Researchers harness the power of deep learning regression to revolutionize the quantitative analysis of aluminum scrap using laser-induced breakdown spectroscopy (LIBS), providing a highly accurate and efficient method for metal sorting and recycling. Proven.
Research published in Spectrokymica Actor Part B: Atomic Spectroscopy The journal has uncovered the potential of deep learning regression to revolutionize the analysis of spent aluminum scrap by laser-induced breakdown spectroscopy (LIBS) (1). This study addresses the limitations of existing methods for estimating alloying element concentrations using LIBS spectra that have not reached the accuracy required by the industry.
deep learning regression is a subfield of machine learning that utilizes deep neural networks to model and predict continuous numbers. This involves training neural networks in multiple layers to learn complex patterns and relationships in your data. Unlike traditional regression techniques that typically rely on manual feature selection, deep learning regression automatically learns relevant features from the input data. Neural networks are trained using large datasets and tune internal parameters to minimize the difference between predicted and actual values. This enables the model to make accurate predictions on new, unknown data, making it particularly effective for complex, high-dimensional problems.
In this study, we compare the performance of new deep learning approaches with traditional linear univariate regression and machine learning methods, using metrics such as root mean square error (RMSE), mean absolute error (MAE), and mean absolute error (MAE). We evaluate its effectiveness based on R.2 (coefficient of determination). Two sample sets were utilized to conduct the evaluation. One consists of 27 certified aluminum reference samples and the other consists of 733 post-consumer scrap pieces, the ground truth concentrations of which are measured by X-ray fluorescence (XRF).
Importantly, the employment of multiple loss functions, each unique to an element, proved to make a big difference in regression performance. This approach improved results across all performance metrics for the scrap sample set. The same trend was observed in the reference sample set, but with the exception of Fe, Mn and Mg, there was no positive effect on the coefficient of determination. Moreover, the proposed methodology effectively addressed the learning prioritization problem, ensuring that the concentrations of base elements were not prioritized over alloying elements.
Impressively, the best-performing deep learning model has an average RMSE of only 0.02 wt% for Al and Si and an average RMSE of Fe, Cu, Mn, Mg, and Zn that does not exceed 0.01 wt%. achieved accuracy. This groundbreaking achievement holds great promise for the future of LIBS in metal sorting applications. This research demonstrates the potential of deep learning regression to meet stringent industry requirements and provides an innovative solution to enhance the sorting and recycling of post-consumer aluminum scrap.
This state-of-the-art research pushes the limits of conventional methods and opens up new avenues for high-precision analysis in the recycling field. This finding not only proves the superior performance of deep learning in LIBS analysis, but also highlights its immense potential in other scientific and industrial applications. With the growing demand for accurate and efficient metal sorting, the progress made in this research paves the way for a more sustainable and resource efficient future.
reference
(1) Van den Eynde, S. Diaz Romero, DJ. Zaprana, I. Peeters, J. Deep Learning Regression for Quantitative LIBS Analysis. Spectrokymica Actor Part B: At. Spectrometer. 2023, 202, 106634. DOI: 10.1016/j.sab.2023.106634
