A study on the contribution of primer layer in car paint analysis using chemometrics approaches / Raja Zubaidah Raja Sabaradin

• Main Author: Raja Sabaradin, Raja Zubaidah
• Format: Thesis
• Language: English
• Published: 2020
• Subjects: Spectroscopy
• Online Access: https://ir.uitm.edu.my/id/eprint/61115/1/61115.pdf; https://ir.uitm.edu.my/id/eprint/61115/

Analysis of car paint is a common practice when involving car road accidents. A car paint system consists of at least 4 layers namely cathodic electro deposition, primer, basecoat and clearcoat. However, not all layers may be present in a car paint sample recovered from the crime scene. Fourier transform infrared spectroscopy (FTIR) has been commonly used for the analysis of car paint as it is a simple and fast analysis. Pyrolysis-gas chromatography-mass spectrometry (Py-GC-MS) is gaining popularity as this technique could analyze the chemical composition of a paint sample. This study is conducted in order to evaluate the contribution of car primer layer in the analysis of car paint sample. Fifty car primer samples of two types (1K and 2K) were applied to aluminum sheets. In addition, eight red basecoat were applied on another 10 selected car primers giving a total of eighty red car paint samples. Car primer samples were analyzed using attenuated total reflectance-Fourier transform infrared spectroscopy (ATR-FTIR) and Py-GC-MS. The pyrolysis temperature of Py-GC-MS method was optimized. The datasets obtained from the analysis were subjected to chemometric techniques. Principal component analysis (PCA) was able to discriminate 50 car primer samples according to their types (1K and 2K). Using discriminant analysis (DA) with the types of car primer (1K and 2K) was set as independent variables, Py-GC-MS datasets gave 100% correct classification for all three DA modes (standard, forward stepwise and backward stepwise) while ATR-FTIR data sets achieved 100% correct classification for standard and backward modes with 98% correct classification for forward mode. The 80 red car paint samples were also analyzed using ATR-FTIR and Py-GC-MS methods. PCA and cluster analysis showed that the analysis using ATR-FTIR grouped the samples based on the shades of red basecoat. Py-GC-MS analysis revealed that most of the samples were also grouped based on the shades of the red basecoat, but some samples were grouped based on the car primer varieties. This observation showed significant contribution of car primer in car paint analysis using Py-GC-MS. The study on the contribution of car primer was extended by exploiting the dataset from Py-GC-MS analysis using artificial neural network (ANN) prediction model to predict the presence of car primer in a car paint sample. The dataset was split into training (103 data), testing (26 data) and validation (15 data). The developed ANN prediction model gave good performance in predicting the presence of car primer with residual error value less than 0.05 and correct classification rate of more than 95%. Evaluation on the application of ANN was done by analyzing 39 real car paint samples. PCA applied to untreated dataset and dataset treated with developed ANN prediction model. Better discrimination was obtained using treated dataset whereby only samples with car primer were included. Thus, the ANN model constructed may be applied as a pretreatment approach on a Py-GC-MS dataset for forensic car paint analysis.