Manufacturing start-up problem solved by mixed-integer quadratic programming and multivariate statistical modelling

We consider a batch process characterized by multiple, correlated process and product variables. The focus is on the identification of optimal process settings in the start-up period. Mathematical programming and multivariate statistical modelling are combined to solve the adjustment problem for the batch start-up. Partial least squares (PLS), a multivariate statistical approach, is used to model the relationship between process and product variables under good operating conditions. The optimal adjustment is identified by solving a mixed-integer quadratic program (MIQP) such that the recommended process settings are consistent with the PLS model. The start-up adjustment algorithm is operator-assisted; i.e. it uses input from the operator to compensate for unavailable but important process and product information. The operator's input is modelled as a constraint of the MIQP, and good production practices are also taken into account. The proposed adjustment algorithm is applied to the start-up period of a filament extrusion batch process and retrospective data indicate the effectiveness of the algorithm.