February 10, 2023 News PiMSlider, Artificial Intelligence and Machine Learning As the amount of data collected during pipeline inspections continues to grow new techniques are being explored to increase the volume of data that is analysed and the speed, accuracy and precision of the analysis. Many companies are turning to Artificial Intelligence and Machine Learning. Using these a huge volume of data from an entire pipeline can be assessed in minutes rather than the days or weeks taken for a more traditional analysis. Improved accuracy and precision leads to increased confidence in the analysis and better threat detection. ATP are probably one of the first companies to have implemented a machine learning approach in their software. In 2005, together with Gasunie, GET and Andrew Francis & Associates Ltd, ATP developed the PiMSlider module, Direct Assessment Expert. Direct Assessment Expert allows engineers to assess the condition of pipelines where inline inspection techniques cannot be applied. The module makes use of Artificial Intelligence principles and Bayesian statistical machine learning to find locations of corrosion or coating defects with high probability. A preliminary hypothesis based on limited historical information (preassessment) is built and the certainty of further hypotheses is improved when new data becomes available. Direct Assessment Expert can calculate the probability of failure and can predict where to find new defects with expected certainty. The more data in the system – the better the certainty of discovering a particular defect in a particular place. ATP has also developed a mobile application, Magellan, for automating the repair/excavation process within the boundaries of the PiMSlider ecosystem. Amongst other functionality Magellan uses Machine Learning techniques to recognise the type and size of defects from photos taken using its host mobile device in the field. Our customers find it especially useful for crack recognition and have provided ATP with hundreds of photos of thousands of real defects to train the Machine Learning algorithms. ATP has all the necessary expertise for Artificial Intelligence and Machine Learning development. Crucially, ATP also has – data. In this respect, ATP are in a unique position. Currently more than 300,000 kilometres of pipelines around the world are under PiMSlider supervision. This means that all information required for analysis is already in our unified database structure. To train the machine learning corrosion prediction models sufficient data is required for soil properties, atmospheric conditions, AC interference (including proximity to high voltage powerlines and proximity to power substations), proximity to roadways, railways, water and other pipelines, pipeline features (including years in service and manufacturer), rectifier amperage etc. ATP has all this data in our databases – connected, cleaned and aligned to pipelines. With a total exposure of around 10 million km*year. ATP also has around 500 million defects in databases and back-to-back defect comparison results over many years. This allows us to test our own, or any third party, corrosion growth or prediction models and compare model results with real corrosion growth for any given number of years and defect types. PiMSlider – is a platform for all aspects of pipeline integrity management, including Artificial Intelligence and Machine Learning. For more information or to discuss PiMSlider’s AI and Machine Learning capabilities please contact us today. Back to News & Events