Abstract
Mining machine simulators involve an IT system based on dedicated software to represent a specific piece of hardware in the working environment of operators, service technicians, and people supervising the operation. In this study, a proprietary roadheader simulator was used to simulate the machine's operating parameters and its dynamic behavior during the mining process in changing conditions. The information collected from the simulation was subsequently fed to a model and analyzed using artificial neural networks (ANN) to enable learning and predictive features regarding the simulated roadheader. This article presents the findings that led to the development of an ANN structure to determine the values of 15 parameters based on four input signals. The ANN models are multilayer perceptron (MLP) networks with one hidden layer. The ANN learning process was carried out using the backpropagation method based on a dataset consisting of 1029 samples/cases generated by computer simulation. Experimentally verified mathematical models were used to prepare the data, which was subsequently divided into training, testing, and validation sets. Three network variants were considered, for which the effectiveness of predicting the values of targeted parameters was assessed. The test set evaluation was carried out on a separate data set. The results indicated that neural networks with a single neuron in the output layer yield the best results. However, the results of the studies do not offer a clear statement as to which of the analyzed ANN structures yields the best results for all the parameters considered. The simulator-generated parameters were compared with the operating parameters of a virtually controlled roadheader in a known operational condition for verification, and also compared with the measured parameters while operating an actual roadheader manually by the operator. The developed predictive model of the roadheader's operating parameters, which was implemented in the {RH-Sim} simulator, allows for tracking the machine condition during the implementation of various automatic control strategies to ensure high efficiency and minimizing its dynamic load. This approach offers new opportunities for the development of roadheader automation in construction and mining applications.