With the current improvements in the capabilities of mathematical simulation models, these models can be used for control applications, opening a new methodology for controlling glass furnaces for optimal glass quality. The method is based on three steps. From classical theories a set of basic models can be derived that roughly describes the relationship between furnace design, glass properties, and operation, with settings on the one hand and the glass quality on the other. These relations are simplified, but they form a good basis for designing control strategies. This type of model can be fine-tuned using a detailed mathematical simulation model. Existing nonlinearities can be built into these simplified models by employing adaptive neuro-fuzzy interference system techniques. The resulting controller is very transparent and robust.