Classification of steel using laser-induced breakdown spectroscopy combined with deep belief network

Combination of plasma acoustic emission signal and laser-induced breakdown spectroscopy for accurate classification of steel

BACKGROUND

Fast and accurate classification of steel can effectively improve industrial production efficiency. In recent years, the use of laser-induced breakdown spectroscopy (LIBS) in conjunction with other techniques for material classification has been developing. Plasma Acoustic Emission Signal (PAES) is a type of modal information separate from spectra that is detected using LIBS, and it can reflect some of the sample's physicochemical information. Existing research has not addressed the use of LIBS in conjunction with PAES for steel classification and identification, thus it is quite interesting to examine a speedy steel classification approach using LIBS and PAES.

RESULTS

In this work, we used LIBS and PAES mid-level data fusion methods to classify and identify eight steel samples. We recorded the LIBS spectral data and PAES data of the eight samples synchronously, respectively, and proposed three novel mid-level data fusion strategies (additive fusion, splicing fusion, and multiplicative fusion). We have discussed the classification results by using machine learning algorithms. The conclusion revealed that the average accuracy of classifying a single LIBS spectrum is 72.5 %, whereas the average accuracy of classifying a single PAES data is 78.75 %. By combining LIBS spectral data and PAES data in the middle layer, the average accuracy of the splicing fusion classification result is 87.5 %, and the average accuracy of the multiplication fusion classification result is 86.25 %. Meanwhile, we have also found that thermal hardness may be an important physical factor affecting the acoustic emission signal of steel plasma.

SIGNIFICANCE

Accurate steel classification is achieved by combining spectral and acoustic data. This approach is anticipated to be used in the future to quickly classify large amounts of steel in industrial settings, leading to a notable increase in the efficiency of industrial production.

Modulate the laser phase to improve the ns-LIBS spectrum signal based on orbital angular momentum

Aiming to enhance the ns-LIBS signal, in this work, we introduced orbital angular momentum to modulate the laser phase of the Gaussian beam into the vortex beam. Under similar incident laser energy, the vortex beam promoted more uniform ablation and more ablation mass compared to the Gaussian beam, leading to elevated temperature and electron density in the laser-induced plasma. Consequently, the intensity of the ns-LIBS signal was improved. The enhancement effects based on the laser phase modulation were investigated on both metallic and non-metallic samples. The results showed that laser phase modulation resulted in a maximum 1.26-times increase in the peak intensities and a maximum 1.25-times increase in the signal-to-background ratio (SBR) of the Cu spectral lines of pure copper for a laser energy of 10 mJ. The peak intensities of Si atomic spectral lines were enhanced by 1.58-1.94 times using the vortex beam. Throughout the plasma evolution process, the plasma induced by the vortex beam exhibited prolonged duration and a longer continuous background, accompanied by a noticeable reduction in the relative standard deviation (RSD). The experimental results demonstrated that modulation the laser phase based on orbital angular momentum is a promising approach to enhancing the ns-LIBS signal.

Laser Ablation Plasmas and Spectroscopy for Nuclear Applications

The development of measurement methodologies to detect and monitor nuclear-relevant materials remains a consistent and significant interest across the nuclear energy, nonproliferation, safeguards, and forensics communities. Optical spectroscopy of laser-produced plasmas is becoming an increasingly popular diagnostic technique to measure radiological and nuclear materials in the field without sample preparation, where current capabilities encompass the standoff, isotopically resolved and phase-identifiable (e.g., UO and UO2 ) detection of elements across the periodic table. These methods rely on the process of laser ablation (LA), where a high-powered pulsed laser is used to excite a sample (solid, liquid, or gas) into a luminous microplasma that rapidly undergoes de-excitation through the emission of electromagnetic radiation, which serves as a spectroscopic fingerprint for that sample. This review focuses on LA plasmas and spectroscopy for nuclear applications, covering topics from the wide-area environmental sampling and atmospheric sensing of radionuclides to recent implementations of multivariate machine learning methods that work to enable the real-time analysis of spectrochemical measurements with an emphasis on fundamental research and development activities over the past two decades. Background on the physical breakdown mechanisms and interactions of matter with nanosecond and ultrafast laser pulses that lead to the generation of laser-produced microplasmas is provided, followed by a description of the transient spatiotemporal plasma conditions that control the behavior of spectroscopic signatures recorded by analytical methods in atomic and molecular spectroscopy. High-temperature chemical and thermodynamic processes governing reactive LA plasmas are also examined alongside investigations into the condensation pathways of the plasma, which are believed to serve as chemical surrogates for fallout particles formed in nuclear fireballs. Laser-supported absorption waves and laser-induced shockwaves that accompany LA plasmas are also discussed, which could provide insights into atmospheric ionization phenomena from strong shocks following nuclear detonations. Furthermore, the standoff detection of trace radioactive aerosols and fission gases is reviewed in the context of monitoring atmospheric radiation plumes and off-gas streams of molten salt reactors. Finally, concluding remarks will present future outlooks on the role of LA plasma spectroscopy in the nuclear community.

Interpretable machine learning assisted spectroscopy for fast characterization of biomass and waste

The combination of machine learning and infrared spectroscopy was reported as effective for fast characterization of biomass and waste (BW). However, this characterization process is lack of interpretability towards its chemical insights, leading to less satisfactory recognition for its reliability. Accordingly, this paper aimed to explore the chemical insights of the machine learning models in the fast characterization process. A novel dimensional reduction method with significant physicochemical meanings was thus proposed, where the high loading spectral peaks of BW were selected as input features. Combined with functional groups attribution of these spectral peaks, the machine learning models established based on the dimensionally reduced spectral data could be explained with clear chemical insights. The performance of classification and regression models between the proposed dimensional reduction method and principal component analysis method was compared. The influence mechanism of each functional group on the characterization results were discussed. CH deformation, CC stretch & CO stretch and ketone/aldehyde CO stretch played essential roles in C, H/ LHV and O prediction, respectively. The results of this work demonstrated the theoretical fundamentals of the machine learning and spectroscopy based BW fast characterization method.

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.