Empirical validation of meta-models of work centres in order release planning

Theoretical insights suggest that modeling the output of a work centre in a period as a function of one independent variable is not sufficient for this type of models because of the time-varying transient states. This paper tests one- and two-dimensional clearing functions on simulation and empirical data obtained from a make-to-order production system. The fit of the clearing functions will be made by a regression through the origin and evaluated by the adjusted R2. The fits of the different clearing functions differ depending on the source of data. The possibilities for improvements by adding independent variables depend both on the period length and on the stationarity of the process. The findings lead to suggestions which additional independent variables should be added to a clearing function in order to improve the estimation of a work centre's future output and hence improve order release planning models.