This page walks you through the process of running products on a drill hole.
Overview
The process of triggering a product is similar regardless of what is being run however, there may be some differences between input requirements and viewing results.
Image Preparation
A pipeline that prepares images for future processing. This pipeline will extract the box(es) and then extract rows for each of the boxes from the image.
Navigate to the Project page of the Datarock Platform
Locate the Upload button (top right of screen)
Images accepted by Datarock need to adhere to the following format:
HOLEID_BOXNUMBER+BOXNUMBER_FROMDEPTH_TODEPTH (optional)
Depth metadata is optional but will require a depth file (a CSV containing the start and end depth of the hole) to be uploaded in its place.
DDH001_1
DDH002_1+2+3
DDH001_1_0_6.2
DDH002_1+2+3_0_18.3
Images can not be rotated in the platform at this time. Prior to uploading, ensure that all images are oriented with core rows increasing in depth from top left to bottom right of the image.
Triggering Image Preparation
To manually trigger Image Preparation:
Navigate into the drill hole
Select 'Rerun Image Preparation from the Actions menu
Hit ‘Rerun’
The above icons are referred to as ‘Pill Boxes’. During the Datarock process, there are three states a pipeline run can be in:
Successfully completed (teal tick)
Failed to complete (red exclamation mark)
Still processing (spinning circle)
Depth Registration
A pipeline that determines the depth where rock came from underground, relative to the start of the drill hole. To achieve this, we run the rows through ML Models (coherent rock, incoherent rock, empty tray, and core block) and OCR and then apply some computational logic.
Select 'Run Depth Registration' from the actions menu and hit 'Run'. Once processing has started, the pill box icon will start spinning (above).
Completing Depth Health Checks
Models requiring depth registration to have run can still be run without depth registration results being approved but it's always a good idea to check the warnings before proceeding. Depth Registration results can be approved by completing a Depth Health Check.
If there are Depth Health Checks to be completed, areas of concern will be displayed in the 'Notifications' panel. Common issues that can increase the number of checks are:
Missing depth metadata
The more ground truth depths, the better
Missing core loss files
Having these helps the depth registration process account for large missing intervals
Poorly performing models
Missed or incorrect predictions can affect the depth registration results
Clicking on the interval or box numbers listed in the notifications panel will bring up the region of interest to inspect. In the example above, the image does not contain any depth metadata, to fix this error the user can add in the from and to depth of the image.
Approving and Revoking Depth Registration Results
Approving Depth Registration can be done via the 'Actions' menu in the drill hole page.
Depth Registration Results can be revoked at anytime via the ‘Revoke’ button located in the drill hole page. Once confirmed and results have been approved, the orange tick will turn green in both the drill hole tab and the pill box icon in the projects page.
Visualising Depth Registration results
Driller's Break and Rock Quality Designation (RQD)
Driller’s Breaks:
Fracture induced by the drilling process. Typically symbolised by a red 'X' on either side of the fracture in the core sample.
RQD:
Rock Quality Designation. A measure of rock quality that counts the pieces of rock that are ≥ 10 cm long. The measurements are made for a series of intervals (standard 1m or customisable).
After Depth Registration Results have been approved or depth results have been deemed appropriate without approval, RQD can be run. Triggering RQD from the 'Actions' menu in the drill hole page, will automatically trigger the Driller's Break model to run prior to RQD.
Once both Driller's Breaks and RQD have successfully run, pill box icons will turn green
RQD charts are available on metre intervals or when an accompanying user interval file has been uploaded, RQD is also calculated on given intervals and compared to ground truth data (if supplied). NB: The charts and visuals are from different drill holes.
Fractures
A general term to include any kind of discontinuity in a body of rock if produced by mechanical failure, whether by shear stress or tensile stress. Fractures include faults, shears, joints, and planes of fracture cleavage. Using machine learning, we then classify the predicted fractures into a series of geotechnical terms (glossary)
Fracture Segmentation
Joint Set Analysis
Datarock’s method of analysing joints to determine joint set number in line with the Q-System to determine
Classification
Fault Characterisation has been used here as an example, however the same process applies for all classification models such as Alteration, Lithology, Fault Classification etc.
