A new instrumentation optimization algorithm is described which combines the iterative nature of simplex optimization with the global aspect of response surface fitting. The model assumes smooth, second order continous functions which are, in most cases, sufficient to characterize instrumental response with changes in parameter values. In contrast to simplex optimization, in which poorest response data values are discarded and replaced by subsequent better fit inputs, optiplex optimization retains both worst and best response data at all times in order to characterize a response surface as completely as possible within the data set used. Once an optimum location has been found, the optiplex algorithm searches the surface in the vicinity of the maximum, in parameter space, to ascertain whether the maximum response found is local or global. This process is iterated and new points taken until a global maximum has been determined. Three parameter optiplex optimization is presented for computer generated test data with and without randomly superimposed noise; atomic absorption (AAS): observation height, sample uptake rate, and fuel flow; and inductively coupled plasma (ICP): observation height, rf power, and nebulizer pressure. Gross signal, absorbance and net signal were the optimization criteria for the computer generated, AAS and ICP data, respectively.
Read full abstract