (1).

The steel industry has been an essential part of commerce and transportation since the Industrial Revolution. Steel has been used to complete various important projects, the most significant of which being rail lines to allow locomotives and trains to transport goods and people to other regions more efficiently. The development of the steel rail resulted in the expansion of public transportation options, as well as the economic development of many countries (2). Even today, trains are regularly used to transport essential cargo across the country, which means it is imperative that the tracks they travel on remain in optimal condition (2). This fact emphasizes the importance of gauging the hardness of the steel rails that are used to construct these tracks.

Technological innovation has spurred countries to examine ways to improve current infrastructure, including how to improve infrastructure that could accommodate faster and heavier trains. This recent study from Zefeng Yang of Southwest Jiaotong University in Chengdu, China, attempted to demonstrate that LIBS and machine learning could be used to gauge the hardness of steel rails (1). Recognizing the pivotal role of rail hardness in ensuring operational safety, the researchers sought to develop a rapid, in-situ method for its measurement.

railway rails go into the distance around the bend against the backdrop of a beautiful sunset | Image Credit: © kazakova0684 - stock.adobe.com

The researchers presented three methodologies in their study: spectral line intensity ratios, machine learning algorithms, and plasma excitation temperature. To analyze the hardness of the U71Mn steel rails, a multivariate model leveraging machine learning approach was used (1).

The study demonstrated that the researchers had to use variance normalization in machine learning to improve information retention during data dimensionality reduction. The researchers examined 12 algorithm combinations, the result of which identified particle swarm optimization in support vector regression (PSO-SVR) (1). PSO-SVR stood out among the algorithms because it yielded the lowest mean squared error (MSE) (1).

Once the researchers determined which algorithm to use, they further refined the data analysis by integrating stochastic weights, resulting in an impressive coefficient of determination (R²) of 0.9876 (1). Subsequent validation on five new samples reaffirmed the model's robust performance, boasting an R² of 0.9864 (1).

Beyond its immediate implications for rail hardness assessment, the study underscores the broader potential of the developed methodology. By enhancing the precision and reliability of LIBS technology in quantitative surface hardness analysis, this approach the researchers developed holds promise for diverse applications across various domains.

As the demand for high-speed rail systems continues to surge globally, the advancement in rail hardness assessment methodologies not only ensures operational safety but also fosters efficiency and reliability in railway transportation networks.

References

(1) Xia, L.; Yang, Z.; Wei, W.; Wu, G. A Rapid In-situ Hardness Detection Method for Steel Rails based on LIBS and Machine Learning. Spectrochimica Acta Part B: At. Spectrosc. 2024, 215, 106908. DOI: 10.1016/j.sab.2024.106908

(2) Ternium, Steel Tracks in Time: The Evolution of Railroads and Trains. Available at: https://us.ternium.com/en/media/news/steel-tracks-in-time--08361156524 (accessed 2024-04-22)