:   .., ..
:  
:  107
:  
:  2024
:   .., .. // . 107. .: , 2024. .66-87. DOI: https://doi.org/10.25728/ubs.2024.107.4
:   , , , , ,
(.):  mesh model, transient heat conductivity, adaptation, gradient descent method, algorithmic complexity, computational stability
:   . , , , . , , , 64%. , 36% . , . , 64% 98%. . , , - .
(.):  This study is devoted to the analysis of algorithmic stability and complexity of the model of transient heat conductivity with implicit adaptation to the thermophysical parameters of the heated solid body. The implicit adaptation method is based on the substitution of such parameters as heat capacity, thermal conductivity and density in the transient heat conduction equation by two dimensionless adjustable coefficients, uniformly discretized over the entire model lifetime, with their further adjustment using a modified stochastic gradient descent method. In order to ensure the stability of calculations of such a model using a computer, some conditions have been defined in previous studies, which allowed us to obtain stability equal to 64%. It was assumed that the remaining 36% was a consequence of violation of these conditions in the process of adjustment. In this paper we propose algorithmic constraints that allow us to solve this problem. The repetition of experiments shows that the application of the proposed approach allows one to increase the stability from 64% to 98%. Also, an analytical comparison of algorithmic complexity classes for models with implicit adaptation and with "group-explicit" adaptation is made. As a result, it is found that the proposed numerical method has a lower complexity in comparison with the finite-difference method with "group explicit" adaptation.

PDF

: 425, : 123, : 9.


© 2007.