Existing optical fibers turned into distributed acoustic sensors alongside pipelines' is no news. Pipelines run through many public areas, and there is a constant chance of false alarms with regular agricultural and transportation activity. The aim is to reduce the rate of these false alarms towards zero for better management and control of pipeline assets.
This paper will discuss using novel algorithmic methodologies and artificial intelligence models to analyze the outstanding amount of collected data in parallel, fusing the results to create a significant decrease in false alert rate toward zero. This next-generation, real-time monitoring makes it a robust tool for better, greener, and digitalized operations. It will detail the fusion process and the real-world test results showcasing the process.
Machine learning mechanisms filter out normal behavior with increasing speed and assurance, training on over 1 petabyte of data collected from gas and liquid pipelines deployed in various set-ups, terrains, areas, and environmental conditions. In contrast, digital signal processing algorithms respond to and recognize different events. The goal is to ensure activities such as nearby railways and highways are not triggering alarms while not masking abnormal events such as pipeline leaks. Both models evolve and improve through countless trials, simulated, and real-life scenarios. The innovative fusion between them decreases false alarm rates to an unprecedented low.
Ultimately, this novel approach toward using AI Applications in large-scale pipeline monitoring allows operators ease of mind, knowing there is a high detection rate without the nuisance of false alarms.