Stability Analysis of Overburden Dumps over Old Underground Workings Using Artificial Neural Networks

Stability of overburden dump slopes is a crucial aspect in designing secure and cost-effective dumps. The Strength Reduction Factor (SRF) serves as a widely used term to assess dump stability. This paper focuses on developing an Artificial Neural Network (ANN) model capable of predicting SRF for overburden dumps situated above existing underground workings. To construct the model, a dataset comprising 96 numerical simulations of overburden dumps generated through the finite element method was utilized. A neural network architecture with three layers of forward-backward propagation was utilized, containing hidden neurons to analyze simulations during training, validation and testing stages. The input parameters for studying overburden dump slopes over underground workings included dump slope height (Sh), dump slope angle ( $$\theta$$ ), cohesion (C), friction angle (Ø), unit weight ( $$\gamma$$ ) of the dump material, depth of working from the surface (D), centre-to-centre pillar distance in underground workings (C-C), and gallery width (Gw). The ANN predicted results were compared with the outcomes derived from numerical simulations of overburden dump slopes above underground workings. The study highlights that the developed ANN model in this research proves highly effective in handling and designing complex overburden dump slopes. The obtained results indicate a Mean Square Error (MSE) of 0.0595 and a coefficient of determination (R) of 0.883, both of which are considered acceptable.