Encyclopaedia Index


April 1998







Sture Lindahl

Jonny Svensson










Table of contents


1. Abstract

2. Introduction

3. Working with the VR system

4. The Cases

5. Working with the MICANET

6. Results

7. Appendix A. Validation, questions and answers

1. Abstract

The MICA system gives the user the opportunity to create CFD cases locally and to run them as PHOENICS simulations on a distant computer where expert assistance is available. Results are presented locally.

In this project, the system was tested carefully by about ten companies and institutions with staffs having different experience using CFD models.

By SMHI and its consultant CFE three oceanographicak/hydraulical cases were set up for tests. The results and conclusions from the SMHI tests are presented in this report, whereas the activities by CFE are presented in a separate report.

The thorough tests of the MICANET used for job submission and data retrieval have shown that the system in its final version is working in a reliable manner. It can even deal with so called Proxy Firewalls, found at many companies with high security de-mands. One remark is that it seems possible to choke the MICANET by sending a lot of cases shortly after each other.

The VR system being used to create cases and to study results has improved and is now fast and easy to use, although the graphical presentation could perhaps be a bit more powerful.

Comparisons with measurements and with simulations on local PHOENICS systems show that cases solved in the MICA system give correct results.

2. Introduction

The MICA project started with an introduction to the system on a meeting at CHAM in Wimbledon in January 1996.

The purpose of the project was to establish and thorougly test a system for local creation of CFD cases, solution of the cases on a distant, powerful computer and retrieval and local study of the results.

Head of the project has been CHAM in Wimledon, assisted by LITEC in Zaragoza and NTUA in Athens. The system has been tested by approximately 10 institutions and companies all over Europe, creating and running their own cases. As an extra test an architect, an experienced CAD user, has checked the system at SMHI.

After a prolongation, the finish of the project was set to April/May 1998.


3. Working with the VR system

Figure 1. The VR Editor with a simple testcase

In the MICA system cases are created in the VR Editor and results are studied in the similar VR Viewer. The Editor gives a wide range of opportunities to design shapes, materials, boundary conditions, solution options etc and the Viewer allows the user to study results as vectors and contours, from different angles, with desired magnification etc.

During the run of the project numerous tests have been carried out with different improved versions of the VR system. The latest version is almost finished with only minor changes to be done. But still one could expect the graphical presentation to be a bit more powerful.

The results of the tests are presented as answers to a lot of questions given by CHAM in the spring of 1997. This material can be found in appendix A.

The architect found the VR Editor/Viewer quite easy to use. After a brief introduction he was able to design his own case asking only a few questions and he had no problems studying his results in the Viewer.


4. The Cases

Two cases were designed for thorough tests and comparisons by SMHI. A third case was designed and studied by the consultancy company CFE. This case will be presented in a stand alone report.

Figure 2. VR representation of the strait Öresund between Denmark and Saltholm.

Length=8500 m ,Width=5000 m , Depth=13 m . Magnified 50 times in the depth direction.

In the area between the cities Malmö in Sweden and Copenhagen in Denmark intensive oceanographical investigations have been carried out in connection with the planning of a road and railway link between Sweden and Denmark. From this region data are available both from continously registrating current meters and from detailed surveys by research vessels.

The area chosen for simulation is the strait between Denmark and the island Saltholm (figure2) where a current meter is moored close to the lighthouse Nordre Röse. The VR representation of the area is designed using the depth data base at SMHI, representing it by "cubes", wedges and spheres from the "SHAPES" library. The figure shows the slope from Denmark on the left side and the slope from Saltholm on the right side. In the middle is the so called Drogden chann el with one rock in it. The Lighthouse with the current meter is situated on the shelf in the upper left part of the VR representation.

Solving this case and comparing vertical velocity profiles from PHOENICS with those registered at the Lighthouse will give information about the success of describing the boundary layer at the bottom. Furthermore, interesting information about the accuracy of the description of turbulence should appear.

Figure 3. VR representation of a case with a jetty (left), outlets, a ship (green), land reclamation(right) etc. Length=1650 m , Width=640m , Depth=9m . Magnified 50 times in the depth direction.

Important and common objects when working with oceanographical cases are land reclamation, outlets, ships hulls, jetties etc. In the case shown in fig 3 a set of such objects are set up in a simple channel (design by CHAM and SMHI). Running this case with a steady flow through the channel will give concentration patterns from the two outlets situated at the jetty to the left half way through the channel. These patterns will be compared to patterns from an identical case set up and solved locally at SMHI in PHOENICS 2.1.3.

Comparing velocity profiles from the two cases will give information about the reliability of the simulation of the boundary layer at the bottom and of the turbulence.


5. Working with the MICANET


Figure 4. The MICA Job Control "Submit case"

One of the most important features of the MICA system is the MICANET. It provides the user with a tool for submitting cases to a computational center and receiving results after simulations.

Being able to design cases in the VR-editor and submitting them to a computational center should mean a possibility of introducing CFD simulations to a wider range of users than today. The user only needs to understand the handling of a VR interface, but not the far more complicated task of running his own PHOENICS simulations.

The first successful submission of a MICA case was demonstrated to the members of the MICA project at the meeting in Zaragoza in April 1997. Shortly after that meeting a new version of the MICA system was shipped to the members. This version enabled them to begin their own submission tests during the summer of 1997.

At SMHI it soon became clear that the MICANET did not work . After unsuccessful tests and extensive contacts between SMHI system managers and system managers of the MICA project the problem was isolated to the so called Proxy Firewall at SMHI. One way of handling this problem was by making a "hole" in the firewall. This gave the opportunity of submitting cases but could not be accepted for security reasons. Since Proxy Firewalls are not uncommon and can be expected to exist at f uture MICA users it was decided to go on trying to get a general solution to the problem. This process with different tests and changes of codes and configurations continued until finally an operational set up was found in February 1998.

During the period when trying to reach a solution of the "Firewall problem" numerous abortive test submissions were carried out. After the successful set up of the connection about 80 submission tests were run. The intentions of these runs were, among other, to check the connections, to test the system for submission and for result retrieval, to see how long one had to wait for results and to evaluate the usefulness of expert intervention.

The MICANET system allows the user or the system to choose among four computational centers. At the moment only two of them, CHAM and INRIA, seem to be operating, Most of the submissions and other tests carried out by SMHI were directed to CHAM. In the few directed to INRIA there always occured some kind of errors. When connecting to CHAM almost all of the submissions were successful. Only at very few occasions did the server refuse to respond.

Submission and data retrieval using the MICA Job Control was straightforward and easy to use. The best way of retrieving data was to use the Job Control. Getting the results by e-mail forces the user to carry out two extra moments: to move the result file from the mail to the working directory and to unpack it. When saving results, one has to be very careful to get them into the considered library.

Figure 5. The Mica Job Control "Status of Jobs"

61 "common" cases were submitted. Less than 10 of these failed, mainly due to problems when asking for "expert intervention". The problems related to the request for expert intervention were caused by server problems at CHAM. These problems were solved and finally some successfull runs including the expert intervention request were carried out. The results from those cases were normally accompanied by informative mails.

A MICA server could be expected to get a lot of submissions from different users during a short period of time when the system is in full scale operation. To simulate such a situation 20 simple cases were submitted shortly after each other, the time step between them being approximately 30 seconds. 14 of these cases were successfully solved while 6 cases failed.

One way of checking the advantage of submitting a case via MICANET compared to running the simulation on a local PC is to compare the time for solution of identical cases. Two cases were designed for this test, a simple one using 30 minutes for solution on a PC and an extensive one using 570 minutes for solution on a PC. The times for running those cases on MICANET were 2 - 3 minutes and 10 - 15 minutes respectively.

For further comments about tests etc see Appendix A.

The testing architect found that Submission of cases and Data retrieval was easy to handle.


6. Results

- The first case, describing the flow in the strait Öresund, converged after 100 sweeps. The simulations were run both on a local PC and on the MICANET. Those simulations gave the same results although carried out 35 times faster on the MICANET. The driving forces were an incoming current of 0.4 m/s on the southern border and a northeasterly wind of 10 m/s.

Figure 6. The case Öresund, Current field at the surface

The current field at the surface after 100 sweeps is presented in figure 6. It shows a distribution of the currents beginning with the speed 0.4 m/s on the boundary in the south. Correctly it then rises when going northward reaching its highest value, 0.9m/s, halfway through the channel. In this area the channel presents its narrowest section with a small depth, narrowing slopes and a rock in the middle. Moving further northwards the channel widens and the current speed decreases reaching its lowest value on the wide northern boundary.

Figure 7. Current profiles at Nordre Röse Lighthouse. Measured at an almost steady situaion (blue line) and simulated in the MICA system (red line).

Figure 7 compares the measured and the simulated current profiles at the lighthouse Nordre Röse (for location see figure 2). Although not of exactly the same magnitude the profiles show a good correspondence both in slope at the bottom and closer to the surface. This indicates that the logaritmic boundary layer and the turbulence are correctly described in this simulation.

- The second case describes an idealized channel with jetties, outlets etc. This case was set up both in the VR system and on a PHOENICS system at SMHI. Steady and time dependent runs were carried out. It was of great interest to compare simulated concentrations of species from the outlets.

The concentration pattern of species from one outlet simulated at SMHI is presented in

figure 8(top). The outlet is situated at the end of the jetty, halfway between the bottom and the surface and directed in the x-direction. 15 m3/s are let out. The concentration is 100 at the outlet. This is a steady case which has been run 100 sweeps. The corresponding results from the MICA run are found in figure 8(bottom).

It is obvious that the results are almost similar.The differences are probably due to use

of different convection schemes. Assuming that the case set up and solved at SMHI is correct

leads to the conclusion that the MICA solution is reliable.

Comparing the current profiles shows that the logarithmic boundary layer and the turbulence

are treated correctly in both solutions.

Figure 8a. Concentrations simulated with PHOENICS 2.1.3

Figure 8b. Concentrations simulated with MICA system.