Prediction uncertainty of new product development based on the T-PLS model

The key element in designing a new product is to find out the process conditions (including raw materials and operating conditions) that can get a specified output. This study develops a method to design new products based on the T-PLS model. The T-PLS model divides the latent variable space of the PLS model into two subspaces, which are output-correlated subspace and output-uncorrelated subspace. Output-correlated subspace serves as a basis for product design, and output-uncorrelated subspace provides freedom for product design. Meanwhile, this paper considers the influence of prediction uncertainty on the model, and provides the operator with design space. The effectiveness of the method is verified in the numerical simulation and the simulation of penicillin fermentation.