For these purposes, a model is constructed in which data describing network elements of pipes, junctions, valves, pumps, tanks, and reservoirs are assembled in a systematic manner to predict pipe flow and junction hydraulic grade lines (HGL) or pressures within a water distribution system.

Computer models are significant investments for water companies. To ensure a good investment return and correct use of the models, the model must be capable of correctly simulating flow conditions encountered at the site. This is achieved by calibrating the models. A calibration involves the process of adjusting model characteristics and parameters so that the model's predicted flows and pressures match actual observed field data to some desirable or acceptable level. This is described in more detail in Walski, Chase and Savic (2001).

Calibration of a water distribution model is a complicated task. There are many uncertain parameters that need to be adjusted to reduce the discrepancy between the model predictions and field observations of junction HGL and pipe discharges. Pipe roughness coefficients are often considered for calibration. However, there are many other parameters that are uncertain and affect junction HGL and pipe flow rate. To minimize errors in model parameters and eliminate the compensation error of calibration parameters (Walski 2001), you should consider calibrating all the model parameters, such as junction demand, operation status of pipes and valves, and pipe roughness coefficients.

Calibrating water distribution network models relies upon field measurement data, such as junction pressures, pipe flows, water levels in storage facilities, valve settings, pump operating status (on/off), and pump speeds. Among all the possible field observation data, junction HGL and pipe flows are most often used to evaluate the goodness-of-fit of the model calibration. Other parameters, such as tank levels, valve settings, and pump operating status/speed are used as boundary conditions that are recorded when collecting a set of calibration observations of junction pressures and pipe flow rates.

Field observation data are measured and collected at different times of the day and at various locations on site, which may correspond to various demand loadings and boundary conditions. In order for the model simulation results to more closely represent observed data, simulation results must use the same demand loading and boundary conditions as observed data. Thus, the calibration process must be conducted under multiple demand loading and operating boundary conditions.

Traditional calibration of a water distribution model is based on a trial-and-error procedure by which an engineer or modeler first estimates the values of model parameters, runs the model to obtain a predicted pressure and flow, and finally compares the simulated values to the observed data. If the predicted data does not compare closely with the observed data, the engineer returns to the model, makes some adjustments to the model parameters, and calculates it again to produce a new set of simulation results. This may have to be repeated many times to make sure that the model produces a calibrated prediction of the water distribution network in the real world. The traditional calibration technique is, among other things, quite time consuming.

In addition, a typical network representation of a water network may include hundreds or thousands of links and nodes. Ideally, during the water distribution model calibration process, the roughness coefficient is adjusted for each link and demand is adjusted for each node. However, only a small percentage of representative sample measurements can be made available for the use of model calibration due to the limited financial and labor requirements for data collection. Therefore, it is of utmost importance to have a comprehensive methodology and efficient tool that can assist the engineer in achieving a highly accurate model under practical conditions, including various model parameters such as pipe roughness, junction demand, and link status, and also multiple demand and boundary conditions.