DESCRIPTION AND EVALUATION OF PROGRAM: DUFLOW ... - PubAg

D E S C R I P T I O N AND E V A L U A T I O N O F PROGRAM: DUFLOW By A. J. Clemmens! and F. M. Holly Jr. 2, Members, ASCE, and W. Schuurmans3 ABSTRACT: DUFLOWis a microcomputersoftware package for simulatingonedimensionalunsteady flow in open-channelsystems. The program is designed for simple networks of channels with simple structures. Water levels and flow rates are determined by solvingthe St. Venant equations of continuityand momentum with the Preissmannscheme. The program has a user interface for conveniently entering a descriptionof the network conditions.DUFLOW is public-domainsoftware and is distributed at a nominal cost. Although the program cannot handle some of the more-sophisticatedmodelingneeds (e.g., dry beds, automatic gates), it is useful for first-timeusers of canal-networksoftware. Limited user support is available.

INTRODUCTION

The program D U F L O W (from "Dutch flow") was examined by the ASCE Task Committee on Irrigation Canal System Hydraulic Modeling. The main computational procedures in D U F L O W are based on IMPLIC, a mainframe-based F O R T R A N program written in the 1970s by the public-works department (Rijkswaterstaat) in The Netherlands. The Delft University of Technology, the International Institute for Hydraulic and E n v i r o n m e n t a l Engineering (IHE), and Rijkswaterstaat collaborated in developing D U F L O W to provide a more user-friendly, free-surface-flow program for general use. The program is intended for use in all types of open channels (e.g., rivers, navigation channels, drainage canals), not just irrigation canals. The IMPLIC computational code was combined with a user interface written in QBASIC to produce D U F L O W . The program is available from the Bureau S A M W A T in The Netherlands (DUFLOW 1989) and only for MS-DOS operating systems. This paper describes the D U F L O W model according to the evaluation categories given in A S C E (1993). TECHNICAL MERIT

Computational Accuracy A four-point implicit Preissmann scheme (Cunge et al. 1980) is used to solve the complete St. V e n a n t equations of continuity and m o m e n t u m . The user can select solution of linearized or fully nonlinear versions of the equations, the latter being solved with a Newton-Raphson-type scheme that starts from the linearized results. External and internal b o u n d a r y conditions are iRes. Hydr. Engr., U.S. Water Conservation Lab., 4331 E. Broadway, Phoenix, AZ 85040. 2prof. of Civ. and Envir. Engrg. and Res. Engr., Iowa Inst. of Hydr. Res., Univ. of Iowa, Iowa City, IA 52242. 3Res. Assoc., Delft Univ. of Technol., Dept. of Civ. Engrg., P.O. Box 5048, 2600 GA Delft, The Netherlands. Note. Discussion open until January 1, 1994. To extend the closing date one month, a written request must be filed with the ASCE Manager of Journals. The manuscript for this paper was submitted for review and possible publication on March 3, 1992. This paper is part of the Journal of Irrigation and Drainage Engineering, Vol. 119, No. 4, July/August, 1993. 9 ISSN 0733-9437/93/0004-0724/$1.00 + $.15 per page. Paper No. 3522. 724

expressed and solved within the Preissmann scheme. The linearized option nearly always gives results very close to the nonlinear results. The set of equations is solved by Gaussian elimination so that any network or loop configuration can be included. D U F L O W allows the user three options related to the flow inertia. The user can choose to retain the Froude term (inertia fully accounted for), eliminate the Froude term (zero-inertia), or use the Froude-term effect not to exceed frictional-resistance effects (damped). This term has a small impact for large, slow-moving irrigation canals, but it can be significant for highervelocity canal flows. Removing the inertial terms allows waves to travel only in the main flow direction, and it removes reflected waves from downstream structures. Numerical Solution Criteria Mass conservation is always guaranteed to within the computational precision of the machine when the nonlinear solution option is chosen. With the linear option, mass conservation is nearly always well within 1% and does not accumulate over time. Conservation of momentum is similarly guaranteed with the nonlinear option but applies to the cells of the Preissmann scheme. The Preissmann scheme has been demonstrated to be consistent with the governing equations through comparison with other, moremathematically precise, methods (e.g., method of characteristics). The Preissmann scheme has significant advantages over the method of characteristics where boundary and internal conditions exist, such as gates and weirs, and is therefore well suited to irrigation-canal-system modeling. Its disadvantage is that it tends to smooth out or attenuate real phenomena (Strelkoff and Falvey 1993). Robustness DUFLOW is designed to handle mild unsteadiness; and it always provides a solution as a result of the methodology employed. High amounts of unsteadiness or time and distance increments that are too large may cause the program to give poor results, but a solution will nearly always be provided. An exception is the dry-bed problem, which causes the program to cease computation and warn the user of this condition. No program logic exists for smoothing out irregularities in the solution. Error messages are provided if the computations fail. Initial Conditions Users of D U F L O W must specify the initial water levels at all defined nodes within a system. The user may also specify discharge values at the nodes. The recommended approach is not to specify discharges and to allow DUFLOW to determine discharges through a brief initial simulation period. DUFLOW cannot start wifh zero flow depth at any point within the system. DUFLOW does not include a separate steady-flow solution procedure. The unsteady procedure provides results that approach steady results over time. Internal and External Boundary Condition Analysis As mentioned earlier, boundary conditions, internal or external, are modeled as relations between head and discharge and are expressed in linear form, just like the continuity and momentum equations. Thus, they provide no particular difficulty in terms of convergence and stability. This is significantly better than explicit procedures, which use values of depth or dis725

charge from previous time steps. Problems can be encountered when a change in flow conditions results during the iterations such as unsubmerged to submerged gate flow or flow reversals. The linearization assumptions are not valid for these conditions and can result in convergence failures. Overall, the Preissmann implicit scheme handles internal and external boundary conditions with less difficulty than other methods.

Special Hydraulic Conditions No model can handle every hydraulic condition that can exist in nature. The D U F L O W program cannot handle advance of water on a dry bed, full dewatering of the channel, supercritical flow, hydraulic jumps, or hydraulic bores. D U F L O W can handle reversal of flows. MODELING CAPABILITIES

System Configuration There are no real limitations in the layout that can be employed in DUFLOW. The user defines nodes and is free to connect any node with any other node by any number of channel reaches or structures. Looping is automatically accounted for with no difficulty. Multiple structures at a location also pose no difficulty. Weirs and orifices are the most common type of structure modeled. Culverts, siphons, and pumps are also modeled as structures. Culverts require the addition of a frictional-resistance coefficient. Siphons need, in addition, starting and stopping levels upstream and downstream. Pumps are assumed to operate at a constant, positive discharge. Cross sections are defined at each node in terms of top width of flow at given depths. D U F L O W also has the capability of water storage, which does not contribute to the flow cross section. This water storage is included in the equations of continuity, but does not contribute to flow rate, velocity, or momentum. D U F L O W does not have any standard cross-section shapes to choose from. Longitudinal channel sections are defined by start and end nodes with associated invert elevations (rather than invert slope). The user must specify cross-section properties at each node in the system. If a pool or canal reach between structures is too long to model as one computational reach, D U F L O W provides no mechanism for defining nodes between user-defined nodes. All nodes are user defined. Thus the user should have a good idea of whether multiple nodes are needed in each pool before assigning nodes to the network. Nodes need not be consecutive, but consecutive node numbering avoids confusion and simplifies computations.

Frictional Resistance The Chezy equation is used as the standard frictional resistance equation for channels, culverts, and siphons. The Manning-Strickler equation can be used for channel resistance, with input of the Strickler k (inverse of Manning's n). D U F L O W has the added capability of adding wind shear, which can be applied in any direction relative to the channel reach.

Boundary Condition Types In DUFLOW, boundary conditions can be defined at nodes at any location within the network and refer to external conditions that influence flow within the network. Boundary conditions can include a fixed water level; a fixed inflow or outflow to the network; a relation between water level and inflow to the network; rain, which adds flow to the network 726

according to precipitation, catchment area, and runoff proportion of rainfall; and wind velocity and direction. A boundary condition cannot be defined as a structure. Boundary conditions can be specified as a time series, as a Fourier series, or as a fixed value. Other internal boundary conditions are classified as structures. A single equation is used to define orifices, weirs, and entrance and exit losses of culverts and siphons. Gates can be modeled as rectangular orifices. Submergence is determined by the level of the downstream water surface relative to the top of the opening rather than by m o m e n t u m considerations or user definition. Some care should be taken by the user to assure that the different forms of this equation and associated assumptions are fully understood.

Turnouts Turnouts are modeled as boundary conditions and are described in the previous section. This would typically be defined as either a fixed outflow or a relation between water level and outflow. Flows out of the network are given a negative sign. A structure cannot be used as an external boundary condition. However, a turnout could be defined as a structure with a short reach downstream and a boundary condition relating discharge to water level at the end of this reach.

Operations Duplication Structure operations are a standard routine in D U F L O W . For any selected structure, parameters (e.g., gate opening) can be altered in a number of ways. The parameter can be changed to a new constant value or to a time series of values. A number of trigger conditions that cause the parameter to change values can be specified, including: (1) Time when depth at one node is greater than (or less than) a specified level; or (2) when the level at one node is greater (or less) than that at another node by a specified amount. These routines are very useful for duplicating known operator controls while calibrating model parameters.

Automatic Control A number of types of automatic controls can be simulated with D U F L O W through the structure operations modes described earlier. One can model emergency conditions or on/off type controls where the new control setting is known before hand. With automation, the intent is to model automatic canal-gate control mechanisms and software. D U F L O W cannot model this type of operation. There is no interface whereby a user could simulate a gate-control algorithm.

Miscellaneous Limitations D U F L O W has several minor limitations in modeling canal networks. D U F L O W is limited to about 250 channel sections and structures. This is not too serious a limitation since a network of this size would take a very long time to execute even for short time periods. Most networks have locations where they can be conveniently subdivided without too much loss of information. Cross sections can be defined with up to 15 depth-width pairs. The number of boundary conditions multiplied by the number of time steps may not exceed 50,000. The discharge-water level relationships for boundary conditions are limited to 20 pairs of values. Each structure operation can have up to 99 triggers, but the number of operations times the number of structures per operation cannot exceed 16. 727

Data are input and output only in metric units. Dimensional inputs are limited to __0.01 m and __0.01 m3/s, which are a limitation only for comparison to laboratory-size channels. Time steps are limited to 1-min increments, which again are too large for laboratory-size channel comparisons but place no limit on modeling operational-size canals. USER CONSIDERATIONS

User Interface D U F L O W is a menu-driven program. The menus are arranged in hierarchical fashion so that the user can easily navigate though the system. A diagram of the menu structure is provided in the user's manual. All data are entered in simple tables on the screen. These data describe the network, boundary conditions, structures, and computational needs (e.g., simulation time, time steps, data to be saved for output, etc.). For many of the tables, default values are entered for the user (e.g., roughness for current longitudinal section is copied from previous section), or the user can copy previously entered data (e.g., a previously entered cross section can be copied to the current cross section being edited). The user can have the network layout plotted to double check proper network definition. The program automatically creates data files, with a name specified by the user, for the network definition (profile and cross-section data), for the initial conditions, and for the boundary conditions and computational requirements. The data from these files can be sent to a disk file or printed so that a record of the input can be kept. Once the network, initial, boundary, and computational parameters are all entered, D U F L O W can perform the simulation. The user cannot interrupt the computations. Once completed, the user can view the results either in tabular or graphical form. D U F L O W allows the user to define the data displayed, which can be elevation or discharge as a function of either time or location. Tabular data can be output to user-defined files or to a printer. Graphs can be output to a printer, as well. The user can alter the scales of the graph and the data that are displayed. These choices are all made in a convenient menu-driven fashion. The user also has the option of entering field-measured data for comparison to the simulated data. Although the program does not suggest entry of field data from a file, we were able to append our field-data file to the data file created by D U F L O W to avoid manual data entry. Overall, we found D U F L O W ' s menu system very convenient and easy to follow. Without the manual, it would be difficult to fully understand the data-input system and what constitutes boundary conditions and structures, for example. Some experienced users have complained about having to travel up and down the hierarchy of menus to make simple changes to certain fields. However, the alternative would be to have many options at one level, which might make data input less logical and less clear to novice users. Documentation and Support The D U F L O W manual (DUFLOW 1989) provides a thorough explanation of the program and menu system. Examples that give a good picture of the types of applications for which D U F L O W is suited are given. One example goes through a complete explanation of data entry for a sample network, which is very useful for novice users. The manual provides no error documentation to aid the user in handling errors that may occur. 728

Telephone support (for users in The Netherlands) is provided only for startup of the program (getting it working initially). No other support is available. Direct Costs The program cost is nominal, --~$150 US in 1990 including shipping. Users have nonexclusive, nontransferable rights to use the program for their own purposes. The distributors of D U F L O W do not assume any liability for application of the program. The program requires a MS-DOS machine (IBM-XT or better) with 640k of memory and a graphics card. A hard-diskdrive system for program and data storage is not required but is recommended. Alternatively, two floppy-disk drives can be used. A math coprocessor is highly recommended. D U F L O W is not limited by the type or quality of the display monitor; however, higher-resolution graphics monitors will provide better output displays. Indirect Costs The main cost in the use of canal hydraulic models has usually been associated with the time involved--to train the users, to input the canal data into the program, to debug the input, and to calibrate the structures' parameters so that simulation output matches the field data as closely as possible. The D U F L O W program can be learned in just a few days of fulltime effort. The graphics output of D U F L O W makes the calibration procedures fairly quick, although some experience is needed in doing this properly. A typical 24-hr simulation takes from 1 to 5 rain depending on the size of the network. APPLICATIONS

The Task Committee developed a number of example problems and data sets to test various aspects of the models tested [see ASCE (1992)]. Several of these example problems are discussed in the following. Sudden Change in Discharge An example problem was developed to test the model's response to a sudden change in discharge at the head of a canal section. The particulars are: Bottom width 9.14 m (30 ft); side slope 2:1 (horizontal to vertical); Manning roughness 0.020; length 3,219 m (2 mi); and slope 0.0005. For the upstream boundary; initial discharge is 28.32 m3/s (1,000 cfs); new discharge 141.6 m3/s (5,000 cfs); and time for change 10 min. For the downstream boundary, flow depth corresponds to normal depth for the discharge there. The time step is 2 rain. ; the number of distance steps 20. The initial condition is depth = 1.71 m (5.6 ft). The time averaging coefficient is 0.6. The D U F L O W manual recommends starting the simulation with discharge values at internal nodes defaulted to zero. The channel-discharge response with zero initial discharges is shown in Fig. 1. When the steadystate discharge values are entered for the internal nodes, the response is shown as in Fig. 2. After about 10 min, the responses are nearly identical. If an initial steady period of 30 min is provided prior to the sudden increase in discharge, the response (Fig. 3) is essentially identical to inputting the steady-state discharges at internal nodes (Fig. 2), even though steady conditions have not quite been reached after the 30 rain. 729

160

Section 1

,..yio .......--S~ ........................... "r~ -"

120

I '

................ ~'---.~ /,,.15 ...... ...'"7"~

..........................

f-- .

.

.

.

.

...........................

CO v

80

-

i- - ,,-'- -~ .........

~...................................................

~..................................

cO

.m_ I~

40

0

~

0

i

i

30

I

i

i

60

90

Time (min) FIG. 1.

DUFLOW-Computed Discharges for Step Change in Inflow with Interme-

diate Discharges Defaulted to Zero

160 Section 1

5 ..........................i.2}%::::::::!!!!~s

.............

...........io ...................... i]........................ __............................................................... ............................ T7;;}77-i .................................................. i

120

/

/

.....

//

'3'

i

../i...................................................................................................................................................

80

iii

~

4o

0

i

0

i

i

i

i

30

r

60

i

90

Time (min) FIG. 2. DUFLOW-Computed Discharges for Step Change in Inflow with Intermediate Discharges Set at Initial Inflow Rate 730

160 Section

1 /

~ ........... ~ - ~ - f f ~ - - , / ....

12o

v

t~

I

..t l

.......... j"

..............

E cl:l e,(9

.f~"

I ,..."'Io ...... -L ............ ~,

"

i

80

i

i i

40-

i i O

i

0

i

I

i

i

30

i

60

Time

i

90

(min)

FIG. 3. BUFLOW-Computed Discharges for Step Change in Inflow after Initial Steady Period

Volume Conservation A level canal was assumed to have standing water in it. A single cycle of sinusoidal variation in discharge was introduced at one end, with an impenetrable boundary at the other end. The result should be that the final and initial volumes are the same. The details are: The bottom width is 10 m (33 fl); side slope 2:1 (horizontal to vertical); Manning roughness 0.040; length 10 km (6.2 mi); and slope 0.0000. For the upstream boundary, one cycle of a sin wave discharge was input. The duration was 6 hr; and the amplitude 200 cfs. The downstream boundary was blocked, with zero discharge. The time step was 6 min.; and the number of distance steps was 10. For the initial condition, the depth was 7 m (23 ft) everywhere. The time averaging coefficient was 0.6. The volume error after the cycle was 0.01% of the volume in the canal and 0.02% of the positive volume added. Thus D U F L O W does a good job of long-term volume conservation. Cal Poly Test Canal A small-scale physical model of a canal system was operated for several specific conditions of interest to the Task Committee. It was much easier for the Task Committee to develop reasonable test cases to study different phenomena than to interrupt operation of in-service canals. The model canal at California Polytechnic Institute (Cal Poiy) is described by Parrish and Burt (1993). D U F L O W had dimensional limitations that prohibited the direct modeling of the Cal Poly canal. Dimensional inputs were limited to 0.01 m, but the test canal was narrow (0.07 m). The time step was limited to 1 min, but test-canal response was very fast, requiring about a 3-s time 731

step. Flow rates were limited to 0.01 m3/s, but canal flow was only 0.043 m3/s. One digit of precision is not adequate for this type of modeling. While the dimensional limitations of D U F L O W should not adversely affect most operating canals of interest, it could not be conveniently used for the laboratory canal studied here. (Froude modeling could have been used to scale the laboratory canal up, but this was not done.) Observation and Comments on Examples

Each software package for unsteady flow has its own conventions for defining inputs and outputs. This can cause some confusion in trying to Nodes St 2

...........|

| St 1

Sections

i St 3

s4

/

Structures ;5

Boundary

(,, -Q FIG. 4.

-Q

Conditions

Typical DUFLOW Topology

Sec on

. -

......

~"

/' / " /

/

/"

20

t

8o r

///

~

9~tm

..../ / /

40

0

=

0

i

i

30

i

60

i

90

T i m e (min) FIG. 5. DUFLOW-Computed Discharges for Step Change in Inflow with Intermediate Discharges Set at Initial Inflow Rate (Time Step = 5 rain; Courant Number = 9.2) 732

160 Section

1

U') 03 v

ii

i

.,.. . . . . . . . . . . . .

II

-

............

////)'/ . . . . .//

12C

E

Q)

-,

/ / / 80

-

-4---,--