The effect of micronization (at 130 and 150 °C) as a potential heat treatment to reduce ‘beany’ aroma and flavor of cooked chickpea (Cicer arietinum) and green lentil (Lens culinaris) flours was investigated. A simultaneous distillation solvent extraction method was developed to extract key volatile compounds with potential contribution to ‘beany’ aroma and flavor notes in micronized pulse flours and analyzed by gas chromatography-mass spectrometry. Concentrations of volatile compounds such as pentanol, hexanal, 2-hexenal, hexanol, heptanal, furan-2-pentyl, 2-octenal, nonanal, 2,4 decadienal, and 2,4- undecadienal were significantly (P<0.05) decreased with micronization. Low fat burgers fortified with 6% micronized chickpea and green lentil flours showed significantly higher acceptability for aroma, flavor, texture, color and overall acceptability (p<0.05) compared to non-micronized samples in a consumer acceptability test with 101 consumers. In addition, fatty acid analysis of burgers showed burgers containing micronized pulses had higher level of linoleic and linolenic acid content.

A systematic methodology for quantitative structure-activity relationship (QSAR) development in environmental toxicology is provided. The methodology is summarized in a strategy with six sequential steps. The strategy rests on two cornerstones, namely (1) the use of statistical design to select a series of representative compounds (the so-called training set) on which to base a QSAR, and (2) the multivariate modelling of the relationship between physicochemical and biological properties of the training set compounds. The first step of the strategy is the division of chemicals into classes of structurally similar compounds. Briefly, steps 2 to 6 are: (2) physico-chemical and structural characterization of the compounds in a class, (3) selection of a training set of representative compounds, (4) biological testing of the selected training set, (5) QSAR model development, and (6) experimental validation of the QSAR and predictions for non-tested compounds. The thesis summarizes the results obtained from the application of the strategy to the class of halogenated aliphatic compounds. Biological measurements were made in four biological test systems, reflecting acute toxicity, mutagenicity, relative cytotoxicity and genotoxicity. QSARs were developed relating each biological endpoint to the structural descriptors of the compounds. Multivariate PLS modelling was used in the data analysis. The developed QSARs were used for predicting the biological activity pattern of the non-tested compounds in the class. These predictions may be used as a starting point for a priority ranking for further biological testing of these compounds. The strategy has not been developed solely for establishing QSARs for the halogenated aliphatics class. On the contrary, this work is intended to demonstrate a generally applicable QSAR methodology. / <p>Diss. (sammanfattning) Umeå : Umeå universitet, 1991</p> / digitalisering@umu

The effect of micronization (at 130 and 150 °C) as a potential heat treatment to reduce ‘beany’ aroma and flavor of cooked chickpea (Cicer arietinum) and green lentil (Lens culinaris) flours was investigated. A simultaneous distillation solvent extraction method was developed to extract key volatile compounds with potential contribution to ‘beany’ aroma and flavor notes in micronized pulse flours and analyzed by gas chromatography-mass spectrometry. Concentrations of volatile compounds such as pentanol, hexanal, 2-hexenal, hexanol, heptanal, furan-2-pentyl, 2-octenal, nonanal, 2,4 decadienal, and 2,4- undecadienal were significantly (P<0.05) decreased with micronization. Low fat burgers fortified with 6% micronized chickpea and green lentil flours showed significantly higher acceptability for aroma, flavor, texture, color and overall acceptability (p<0.05) compared to non-micronized samples in a consumer acceptability test with 101 consumers. In addition, fatty acid analysis of burgers showed burgers containing micronized pulses had higher level of linoleic and linolenic acid content.

The desire to use the tools and concepts of chemometrics when studying problems in the life sciences, especially biology and medicine, has prompted chemometricians to shift their focus away from their field‘s traditional emphasis on model predictivity and towards the more contemporary objective of optimizing information exchange via model interpretation. The complex data structures that are captured by modern advanced analytical instruments open up new possibilities for extracting information from complex data sets. This in turn imposes higher demands on the quality of data and the modeling techniques used. The introduction of the concept of orthogonal variation in the late 1990‘s led to a shift of focus within chemometrics; the information gained from analysis of orthogonal structures complements that obtained from the predictive structures that were the discipline‘s previous focus. OPLS, which was introduced in the beginning of 2000‘s, refined this view by formalizing the model structure and the separation of orthogonal variations. Orthogonal variation stems from experimental/analytical issues such as time trends, process drift, storage, sample handling, and instrumental differences, or from inherent properties of the sample such as age, gender, genetics, and environmental influence. The usefulness and versatility of OPLS has been demonstrated in over 500 citations, mainly in the fields of metabolomics and transcriptomics but also in NIR, UV and FTIR spectroscopy. In all cases, the predictive precision of OPLS is identical to that of PLS, but OPLS is superior when it comes to the interpretation of both predictive and orthogonal variation. Thus, OPLS models the same data structures but provides increased scope for interpretation, making it more suitable for contemporary applications in the life sciences. This thesis discusses four different research projects, including analyses of NIR, FTIR and NMR spectroscopic data. The discussion includes comparisons of OPLS and PLS models of complex datasets in which experimental variation conceals and confounds relevant information. The PLS and OPLS methods are discussed in detail. In addition, the thesis describes new OPLS-based methods developed to accommodate hyperspectral images for supervised modeling. Proper handling of orthogonal structures revealed the weaknesses in the analytical chains examined. In all of the studies described, the orthogonal structures were used to validate the quality of the generated models as well as gaining new knowledge. These aspects are crucial in order to enhance the information exchange from both past and future studies.

In past decades, the expectation of synergy has fueled many thousands of mergers and acquisitions. Meanwhile, economists and analysts have reported a large proportion of merger failures. This apparent contradiction has provided researchers with a rich source of studies. One of the likely causes of a merger failure is the “people factor”. Revolving around the axis of mergers and acquisitions, the peoples affected are, on the one side, the stockholders, top management, and economists who “talk the project” and tend to have a positive attitude and on the other side, the people who “walk the project” – the employees - who have a more hesitant attitude. / This empirical study adopted the construct of Psychological Contracts to measure the expectations of employees who are influenced by mergers and acquisitions. Based on this construct, a model was developed to study employees’ behaviour after a merger, examining it from a multitude of dimensions. Using the PLS-Graph analysis tools, the model was tested with the aim of assessing the factors’ impact on employees’ behaviour. Apart from the direct causal relationship between two variables, the indirect effects caused by other variables are assessed as well. / The first contribution made by this research is the fact that it examines the relevance of a psychological contract in a non-Western geographical region. Next, the study clearly confirms some of the existing conceptualizations regarding psychological contracts and reveals some additional insights, particularly in relation to the consideration of psychological contracts in a non-Western socio-cultural context. / The research aspires to generalize the model for predicting the post-merger behaviour of employees anywhere, across any industry, business segment and profession.

In this study, groundwater resource potential (GRP) for Västervik municipality has been investigated using the methodology developed in the article by Earon et al. (2015). The aim was to test the reliability of the methodology for groundwater mapping, as to further add to the knowledge base of groundwater access. The GRP-methodology is a linear additive multicriteria analysis where geohydrological indicators are scored, classified into groups, then multiplied by weights calculated using Principal Component Analysis (PCA). The classification and validation were performed against specific capacity [L/(h*m)], which is a well's capacity calculated at drilling, per meter well. GRP was calculated in different sets based on 13 geohydrological variables such as altitude and Topographic Wetness Index (TWI). The results included correlation tests for Kendall's tau (0.06-0.13), Spearmans rho (0.09-0.19) with a total accuracy of 52-55%. Positive but low values ​​for Cohen´s kappa indicated that all calculations performed better than a random generator, but not by margin. Calculations of VIP (Variables importance on PLS projection), based on Partial Least Squares (PLS), identified Altitude, Earth type, Drainage density and TWI as the most influential indicators for the analysis.The conclusions of this study were, among other things, that the GRP methodology had low predictivity due to the weak relationships between the indicators and the specific capacity. The weaknesses could also be due to the fact that specific capacity has weaknesses as a validation variable for groundwater resource potential linked to uncertainties of the capacity measured at wellbore. The study shows that further development of the weighting scheme by integrating PLS would be beneficial, as PLS calculates the variance of the indicators based on the specific capacity, instead of assuming it as a PCA. / I denna studie har grundvattenresurspotential (GRP) för Västerviks kommun undersökts med hjälp av metodik från artikeln av Earon et al.(2015). Metodiken användes för att ta fram översiktligt underlag till grundvattenresursplanering på ett billigt och effektivt sätt och är därför av intresse för beslutsfattare och konsulter. Exempel för dessa användningsområden är bland annat vid grundvattenresursplanering och vid brunnsborrning. Syftet var att testa metodikens pålitlighet för grundvattenkartläggning. Endast öppen data användes i undersökningen vilket krävde små insatser jämfört med traditionell kartering. Stora möjligheter finns därmed vid användande av metodiken för grundvattenresursplanering och brunnsborrning.  Metodiken är baserad på en linjärt additiv multikriterieanalys där indikatorer såsom altitud, topografiskt fuktighetsindex (TWI) och bergarter poängsätts, klassificeras i grupper, för att sedan multipliceras med vikter beräknade med hjälp av principalkomponent analys (PCA). Klassificeringen och valideringen utfördes gentemot specifik kapacitet [L/(h*m)], vilket är en brunns kapacitet beräknad vid borrning, per meter brunnsdjup, hämtad från SGUs brunnsarkiv. Brunnarna delades slumpartat upp i 70 % kalibreringsbrunnar och 30 % valideringsbrunnar. GRP beräknades i olika varianter utifrån 13 variabler (altitud, lutning, relativ altitud, avstånd till sjöar/hav, avstånd till vattendrag, avstånd till lineament, avstånd till nodpunkter, bergarter, jordarter, dräneringstäthet, avstånd till geologiska övergångslinjer, jorddjup och TWI). Resultaten från korrelationstester för Kendalls tau var mellan 0,06-0,13 och för Spearmans rho mellan 0.09-0.19, samt en total noggrannhet på 52-55 %. Positiva men låga värden för Cohens kappa indikerade att alla beräkningar överträffade slumpen, men inte med marginal. Beräkningar av VIP (Variables importance on PLS projektion), utifrån PLS (Partial Least Squares) identifierade altitud, jordart, dräneringsdensitet och TWI som de mest inflytelserika indikatorerna för analysen.  Slutsatserna i denna studie var bland annat att GRP-metodiken hade låg prediktionsförmåga som sannolikt berodde på de svaga sambanden mellan indikatorerna och den specifika kapaciteten. De svaga sambanden kunde också bero på att specifik kapacitet har svagheter som valideringsvariabel för grundvattenresurspotential, kopplat till hur kapaciteten mätts vid brunnsborrning. Studien visar att en vidareutveckling av viktningsmetodiken genom att integrera PLS skulle vara fördelaktigt, eftersom PLS beräknar indikatorernas varians utifrån den specifika kapaciteten, istället för att anta sambandet som PCA.

Submitted by Uillis de Assis Santos (uillis.assis@ufba.br) on 2018-01-26T13:26:54Z No. of bitstreams: 1 AnaRosaCaribédeGóesMassa_Dissertação-UFBA-Politécnica-EngenhariaIndustrial-2017.pdf: 4582890 bytes, checksum: 3c5c6099462e9b7f61ff397081c5dd6a (MD5) / Approved for entry into archive by Vanessa Reis (vanessa.jamile@ufba.br) on 2018-01-26T16:03:46Z (GMT) No. of bitstreams: 1 AnaRosaCaribédeGóesMassa_Dissertação-UFBA-Politécnica-EngenhariaIndustrial-2017.pdf: 4582890 bytes, checksum: 3c5c6099462e9b7f61ff397081c5dd6a (MD5) / Made available in DSpace on 2018-01-26T16:03:46Z (GMT). No. of bitstreams: 1 AnaRosaCaribédeGóesMassa_Dissertação-UFBA-Politécnica-EngenhariaIndustrial-2017.pdf: 4582890 bytes, checksum: 3c5c6099462e9b7f61ff397081c5dd6a (MD5) / Analisadores em linha fornecem uma resposta rápida de composição em comparação às análises laboratoriais. Porém, esses estão sujeitos a frequentes interferências e contaminações devido às substâncias utilizadas nas correntes reais, que agridem, contaminam e comprometem o funcionamento do equipamento, exigindo manutenções rotineiras. Durante estas, há perda de informações vitais que podem culminar numa parada da produção, a não ser que outro equipamento permita estimar tais informações de maneira confiável. Desta forma, o presente trabalho tem por objetivo desenvolver um analisador virtual para estimar a concentração dos contaminantes metilacetileno e propadieno (MAPD) em um reator trickle bed em uma indústria de propileno no Brasil. A partir da coleta de dados de processos de uma campanha catalítica, coletados por cromatógrafos a gás, termopares e medidores de vazão, foram desenvolvidos modelos de calibração multivariada utilizando a técnica de Regressão por Mínimos Quadrados Parciais (PLS), para dois leitos catalíticos, A e B. Dados de treze variáveis de processo, monitoradas a cada 10 minutos durante uma campanha catalítica para cada leito, foram utilizados. Os modelos PLS foram desenvolvidos e validados e foram capazes de fornecer valores preditos confiáveis, com R2 de 0,84 para o leito A e 0,92 para o leito B. Normalidade e homocedacidade dos resíduos foram observadas em ambos modelos. Além disso, foi realizada uma seleção de variáveis utilizando o gráfico de escores VIP (Variable Importance in Projection) obtido durante o desenvolvimento dos modelos PLS. As variáveis mais importantes foram selecionadas e os modelos PLS construídos apenas com essas variáveis mantiveram a capacidade de predição em ambos os leitos, com valores de R2 de 0,82 para o leito A e 0,87 para o leito B. Normalidade e homocedacidade dos resíduos foram mantidas, e um teste F não gerou evidência que indicasse diferença significativa entre os modelos desenvolvidos antes e após a seleção de variáveis. Dessa forma, os modelos PLSVIP fornecem uma estimativa confiável do teor de MAPD no reator trickle bed na planta de propeno estudada. Esses resultados mostraram que os métodos desenvolvidos possuem um alto potencial de aplicação nos reatores estudados caso haja necessidade, evitando assim uma parada da planta e subsequente perda de capital investido. / Online analysers grant a faster answer on the composition of products when compared with laboratory analysis. However, the former is often affected by substances in the stream line which harm and compromise its normal working state, calling for frequent maintenance. During those, the loss of vital information could lead to a halt in production, unless another device allows for such information to be carefully estimated.As such, this paper aims at developing a Virtual Analyser that can estimate the concentration of methylacetylene and propadiene (MAPD) contaminants in a trickle bed reactor at a propene industrial plant in Brazil. Process data collected in the reactor by gas chromatographers,temperature probes and flowmeters were employed to build multivariate calibration models by using the Partial Least Square Regression(PLS)technique, for two bed reactors, A and B.Data from thirteen process variables, monitored every 10 minutes during one catalytic campaign for each bed, of about three months each, were used. The developed PLS models for both beds have shown a great prediction capacity and remarkable performances, with R2 of 0.84 for bed A and 0.92 for bed B. Residual normality and homoscedasticity were observed for both models. In addition, a variable selection approach was carried out using the VIP (Variable Importance in Projection) score plot obtained during the developed PLS models. The most important variables (process variables)were selected and the PLS models built with only these variables were still able to keep are markable prediction ability for both beds, with a R2 of 0.82 for bed A and 0.87 for bed B. Residual normality and homoscedasticity were kept, and an F test did not provide evidence for significant difference between the models developed before and after the variable selection. Therefore, the PLS-VIP models provided a reliable estimate of the MAPD content in the trickle bed reactors at the studied propene plant. These outcomes showed that the developed methods present a high potential for application in the studied reactors, if necessary, in order to prevent a halt in production and its subsequent loss of invested capital.

134 f. / Submitted by Ana Hilda Fonseca (anahilda@ufba.br) on 2013-09-19T14:49:05Z No. of bitstreams: 1 Tese_Vanjoaldo Lopes.pdf: 3691974 bytes, checksum: 4acbb7d5e3b7ac76fe0b2ff6e4a01cc2 (MD5) / Approved for entry into archive by Ana Hilda Fonseca(anahilda@ufba.br) on 2013-09-19T15:13:57Z (GMT) No. of bitstreams: 1 Tese_Vanjoaldo Lopes.pdf: 3691974 bytes, checksum: 4acbb7d5e3b7ac76fe0b2ff6e4a01cc2 (MD5) / Made available in DSpace on 2013-09-19T15:13:57Z (GMT). No. of bitstreams: 1 Tese_Vanjoaldo Lopes.pdf: 3691974 bytes, checksum: 4acbb7d5e3b7ac76fe0b2ff6e4a01cc2 (MD5) Previous issue date: 2012 / CAPES, FAPESB, CNPQ / Buscando minimizar a dependência dos combustíveis fosseis, fontes energéticas alternativas e renováveis estão em franco desenvolvimento, ganhando cada vez mais espaço na matriz energética mundial e levando à diminuição dos impactos ambientais. Neste contexto, muitas pesquisas são realizadas, buscando o aumento da produção de biodiesel e a garantia da sua qualidade. Em um país como Brasil, que possui vasto território, onde muitas oleaginosas e gorduras animais podem ser usadas para produzir biodiesel, investigar a origem do biodiesel é uma necessidade, uma vez que a qualidade do biodiesel está relacionada às matérias-primas que o produziu. Neste trabalho, determinou-se a composição de amostras de biodiesel em relação aos tipos de matéria-prima através da cromatografia a gás associada à calibração multivariada. Para tal, foram desenvolvidas três modelagens por PLS, a partir de cromatogramas de amostras de biodiesel obtidos empregando a norma EN 14103, onde foi possível identificar, com segurança estatística, a composição das matérias-primas de amostras de B100. As duas primeiras modelagens foram realizadas com as áreas dos picos cromatográficos e com as áreas normalizadas, respectivamente; nestas duas modelagens, os picos foram integrados manualmente. Na terceira modelagem, os cromatogramas foram previamente alinhados usando o algoritmo COW, e todo o cromatograma foi explorado pelo PLS. Os procedimentos foram aplicados a misturas de biodiesel quaternárias (soja, sebo bovino, girassol e dendê), sendo alcançados parâmetros similares para as duas primeiras modelagens na determinação da composição do biodiesel quanto as suas matérias-primas. Na modelagem das áreas não normalizadas, nos conjuntos de calibração encontraram-se valores de RMSEC < 4,40%; RMSECV < 5,36%; R2 (calibração) > 0,973 e R2 (Validação Cruzada) > 0,962; o conjunto de previsão apresentou correlação > 0,992 e RMSEP < 3,57%. Para a modelagem das áreas normalizadas, encontraram-se valores: RMSEC < 4,70%; RMSECV < 5,34%; R2 (calibração) > 0,985; R2 (Validação Cruzada) > 0,963; correlação > 0,989 e RMSEP < 3,43%. A terceira modelagem, com os cromatogramas totais alinhados, encontraram-se valores: RMSEC < 5,35%; RMSECV < 9,14%; R2 (calibração) > 0,955 e R2 (Validação Cruzada) > 0,875; Correlação > 0,966 e RMSEP < 7,19%. As duas primeiras modelagens estabeleceram parâmetros análogos, exatos e seguros para avaliar o teor de cada tipo de biodiesel na composição da mistura B100. A terceira modelagem apresentou parâmetros inferiores aos das duas primeiras modelagens; este fato pode ser explicado porque, nesta modelagem, a altura (sinal) foi a variável avaliada, e por sua vez a altura é um dado cromatográfico de menor confiabilidade que a área dos picos. / Salvador

DANTAS, Allan Nilson de Sousa. Estudo de novos parâmetros para aprimoramento do controle de qualidade da cera de carnaúba. 2014. 113 f. Tese (Doutorado em química)- Universidade Federal do Ceará, Fortaleza-CE, 2014. / Submitted by Elineudson Ribeiro (elineudsonr@gmail.com) on 2016-05-31T18:15:29Z No. of bitstreams: 1 2014_tese_ansdantas.pdf: 5176120 bytes, checksum: e189ee73179294e707b53c7a859776d5 (MD5) / Approved for entry into archive by José Jairo Viana de Sousa (jairo@ufc.br) on 2016-05-31T23:21:09Z (GMT) No. of bitstreams: 1 2014_tese_ansdantas.pdf: 5176120 bytes, checksum: e189ee73179294e707b53c7a859776d5 (MD5) / Made available in DSpace on 2016-05-31T23:21:09Z (GMT). No. of bitstreams: 1 2014_tese_ansdantas.pdf: 5176120 bytes, checksum: e189ee73179294e707b53c7a859776d5 (MD5) Previous issue date: 2014 / Carnauba wax is an important economic product and its industrial production has over the years suffering from the scarcity of scientific studies to aid in the understanding of the chemical differences between different types of wax. The aim of this work was characterize the different types of carnauba waxes (Type 1, Type 3 and Type 4) in relation to the chemical composition. The organic matrix of the samples were investigated by Infrared Spectroscopy (MID and NIR), Thermal Analysis and Gas Chromatography (CGMS). The results showed that the three types of waxes are very similar in its organic composition, presenting in its composition mainly paraffin, alcohols and esters of long chain fatty acids. To evaluate inorganic content, an optimization of the conditions of the sample preparation by microwave assisted wet digestion implemented. The digests showed residual carbon content less than 5%. The analysis was performed by ICP-OES. The elements data set was processed using the software The Unscrambler 7.1 and X10.3. The mean values for the elements in the samples investigated wax Type 1 levels were: K (22.3 mg kg -1), Al (20.5 mg kg-1), Fe (11.5 mg kg-1), Mg (4.0 mg kg-1), Na (3.5 mg kg-1), Pb (2.2 mg kg-1), Ca (1.2 mg kg-1), Cu (0.5 mg kg-1), Mn (0.4 mg kg-1) , Zn (0.2 mg kg-1) and Co (0.2 mg kg-1). The levels observed for Ni (67.2 mg kg-1) were higher in samples from three industries, raising the average value of the element. For samples carnauba wax Type 3, the mean concentrations of elements were: Al(122.0 mg kg-1), K(112.1 mg kg-1), Ca (103.9 mg kg- 1), Fe(93.4 mg kg-1), Mg(67.1 mg kg-1), Na(39.6 mg kg-1), Mn(9.6 mg kg-1), Pb(1 7 mg kg-1), Zn (0.6 mg kg-1), Ni (0.3 mg kg-1), Cu (0.2 mg kg-1) and Co (0.1 mg kg-1). Finally, the samples with wax to Type 4, the mean concentration values were observed : K (185.3 mg kg-1), Ca (116.8 mg kg-1), Al (97.6 mg kg-1), Fe (84.5 mg kg-1) mg (76.9 mg kg-1), Na (37.6 mg kg-1), Mn (8.7 mg kg-1), Pb (2.3 mg kg-1), Zn (0.7 mg kg -1), Cu (0.2 mg kg-1) and Co (0.1 mg kg-1). Principal Component Analysis showed that the three types of waxes could be distinguished based on the levels of inorganic elements. In this view point waxes Type 4 showed the highest inorganic levels than the waxes Type 3 and Type 1. New wax samples were acquired with a view to developing a model of multivariate calibration (NIR / PLS) for prediction of Al, Ca, Cu, Fe, K, Mg, Mn and Na in wax samples. The spectra were preprocessed applying first derivative and then the models PLS-1 were obtained. The spectra data set were divided into two groups: one for the calibration and the other for validation. The models obtained were promising for predicting the content of elements such as Al, Fe and Cu in samples of wax. Finally, the results obtained in this work can be used as parameter for sample classification and as well as used for the development of standards rules for quality assurance of the carnauba waxes produced by refining industrial processes. / A cera de carnaúba consiste em um importante produto econômico e sua produção vem ao longo dos anos sofrendo com a escassez de trabalhos científicos que auxiliem no entendimento das diferenças químicas existentes entre os tipos de cera. Deste modo, o presente trabalho teve por objetivo propor uma caracterização dos diferentes tipos de cera de carnaúba (Tipo 1, Tipo 3 e Tipo 4) quanto à sua composição química, em especial em relação aos constituintes inorgânicos. As amostras de cera foram avaliadas quanto à composição orgânica por FT-IR, TG e CG-MS. Os resultados obtidos mostraram que as ceras dos Tipos 1, 3 e 4 possuem pequenas diferenças são quanto a sua composição orgânica fator que pode ser responsável pela diferença de coloração entre os materiais, apresentando em sua composição basicamente parafinas, álcoois de cadeia longa e ésteres de ácidos graxos. Para avaliação dos inorgânicos, uma etapa de otimização das condições de preparo de amostra por via úmida assistida por micro-ondas foi implementada. Os digeridos apresentaram teores de carbono residual inferior a 5%, sendo feita a quantificação dos elementos por ICP OES. Os teores médios para os elementos investigados nas amostras de cera do Tipo 1 foram: K (22,3 mg Kg-1), Al (20,5 mg Kg-1), Fe (11,5 mg Kg-1), Mg (4,0 mg Kg-1), Na (3,5 mg Kg-1), Pb (2,2 mg Kg-1), Ca (1,2 mg Kg-1 ), Cu (0,5 mg Kg-1), Mn (0,4 mg Kg-1), Zn (0,2 mg Kg-1) e Co (0,2 mg Kg-1). Os teores observados para Ni (67,2 mg Kg-1 ) se mostraram elevados nas amostras de três industrias, elevando o valor médio do elemento. Para as amostras de cera de carnaúba do Tipo 3, os valores médios das concentrações dos elementos foram: Al (122,0 mg Kg-1), K (112,1 mg Kg-1), Ca (103,9 mg Kg-1), Fe (93,4 mg Kg-1), Mg (67,1 mg Kg-1), Na (39,6 mg Kg-1), Mn (9,6 mg Kg-1 ), Pb (1,7 mg Kg-1), Zn (0,6 mg Kg-1), Ni (0,3 mg Kg-1), Cu (0,2 mg Kg-1) e Co (0,1 mg Kg-1). Por fim, para as amostras de cera do Tipo 4, os valores médios de concentração observados foram: K (185,3 mg Kg-1), Ca (116,8 mg Kg-1), Al, (97,6 mg Kg-1), Fe (84,5 mg Kg-1 ), Mg (76,9 mg Kg-1), Na (37,6 mg Kg-1), Mn (8,7 mg Kg-1), Pb (2,3 mg Kg-1), Zn (0,7 mg Kg-1), Cu (0,2 mg Kg-1) e Co (0,1 mg Kg-1 ). Os dados foram tratados utilizando o software de Quimiometria The Unscrambler 7.1 e X10.3. Análise de Componentes Principais mostrou que as ceras podem ser classificadas e distinguidas em função dos teores das espécies avaliadas. Nesse contexto as ceras Tipo 4 apresentaram os maiores teores de inorgânicos, seguido das ceras do Tipo 3 e por fim das ceras do Tipo 1. Novas amostras de cera foram adquiridas com o intuito de desenvolver um modelo de calibração multivariada (NIR/PLS) para previsão de Al, Ca, Cu, Fe, K, Mg, Mn e Na nas amostras de cera. Os espectros foram pré-processados aplicando 1º derivada ao conjunto de dados espectrais e em seguida foram obtidos os modelos utilizando PLS-1. As amostras de cera foram divididas em dois conjuntos de dados: um para calibração e o outro para validação. Os modelos obtidos se mostraram promissores para a previsão do teor de elementos como Al, Fe e Cu nas amostras de cera de carnaúba. Por fim, os resultados obtidos neste trabalho podem ser utilizados como parâmetros de classificação das amostras de cera, bem como podem ser utilizados como aporte para a elaboração de normas para classificação dos teores de qualidade das ceras de carnaúba produzidas pelas indústrias refinadoras.

A presente dissertação propõe métodos para seleção de variáveis preditivas com base em índices de importância das variáveis e regressão PLS (Partial Least Squares). Partindo-se de uma revisão da bibliografia sobre PLS e índices de importância das variáveis, sugere-se um método, denominado Eliminação Backward (EB), para seleção de variáveis a partir da eliminação sistemática de variáveis de acordo com a ordem definida por índices de importância das variáveis. Um novo índice de importância de variáveis, proposto com base nos parâmetros da regressão PLS, tem seu desempenho avaliado frente a outros índices reportados pela literatura. Duas variações do método EB são propostas e testadas através de simulação: (i) o método EBM (Eliminação backward por mínimos), que identifica o conjunto que maximiza o indicador de acurácia preditiva sem considerar o percentual de variáveis retidas, e (ii) o método EBDE (Eliminação backward por distância euclidiana), que seleciona o conjunto de variáveis responsável pela mínima distância euclidiana entre os pontos do perfil gerado pela eliminação das variáveis e um ponto ideal hipotético definido pelo usuário. A aplicação dos três métodos em quatro bancos de dados reais aponta o EBDE como recomendável, visto que retém, em média, apenas 13% das variáveis originais e eleva a acurácia de predição em 32% em relação à utilização de todas as variáveis. / This dissertation presents new methods for predictive variable selection based on variable importance indices and PLS regression. The novel method, namely Backward Elimination (BE), selects the most important variables by eliminating process variables according to their importance described by the variable importance indices. A new variable importance index is proposed, and compared to previous indices for that purpose. We then offer two modifications on the BE method: (i) the EBM method, which selects the subset of variables yielding the maximum predictive accuracy (i.e., the minimum residual index), and (ii) the EBDE, which selects the subset leading to the minimum Euclidian distance between the points generated by variable removal and a hypothetical ideal point defined by the user. When applied to four manufacturing data sets, the recommended method, EBDE, retains average 13% of the original variables and increases the prediction accuracy in average 32% compared to using all the variables.