Context
In the demanding environment of offshore oil operations, quick connectors play a critical role in assembling long drilling riser pipes, often 10 meters or more in length. These components must be fastened and unfastened rapidly while withstanding significant mechanical loads and high-pressure conditions. The reliability of such systems is essential to ensure both the safety and profitability of offshore operations.
To meet these requirements, a technology called “Clip Riser” was developed to enable fast and secure connections between drilling riser joints. These connectors significantly reduce downtime at sea. To guarantee their durability and performance over time, the devices undergo rigorous lab testing under stress, including aging tests that simulate extreme fatigue conditions.
In addition to testing in controlled environments, non-destructive testing (NDT) is a key priority. It is essential for continuously monitoring equipment health without altering it. Very high-frequency acoustic emission (AE) technology is used for this purpose. This method allows for the evaluation and analysis of noise levels linked to industrial activities and installations, capturing early signs of damage—such as the appearance of micro-cracks invisible to the naked eye.
The ultimate goal is to transfer these NDT methods directly into operational conditions. This would enable real-time, proactive monitoring of risers, without the need for disassembly or operational interruptions. Field deployment of such NDT methods is a strategic advantage to enhance safety, optimize maintenance, and reduce the risk of failure in offshore environments.
Needs
The project had several goals:
- Evaluate the performance of the DiagFit software, using Machine Learning to detect damage during fatigue testing.
- Assess whether acoustic emission (AE) analysis could outperform traditional expert-based analysis in identifying damage.
- Automate the analysis of NDT data, accelerating the test campaign reporting process.
The study was based on data from acoustic emission sensors. Thanks to very high-frequency capabilities, these sensors can detect weak signals that indicate the early formation of micro-damage.
Solutions
To address these objectives, DiagFit, a no-code software platform specialized in the intelligent analysis of industrial time series, was used to process AE data. The process was carried out in several stages:
- Data import:
- Initial data preparation
- Automatic detection and categorization of imported data using DiagFit
- Dataset annotation based on established business rules and domain expertise
- Model training:
- An unsupervised model was trained to learn normal behavior and detect significant deviations
- A finalized anomaly detection model was created, aligned with the anomaly zones defined by the user
- Model application:
- The model was applied to other datasets, enabling automatic anomaly detection across test scenarios
Results
In an initial test on a connector subjected to hundreds of thousands of fatigue cycles, the software enabled :
The presence of anomaly zones detected by DiagFit (anomalies in blue) upstream of the red zone + user-defined absence zones (in red: based on EA analysis results)
- Anomalies detected by the software appear before the red zone; anomaly-free zones were computed during testing.
- Traditionally, defining an anomaly zone required about 38,000 cycles and several hours of expert analysis. With DiagFit, this task was completed in around ten minutes → Optimized NDT process.
- The client particularly valued the ease of use and the relevance of the results.
Interested in testing DiagFit on your own industrial equipment?
