From the recorded process data by the edge, digital models are generated using algorithms from artificial intelligence and machine learning. Conventional protocols from the measuring machines are used to train the models or to calculate the model parameters. In online operation, the model is used for the quality prognosis of the manufacturing tolerances, the automated cause-effect analysis, the real-time parameter adaptation for zero-defect production and the automated cause-effect analysis of productivity. The machine learning systems are capable of learning and adapt themselves better and better to the constantly changing production conditions over longer periods of time.