Description
For the past several decades, there has been little change in the way we conduct maintenance for ventilation systems of underground spaces such as metro systems. Typically, a tunnel ventilation system (TVS) maintenance procedure involves set dates at which select components are checked or tested. Even though it is a life safety system, maintenance of tunnel ventilation fans (TVF) is often neglected; particularly in systems where it is only used during emergency operations.
With the advancement of technology and the increasing use of artificial intelligence, could we not implement new methods using existing technologies to change the narrative? In temperate climates where the TVS may never be used to its full capacity whilst it is present on site, there may still be a lot of ‘life’ left in the system beyond its design life. By switching to alternative means of maintenance and component monitoring, there is the potential to steer away from human-based time discrete maintenance to something more proactive such as condition-based maintenance. This is when maintenance is only conducted when the TVF requires it. Is there not a better way to determine the prognosis of a TVF rather than just replace it when it completes its prescribed design life?
This paper sets out to find solutions to this question. By conducting multiple sensor data fusion using extended Kalman filters and autonomous damage detection using thresholding, this study discusses a maintenance methodology that uses readily available sensors and computational power to perform remote damage detection of a TVF. The methodology is universal in its application to new and ageing systems.
To demonstrate the validity of the proposed methodology, the paper explores a case study scenario where a theoretical multiple degree of freedom is set up to represent rotor excited vibrating components in a TVF during its operation. Synthetic faults are introduced into the system to test the robustness of the methodology in its ability to handle the introduction of non-linearities within the system caused by component degradation. The findings suggest that through the implementation of multiple sensor data fusion for structural health monitoring (SHM), constant checking of a TVF is possible thus allowing the design life to be extended with confidence of autonomous early damage detection. A cost benefit study has also been conducted to show the potential commercial merit to the client.
