Enhancing the Corrosion Assessment and Inspection management in Upstream Piping Systems Using Machine Learning
TWI Industrial Member Report 1205-2024
By Mehran Izadkhah
Industrial Need
Managing the integrity of piping systems and pressurized components specifically on offshore facilities is always challenging due to environmental conditions, limited accessibility, and costly logistics. Many components, especially those running at height, under insulation, or in congested areas, cannot be regularly accessed for inspection. As a result, process owners and integrity engineers often lack timely and reliable data to assess corrosion risks and plan maintenance activities proactively.
Traditionally, corrosion prediction has relied on practical formulas, conservative corrosion allowances, and presence of previous inspections such as ultrasonic thickness measurements (UT) or material degradation assumptions. In the offshore platforms, these methods are not always easily applicable due to the lack of data (for using formulas), cost and budget issue for site inspection and underestimation of risk, potentially compromising safety.
With the rapid advancement of digital technologies, machine learning (ML) and Artificial Intelligence (AI) models are now considered as powerful tools for predictive analytics in asset integrity. These models coinciding with technical data analytics can identify patterns in large and diverse datasets that traditional methods are not designed to detect. More importantly, ML models can predict degradation even in the absence of prior inspection data, by learning from background information such as design parameters, process conditions, and historical trends across similar assets.
In Phase 1 of our development, we introduced CorrosionWISE, an application built around an ML model that predicts internal corrosion rates in upstream offshore piping systems. This tool can help clients improve inspection prioritization and resource planning. This feature is particularly valuable for components that are difficult to access or inspect, giving operators a practical estimate of remaining life without the need for previous UT data.
Now in Phase 2, CorrosionWISE will be enhanced by integrating a new ML-based model focused on external pitting corrosion prediction, specifically presence and estimation of pitting depth.
The simplicity of this application lies in its ease of use and interpretability: users input readily available assets and process information, and the model returns estimated pitting risk levels and depth. The results are displayed through a clean, user-friendly interface designed for fast decision-making, even by non-expert users. This allows for early intervention and better prioritization of barge visits, avoiding unplanned downtime or unnecessary inspections.
Looking ahead, this model is not limited to piping systems alone. Its predictive framework can be extended to other onshore and offshore assets, including pressure vessels, risers, heat exchangers, and structural components. This scalability opens the door to a more comprehensive, digitalized integrity management strategy where prediction, rather than reaction, drives inspection and maintenance planning across the platform.
By adopting AI and machine learning in this way, we not only address a pressing need in the industry but also move towards a smarter, more efficient, and risk-informed future for offshore asset management. Furthermore, the project aims to integrate Fitness for Service (FFS) evaluations, offering a digitalized approach to prioritize inspection budgets effectively on-site. Asset owners can extend critical infrastructure lifespan, contributing to a greener and more sustainable future.
The integration of this model acts as an effective bridge between standardized code-based RBI methodologies (e.g., API 581) and the practical insights of integrity engineers. By doing so, it enables data-driven prioritization of inspection and maintenance activities, even in the absence of conventional UT data, ultimately supporting smarter and more proactive asset integrity management.
This investment opportunity in CorrosionWISE embodies financial returns alongside a commitment to sustainability and efficiency. With a legacy of over 70 years in innovation, TWI stands at the forefront of driving industry progress and shaping the future of corrosion assessment and inspection management. This visionary capability is a proof to TWI's commitment to efficiency, practical engineering, and a transformative shift towards a sustainable industry landscape.
Key Findings
- The ML model demonstrates robust predictive capabilities with a 95% F1 score on the training data and 86% on the test data.
- The project successfully classified external pitting depth within an ML framework led by TWISEA, emphasizing non-zero, highly correlated features for valuable insights.
- Challenges exist in predicting categories exceeding 5 mm of external pitting depth, constituting only 0.8% of the dataset. Improvements, such as incorporating more external environment-related features, are suggested for enhanced accuracy beyond 90%.
- Correlation analysis indicates weak linear relationships between numerical features and external pit depth, with a negative correlation between remaining life and pit depth.
- Component type analysis highlights varying pit depth associations, with certain components consistently reporting zero pit depth while others, like Tee components, are strongly linked to deeper external pitting.
Impact
The integration of Internal an external corrosion prediction models into one software – CorrosionWISE- marks a significant moving forward in the field of corrosion assessment and integrity management. Developed by TWI, CorrosionWISE has already proven effective in predicting internal corrosion rates in offshore piping systems, especially in areas where inspection is challenging or impossible. Now, with the addition of external pitting prediction capabilities, the platform is prepared to transform how asset owners manage risk and plan inspections.
This advancement introduces a data-driven alternative to conventional methods that typically rely on conservative assumptions and generalized reference values. By leveraging a robust dataset of real-world corrosion measurements, the new model accounts for the complex environmental conditions found in marine and offshore settings providing more accurate, asset-specific corrosion predictions.
Incorporating both internal and external corrosion rate predictions, the enhanced model delivers a more realistic estimate of remaining life, even in the absence of prior inspection data. This directly supports smarter maintenance decisions, better prioritization of inspection scopes, and more strategic use of barge visits and manpower.
Additionally, the model improves coating condition evaluation and supports compliance with standards such as API 581, contributing to more reliable Risk-Based Inspection (RBI) planning. Its integration into a user-friendly software platform enables engineers and decision-makers to access predictive insights without needing deep data science expertise.
Most importantly, this model sets a foundation for broader application beyond piping systems—extending predictive capabilities to other offshore assets such as vessels, structural components, and risers. As the industry continues to move towards digitalization, tools like CorrosionWISE are not just enhancing safety and operational efficiency, but also enabling a more intelligent, proactive approach to asset integrity management across the entire lifecycle.