WP5 ROCK MASS MODELLING SOFTWARE

3GSM (lead), UPM, VA Erzberg GmbH, MUL, TU GRAZ

WP5 will provide a software component for measuring muck pile properties based on aerial imagery. A new algorithm for fragmentation analysis and the presence of boulders will be developed as well as a new automatic blast design software based on surface geometry and automatically characterised
rock mass structures.

5.1. COMPUTATIONAL MUCK PILE CHARACTERISATION

The main task is to determine muck pile properties like shape, presence of boulders and fragmentation in open cut blasting sites by using a camera carrying UAV (unmanned aerial vehicle) combined with modern machine learning and photogrammetric computer vision systems. Work includes the development of a new algorithm for interpreting visual and geometric data from muck piles that allow for identifying (i) the fragment size distribution, (ii) the presence and
location of boulders, (iii) the identification of partially covered fragments, and (iv) fine material below image resolution

Subtask 5.1.1: Planning, Specifications and Design
Literature review and specification of experimental conditions such as geometric boundary conditions of blasts

Specifications of key parameters describing muck pile properties for the given context
Specification of imaging system

Subtask 5.1.2: Data acquisition
Provision of an adequate UAV including required legal permissions
Data capture flights at ERZBERG and ARNO with the UAV to gather test data for development and reference data for evaluation purposes

Subtask 5.1.3: Algorithm and software development
The goal is to utilize state-of-the-art findings in machine learning and 3D computer visions do develop computational methods for muck pile characterisation. Deep learning and convolutional neural networks (CNN’s) will be developed to interpret images of muck piles, e.g. for boulder delineation. In particular the possibility of a joint classification using 2D image information and 3D information from images will be investigated. To achieve these goals, first the
photogrammetric computer vision system of TUG will be adapted to the application domain. Then methods for deep learning will be investigated and a combined 2D/3D system will be investigated.
Muck pile analysis and development of application software (3GSM)
Integration with existing application software (3GSM)

Subtask 5.1.4: On site tests and evaluation
Evaluation of the investigated algorithms on relevant test data and final one site tests close to the end of the task. For
this purpose fragmentation data from sieving in WP2.3 and aerial photogrammetry will be considered

5.2. AUTOMATIC BLAST DESIGN SOFTWARE

3GSM (lead), ERZBERG

The main objective of this task is to provide a software application that allows for an automatic blast design based on the precise knowledge of the blast site geometry and structural information of the rock mass. A sound 3D model of the rock mass that includes automatically determined geological information is used to determine a drill pattern that optimises blasting results in terms of energy consumption and desired fragmentation. Another fundamental requirement for achieving the objective is to establish an according loading of the boreholes analytically. The following actions shall be carried out in this task:

  • Development of an algorithm for an automatic identification of rock mass structures (discontinuities)
  • Provision of a software component for field testing the automatic structure determination
  • Development of an algorithm for the estimation of in situ block sizes
  • Provision of a software component that determines blast hole locations automatically considering the actual geometry of the free face
  • Extension of the software component by allowing updates of the blast layout according to determined rock structures and in-situ block sizes
  • Extension of the software component for an automatic loading scheme for the determined blast layout
  • Testing of the software component on real blast sites
  • Evaluation report

 

WP OVERVIEW