Automation of the process of dynamic signals of tensometric systems using convolutional neural network

Illia Kolysnychenko, Victor Tkachov

Abstract


Purpose. Based on empirical data obtained from the system of weighing on single-platform railway scales in motion and numerical methods, create software to generate an array of input data to the wrapped neural network. Using the Python programming language, build a model of a wrapped neural network that will allow you to recognize the type of object (auto-coupling, cart) that has passed through the strain gauge platform. Generate test data of different quality, with noise overlay, and test the recognition quality of the obtained model on different test input data. Research methods. The Python programming language and keras library were used to build plot diagrams of objects through the weighing platform, create a model of a convolutional neural network and generate test plots with noise overlay. Results. Empirical data from the railway strain gauge system were approximated for all types of moving objects used by the enterprise using the algorithm of approximation to empirical data by the Heaviside function. Software for data generation based on the obtained approximation equations is developed. Using a noise generator, we managed to create a dataset for learning and testing a convolutional neural network with different levels of input signal quality. A convolutional neural network has been built and trained, the test of which shows a high level of recognition of test objects, even when simulating an incorrectly configured weighing platform. Scientific novelty. The novelty is a new method of processing dynamic signals for tensometric systems based on machine learning, which with minimal error can recognize the types of objects that have passed through the weighing system. This used algorithm for processing dynamic signals is universal and can be so in many dynamic systems. Practical meaning. The method presented in the article is a key to building an intelligent identification system for evaluating wagons in dynamics. This system allows you to increase the speed of weighing cars and reduce errors in the system as a whole.

Keywords


scales; strain gauge; artificial intelligence; weight measuring systems; convolutional networks; dynamics; trolley; signal

References


Shahid, N., Rappon, T., Berta, W. Applications of artificial neural networks in health care organizational decision-making: A scoping review. PLoS One, 2019, vol. 14, iss. 2, article no. 0212356. DOI: 10.1371/journal.pone.0212356.

Tian, Y. Artificial intelligence image recognition method based on convolutional neural network algorithm. IEEE Access, 2020, vol. 8, pp. 125731-125744. DOI: 10.1109/access.2020.3006097.

Gevaert, W., Tsenov, G., Mladenov, V. Neural networks used for speech recognition. Journal of automatic control, 2010, vol. 20, iss. 1, pp. 1-7. DOI: 10.2298/jac1001001g.

Dixit, R., Kushwah, R., Pashine, S. Handwritten digit recognition using machine and deep learning algorithms. International journal of computer applications, 2020, vol. 176, no. 42, pp. 27-33. DOI: 10.5120/ijca2020920550.

Karakaya, R., Kazan, S. Handwritten digit recognition using machine learning. Sakarya university journal of science, 2021, vol. 25, iss. 1, pp. 65-71. DOI: 10.16984/saufenbilder.801684.

Shrivastava A., Jaggi I., Gupta S., Gupta D. Handwritten digit recognition using machine learning: A review. 2nd international conference on power energy, environment and intelligent control (PEEIC), Greater Noida, India, 18–19 October 2019, 2019, pp. 322-326. DOI: 10.1109/peeic47157.2019.8976601.

Kocer, Erdinc H., Cevik, Kursat K. Artificial neural networks based vehicle license plate recognition. Procedia Computer Science, 2011, vol. 3, pp. 1033–1037. DOI: 10.1016/j.procs.2010.12.169.

Ma, L., Zhang, Y. Research on vehicle license plate recognition technology based on deep convolutional neural networks. Microprocessors and Microsystems, 2021, vol. 82, article no. 103932. DOI: 10.1016/j.micpro.2021.103932.

Fennhagen, D., Johansson, T. Vehicle Pose estimation using machine learning. Örebro, Sweden, Institutionen för neurovetenskap och teknik Publ., 2020. 57 p.

Kolysnychenko, I. Yu. Doslidzhennya dynamichnykh syhnaliv odnoplatformnykh zaliznychnykh vah [Research of dynamic signals of single platform railway scales]. Zbirnyk naukovykh prats' natsional'noho hirnychoho universytetu – Collection of research papers of the national mining university, 2022, vol. 68, pp. 174–183. DOI: 10.33271/crpnmu/68.174.




DOI: https://doi.org/10.32620/aktt.2022.4.10