Real-time monitoring of carbon concentration using laser-induced breakdown spectroscopy and machine learning

Rare earth metals detection and recognition based on laser induced breakdown spectroscopy and machine learning

The rapid detection and identification of the electronic waste (e-waste) containing rare earth (RE) elements is of great significance for the recycling of RE elements. However, the analysis of these materials is extremely challenging due to extreme similarities in appearance or chemical composition. In this research, a new system based on laser induced breakdown spectroscopy (LIBS) and machine learning algorithms is developed for identifying and classifying e-waste of rare-earth phosphors (REPs). Three different kinds of phosphors are selected and the spectra is monitored using this new developed system. The analysis of phosphor spectra shows that there are Gd, Yd, and Y RE element spectra in the phosphor. The results also verify that LIBS could be used to detect RE elements. An unsupervised learning method, principal component analysis (PCA), is used to distinguish the three phosphors and training data set is stored for further identification. Additionally, a supervised learning method, backpropagation artificial neural network (BP-ANN) algorithm is used to establish a neural network model to identify phosphors. The result show that the final phosphor recognition rate reaches 99.9%. The innovative system based on LIBS and machine learning (ML) has the potential to improve rapid in situ detection of RE elements for the classification of e-waste.

Identification of the interference spectra of edible oil samples based on neighborhood rough set attribute reduction

Due to numerous edible oil safety problems in China, an automatic oil quality detection technique is urgently needed. In this study, rough set theory and Fourier transform spectrum are combined for proposing a digital identification method for edible oil. First, the Fourier transform spectra of three different types of edible oil samples, including colza oil, waste oil, and peanut oil, are measured. After the input spectra are differentially and smoothly processed, the characteristic wavelength bands are selected with neighborhood rough set attribution reduction (NRSAR). Moreover, the classification models are established based on random forest (RF) and extreme learning machine (ELM) algorithms. Finally, confusion matrix, classification accuracy, sensitivity, specificity, and the distribution of judgment are calculated for evaluating the classification performances of different models and determining the optimal oil identification model. The results show that by using the third-order difference pre-processing method, 193 wavelength bands in the visible range can be reduced to 10 characteristic wavelengths, with a compression ratio of over 88.61%. Using the established NRS-RF and NRS-ELM models, the total identification accuracies are 91.67% and 93.33%, respectively. In particular, the identification accuracy of peanut oil using the NRS-ELM model reaches up to 100%, whereas the identification accuracies obtained using the principal component analysis (PCA)-based models that are commonly used in information processing (PCA-RF and PCA-ELM) are 81.67% and 90.00%, respectively. As compared with feature extraction methods, the proposed NRSAR shows directive advantages in terms of precision, sensitivity, specificity, and the distribution of judgment. In addition, the execution time is also reduced by approximately 1/3. Conclusively, the NRSAR method and NRS-ELM the model in the spectral identification of edible oil show favorable performance. They are expected to bring forth insightful oil identification techniques.

Simultaneous measurement of carbon emission and gas temperature via laser-induced breakdown spectroscopy coupled with machine learning

A method, which can accurately measure carbon emission and gas temperature simultaneously in real-time from a laser-induced breakdown spectrum (LIBS) via machine learning, is proposed in this study. In typical, peak intensity ratios had been used to map species concentrations prior to plasma formation, after removing the broadband continuum of the spectrum; however, the dependence of these peak intensity ratios on the concentration changes with the change in gas density. Therefore, considering the fact that the strength and shape of this broadband continuum is a function of the gas density for a given optical setup, we attempted to collect a spectrum by shortening the time delay after the laser fire, such that the spectrum can contain some of the broadband continuum. Since the analytical quantification of this broadband continuum is not trivial, we employed a machine learning approach to acquire a model that simultaneously predicts the gas temperature and CO₂ concentration. The predictive performance of the model trained with spectra that contain the broadband continuum was much better than that without it; the gradient-weighted regression activation mapping (Grad-RAM) analysis revealed that the model utilizes the broadband spectrum for temperature prediction and correction of changes in peak intensity due to temperature changes in the concentration prediction process.

Real-time in-situ optical detection of fluid viscosity based on the Beer-Lambert law and machine learning

As an important physical quantity to describe the resistance of fluid to flow, viscosity is an essential property of fluids in fluid mechanics, chemistry, medicine, as well as hydraulic engineering. While traditional measurement methods, including the rotating-cylinder method, capillary tube method and falling sphere method, have significant drawbacks especially in terms of accuracy, response time and the sample must be made to move. In this work, a novel Beer-Lambert law-based method was proposed for the viscosity measurement. Specifically, this work demonstrates that viscosity can be quantitatively reflected by spectral line intensity, and castor oil was selected due to its viscous temperature properties (viscosity has been accurately measured under different temperature), and its transmission spectrum at different temperatures ranging from 10 to 50°C was detected firstly. Then, the principal component analysis (PCA) was employed to obtain the intrinsic features of the transmission spectrum. Finally, the processed data was utilized to train and verify the radial basis function (RBF) neural network. As a result, the accuracy of the predictions conducted by means of the RBF reached 98.45%, which indicates the complicated and non-linear relationships between spectra formation and viscosity can be depicted well by RBF. The results show that the real-time in-situ optical detection method adopted in this work represents a great leap forward in the viscosity measurement, which fundamentally reforms the traditional viscosity measurement methods.

Construction of classification models for pathogenic bacteria based on LIBS combined with different machine learning algorithms

Bacteria, especially foodborne pathogens, seriously threaten human life and health. Rapid discrimination techniques for foodborne pathogens are still urgently needed. At present, laser-induced breakdown spectroscopy (LIBS), combined with machine learning algorithms, is seen as fast recognition technology for pathogenic bacteria. However, there is still a lack of research on evaluating the differences between different bacterial classification models. In this work, five species of foodborne pathogens were analyzed via LIBS; then, the preprocessing effect of five filtering methods was compared to improve accuracy. The preprocessed spectral data were further analyzed with a support vector machine (SVM), a backpropagation neural network (BP), and k-nearest neighbor (KNN). Upon comparing the capacity of the three algorithms to classify pathogenic bacteria, the most suitable one was selected. The signal-to-noise ratio and mean square error of the spectral data after applying a Savitzky-Golay filter reached 17.4540 and 0.0020, respectively. The SVM algorithm, BP algorithm, and KNN algorithm attained the highest classification accuracy for pathogenic bacteria, reaching 98%, 97%, and 96%, respectively. The results indicate that, with the support of a machine learning algorithm, LIBS technology demonstrates superior performance, and the combination of the two is expected to be a powerful tool for pathogen classification.