Prediction of rock mass rating using neural network with an improved rider optimization algorithm

Evaluation of TBM performance is critical for the choice of TBM specifications and tunnel design. In the past decades, the hypothetical schemes depending on the rock fragmentation process and the experimental models up to field surveillance as well as machine performance are the two main methods. Traditional and conventional approaches for rock mass rate (RMR) prediction usually consider excessive parameters and the accuracies are far from actual values. A new RMR prediction model based on the optimized neural network (NN) is designed. To improve the prediction accuracy, this paper proposed a new self-adaptive rider optimization algorithm (SA-ROA), which applied optimization logic to train the NN by updating the weight as wave velocity (Vp), transverse wave velocity (Vs), Vp/Vs, statistics (Stat), orientation, magnitude, polarity, wave type, and metre. Finally, the RMR prediction analysis of the adopted NN-SA-ROA model is compared to the conventional and traditional classifiers with varied learning percentages: 50%, 60%, 70%, and 80% for three data sets, respectively. Subsequently, the performance of the proposed work is verified using other approaches based on error analysis. The predicted mean absolute errors (MAEs) and the mean absolute percentage errors (MAPEs) of SA-ROA are smaller than conventional and traditional schemes. The results show that the proposed method can successfully predict the actual RMR.