Loading...
Publication
Identificação de veículos a partir de tintas automóveis transferidas em acidentes de viação
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 5.11 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A análise de tintas automóveis pode tornar-se complexa, devido às composições físicas e químicas destas, bem como à pequena quantidade disponível. Para tal, recorre-se a técnicas precisas para a análise deste tipo de amostras, onde seja possível a análise de pequenas porções, como o FTIR-ATR.
O presente trabalho tem como objetivo avaliar o potencial discriminatório do FTIR-ATR na análise de tintas trituradas como um todo, ou seja, sem qualquer distinção de camadas, desenvolvendo, assim uma metodologia de identificação de veículos envolvidos em acidentes de viação através da análise de amostras resultantes de transferências de tinta.
No decorrer do trabalho, foram analisados dois tipos de amostras: amostras de referência e amostras de casos reais. As primeiras foram trituradas e efetuaram-se misturas entre pares de tinta em diferentes rácios para simular as transferências de tinta. As amostras de casos reais, foram obtidas nos locais onde ocorreram os acidentes de viação com o auxílio do Destacamento Territorial de Almada da GNR, sendo igualmente trituradas. Todas as amostras foram analisadas seguindo a mesma metodologia.
Os resultados obtidos a partir das amostras de referência apresentam diferenças nos espetros entre as amostras puras que se propagam nos espetros dos rácios de cada mistura. O modelo preditivo desenvolvido para estas amostras indica uma boa capacidade identificativa, tendo sido obtidos valores de exatidão entre 84,44% e 97,78%. No entanto, este modelo tem uma exatidão de apenas 40% para as amostras puras (melhor do que a exatidão inicial de previsão, cerca de 10%).
Como não se obtiveram amostras de casos reais em sintonia com as amostras de referência, o modelo preditivo desenvolvido não foi aplicado aos casos reais.
Os resultados obtidos são promissores, revelando que a metodologia desenvolvida poderá permitir futuramente a identificação de automóveis a partir de amostras de tinta transferida recorrendo às técnicas descritas.
The analysis of automotive paints can become complex due to their physical and chemical compositions as well as the short amount available. Hence, precise techniques are used for the analysis of this type of samples, where it is possible to analyse of small portions, such as ATR-FTIR. The thesis aim is to evaluate the discriminatory potential of ATR-FTIR for the analysis of milled automotive paints as a whole, i.e. without any layers distinction, thus developing a methodology for identifying vehicles involved in road accidents through analysis of samples resulting of paint transfers. Two types of samples were analysed: reference samples and real case samples. The reference samples were milled and mixtures of paint pairs were mixed at different ratios to simulate the transfers of paint. The real case samples were obtained from places where traffic accidents occurred with the help of Destacamento Territorial de Almada da GNR. All samples were analysed using the same methodology. The results obtained from the reference samples show differences in spectra between pure samples that propagate in the spectra of each mixture for different racios. The predictive model developed for the reference samples indicates a good identification ability, with accuracy values ranging from 84,44% and 97,78%. However, this model for pure reference samples only indicates an accuracy of 40% (this value is considered positive since the initial accuracy was 10%). Since it wasn’t possible to obtain real case samples in line with the reference samples, the developed predictive model was only used in the reference samples. The results obtained are promising, revealing that the developed methodology could allow, in the future, the identification of vehicles from transferred paint samples using described techniques.
The analysis of automotive paints can become complex due to their physical and chemical compositions as well as the short amount available. Hence, precise techniques are used for the analysis of this type of samples, where it is possible to analyse of small portions, such as ATR-FTIR. The thesis aim is to evaluate the discriminatory potential of ATR-FTIR for the analysis of milled automotive paints as a whole, i.e. without any layers distinction, thus developing a methodology for identifying vehicles involved in road accidents through analysis of samples resulting of paint transfers. Two types of samples were analysed: reference samples and real case samples. The reference samples were milled and mixtures of paint pairs were mixed at different ratios to simulate the transfers of paint. The real case samples were obtained from places where traffic accidents occurred with the help of Destacamento Territorial de Almada da GNR. All samples were analysed using the same methodology. The results obtained from the reference samples show differences in spectra between pure samples that propagate in the spectra of each mixture for different racios. The predictive model developed for the reference samples indicates a good identification ability, with accuracy values ranging from 84,44% and 97,78%. However, this model for pure reference samples only indicates an accuracy of 40% (this value is considered positive since the initial accuracy was 10%). Since it wasn’t possible to obtain real case samples in line with the reference samples, the developed predictive model was only used in the reference samples. The results obtained are promising, revealing that the developed methodology could allow, in the future, the identification of vehicles from transferred paint samples using described techniques.
Description
Dissertação para obtenção do grau de Mestre no Instituto Universitário Egas Moniz
Keywords
Tintas automóveis FTIR-ATR Acidentes de viação Modelo preditivo