CASE STUDY
Pipeline Welding Inspection
Our client, a leading global pipeline supplier, enlisted our support in elevating the quality control of defect inspection in their welding lines. They specifically emphasized the need for advanced x-ray image analysis to accurately identify pores. Mindtrace, acknowledged for its expertise in this field, was approached to deliver a customized solution to meet their precise requirements.
Challenges
- A limited quantity of unlabelled training data (15 images)
- Only 5 available images for testing
- Several types of pipeline defects
Radiography film image collection
Mindrace collected all available welding images, all of which were radiographic film images.
Radiography film quality classification
Our Classification brain (C-Brain) assessed the quality of radiographic film in order to accept or reject images, resulting in a final selection of 10 training images.
Defect detection and classification
The Detection Brain (D-brain) proficiently identified various types of defects present in the welding lines of the pipelines. These defects encompassed cracks, penetration issues, fusion irregularities, porosity, and other anomalies.
Outcome
Mindtrace effectively demonstrated the Brain-Sense™ platform’s anomaly detection capabilities. Despite constraints with a small dataset (10 training images and 5 testing images), the delivered solution achieved an impressive 91% accuracy. As the model processes more images, its accuracy is anticipated to improve further, leveraging the continuous learning capabilities of the AI brain.
Training Samples
Accuracy
Week Project