ANÁLISIS Y EVALUACIÓN DE PROCESOS POR LOTES APLICANDO LA ESTADÍSTICA T2 DE HOTELLING

Manufacturing processes can be classified in two groups: continuous and batch production processes. The statistical techniques used to control continuous production processes are not applicable to batch production processes because the latter present an additional dimension to deal with: the batch size. Trying to control the variability of a batch production process, the total variability of the observations was divided in two independent sources: one to explain the variability between measures at different times, that is, between batches, and the other to explain the variability between measures at the same time, that is, within batches. Likewise, when working with batch production processes it is possible that each batch had a different mean vector. This situation affects the control setting of the process, so it is necessary to know if the data have a multivariate normal distribution with common mean vector and variance-covariance matrix (category 1) or if they have a multivariate normal distribution with different mean vectors and common variance covariance matrix (category 2).To implement this methodology was necessary to use the modified Hotelling T2 statistics and to simplify the calculation of statistics had to develop a free software program R. This way it could reach a proper analysis of the variability of a batch process. The implementation of the proposed methodology was carried out in the company “Seafood” specialized in the export of bivalve molluscs.