Design of Models Flashcards
Describe the steps in the modelling process
What are the steps involved in naive modelling?
Why, how, assume, formulate, use, recommend
What are the steps involved in robust modelling?
Why, how, given, assume, formulate, working, suitable, improve, use, unknown, recommend
Describe the WHY section of the process
What are the broad dimensions of the problem? What is our goal? What are the stakeholders’ opinions/preferences?
Describe the HOW section of the process
Define the scope of the problem. Find the small problem inside the big problem, that provides good insight into the bigger problem. Define what is to be left out and what is to be learned.
Describe the GIVEN section of the process
What information is available to us and what do we already know? Literature review for modelling approaches and potential parameter values. Compile data for driving and checking the model.
Describe the ASSUME section of the process
What reasonable assumptions can be made to simplify the problem? Create a conceptual model with the physics, domains, parameters, and BCS and ICS
Describe the FORMULATE section of the process
How should the model be constructed, numerically and mathematically? Define governing equations, and analytical or numerical solutions.
Describe the WORKING section of the process
How will we know if the model is working correctly? Is the model right? Verification (are there errors, hunting for bugs, benchmarking, unit tests).
Describe the SUITABLE section of the process
How will we know if the model fulfils its purpose? Is it the right model? Validation (is the model suitable to physically explain, experimental comparisons, calibration, model misfit).
Describe the IMPROVE section of the process
How can the model be made better? Critically iteratively evaluate the assumptions, the conceptual model, the software, and the calibration data. Remember diminishing returns.
What parts of the process are considered quality control?
Everything before use (why, how, given, assume, formulate, working, suitable)
Describe the USE section of the process
How will we obtain new insights from this mode? What-if scenarios, optimal outcomes, forecasting, inverse modelling (parameter estimation).
Describe the UNKNOWN section of the process
What are the limitations of the insights provided by the model? Parameter uncertainty, structural/model error, observation/data error, model ensembles
Describe the RECOMMEND section of the process
Based on insights from the model, and considering non-model factors, what course of action should be taken? Non-model factors: financial, legal, environmental, societal, cultural.
What is the difference between the WHY and HOW questions?
WHY defines the goal and the broad dimensions of the problem (e.g problem statement, outcomes/solutions, stakeholders). How focuses on how the model can be used to help address the problem (e.g the specific scope of the model and the insights to be learned are defined).
When preparing to undertake a modelling study, which questions are we focused on?
Why, how, given
When developing a model, which questions are we focused on?
Assume, formulate, working, suitable, improve
When getting insight from the model, which questions are we focused on?
Use, unknown, recommend
Which questions could fall under the heading of quality assurance?
Everything before use (why, how, given, assume, formulate, working, suitable, improve)
Do models make real-world decisions?
No, we make real world decisions with the help of models. We have to use model insights alongside non-model factors such as legal, environmental, societal, cultural, and financial to make real-world decisions.
Describe the difference between “is the model right?” and “is it the right model?”
Is the model right refers to the process of verification. It involves debugging the code (or software), and ensuring that it is running as expected (e.g use of unit testing or benchmarking to ensure the correct math is implemented). Is it the right model refers to validation. It involves comparing and calibrating the model to experimental data to ensure that the model is the correct one to explain the physical process being modelled.
The modelling process is a guide, not a recipe. What is meant by this?
Following the exact steps provided will not guarantee a perfect model. Engineering judgement is required through the process to ensure a suitable model is created.
How can concepts of engineering design - requirements, identify options, define configuration, deploy solution - be applied to the design of computer models?
At each step of the modelling process, the engineering design requirements should be used. This will result in a better designed computer model.