A fundamental problem of machine learning (ML) is measurement uncertainty and the influence on ML results. Measurement uncertainty, which is critical in hazardous gas detection, is directly addressed in this paper. A previously published toolbox is extended for regression. One of the benefits of this approach is obtaining a better understanding of where the overall system should be improved. This can be achieved by either improving the trained ML model or using a sensor with higher precision.

Industrial X-ray computed tomography is a holistic measurement technique for dimensional metrology. However, the relatively long measurement time is a hindrance to its application. A possibility to reduce measurement times, the continuous scan mode, is characterized in this work. The question is how much the time reduction impacts the accuracy of the dimensional measurements. The core result of the paper is an estimate of the effect along with experimental proof that this estimate is reasonable.

The paper proposes a structure for the creation of new simulation software for uncertainty estimation. The structure was derived from the well-established VCMM. To make it easy to apply the software structure to specific projects, a supporting software library (written in C++) was created. The library provides all the components necessary to create software for uncertainty estimation. The software structure and library proposed can be used in different domains of metrology.

A TIR100-2 emissometer is used to measure the emissivity of thermal insulation products using the low-emissivity effect to increase the thermal resistance, but the uncertainties of measurements on low-emissivity foils were not known. The study allowed the determination of the uncertainty through the detailed analysis of the measurement principle and the determination of the influences of the parameters. The measurement results were validated by comparison to a reference measurement setup.

Virtual assembly (VA) is a method for the quality prediction of assemblies considering local form deviations of relevant geometries. Point clouds of measured objects are registered in order to recreate the objects’ hypothetical physical assembly state, which is strongly influenced by the measurement uncertainty of individual points. Thus, we studied the propagation of uncertainties by VA. The results reveal larger propagated uncertainties by VA compared to the unconstrained Gaussian best fit.