|
Near infrared spectroscopy can be used to determine the amount of ash and char present in various types of biomass, according to researchers in Japan and the US. An informatics model based on multivariate analysis of the NIR data will allow those working with and using biomass to predict the levels of such burnt materials in advance of processing. Nicole Labbé and Nicolas André of the Forest Products Center and Hyun-Woo Cho and Myong Jeong of the Department of Industrial and Information Engineering, at the University of Tennessee, Knoxville, Tennessee, together with Seung-Hwan Lee of the Biomass Technology Research Center, at the National Institute of Advanced Industrial Science and Technology, in Hiroshima, Japan, describe a predictive model for improving the outcome of biomass conversion. Oil fuels politics, but biomass lights fires. "The utilization of biomass has gained increased importance due to threats of uncertain petroleum supply in the near future and concerns about environmental pollution," say Labbé and colleagues. They point out that cellulose plant matter could be the most exploitable form of biomass, however, its widespread adoption is partly hindered by factors that affect it conversion, such as physico-chemical, structural and compositional factors. The researchers explain that mineral impurities can stifle the complete pyrolysis and gasification of biomass, for instance. The end result is formation of metallised carbon deposits within processing equipment. This slagging can leads to corrosion and damage to machinery reducing life and adding costs to what should be an effectively renewable process. Conversely, some mineral elements might actually play a role in catalysing the biomass conversion and so could be beneficial in terms of the overall energy/cost balance. Understanding the exact factors that might be fine-tuned to optimise the processing of biomass from various source to make the whole process as clean and efficient as possible requires insights into the chemical nature of this often highly complex raw feedstock. By turning to NIR spectroscopy, which has been used previously to characterize the organic composition of biomass, the international team hoped to develop a model using cheminformatics techniques to classify biomass materials. Such materials might include preservative-treated wood, which can have varying levels of mineral and other components. The researchers point out that NIR spectroscopy, unlike ultraviolet/visible and mid-infrared spectroscopy, is perfectly suited for statistics and chemometrics approaches. Indeed, because the majority of peaks in an NIR spectrum are overtones and combination peaks of molecular vibrations, principal component analysis (PCA) and partial least squares (PLS), and other multivariate data analytical techniques, are essential for extracting information from these spectra. This is particularly true for complex resources including wood and wood composites being developed as raw biomass materials. Aside from mineral content, char and ash content is another important aspect of the biomass process that requires the same kind of analytical approach. The team explains that predicting the levels of burnt carbon content is essentially a multivariate calibration problem too. "A main objective of the calibration model is to predict these properties from experimental or historical NIR spectral data before any processing, such as gasification or fermentation, [is carried out]" they explain. However, constructing a calibration model with so many different types of variable involved has proved difficult or impossible in other studies. Now, the team has tested several statistical approaches such as principal component analysis (PCA), orthogonal signal correction (OSC) treated PCA and partial least squares (PLS), Kernel PCA and PLS. Each was tested on the raw near infrared spectroscopic data from wood source materials and on data subjected to pre-treatments. This included multiplicative scatter correction and orthogonal signal correction. "The model with the highest coefficient of correlation and the lowest root mean square error of prediction was obtained with OSC-treated Kernel PLS method," they say. The work demonstrates a proof of principle in the prediction of ash and char content in unknown biomass materials. "The number of samples and types of biomass was limited for this specific study," the researchers concede. In order to produce robust calibration model of widespread application, they add that, "More samples and types of biomass need to be analysed." Related links:
|
|
