Abordagem bayesiana da isoterma de langmuir

Abstract

The trace metals are highly toxic contaminants particularly in the aquatic environment. The adsorption process is an efficient technique for removing these trace metals. There is a generation of contaminated materials in studies laboratory related to adsorption, it this implies social and environmental risks and additional costs associated with treatment and proper disposal of toxic waste. This fact brings a practical limitation on the amount of sample to be used. The adsorption process can be described by functions called adsorption isotherms. The Langmuir isotherm is one of the most used. This isotherm shows the parameter k related to the energy of adsorption, and the M parameter which indicates the maximum capacity metal adsorption. The parameter estimation is generally based on asymptotic methods therefore for small sample sizes estimates tend to have low precision and accuracy. Thus, there is a difficulty in obtaining precise and accurate estimates at the same time be used for small sample size. Aiming to overcome this problem has been studied in this work, the application of Bayesian inference in fitting isotherm Langmuir considering informative and uninformative priors. Then carried out a study of data simulation with sample sizes of 5, 6, 7, 9, 12, 15, 20, 25 and 30, they were discussed in two scenarios considering different values of M and k. For the evaluation of the precision was used the mean square error and the length of a 95% HPD credible intervals, for to evaluate the accuracy was used the mean bias relative absolute. The samples of the marginal posterior distributions the parameters were obtained by the Gibbs sampler. The inferences were made. The results indicated that the estimates obtained with use of informative priors presented greater precision and accuracy even at lower sample sizes. Subsequently the isotherm was adjusted on experimental data of adsorption lead (Pb2+) in orange peel considering the priors proposals and the results corroborate whith the results obtained in the simulation study. Thus, the methodology bayesian was efficient for the estimation of isotherm parameters of Langmuir and the use of informative priors, it allowed bypass practical limitation on the amount of sample available in real adsorption studies.