Courtney Phillips Leap Technologies John Edwards Process NMR - PowerPoint PPT Presentation

High Throughput Petroleum Stream Analysis in Refinery Process Laboratories: Benchtop NMR Offers Timely Results with Automation & Chemometrics Courtney Phillips Leap Technologies John Edwards Process NMR Associates

Residual Fluidized Catalytic Cracking Feed-stream Analysis Traditional Analysis – Refractive Index, Distillation,Viscosity Specific Gravity Calculation – Watson K-Factor Outcome: aromatic carbon number aromatic hydrogen number total hydrogen content NMR Proposition: Detailed hydrocarbon analysis for kinetic model development

CH 2 Alpha Aromatic CH 2 CH 3 DiAromatic Di+Tri Monoaromatic Aromatic Tri+Tetra Ar-CH 2 -Ar Aromatic CH 3

Data Stacked – Referenced – TMS excluded – spectrum binned (integrated every 0.04 pm from -1 to +11 ppm) - Normalized Y-Block X-Block 0.9399 0.9574 0.9230 0.9255 0.9354 0.9122 0.9000 0.8987 0.8823 0.9786 0.9112 0.9 0.8781 0.9270 0.9255 0.9359 0.9006 0.8745 0.9271 0.9215 0.8882 PLS 0.9233 0.8802 0.8847 0.8842 0.8628 0.8753 0.9217 0.9208 0.9224 0.8888 0.9258 0.8795 0.8663 0.8990 0.8757 0.9488 0.9272 0.9021 0.8762 0.9233 0.9390 0.8916 0.9042 0.9213 0.8829 0.8779 0.8930 0.9130 0.9208 0.9218 0.8979 0.9364 0.8946

Data after importation into Eigenvector Research Chemometrics Package – Note Reversal of Spectrum 160 140 6 120 5 100 4 Data Data 80 3 2 60 1 40 0 60 80 100 120 140 160 180 200 220 Variables 20 0 50 100 150 200 Variables

Mod odel Linear regression model using 1 H NMR Prediction Model for Density (g/ml) of RCC Feed Partial Least Squares calculated with the SIMPLS algorithm Developed 07-Oct-2014 23:15:09.003 Samples/Scores - PLS 9 LVs - Binned Data.csv, Density.xlsx Author: John@JCEPNA X-block: Binned Data.csv 46 by 248 3 3 Included: [ 1-6 8-11 14-32 36-38 41-46 48-54 56 ] [ 1-248 ] 2.5 Preprocessing: Mean Center 2 Q Residuals (0.06%) Y Stdnt Residual 1 Y-block: Density.xlsx 46 by 1 2 1 Included: [ 1-6 8-11 14-32 36-38 41-46 48-54 56 ] [ 1 ] Preprocessing: Mean Center 1.5 0 Num. LVs: 9 Cross validation: venetian blinds w/ 7 splits -1 1 RMSEC: 0.0034 g/ml RMSECV: 0.0052 g/ml -2 Bias: 3.33067e-016 CV Bias: -9.35779e-005 0.5 R^2 Cal: 0.980 R^2 CV: 0.954 -3 0 0 10 20 30 0 0.2 0.4 0.6 0.8 SSQ Table le Hotelling T^2 (99.94%) Leverage Percent Variance Captured by Regression Model -----X-Block----- -----Y-Block----- 1 20 Comp This Total This Total ---- ------- ------- ------- ------- Scores on LV 2 (3.37%) 10 1 89.72 89.72 71.31 71.31 Y CV Predicted 1 0.95 2 3.37 93.09 15.87 87.18 3 3.77 96.86 2.61 89.79 0 4 1.85 98.71 1.10 90.89 5 0.39 99.10 3.35 94.24 0.9 6 0.40 99.50 1.18 95.42 -10 7 0.33 99.83 0.82 96.24 8 0.04 99.86 1.54 97.78 9 0.08 99.94 0.24 98.02 0.85 -20 0.85 0.9 0.95 1 -100 -50 0 50 Y Measured 1 Scores on LV 1 (89.72%)

Variables/Loadings Plot for Binned Data.csv 0.1 0 Examples of Latent Variable Loadings for LV1, LV4, and LV9 -0.1 -0.2 Variables/Loadings Plot for Binned Data.csv 0.4 LV 1 (89.72%) -0.3 -0.4 0.2 -0.5 0 -0.6 Variables/Loadings Plot for Binned Data.csv 0.4 -0.7 -0.2 LV 4 (1.85%) -0.8 0.3 50 100 150 200 Variable -0.4 0.2 -0.6 0.1 LV 9 (0.08%) -0.8 0 -1 50 100 150 200 Variable -0.1 -0.2 -0.3 50 100 150 200 Variable

Mod odel 1 H NMR Prediction Model for API Gravity of RCC Feed Linear regression model using Partial Least Squares calculated with the SIMPLS algorithm Developed 06-Oct-2014 15:54:029.09 2 3 Author: John@JCEPNA X-block: 54 Samples - Binned Data - X-Block.xlsx 47 by 248 2 Q Residuals (0.05%) 1.5 Y Stdnt Residual 1 Included: [ 2-10 12-14 16-35 38-46 48-53 ] [ 1-248 ] 1 Preprocessing: Mean Center Y-block: API Gravity - 54 Samples.xlsx 47 by 1 1 0 Included: [ 2-10 12-14 16-35 38-46 48-53 ] [ 1 ] Preprocessing: Mean Center -1 Num. LVs: 9 0.5 Cross validation: venetian blinds w/ 7 splits -2 RMSEC: 0.51 deg RMSECV: 0.77 deg -3 Bias: -1.42109e-014 CV Bias: 0.0102836 0 0 5 10 15 20 0 0.2 0.4 0.6 0.8 R^2 Cal: 0.98737 R^2 CV: 0.971025 Hotelling T^2 (99.95%) Leverage SSQ Table le Percent Variance Captured by Regression Model 35 40 -----X-Block----- -----Y-Block----- Scores on LV 2 (5.66%) 30 Comp This Total This Total 20 Y CV Predicted 1 ---- ------- ------- ------- ------- 25 1 88.67 88.67 72.12 72.12 0 2 5.66 94.33 11.86 83.98 3 2.69 97.02 5.02 89.00 20 4 1.88 98.90 1.12 90.12 -20 5 0.41 99.30 2.74 92.86 15 6 0.39 99.70 1.84 94.70 7 0.17 99.87 1.57 96.27 10 -40 8 0.03 99.91 2.06 98.33 10 15 20 25 30 35 40 -50 0 50 100 9 0.04 99.95 0.41 98.74 Y Measured 1 Scores on LV 1 (88.67%)

Mod odel 1 H NMR Prediction Model for Fraction Carbon Aromaticity of RCC Feed Linear regression model using Primary Test Method: 13 C NMR Partial Least Squares calculated with the SIMPLS algorithm Developed 07-Oct-2014 23:45:055.39 Analysis - PLS 3 LVs - 54 Samples - Binned Data - X-Block.xlsx, Fa_C13.xlsx Author: John@JCEPNA 0.08 3 X-block: 54 Samples - Binned Data - X-Block.xlsx 51 by 97 2 Included: [ 2-35 38-54 ] [ 1 153-248 ] Q Residuals (0.65%) 0.06 Y Stdnt Residual 1 Preprocessing: Mean Center 1 Y-block: Fa_C13.xlsx 51 by 1 0 0.04 Included: [ 2-35 38-54 ] [ 1 ] -1 Preprocessing: Mean Center -2 0.02 Num. LVs: 3 -3 Cross validation: venetian blinds w/ 7 splits RMSEC: 0.0067 RMSECV: 0.0074 0 -4 0 5 10 15 0 0.1 0.2 0.3 0.4 Bias: -2.77556e-017 CV Bias: 0.0002 Hotelling T^2 (99.35%) Leverage R^2 Cal: 0.975 R^2 CV: 0.970 0.25 1 SSQ SSQ Tab able Percent Variance Captured by Regression Model Scores on LV 2 (3.41%) 0.2 0.5 Y CV Predicted 1 -----X-Block----- -----Y-Block----- 0.15 0 Comp This Total This Total ---- ------- ------- ------- ------- 0.1 -0.5 1 93.95 93.95 88.54 88.54 2 3.41 97.36 5.84 94.38 3 2.00 99.35 3.15 97.53 0.05 -1 0.05 0.1 0.15 0.2 0.25 -4 -2 0 2 4 Y Measured 1 Scores on LV 1 (93.95%)

Mod odel Linear regression model using 1 H NMR Prediction Model for Sulfur Content (Wt%) of RCC Feed Partial Least Squares calculated with the SIMPLS algorithm Primary Method - XRF Developed 08-Oct-2014 00:26:012.79 Author: John@JCEPNA 0.08 3 X-block: 54 Samples - Binned Data - X-Block.xlsx 48 by 88 Included: [ 2-14 16-30 32-35 38-51 53-54 ] [ 1 162-248 ] 2 Q Residuals (0.37%) 0.06 Y Stdnt Residual 1 Preprocessing: Mean Center 1 Y-block: Sulfur - 54 Samples.xlsx 48 by 1 0.04 0 Included: [ 2-14 16-30 32-35 38-51 53-54 ] [ 1 ] Preprocessing: Mean Center -1 Num. LVs: 5 0.02 -2 Cross validation: venetian blinds w/ 7 splits -3 RMSEC: 0.17 wt% RMSECV: 0.26 wt% 0 0 5 10 15 0 0.1 0.2 0.3 0.4 Bias: -1.33227e-015 CV Bias: 0.004 Hotelling T^2 (99.63%) Leverage R^2 Cal: 0.958 R^2 CV: 0.902 4 1 SSQ SSQ Tab able Percent Variance Captured by Regression Model 0.5 Scores on LV 2 (4.35%) 3 Y CV Predicted 1 0 -----X-Block----- -----Y-Block----- Comp This Total This Total 2 -0.5 ---- ------- ------- ------- ------- -1 1 93.93 93.93 55.11 55.11 1 2 4.35 98.28 17.67 72.78 -1.5 3 1.11 99.39 12.16 84.94 0 -2 4 0.17 99.56 7.02 91.96 0 1 2 3 4 -4 -2 0 2 4 Y Measured 1 Scores on LV 1 (93.93%) 5 0.08 99.63 3.86 95.82

Mod odel 1 H NMR Prediction Model for Total Aromatic Content (Wt%) of RCC Feed Linear regression model using Primary Method – HPLC-UV-DAD Partial Least Squares calculated with the SIMPLS algorithm Developed 16-Sep-2014 00:39:028.36 0.2 4 Author: John@JCEPNA X-block: 54 Samples - Binned Data - X-Block.xlsx 49 by 73 Q Residuals (1.04%) 0.15 2 Y Stdnt Residual 1 Included: [ 1 3-33 35 38-51 53-54 ] [ 1-2 159-229 ] Preprocessing: Mean Center 0.1 0 Y-block: Total Aromatics - 54 Samples.xlsx 49 by 1 Included: [ 1 3-33 35 38-51 53-54 ] [ 1 ] Preprocessing: Mean Center 0.05 -2 Num. LVs: 3 Cross validation: venetian blinds w/ 7 splits 0 -4 0 5 10 15 0 0.1 0.2 0.3 0.4 RMSEC: 0.80 wt% RMSECV: 0.88 wt% Hotelling T^2 (98.96%) Leverage Bias: 3.55271e-015 CV Bias: -0.02 R^2 Cal: 0.949 R^2 CV: 0.937 25 0.5 SSQ SSQ Tab able Scores on LV 2 (0.85%) 20 Y CV Predicted 1 Percent Variance Captured by Regression Model 15 0 -----X-Block----- -----Y-Block----- Comp This Total This Total ---- ------- ------- ------- ------- 10 1 93.82 93.82 88.47 88.47 2 0.85 94.67 6.38 94.85 5 -0.5 5 10 15 20 25 -4 -2 0 2 4 3 4.29 98.96 0.07 94.92 Y Measured 1 Scores on LV 1 (93.82%)