Grain-e Reviews & Product Details

Digital Twin of Wellbore Geology Digitalising Sediment Description using AI: From Grain Images to Reservoir Insights Our team has developed a powerful new approach that uses machine learning to transform how geologists interact with cuttings samples. By combining high-resolution imagery, artificial intelligence, and interactive filtering tools, we’re making it possible to measure thousands of individual grains automatically and accurately in each cutting sample, delivering detailed sedimentological data in a fraction of the time. From Cuttings to Clean Data To create consistent data, samples are processed using the proprietary approach of Rockwash Geodata. The samples go through a fully automated washing process. Each samples is photographed under controlled lighting. Each image captures fine details of a 43 mm by 28 mm area at extremely high resolution (4000 x 6000 pixels). Smart Grain Segmentation (AI) The real breakthrough lies in what happens next. Using a proprietarily trained segment anything model, we segment the image—identifying and outlining individual grains. This gives us a digital map of the grains in each sample. We then calculate properties like grain size, roundness, sphericity, and average colour for each grain. However, the process isn’t perfect. Sometimes the model mistakes a bit of shale or paint for a grain, or combines multiple grains into one. That’s why we’ve developed an interactive filtering system that lets geologists review and refine the output. With sliders and visual tools, users can filter out grains based on colour, size, shape, brightness, and whether the grain is clearly visible in the foreground. This ensures only representative grains are included in the final datasets. Sedimentary statistics The ability to get grain size, sorting, sphericity and other characteristics on thousands of grains allows for a vast dataset to understand the sedimentology. With the ability to filter out shale contamination, cavings, drilling additives and partially obscured background grains allows for a clean high confidence sedimentological information in even highly contaminated samples. Geomechanics (Pore Pressure module) An exciting feature is the ability to identify “splintery” cavings—elongated fragments that can indicate high pore pressure zones in the subsurface. These features often appear in over-pressured formations where the rock breaks due to higher pore pressure than mud weight. Automatically detecting and quantifying these features identifies early warning signs for drilling hazards and informs geomechanical models in offset wells. Stratigraphic subdivision With the ability to extract colours from each grain, given the high-quality consistent lighting we can build stratigraphic logs based on bulk colour (majority of the grains) and trace colour (few or single occurrence grains in a sample). These colour logs can reveal subtle changes in bulk or trace mineralogy relating to provenance and depositional environments. Why It Matters This approach has several big advantages. First, it creates a consistent, repeatable way to describe sediments. Whether you’re working on a single well or comparing data across a basin, the measurements are objective and data-driven. Second, it dramatically speeds up analysis. Instead of estimating colours and composition under the microscope, the system analyses thousands of grains in each sample, with hundreds of samples in a well. Third, it opens up new types of data that were previously hard to capture at scale. Properties like sphericity, trace colour, or the presence of splintery cavings can now be logged quantitatively and consistently. Applications Reservoir Characterisation: Detailed grain properties help assess sorting, depositional energy, and potential permeability. Drilling Safety: Detection of splintery cavings can support real-time pore pressure prediction. Stratigraphy and Correlation: Colour logs and grain statistics assists correlation between wells and identification of subtle lithological changes. Cuttings Digitisation: With a growing move towards digital well data, this method provides a scalable way to digitise physical samples at scale and integrate all your data into subsurface models. Looking Ahead We’re refining the system and expanding its capabilities. Talk to us about your specific problems and we will develop the tools to extract the statistics you most need!