Ruosen Qi
Developing improved models for fault prognosis in industrial systems.
Email: [email protected]
Supervisor
Project description
In recent years, there has been a rapid development of modern science and industrial technology. Industrial systems are becoming more and more complicated.
There is continuous improvement of the complexity, integration, and intelligence of industrial systems. The cost of their development and production, especially maintenance and protection, is increasing. Due to the increase of components and influencing factors, the probability of failure and functional failures also increases. Thus, the prognosis and maintenance of industrial systems faults have become a focus of attention.
Existing methods are very effective in dealing with sensor faults where the fault direction is easy to determine. But implementing reconstruction methods for process faults is quite challenging. The fault direction vectors are usually difficult to specify. Process faults usually affect many process variables to various extents. This research introduces a principal component analysis (PCA) based fault reconstruction method. We use PCA to analyse historical process data with faults to extract fault directions. These are then used for fault reconstruction.
The reconstructed fault magnitudes can be used to develop data-driven fault prognosis models. We are developing both linear autoregressive models and extreme learning machine (ELM) models for fault prognosis.
Publications
- Qi R, Zhang J. Process Fault Detection and Reconstruction by Principal Component Analysis. In: 2019 24th International Conference on Methods and Models in Automation and Robotics (MMAR), 2019. IEEE.
Interests
Fault prognosis; prognostic and health management