Scalable diagnosis and health assessment

Scalable diagnosis and health assessment


We developed a methodology to generate accurate diagnosis and prognosis systems efficiently, thus helping operators of high-tech systems to achieve maximum system availability with minimal cost.

The methodology delivers system health assessments based on probabilistic reasoning on an industrial scale, using engineering as well as physical models plus operational data as key components. The figure below depicts how these parts are combined within three sections: the building of the reasoning (left), the physical models that ensure data quality (upper right) and the operation of the health assessment (lower right).

Building a reasoning engine for diagnostics

The probabilistic reasoning models at the core of the methodology are built within an object-oriented approach based on expert knowledge and the specifications of the system.

First, we compile a library of sub-models for each type of component in the complex system, ensuring re-use and thus reduced efforts. Then, we assemble the sub-models into the complete probabilistic reasoning model by matching the system’s specifications and schematics using generative techniques.

This structured procedure ensures scalability and maintainability of the reasoning model, together with limiting the modelling efforts for the experts. It is domain independent and applicable to a variety of complex industrial systems.


We use operational data to fine-tune our reasoning models and to drive the diagnosis. Coming from electromechanical systems in highly complex environments and settings, such data is often incomplete and prone to noise. We therefore clean and complement the data based on physical models. Such models also explain the degradation and subsequent remaining useful life of key system components, greatly improving the diagnosis and prognosis power of our reasoning models.


The resulting reasoning model produces a novel system health assessment. Together with visualizations that clarify the inferred probabilistic predictions, it lifts condition monitoring capabilities up towards the system level and provides for mission readiness investigations. Finally, a stepwise regression that predicts the long-term developments of key system’s components enables prognosis and assists operators to plan maintenance in advance.

Additional information