Overview

Strayos provides you the most accurate fragmentation predictions possible by automatically calibrating them using previous blasts. The prediction models continually learn and improve as you use the Fragmentation Prediction and Fragmentation AI modules.

What you have to do

Most of the process is automated behind the scenes however there are a few things you should check to ensure that your predictions are being properly calibrated. An outline of the end to end process is shown below:

1. Design a blast

Use the Drilling Design, Loading Design (optional), and Timing Design (optional) modules to create an initial blast design. This will allow the Fragmentation Prediction module to automatically populate the input fields based on your actual design. (Note: if you would like to skip this step it is possible to just create a "dummy" drilling design and then manually enter values to the input fields later.)

2. Predict the fragmentation

Launch the Fragmentation Prediction module from the Blast Prediction folder in the sidebar. 

You will see some or all of the input fields populated automatically from your blast design and you will see the predicted fragmentation distribution for the current input field values.

From here you can edit the inputs to find the optimal design parameters to achieve your target fragmentation. For example, if you know that all rocks over 1000mm require secondary breakage, you could vary the burden and spacing until you reach an optimal balance between D&B cost and secondary breakage cost.

You can also go back and change the blast design and see how that impacts the fragmentation prediction. For example, you could:

1. Make one prediction with ANFO
2. Click "Save for comparison"
3. Go back to the Loading Design module and switch the ANFO with Emulsion
4. Return to the Fragmentation Prediction module and see the new Emulsion prediction
5. Toggle back and forth between your two predictions using the tabs above the graph

For the AI calibration to work, once you have finalized your blast parameters you must click "Save for Comparison" and set the correct prediction as the Primary Prediction.

You will see it appear in the sidebar list under "Compare Saved Predictions".

The first prediction you save will automatically be set as the "Primary Prediction". This is signified by the star next to the prediction name. The Primary Prediction is the one that will be used to calibrate future predictions.

To set the final prediction as the Primary Prediction you can click the three dots menu next to the prediction name and select "Set as Primary Prediction".

Interpreting the prediction results

The chart shows two cumulative lines representing the percentage of material which is predicted to pass through a sieve of a particular size. E.g. you can read it as "What percentage of material is smaller than a certain size"

Each line represents a different industry standard prediction model (Kuz-ram and xp-frag). You can decide to use either model as the basis for your analysis. Both models will be calibrated using AI to match the fragmentation measurements made for that site.

D10, D50 and D80 represent points on the graph corresponding to 10%, 50% and 80% passing. E.g. D50 is the sieve size that 50% of material would pass through.

The table shows the graph data at discrete sieve sizes so you can read what percentage of material is smaller than each certain sieve size. 

The sieve sizes in the table can be customized using the "Customize Sieve Sizes" toggle. This lets you make predictions based on sizes that are most meaningful for your operations e.g. maximum crusher jaw size or shovel bucket size.

3. Measure the fragmentation after the shot (to calibrate the prediction model)

Create a new project using your muckpile images and ensure you switch on "Post-blast: Fragmentation and Muckpile AI" and link it to your pre-blast project containing the prediction.

Once the new project has finished processing, launch the Fragmentation AI module from the Blast Performance folder in the sidebar.

Any muckpiles that have been detected by the AI will be listed in the sidebar. The muckpile with the star next to the name is the Primary Muckpile. This is the one which is used to calibrate future predictions. 

By default, the closest muckpile to the pre-blast shot design is set as the Primary Muckpile. If you would like to change this, you can do it by clicking the three dots menu and selecting "Set as Primary Muckpile"

To view the fragmentation measurement results, toggle on a muckpile and then toggle on "Fragmentation Graph" in the sidebar.

You can compare it with the Primary Prediction for the linked pre-blast project by toggling on "Compare with Prediction". Note: the first few times you do this the comparison may not be close as the prediction model is not yet calibrated to your site. As you enter more data, the predictions will become more accurate.

4. Repeat the process for the next shot

As long as you set a Primary Prediction and Primary Muckpile for each blast, the prediction models will continually learn and improve for your site.

Note: it is possible to turn off the calibration for a prediction and use the basic un-calibrated xp-frag or Kuz-Ram models. You can do this by turning off the "Calibration" toggle in the Fragmentation Prediction module. This will not affect any of your other saved predictions.