Simulated Annealing Fitting of Chemical Analysis Data

New methods for the fitting of experimental analytical chemical data are required for quantitative mixture analysis. Frequently, the measured properties of a multi-component system are very convoluted and identification of the systems components from the pure component measurements is non-trivial. Many current chemometric methods of data fitting are not robust enough to handle complex data. Problems with these methods include susceptibility to local minima trapping, not being applicable to nonlinear systems, and inability to constrain the produced results to a realistic range and not being able to handle data that falls outside the domain of previous data. Thus, the global best fit for the system if often not found. The proposed research overcomes all these problems through the use of a global search technique known as simulated annealing. This technique has found wide use in optimization problems in many other fields but has seldomly been applied to data fitting problems.