Mining Contract: Ground Hazard Mitigation Using Neural Networks

This page is archived for historical purposes and is no longer being maintained or updated.

Contract #	75D30121C12372
Start Date	9/1/2021
End Date	8/31/2023
Research Concept	Roof falls are a major safety concern in underground mines, causing deaths, injuries, and lost work times to miners in the U.S. According to accident, Illness, and injury and employment Part 50 data from the Mine Safety and Health Administration (MSHA), from 2010 through 2020, roof fall events resulted in a total of 2443 injuries, including 19 fatalities, 24 permanent total or partial disabilities, 1130 days away from work only, 142 days away from work and restricted activity, and 255 days restricted activity only.

Contract Status & Impact

This contract is complete. To receive a copy of the final report, send a request to mining@cdc.gov.

Under this contract, the overarching objective of this research is to minimize roof fall accidents in underground mines through proper strata characterization using the Artificial Neural Network (ANN) technique. The research will develop ANN models for underground mine strata characterization.

The research will have two major impacts on the mining community. First, the knowledge gained by using ANN technologies will lead to improved management of safety hazards related to roof falls in underground mines. Second, improved safety of mine works will be economically beneficial to the industry.

Using ANN models to reduce roof falls, the research will be performed in three stages:

A portion (typically 70%) of the collected borehole sample data (X, Y, Z coordinates, borehole depth, assay, RQD, joints, lithology, UCS) will be used to train (or teach) the ANN models for each strata characteristic individually.
Trained models will be validated for the corresponding strata characteristic using the remaining (30%) collected borehole sample data.
Validated models will be utilized to predict the strata characteristic results for the new boreholes drilled in 2021, and the predicted results will be compared against data from the new borehole samples to re-verify the accuracy of the models.

The ANN models can then be used to predict strata characteristics at unknown locations and provide guidance for mine design and ground support. The level of prediction accuracy can be increased by validating the model again using data from boreholes drilled in 2022. The same model will be applicable to other mines if the mines have collected relevant rock strata information to train and validate the model.