HYDRAULIC CASE STUDY OF A VAULTED COMPOUND CHANNEL

Reinhard Prenner¹, Boris Huber¹ and Helmut Drobir¹

¹ Institute of Hydraulic Engineering, Vienna University of Technology

A-1040 Wien, Karlsplatz 13, Austria/Europe

E-mail:  reinhard.prenner@kw.tuwien.ac.at, 

Tel: +43-1-58801-22244,  Fax: +43-1-504-5928

Abstract: About 100 years ago the Wien River was regulated in an outstanding engineering achievement. The approximately rectangular cross-section of the river was constructed of concrete and partially of stone. Now, rehabilitation works along the 13-km-long river have made it necessary to investigate the hydraulic discharge capacity of a 2.1-km-long vaulted stretch during flood events. To calculate the water surface profiles caused by different flood loads an one-dimensional computer program was used and to verify the calculation assumptions with the actual flow behavior a straight hydraulic model was built. It had a scale of 1:15, a width of 1.4 m, a length of 15 m and operated with different slopes (1.5 %o and 3 %o). Friction losses in the planned ¡°new¡± river bed were determined. The influence of mixing effects in the interaction region between the rough main channel and the smooth overbank on the velocity distribution was investigated additionally. A variant study with a reduced cross-section - caused by the insertion of a ceiling to create an additional multipurpose area - was carried out as well. The suitability of application of the 1-dimensional calculation with the results of the experiments will be described and evaluated.

Keywords: channel hydraulics, compound channel, velocity profile, flow resistance, fricton factor, roughness parameter, hydraulic model, one-dimensional modelling

1    INTRODUCTION AND AIM OF THE INVESTIGATION

General: The planned improvement of flood protection measures and the ecological renaturalization of the Wien River comprises several river sections over a length of about 13 km. This revitalisation project should also improve the water quality in the urban area during heavy rainfalls by a new sewage system within the existing cross-section, create recreational areas and enhance the urban environment. In 1999 the Institute of Hydraulic Engineering at the Technical University Vienna was commissioned by Vienna¡¯s Department of Water Resources to investigate hydraulic efficiency during flood events in the 2.1-km-long and 17 to 21-m-wide vaulted section of  the Wien River in the inner city of Vienna (Fig. 1).

Flood protection demands are very important in urban areas because of the high damage potential. In order to direct river floods safely through the city, it was necessary to reduce peak runoffs by retention basins. These existing retention basins upstream of the Wien River had been enlarged and reconstructed in the past few years. The total retention system is controlled by a computer system which takes into account the total runoff regime and the water level in the individual basins according to the weather forecast. The normal average discharge of the Wien River is about 800 l/s, the annual flood is about 30 m³/s. In the case of a discharge of more than 244 m³/s (100-year return period), the sewer canals in the vaulted section will be used as bypasses for flood flow, so that the design flood with a return period of 5000 years is about 380 m³/s in the vaulted section.

Fig. 1    Typical cross-section of the Wien river in the vaulted section before and after rehabilitation, variant with a multipurpose area

Problem: The present investigation is concerned with the flood discharge capacity of an unusual vaulted compound river cross-section because there is still uncertainty regarding the influence of different roughness subdivisions on the total flow capacity. For the calculation of the water levels during peak runoffs, a 1-dimensional computer program, HEC-RAS 6.2, was used. This program allows computation of the discharge using a mean conveyance of the entire cross-section (Horton-Einstein) or the summed conveyances of all incremental sections in the overbanks and the main channel. Friction losses are calculated using the Gaukler-Manning-Strickler formula. Because of the empirical background of the roughness coefficients (k_st) great care must be taken herein.Therefore it was necessary to determine the roughness parameters of the ¡°new¡° river bed in hydraulic experiments to obtain precise input data for the calculations. An important question was furthermore, whether the assumptions of uniform velocity distribution in the 1-dimensional calculation are valid especially for higher discharge rates.

2    BASIC ASPECTS

Steady flow in open compound channels has been investigated in numerous studies. The literature on open channel hydraulics is extensive. There have also been many attempts to simulate flow in a two-stage river cross-section with sophisticated 2-d or 3-d numerical models. Usually, for steady or unsteady hydraulic calculations of longer river reaches or complicated networks, more or less advanced 1-dimensional models are used. In all studies known to the authors, the conditions of the published investigations were not directly suitable for and comparable with the given problem.

According to the 1-dimensional theory water surface profiles are computed from one cross-section to the next by solving the energy or momentum and continuity equations with an iterative procedure. HEC-RAS software computes a single water surface profile with a mean energy head in each cross-section. The mean energy head is obtained by computing a flow-weighted energy from the subsections by a velocity oefficient a (Fig. 2).

The energy head loss between two cross-sections consists of friction, contraction and expansion losses. Friction loss between two sections is taken into account by Manning-Gauckler-Strickler¡¯s formula.  Strickler¡¯s mean composite roughness coefficient k_st for each overbank and the main channel is calculated by the Horton-Einstein formula. To determine the roughness of a natural river bed, Strickler proposed following empirical correlation for  k_st=21/⁶Öd_m. Actually, this formula is impaired by some uncertainties, therefore experimental investigations were carried out to obtain more reliable and accurate data on the roughness of the river bed. Additional velocity measurements in a full model show the influence of the interaction zone between both sections with distinct different roughness.

Fig. 2    Theoretical mean kinetic energy head in an assymetric compound channel

3    EXPERIMENTAL INVESTIGATIONS

The investigations were carried out in our labratory in two parts. Part I was a pre- experiment (scale 1 : 10) to determine the bottom resistance (grain and bed form roughness) of the „new¡° river bed in a glass-walled tilting flume 0.5 m in width and 20 m long. Part II aims to verify the results of the first investigation and also show the actual flow conditions (velocity distribution) in the compound channel. This hydraulic model (scale 1 : 15, 1.4 m in width and 15 m long) was also built in a tilting flume (Fig. 3).

Fig. 3    Experimental setup of the vaulted two-stage river cross section in the flume

The artificially armored riverbed was modeled by larger and smaller quarry stones. The riprap lay on a structured sand bed which gave the basic bed form. Gaps between quarry stones were only partially filled with finer substratum to minimize sediment transport. Structural elements, such as small ponds were connected with a meandering secondary low-level river bed for fish migration. Between these, bed groynes were formed by larger quarry stones. The riprap of the new river bed is dimensioned to be stable for floods up to the design flood and slightly more. An impression of the actual conditions can be gained from Fig. 4. 

This riprap-covered river bed had a thickness of 1 to 2 stone layers with an average weight of 35 kg/m² to 55 kg/m². The quarry stones were sharp-edged. The grain-size distribution of the riprap lay in a range of 13mm to 40 mm (equivalent diameter of a sphere). The average sphere diameter (d₅₀) of the smaller stone fraction was 19 mm; that of the greater one was 25 mm.

Fig. 4    Significant cross-section of a compound channel, view to upstream

The hydraulic model was built according the application of Froude¡¯s law of similarity for steady flow conditions. The similarity of the flow process between nature and model could be reproduced very well due to the high roughness of the river bed and the Reynolds numbers from 200000 to 800000 (fully developed turbulent flow). The significant discharges that were investigated can be seen in Tab. 1.

    Table 1    Flow discharges in nature and model depending on return period of floods

Fig. 5   Compound channel ¨C intake to tunneling section, view to downstream

Measurement of the flow discharges was carried out via a magnetic inductive flow meter. Velocity measurements were taken by an ultrasonic velocity profiler (UVP- MET-FLOW, SA,) by means of 3 ultrasonic transducers. Measurement profiles were located after a distance of about 10 m from the channel entrance where uniform flow has already developed. The ultrasonic probes had to dive into the water for 1 cm during the measurements and were fixed to a sliding bar. Velocity profiles in the conduit section were taken at the contraction zone of the intake and about 3m downstream of this profile. Further investigations are under preparation on a full hydraulic model (scale 1:50, length 60 m) to observe flow behavior in curves, sudden expansions in cross-sections and more.

4    TEST RESULTS

Recalculations of the model pre-tests in the glass-walled flume yield a riverbed roughness (after Strickler) of k_st=41. Water surface profiles in the full model were computed with the HEC-RAS program with following input data: riverbed k_s=41 (in nature k_st =26), surface of the bicycle trail, walls (vault) with k_st=95 (in nature ks=60) and ceiling k_st =125 (in nature k_st =80). Calculated mean k_st-values of the compact compound channel are depicted in Fig. 6.

Fig. 6    Mean k_st¨Cfactors depending on discharge and water depth

Velocity measurements in the open channel profile under uniform flow conditions with lower discharges showed insignificant higher shear friction losses in the contact region between trail and riverbed than increasing discharges (Fig. 7). Seperate calculations of velocity profiles according to the theoretical log- profile are suitable in the lower region of the river bed.

In case of a pressurized discharge in the conduit section larger shear stresses occur in the interaction region than in the open channel flow (Fig. 8). The reason of this increase could be found in the development of the velocity profiles in the smooth and the rough section. When flow gets pressurized a sudden change of hydraulic radius takes place. Nevertheless, the calculations were satisfying for practical use.

Control calculations with HEC-RAS have shown that the experimental results correspond very good to results computed with the compact cross-section method. Subdivided calculation yields a slight difference to the actual mean velocity distribution coefficient a. Influences of  the vault on the flow behavior in the constantly reducing cross-section leads only to an increase of the waterdepth in the investigated range of discharges.

Fig. 7    (a) Open channel flow, velocity distribution, Q=133 l/s, I_e=3%o, a=1.09
(b) Open channel flow, velocity distribution, Q=357 l/s, I_e=3%o, a=1.05

Fig. 8    Conduit flow (with ceiling), velocity distribution, Q=438 l/s, a=1.15

5    CONCLUSIONS AND PRACTICAL IMPLICATIONS

To ensure an accurate calculation of water surface profiles in the Wien river by an 1-dimensional computerprogram it was necessary to carry out hydraulic model tests to get confidential input data and to validate the assumptions in the calculation. The use of this commercial 1-d computerprogram HEC-RAS showed that it is possible to obtain very reliable results in combination with basic experiments. Confidient input data are most important for the results. If the shear stresses in the interaction zone are unconsidered, calculation as a compact cross-section (Horton-Einstein with a mean k_st-value) give slight better results than with subsections. Influences of the velocity distribution (a-coefficient) on the computed results are negligible for the investigated compound channel or for similiar cross-section conditions. Roughness (grain and bedform) parameters of the structured river bed were determined  and verified in 2 experiments (model k_st=41). Pressurized flow could be simulated with the actual roughness parameters in the program as well.

References

[1]    Ladinig G.: Hochwasserschutz und Ökologie am Wienfluss. Aqua Press International 1/2000, pp. 6-9.

[2]    US Army Corps of Engineers, Hydrologic Engineering Center: HEC-RAS River Analysis System, Hydraulic Reference Manual, Version 2.2 September 1998.

[3]    Naudascher, E.: Hydraulik der Gerinne und Gerinnebauwerke. Springer Verlag, 1992.

[4]    Met-Flow SA, UVP Monitor Model UVP-XW: User Guide, Release 1, March 2000, Lausanne, Switzerland.