Comparing RWIND with ABAQUS, ANSYS, and Master's Thesis Based Model Test
The following article deals with the comparison of a reference study conducted by the Graz University of Technology in Austria  by means of an imitated simulation performed in RWIND.
Within the scope of the study, a cuboid was analyzed in a wind tunnel test with regard to pressure distribution. Then, the test was remodeled in ABAQUS and ANSYS, and results were compared.
Both the longitudinal as well as the transversal flow were analyzed using two flow profiles each. Because the results of both wind profiles are very similar, only the higher inlet velocity was selected for comparison.
As in the reference study, the Cp values were evaluated on the vertical center line for the longitudinal and transversal flow. In the experiment, sensor data of the central radiosondes were used. Below, the results are given in comparison. The points of the curves from the reference study were read out by the ENGAUGE Digitizer tool and therefore not connected to one another.
First of all, it can be seen that all three mesh densities correlate well in RWIND. Therefore, a denser mesh is not advisable. The three simulations performed in RWIND seem to hit the points of the experimental benchmark better than the simulations in ABAQUS and ANSYS. The deviation from the other two solvers is very large, especially at the medium height of the test object.
The Cp values' increase towards the ground is noticeable in ABAQUS and ANSYS. The experiment cannot confirm it due to the lack of sensors placed in that area, but the results seem physically illogical. However, this anomaly is not shown by RWIND. Overall, RWIND seems to better hit the experiment's Cp values, despite simpler approach functions and thus less computational effort.
In relation to the longitudinal flow, the same conclusions can be drawn for the transversal flow. The results of the comparison are closer to the experimental benchmark than for the longitudinal flow, but the RWIND results show a comparably good fit. Here, again, an anomaly of the Cp curves can be seen near the ground.
In addition to the evaluation on the vertical center line, the comparison provides false-color images for the Cp value tensor. It is therefore advisable to compare the Cp distribution on various surfaces.
The images of the experimental benchmark should be seen with caution. Due to the relatively coarse distribution of pressure sensors, the representation of the level lines is more an assumption than a measurement. The real Cp value is likely to be significantly lower at the edges, since there are no sondes in this critical area.
Note that only half as many discrete color-value ranges are available in the RWIND result display as in ABAQUS and in the experimental analysis, which is why it is less the colors than the corresponding value ranges that should be compared visually. Furthermore, for reasons of clarity, only the ANSYS simulation of the reference study is used below. ANSYS tends to have better results; however, overall the results are much more similar than, for example, ABAQUS and RWIND.
For the longitudinal flow of the front side, the findings from the plots of the center line are confirmed. The RWIND simulation seems to hit the specific Cp values over a certain area better than ANSYS, but the distribution is very similar in all sample images. The increase of values in the ground area in the ANSYS image described above is also shown here and is not confirmed by either the experiment or RWIND. Again, it is also noticeable that in all simulations the Cp value decreases sharply towards the edge of the object, but not in the experiment. The difference is probably due to the lack of sensors in this area. Overall, the simulation of the pressure side in RWIND is very convincing, especially with regard to the relatively low resources and modeling effort compared to other programs.
For the suction side, the pressure side's trends do not continue. While the level distribution of the Cp values in RWIND conforms slightly better to the simulation than the ANSYS simulation, the specific values are significantly lower. With the exception of the edge on the flow side, the level distribution in RWIND tends to comform better to the tests than ANSYS. In both CFD simulations, the local minima are on the front edge but are clearly concentrated in ANSYS and distributed over a certain area in RWIND. In ANSYS it is about -1.2, whereas RWIND is again considerably lower at -2.47. Here, ANSYS clearly shows the more physical behavior. At the end of the flow's far side, this trend is directly reversed. Here, RWIND and ABAQUS are below the value of the tests. Again, however, the observation is doubtful because of the pressure sonde distribution. The Cp minimum in the ground area of the ANSYS simulation follows the trend already described. Overall, RWIND seems to provide worse results than ANSYS or ABAQUS on the sides dominated by suction. The simpler element approach could possibly be responsible. An unsteady flow analysis in RWIND could address, for example, the clearly overestimated suction areas.
In fact, the results in the RWIND's unsteady flow simulation are much better. Both level distribution and point values show good conformity with the wind tunnel tests. However, a comparative evaluation using ANSYS would make no sense, since a stationary flow model is available there.
For the rear side, representing another suction area, the differences are even greater. While ANSYS and the tests show good conformity, RWIND calculates significantly lower Cp values and spreads them, according to the amount, higher than the other two images. Here again, an unsteady flow calculation provides significantly better results comparable to ANSYS, but this was not the subject of the comparison.
The knowledge gained from the comparisons of false-color images showing longitudinal flows also applies to the transversal flow. The comparative images of the lateral flow are shown below but are not described in detail due to doubling.
Do you have further questions or need advice? Contact us via phone, email, chat, or forum, or search the FAQ page, available 24/7.
AIJ Examples Case E - Building Complex in Actual Urban Area with Dense Concentration of Buildings in Niigata City
Creating a validation example for Computational Fluid Dynamics (CFD) is a critical step in ensuring the accuracy and reliability of simulation results. This process involves comparing the outcomes of CFD simulations with experimental or analytical data from real-world scenarios. The objective is to establish that the CFD model can faithfully replicate the physical phenomena it is intended to simulate. This guide outlines the essential steps in developing a validation example for CFD simulation, from selecting a suitable physical scenario to analyzing and comparing the results. By meticulously following these steps, engineers and researchers can enhance the credibility of their CFD models, paving the way for their effective application in diverse fields such as aerodynamics, aerospace, and environmental studies.
- How can I connect surfaces to other surfaces or members in a hinged/semi-rigid way? What are Line Hinges and Line Releases?
- How is the dimension for an equivalent round section calculated in the Timber Design add-on?
- I have a 3D model and try to split it up into "walls" and "surfaces" with the Building Model and to calculate them separately. However, my structural system becomes unstable then. What can be the reason?
RWIND 2 is a program (digital wind tunnel) for the numerical simulation of wind flows around any building geometries with determination of the wind loads on their surfaces. It can be used as a stand-alone application or used together with RFEM and RSTAB for complete structural analysis and design.
RWIND is a program (digital wind tunnel) for the numerical simulation of wind flows around any building geometries with determination of the wind loads on their surfaces. It can be used as a stand-alone application or together with RFEM and RSTAB for complete structural analysis and design.