Laser-induced breakdown spectroscopy (LIBS) and near-infrared spectroscopy (NIRS) are complementary spectroscopic techniques that provide elemental and molecular fingerprint information of materials, respectively. The integration of LIBS and NIRS enhances analytical accuracy and broadens application potential, attracting growing interest in scientific and industrial communities. However, existing data fusion methods for LIBS and NIRS often process each modality separately, neglecting cross-modal correlations and leaving the intrinsic connections between atomic emission lines and molecular absorption bands unexplored. To address this limitation, we propose a bidirectional cross-attention feature fusion network (Bi-CAFF), a novel data fusion method for improving quantification accuracy in LIBS-NIRS fused data analysis. Bi-CAFF utilises a bidirectional cross-attention mechanism to establish interactive feature correlations between LIBS and NIRS modalities. Importantly, we interpret these interactions based on physical knowledge, revealing a meaningful relationship between LIBS atomic emission lines and NIRS molecular absorption bands. Additionally, we introduce spectral feature distillation (SFD), where a student network trained on NIRS data learns refined features from a LIBS-based teacher model. This approach enhances the quantification accuracy of portable NIRS systems while preserving their cost-efficiency and field-deployment capabilities. Evaluations on industrial datasets show that Bi-CAFF outperforms baseline fusion methods, reducing mean absolute error (MAE) by 12.1-60.3 % and root-mean-square error (RMSE) by 12.4-54.1 %. Moreover, SFD achieves MAE and RMSE reductions of 6.6-49.2 % and13.1-46.7 %, respectively, compared to standalone NIRS-based models. Together, these innovations advance the state-of-the-art in machine learning-assisted chemical analysis, laying the groundwork for a new paradigm in multimodal spectral data analysis.

The growing demand for rapid, on-site chemical analysis has driven the development of cost-effective smartphone-based sensors. However, such methods are typically confined to the visible range and lack spectral selectivity and chemical specificity, which limits their accuracy and reliability for elemental quantification and mapping. In this work, we propose a laser-induced breakdown spectroscopy (LIBS)-enhanced smartphone video imaging (LE-SVI) approach that leverages LIBS-derived spectral fingerprints to overcome these limitations. By establishing a cross-modal association between videos and spectra, LE-SVI transfers spectral knowledge to the imaging modality, allowing colour patterns to be interpreted as compositional information. During inference, only a smartphone is required for rapid measurements, yielding elemental content predictions and spatial distribution maps. Experimental results demonstrate that LE-SVI significantly improves the accuracy of NiO content prediction in composite samples compared to a smartphone-only benchmark, with the coefficient of determination increasing from 0.957 to 0.979 on the test set. Furthermore, the method enables fast and high-resolution mapping of relative Mg content in rock samples by utilising the reconstructed LIBS characteristic line intensity. LE-SVI offers a cross-modal knowledge transfer strategy to compensate for the hardware limitations of smartphone-based sensors, enhancing their analytical performance beyond conventional constraints. It achieves an advantageous trade-off between speed, cost and accuracy, providing a high-throughput and reliable solution for rapid elemental analysis in field-deployable applications.

Exploration for critical minerals, such as beryllium (Be) and uranium (U), requires accurate reserve assessment, for which drill core analysis is essential. Techniques like laser-induced breakdown spectroscopy (LIBS) and X-ray fluorescence spectroscopy (XRF) are widely used for rapid core analysis but have some limitations. LIBS suffers from poor sensitivity for low-concentration U, while XRF cannot detect Be. Furthermore, matrix effects in both techniques hinder the accurate simultaneous quantification of Be and U. We introduce a novel LIBS-XRF method for the simultaneous measurement of Be and low-concentration U. The methodology involves first analyzing samples with XRF and LIBS. Subsequently, a support vector machine (SVM) algorithm classifies the samples based on the XRF data. A separate predictive model is then developed for each category. A basic linear model is constructed using the spectral line of the target elements as the dominant factor based on dominant factor (DF) modeling strategy, and machine learning algorithms are then used to compensate for the residuals of this basic model. Tests on ore cores demonstrated that this method significantly reduces quantification errors. The achieved mean relative errors were 7.58% for Be and 7.02% for U. These results represent improvements of 61.42%/77.20% and 69.77%/72.48% over conventional unclassified and experience-based methods, respectively. This work is the first to use a LIBS-XRF approach for the highly accurate and simultaneous detection of Be and low-concentration U in ore cores, proving its high practical utility for this application.

Ideal decomposition of severely overlapped spectra due to limited spectral resolution remains an unsolved puzzle for spectroscopic technologies. Although ultra-high-resolution spectrometers can resolve spectral overlap problems, they often entail high costs and significant signal loss, which naturally leads to relative high limit of detection (LOD). In this work, based on the assumption that the measured spectrum is the convolution of the much less overlapped real plasma emission profile and the spectrometer instrumental response, we propose a center-wavelength coupled with broadening-width-ratio constrained decomposition (CC-BCD) method for severely overlapped peaks from relatively low-resolution spectrometers. More specifically, the method incorporates the extra supporting information (the central wavelengths and the ratio of broadening width of the overlapped peaks obtained from ultra-high-resolution spectra) as hard constraints into the decomposition model, transforming the model from an otherwise underdetermined mathematical fit into a physically guided reconstruction and enabling accurate and stable resolution of severely overlapped peaks. The method was successfully applied for plasma emission technology such as laser-induced breakdown spectroscopy (LIBS) and spark-discharge optical emission spectroscopy (SD-OES). For uranium ores analysis using LIBS, the completely overlapped peaks (U II 385.957 nm and Fe I 385.991 nm) were fully resolved, reducing LOD to 7.3 mg/kg, two orders of magnitude compared with that of using ultra-high-resolution spectrometer and the lowest record ever for ores. For brass sample detection using SD-OES, the severely overlapped peaks (Zn I 328.233 nm and Cu I 328.272 nm) were also resolved, reducing LOD for Zn from 0.39 wt% to 0.11 wt%. The proposed method enables relatively low-resolution spectrometers to achieve high resolution capabilities while retaining the high optical throughput, thereby providing a highly sensitive and low-cost approach for scenarios where the analysis heavily relies on severely overlapped lines, such as ultra-high-sensitivity analysis of uranium in complex matrices.

This paper proposes a new method to improve the accuracy of analysing low-quality spectral data, namely twin spectral reconstruction network. It consists of two neural networks with shared weights and generates useful spectral fingerprints in low-quality spectra by learning from high-quality spectra. The proposed method is tested on a new and challenging task of identifying fire-retardant coating (FRC) brands using low-quality spectra under small sample conditions. It significantly improves the identification accuracy compared to the baseline classifiers, and the reconstructed high-quality spectra closely resemble the target spectra. In addition, this paper presents a low-cost approach for FRC identification using smartphone videos and machine learning. It records short videos of samples being illuminated by a colour-changing screen and converts them into spectral data. As a pre-screening tool, it yields an accuracy of 87 % and can greatly reduce the cost and complexity of FRC identification compared to baseline techniques.

In recent years, food fraud issues related to whey protein supplements have disrupted the market and caused significant concern among consumers. Conventional analytical methods such as HPLC and ion exchange chromatography are commonly used to detect adulteration in whey protein supplements. However, these methods are costly, time-consuming and require specialised operation, making them less suitable for a wider range of users. This study presents a rapid and reliable approach for verifying the authenticity of whey protein supplements using laser-induced breakdown spectroscopy (LIBS) and machine learning. Specifically, this approach is employed to identify 15 brands of whey protein concentration (WPC), quantify protein and carbohydrate concentrations, distinguish three types of adulterants, and predict the level of adulteration. The relationship between LIBS data and analyte labels is established using machine learning methods, including partial least squares regression (PLSR), partial least squares discriminant analysis (PLS-DA), and kernel extreme learning machine (K-ELM). The accuracy for identifying WPC brands was over 0.977, and the highest coefficient of determination (R2) for quantifying protein and carbohydrate contents was 0.984 and 0.978, respectively. In addition, different adulterants can be differentiated with accuracies exceeding 0.986, and the R2 values for adulteration prediction are above 0.967 in most cases. These results suggest that LIBS combined with machine learning can serve as a viable and efficient solution for detecting adulteration in whey protein supplements.

Whey protein supplements are gaining increasing popularity among health and fitness enthusiasts due to their ability to enhance protein anabolism and promote muscle recovery and building. The growing demand for whey protein supplements has led to a high incidence of food fraud, including the addition of cheap proteins and non-protein nitrogen sources, posing significant health risks and economic losses. This study presents the use of portable near-infrared (NIR) spectroscopy and visible near-infrared hyperspectral imaging (HSI) combined with machine learning to evaluate the quality and authenticity of whey protein supplements. Specifically, NIR and HSI data from 15 brands of whey protein concentration (WPC) samples were analysed using principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA) and kernel extreme learning machine (K-ELM), demonstrating distinct class separability and excellent classification accuracy. The protein and carbohydrate contents of the samples were effectively quantified using partial least squares regression (PLSR) and K-ELM, yielding the lowest root mean square error (RMSE) of 0.023 for both predictions. Moreover, useful spectral fingerprints related to protein and carbohydrate contents were identified based on the regression coefficients. In addition, three common adulterants, including maltodextrin, wheat flour and milk powder, at concentrations ranging from 5% to 50% (w/w) in WPC, were accurately detected and quantified. The RMSE for quantifying adulterant levels ranged from 0.009 to 0.026. These results suggest that NIR spectroscopy and HSI, in combination with machine learning, can provide a reliable and practical solution for assessing the quality and authenticity of whey protein supplements.

Creatine monohydrate is an important sports nutrition supplement that enhances energy and promotes muscle growth. Recent concerns about the quality and authenticity of creatine monohydrate have highlighted the urgent need for rapid and cost-effective assessment methods. This study presents a new approach for assessing the quality of creatine monohydrate using spectroscopy combined with machine learning. Spectral data of creatine monohydrate samples from 15 brands are acquired using portable near-infrared (NIR) spectroscopy and benchtop hyperspectral imaging (HSI). Machine learning methods are employed to extract high-level features from the spectral data and model the relationship between the data and creatine concentrations. The root mean square error (RMSE) for models based on NIR data ranges from 0.258 to 0.291, whereas those derived from HSI data vary between 0.468 and 0.576. To improve the accuracy and reliability of spectral data analysis, multi-model fusion and data fusion strategies are used to integrate the outputs of different models and data from different sources, respectively. By combining NIR-HSI data fusion with multi-model fusion, the lowest RMSE for creatine quantification is reduced to 0.18. These results demonstrate that spectroscopic techniques coupled with machine learning can provide a rapid and cost-effective solution for assessing the quality and authenticity of creatine monohydrate.

Laser-induced breakdown spectroscopy (LIBS) has been demonstrated as a promising technique for real-time combustion diagnosis due to its capacity for simultaneous multi-species analysis. The gradient of species concentration in reacting flows often coupled with variations in gas temperature which prevented accurate concentration measurement. The effect of gas temperature on spectral emission intensity and plasma property was comprehensively investigated by employing a Bunsen flame. With the increasing of gas temperature, less laser energy was deposited into the plasma, resulting in a monotonic decline in plasma volume, brightness and atomic emission intensity. Conversely, the plasma temperature improved due to fewer gas molecules being excited. The intensity of ionic lines and electron density were initially increased but subsequently decreased, with the reduction in gas density playing a dominant role at higher gas temperature. Intensity ratio pairs of C/O and H/O were found to be susceptible to gas temperature. The deviation of C/O ratio caused by gas temperature from burner nozzle to Bunsen tip (similar to 1100 degrees C) was about 28.4%. Clear elucidation of the effect of gas temperature provides reliable basis to accurate combustion diagnosis with LIBS.

With the growing interest in health and fitness, whey protein supplements are becoming increasingly popular among fitness enthusiasts and athletes. The surge in demand for whey protein supplements highlights the need for cost-effective methods to characterise product quality throughout the food supply chain. This study presents a rapid and low-cost method for authenticating sports whey protein supplements using smartphone video imaging (SVI) combined with machine learning. A gradient of colours ranging from purple to red is displayed on the front screen of a smartphone to illuminate the sample. The colour change on the sample surface is captured in a short video by the front-facing camera. Then, the video is split into frames, decomposed into RGB colour channels, and converted into spectral data. The relationship between video data and sample labels is established using machine learning models. The proposed method is tested on five tasks, including identifying 15 brands of whey protein concentrate (WPC), quantifying fat content and energy levels, detecting three types of adulterants, and quantifying adulterant levels. Moreover, the performance of SVI was compared to that of hyperspectral imaging (HSI), which has an equipment cost of around 80 times that of SVI. The proposed method achieves accuracies of 0.933 and 0.96 in WPC brand identification and adulterant detection, respectively, which are only around 0.05 lower than those of HSI. It obtains coefficients of determination of 0.897, 0.906 and 0.963 for the quantification of fat content, energy levels and milk powder adulteration, respectively. Such results demonstrate that the combination of smartphones and machine learning offers a low-cost and viable preliminary screening tool for verifying the authenticity of whey protein supplements.

BACKGROUND:Cement composition, including key oxides such as CaO, SiO2, Al2O3, and Fe2O3, plays a critical role in determining cement's strength and durability. Real-time monitoring of these components during cement production is essential for ensuring optimal raw material ratios. Spectroscopic techniques, such as Laser Induced Breakdown Spectroscopy (LIBS) and Near Infrared Spectroscopy (NIRS), offer significant potential for rapid and non-destructive cement analysis, but their individual limitations, such as matrix effects in LIBS and spectral overlap in NIRS, necessitate an integrated method to achieve accurate and stable results. RESULTS:In this study, we propose a novel fusion method based on a dual-branch convolutional neural network with an attention module (DBAM-CNN) to synergize LIBS and NIRS data for enhanced cement component quantification. The dual-branch CNN structure enables feature extraction of atomic and molecular information from LIBS and NIRS data, respectively, optimizing the global task of improving quantitative analysis by capturing complementary information from both spectroscopic techniques. These features are then fused, and spatial and channel attention modules are used to refine the feature weights, enabling the model to effectively capture spectral fingerprint information. Experimental results show that the DBAM-CNN outperforms both existing fusion strategies and single technologies, demonstrating exceptional performance in real-time, high-precision cement composition analysis. SHAP analysis further reveals that the method highlights key features in LIBS and NIRS, leading to enhanced quantitative outcomes. SIGNIFICANCE:The proposed DBAM-CNN method significantly enhances cement composition analysis by effectively integrating complementary information from LIBS and NIRS. By addressing issues such as information redundancy and feature loss that are common in existing fusion strategies, this approach offers a more reliable and efficient solution for real-time, high-precision monitoring in cement production. It represents an advancement in spectroscopic data fusion techniques, paving the way for improved cement quality control.

Accurate estimation of heating temperatures experienced by fire retardant coatings (FRCs) is crucial in identifying the ignition source during fire investigations. While traditional methods, such as spectroscopy, effectively capture the compositional changes in FRC at various heating temperatures, they are typically bulky, costly, and unsuitable for rapid field analysis. This study proposes the use of smartphone and machine learning to predict the heating temperatures of FRC. A smartphone is employed to capture short videos of FRC samples illuminated by its color-changing screen. Video frames are then decomposed into color images and converted into spectral data for further processing. Linear and nonlinear regression models are applied to identify key variables and enhance predictive accuracy. The performance of smartphone-based temperature estimation is compared to that of hyperspectral imaging and laser-induced breakdown spectroscopy. In the test phase, the coefficient of determination for smartphone-based estimation ranges from 0.946 to 0.962, often surpassing that of benchmark methods. These results indicate that smartphones can provide a low-cost, effective means for estimating heating temperatures of FRC in fire investigations.

This study presents a low-cost smartphone-based imaging technique called smartphone video imaging (SVI) to capture short videos of samples that are illuminated by a colour-changing screen. Assisted by artificial intelligence, the study develops new capabilities to make SVI a versatile imaging technique such as the hyperspectral imaging (HSI). SVI enables classification of samples with heterogeneous contents, spatial representation of analyte contents and reconstruction of hyperspectral images from videos. When integrated with a residual neural network, SVI outperforms traditional computer vision methods for ginseng classification. Moreover, the technique effectively maps the spatial distribution of saffron purity in powder mixtures with predictive performance that is comparable to that of HSI. In addition, SVI combined with the U-Net deep learning module can produce high-quality images that closely resemble the target images acquired by HSI. These results suggest that SVI can serve as a consumer-oriented solution for food authentication.

Camellia oil (CO) is known for its nutritional value and health benefits, but its high price makes it susceptible to adulteration. This study developed a binary adulteration system for CO in response to the adulteration of rapeseed oil (RO) into CO that been observed in the market. A total of 243 oil samples adulterated with various concentrations of RO were prepared. The spectral information of the adulterated oil samples was obtained using near-infrared (NIR) spectroscopy. Additionally, visual data obtained from smartphone-captured images and videos were analysed. Deep-learning models trained on video data reached the highest accuracy of 96.30 %. To improve detection accuracy, a multimodal approach was adopted by combing spectral and visual data. Generally, this study presented a novel method for detecting the authenticity of CO in real time, providing technical support to address increasingly serious food safety concerns and laying the foundation for future rapid online detection using smartphones.

Laser-induced breakdown spectroscopy (LIBS) has long been regarded as an ideal analytical technology with the unique capabilities of real-time and multielement sensing. However, the lack of a clear understanding of the impact of ambient gas properties on the LIBS signal has severely hindered LIBS quantification improvement. We proposed an innovative approach by applying neural networks to discover the dependence of the LIBS signal on the ambient gas properties supported with a series of purposely designed experiments. For the first time, the full picture of the dependence of the LIBS signal on the main gas properties was clearly discovered, and the impact mechanism was further clarified. It is not only the first time that AI was used for complicated physical dependence rather than quantification in LIBS and the spectroscopic field but also established a new paradigm for the application of AI in complicated physical dependence by constructing comprehensive data points that are virtually impossible to attain through traditional experimental methods.

BACKGROUND:Laser-induced breakdown spectroscopy (LIBS) has emerged as a powerful technique for equivalence ratio measurement, offering deeper insight into chemical reaction processes in combustion systems. However, in practical flame conditions, variations in the equivalence ratio are inherently coupled with variations in local gas temperature, which significantly affect the accuracy of LIBS measurements by altering spectral line intensity ratios. RESULTS:To mitigate the impact of local gas temperature, a hybrid correction method comprising two key steps is proposed. First, plasma-based reference signals are exploited to accurately characterize local gas temperature and correct the spectral line intensities accordingly. Second, the modified spectrum standardization method is applied to compensate for deviations in spectral line ratios. This approach significantly reduces deviations in line intensity ratios across different gas temperatures. Specifically, the AS_RSD of H/O ratio decreases from 25.96 % to 3.94 %. The H/O ratio measured at different flame positions can be regressed onto a unified calibration curve, with an R2 greater than 0.95. The robustness of the method is demonstrated on various line ratio pairs, such as H/O, C/O, H/N and C/N. This calibration model enables precise equivalence ratio determination throughout the entire flame field. SIGNIFICANCE:The hybrid correction method allows for the simultaneous and accurate measurement of both local gas temperature and equivalence ratio. By effectively mitigating the influence of temperature variations, this method represents a great progress in enhancing the accuracy of LIBS for equivalence ratio measurement in complex combustion environments.

LIBS with beam shaping plasma modulation technology can reduce spectral interference and improve the quantitative performance of uranium in di-uranate.

Laser-induced breakdown spectroscopy (LIBS) is regarded as the future superstar for analytical chemistry and widely applied in various fields. Improving the quality of LIBS signal is fundamental to achieving accurate quantification and large-scale commercialization of LIBS. To propose control methods that improve LIBS signal quality, it is essential to have a comprehensive understanding of the influence of key parameters, such as ambient gas pressure, temperature, and sample temperature on LIBS signals. To date, extensive research has been carried out. However, different researchers often yield significantly different experimental results for LIBS, preventing the formation of consistent conclusions. This greatly prevents the understanding of influencing laws of key parameters and the improvement of LIBS quantitative performance. Taking ambient gas pressure as an example, this paper compares the effects of ambient gas pressure under different optimization conditions, reveals the influence of spatiotemporal window caused by inherent characteristics of LIBS signal sources, i.e., intense temporal changes and spatial non-uniformity of laser-induced plasmas, on the impact patterns of key parameters. From the perspective of plasma spatiotemporal evolution, the paper elucidates the influence patterns of ambient gas pressure on LIBS signals, clarifying seemingly contradictory research results in the literature.

Laser-induced breakdown spectroscopy(LIBS)is an emerging atomic spectroscopy technique that has the advantages of low sample pre-treatment and rapid,in situ,and simultaneous multi-element measurements.LIBS demonstrates good prospects in the field of coal analysis.In recent years,chemometric and machine learning models have been widely used to improve the quantitative accuracy of LIBS in coal analysis.Generally,these models rely on a certain number of training samples to ensure the reliability of the prediction results.However,obtaining the certified content(label information)of coal samples used for model training requires traditional chemical analysis,which is complex and time-consuming.This may lead to insufficient training samples and poor model performance.To tackle the small sample problem in LIBS-based coal analysis,this work proposes a semi-supervised learning method based on the ensemble of multiple models.5 baseline models are first established based on the initial training set,including multiple linear regression(MLR),partial least squares regression(PLSR),locally weighted partial least squares regression(LW-PLSR),support vector regression(SVR),and kernel extreme learning machine(K-ELM).The unlabelled data are processed using the 5 models,and 5 prediction values are obtained.For each unlabelled sample,the standard deviation of the 5 prediction values is calculated,and the unlabelled sample corresponding to the smallest standard deviation is added to the training set.Its pseudo label is the average of the 5 prediction values.As the training set is iteratively expanded,its corresponding training model is updated.The final training model is optimized and used to analyse the test samples.The proposed method is tested on a coal dataset containing 20 training samples,39 test samples and 280 unlabelled samples.The results show that the proposed method improves the coefficient of determination(R2)for content prediction of fixed carbon,ash,and volatile by 0.033,0.102 and 0.118,respectively.Therefore,if the number of training samples is insufficient,semi-supervised learning can effectively improve the accuracy and reliability of LIBS quantification.

Saffron (Crocus sativus L.) is the most expensive spice in the world and is highly susceptible to economically motivated adulteration. Hyperspectral imaging (HSI) and near-infrared (NIR) spectroscopy are recently popular techniques for rapid, non-destructive detection of saffron adulteration. However, the high cost and measurement complexity of these techniques make them less suitable for consumer-level applications. In this work, a smartphone video-based system combined with chemometrics is used to detect saffron adulteration. A smartphone is utilised to capture short videos of pure and adulterated saffron powder samples illuminated by its colour-changing screen. The video frames are decomposed into RGB colour images and converted into spectral-like data. Partial least squares regression is used to model the relationship between data and saffron purity. The smartphone video-based system obtains a coefficient of determination of 0.9774 for prediction, which is comparable to reference techniques such as NIR spectroscopy and HSI. Moreover, it is efficient in presenting distribution maps related to saffron purity. In addition, when used in conjunction with the smartphone video-based system, HSI and NIR spectroscopy achieve higher performance without significantly increasing measurement cost and complexity. Such results suggest that the smartphone video-based system has the potential to be a viable primary screening and auxiliary tool for detecting saffron adulteration.