Identification of accident classes in coal mines using the neural network technology

The article addresses prompt decision-making on coal mine safety in case of accidents by the dispatching personnel. The discussion deals with issues connected with identification of underground accidents using the learning model of fuzzy neural network. The algorithm of fuzzy neural network classification of off-optimum situations (OOS) in coal mines is proposed. Belonging of the learning image of OOS to the existing or a new cluster is determined from the maximum similarity concept. The function of difference between the learning image of OOS and the cluster in the neural network is defined as the Euclidean distance. This function displays two vectors (the learning image of OOS and the cluster) as a real number. Images of each cluster (average values of OOS images in a cluster) are stored in the communication links (weights) of the neural network during the whole process of classification. Application of the classification algorithm in the conditions of partly overlapped classes of accidents is illustrated. If classified clusters are partly overlapped, each image of OOS in the learning family may belong to more than one cluster. Binary matrix serves for recording each image of OOS in a cluster. The algorithm based on a hypothesis that there exists a model for each cluster, which is determined as an average value of models of all images of OOS in this cluster, is used to determine the membership function. Membership of each image of OOS in a cluster is determined from similarity of this image to the cluster model. The similarity can be found as the function of the Euclidean distance between the image and the cluster model. This article uses the triangle membership function in fuzzy set. This function is simple for the mathematical manipulation in equivalent application domain. The competitive learning was applied. In this case, the learning processes is corrected by the weight factor connected with the input and output nodes of neural network. The learning model was applied to identification of existing classes of standard accidents, to modeling these classes and to classification of OOS in coal mines. The quality of this model performance was estimated using a learning set composed of many learning images. The learning model is the base model, and is applicable to dealing with large-scale problems connected with identification of OOS, as well as causes and initiation conditions of accidents. The idea of such identification is formulated.

Kupriyanov V. V. Identification of accident classes in coal mines using the neural network technology. MIAB. Mining Inf. Anal. Bull. 2022;(8):148-157. [In Russ]. DOI: 10.25018/0236_1493_2022_8_0_148.

V.V. Kupriyanov, Dr. Sci. (Eng.), Professor, National University of Science and Technology «MISiS», 119049, Moscow, Russia, e-mail: Kupriyanov.VV@misis.ru, ORCID ID: 000-0003-3793-8361.

1. Kupriyanov V. V. Methodology of gas-nonregular situations operative discovery at the panels of mine. Vestnik Universiteta. 2020, no. 3, pp. 76—83. [In Russ]. DОI: 10.26425/18164277-2020-3-76-83.

2. Kupriyanov V. V. Automation of nonregular situations recognition in coal mines on the basis of topology and weight coefficients — change neural network. Trudy XVIII Vserossiyskoy nauchnoy konferentsii «Neyrokomp'yutery i ikh primenenie» [Proceedings of XVIII All-Russian scientific conference «Neuralcomputers and their application»], Moscow, MGPPU, 2020, pp. 59—60. [In Russ].

3. Murashchev S. I., Timoshenko A. M., Sukhorukov V. A. About analysis of technogeneous accidents with catastrophic effects which took place at coal enterprises in Russia. Industrial Safety. 2012, no. 1, pp. 55—60. [In Russ].

4. Smirnyakov V. V., Smirnyakova V. V. Unhandy factors in statistics of accidental gas and dust explosions in coal mines in Russia. Gornyi Zhurnal. 2016, no. 1, pp. 30—34. [In Russ]. DОI: 10.17580/gzh.2016.01.07.

5. Temkin I. O., Do Chi Thanh, Agabubaev A. Enhancement of safety in Veitnam mines with high methane content using models of prediction analytics. MIAB. Mining Inf. Anal. Bull. 2017, no. 8, pp. 142—151. [In Russ]. DOI: 10.25018/0236-1493-2017-8-0-142-151.

6. Svetlov V. A., Persiantsev I. G., Shugay I. S. A new implementation of the algorithm of adaptive construction of hierarchical neural network classifiers. Optical Memory and Neural Networks (Information Optics). 2015, vol. 24, no. 4, pp. 288—294.

7. Adeli H., Yeh C. Neural network learning in engineering design. Proceedings of International Neural Network Conference, 2000, pp. 412—415.

8. Krizhevsky A., Sutskever I., Geoffrey H. ImageNet classification with deep convolutional neural networks. Communications of the ACM. 2017, vol. 60, no. 6, pp. 84—90.

9. Kolenov D., Davidse D., Le Cam J., Pereira S. F. Convolutional neural network applied for nanoparticle classification using coherent scatterometry data. Applied Optics. 2020, vol. 59, no. 27, pp. 8426—8433. DOI: 10.1364/AO.399894.

10. Ahammed A. K., Azeem M. F. Robust stabilization and control of Takagi-Sugeno fuzzy systems with parameter uncertainties and disturbances via state feedback and output feedback. International Journal of Fuzzy Systems. 2019, vol. 21, no. 8, pp. 2556—2574.

11. Hung S. L., Adeli H. A model of perceptron learning with a hidden layer for engineering design. Neurocomputing. 1991, vol. 3, no. 1, pp. 3—14. DOI: 10.1016/0925-2312(91)90016-5.

12. Jafari M., Mobayen S. Second-order sliding set design for a class of uncertain nonlinear systems with disturbances: An LMI approach. Mathematics and Computers in Simulation. 2019, vol. 156, pp. 110—125. DOI: 10.1016/j.matcom.2018.06.015.

13. The Coalmining History Resource Centre, available at: http://www.cmhrc.co.uk/site/disasters/ [accessed 30.01.2021].

14. Jong-Chih Chien, Ming-Tao Wu, Jiann-Der Lee Inspection and classification of semiconductor wafer surface defects using CNN deep learning networks. Applied Sciences. 2020, vol. 10, no. 15, pp. 13—21, article 5340. DOI: 10.3390/app10155340.

15. Aristov A. O. Quasi-cellular nets based on models of flow-systems. Journal of Physics: Conference Series. 2019, vol. 1392, no. 1, pp. 1—5, article 012064. DOI: 10.1088/1742-6596/ 1392/1/012064.

16. Montana D. J., Davis L. Training feedforward networks using genetic algorithms. Proceedings of the 11th International Joint Conference on Artificial Intelligence, vol. 1. Morgan Kaufman Publishers, San Mateo, California. 1989, pp. 762—767.

17. Takagi T., Sugeno M. Fuzzy identification of systems and its applications to modeling and control. IEEE Transactions on Systems, Man, and Cybernetics. 1985, vol. 15, no. 1, pp. 116— 132. DOI: 10.1109/TSMC.1985.6313399.

18. Ramirez I. E., Minami Y. Design of neural network quantizers for networked control systems. Electronics. 2019, vol. 8, no. 3, pp. 318—326. DOI: 10.3390/electronics8030318.