This article was originally published in October 2026 by Gas International, the official journal of the Institution of Gas Engineers and Managers, available here
Antonio Caraballo, Pipelines Discipline Manager at Costain, Dr Ollie Folayan, Costain’s Head of Process Engineering and Professor Nilay Shah from Imperial College London, examine why artificial intelligence and machine learning mark an inflection point in pipeline engineering, and how industry and academia can work together to capitalise on the opportunities these technologies present.
The integration of artificial intelligence (AI) and machine learning (ML) into the engineering and operations of energy pipelines represents one of the most significant technological shifts in the pipeline industry. Drawing on Costain’s end-to-end engineering expertise and joint analysis with Imperial College, this article explores how AI and ML are being used to improve pipeline systems and processes, and the challenges that remain.
Much of traditional engineering design, construction, operations and pipeline integrity management has relied heavily on periodic inspections, manual data analysis, and corrective maintenance strategies. However, the complexity of modern pipeline networks, combined with ageing pipeline infrastructure in the UK and heightened environmental concerns, demand a more sophisticated approach to their design, management and operation. AI and ML technologies offer unprecedented capabilities for designing, monitoring, predicting, and optimising pipeline engineering operations in real time. The technology is already being implemented on live projects and throughout system operations. It is set to fundamentally change how the pipeline industry approaches all aspects of its lifecycle, including, design, safety, integrity management, and efficiency.
Strategic implications and future directions
The adoption of AI/ML in pipeline engineering operations carries strategic business implications that extend beyond technical and operational performance.
Firstly, it represents a competitive differentiator in an industry facing pressure to improve both economic efficiency and environmental performance. Operators who successfully implement AI/ML systems can achieve lower operating costs, reduced emissions, and enhanced safety records – all critical factors in an increasingly regulated and scrutinised environment.
Secondly, AI/ML adoption necessitates workforce transformation. The industry must develop talent capable of bridging traditional pipeline engineering with data science and AI expertise. It is not about replacing human operators and engineers but augmenting their capabilities with intelligent decision support systems. In the UK, this augmentation will be accelerated by the government’s commitment to long-term major infrastructure projects through its 10-year infrastructure strategy and pipeline, as well as national and regional workforce plans led by the Construction Skills Mission Board. The most successful implementations will be those that integrate AI/ML tools into existing workflows while providing operators with the training and support needed to leverage these capabilities effectively.
Third, the technology enables new business models and operational paradigms. Digital twin frameworks, which create virtual replicas of physical pipeline systems, allow operators to simulate scenarios and optimise operations in a risk-free environment before implementing changes in the real world.
They are commonplace on major pipeline projects in the water sector. Anglian Water’s Strategic Pipeline Alliance, which is delivering hundreds of kilometres of new water pipeline, has applied hydraulic models to identify how the existing network and storage could be optimised to avoid the need to lay new pipes. In addition to supporting decarbonisation objectives through lower material use, this capability fundamentally changes how operators approach planning, maintenance scheduling, asset integrity management, safety, and emergency response. Several other areas of research warrant attention.
Advanced algorithms that can handle sparse data and transfer learning approaches that enable knowledge sharing across pipeline systems represent promising avenues for deploying critical carbon capture projects to decarbonise UK industrial clusters, such as the Northern Endurance Partnership’s onshore CO₂ gathering system. Additionally, integrating AI/ML with emerging technologies, such as robotics for autonomous inspection, could further enhance pipeline operations and integrity management.
Finally, as AI and ML become increasingly embedded in energy pipeline systems, their implementation must be aligned with robust ethical principles and transparent governance. Decisions impacting safety, risk management, integrity management, and environmental performance should remain subject to specialist human oversight. There should also be clear accountability for data quality, model uncertainties, bias and understanding of failure, supported by auditable governance frameworks that ensure innovation is delivered safely, responsibly and in the public interest. It must augment engineering judgement rather than replace it.
The core value proposition: from reactive to predictive
Pipeline engineering is a data-rich specialism, with traditional approaches typically detecting problems during or after they have occurred. The most compelling argument for AI/ML adoption in pipeline engineering lies in the paradigm shift from reactive to predictive management of these systems through the different stages of their lifecycle. AI and ML systems, when correctly implemented, can anticipate failures or system vulnerabilities before they materialise. This transformation is evident across almost all domains of pipeline engineering, including promising emerging applications in hydrogen and carbon capture pipelines.
There are multiple areas in which AI/ML can enhance current systems and processes:
Pipeline route selection and route optimisation
AI and ML are transforming pipeline route selection and optimisation across onshore and offshore environments. ML algorithms have been applied to identify minimum-cost routes through complex terrain and obstacles, as well as to integrate hydraulic constraints for a gas-liquid two-phase flow system.
Similarly, AI/ML algorithms combined with geographical information systems (GIS) have been used to support cost minimisation by evaluating key topographical, environmental, constructability, crossing, and natural hazard variables simultaneously. Despite these advances, challenges remain in ensuring that the technology helps projects meet regulatory compliance and in developing its application beyond the conceptual and pre-front-end engineering and design (FEED) stages.
Operational optimisation
ML algorithms excel at analysing complex systems to identify optimal operating parameters and augment human ability to evaluate multiple dynamic attributes. For instance, in 2024, a North American energy company demonstrated how ML-driven business intelligence applications increased daily natural gas throughput while reducing greenhouse gas emissions.
Successful examples can also be found in the UK. INEOS FPS, to whom Costain is providing engineering services to optimise operations at its Kinneil Terminal in Grangemouth, is using an emissions analysis AI platform to monitor energy use, minimise fuel consumption and reduce carbon emissions with the aim of lowering the terminal’s footprint. This dual benefit – enhanced efficiency and environmental performance – illustrates the strategic value of AI/ML for critical national infrastructure beyond traditional safety and integrity applications.
Building predictability into maintenance and integrity management
AI-powered predictive maintenance systems analyse historical data, sensor readings and operational patterns to forecast equipment failures before they occur. This capability transforms maintenance from a scheduled, time based activity to a condition-based, optimised process.
The economic implications are substantial: operators can maximise asset utilisation, reduce unplanned downtime, and extend equipment lifespan while simultaneously improving efficiency, safety and asset integrity management.
Techno-economic analysis of pipeline projects
Capital expenditure (CAPEX) and operational expenditure (OPEX) predictions constitute fundamental aspects of pipeline techno-economic analysis. AI/ML methods can be applied to develop predictive models that estimate cost based on design parameters, operational variables, and project-specific attributes.
These AI/ML methods can support CAPEX prediction by applying regression analysis models, OPEX forecasting, lifecycle cost (LCC) integration, demand forecasting, integrity-cost coupling, and automated financial appraisal. This breadth suggests the potential for integrated techno-economic decision-support systems that optimise across design, operations, and maintenance.
Remaining realistic on the pace of implementation
AI/ML is already maturing at pace thanks to proven use cases across all aspects of pipeline engineering. However, it is essential to maintain realistic expectations about current capabilities and implementation challenges.
Indeed, successful deployment requires more than sophisticated algorithms. This includes implementation of strategic business frameworks covering the following aspects:
Data quality as foundation
AI/ML systems are only as good as the data they consume. Pipeline operators must invest in comprehensive sensor networks, data integration platforms, talent development, and quality assurance processes. The challenge is particularly acute for mature operators with outdated legacy systems and/or fragmented data sources. Without clean, consistent, and comprehensive data, even the most advanced algorithms will underperform.
Integration with physical models
The most effective AI/ML systems do not operate in isolation but rather augment physics-based models with data-driven insights and understanding of physical constraints. This hybrid approach combines the interpretability and theoretical grounding of physical models with the pattern recognition capabilities of ML algorithms. Such integration is particularly valuable for complex phenomena such as multiphase flow in pipelines and impurity concentration in CO₂ fluid streams, where purely data-driven approaches may struggle with generalisation.
Explainability and trust
In safety critical applications like pipeline operations, black-box AI systems face legitimate scepticism from operators and regulators. The industry needs explainable AI approaches that provide transparent reasoning for their predictions and recommendations.
Building trust requires technical transparency and demonstrated reliability and repeatability across diverse operating conditions. But arguably most important is the role of human oversight in monitoring technological outputs. Asset owners will be looking to construction and delivery partners for reassurance and expertise, and ultimately those that become long-term, trusted partners on the implementation journey.
The opportunities remain transformative
AI and ML technologies represent a transformative opportunity for the energy pipeline industry. The industry stands at a juncture where early adopters are already demonstrating competitive advantages through improved safety, operational efficiency, and environmental performance. As the technology matures and best practices emerge, AI/ML will likely transition from a competitive differentiator to a baseline expectation for modern pipeline engineering and operations.
The question is not whether AI/ML will transform pipeline management, but how quickly and effectively individual operators will navigate this transformation. Collaborative partnerships among delivery partners, engineers, and asset owners are required to make this a normal business practice and ultimately deliver fit-for-purpose infrastructure that serves the end user.
For industry leaders, the imperative is clear: invest in the foundational capabilities of data infrastructure, talent development, and organisational readiness — key enablers of effective AI/ML deployment. For researchers and technology providers, the challenge is to develop solutions that are not only technically sophisticated but also practical and pragmatic, explainable, and aligned with the operational realities of pipeline operations management. Together, these efforts can realise the full potential of AI/ML to create safer, more efficient, and more sustainable energy pipeline systems.
-ENDS-
Contact