An on-line process monitoring scheme to improve accuracy of the tube boiler inspection process

Abstract

Tube boilers are deployed in heavy industry plants. These tube boilers suffer from defects caused by corrosion. Current maintenance works can only be carried out after shutting down the plants, increasing time-consuming and labor intensive. An on-line condition monitoring scheme for U-bend tubes is designed based on Fiber Bragg grating (FBG) sensors. As the data collected from sensors would be large volume in every second, we will store data in the distributed computing framework. The task is considered as an on-line learning task, where we can identify and investigate crucial failure causes and positions. With collection of big data from sensors, a predictive model to monitor tube boilers is developed. The thickness of U-bend tubes will be predicted to track the health status of tubes, leading to more accurate and intelligent maintenance. When an alarm from the model is reported, we can recommend preventive maintenance. The online-monitoring scheme can detect defects and increase the speed of inspection process for defect localization. Cost-benefit analysis is conducted for the current maintenance policy adopted by Sembcorp maintenance team and possible maintenance policies leveraging our monitoring scheme.