Machine Learning/Artificial Intelligence Driven Computer Vision for Cuttings Analysis Under Drilling and Completions

Abstract

The shape and size of formation cuttings passing through a shaker screen can provide valuable insights about any potential downhole problems. Large size cuttings or carvings may indicate the presence of an abnormal pressure zone and hole size may be enlarged which may lead to NPT events (stuck pipe, loss circulation, etc.), asset loss or HSE incidents. We proposed a new method of real-time automated analysis of cuttings in the shale shaker enabling faster reaction to mitigate risks associated with drilling operations. The solution uses a camera on the shaker screen, capturing the cuttings images and applying computer vision and convolutional neural networks algorithms to identify and classify individual cuttings shape, size and type combined with wireline data to raise alarms on specific conditions and prescribe actions to mitigate the problem. The solution showed a remarkably high confidence in identifying the cutting types and size and in detecting potential problems at their early stage enabling the drilling engineers to take the corrective actions at the onset of an event.