Project Area Description
In this section we present the projects developed by ARAMIS for the Manufacturing industry.
One of the most important objective to achieve in the manufacturing industry is the availability of the production plant. Therefore, to avoid plant unavailability and the related losses of revenues, it is necessary to online monitoring the industrial machine for accurately assessing its degradation state.
To reach this goal, Aramis has considered the vibration signals collected on the machine, developing a statistical approach for the automatic identification of the signals related to the machine degradation and a machine learning approach for the automatic classification of the machine degradation state.
The developed hybrid approach (statistical and machine learning) provides accurate classification of the machine degradation state, allowing avoiding unexpected machine failures which might lead to the unavailability of the plant and to significant economic losses.
Developing the physical model of the degradation of industrial components and performing experimental tests ca be very difficult, or even impossible due to the required costs and knowledge of the process.
These situations are typically encountered for the safety-critical and high-value components which are characterized by very high reliability, unique or new designed material composition, and for which performing run-to-fail test is too expensive or not feasible.
To overcome this problem, Aramis has developed a novel methodology to identify component degradation state when only few measurements of a physical quantity indirectly related to the component degradation are available at the beginning and the end of the component life, whereas no information is available on the component degradation state during its operation life.
The developed diagnostic tool allows accurately assessing the component degradation state avoiding unnecessary maintenance interventions and component replacements.
The availability of only small amounts of field-data significantly complicates the accurate prediction of the failure time and of the remaining useful life of industrial equipment.
To overcome in this problem, Aramis has developed a locally weighted ensemble of different predictive models: the outcomes of each individual model are aggregated in an adaptive way, allowing obtaining for good predictive performance throughout the degradation progression.
The developed adaptive ensemble accurately monitors the degradation evolution of the component, providing more accurate failure predictions and allowing the optimal scheduling of maintenance intervention.