Nicolas Andre

Extraction of information from laser-induced breakdown spectroscopy spectral data by multivariate analysis

Laser-induced breakdown spectroscopy (LIBS) is being proposed more and more as a high-throughput technology to assess the elemental composition of materials. When a specific element is of interest, semiquantification is possible by building a calibration model using the emission line intensity of this element for known samples. However, a unique element has usually more than one emission line, and there are many examples where several emission lines used in combination give dramatically better results than any of the individual variables used alone. With a multivariate approach, models can be constructed that take into account all the emission lines related to a specific element; therefore more robust models can be developed. In this work, chemometric methods such as principal component analysis and partial least squares are proposed to resolve and extract useful information from the LIBS spectral data collected on biological materials.

Enhanced discrimination and calibration of biomass NIR spectral data using non-linear kernel methods.

Rapid methods for the characterization of biomass for energy purpose utilization are fundamental. In this work, near infrared spectroscopy is used to measure ash and char content of various types of biomass. Very strong models were developed, independently of the type of biomass, to predict ash and char content by near infrared spectroscopy and multivariate analysis. Several statistical approaches such as principal component analysis (PCA), orthogonal signal correction (OSC) treated PCA and partial least squares (PLS), Kernel PCA and PLS were tested in order to find the best method to deal with near infrared data to classify and predict these biomass characteristics. The model with the highest coefficient of correlation and the lowest RMSEP was obtained with OSC-treated Kernel PLS method.

Prediction of internal bond strength in a medium density fiberboard process using multivariate statistical methods and variable selection

This paper presents new data mining-based multivariate calibration models for predicting internal bond strength from medium density fiberboard (MDF) process variables. It utilizes genetic algorithms (GA) based variable selection combined with several calibration methods. By adopting a proper variable selection scheme, the prediction performance can be improved because of the exclusion of non-informative variable(s). A case study using real plant data showed that the calibration models based on the process variables selected by GA produced better performance than those without variable selection, with the exception of the radial basis function (RBF) neural networks model. In particular, the calibration model based on supervised probabilistic principal component analysis (SPPCA) yielded better performance (only when using GA) than partial least squares (PLS), orthogonal-PLS (O-PLS), and radial basis function neural networks models. The SPPCA model benefits most from the use of GA-based variable selection in this case study.

On-Line Monitoring of the Buffer Capacity of Particleboard Furnish by Near-Infrared Spectroscopy

This paper reports on a study of on-line monitoring of the buffer capacity of particleboard furnish using near-infrared (NIR) spectroscopy and multivariate analysis models (chemometrics). The buffer capacity of wood furnish is known to affect the quality of polymerization and the curing rates of urea-formaldehyde (UF) resins, which may affect the mechanical properties of manufactured panel. The first phase of the study consisted of building multivariate calibration and validation models from NIR spectroscopy data to predict the buffer capacity of particleboard furnish in a laboratory environment. During this phase, a spectrometer (Ocean Optics USB2000) operating in the 550–1100 nm spectral range was evaluated. The second phase of the study took place at a North American particleboard plant over several weeks. Several multivariate calibration models were constructed and tested on-line during a four-day test period. The on-line root mean square error of prediction (RMSEP) and the coefficient of variation (CV) for buffer capacity predictions ranged from 3.45 to 0.92 and 22.4% to 5.8%, respectively.

Two-dimensional homo- and hetero-correlation technique applied to NIR and py-MBMS spectra of wood

Abstract In this work, near infrared (NIR) and pyrolysis-molecular beam mass spectra (MBMS) of loblolly pine (Pinus taeda) were studied by means of generalized two-dimensional correlation spectroscopy in order to assign specific contributions of cellulose in the two spectral domains. Homo- and hetero-correlation techniques were employed to analyze the concentration-dependent spectral variations of cellulose. Specific bands of cellulose were assigned in the NIR and MBMS spectra, and moreover two masses m/z at 114 and 173, commonly assigned to cellulose fragments, were found to not originate from the pyrolysis of cellulose.

Real-time process modeling of particleboard manufacture using variable selection and regression methods ensemble

This study focuses on the real-time prediction of mechanical properties such as internal bond strength (IB) and modulus of rupture (MOR) for a wood composite panels manufacturing process. As wood composite panel plants periodically test their products, a real time data fusion application was developed to align laboratory mechanical test results and their corresponding process data. Fused data were employed to build regression models that yield real-time predicted mechanical property values when new process data become available. The modeling algorithm core uses genetic algorithm to preselect a meaningful subset of process variables. Calibration models are then built using several regression methods: multiple linear regression, ridge regression, neural networks, and partial least squares regression (PLS). Four different predicted response values were generated for each new record of real time process variables. On-line validation results showed good performance of the ridge regression method with a 0.89 correlation coefficient between actual and predicted MOR values, a root mean square error (RMSEP) of 1.05xa0MPa and a mean normalized error of 9xa0%. IB was best predicted by PLS with a 0.81 correlation coefficient between actual IB and PLS predicted IB values, a RMSEP of 75.1xa0kPa, and a mean normalized error of 15xa0%.ZusammenfassungDiese Studie beschäftigt sich mit der Echtzeitvorhersage mechanischer Eigenschaften wie der Querzugfestigkeit (IB) und der Biegefestigkeit (MOR) bei der Herstellung von Holzwerkstoffen. Da die Holzwerkstoffindustrie regelmäßig ihre Produkte prüft, wurde ein Verfahren zur Echtzeit-Datenfusion entwickelt, um die Ergebnisse mechanischer Laborprüfungen mit entsprechenden Prozessdaten zusammenzuführen. Diese Datensätze wurden zur Entwicklung von Regressionsmodellen verwendet, welche in Echtzeit vorhergesagte mechanische Kennwerte liefern, wenn neue Prozessdaten verfügbar sind. Der Modellalgorithmus bedient sich eines genetischen Algorithmus zur Vorauswahl einer aussagekräftigen Teilmenge von Prozessvariablen. Anschließend werden anhand verschiedener Regressionsverfahren (multiple lineare Regression, Ridge-Regression, neurale Netzwerke sowie Partial-Least-Square Regression (PLS)) Kalibrierungsmodelle erstellt. Für jeden neuen Satz von Echtzeit-Prozessvariablen wurden vier verschiedene Response-Variablen generiert. Online-Validierungsergebnisse zeigten ein gutes Ergebnis für das Ridge-Regressionsverfahren mit einem Korrelationskoeffizienten von 0,89 zwischen den im Labor bestimmten und den vorhergesagten Festigkeitswerten, einem mittleren vorhergesagten Fehler (RMSEP) von 1,05xa0MPa und einem mittleren normalisierten Fehler von 9xa0%. Die Querzugfestigkeit wurde am besten mit PLS vorhergesagt. Der Korrelationskoeffizient zwischen der im Labor bestimmten und der mittels PLS vorhergesagten Querzugfestigkeit betrug 0,81, der mittlere vorhersagbare Fehler 75,1xa0kPa und der mittlere normalisierte Fehler 15xa0%.

Spectral Analysis of Rare Earth Elements using Laser-Induced Breakdown Spectroscopy

There is growing interest in rapid analysis of rare earth elements (REEs) both due to the need to find new natural sources to satisfy increased demand in their use in various electronic devices, as well as the fact that they are used to estimate actinide masses for nuclear safeguards and nonproliferation. Laser-Induced Breakdown Spectroscopy (LIBS) appears to be a particularly well-suited spectroscopy-based technology to rapidly and accurately analyze the REEs in various matrices at low concentration levels (parts-per-million). Although LIBS spectra of REEs have been reported for a number of years, further work is still necessary in order to be able to quantify the concentrations of various REEs in realworld complex samples. LIBS offers advantages over conventional solution-based radiochemistry in terms of cost, analytical turnaround, waste generation, personnel dose, and contamination risk. Rare earth elements of commercial interest are found in the following three matrix groups: 1) raw ores and unrefined materials, 2) as components in refined products such as magnets, lighting phosphors, consumer electronics (which are mostly magnets and phosphors), catalysts, batteries, etc., and 3) waste/recyclable materials (aka e-waste). LIBS spectra for REEs such as Gd, Nd, and Sm found in rare earth magnets are presented.

Real-time dynamic simulation of continuous bulk material flow to improve the statistical modelling of final product strength properties

In regression analyses, correlations between independent variables (e.g. process variables) and dependent variables (e.g. product quality) are of major interest. However, only statistically significant correlations ensure a reliable interpretation of how process variables affect product qualities. In this respect, accurate time alignment of independent variables is crucial to obtain regression models with acceptable validation that are influenced by temporal phenomena (e.g. industrial processes) only. In this study, the commonly used static form of time alignment, where only the distances between consecutive process parameters and the average production speed are considered, is compared to a newly developed dynamic calculation of time lags. The dynamic calculation of time lags was achieved by modelling the continuous bulk material flow. The two different methods of calculation were then applied on an industrial production of particleboards to predict final board strength properties. Results of regression models showed that the use of dynamically calculated time lags improved the predictability of the internal bond strength of boards by 67% compared to statically calculated time lags. Consequently, final product strength properties could be predicted more accurately, which should lead to lower costs of rejects and a higher efficiency of material inputs.

Blended Feedstocks for Thermochemical Conversion: Biomass Characterization and Bio-Oil Production From Switchgrass-Pine Residues Blends

An abundant, low-cost, and high-quality supply of lignocellulosic feedstock is necessary to realize the large-scale implementation of biomass conversion technologies capable of producing renewable fuels, chemicals, and products. Barriers to this goal include the variability in the chemical and physical properties of available biomass, and the seasonal and geographic availability of biomass. Blending several different types of biomass to produce consistent feedstocks offers a solution to these problems and allows for control over the specifications of the feedstocks. For thermochemical conversion processes, attributes of interest include carbon content, total ash, specific inorganics, density, particle size, and moisture content. In this work, a series of switchgrass and pine residues blends with varying physical and chemical properties were evaluated. Physical and chemical properties of the pure and blended materials were measured, including compositional analysis, elemental analysis, compressibility, flowability, density, and particle size distribution. To screen blends for thermochemical conversion behavior, the analytical technique, pyrolysis gas chromatography mass spectrometry (Py-GC/MS), was used to analyze the vapor-phase pyrolysis products of the various switchgrass/pine residues blends. The py-GC/MS findings were validated by investigating the bio-oils produced from the selected blends using a lab-scale fluidized-bed pyrolysis reactor system. Results indicate that the physical properties of blended materials are proportional to the blend ratio of pure feedstocks. In addition, pyrolysis of pine residues resulted in bio-oils with higher carbon content and lower oxygen content, while switchgrass derived pyrolysis products contained relatively greater amount of anhydrosugars and organic acids. The distribution of the pyrolysis vapors and isolated bio-oils appear to be a simple linear combination of the two feedstocks. The concentration of alkali and alkaline earth metals (Ca, K, Mg, and Na) in the blended feedstocks were confirmed to be a critical parameter due to their negative effects on the bio-oil yield. This work demonstrates that blending different sources of biomass can be an effective strategy to produce a consistent feedstock for thermochemical conversion.