Paint film quality predictive model development in an automotive paint shop

Laburpena

This doctoral thesis studies the requirements and proposes the necessary procedures for the development of a predictive model of product quality in an automotive paint shop. The requirements identified and the proposed procedures for the development of the predictive model have been divided into three parts, deployed sequentially. The first part, definition, is focused on the theoretical needs of the identification of the variables that will serve to train the predictive model. The second part, generation, is focused on the requirements and procedures that lead to the creation of a valid data set with the values of the variables defined in the previous step. In the third part, analysis, the requirements that said data must meet in order to train a model effectively are described. Then, the appropriate actions, based on advanced analytics, that are necessary for the development of the predictive solution are defined. Once the requirements were identified and the theoretical procedures were developed for the development of the predictive model, an application case was carried out in a real automotive paint shop. In this experiment, the limits of the theoretical proposal have been identified and methodologies based on the implementation of the state of the art applied to this particular case have been proposed so that they lead to surpassing said limits.