Software for on-line testing of pipeline modeling methods

Abstract

The paper considers the problem of comparative testing of mathematical methods for modeling the pipeline systems. Causes of existence of multiple alternative methods for solving the same problem are shown on the example of flow distribution problems. Many studies proposing new methods or their modifications give, as a rule, one or two numerical examples demonstrating advantages of proposed methods over others, which is surely not convincing. Practice shows that one method can be efficient for one example and fail for the other one. For this reason the issue of comparative testing of the availability and efficiency of computation methods for defining the areas of their preferred application is still open. Defined are the objectives, criteria and capabilities of such testing. The paper gives brief characteristic of the on-line IHCS software as a basic tool for gathering and interpretation of a large array of test examples, as well as techniques and initial results of its application. These results can presumably be used for forecasting the computational intensity of modeling the pipeline systems in specific computing environment using machine learning (ML) methods.