Jacobs is participating in the 94th Annual Water Environment Federation Technical Exhibition and Conference (WEFTEC 2021) being held October 16 – 20 at McCormick Place in Chicago, Illinois.
WEFTEC is the world’s largest wastewater conference, drawing more than 20,000 attendees from around the globe. This is the first major in-person water event since the COVID-19 pandemic hit and we’re a proud Partner-level sponsor.
Numerous Jacobs innovators will be presenting in workshops, technical sessions and panels throughout the event, such as Jacobs’ Global Technology Leader for Wet Weather and Collection Systems Management Bill McMillin who will talk about Jacobs’ Water Model Autocalibration tool on Wednesday Oct 20. Bill is presenting on behalf of John Myers, Jacobs water conveyance modeler, who authored the paper and created the presentation for WEFTEC.
Read more about the Water Model Autocalibration tool from John in the article below.
Modeling drives design and decision making within collection systems engineering. Within models, there are inevitably parameters that cannot be directly determined and must be inferred from observed flow and depth data during model calibration. Model calibration is a ubiquitous challenge that is typically approached as a tedious, time-consuming trial-and-error process. Model calibration is a critical path task that often demands significant labor hours and controls project schedules.
Over the past three years, Jacobs developed the Water Model Autocalibration tool to calibrate collection system models efficiently and automatically. This autocalibration tool expedites multi-objective optimization using a procedure called response surface characterization.
The tool quantifies the mathematical relationship between model inputs and outputs – the model’s response surface – using a small number of strategically selected model simulations. It then executes a multi-objective optimization using the computationally inexpensive response surface instead of the model itself to minimize the difference between modeled and observed outputs. The foremost advantage to this approach is the computational cost reduction: the number of model simulations required to arrive at an optimal solution is significantly lower than would be required using traditional optimization algorithms. A single run of the autocalibration tool on a collection system model allows the user to find the set of input parameters that produces the optimal outputs, such as modeled flow and depth, across multiple meters and calibration events.
What emerged as one of the greatest benefits of using the autocalibration tool was its ability to continuously calibrate the model overnight and during weekends with no user interaction. When this automation is paired with the elimination of redundant model runs and the ability to calibrate multiple meters of the same order simultaneously, the benefit of this autocalibration tool is substantial. The clear caveat is that, just as in manual calibration, the quality of the result is only ever as good as the quality of the inputs. Before calibration is initiated, modeler time must be invested in ensuring that the non-calibration parameters and structure of the model are representative of the real collection system, and that the tool is calibrating to good quality data.
Additionally, significant effort has been invested into functionalities that allow for engineering judgement, historical knowledge, and client requirements to be central to the autocalibration tool’s operation. This is meant to reduce the “black-box” nature that is present in certain machine learning implementations, making it easier to apply this tool and accept its results. When paired with initial quality assessment of calibration data, the autocalibration tool is able to save modelers a significant amount of time and money on complex calibrations.
John Myers is an environmental Engineer with a demonstrated history of working in higher education and private industry. As a water conveyance modeler and data scientist at Jacobs, John is currently working in integration of machine learning and artificial intelligence with water modeling. A strong research professional, John holds a Master of Science (MS) in Environmental Engineering and BS in Civil Engineering from University of Cincinnati.