The development of a knowledge based front end for a computational fluid dynamics package

Hartle, S.L., 1993. The development of a knowledge based front end for a computational fluid dynamics package. PhD, Nottingham Trent University.

10290065.pdf - Published version

Download (51MB) | Preview


The overall aim of this study was to establish a knowledge based approach to the preparation of data for complex computer programs. This was achieved through the development of a Knowledge Based Front End which interacts with a user to extract data, performs inference on this data and then synthesises the data to generate appropriate commands acceptable to the original program.

Initial development of a Knowledge Based Front End to a Computational Fluid Dynamics package, PHOENICS, using a commercial expert system shell, LEONARDO, was found to be inadequate. The limitations of the shell lead to the re-development of the Front End using the traditional Artificial Intelligence language, LISP.

LISP was used to create knowledge representation formalisms, data storage techniques and a purpose built inference engine for the target application. Knowledge representation formalisms included factual templates, objects and a specifically designed rule base language. The creation and implementation of inference networks reduced the number of rules the system needs to consider when using a specific rulebase. Base rules within each rulebase are used as the roots with which forward chaining commences. Antecedents that cannot be proved through forward chaining are then used as the goal for backward chaining throughout the associated inference network. The Knowledge Based Front End for PHOENICS improved the accuracy and consistency of the prepared data file. The system synthesises the user entered data and inferred data into appropriate PHOENICS commands to fully describe a computational analysis of fluid flow. A knowledge domain for jet impingement was used as a vehicle to demonstrate the concepts incorporated within the system.

The program architecture was carefully designed to enable future extensibility. Replacement, or extension, of the existing database and knowledge bases with new assertion templates, objects and rules, which would be inferred upon by the same inference engine is feasible. This potential for extensibility allows the system to be applied to different knowledge domains.

An important aspect of Computational Fluid Dynamics is the correct specification of the meshed geometry. Aspect ratios within the grid can have disastrous effects on the convergence of the solution and the accuracy of the results, and are therefore of paramount importance. A novel method of aspect ratio dependent finite volume grid generation is presented which utilises a generalised Fourier Series profile function. This technique ensures that given an arbitrary, one dimensional, region, its overall height, the minimum cell size, and the maximum allowed cell aspect ratio, the region can be meshed using grid clustering near a wall or within a duct. Meshing each axis as a one dimensional region enables a complete mesh to be obtained by superimposing the axes together. Within the final domain, the cell aspect ratio will not exceed the predefined maximum. Feasibility studies into the monitoring and control of the PHOENICS solution algorithm and results analysis through post processing grid optimisation, were performed. The potential for the latter two areas to be integrated into the KBFE looks promising.

Item Type: Thesis
Creators: Hartle, S.L.
Date: 1993
ISBN: 9781369323146
Rights: This copy of the thesis has been supplied under the condition that anyone who consults it is understood to recognise that its copyright rests with its author and that no quotations from the thesis and no information derived from it may be published without the author’s prior consent.
Divisions: Schools > School of Science and Technology
Record created by: Linda Sullivan
Date Added: 16 Jun 2021 11:37
Last Modified: 22 Sep 2023 13:40

Actions (login required)

Edit View Edit View


Views per month over past year


Downloads per month over past year