Abstract: A glass blank is sent to a heating zone where it is softened region by region and drawn continuously in a controlled manner as a drawing bulb is formed. Estimates of at least one controlled variable are predicted free of dead time; in addition, at least one geometric variable of the component which can be correlated with the controlled variable is measured continuously, and the measurement values thus obtained are used to adjust the predicted estimates. On the basis of the predicted estimates thus adjusted, the nominal-actual deviation of the controlled variable is determined and converted to a change in a controlling variable. To ensure the production of a component with especially high dimensional accuracy, a geometric variable is measured at a first measurement site in the area of the drawing bulb for the prediction of the estimate, and the temperature of the drawing bulb be measured and used as a controlled variable.