Laser-Induced Breakdown Spectroscopy Quantitative Analysis Using a Bayesian Optimization-Based Tunable Softplus Backpropagation Neural Network

In this paper we exploit Artificial Neural Networks (ANN) to model the functional relationship between LIBS spectra and the corresponding composition of bronze alloys, expressed in terms of concentrations of the four elements constituting the alloy. The typical approach to Laser-Induced Breakdown Spectroscopy (LIBS) quantitative analysis uses calibration curves, suitably built based on appropriate reference standards. More recently, statistical methods relying on the principles of ANNs are increasingly used. In particular, an ANN can be used for a preliminary exploration of the LIBS spectra in order to find out the most significant areas of the spectrum, which will be used by another ANN dedicated to the calibration. In this paper we will show that the use of ANNs to deal with LIBS spectra provides a viable, fast and robust method for LIBS quantitative analysis. Actually, this approach requires a relatively limited number of reference samples for the training of the network, with respect to the current approaches, and can automatically analyze a large number of samples.

Precision agriculture aims to increase yield and profits while reducing costs, waste, and environmental side-effects. This is achieved through a process of measuring, modelling and acting; for example, laser-induced breakdown spectroscopy (LIBS) can be used to measure macro and micro nutrients in crops to determine nutrient requirements. The limiting factor with quantitative LIBS analysis of plant nutrient levels is the variation between shots on the same sample. Following a review of current literature relevant to LIBS for agriculture, this work investigates whether different chemometric methods can mitigate these variations and can create quantitative calibrations for nutrient levels in fresh and dried pelletised pasture under laboratory conditions. The methods explored were Savitzky Golay filtering, multiple linear regression, principal component regression, partial least-squares regression, gaussian process regression, and artificial neural networks. The algorithms that performed best were partial least-squares with gaussian process regression (R2 of 0.93, 0.95, and 0.92 for K, Na, and Mn, respectively), principal components analysis with artificial neural networks (R2 of 0.94, 0.83, and 0.80 for Fe, Ca, and Mg, respectively), and partial least-squares with artificial neural networks (R2 of 0.77 for B). Removing the moisture from the pasture improved model R2 values by 4–5% on average. Acquiring spectra under an argon purge produced a small reduction in accuracy for some nutrients compared to models acquired in air. Including categorical data in the principal component regression and the artificial neural networks produced negligible improvements in prediction. This chapter will give an introduction of using different types of chemometric analyses on spectra generated by LIBS to measure micro and macro nutrients in pasture under laboratory conditions. It discusses the challenges faced when building models for each nutrient.

The feasibility of in situ quantitative multielemental analysis and production failures detection by laser induced breakdown spectroscopy (LIBS) has been demonstrated during direct energy deposition process in additive manufacturing. Compact LIBS probe was developed and equipped with the laser cladding head installed at industrial robot for real-time chemical quantitative analysis of key components (Ni, W) during the synthesis of high wear resistant coatings of nickel alloy reinforced with tungsten carbide particles. Owing to non-uniform distribution of tungsten carbide grains in the upper surface layer the only acceptable choice for LIBS sampling was made to the melt pool at growing clad. Laser ablation at powder particles above melt pool was insignificant for LIBS plasma properties due to low intensity and low probability of plasma breakdown at powder particles. No impact of LIBS sampling on cladding process and clad properties was observed according to optical and scanning electron microscopies. The feasibility of in situ LIBS quantitative elemental analysis of key components (tungsten and nickel) has been demonstrated during the cladding process. LIBS analysis results were in good agreement with offline measurements by electron energy dispersive X-ray spectroscopy and X-ray fluorescence spectroscopy. Finally, LIBS technique was demonstrated to be a good tool for real-time detection of cladding process failures (poor laser beam quality, undesirable variation of components concentrations).

Application of elastic net in quantitative analysis of major elements using Martian laser-induced breakdown spectroscopy datasets

Rapid and accurate measurement of coal quality has great significance for efficient use of coal at thermal power plants. Laser-induced breakdown spectroscopy (LIBS) combined with chemometric methods has many unique advantages in coal analysis. In this study, four calibration models, based on partial least squares regression (PLSR), support vector regression (SVR), artificial neural network (ANN), and principal component regression (PCR), were applied assisted by the LIBS technique for the quantitative analysis of coal quality. In order to find the optimal calibration method with LIBS for coal analysis, the spectral data of 40 standard coal samples with pressed-pellet pretreatment were acquired through a LIBS experimental setup, and the modeling efficiency and prediction accuracy of the four chemometric methods were compared based on these spectral data. As a result, the modeling efficiency of PLSR was found to be the highest, that of SVR was the lowest, and that of ANN ranked third. In terms of prediction performance, ANN was found to work better than the other three chemometric methods, and the average absolute error (AAE) of prediction of ash content, volatile matter content and calorific value were 0.69%, 0.87%, and 0.56 MJ kg-1, respectively. ANN can seek the best compromise of modeling efficiency and prediction accuracy and is demonstrated to be an optimal multivariate calibration method with LIBS for online measurement of coal quality at thermal power plants.

With the development of miniaturized Laser-induced breakdown spectroscopy (LIBS) instruments, the desire for accuracy and stability of quantitative analysis of portable LIBS systems is growing. In this research, a hybrid deep learning model of a convolutional block attention module combined with a convolutional neural network and a long short-term memory (CBAM-CNN-LSTM) was for the first time used for quantitative analysis in a LIBS system with a portable spectrometer. Among them, the spectra collected by LIBS were in the form of the image, the CNN module was responsible for feature extraction of image spectra, mining deep features through the CBAM, and the simultaneous accurate quantitative analysis of Ca, Mg, Na and Ba was realized by the LSTM module. Meanwhile, a linear regression (LR), a CNN and a CNN-LSTM model were compared to the CBAM-CNN-LSTM model. The results showed that the performance of this model was much better than the LR model. More importantly, compared to CNN and CNN-LSTM, the average relative error of CBAM-CNN-LSTM was reduced by 80.5% and 68.1%, the average root mean square error was reduced by 56.7% and 53.4%, and the average stability was increased by 62.3% and 58.8%, respectively. In addition, the feature visualization results of CNN-LSTM and CBAM-CNN-LSTM displayed that CBAM can more effectively extract the features of characteristic peaks of corresponding elements and suppress the irrelative features. It indicated that the model can achieve an accurate and stable quantitative analysis of LIBS, which has the potential to be applied in the in-site analysis.

Spectral knowledge-based regression for laser-induced breakdown spectroscopy quantitative analysis

A feasibility of real-time quantitative multielemental analysis during parts growth by additive manufacturing technology has been demonstrating for the first time using laser induced breakdown spectroscopy (LIBS). A low weight and compact LIBS probe was developed to equip the laser cladding head installed at six axis industrial robot. LIBS quantitative analysis of key components (carbon and tungsten) has been demonstrated in real-time conditions during synthesis of high wear resistant coatings of nickel alloy reinforced with tungsten carbide particles. Hot solidified clad as well as a melt pool surface was sampled by LIBS probe but better analytical results were achieved for the latter case due to low partial sinking of heavy tungsten carbide particles in nickel matrix. LIBS system was calibrated for quantitative measurements by offline analysis with energy dispersive X-ray spectrometry (EDX). We didn’t observe any impact of laser ablation (LIBS sampling) on clad dimensions even when LIBS probe ablated melt pool. LIBS system was capable to quantify concentrations for both light (carbon, silicon) and heavy (nickel, tungsten, chromium, etc.) elements at frequency of 10 Hz. A good correspondence was achieved between online LIBS measurements and offline analysis opens wide perspectives for LIBS technique as a powerful tool for controlling process of parts growth by additive manufacturing

We present results indicating mass loading induced plasma matrix effects on the application of quantitative laser-induced breakdown spectroscopy (LIBS) for estimation of carbon contents in aerosols. An in-house flow-controlled powder-dispersion system generated carbonaceous aerosols with varying bi-modal particle size distributions (∼1 µm and 10 µm median diameters), thereby resulting in a wide mass loading range. For ease of chemical handling and to eliminate toxic effects, common talcum powder was used as our standard aerosol. Normalized atomic species concentrations of C, i.e., [C]/[Si] ratios, were calculated from atomic emission lines of C I (248 nm), Si I (252 nm), and plasma temperatures estimated from a series of Mg I lines. The results show a decrease in [C]/[Si] ratio to about 65% of the initial value as relative mass loadings increased (5.5–100%) due to the increase in number concentrations of larger sized particles (∼10 µm median diameter). As a comparison, normalized ratio of [Mg]/[Si] did not exhibit any marked change with increased mass loading. The normalized total absorption of photon flux across the C I (248 nm) spectral line indicated a strong correlation to the percentage decrease in [C]/[Si] ratio. We used an impactor with a cut-off size of around 10 µm diameter to generate mono-modal aerosolized powders (∼1 µm median diameter) that had lower relative mass loadings (0.32–0.16%). Similar LIBS analysis on these did not indicate any of the matrix effects. We conclude that for aerosol systems with widely varying mass loadings, quantitative LIBS analysis can be significantly affected by plasma matrix effects, specifically for the C I (248 nm) emission line as noticed in this study. This bears significance for the application of quantitative LIBS in the chemical characterization of all forms of carbonaceous aerosols.

Laser-induced breakdown spectroscopy (LIBS) has been applied to many fields for the quantitative analysis of diverse materials. Improving the prediction accuracy of LIBS regression models is still of great significance for the Mars exploration in the near future. In this study, we explored the quantitative analysis of LIBS for the one-dimensional ChemCam (an instrument containing a LIBS spectrometer and a Remote Micro-Imager) spectral data whose spectra are produced by the ChemCam team using LIBS under the Mars-like atmospheric conditions. We constructed a convolutional neural network (CNN) regression model with unified parameters for all oxides, which is efficient and concise. CNN that has the excellent capability of feature extraction can effectively overcome the chemical matrix effects that impede the prediction accuracy of regression models. Firstly, we explored the effects of four activation functions on the performance of the CNN model. The results show that the CNN model with the hyperbolic tangent (tanh) function outperforms the CNN models with the other activation functions (the rectified linear unit function, the linear function and the Sigmoid function). Secondly, we compared the performance among the CNN models using different optimization methods. The CNN model with the stochastic gradient descent optimization and the initial learning rate = 0.0005 achieves satisfactory performance compared to the other CNN models. Finally, we compared the performance of the CNN model, the model based on support vector regression (SVR) and the model based on partial least square regression (PLSR). The results exhibit the CNN model is superior to the SVR model and the PLSR model for all oxides. Based on the above analysis, we conclude the CNN regression model can effectively improve the prediction accuracy of LIBS.

As a reliable method, Laser induced breakdown spectroscopy (LIBS) is widely applied to elemental analysis in different fields. In this research, LIBS was used to analyze pharmaceutical tablets. For this purpose, spectra from eight samples of three pharmaceutical classes were acquired; and after peaks recognition, they were investigated by principal component analysis (PCA) and artificial neural network (ANN). According to the results, due to the similarity of the main elements of different samples, it is difficult to distinguish them by spectra; therefore, PCA was used to make a better comparison between the samples. Apart from that, unknown samples were identified and predicted through the ANN.

Laser induced breakdown spectrometry (LIBS) was applied for the determination of macro (P, K, Ca, Mg) and micronutrients (B, Cu, Fe, Mn and Zn) in sugar cane leaves, which is one of the most economically important crops in Brazil. Operational conditions were previously optimized by a neuro-genetic approach, by using a laser Nd:YAG at 1064 nm with 110 mJ per pulse focused on a pellet surface prepared with ground plant samples. Emission intensities were measured after 2.0 μs delay time, with 4.5 μs integration time gate and 25 accumulated laser pulses. Measurements of LIBS spectra were based on triplicate and each replicate consisted of an average of ten spectra collected in different sites (craters) of the pellet. Quantitative determinations were carried out by using univariate calibration and chemometric methods, such as PLSR and iPLS. The calibration models were obtained by using 26 laboratory samples and the validation was carried out by using 15 test samples. For comparative purpose, these samples were also microwave-assisted digested and further analyzed by ICP OES. In general, most results obtained by LIBS did not differ significantly from ICP OES data by applying a t-test at 95% confidence level. Both LIBS multivariate and univariate calibration methods produced similar results, except for Fe where better results were achieved by the multivariate approach. Repeatability precision varied from 0.7 to 15% and 1.3 to 20% from measurements obtained by multivariate and univariate calibration, respectively. It is demonstrated that LIBS is a powerful tool for analysis of pellets of plant materials for determination of macro and micronutrients by choosing calibration and validation samples with similar matrix composition.

Rapid and successful clinical diagnosis and bacterial infection treatment depend on accurate identification and differentiation between different pathogenic bacterial species. A lot of efforts have been made to utilize modern techniques which avoid the laborious work and time-consuming of conventional methods to fulfill this task. Among such techniques, laser-induced breakdown spectroscopy (LIBS) can tell much about bacterial identity and functionality. In the present study, a sensitivity-improved version of LIBS, i.e. nano-enhanced LIBS (NELIBS), has been used to discriminate between two different bacteria (Pseudomonas aeruginosa and Proteus mirabilis) belonging to different taxonomic orders. Biogenic silver nanoparticles (AgNPs) are sprinkled onto the samples’ surface to have better discrimination capability of the technique. The obtained spectroscopic results of the NELIBS approach revealed superior differentiation between the two bacterial species compared to the results of the conventional LIBS. Identification of each bacterial species has been achieved in light of the presence of spectral lines of certain elements. On the other hand, the discrimination was successful by comparing the intensity of the spectral lines in the spectra of the two bacteria. In addition, an artificial neural network (ANN) model has been created to assess the variation between the two data sets, affecting the differentiation process. The results revealed that NELIBS provides higher sensitivity and more intense spectral lines with increased detectable elements. The ANN results showed that the accuracy rates are 88% and 92% for LIBS and NELIBS, respectively. In the present work, it has been demonstrated that NELIBS combined with ANN successfully differentiated between both bacteria rapidly with high precision compared to conventional microbiological discrimination techniques and with minimum sample preparation.

The increasing demand for organic chicken eggs has led to cases of counterfeit organic eggs being sold, often produced using commercially-fed chickens. A reliable and accurate method is therefore required to verify the authenticity of organic eggs. This study aims to determine the authenticity of organic eggs by analyzing emission spectra from eggshells using Laser Induced Breakdown Spectroscopy (LIBS) combined with Artificial Neural Networks (ANN). LIBS was chosen for its modern analytical capability with minimal sample preparation, while ANN provides robust classification support. The samples consisted of organic and non-organic eggshells, collected from five groups with controlled feeding: one entirely organic-fed group (100% organic), a completely non-organic group (0% organic), and three other groups with combined feed. A neodymium-doped yttrium aluminum garnet (Nd-YAG) laser was applied to the samples at 1 ATM pressure, producing plasma in an atmospheric air environment. Emission spectra from the plasma were captured by an Optical Multichannel Analyzer (OMA) for elemental analysis. The ANN model was then built to classify the samples based on the spectral data, successfully demonstrating a good classification ability in distinguishing between sample groups. The elemental composition of eggshell samples included Ca, C, O, P, S, Mg, V, Pb, Cr, Hg, Na, Zn, Cu, N, K, and Sr, with Ca showing the highest intensity due to its abundance. These findings demonstrate the potential of LIBS integrated with ANN for organic egg authenticity verification and suggest its applicability in detecting compositional variations at different authenticity levels.

The convolutional neural network (CNN) has been widely used in image recognition field due to its good performance. This paper proposes a facial expression recognition method based on the CNN model. Regarding the complexity of the hierarchic structure of the CNN model, the activation function is its core, because the nonlinear ability of the activation function really makes the deep neural network have authentic artificial intelligence. Among common activation functions, the ReLu function is one of the best of them, but it also has some shortcomings. Since the derivative of the ReLu function is always zero when the input value is negative, it is likely to appear as the phenomenon of neuronal necrosis. In order to solve the above problem, the influence of the activation function in the CNN model is studied in this paper. According to the design principle of the activation function in CNN model, a new piecewise activation function is proposed. Five common activation functions (i.e., sigmoid, tanh, ReLu, leaky ReLus and softplus–ReLu, plus the new activation function) have been analysed and compared in facial expression recognition tasks based on the Keras framework. The Experimental results on two public facial expression databases (i.e., JAFFE and FER2013) show that the convolutional neural network based on the improved activation function has a better performance than most-of-the-art activation functions.