Ensemble Neural Network 3D-CNN and LSTM in the Problem of Assessing the State of a Technological System for Processing Ore Waste

The study deals with the problem of improving information support for a complex energy and technological system of processing ore waste which accumulates in large volumes in areas adjacent to mining and processing plants being a great environmental threat to the environment. The use of modern artificial intelligence methods in the process of such improvement is a current area of research, as it allows improving the quality of technological data analysis for monitoring and control systems. The novelty of the research results lies in the proposed model for analyzing technological information coming from the system for processing fine ore raw materials as well as the architecture of the ensemble deep neural network, being the base of the model. An ensemble neural network solves the problem of classifying the state of a technological system. The network consists of 3D convolutional networks and a recurrent neural network. This makes it possible to process multimodal information coming from multispectral video cameras and measuring equipment that aggregates data in the form of multidimensional numerical sequences. The use of 3D convolution ensures the dynamics of technological processes to be taken into account on the basis of video data processing. To test the proposed model software was created in MatLAB, which showed that the quality of state estimation by an ensemble neural network on a synthetic data set is higher than when using convolutional and recurrent neural networks separately.