Darwin Calibrator does a great job with calibration if you 1. are adjusting the right parameter and 2. have good quality data. The key place where the modeler needs to think is identifying what is the reason that the model initially doesn't agree with the field data. If, for example, you adjust demands when the error is in elevation data, then you get what I call "calibration by compensating errors". Next you need to review your field data. If you are measuring pressure to +/- 2 psi (5 ft) and elevations to +/- 10 ft and the head loss between the source and the measuring point is 7 ft, you are basically calibrating the noise in the data. Darwin can't tell if you are calibrating to an accurate or inaccurate field data point. If you try to make every possible parameter an unknown, then you grossly increase the size of the solution space and make it less likely that Darwin can find a good solution. This is why you need to group unknowns. However, if you group parameters that should not be grouped, you make it harder to find a great solution. As I've said on numerours occasions, Darwin Calibrator can help but the most important tool for model calibration lies between your ears. Best of luck, Tom
↧