Carlos Rivera Villarreyes, DHI
FEFLOW brings a wide spectrum of tools to overcome the most simple challenges as well as complex flow and mass transport problems. This workshop will review the multiple modeling options for flow and mass transport using simple yet educational examples, and will focus mainly on higher-level “why” and “when” questions rather than narrow technical topics. This session is practical for both new and experienced users since it will give a quick overview of the software options as well as address punctual conceptual questions. Topics that will be covered:
Dr. John Doherty, Watermark Numerical Computing Pty. Ltd.
This session will focus on the use of models in decision-support, and the role that model-value-adding software can play in optimizing their potential in this role. A brief overview of the importance of risk (and therefore uncertainty) assessment in decision-making will be given. Highly parameterized inversion and the use as pilot points as a parameterization device will be discussed. The session will also cover some exciting new developments such as (linear and nonlinear) optimization under uncertainty, and Kalman smoother technology offered by PESTPP-IES that supports calibration-constrained uncertainty analysis at a very low numerical cost. The new, open source PANTHER parallel run manager used by programs of the PEST++ suite will also be addressed.
Dr. Fabien Cornaton, DHI
The FEFLOW package utility FePEST is commonly used for calibration, prediction and uncertainty analysis operations. These operations require multiple model runs which may be run in parallel, distributed on multiple computers. Modern cloud computing provides an opportunity to temporarily add virtual machines to the pool of computing resources for the duration of a FePEST operation in order to reduce the total run time.
The use of the container platform software Docker (https://www.docker.com/what-docker) allows for a Cloud solution deployment to not be reliant on any one specific Cloud platform. With this technology, a virtualized machine with a preconfigured FePEST can be easily deployed on a variety of environments, such as Amazon Web Services, Microsoft Azure, Google Compute Engine, and many others.
This highly facilitates the access for customers to speedup of multiple solutions engaged in model calibration, sensitivity analysis (e.g. through Monte-Carlo analysis) and uncertainty quantification assessments.
The proposed workshop reviews the necessary requirements to have FEFLOW / FePEST functionalities easily accessible in the Cloud through comfortable user-experience:
Carlos Rivera Villarreyes, DHI
During modeling work we are continuously repeating tasks related to the model set-up and/or result evaluation. This workshop provides a smooth introduction to the FEFLOW Python Interface with the aim to minimize operational costs and automatize several workflows. This modular, instructor led, hands-on workshop provides fundamentals and approaches for FEFLOW customization using the Interface Manager IFM (C/C++) and the Python interface. The module is mainly oriented in Python operations, but it also brings a basic background in C++ Interface, so you will come away with a complete understanding of the FEFLOW Application Programming Interface (API).
You will get an overview of the necessary tools for optimizing your daily workflow as a groundwater modeler. With simple, but practical applications, Carlos demonstrate pre- and post-processing operations such as working with selections, adjusting parameter values, running model scenarios and extracting model results in a customized manner. The development of automatic workflows may boost your efficiency in pre- and post-processing numerical models. Moreover, the workshop will give multiple several hints about function documentation, user-tricks and best approaches.
Among several topics, the workshop will address the following:
• Why and when to use the Interface Manager (C/C++) and when to use Python
• Setting up Visual Studio and Python for getting started
• Using Callbacks, Kernel Control Methods and API
• Working with selections and manipulating material properties
• Export of model data (e.g., process variables, velocities)
• Running model scenarios and automated budget analysis
• Usage of nodal or elemental distributions for in– and output of customer data
• Working with external Python libraries (Pandas, Matplotlib, Bokeh, etc.)
• Several examples
• Open discussion of capabilities and participant optimization “wishes”
Peter Schätzl, DHI
Feel free to bring example models/data illustrating a certain question (please no real ones). Questions can also be submitted in advance to help Peter prepare.