Predicted and Measured Flow Patterns In Channels Using

Advances in computer speed and availability of user friendly software has made computational fluid dynamics (CFD) a cost effective compliment and at times alternative to physical modeling in the field of civil engineering hydraulics. The use of CFD is wide spread in mechanical and aerospace engineering, however, in the field of civil engineering and particularly in hydraulics, there is minimal utilization of CFD. A lack of validation and test cases, and the potential added complexity of 2-phase flow models required for some free surface flow problems, appears to make hydraulic engineers reluctant to accept modeling results. Validation of CFD models for open channel flow conditions remains limited due to the significant cost of obtaining data. This is aggravated by the fact that once a flume study has been conducted, a numerical study typically appears redundant. This presentation demonstrates the strengths and weaknesses of CFD in reproducing free surface flow in a curved trapezoidal channel. Laboratory data was collected at MIT during the late 1950's and early 1960's to obtain an understanding of shear stress distribution and flow patterns in bends, a portion of those studies was reproduced using CFD and a comparison was made between observed and predicted values. The presentation further demonstrates that CFD can offer a cost effective alternative and compliment to physical modeling of some smooth, rigid boundary conditions. Extensive flow visualization is used to show how CFD can offer insight into flow phenomena that are difficult to obtain using only laboratory data. The CFD software Fluent was used to model the flow in the trapezoidal channel using two turbulence models, K-Epsilon and Reynolds Stress. Model results are compared to each other to assess the differences in predicted flow due to the selection of turbulence model. In addition, model results are compared to flume data to assess the overall predictive capabilities of the numerical model. Model results show that observed and predicted water surface elevations typically differ by less than 2.5 percent while predicted shear stresses differ by less than 10 percent. Furthermore, results illustrate the limitations of the K-Epsilon model as well as conditions under which it produces very similar results to the computationally more expense Reynolds Stress model. Results of the study clearly demonstrate the strengths and weaknesses of using CFD as an alternative or compliment to physical modeling.