A comparison between Recurrent Neural Networks and classical machine learning approaches in Laser Induced Breakdown Spectroscopy - LIBS concentration prediction

Aug 4, 2026

Upload by Logic

Recurrent Neural Networks are evaluated for temporal modeling in laser-induced breakdown spectroscopy (LIBS) to enable quantitative concentration analysis of aluminum alloys. Using the 1064 nm fundamental harmonic of a nanosecond Nd:YAG laser pulse, LIBS plasma emissions are used to predict constituent concentrations of aluminum standard samples. Multiple recurrent architectures—LSTM, GRU, SimpleRNN—and convolutional recurrent variants (Conv-SimpleRNN, Conv-LSTM, Conv-GRU) are compared against classical machine learning regressors including SVR, MLP, Decision Tree, GBR, RFR, Linear Regression, and KNN. Results indicate convolutional recurrent networks provide the highest prediction efficiency across most elements.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages28

Reading Time1 h 11 min

AccessFree to download

Abstract

Introduction

LIBS fundamentals and plasma emission

Classical and deep learning in LIBS

Study objective and model comparison

A comparison between Recurrent Neural Networks and classical machine learning approaches In Laser induced breakdown spectroscopy Fatemeh Rezaei 1*, Pouriya Khaliliyan 1,2, Mohsen Rezaei 3, Parvin Karimi 4, Behnam Ashrafkhani 5 1 Department of Physics, K. N. Toosi University of Technology, Tehran, Iran 2 PDAT Laboratory, Department of Physics, K. N. Toosi University of Technology, Tehran, Iran 3 Department of Industrial Engineering, University of Science and Technology of Mazandaran, Behshahr, Iran 4 Department of Physics, South Tehran Branch, Islamic Azad University, Tehran, Iran 5 Department of Physics, University of Calgary, Canada, Calgary, Alberta Corresponding author: [fatemehrezaei@kntu.ac.ir](fatemehrezaei@kntu.ac.ir) Abstract Recurrent Neural Networks are classes of Artificial Neural Networks that establish connections between different nodes form a directed or undirected graph for temporal dynamical analysis. In this research, the laser induced breakdown spectroscopy (LIBS) technique is used for quantitative analysis of aluminum alloys by different Recurrent Neural Network (RNN) architecture. The fundamental harmonic (1064 nm) of a nanosecond Nd:YAG laser pulse is employed to generate the LIBS plasma for the prediction of constituent concentrations ofthe aluminum standard samples. Here, Recurrent Neural Networks based on different networks, such as Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU), Simple Recurrent Neural Network (Simple RNN), and as well as Recurrent Convolutional Networks comprising of Conv-SimpleRNN, Conv-LSTM and Conv-GRU are utilized for concentration prediction. Then a comparison is performed among prediction by classical machine learning methods of support vector regressor (SVR), the Multi Layer Perceptron (MLP), Decision Tree algorithm, Gradient Boosting Regression (GBR), Random Forest Regression (RFR), Linear Regression, and k-Nearest Neighbor (KNN) algorithm. Results showed that the machine learning tools based on Convolutional Recurrent Networks had the best efficiencies in prediction of the most of the elements among other multivariate methods. Keywords: LIBS, Recurrent Neural Networks, Concentration prediction, Machine learning, LSTM, GRU, SimpleRNN, Convolutional Recurrent Networks. Introduction Aluminum is one of the most abundant metals on the earth, and the most widespread element in the earth's crust, after silicon and oxygen. The main properties of aluminum are its low density, high heat conductivity, ductility, corrosion resistance, being a catalyst, low temperature resistance, high reflectivity, and sound absorbing. Studying the characteristics of aluminum’s concentration can be performed by different methods. Laser induced breakdown spectroscopy (LIBS) as a type of atomic emission spectroscopy is a fast, online optical technique that employs a highly energetic laser pulse as the excitation source. These advantages make LIBS a powerful tool for the quantitative analysis of different materials, especially metal samples 1,2. In this method, the laser is focused to produce hot plasma, which atomizes and excites the targets. Notice that the formation of plasma only starts when the focused laser energy reaches a certain threshold for optical breakdown. Then the emissions of plasma are conducted to a spectrometer for identification of the constituent elements of the analyzed sample 3-4. Various research groups have studied the classical 5-14 and deep machine learning algorithms 15-18 in LIBS technique for improving the state of the art quantitative estimations by using simple Artificial Neural Networks (ANN) 19-23, Convolutional Neural Network 15,18,19 , and so on. For instance, Xu et al. 15 used the algorithm of Convolutional Neural Network for studying the LIBS spectra of two-dimensional soil samples. They demonstrated that Convolutional Neural Networks outperformed traditional preprocessing approaches by preventing overfitting. They presented that th

FAQs

Which laser and LIBS setup is used for the concentration prediction task?

The study uses the 1064 nm fundamental harmonic of a nanosecond Nd:YAG laser pulse to generate LIBS plasma for aluminum alloy standard samples.

Which recurrent neural network variants are tested?

LSTM, GRU, SimpleRNN, and convolutional recurrent networks including Conv-SimpleRNN, Conv-LSTM, and Conv-GRU are used for concentration prediction.

How do recurrent convolutional networks compare with classical machine learning methods?

The results show convolutional recurrent networks achieve the best efficiencies for predicting most elements compared with multivariate classical approaches such as SVR, MLP, Decision Trees, GBR, RFR, Linear Regression, and KNN.

Aug 4, 2026

Upload by Logic