An efficient approach towards predictive kinetic models. W. Hauptmann, M. Votsmeier, J. Gieshoff, D. G. Vlachos, A. Drochner, H.

Transcription

1 An efficient approach towards predictive kinetic models W. Hauptmann, M. Votsmeier, J. Gieshoff, D. G. Vlachos, A. Drochner, H. Vogel

2 Introduction - Literature Mechanisms heating Experiment Run1 Olsson 1999 Koop 2008 Olsson T / C, eq L. Olsson, B. Westerberg, H. Persson, E. Fridell, M. Skoglundh, B. Andersson, Journal of Physical Chemistry B 1999, 103, J. Koop, Dissertation, Universität Karlsruhe L. Olsson, H. Persson, E. Fridell, M. Skoglundh, B. Andersson, Journal of Physical Chemistry B 2001, 105,

3 Modeling Process - Process Steps - Model Model Development Experiments Simulation Model Model Analysis Optimization Predictive Model Model 3

4 Experiments Model Model Development Experiments Simulation Model Model Analysis Optimization Predictive Model Model 4

5 Experiments - Test Bench - Gas supply Analytics Monolith holder Oven 5

6 Experiments - Inverse Hysteresis heating cooling Experiment Run1 Experiment Run2,eq 400 T Ofen T / C Pre- Treatment Run1 Run T / C x x x+60 x x +74 x x +74 x x +74 x +74 t / min 0 (NO) = 430 vol.-ppm; 0 (O 2 ) = 6 vol.-%; balance N 2 ; = ±5 K min -1 ; 6

7 Model Development Model Model Development Experiments Simulation Model Model Analysis Optimization Predictive Model Model 7

8 Model Development - Mechanism - R1 R2 R3 R4 R5 R6 R7 R8 O Pt 2 Pt-O NO + Pt Pt-NO NO 2 + Pt Pt-NO 2 Pt-NO + Pt-O Pt-NO 2 + Pt 2 Pt-O O Pt Pt-NO NO + Pt Pt-NO 2 NO 2 + Pt Pt-NO 2 + Pt Pt-NO + Pt-O L. Olsson, H. Persson, E. Fridell, M. Skoglundh, B. Andersson, Journal of Physical Chemistry B 2001, 105, W. Hauptmann, M. Votsmeier, J. Gieshoff, D. G. Vlachos, A. Drochner, H. Vogel, Topics in Catalysis 2009, doi: /s z. 8

9 Model Development - Pre-exponential Factors - Collision Theory Theory Transition State State Theory Theory Adsorption Desorption Bimolecular Surface Reaction Sticking Coefficient A = s -1-1 (DoF (DoF Transition State > DoF Transition State > DoF Adsorbates ) Adsorbates ) A = s -1-1 (Mobile adsorbates (Mobile adsorbates without rotation) without rotation) J. A. Dumesic et. al., The Microkinetics of Heterogeneous Catalysis, American Chemical Society, Washington DC

10 Model Development - Activation Energy - Surface reaction (AB (AB A + B): B): Activation energy energy (forward reaction): Surf Ea f xab, TS H H H Surf Surf :: Dissociation Surf H Q D AB QAQB Q Q AB A Q A C B C Q B Adsorption: E a = a kj kj mol mol -1-1 Desorption: E a = a Heat Heat of of adsorption UBI-QEP Bimolecular reaction H Surf Q C Q AB D AB D BC Q A Q BC Activation energy energy (back (back reaction): Surf Ea Ea H b f E. Shustorovich, H. Sellers, Surface Science Reports 1998, 31,

11 Simulation Model Model Development Experiments Simulation Model Model Analysis Optimization Predictive Model Model 11

12 Simulation - Initial Mechanism heating Experiment Run1 Simulation Run1,eq T / C 0 (NO) = 430 vol.-ppm; 0 (O 2 ) = 6 vol.-%; balance N 2 ; = ±5 K min -1 ; 12

13 Modeling Process - Process Steps - Model Model Development Experiments Simulation Model Model Analysis Optimization Predictive Model Model 13

14 Model Analysis - Sensitivity Analysis - Objective: Identification of of relevant parameters and/or and/or reactions. 14

15 Model Analysis - Sensitivity Analysis - Objective: Identification of of relevant parameters and/or and/or reactions. Procedure: Iterative variation of of each each parameter value. value. Calculation of of normalized sensitivity coefficients (NSC). (NSC). Calculation: Resi SCi, r Para NSC i, r r Para Res i r Res i Para r 15

16 Model Analysis - Sensitivity Analysis - Objective: Identification of of relevant parameters and/or and/or reactions. Reaction R7 R6 R5 R4 R3 R2 R NSC A' = A Original / 10 Ea' = Ea Original +-20 kj mol -1 Ea' = Ea Original +-30 kj mol -1 Procedure: Iterative variation of of each each parameter value. value. Calculation of of normalized sensitivity coefficients (NSC). (NSC). Calculation: Resi SCi, r Para 16 NSC i, r r Para Res i r Res i Para Interpretation: NSC-value corresponds to to parameter impact. impact. Sign Sign (+/-) (+/-) shows shows proportionality. r

17 Model Analysis -Result - Pt-NO + Pt-O Pt-NO 2 + Pt Pt-NO 2 NO 2 + Pt NO 2 + Pt Pt-NO 2 Pt-NO NO + Pt NO + Pt Pt-NO 2 Pt-O O Pt O Pt 2 Pt-O A' = A Original / 10 Ea' = Ea Original +-20 kj mol -1 Ea' = Ea Original +-30 kj mol NSC 17

18 Optimization Model Model Development Experiments Simulation Model Model Analysis Optimization Predictive Model Model 18

19 Optimization - Iterative Process - Model Initial Guess DoE Use last Initial Guess. Reduce Size of Design. Use opt. Parameters as Initial Guess. Keep Size of Design. Fit Polynomial Reference Data No Res New < Res Old Yes Minimize Objective Function Res < SC No Yes Stop Optimization 19

20 Optimization - Optimization Results heating 0.2 Experiment Run1 Simulation Run1,eq T / C 0 (NO) = 430 vol.-ppm; 0 (O 2 ) = 6 vol.-%; balance N 2 W. Hauptmann, M. Votsmeier, J. Gieshoff, D. G. Vlachos, A. Drochner, H. Vogel, Topics in Catalysis 2009, doi: /s z. 20

21 Optimization - Model Assessment heating Experiment Run1 Simulation Run1,eq T / C 0 (NO) = 280 vol.-ppm; 0 (O 2 ) = 6 vol.-%; balance N 2 W. Hauptmann, M. Votsmeier, J. Gieshoff, D. G. Vlachos, A. Drochner, H. Vogel, Topics in Catalysis 2009, doi: /s z. 21

22 Optimization - Model Assessment Light-Off/Light-Out Mechanism can t can t describe inverse inverse hysteresis Experiment Run1 Simulation Run1,eq T / C 0 (NO) = 430 vol.-ppm; 0 (O 2 ) = 6 vol.-%; balance N 2 22

23 Model Elaboration Model Model Development Experiments Simulation Model Model Analysis Optimization Predictive Model Model 23

24 Model Elaboration - Implementation of Catalyst Deactivation - R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 O Pt 2 Pt-O NO + Pt Pt-NO NO 2 + Pt Pt-NO 2 Pt-NO + Pt-O Pt-NO 2 + Pt Pt-NO 2 PtOx + NO 2 Pt-O O Pt Pt-NO NO + Pt Pt-NO 2 NO 2 + Pt Pt-NO 2 + Pt Pt-NO + Pt-O PtOx + NO Pt-NO 2 W. Hauptmann, M. Votsmeier, J. Gieshoff, A. Drochner, H. Vogel, Applied Catalysis B: Environmental 2009, accepted. 24

25 Model Elaboration - Model Performance heating cooling Experiment Run1 Experiment Run2 Simulation Run1 0.2 Simulation Run2, eq T / C 0 (NO) = 430 vol.-ppm; 0 (O 2 ) = 6 vol.-%; balance N 2 W. Hauptmann, M. Votsmeier, J. Gieshoff, A. Drochner, H. Vogel, Applied Catalysis B: Environmental 2009, accepted. 25

26 Model Elaboration - Model Assessment heating cooling Experiment Run1 Experiment Run2 0.2 Simulation Run1 Simulation Run2,eq T / C 0 (NO) = 280 vol.-ppm; 0 (O 2 ) = 6 vol.-%; balance N 2 W. Hauptmann, M. Votsmeier, J. Gieshoff, A. Drochner, H. Vogel, Applied Catalysis B: Environmental 2009, accepted. 26

27 Model Elaboration - Model Assessment II heating cooling t / h T / C300 Exp. Part1 Exp. Part2 Exp. Part3 Sim. Part1 Sim. Part2 Sim. Part3,eq T T Exp. Sim T / C T / C t / h 0 (NO) = 430 vol.-ppm; 0 (O 2 ) = 6 vol.-%; balance N 2 W. Hauptmann, M. Votsmeier, J. Gieshoff, A. Drochner, H. Vogel, Applied Catalysis B: Environmental 2009, accepted. 27

28 Summary Experiments: Experiments: The The NO NO oxidation oxidation on on Pt Pt shows shows an an inverse inverse hysteresis hysteresis Model Model Development: Development: Applied Applied theories theories are are appropriate appropriate to to parameterize parameterize kinetic kinetic models. models. Simulation: Simulation: Low Low temperature temperature activity activity of of the the initial initial mechanism mechanism is is to to high. high. Model Model Analysis: Analysis: Sensitivity Sensitivity analysis analysis and and Shainin Shainin analysis analysis are are able able to to determine determine the the active active parameters. parameters. Optimization: Optimization: Iterative Iterative process process is is appropriate appropriate for for parameter parameter estimation. estimation. Predictive Predictive Model: Model: Model Model without without catalyst catalyst deactivation deactivation is is able able to to describe describe the the light-off-behavior. light-off-behavior. Model Model including including catalyst catalyst deactivation deactivation is is able able to to describe describe the the light-off/light-out light-off/light-out behavior. behavior. 28

29 Special thanks go to Alfons Drochner Herbert Vogel Martin Votsmeier Jürgen Gieshoff Dion Vlachos 29

30 Thank you for your kind attention!