Fitting multiple Datasets with shared and individual Parameters.

Question

Fabian Bienen 2018-2-24

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/384582-fitting-multiple-datasets-with-shared-and-individual-parameters

回答： Alex Sha 2019-4-28

Hello Matlab Community,

I wan't to fit Multiple Datasets to one Modellequation. Some Parameters are shared and should be fitted to all datasets, while there are some parameters which should be fitted individual for each dataset. I think i found a code in this board which can help me do that, but i do not understand how my raw data has to be supplied. In detail, how should my .xls-File look like? My xdata will be the Time and the y data will be the Mass. Here is the code:

    % function sharedparams
 ds = dataset('xlsfile', 'Test.xlsx') ; 
 ds.dataset = nominal(ds.dataset);
 % Initial estimates for [N, D, M_C11, M_C12, M_C21 and M_C22] 
 p0 = [120 , .01 , 100, 100, 100, 100];
 % Estimate parameters 
 fn = @(p , f ) objFcn(p , f , ds.dataset) ;
 % Function handle
 lb = [100 , 0 , 0, 0, 0, 0]; ub = [180 , 0.3 , 300, 300, 300, 300]; options = optimoptions('lsqcurvefit');      options.TolFun = 1e-10;
 pFit = lsqcurvefit(fn , p0 , ds.Frequency , ds.Magnitude,lb, ub, options);
 assignin('base','pFit',pFit);
 end
 function yfit = objFcn(p , f, dataset)
 N = p(1) ; % location-independent parameter 1 
 D = p(2) ; % location-independent parameter 2
 M1 = p(3) ; % Location dependent parameter (C11) 
 M2 = p(4) ; % Location dependent parameter (C12) 
 M3 = p(5) ; % Location dependent parameter (C21) 
 M4 = p(6) ; % Location dependent parameter (C22)
 f1 = 2*pi*f(dataset == 'A'); 
 yfit_C11 = (0.0035*0.027*f1.^2*N^2)./(M1*sqrt((N^2-f1.^2).^2+(2*D*N*f1).^2));
 f2 = 2*pi*f(dataset == 'B'); 
 yfit_C12 = (0.0035*0.027*f2.^2*N^2)./(M2*sqrt((N^2-f2.^2).^2+(2*D*N*f2).^2));
 f3 = 2*pi*f(dataset == 'C'); 
 yfit_C21 = (0.0035*0.027*f3.^2*N^2)./(M3*sqrt((N^2-f3.^2).^2+(2*D*N*f3).^2));
 f4 = 2*pi*f(dataset == 'D'); 
 yfit_C22 = (0.0035*0.027*f4.^2*N^2)./(M4*sqrt((N^2-f4.^2).^2+(2*D*N*f4).^2));
 % Combine prediction 
 yfit = [yfit_C11; yfit_C12; yfit_C21; yfit_C22] ;
 end