Darwin Calibrator Troubleshooting Tips

If you've found your way to this section, then you are probably looking for an answer to a problem that you cannot find elsewhere. Please refer to the list below if you are having problems running Darwin Calibrator (you keep getting unsatisfactory solutions) or if you receive this message while running a calibration: The calibration engine was unsuccessful. See the help system for troubleshooting tips.

If you are receiving the engine unsuccessful message, try the following:

Take note of the error message that is provided along with the calibration engine was unsuccessful message. It may provide a clue as to why your calibration didn't run and save you from having to go any further through this list!
Ensure that the scenario model upon which the calibration is based will run properly in WaterCAD 2024 . Select Analysis > Compute, select the steady state button, and click GO. If the run obtains either a yellow or green light, then the hydraulic model runs and this is not the problem.
Ensure that all your roughness and demand group settings are valid and reasonable. For example, ensure that roughness adjustments and/or demand adjustments are not such that your hydraulic model might have difficulty converging. For example, make sure that you are not allowing demands to be set too high or pipes too rough, causing excessive amounts of head loss.
If you have a large number of pipes assigned to status groups, review the need to include all of those pipes as status decisions and try to minimize the number of pipes in status groups.

Note: Virtual memory settings should only be adjusted by advanced users or system administrators.
You may be experiencing low system memory. When running Darwin Calibrator, be sure to close any other unused applications and if adjusting advanced GA parameters ensure that you are using a cut probability of more than a few percent, and a splice probability of less than 90 percent. If your system doesn't have much RAM (<128Mb), you may also wish to increase the amount of allocated virtual memory that your system is using. Please see your Microsoft Windows documentation for information on virtual memory settings specific to your operating system.

If you are having problems getting reasonable calibration solutions, try the following:

Ensure that the Time field for each of your field data measurement sets corresponds to the time of day that your measurements were taken. The reason being that the time entered in your field data set is used to determine demand multipliers (from hydraulic patterns), which are in turn used to calculate the junction demands that will be simulated within the GA calibration engine. (The demand at a junction during a GA calibration run is the product of its baseline demands and the demand factors at the time specified for the field data set.) Pump settings and control settings, etc., are also determined from the time setting you specify. Demand multiplier adjustments and additional junction demands (e.g., fire flow tests) are in addition to, not in lieu of, junction demands already calculated from pattern multipliers. Also note that a steady state run in WaterCAD will run with only junction baseline demands applied, whereas a GA calibration run based on a steady state scenario will still use pattern multipliers for the specified time.
Modifying the status of a link can have significant effects on hydraulic results and your chances of finding good calibration solutions. If you are using a number of status group adjustments, you should review why you need those adjustment groups. It may be better to experiment with these kinds of adjustments manually, or get somebody to find out whether that valve really is closed and remove the status decision from the GA calibration. In general, try to keep status adjustment decisions to a minimum.
Make sure that your adjustment groupings are logical. For example, junctions are grouped by similar pattern or demands for demand groups and pipes are grouped by similar size, age and location for roughness groups.
Ensure that you do not have too many adjustment groups or the allowable ranges and increments for those groups do not allow too many choices for each group. For example, a roughness group allowed to vary between a Hazen-Williams C of 80 and a Hazen-Williams C of 130, with an increment of 0.1 equates to 500 different possible roughness settings for one group. This is far too high! Try to choose lower and upper bounds, and an increment that will give you no more than 10-12 possible values. If need be, you can start off with course settings (say 80 to 130 with an increment of 5) initially, and gradually refine the allowable range and increment to refine your calibration solutions. This applies to both roughness adjustment groups and also to demand adjustment groups.
Make sure that you have sufficient and quality field data and that it has been entered correctly. In general, it is a good idea to have as many (or more) field data measurements as adjustment groups for the calibration, or else your calibration problem is under-specified. This means that there is likely to be multiple calibration solutions that produce the same or very similar hydraulic results (e.g., solutions that exhibit compensating errors). In theory, there is only one correct solution, however, due to limits observed for many practical model calibrations, the more quality field data you can provide, the better chance you have of finding a solution that is close to the real situation. When assessing the number of field observations that you have, consider that each individual observation should contribute unique and accurate information to the calibration. For example, pressure measurements made at two junctions in different parts of the distribution system are likely to be more valuable than two measurements made at locations close to each other in the distribution system. In fact, the two measurements taken at points close together may only be as good as one measurement. That is, both measurements say the same thing about the system. Simply, the field data you collect and enter into Darwin Calibrator should be data that represents times when your system is experiencing high demand, even if it is only the result of such activities as fire flow tests. The reason for this is that during times of normal demands, the head loss across the system is usually on the same order of magnitude as the error in measuring head loss. Therefore, small errors in measurement can lead to huge errors in roughness coefficient or demand.
Make sure that you haven't entered field data observations that are made impossible to achieve by any observed boundary conditions, such as an observed grade out for a PRV set to a different grade.

Note: Tank levels, pump speed settings, valve settings, and reservoir HGL are all used by the calibration engine as boundary conditions and as such these field data entries will not appear in the calibration report summary. That is, these quantities are set as fixed in the calibration simulations and the calibration does not try to match these data. All other quantities are used as observed quantities that the calibration engine tries to match by adjusting parameters defined in your adjustment groups.
Make sure you are using the correct boundary conditions. If you have entered observations for tank levels etc., ensure that you have not made any errors in entering the data.