Automatic fault identification of process and sensors

Important results from the project

The project evaluated and validated fault detection methods to automatically identify process and sensor errors in industrial processes. A proposal for a flexible software framework which enables an iterative process for the development and evaluation of fault detection methods in full-scale was developed. The project also investigated what technology is available on the market today.

Expected long term effects

This feasibility study has provided the possibility to develop this concept into an implementable system to identify process abnormalities on a short and long time scale as well as sensor faults and errors. It has also given insight in where further development should be performed and what is available today. In the future, these results can contribute to a reduction of unforeseen production stoppages that cause substantial costs in the Swedish process industry due to loss of production and expensive repairs. This will also have impact on the environment and on the local community.

Approach and implementation

Data from the production process was compiled and analyzed together with the industrial partners. Existing and new fault detection methods were tested and updated based on these data. In parallel, the proposal of the software framework was developed. The results from the project have been disseminated internally to the participating partners, as well as through participation from the project at PiiA Summit 2016.