Kirsch, H.H.Kirsch2022-03-082022-03-081993https://publica.fraunhofer.de/handle/publica/320753Typical industrial monitoring and diagnosis tasks often require to detect type and place of a fault, based on data originating from sensor readings suitably processed, as well as from operator observations. If thresholds do not suffice and exact mathematical models are missing or too complex to handle, a reasoning tool is needed, able to work with qualitative, circumsribing data and data describing only general tendencies. Bayesian Nets offer a clear and concise way to do so, based on safe mathematical grounds. This article describes along a combustion engine diagnosis example the construction, properties and usage of Bayesian Nets, while highlighting their benefits and not concealling their deficiencies.enBayesian netBayes'sches NetzDiagnosediagnosisprobabilistic reasoningprobabilistisches Schließen004conference paper