Automated determination of poplar chip size distribution based on combined image and multivariate analyses. (February 2015)

• Record Type: Journal Article
• Authors: Febbi, Paolo; Menesatti, Paolo; Costa, Corrado; Pari, Luigi; Cecchini, Massimo
• Is Part Of: Biomass and bioenergy. Volume 73(2015:Feb.)
• Page Start: 1
• Page End: 10
• Publication Date: 2015-02
• Subjects: Cumulative size distribution curve -- Sieving -- Size classification -- Biofuel quality determination -- Modeling -- Partial least squares; Cmin chip width measured by digital caliper -- Cmax chip length measured by digital caliper -- Dmin minimum Feret diameter -- Dmax maximum Feret diameter -- FD Fourier descriptor -- LV latent variable -- P designation for particle size distribution -- PLS-DA partial least squares discriminant analysis -- PLS-R partial least squares regression -- RMSEC root-mean-square error of calibration -- RMSECV root-mean-square error of cross-validation -- RPD ratio of percentage deviation -- SRF short rotation forestry -- VIP variable importance in the projection; Biomass energy -- Periodicals; Biomass -- Periodicals; Energy-Generating Resources -- Periodicals; Bioénergie -- Périodiques; 333.9539
• DOI: 10.1016/j.biombioe.2014.12.001
• Languages: English
• ISSNs: 0961-9534

Abstract

Abstract: The European technical standard EN 14961 on solid biofuels determines the fuel quality classes and specifications for wood chips. Sieving methods are currently used for the determination of particle size distribution. Some authors suggested that image analysis tools could provide methods for a more accurate measure of size integrated with shape. This work for the first time analyzes how image analysis combined with multivariate modeling methods could be used to construct cumulative size distribution curves based on chip mass (or weight). This has been done through a Partial Least Squares Regression model for the weight prediction of poplar chips and Partial Least Squares Discriminant Analysis models for estimation of chips size classification. Images of 7583 poplar chips were analyzed to extract size and shape descriptors (area, major and minor axis lengths, perimeter, eccentricity, equivalent diameter, fractal dimension index, Feret diameters and Fourier descriptors). The weight prediction model showed a high accuracy ( r = 0.94). The chip classification based on three size fractions (8–16 mm, 16–45 mm and 45–63 mm), with or without Fourier descriptors, showed accuracies equal to 92.9% of correct classification for both models in the independent test. The combination of image analysis with multivariate modeling approaches allow a better conversion of image analysis results to sieve results using the esteemed weight. The proposed method will allow to standardize processes applicable by biofuels laboratories and machinery certifiers. Highlights: Image analysis protocols were used to determine quality classes and dimensions of wood chips. 2-D shape and size descriptors were extracted. PLS-R model was adopted to predict chips' weight. PLS-DA models were adopted to predict chips' size fraction. Cumulative size distribution curves based on predicted chip mass were constructed.