Parameter Estimation for a system of ODE using patternsearch and surrogateopt

Hello there.

I am currently trying to estimate the parameters (28 in total) of my kinetic model (based on LHHW reaction kinetics) using a system of ODEs (using ODE 15s as the solver) and a solver from the global optimisation toolbox (GA, surrogateopt and patternsearch). I am trying different global optimisation solvers to see which one helps me get the lowest objective funtion value. However, GA doesnot proceed even to calculate the objective function at the initial point (does not throw error or a point on the gaplotbestf plot) and the fval in the patternsearch do not change over iterations and over several runs with different x0 values. Out of these three solvers, surrogateopt shows a decrease in the objective function value but it also get stuck after 1000 iterations (doesnot proceed further in terms of iterations and doesnot terminate with an exitflag as well). I have tried everything I know to help me lower the objective function value but all in vain. Can you please help me to solve this?

I appreciate any suggestions in this matter and thank you in advance.

Additional Information: I am attaching my script files to this post. A brief description of each of the script files are given below:

(1) Reaction Rate files (rate_CO2_form10.m,rate_NO2_form10.m and rate_N_form10.m) : These are the script files which contain the reaction rate equations. I am dealing with three reactions and therefore, I have three different reaction rate expressions for different reactions. The reaction rates are a function of reaction temperature T and partial pressures of each of the reactant/product components. The unknown parameters in these equations are the kinetic/adsorption/ desorption parameters. The range of these parameters are not very clear since I am using a novel catalytic material and not many people have studied it before in the literature.

(2) ODE solver file (pfr_model_v3.m) : Contains the system of ODEs that will be solved. The ODEs are related to the molar flow rate of different reactaant and product content in the reactor which varies with catalyst weight (dFdW).

(3) Objective Function File (objfun_pfr_v3.m) : This script files calls the pfr_model_v3.3 file for ODE calculations and calculates the molar conversions of reactants. Further, the funtion returns the sum of the square of difference between the calculated and experimental conversion values.

(4) Solver file (Optimise.m) : This script is the main script that needs to be executed. It retrives the experimental data and upper and lower bounds from their respective Excel files. This script also calls the solver for the minimisation of the objetive function.

(5) Nonlinear Constraint Function (nonlcon_form10.m) : The objective of the nonlinear constraint function is to make the molar flow rates of the components positive.

[Madan Behera]

In addition to this post, I was not able to upload the experimental data due to daily limit. So, here is the excel file containing the experimental data uploaded on Google Drive : Google Drive Expt Data.

[Alan Weiss]

I do not see an attachment Optimise.m. Google does not give me access to your experimental data files.

Alan Weiss

MATLAB mathematical toolbox documentation

[Madan Behera]

Hello @Alan Weiss... Please find the Optimise.m file attached to this comment box.

Let me know if you have any suggestions. Thank you in advance.

Regards,

Madan