NEURAL MODELING BY DEEP LEARNING, PREDICTING THE OPERATION OF AN ECCENTRIC INJECTION PUMPS USED IN NATURAL GAS ODORIZATION

Emil TEUȚAN, Vasile RAFA

Abstract


The paper presents a predictive modeling in the operation of an odorizing pump operated with an eccentric mechanism, using a field of artificial intelligence (AI) called Deep Learning according to the parameters of flow and pressure in a piping system. natural gas transportation; odorization is necessary because natural gas is odorless, and in case of a leak in the transmission or distribution system for natural gas and if this oil spill is not detected, monitored, there is a danger of explosion, the worst danger being in urban areas. The operation of this pump with the eccentric using a DC motor without brushes, this motor being constructively adapted for a potentially explosive environment - according to the Ex zoning explosion described in stas SR 60079-10: 2016 through which the variation of the odorant flow injected into the pipe is determined. Using an application from the MATLAB program, more precisely an additional package of this program, Simulink, through which simulations of dynamic systems can be performed using mathematical models in order to optimize them, we will analyze the predictive operation of the injection pump. The neural model, which is implemented in the Deep Learning Toolbox software, and through a non-linear programming will predict the operation of this odorant pump used in the system of natural gas transmission pipelines. The controller, used in Simulink, will then model the control input that will optimize the operating performance of the eccentric injection pump in a defined time range.


Full Text:

PDF

References


Florea A.- M., Elemente de Inteligenta Artificiala, Ed. Universitatea Politehnica, București,1993.

Toderean, M., Coșteiu M., Giurgiu M., Rețele neuronale, Ed. Microinformatica, Cluj-Napoca, 1994.

Tiponuț V., Căleanu C-D., Rețele Neurale Arhitecturi și algoritmi, Ed. Politehnica, Timișoara, 2002

Dzițac. I., Inteligență Artificială, Ed. Universității Aurel Vlaicu, Arad, 2008.

Dumitrescu, D., Principiile inteligenței artificiale,Editura Albastră, Cluj Napoca, 1999.

* * * Statistical Mechanics of Deep Learning-Department of Applied Physics, Stanford University, Stanford, California 94035, USA, 2020.

Hinton G. E., Dayan, P. Frey B.-. J, Radford M. N.: -The wake-sleep algorithm for unsupervised neural networks Department of Computer Science University of Toronto6 King’s College Road Toronto M5S 1A4, Canada, 1995.

Simescu,N., Proiectarea construirea și exploatarea conductelor magistrale de transport gaze naturale, Ed. Lucian Blaga, Sibiu 2001.

Săndulescu,D.: Chimie fizică, vol I, Editura Științifică și Enciclopedică București, 1979.

Teuţan E., Rafa V., Analysis and fuzzy simulation of a pump with eccentric for natural gases odorized, Acta Technica Napocensis, 2018.

Atkins,P.W., Tratat de chimie fizică, Editura Tehnică București, 1994.

www.mathworks.com/products/ Design Neural Network Predictiv, A campus wide license is provided by Matlab to al staff and researches at TU Cluj-Napoca.

Rusu, A., Caracterizarea calitativă și condiționarea prin odorizare, separarea pe faze și uscarea gazelor naturale, Editura Lucian Blaga , Sibiu 200


Refbacks

  • There are currently no refbacks.


JOURNAL INDEXED IN :