Application of the neural network training procedure to reconstruct an unknown influence on the temperature distribution of the boiler drum wall

The paper proposes a novel method for reconstructing an unknown input influence based on output measurements with a known dynamic model of the system. A feedforward neural network acts as a device for forming the vector of reference values of the unknown input influence. The dynamic model of the sys...

Full description

Saved in:
Bibliographic Details
Published in:Vestnik Samarskogo universiteta. Aèrokosmičeskaâ tehnika, tehnologii i mašinostroenie (Online) Vol. 24; no. 3; pp. 171 - 183
Main Authors: Danilushkin, I. A., Diligenskaya, A. N., Kolpashchikov, S. A.
Format: Journal Article
Language:English
Published: Samara National Research University 23.10.2025
Subjects:
back propagation algorithm
feed forward neural network
identification
input influence
object with distributed parameters
steam boiler drum
temperature distribution
ISSN:2542-0453, 2541-7533
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The paper proposes a novel method for reconstructing an unknown input influence based on output measurements with a known dynamic model of the system. A feedforward neural network acts as a device for forming the vector of reference values of the unknown input influence. The dynamic model of the system is included in the loop of formation of the error vector, which is minimized in the neural network training. The error vector is formed anew each time after the next epoch of neural network training. For this purpose, an intermediate estimate of the input influence is fed into the input during the simulation of the system's dynamic model. The paper proposes a method of error vector formation for the realization of the back propagation algorithm when training a neural network by the mismatch between the model output and real data. A set of hyperparameters of the training procedure is defined. As an example, we consider the problem of reconstructing an unknown control influence based on data from the temperature change process in the metal of a steam boiler drum. The example uses a model of temperature distribution across the thickness of the drum wall, which takes into account the influence of the ambient temperature at the boundaries of the spatial coordinate domain. The results of input influence reconstruction are presented, demonstrating the performance and sufficient accuracy of the proposed method.
ISSN:2542-0453
2541-7533
DOI:10.18287/2541-7533-2025-24-3-171-183