Combining different hydraulic methods to estimate the discharge from Combined Sewer Overflows (CSO) into streams.

Combined Sewer Overflows (CSOs) in cities can play a significant role in the morphology and hydraulic performance of streams near urban areas. A complete urban drainage system is often modelled by a dedicated hydrological/hydraulic model (e.g., SWMM or Mike+). However, these models must be calibrated against observations. Especially the flow from CSOs is difficult to estimate. The quality of the data and the drainage models depend on the accuracy of the overall mass balance of the drainage system. If it is not possible to estimate the discharge from, for example, a CSO, the results from other parts of the system become unreliable.

This research evaluates flow dynamics through a multi-methodological approach where the CSO is evaluated by theoretical models, CFD models, experimental work in a laboratory, and a new innovative method where the noise from a CSO is analyzed. The sound is both analyzed directly and by training a machine learning model on the laboratory experiments. The result is a hybrid model filtering all the estimates to one flow estimate. CFD has been used to model the specific CSO to take a Q-h relationship into account, and to generate a so-called catalog method. In this method, multiple variations of geometry are simulated in a free-surface CFD model to cover many different geometries, and general equations are extracted from these simulations.

The hybrid approach opens the door to a new way of estimating interactions between the urban water cycle and the receiving waters. Applying edge processing makes it possible to continuously adapt to local conditions that were not present during the calibration and validation of the model. Edge processing involves signal processing and modeling at the measuring point, where the maximum bandwidth of the sensor data is available and can be used for the most accurate data estimation.