Pipelines transport liquid and gas through a system of pipes, typically across long distances, supporting our modern living standard. Over time, pipelines are subject to a variety of integrity threats, from corrosion, to dents, to gouging. In order to identify and understand these threats, In Line Inspection (ILI) tools also known as ‘pigs’ are widely-used to scan the pipe’s interior, measuring and recording irregularities including corrosion, cracks, deformations, or other defects. The amount of ILI data is however immense and dealing with this data from a fitness-for-service point of view poses a significant challenge to the industry, resulting in inaccuracies and uncertainties of the depth, size and location of the pipe wall. Similarly, defects are often identified, yet classified generically, which can be problematic, leading to incorrect dig and repair prioritisations.
This Challenge, set in conjunction with Phillips 66, the US energy manufacturing and logistics company, sought innovative IoT, machine learning and AI solutions to more effectively identify anomalies from the immense ILI data sets. The solution should have been capable of processing both raw ILI and historical data, to identify pipeline integrity threats with higher accuracy, reducing the risk of pipeline leaks and major incidents.
Phillips 66 selected Conundrum as the winner for this challenge. The startup provides an Industrial AI platform with a defined set of applications to address the critical industrial challenges: Predictive Maintenance, Quality control, Quality Optimization and Anomaly Detection. Their AI-driven technologies and software are based on deep learning and automated machine learning for industrial time series.