BASILISK – Bayesian model migration for effort-reduced model building in measurement uncertainty determination with initial model building using artificial neural networks

• Applicant: Professor Dr.-Ing. Robert Schmitt
• Subject Area: Measurement Systems
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 542649823

Since every measurement process is inherently subject to uncertainty, the further use of measurement data subject to uncertainty means that derived statements are also uncertain. This can lead to erroneous decisions. If the measurement uncertainty when inspecting a produced feature is too high and the feature is close to the specification limits, the decision as to whether the feature is within or outside the specification is risky. As a result, parts that are within specification may be rejected during inspection (alpha-failure) or vice versa (ß-failure). In the second case, this means that defective parts are passed on to the customer, where they may cause damage for which the producer is responsible. Both cases, depending on the production and defect costs, can lead to a high economic loss. Since the risk of wrong decisions cannot be determined and thus controlled without knowledge of the measurement uncertainty when collecting measurement data, measurement values without measurement uncertainty information are worthless. Previous methods for determining the measurement uncertainty use a model of the measurement as the basis for specifying the measurement uncertainty. With this model, the measurement uncertainty is determined based on the natural fluctuations of the input variables. The determination of the model of the measurement is often connected with a high expenditure. Existing procedures for the determination of the model of the measurement with the help of procedures from the area of the artificial intelligence offer the potential for the reduction of the expenditure, do not include however previous knowledge in the form of an estimate of the similarity to other measuring processes, for which if necessary already a model of the measurement exists. In order to reduce the effort in the determination of the measurement uncertainty, it is therefore necessary to develop a procedure which uses existing models of the measurement, taking into account an estimate of the effort, in order to enable low-effort model building for new, similar measurement processes.

• DFG Programme: Research Grants