You may consider fire flows by one of two methods:

1. Use the demand adjustments feature in the required design event to add additional demand to the specific junctions at which fires are to be fought.
2. In WaterGEMS CONNECT, create a child demand alternative of the demand alternative referenced by the representative scenario, and then add the fire flows as fixed pattern flows to the appropriate junctions. Next, in Darwin Designer, set up a design event and select the Override Scenario Demand Alternative check box, and select the new child demand alternative you created.

   Of the two methods, the second one is preferred, since, after you have exported your design from Darwin Designer to a new scenario, you will most likely want to verify the performance of the design directly within WaterGEMS CONNECT. If you have used method one to add fire flows, then you will have to add those fire flows to your current (or new) demand alternative in order to simulate the design against the same demands as in your design event. If you had used method two, however, then you would not need to create any additional demand alternatives, since you had already done that.

Fire-flow design event can be set up by using one of two methods in Question 1. To achieve a good design, you need to ensure that a design can funcion under the most important fire-fighting scenarios. This will be different from system to system. When you set a fire-flow design event, Darwin Designer optimizes the system capacity (pipe sizes) to meet the additional demand requirement for the portion of a system where a fire flow is set up. The other portion of the system may have inadequate capacity. To improve the system-wide emergency response capability, it is recommened that fire flows are set at the outskirts of a distribution grid; this will allow Darwin Designer to optimize the systemwide supply capacity.

Emergency conditions, such as pipe breaks and facility outages, can be handled in Darwin Designer by using the boundary-conditions feature of a design event to close pipes that would normally be open. For example, you may want to consider the effect of a water treatment plant being out of service. This can be achieved by adding any connecting pipes to the design-event boundary conditions and setting their status to closed.

Selection of new facilities may be achieved by using various modeling techniques, an example of which follows.

Selecting the location of a new tank:

1. You can select the location of a new tank modeling the new proposed tank in the representative scenario. Given a specific tank location you will need to enter the tank elevation, diameter, and other size information as if it existed—but, connect the tank to the system with a short small diameter pipe. Give the new pipe an obvious label such as New Tank Connector.

   The pipe that connects the tank to the system should have a length of 1 and a diameter of 0.01.

2. Create a new Design group and label it as New Tank Connector, or something similar, and add the connecting pipe to the new group.
3. In Darwin Designer, create a new pipe option group, label it New Tank, or something similar, and add the following data:

   Diameter	Cost
   0	0
   X	Cost of Tank

   Where, X is some large diameter sufficient for the expected flows to and from the tank.

4. In your local design run group, enable the new pipe group by clicking Active and select the New Tank option group.

   Darwin Designer can now connect the tank to the system and incur the cost specified in the above table, or it will construct a 0 diameter pipe (no pipe) and the tank will not be included in the system. Note that it is up to you to make sure that sufficient demand cases are investigated to verify the tank's design and that tank operation is independently verified through an EPS simulation.

   Using similar logic Designer could be used to consider the inclusion exclusion of pump stations, valves, water treatment facilities, reservoirs and so on.

There are a number of things that could be causing you get strange or unexpected results with Darwin Designer. Before calling technical support, please take the time to review this list to see if any of these things may apply to you.

1. Make sure you are using the correct design data. Make sure you are using the correct representative design scenario and that scenario includes all pipes to be sized by Darwin Designer.
2. Make sure that the representative design scenario runs successfully within WaterGEMS CONNECT . If it does not, then Designer will not be able to function correctly.
3. Make sure that the correct demands are present. For EPS representative scenarios, make sure your patterns are correct and that you are using the correct time from start value in your design events.
4. Make sure that you have applied the correct and necessary boundary conditions. For example, if you are designing for a 7 a.m. peak-flow condition, make sure that you have boundary conditions specified for all necessary tank levels, pump operation, etc. For designs that include a significant amount of new infrastructure or completely new designs, tank levels have to be assumed tank levels.
5. Make sure that the range of pipe sizes and rehab actions you are using are reasonable. For example, make sure that you are allowing Darwin Designer a sufficient range of pipe diameters to come up with a reasonable design. While Darwin Designer does perform an initial feasibility check (it uses the largest pipe sizes and checks minimum pressures), too few pipe choices may artificially restrict the flexibility of the optimization. Conversely, too many choices may affect the convergence of the optimization on to a good solution. It doesn't make sense, for example, to allow a rising main from a pump station to be 6 in. or 8 in.
6. Make sure that you have a reasonable number of design and/or rehab groups. As an extreme example, consider that every pipe to be design was in the same group. Then the only possible solution that the optimization can arrive at is to construct all of the pipes the same size. While it may still be possible to find a feasible solution, only having a single design group will restrict the flexibility of the optimization and the ability of Darwin Designer to find cheaper solutions. Conversely, too many design groups will hinder the convergence of the optimization and result in sub-optimal solutions. A good number of design groups will depend on the actual model and design situation, but would lie somewhere between 10 and 100.
7. Make sure you have sufficient and reasonable design constraints in place. The genetic algorithm optimization engine in Darwin Designer is very powerful. If the objective of the optimization is to minimize cost, the optimization engine will do everything in its power to minimize cost including unwanted things that may not have been disallowed by the designer. The worst case scenario is a design with no constraints. If the design does not have any performance requirements, then the cheapest design is no design at all.

   The optimization algorithm only knows the problem that is defined for it, and to that end if you wish to get meaningful designs from Darwin Designer, you need to constrain your designs appropriately. The idea is to set up design constraints that corner the optimization algorithm into a region of the solution space (region of all possible solutions) that makes the most practical sense.

   Design constraints can be applied in Darwin Designer by pressures (max. and min.) and also pipe velocities (max. and min.). An example of an impractical situation in a hydraulic model might be a 1 MG tank that is draining at far too high a rate. In order to save costs on constructing pipes to a more distant source, the optimization algorithm may over-use a closer water source.

   Another example of a design constraint-other than the pressure and flow constraints-is the number of design events (and hence demand/operational cases) that the design must meet. The optimal solution to a single demand case does not fully reflect the real system operating scenarios. If a single load condition is used along with a zero-diameter as one of possible sizes in a option group, it will most likely result in a branched network design. Thus, it is necessary for reliability reasons to design systems for multiple demand conditions.

   It is up to the engineer to recognize any impracticality of an optimized design and set up the necessary design constraints to prevent that type of design from being feasible, thus removing that design possibility from the grasp of the optimization algorithm.

To do this you need to do three things:

1. • Group together the pipes that will attract the special cost. These pipes can be each in their own groups or all in one group, but they should be grouped such that they are separate from pipes that won't attract the special cost.
2. Create a option group (new pipe or rehabilitation option group) that includes the special cost premiums.
3. Assign the special option groups to the associated design groups locally, for the design run you wish to use with the special costs.

Darwin Designer applies a competent genetic algorithm to optimize the design. It does not require or have any domain-specific knowledge about the water system, which ensures it is a generic tool, but also causes some side-effect for some design cases-like giving up-or-down pipe sizes. In particular, the solutions are evaluated by comparing the fitness values of solutions. Darwin Designer will assume a pipeline with pipe sizes that go up and down (to meet required pressures as closely as possible) is better than one that has a constant size that exceeds the pressures at some locations, since there is no specific penalty assigned to the fitness of a solution that has pipes that change up and down in size. It is, therefore, up to you to control the eventual design and this can be done by different means, as follows:

1. The first means is simply to make manual adjustments to a design after Darwin Designer has finished, in order to clean up the design and make it a practical design. Cleaning up a design may technically move you away from the cheapest design, but an inexpensive design that won't be constructed is of little use. You may find that not much cleaning up is necessary. Quick edits to diameters or rehab actions like can be performed effectively in Darwin Designer by using a manual design run.
2. Another thing to consider when analyzing a Darwin Designer design is whether the chosen pipe sizes are a function of the lengths of pipe in your model.

   To better illustrate this concept, consider a run of four pipes in series, each with different lengths. For these four pipes, the controlling pressure is the downstream-most junction, and all intermediate junctions are well above the required pressure. Now, after Darwin Designer finishes designing the run of pipe, it selects the first pipe as a 16 in., the second as 12 in., the third as 16 in. and the fourth as 12 in. It is unlikely that this design would be constructed as-is, but if the pipes themselves represented sufficient length of pipe, then it may be practical to construct a portion of the pipeline as 16 in. and a portion as 12 in. If this is the case, then you need to look at the model to determine why Darwin Designer is changing the third pipe back up to 16 in. It may be that since the downstream-most junction is the only controlling node, that Darwin Designer is merely trying to achieve the right head-loss in the total pipe length, by choosing the length of pipe that should be 16 in. and the length that should be 12 in. Of course, it is still constrained by the individual pipe lengths in the model, but if they are different, the optimization algorithm will use this fact to its advantage. In this case, it may very well be that Designer is saying construct a total of 1500 ft. of 16-in. and 1000 ft. of 12-in. pipe, and not necessarily 850 feet of 16-in., 600 feet of 12-in., 650 feet of 16-in., and 400 feet of 12-in. pipe in sections. Use engineering judgment when analyzing the results.

3. Another means of achieving more constructible designs from Darwin Designer is to group in the same group pipes that would be constructed the same size. For example, a rising main would most likely be constructed a single size, and it would thus make sense to include all the model pipes that make up the rising main in the same design group. What you don't want to do by grouping pipes is artificially design the system even before you have had a chance to optimize it.

It is possible to do the same thing for rehabilitation actions. To create a rehabilitation action that represents a Do Nothing option, simply follow these steps:

1. Create a pre-rehab diameter versus post-rehab diameter function that defines at least two diameters that cover the domain of diameters in your model. For example, mi.n pipe size through max. pipe size and make the pre-rehab diameter the same as the post-rehab diameter. This function will define that the diameter of any single pipe remains the same before and after the rehab action.
2. Create a diameter versus unit cost function that defines at least two diameters that cover the domain of diameters in your model. E.g., min. pipe size through max. pipe size and make the cost for each diameter zero. This function will thus define that the cost for the rehab action, regardless of pipe size is zero.
3. Create a pre-rehab diameter versus post-rehab roughness function that defines at least two diameters that cover the domain of diameters in your model. E.g., min. pipe size through max. pipe size and make the post-rehab roughness, the roughness of the current pipes to which the Do Nothing option will be an option. This function will thus define that the resulting roughness stays the same as the original values.

   Create a Do Nothing rehab action that references each of the above functions. If selected by Designer, the Do Nothing action will leave the same diameter, cost nothing, and leave the same roughness: in effect, doing nothing.

In most circumstances it is not necessary to change the parameters in order to run Darwin Designer, however, you may wish to modify certain values as follows:

1. Random Seed-The Darwin Designer optimization algorithm depends on the generation of pseudo-random numbers through a random number generator. The reason the numbers are pseudo-random is that they are generated by a mathematical formula, and hence the resulting series of numbers is not actually random at all. In order to make the random numbers different the random number algorithm is initialized with what is known as a seed. For a different seed value, a different series of pseudo-random numbers will be produced. When no parameters in the Designer optimization problem change (i.e., no changes at all, including hydraulic model changes, constraint changes, etc.), running Darwin Designer twice will result in exactly the same result. Darwin Designer results are therefore repeatable in this way. One way of ensuring a different result (or at least a different progression to the same result) is by changing the random number seed. Doing this will result in different optimization results for different runs. By the nature of genetic algorithm optimization, you should not just accept the result of a single optimization run, but run several runs and make sure that all runs produce similar results. An easy way to run multiple runs and achieve different results is to change the random number seed.
2. Penalty Factor-Penalty factor is a weighting that is used in the determination of the fitness value for an hydraulic solution. In particular the penalty factor is used to discourage the survival of designs that fail the design constraints. A higher value for penalty factor will put designs that fail the design constraints in greater disfavor, where as a lower value for penalty factor will place designs that fail the design constraints in less disfavor. A reasonable default for penalty factor has already been selected for you. However, if you find that Darwin Designer keeps settling on designs that contain constraint violation, then you may wish to increase the penalty factor value.
3. Probabilities, Era Numbers, and Population Size-Good defaults have already been selected for you for these values, but instead of changing the random number seed when conducting multiple optimization runs of the same design, you may want to change these values. Good ranges for the values are therefore listed below for your convenience.
   Note: The upper limit values for population size, maximum era number, and era generation number are problem-dependent. For larger design models, you should use greater values than for smaller models.

   Population Size: 40 to 200

   Cut Probability: 0.5 to 2.5%

   Splice Probability: 50 to 80%

   Mutation Probability: 0.5 to 2%

   Maximum Era Number: 4 to 10

   Era Generation Number: 50 to 200

You cannot select pipes directly from the drawing in this first release of Darwin Designer. For this reason, we recommend you identify pipe groups and create appropriately-named selection sets before starting Darwin Designer. When you have defined the necessary selection sets, they can be used directly within Darwin Designer. Selection sets can also be used to define pressure and flow constraints, and to select boundary condition elements.

It is very likely that in using Darwin Designer, you will encounter situations where Darwin Designer cannot find a feasible solution. This happens even to those experienced in genetic-algorithm optimization and is due to the fact that the determination of which designs are feasible and which aren't is assessed by a computer subject to the information you tell it. That is, the rules are applied, with no exceptions.

For example, if you want a minimum of 20 psi across the board, Darwin Designer will determine as infeasible any solution that does not have 20 psi at every junction. If you have a couple of junctions that are part of the detail of a tank inlet valving, for example, then maybe you don't really require 20 psi at those junctions. Perhaps what you really mean is that you want 20 psi at all service junctions. In that case, you'll find where an engineer would have said the design is feasible (because the design only fails the 20 psi requirement at non-service junctions), but Darwin Designer is unable to make that determination, since it was told 20 psi was required at all junctions. The process by which you can get around these kinds of issues is simply to identify them, correct them, and then re-run the optimization. For the case of the 20 psi junction example, the fix might be to create a selection set (in WaterGEMS CONNECT ) of the junctions that are service junctions, and only use those junctions as pressure constraint junctions. (The selection set can be selected from within Darwin Designer.)

Along these same lines, you may also want to consider if any of the following things might be causing trouble, before calling technical support:

1. Check for constraint violations in the results. Check both pressure and flow constraints for the presence of constraint violations. If any violations exist, you will need to determine why the junctions and/or pipes at which the violations occur are problematic. Maybe a minimum pressure constraint is simply impossible to meet due to the junction elevation, etc. Other things to check for are the applicability of blanket minimum and maximum pressures and velocities to modeling elements in detail models of pump stations, and the like. If you find anything, then you need either to change the model, or modify/remove the offending constraint and run the optimization again.
2. Make sure you have sufficient design options for a feasible design. That is, make sure that you have a sufficient range of pipe sizes and/or rehabilitation actions available to Darwin Designer to find a valid design.
3. Make sure that you haven't specified competing design events. While it may be possible to meet one design event or another separately, it may be impossible to meet two together if they compete with each other. For example, one design event might specify that a minimum pressure is required, and as such the corresponding pipe taking the flow to that location needs to be large, however, in the next design event with similar demands, a minimum velocity constraint means the pipe has to be sized smaller. It may be impossible to meet both design events with the single pipe size. To test this, build runs up by performing initially with only one design event, then adding more in. If all of a sudden after adding in a design event no more feasible solutions can be found, then you can try to work out what in the most recently added design event is causing the problem.
4. For multi-objective and maximum benefit optimizations, make sure you have sufficient budget specified. It may just be that you have not given Darwin Designer sufficient budget to allow a feasible design to be found. Try increasing the budget.