Predicting rock permeability from well logs in uncored wells is an important task in reservoir characterization. Due to the high costs of coring and laboratory analysis, typically cores are acquired in only a few wells. Since most wells are logged, the common practice is to estimate permeability from logs using correlation equations developed from limited core data. Most commonly, permeability is estimated from various well logs using statistical regression. For sandstones, often the logs of permeability can be correlated with porosity, but in carbonates the porosity permeability relationship tends to be much more complex and erratic. For this reason permeability prediction is a critical aspect of reservoir characterization in complex reservoirs such as carbonate reservoirs. In order to improve the permeability estimation in these reservoirs, several statistical regression techniques have already been tested in previous work to correlate permeability with different well logs. It has been shown that statistical regression for data correlation is quite promising in predicting complex reservoirs. But using all the possible well logs to predict permeability is not appropriate because the possibility of spurious correlation increases if you use more well logs. In statistics, variable selection is used to remove unnecessary independent variables and give a better prediction. So we apply variable selection to the permeability prediction procedures in order to further improve permeability estimation. We present three approaches to further improve reservoir permeability prediction based on well logs via data correlation and variable selection in this research. The first is a combination of stepwise algorithm with ACE technique. The second approach is the application of tree regression and cross-validation. The third is multivariate adaptive regression splines. Three methods are tested and compared at two complex carbonate reservoirs in west Texas: Salt Creek Field Unit (SCFU) and North Robertson Unit (NRU). The result of SCFU shows that permeability prediction is improved by applying variable selection to non-parametric regression ACE while tree regression is unable to predict permeability because it can not preserve the continuity of permeability. In NRU, none of these three methods can predict permeability accurately. This is due to the high complexity of NRU reservoir and measurement accuracy. In this reservoir, high permeability is discrete from low permeability, which makes prediction even more difficult. Permeability predictions based on well logs in complex carbonate reservoirs can be further improved by selecting appropriate well logs for data correlation. In comparing the relative predictive performance of the three regression methods, the stepwise with ACE method appears to outperform the other two methods.

Submitted by Marcos Anselmo (marcos.anselmo@unipampa.edu.br) on 2016-09-21T20:29:32Z No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Natana Becker.pdf: 1266893 bytes, checksum: 018f6e1bf563008837c534403e5c801e (MD5) / Approved for entry into archive by Marcos Anselmo (marcos.anselmo@unipampa.edu.br) on 2016-09-21T20:29:52Z (GMT) No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Natana Becker.pdf: 1266893 bytes, checksum: 018f6e1bf563008837c534403e5c801e (MD5) / Made available in DSpace on 2016-09-21T20:29:52Z (GMT). No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Natana Becker.pdf: 1266893 bytes, checksum: 018f6e1bf563008837c534403e5c801e (MD5) Previous issue date: 2015-08-21 / Os fármacos valsartana (VAL), hidroclorotiazida (HCT) e besilato de anlodipino (ANL) são utilizados em associação e comercializados no Brasil como agentes anti-hipertensivos. Geralmente a determinação simultânea destes fármacos é realizada por cromatografia líquida de alta eficiência (CLAE). Este trabalho teve por objetivo a determinação simultânea de VAL, HCT e ANL em uma formulação comercial de comprimidos através da técnica de espectroscopia no infravermelho próximo com transformada de Fourier e acessório de esfera de integração (FT-NIR) associadas a métodos de análise multivariada. Os modelos de calibração foram construídos utilizando mínimos quadrados parciais (PLS) e seleção de variáveis através dos algoritmos mínimos quadrados parciais por intervalo (iPLS) e mínimos quadrados parciais por sinergismo de intervalos (siPLS). Um total de 36 amostras sintéticas e 1 amostra real (26 amostras para o conjunto de calibração e 11 amostras para o conjunto de previsão), foram utilizadas as faixas de concentração de 261,9-500,0 mg g-1 para VAL; 20,2-83,3 mg g-1 para HCT e 11,6-49,6 mg g-1 para ANL. Os dados espectrais foram adquiridos na faixa de 4000 a 10000 cm-1 com resolução de 4 cm-1 por FT-NIR. Os melhores modelos foram obtidos através da utilização do pré-processamento centrado na média (CM) e do tratamento de correção do espalhamento de luz (MSC). O erro relativo de previsão (RSEP%) de 1,27% para VAL, 1,92% para HCT e 5,19%para ANL, foi obtido após seleção dos melhores intervalos por siPLS para dados obtidos por FT-NIR. Não foi encontrada diferença significativa (teste t-pareado, 95% de confiança) entre os valores do método de referência e do método proposto. Os resultados mostraram que modelos de regressão PLS (associados a métodos de seleção de variáveis, como iPLS e siPLS) combinados com FT-NIR são promissores no desenvolvimento de metodologias mais simples, rápidas e não destrutivas. Estes modelos permitem a determinação simultânea de VAL, HCT e ANL na formulação farmacêutica. / Valsartan (VAL), hydrochlorothiazide (HCT) and amlodipine besylate (ANL) drugs are used in combination and they are commercialized in Brazil as antihypertensive agents. Generally, the simultaneous determination of these drugs is carried out by high performance liquid chromatography (CLAE). This study aimed to the simultaneous determination of VAL, HCT, and ANL in a comercial tablet formulation through the technique near infrared spectroscopy with Fourier transform and integrating sphere accessory (FT- NIR) associated with methods of multivariate analysis. The calibration models were built using partial least squares (PLS) and variable selection through partial least squares algorithms for interval (iPLS) and partial least squares by synergism intervals (siPLS). A total of 36 synthetic samples 1 and commercial sample (26 samples for the calibration sample set and 11 for the prediction set), were used the concentration ranges of 261.9-500.0 mg g-1 for VAL; 20.2-83.3 mg g-1 for HCT and 11.6-49.6 mg g-1 for ANL. The spectral data were acquired in the range 4000-10000 cm-1 with resolution of 4 cm-1 by FT-NIR. Multiplicative scatter correction (MSC) and the data centered in the media (CM) produced the best models. A relative standard error of prediction (RSEP%) of 1.27% for VAL, 1.92% for HCT and 5.19% for ANL was obtained after selection of the best intervals for data obtained by siPLS FT-NIR. There was no significant difference (paired t-test, 95% confidence) between the values of the reference method and the proposed method. Results showed that PLS models regression (associated with iPLS and siPLS regression models) combined with FT-NIR are promising in the development of simpler methods, rapid and non-destructive. These models allow simultaneous determination of VAL, HCT, and ANL in the pharmaceutical formulation.