In chemometric studies all predictor variables are usually collected in one data matrix X. This matrix is then analyzed by PLS regression or other methods. When data from several different sub-processes are collected in one matrix, there is a possibility that the effects of some sub-processes may vanish. If there is, for instance, mechanic data from one process and spectral data from another, the influence of the mechanic sub-process may not be detected. An application of multi-block (MB) methods, where the X-data are divided into several data blocks is presented in this study. By using MB methods the effect of a sub-process can be seen and an example with two blocks, near infra-red, NIR, and process data, is shown. The results show improvements in modelling task, when a MB-based approach is used. This way of working with data gives more information on the process than if all data are in one X-matrix. The procedure is demonstrated by an industrial continuous process, where knowledge about the sub-processes is available and X-matrix can be divided into blocks between process variables and NIR spectra.

No abstract

In this work, we investigated the possibility to perform wavelength selection by exploiting the metric structure of the spectrophotoscopic measurements. The topologically preserving representation of the data is performed using the self-organizing map (SOM) where the inputs' significance to the output is computed with the measure of topological relevance (MTR) on SOM. The MTR on SOM is a metric measuring the similarity between local distance matrices and we found that spectral inputs with a topology, which is, close to the output's are also associated to the wavelengths that chemically explain the influence of the spectra to the property of interest. As a result, we suggest a wavelength selection strategy based on the MTR on SOM, that is, interpretable to the domain experts and independent on the regression technique subsequently used for estimation. To support the presentation, a full-scale application from the oil refining industry is illustrated on the problem of estimating standard properties in a complex hydrocarbon product starting from spectrophotoscopic measurements. The method is further validated on the problem of octane number estimation in finished gasolines, under small sample conditions. The application led to accurate, parsimonious and understandable models. Copyright (C) 2008 John Wiley & Sons, Ltd

In chemometric studies all predictor variables are usually collected in one data matrix X. This matrix is then analyzed by PLS regression or other methods. When data from several different sub-processes are collected in one matrix, there is a possibility that the effects of some sub-processes may vanish. If there is, for instance, mechanic data from one process and spectral data from another, the influence of the mechanic sub-process may not be detected. An application of multi-block (MB) methods, where the X-data are divided into several data blocks is presented in this study. By using MB methods the effect of a sub-process can be seen and an example with two blocks, near infra-red, NIR, and process data, is shown. The results show improvements in modelling task, when a MB-based approach is used. This way of working with data gives more information on the process than if all data are in one X-matrix. The procedure is demonstrated by an industrial continuous process, where knowledge about the sub-processes is available and X-matrix can be divided into blocks between process variables and NIR spectra.

One option of recycling used contaminated packaging is to recover its high energy content. This can be performed in a normal multi-fuel power plant by co-combustion of packaging-derived fuel (PDF) or refuse-derived fuel (RDF) with fossil fuels, such as coal or peat. This work includes the results of 17 co-combustion tests and an evaluation of the results by the Principal Component Analysis (PCA) and the Partial Least Squares Projections to Latent Structures (PLS).PCA and PLS calculations showed that especially Pb, but also Cr, and Cu correlated with lower chlorinated furans (PCDFs) in the fly ash. Correlation between Sn and lower chlorinated dioxins (PCDDs) in the fly ash was also noticed. CO and PAH emission in the flue gas correlated with total PCDD/Fs in the flue gas. In a real full-scale combustion process, a single parameter in fuel, flue gas or a combustion parameter did not provide a guide to PCDD/F formation or to a level of the total PCDD/F emission, but correlations between different parameters and PCDD/Fs could be found. Although PDFs and RDF had catalytic heavy metals and chlorine, the co-combustion results showed that they can be co-combusted with peat and coal in a fluidized-bed boiler at least up to 26 % with very low total PCDD and PCDF emissions.