Automating hydraulic transient model to enable efficient sensitivity analysis and calibration

Hydraulic transient pressure is a critical parameter in designing surge mitigation devices, pipe material and layout of forcemains in wastewater/stormwater forcemain systems and feedermains in water distribution systems. During conditions such as pump trip, control valve closure, or power failure, pressure fluctuation in the pipeline and pumping station can cause significant vibration to pipeline and noise to surroundings, and transient pressure extremes (high and low) can even burst or collapse pipes. Hydraulic transient behavior in a pipeline system is governed by fluid continuity and momentum equations. These equations can be solved in one-dimension along the pipeline with finite difference, finite element, method of characterization (MOC), or wave method (WM). Finite difference and finite element were proven not to be computationally effective, while MOC and WM are relatively effective and implemented in commercial software packages (eg Bentley Hammer, InfoSurge, KYPipe, MIKE URBAN FGDHT, and TransAM).

In model sensitivity and calibration studies, it is often necessary to run a numerical model hundreds of times. Sensitivity studies could be completed manually by creating hundreds of scenarios that enumerate parameters possibilities, which is extremely time consuming and often impractical.   Manual calibration could be completed by trial-and-error process, which randomly picks up the next set of parameters and is thus very difficult to capture the optimal set for a model. Model automation makes such studies effective, efficient and practical. A model required for automation should support: a) updating model input parameters, b) running transient numerical engine, and c) retrieving the results (both profile and time history). Bentley exposes its software functionality (both user interface and numerical engine) through Application Programming Interface (API) that resides in WaterObjects.NET, and provides software customization opportunity to end users. Specifically, Bentley Hammer APIs support the three requirements, thus used for sensitivity and calibration study.

This paper proposes an automation procedure developed for Bentley Hammer to perform a sensitivity study for three model parameters: pump inertia, wave speed, and C factor, and a model calibration study with genetic algorithm. The proposed automation procedure can also be applied for any other model parameters (eg air release valve size and location). The proposed automation procedure is demonstrated through a case study in the City of Hamilton, and shows its ability in saving significant time in sensitivity analysis and enables calibration study, which subsequently leads to better understanding and transient model representation of the system. In this study, pressures recorded with high frequency sensor at the outlet of a pumping station were utilized in the model calibration.