PLS modelling of detergency performance for some technical nonionic surfactants

The detergency effect of a series of technical nonionic surfactants has been examined. The detergency experiments have been performed according to different central composite circumscribed statistical designs. The data from the differently designed investigations were combined, and a response surface model was estimated for each surfactant with the use of partial least squares of latent structures. The combination of the designed series gave a new insight into the importance of a variable not considered before. The effects of different transformations and model complexities were investigated and a new approach to compare the efficiency of the different surfactants was found. In the evaluation of the individual designs all replicates were used in the calculations, while the evaluation of the combined designs was made with the use of the mean values of each experimental point. The reason for this is presented.