TDNN neural network for diagnosing the state of the fan installation of the main airing

The article reviews and analyses existing methods for diagnosing a mine fan. Based on the main advantages and disadvantages of these methods, a neural network method for diagnosing the state of the main ventilation fan has been developed and implemented. The basis of this method is the proposed TDNN neural network, which is associated with its providing the best diagnostic accuracy. The paper defines the structure of the neural network model, selects and justifies a feature for evaluating its effectiveness, and performs training using addition reduction on the training set. To accelerate the process of training the author's neural network, a batch mode of training is proposed; it should be noted that the speed of forward and backward travel increases when using it. The architecture of this diagnostic model is determined on the basis of experimental studies. It has been found that an increase in the number of modules of the input layer of the neural network is accompanied by a decrease in the value of the root mean square error of diagnosis. The results obtained indicate that it is inexpedient to use more than sixteen modules in the input layer of the network. With a larger number of modules, a rather insignificant change in the value of the error will be observed. The adequacy of the above model is characterised by the choice of parameter values that ensure minimisation of the root mean square error, i.e. minimisation of the difference between the model output and the desired output. The time-delay neural network chosen by the authors allows achieving a diagnostic result with the smallest deviation and ensuring an improvement in the quality of the results of the fan diagnostic process. The proposed approach can be applied in various intelligent systems that provide diagnostic processes of various kinds

diagnostics; fan installation of the main airing; neural network; TDNN; operational safety; batch training mode