Brian Critchfield Uchenna Paul Prof. Calvin Bartholomew Prof. Dennis Tolley Design of Kinetic Experiments for Fischer-Tropsch Synthesis on Supported Fe.

Slides:



Advertisements
Similar presentations
CO x -Free Hydrogen by Catalytic Decomposition of Ammonia on Commercial Fe and Ru Catalysts: An Experimental and Theoretical Study Caitlin Callaghan Barry.
Advertisements

CmpE 104 SOFTWARE STATISTICAL TOOLS & METHODS MEASURING & ESTIMATING SOFTWARE SIZE AND RESOURCE & SCHEDULE ESTIMATING.
Experimental Design, Response Surface Analysis, and Optimization
Dimension reduction (1)
Nonlinear Regression Ecole Nationale Vétérinaire de Toulouse Didier Concordet ECVPT Workshop April 2011 Can be downloaded at
P M V Subbarao Professor Mechanical Engineering Department
11.1 Introduction to Response Surface Methodology
Response Surface Method Principle Component Analysis
Classification and Prediction: Regression Via Gradient Descent Optimization Bamshad Mobasher DePaul University.
Curve-Fitting Regression
Development of Reliable, Simple Rate Expressions from a Microkinetic Model of FTS on Cobalt Calvin H. Bartholomew, George Huber, and Hu Zou Brigham Young.
L Berkley Davis Copyright 2009 MER301: Engineering Reliability Lecture 14 1 MER301: Engineering Reliability LECTURE 14: Chapter 7: Design of Engineering.
Multivariate Data Analysis Chapter 4 – Multiple Regression.
Generalized Kinetics of Fischer-Tropsch Synthesis on Supported Cobalt
Engineering Computation Curve Fitting 1 Curve Fitting By Least-Squares Regression and Spline Interpolation Part 7.
Lecture 17 Today: Start Chapter 9 Next day: More of Chapter 9.
© 2014 Carl Lund, all rights reserved A First Course on Kinetics and Reaction Engineering Class 10.
1 1 Slide © 2008 Thomson South-Western. All Rights Reserved Slides by JOHN LOUCKS & Updated by SPIROS VELIANITIS.
Calibration & Curve Fitting
Objectives of Multiple Regression
1 Statistical Tools for Multivariate Six Sigma Dr. Neil W. Polhemus CTO & Director of Development StatPoint, Inc. Revised talk:
Integration of the rate laws gives the integrated rate laws
Multiple Linear and Polynomial Regression with Statistical Analysis Given a set of data of measured (or observed) values of a dependent variable: y i versus.
Visualization, reduction and simplification of a water gas shift mechanism through the application of reaction route graphs CA Callaghan, I Fishtik, and.
Stats for Engineers Lecture 9. Summary From Last Time Confidence Intervals for the mean t-tables Q Student t-distribution.
MODEGAT Chalmers University of Technology Use of Latent Variables in the Parameter Estimation Process Jonas Sjöblom Energy and Environment Chalmers.
Non-Linear Models. Non-Linear Growth models many models cannot be transformed into a linear model The Mechanistic Growth Model Equation: or (ignoring.
CJT 765: Structural Equation Modeling Class 7: fitting a model, fit indices, comparingmodels, statistical power.
Discovering Dynamic Models Lecture 21. Dynamic Models: Introduction Dynamic models can describe how variables change over time or explain variation by.
Non-Linear Models. Non-Linear Growth models many models cannot be transformed into a linear model The Mechanistic Growth Model Equation: or (ignoring.
Practical Statistical Analysis Objectives: Conceptually understand the following for both linear and nonlinear models: 1.Best fit to model parameters 2.Experimental.
Quality of Curve Fitting P M V Subbarao Professor Mechanical Engineering Department Suitability of A Model to a Data Set…..
Application of reaction route graphs to microkinetic analysis and design of water-gas-shift catalysts Fuel Cell Center Chemical Engineering Department.
Various topics Petter Mostad Overview Epidemiology Study types / data types Econometrics Time series data More about sampling –Estimation.
© 2014 Carl Lund, all rights reserved A First Course on Kinetics and Reaction Engineering Class 6.
1 Handling Uncertainty in the Development and Design of Chemical Processes David Bogle, David Johnson and Sujan Balendra Centre for Process Systems Engineering.
Response surfaces. We have a dependent variable y, independent variables x 1, x 2,...,x p The general form of the model y = f(x 1, x 2,...,x p ) +  Surface.
Chemical Reaction Engineering Lecture (1) Week 2.
Lesson Multiple Regression Models. Objectives Obtain the correlation matrix Use technology to find a multiple regression equation Interpret the.
Bashkir State Univerity The Chair of Mathematical Modeling , Ufa, Zaki Validi str. 32 Phone: ,
Propagation of Error Ch En 475 Unit Operations. Quantifying variables (i.e. answering a question with a number) 1. Directly measure the variable. - referred.
Non-Linear Models. Non-Linear Growth models many models cannot be transformed into a linear model The Mechanistic Growth Model Equation: or (ignoring.
Validation Defination Establishing documentary evidence which provides a high degree of assurance that specification process will consistently produce.
Reactor Design S,S&L Chapter 6. Objectives De Novo Reactor Designs Plant Improvement –Debottlenecking –Increase Plant Capacity –Increase Plant Efficiency.
Advanced Residual Analysis Techniques for Model Selection A.Murari 1, D.Mazon 2, J.Vega 3, P.Gaudio 4, M.Gelfusa 4, A.Grognu 5, I.Lupelli 4, M.Odstrcil.
Statistics Presentation Ch En 475 Unit Operations.
L Berkley Davis Copyright 2009 MER301: Engineering Reliability Lecture 12 1 MER301: Engineering Reliability LECTURE 12: Chapter 6: Linear Regression Analysis.
More on regression Petter Mostad More on indicator variables If an independent variable is an indicator variable, cases where it is 1 will.
ESTIMATION METHODS We know how to calculate confidence intervals for estimates of  and  2 Now, we need procedures to calculate  and  2, themselves.
Parameter Estimation. Statistics Probability specified inferred Steam engine pump “prediction” “estimation”
MathematicalMarketing Slide 5.1 OLS Chapter 5: Ordinary Least Square Regression We will be discussing  The Linear Regression Model  Estimation of the.
Kinetic analysis of Temperature Programmed Reduction R. Jude vimal Michael National Centre for Catalysis Research 31 January 2009.
Computacion Inteligente Least-Square Methods for System Identification.
Using COMSOL for Chemical Reaction Engineering Your name COMSOL.
IC-1/38 Lecture Kinetics IC-2/38 Lecture What is Kinetics ? Analysis of reaction mechanisms on the molecular scale Derivation.
A.N.Zagoruiko. Anaerobic catalytic oxidation of hydrocarbons in moving heat waves. Case simulation: propane oxidative dehydrogenation in a packed adiabatic.
CSE 330: Numerical Methods. What is regression analysis? Regression analysis gives information on the relationship between a response (dependent) variable.
IC-1/18 Lecture Kinetics. IC-2/18 Lecture What is Kinetics ? Analysis of reaction mechanisms on the molecular scale Derivation.
ChE 402: Chemical Reaction Engineering
Deep Feedforward Networks
Regression Analysis Part D Model Building
Microkinetic Study of CO Adsorption and Dissociation on Fe Catalysts
Progress in Development of Activity Models of FT Synthesis on Cobalt
CJT 765: Structural Equation Modeling
D. Foster, T. Root, T. Kawai, E. Wirojsakunchai, E. Schroeder, N
CO Adsorption, Dissociation and Hydrogenation on Fe
Statistical Methods For Engineers
A Systematic And Mechanistic Analysis Of The Water-Gas Shift Reaction Kinetics On Low And High Temperature Shift Catalysts CA Callaghan, I Fishtik, and.
ESTIMATION METHODS We know how to calculate confidence intervals for estimates of  and 2 Now, we need procedures to calculate  and 2 , themselves.
The TSR A Kinetics Instrument
Presentation transcript:

Brian Critchfield Uchenna Paul Prof. Calvin Bartholomew Prof. Dennis Tolley Design of Kinetic Experiments for Fischer-Tropsch Synthesis on Supported Fe Catalysts Chemical Engineering and Statistics Brigham Young University Provo, Utah

Introduction Langmuir-Hinshelwood models derived from mechanisms are generally found to fit rate data well for a number of catalytic reactions, e.g., for Fischer-Tropsch synthesis: This model is nonlinear and, as a result, there is typically a high correlation between kinetic parameters.

Challenges in Collecting/Fitting Rate Data Collecting enough data to regress the model parameters can be time consuming. Without an appropriate experimental plan, parameter estimates may be poor; parameters may be highly correlated. Due to the nonlinear nature of the model, the best experimental design is not apparent.

Sequential/D-optimal Experimental Design Can Be Very Helpful

a form of response surface design – using optimization methods (Other forms include: A-Optimal, E-Optimal, G- Optimal, and V-Optimal). a proven tool for obtaining the most precise estimates of model parameters in the least number of experiments. enables selection of conditions that minimize the overall variances of the estimated parameters by spreading out design variables over available variable space. reduces the volume of the confidence region for estimated parameters. substantially reduces correlation among parameters. D-Optimal Design (DOD)

A rate function is specified, y i = f(x i,  ) where x i are the set of design parameter inputs and  is the set of kinetic coefficients. Calculus and matrix algebra are used to maximize the following determinant: where F is the Jacobian matrix and F T is the transpose of the Jacobian matrix, where ….. How Does DOD Work?

The Jacobian the Jacobian of the rate function r FT is: where f N,p is the set of partial derivatives of r FT with respect to the p th parameter at the N th set of experimental conditions; the N+1 set is the new experimental conditions. For example, f 1,1 = ∂r FT /∂A where A = the preexponential factor

Sequential Design using DOD

Response surface (i.e., value of the determinant D as a function of P H2 and P CO indices) for D-optimal design of rate expression for C 2+ hydrocarbons

Sequential Design Summary 1.Obtain initial estimates of parameters 2.Determine process condition that maximize D, i.e., minimize | F T F| -1/2 3.Run experiment at calculated optimal conditions 4.Nonlinear regression to estimate parameters 5.Repeat until |F T F| -1/2 reaches asymptotic value

Thus, statistical methods provide a map of the experiments, while optimization serves as a compass.

Overall Research Approach Microkinetic Model Adsorption/Desorption TPD/TPH Heats, Coverages DFT Electronic structure of stable species, intermediates and transition states Detailed Kinetics Activity, Selectivity, Stability XPS, XRD, Mössbauer Alloy formation, oxidation states, surface composition IR Surface species Microscopy Surface morphology and composition

Collaboration with Manos Mavrikakis and Jim Dumesic Objective: develop data for validation of microkinetic and LH models More than a dozen previous kinetic studies Most did not meet basic criteria of Ribeiro et al. (1997) for absence of heat/mass transfer effects, deactivation, etc. None used optimal statistical design of experiments. Data were fitted to power law and Eley-Rideal expressions mostly covering narrow ranges of operating conditions. Few reported TORs, thus preventing valid comparisons. Thus, much of previous work is unreliable or unusable FTS Reaction Kinetics on Fe

Derive LH and ER rate forms from a logical mechanism. Use D-optimal/sequential design to optimize experimental conditions, minimize errors in rate parameters, and minimize number of experiments Collect intrinsic rate data on a stable Fe-Pt/Al 2 O 3 catalyst in a Berty CSTR reactor over a wide range of commercially relevant conditions. Pt-promoter and La-stabilized alumina support facilitate Fe reduction and hydrothermal stability. Catalyst washcoated on monolith ensures high effectiveness, enabling operation over wide range of temperature. Use nonlinear regression methods to fit rate data to best mechanisms. Our Approach to Kinetic Study

Application of DOD to FTS on Fe 1. Select a reasonable mechanism.

(Application of DOD to FTS on Fe) 2. Derive LH rate expression from reasonable mechanism. 3.Choose independent variables: temperature, P CO, and P H 2 and model parameters: A, E act, A ads, and  H ads.

(Application of DOD to FTS on Fe) 4. Conduct scoping runs to obtain preliminary values of model parameters. Data are correlated well by the model. Catalyst is quite stable over > 150 h Run 11

(Application of DOD to FTS on Fe) 5.Set up Jacobian matrix with scoping runs and maximize determinant to obtain response surface for experimental parameters (i.e., P CO, and P H 2 at a specified T) for the next set of experiments. P CO P H2 Steep gradient and maximum for D (snow-capped peak) is observed around P CO = 0.75 and P H2 = 10. Our next experiment should be in that region. Run 5

Values of |FTF|-1/2 with respect to the number of runs at 239°C.

Values of k and K with respect to the number of runs at 239°C. k K

Experimental rates versus model predicted rates for sequentially designed experiments at 239°C.

Joint 95% likelihood confidence regions for k, and K at 239°C at different stages of the sequential design procedure

Results of 37 runs at 3 temperatures A Atm 1.5 -mol/g-min E act kJ/mol A ads atm x 10 3  H ads kJ/mol Parameter Estimate Lower 95% Confidence Level Upper 95% Confidence Level Variation in parameters less than 5-10%

Conclusions A stable, well-dispersed 15% FePt/Al 2 O 3 -La 2 O 3 wash- coated monolith catalyst in combination with a CSTR facilitates obtaining intrinsic FTS rates under commer- cially-relevant conditions. An LH rate expression based on C+H and OH + OH as RDSs provides the best fit to the data. A sequential design procedure using DOD resulted in precise parameter estimates in a minimal number of (10- 15) experiments at each of two temperatures. Three data sets at three temperatures (37 total runs) could be combined to obtain a rate law fitting C 2+ production rate data well over a wide range of T and partial pressures of CO and H 2.

Acknowledgments Collaboration with Professors James Dumesic and Manos Mavrikakis of U. Wisconsin Funding from DOE/NETL

Brian Critchfield and Uchenna Paul

Professor James A. Dumesic ACS Somorjai Award Recipient Friend, colleague, and collaborator for 34 years Pioneer and leader in catalysis research Bright, whimsical, youthful, creative, and modest Congratulations!