A general framework for latent variable model inversion to support product and process design

In this paper a general framework to perform the inversion of latent variable regression models (LVRMs) is proposed. The framework exploits the advantages of LVRMs in modeling the driving forces between databases of developed products and the raw materials/process conditions used. These relations are used to develop new products, by estimating the best combinations of input variables to obtain a desired product in output. The procedure can deal with several different constraints both in the predictor and in the quality spaces. A wet granulation particle design problem is used to illustrate the benefits of the proposed framework.