1 / 24

Development of Infrared Library Search Prefilters for Automotive Clear Coats

This study focuses on the development of search prefilters for infrared (IR) library searching of automotive clear coats using simulated attenuated total reflection (ATR) spectra. The aim is to overcome distortions in ATR spectra compared to transmission spectra and improve the accuracy and selectivity of IR library searches.

devonj
Download Presentation

Development of Infrared Library Search Prefilters for Automotive Clear Coats

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Development of Infrared Library Search Prefilters for Automotive Clear Coats from Simulated ATR Spectra Barry K. Lavine1, Undugodage Perera1 and Koichi Nishikida2 1Department of Chemistry, Oklahoma State University 2Materals Science Center, College of Engineering, University of Wisconsin at Madison

  2. Transmission Versus Attenuated Total Reflection ATR Transmission Differs in optical configuration and process

  3. d p = 2pnatr(sin2q) - ] [ nsample 2 ( ) 1/2 natr Depth of Penetration l 1/2 • natr = Refractive index of ATR element • q = Angle of incidence of ATR element • nsample = Refractive index of sample • l = Wavelength 1/2

  4. Motivation • ATR is popular because it requires minimal sample preparation. • As the penetration depth of the ATR analysis beam is shallow, the outer layers of a laminate or multi-layered paint sample can be preferentially analyzed with the entire paint sample intact. • However, the IR spectrum of a paint sample obtained by ATR will exhibit distortions when compared to its transmission counterpart. • This hinders library searching as most library spectra are measured in transmission mode.

  5. Variation of Refractive Index, n, and Absorption Index, k, Across an Absorption Band Optical constant Spectra of simulated infrared absorption band Figure 13.19 from Fourier Transfom Infrared Spectrometry (2nd Edition), by P. R. Griffiths and J. A. de Haseth, Wiley Interscience, Hoboken, NJ (2007)

  6. K-Index The fundamental parameters that govern the absorption of radiation are the real and imaginary components of the complex refractive index: n =n+ ik Typically looks like absorbance spectrum Absorption Index Wavenumber cm -1

  7. N-Index This type of refractive index across absorption bands is known as anomalous dispersion Refractive Index Wavenumber cm-1

  8. ATR Spectral Distortions • IR spectra obtained by ATR exhibit distortions compared to transmission spectra: • Band broadening • Band shifting • Lower relative intensities • An algorithm to convert transmission libraries to ATR spectral libraries is needed. • Spectra only need to be corrected once. Comparison of simulated ATR spectrum with experimentally measured ATR spectrum can reveal problems associated with the measurement.

  9. Transmission to ATR Kramers-Kronig Relationship Fresnel’s reflection coefficients for s and p-polarized beam rp= reflectance of parallel polarized radiation; rs= reflectance of perpendicularly polarized radiation (5)

  10. ATR Simulation 0.90 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50 Absorbance 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 -0.00 4000 3500 3000 2500 2000 1500 1000 500 Wavenumbers (cm-1) Transmission Calculated ATR Observed ATR

  11. Simulated Versus Real 0.35 (a) Simulated 0.15 Absorbance (a.u.) 0.40 Real (b) 0.20 0.00 4000 3500 3000 2500 2000 1500 1000 500 Wavenumber (cm-1)

  12. Pattern Recognition Analysis of Simulated ATR Spectra • Search prefilters (principal component plots that act as discriminants) were developed from ATR transformed IR spectra that can identify the assembly plant of a General Motors vehicle from the ATR spectrum of the clear coat layer recovered from a crime scene. • Use of pattern recognition methods to truncate the PDQ infrared (IR) library to a specific assembly plant increases both the accuracy and the selectivity of the IR search when the search is performed under difficult conditions, e.g., when the IR spectra of the layers are similar and when specific details of identification are required in sparse sample and data conditions.

  13. Search Prefilter Development • 456 transmission spectra from the PDQ database spanning 22 General Motors assembly plants within a limited production year range (2000-2006) that serve as the training set cohort were transformed into ATR spectra using the simulation algorithm. • The spectral region selected to develop these search prefilters for IR library searching was the fingerprint region (1500 cm-1 to 500 cm-1). • Both the transformed ATR spectra (training set) and the experimental ATR spectra (validation set) were preprocessed using the discrete wavelet transform to increase the signal to noise of the data by concentrating the signal in specific wavelet coefficients. • All search prefilters developed in this study were validated by ATR spectra of 14 clear coats collected using a Nicolet iS50 FTIR spectrometer.

  14. Hierarchical Search Prefilters • The 456 simulated ATR spectra comprising the training set cohort were divided into five plant groups based upon the carbonyl band (singlet versus doublet) and a visual analysis of the fingerprint region of each simulated spectrum. • For Plant Groups 1, 3, and 4, the carbonyl band is a singlet indicative of acrylic melamine styrene, whereas for Plant Groups 2 and 5, the carbonyl band is a doublet indicative of acrylic melamine styrene polyurethane. • Although PDQ assigns a text code for each clear coat sample based on its composition (either acrylic melamine styrene or acrylic melamine styrene polyurethane), the assigned text code gives no indication of how much of each constituent is present in the formulation. • Clearly, these five plant groups indicate variations in the constituents comprising the formulations used to prepare modern automotive clear coats. • An unknown clear coat is first classified as to its plant group and then is assigned to a specific assembly plant or assembly plants within the plant group using a second search prefilter.

  15. Training Set for Plant Group Search Prefilters

  16. Principal Component Plot of 1178 Wavelet Coefficients 1 = Plant Group 1 2 = Plant Group 2 3 = Plant Group 3 4 = Plant Group 4 5 = Plant Group 5

  17. Genetic Algorithm for Pattern Recognition Analysis • To identify the wavelet coefficients correlated to plant group, the genetic algorithm for pattern recognition analysis was applied to the training set data. • The pattern recognition GA identifies wavelet coefficients characteristic of plant group by sampling key feature subsets, scoring their principal component plots and tracking those plant groups and/or simulated ATR spectra that are difficult to classify. • The boosting routine used this information to steer the population to an optimal solution. • After 200 generations, the pattern recognition GA identified twenty-seven wavelet coefficients whose PC plot showed clustering of the simulated ATR spectra on the basis of plant group.

  18. Principal Component Plot of 27 Wavelet Coefficients 1 = Plant Group 1 2 = Plant Group 2 3 = Plant Group 3 4 = Plant Group 4 5 = Plant Group 5

  19. Validation of Search Prefilter for Plant Group • To assess the predictive ability of the 27 wavelet coefficients identified by the pattern recognition GA, a validation set of 14 clear coats whose ATR spectra were measured by a Nicolet iS50 FTIR spectrometer was employed. • The 14 ATR spectra comprising the validations set were projected onto the PC plot developed from the 456 simulated ATR spectra of the training set and the twenty-seven wavelet coefficients identified by the pattern recognition GA. • All 14 clear coats were correctly classified. Each projected sample was located in a region of the plot with samples from the same plant group.

  20. Projection of 14 Validation Set Samples onto PC Plot Developed from the 27 Wavelet Coefficients Training Set 1 = Plant Group 1 2 = Plant Group 2 3 = Plant Group 3 4 = Plant Group 4 5 = Plant Group 5 Validation Set A = Plant Group 1 B = Plant Group 2 C = Plant Group 3 D = Plant Group 4 E = Plant Group 5

  21. Without ATR Simulation Algorithm • Without using the ATR simulation algorithm, all validation set samples were incorrectly classified by the discriminant (i.e., principal component plot of the selected wavelet coefficients) developed from the transmission IR spectra of the clear coats from the PDQ library. • Furthermore, a search prefilter developed for plant group using the original 456 IR transmission spectra of the General Motors clear coats (from the PDQ library) wavelet transformed using 8Sym6 was unable to correctly classify any of the 14 validation set spectra that were transformed from ATR spectra into transmission spectra using the advanced ATR correction method in OMNIC (Thermo-Nicolet).

  22. Search Prefilter Developed from Original IR Transmission Spectra Validation Set A = Plant Group 1 B = Plant Group 2 C = Plant Group 3 D = Plant Group 4 E = Plant Group 5 Training Set 1 = Plant Group 1 2 = Plant Group 2 3 = Plant Group 3 4 = Plant Group 4 5 = Plant Group 5

  23. Importance of Simulation Algorithm • Emmons and coworkers in a previous studyreported that frequency shifts for some vibrational modes are observed in IR spectra of polymers measured in a high pressure transmission diamond anvil cell (DAC). • He attributed these observed frequency shifts to the removal of void spaces in the polymer which occurred during the compression of the sample by the DAC. • Although the advanced ATR correction module of OMNIC cannot correct for this type of spectral shift when converting ATR spectra to IR transmission spectra, the ATR simulation algorithm is able to correct for these shifts because the simulation algorithm utilizes both the IR transmission spectra of automotive clear coats from the PDQ library and the corresponding ATR spectra of the same paint samples to develop estimates of the incident angle relative to the normal for the IRE, the refractive index of the clear coat layer, and the thickness of the paint sample after compression by the DAC. • As there are currently no commercial vendors that distribute high pressure diamond cells with sufficient energy throughput for automotive paint analysis, the ATR simulation algorithm described in this study will allow forensic laboratories to utilize the spectral database of PDQ for IR library matching.

  24. Acknowledgements • Dr. Peter Griffiths and Dr. James de Haseth (IR courses) • Mark Sandercock of the Royal Canadian Mounted Police Forensic Laboratory • This study was supported by Award No. 2012-DN-BX-K05, the National Institute of Justice, Office of Justice Programs, United States Department of Justice. The opinions, findings, and conclusions or recommendations expressed in this publication/program/exhibition are those of the author(s) and do not necessarily reflect those of the Department of Justice.

More Related