Early detection of machine and manufacturing defects is important to guarantee a short response time and avoid a lot of rework. Usually these defects are the result of a combination of several process parameters and cannot be identified with the 'classic' if-else programming paradigm.
Given data from the different processes and machines, a large scale optimization can also be performed to reduce costs and waste from an overall standpoint or from a machine perspective. Thus, machine parameters can be optimized with sufficient data. Usually, these parameters are very complex and have highly nonlinear characteristics. With this optimization, a smart conclusion for better parameters (e.g. temperature, voltage) can be chosen.
The high-impact cases where enough data is available to unleash the advanced pattern capture capabilities of deep learning can lead to incredible advances in production optimization.