Transferable belief model in fault diagnosis

Abstract The Transferable Belief Model (TBM) is an approximate reasoning approach which is derived from the Dempster–Shafer mathematical theory of evidence. The key property of TBM is its ability to treat inconsistency in data by a novel concept of assigning belief, referred to as the ‘open-world’ assumption. The aim of this paper is to apply TBM to a diagnostic framework, and to provide a detailed study of its most important features. The TBM approach is compared with fuzzy logic and conventional Boolean technique. It is further shown that the ‘open-world’ assumption can be easily incorporated into the fuzzy logic context, resulting in comparable diagnostic outputs, even in cases of measurement noise and modelling errors. In addition, these novel reasoning approaches provide a measure of confidence in the underlying diagnostic results, which contributes a new quality to the diagnostic practice. Finally, the underlying methods are applied to a DC motor test rig. Practical results clearly show how TBM, along with its fuzzy counterpart, outperforms Boolean reasoning.