Cement classification and characterization using Non-Invasive techniques

• Main Author: Romero, Esteban
• Publication Date: 2025
• Other Authors: Ferreira, Dennis S., Pereira, Fabiola M.V. [UNESP], Olivieri, Alejandro C., Pereira-Filho, Edenir R., Arancibia, Juan A.
• Format: Article
• Language: eng
• Source: Repositório Institucional da UNESP
• Download full: http://dx.doi.org/10.1016/j.talanta.2024.127212; https://hdl.handle.net/11449/302582
• topic: Cement powder; Data fusion; ED-XRF; Figures of merit; LIBS; Multivariate calibration

Cement is one of the most widely used materials in the global construction industry, serving as an adhesive and binder in projects that require strength and durability. Additionally, cement production indicates a country's development and economic activity, with global production reaching approximately 4 billion tons annually. It is a fine powder composed mainly of lime, silica, iron oxide, and alumina. Portland cement is the most common type, although a wide variety of types of cement differ in their chemical composition, providing them with specific properties for different applications. A set of fifty samples, consisting of eleven primary samples and thirty-nine blends formed by the combination of these eleven samples, was prepared. Additionally, twenty-four samples were randomly selected for error covariance calculation. Subsequently, two analytical techniques, laser-induced breakdown spectroscopy (LIBS) and energy dispersive X-ray fluorescence (ED-XRF) were applied to quantify aluminum (Al), calcium (Ca), iron (Fe), potassium (K), magnesium (Mg), sodium (Na), and sulfur (S). Afterward, the samples were analyzed via ICP OES after acid mineralization with 8 mL aqua regia in a digester block. Multivariate calibration strategies such as principal component regression (PCR), maximum likelihood principal component regression (MLPCR), partial least-squares regression (PLS), and error covariance penalized regression (ECPR) were employed. Finally, figures of merit were calculated to verify the most suitable models. The results revealed robust models with notable sensitivity, ranging from 0.3 to 329 signal a.u (% w w−1)−1), low limits of detection (LoD) within the range of 0.00–0.1 % w w−1, and remarkable accuracy ranging from 67.8 % to 140.3 %, particularly for Ca, Fe, Mg, and Na. This research takes an essential step in developing simple analytical methods with low waste generation and less environmental impact, thanks to using novel chemometric techniques to process the data.