Abstracts

Novosibirsk participants

The steady problems of calculation on piping systems (water-supply networks, heat-supply systems etc.), as a rule, are reduced to solution of the algebraic equations' systems, defined on the graph. Both direct problems (calculation of flow distribution in the pipes of system) and inverse problems, such as optimal pipe diameters' finding, and identification problems (definition of a pipe wall roughness by flow parameters' measurement data) are considered.

The usually used "classical" approaches are confronted with difficulties caused by essential nonlinearity of the equations. In addition, the solution of the above mentioned inverse problems is reduced to the not simple problem of the multivariable function minimum search.

The application of modern information technologies and advances in the field of artificial intelligence allows to find the new non-traditional approaches to solution of these problems, which prove to be more effective in a number of cases.

The offered approach uses the concepts of the subdefinite computational models, constraint programming etc., appropriate declarative languages and such integrated software supporting them, as "UniCalc" and "NeMo++" systems. This approach, based on new interpretation of concept "solution", at which the process of solution is interpreted as the process of lowering of a initial level of uncertainty of the problem. The user is released from algorithm making for solution within the framework of such approach, he is only required to describe computational model successfully by the system input language.

The examples of calculations and computing experiments' results are given for different classes of the practical problems.

Note. Abstracts are published in author's edition

Comments

[Home] [Conference]

© 1996-2001, Institute of computational Techologies SB RAS, Novosibirsk
© 1996-2001, Siberian Branch of Russian Academy of Science, Novosibirsk