# Overview

Strayos uses deep learning algorithms to **automatically** detect rock boundaries and calculate fragmentation particle size distributions.

# Activating the analysis

This analysis is triggered at the project upload stage by activating the "Post-blast: Fragmentation and Muckpile AI" toggle

# Accessing the module

Once the model is processed, the fragmentation analysis can be accessed by selecting "Fragmentation AI" from within the "Blast Performance" folder in the sidebar

# Fragmentation AI Screen

### Detect Rock Size

Switches on an off the rock boundaries detected by the AI. These are visualised as lines on the map.

### Particle Size Colour

Switches on and off the coloration of rocks based on their relative size. The rock colors are visualized on the map according to the legend in the side bar.

### Show all/none

These buttons show or hide all of the automatically detected muckpiles in the project.

### Muckpile list

The switches for each automatically detected muckpile show or hide that muckpile from view.

### Fragmentation graph

This switch opens and closes the pop up showing the full fragmentation analysis.

### Fragmentation size disr

This table shows the summary information for the fragmentation analysis of the selected muckpile.

D10, D50 and D80 show the sieve size through which 10%, 50% and 80% of the area of the muckpile would pass (respectively). This is based on the measured particle size distribution rather than the fitted Rosin-Rammler or Swebrec models.

Add DXX allows the user to specify a custom percentage passing for which the platform will calculate the respective sieve size.

n and b represent the fit factors for the Rosin-Rammler and Swebrec functions.

### Rock Diameters CSV

This export is a CSV file containing the diameters of every rock detected by the AI. The rock diameter is defined as the shorter edge of the smallest rectangle that bounds the entire rock area.

### Size Distribution CSV

This export is a CSV file containing the cumulative percentage passing % at each sieve size step modeled.

It is shown for the measured data as well as the fitted Rosin-Rammler and Swebrec models.

It is shown for both the 2D (Area) and 3D (Volume) calculation methods.

### Fragmentation Report

This button exports a pdf report containing the particle size distribution analysis outputs for all muckpiles in the project.

# Fragmentation Graph Screen

This screen is accessed by activating one of the toggle switches labelled "Fragmentation Graph" within an open muckpile in the sidebar.

## Understanding the graph and table

*Figure 1 : Fragment size distribution for data generated in field*

Cumulative percent passing curve is obtained by image analysis. User’s can delineate the area for which rock size distribution is required. This area is fenced by boundary and represented in red as shown in the Figure.1,(legend 1).Percent passing at different rock size is calculated by an algorithm, and displayed in Figure.1, (legend 2).Also,*D*_{01},

*D*_{50},*D*_{80}, uniformity index (n) for Rosin-Rammler equation, and curve-undulation parameter (b) for swebrec function are calculated and provided in Figure.1,(legend 3).

*Figure 2: Cumulative passing curve and histogram plot*

Distribution is represented in following manner:

**Line graphs: **Line graphs shows the cumulative rock mass percent passing at different size. Actual (image based), Rosin-Rammler, and Swebrec graphs are plotted and shown in different color. Cumulative percentage passing are plotted on a log-normal scale to capture the variation in rock size effectively.Figure.2 shows the line graphs for different distributions.

**Histogram: **Histograms represent the percentage of rock mass passing at different sieve size. These rock mass percentage are not cumulative and represents the rock mass passing for a given sieve size. Histogram’s are plotted on a linear scale and represented by blue markings on the x-axis.Figure.2 shows the histogram obtained for analysis of Figure.1 data.

## Statistics

The Statistics table is activated using the "Statistics" toggle below the summary table

**Total muckpile area **represents the area bound by the muckpile polygon

**Minimum diameter **represents the diameter of the smallest rock detected in the muckpile. Flying the drone at lower altitudes closer to the muckpile can help detect smaller fragment sizes.

**Maximum diameter **represents the diameter of the largest rock detected in the muckpile.

**Mean diameter **represents the arithmetic mean of the diameter sizes of all the rocks detected in the muckpile.

**Diameter STD **represents the standard deviation of the rock diameters of all the rocks detected in the muckpile.

**Rocks count **represents the total number of rocks detected in the muckpile.

## 3D Analysis

The 3D Analysis toggle switches the graph and table view between the 2D and 3D particle size distribution calculation methods.

**The 2D method** calculates an area-based view of the fragmentation distribution (100% refers to 100% of the detected muckpile area)

**The 3D method** calculates a volume-based view of the fragmentation distribution (100% refers to 100% of the detected muckpile volume)

The 3D method can help to reduce the impact of large boulders at the edge of the muckpile on the overall distribution results.

## Custom Sieve Sizes

This feature allows the user to define a series of bin sizes for the histogram and table. This allows the user to see, for example, exactly what percentage of material sits in the target size range for their crusher.

This series is defined by

**Min** - the lowest sieve size in the series

**Max** - the largest sieve size in the series

**Step** - the difference between each sieve size in the series

# Fragmentation Report

This printable pdf report contains a summary of all information for the active muckpiles in the project.

The report is downloaded from the "Fragmentation Report" button at the bottom of the sidebar

# Technical methodology to obtain fragment size distribution

**1. Fragment size distribution using image analysis**

Rock fragments are analyzed using image analysis techniques to obtain the distribution within delineated area. Individual rocks are bounded using rectangular boxes and their diameter is calculated as the shorter edge length of this box.

For 2D analysis, rock fragments are arranged in ascending order of their area and a distribution is plotted on a graph based on a % of the total detected area that exists beyond a specified sieve size.

For 3D analysis, rock fragments are arranged in ascending order of a number calculated as the rock area multiplied by the height of the muckpile at the centroid of the rock location. The sum of this series then approximates the total 3D volume in the muckpile. This effectively 'weights' the rocks by the volume of muckpile beneath them which reduces, for example, the impact of boulders on the edge of the muckpile.

The resultant distribution from either method is termed as actual distribution of the plot, and is depicted by the black line graph. **This process is fully automated and does not require user’s intervention.**

**2. Rosin-Rammler curve: **Rosin-Rammler distribution is part of Kuz-Ram model to obtain the rock size distribution, post blasting. It is given by the following equation.

where, R is mass fraction larger than size *X*.X, is the diameter of fragment (mm), *X _{c }*is the characteristic size (mm), n is the Rosin-Rammler exponent, and e the base of natural logarithm. Here,

*X*is approximately 36

_{c }*.*8% size retainment point on the size distribution function.

Uniformity index for the Rosin-Rammler function shows how well are the fragments distributed and usually ranges between (0.6-2.2).A value close to 0.6 shows that the muck is non-uniform, whereas, value close to 2.2 means uniform muckpile with majority of fragments close to mean size. A statistical relationship was developed by (Chung and Katsabanis) to estimate *n*. Coefficient,*n *is calculated using the equation.

**3. Swebrec curve: **Swebrec function is one of industry adopted size distribution function. Swebrec function is given by:

Here,*f*(*x*) = (^{ln(}^{dmax/}^{d}^{)}*/*ln(*d*_{max}*/ _{d}*

_{50}))

*,*

^{b}*d*

_{50 }is the 50% passing value and

*d*is an upper limit of fragmentation size,

_{max }*P*(

*x*) is the percentage of rock mass passing a given sieve size. Curve-undulation parameter (b) is dynamically calculated using by equating slopes at

*d*

_{50 }(Ouchterlony).

**4. Fragment size calculation procedure: **Finally, fragments size distribution is calculated using the following steps:

- User’s delineate the area where the fragmentation analysis is desired. This area is delineated in red portion in Figure.1,(legend 1).
- Algorithm detect the rocks automatically and creates a size distribution. This distribution is termed as the actual distribution and is shown in 1,(legend 2).Further,
*d*_{01},*d*_{50},*d*_{80}, the uniformity factor*n*, and*b*is calculated by procedure mentioned in Part’s 2, and 3. - Histogram is plotted to show the rock passing percentage at different sieve sizes.
- Line graphs are plotted for actual, rosin-rammler, and swebrec curves as shown in Figure.2.

## Comments

0 comments

Please sign in to leave a comment.