Due to the increasing complexity of technical systems, accurate fault identification is crucial in order to reduce maintenance costs and system downtime. Model-based diagnosis has been proposed as an approach to improve fault localization. By utilizing a system model, possible causes, i.e. defects, for observable anomalies can be computed. Even though model-based diagnosis rests on solid theoretical background, it has not been widely adopted in practice. The reasons are twofold: on the one hand it requires an initial modeling effort and on the other hand a high computational complexity is associated with the diagnosis task in general. In this paper we address these issues by proposing a process for abductive model-based diagnosis in an industrial setting. Suitable models are created automatically from failure assessments available. Further, the compiled system descriptions reside within a tractable space of abductive diagnosis. In order to convey the feasibility of the approach we present results of an empirical evaluation based on several failure assessments.