Protocol number:ENANMAN25-616 abstract code:581 on numerical deconvolution for data analysis of tracers experiments

Data analysis of a tracer experiment is performed for studying the plant behavior,impact of procedures changes and several other goals.In all these cases, the observed data is the convolution of an injection function,and the systems transfer function under study.Practical discrete time convolutions may be performed multiplying a matrix of the process transfer function with the input vector,but for deconvolution it is necessary to invert the matrix which is singular in a causal system.Another method for deconvolution is by means of Fourier Transforms.Actual reading are usually corrupted by noise and besides its transforms shows high low frequencies components and high frequency ones mainly due to additive noise.Subjective decisions as cut-off frequency should be taken as well.This paper proposes a deconvolution method based on parameters fitting of suitable models,where they exist,and estimation of values where analytical forms are not available.It is based on the global, non linear fitting of them,with a criteria of maximum likelihood