Maria Papathanasiou Flashcards
What are the three qustions in developing the right model?
- Why:
Scope
Level of Complexity
Future Use
- What: Data already available Data that can be obtained Measurement Capabilities Measurement Frequency Variety of Conditions
- How:
Empirical
Mechanistic
Data-driven
What are the levels of complexity?
- Macroscale (e.g. bioreactor)
- Mesoscale (e.g. cell culture)
- Microscale (e.g. individual cell)
What are the 4 steps in the systeamtic approach to developing a model
- Model Development
- Sensitivity Analysis
- Parameter Estimation
- Final Model Formulation
What are the steps involved in model development?
- Pose the question
- Choose the modelling approach
- Develop the equations and make assumptions
- Choose initial conditions
- Choose initial values for the parameters
What are the three methods of sensitivity analysis
- Screening Methods (1 parameter at a time)
- Local Methods (Checks the vicinity of a nominal parameter value)
- Global Methods (All parameters are varied together)
What are the three checks during sensitivity analysis
- Parameters have biological meeting
- Problem is unbiased
- Perform meaningful grouping
3 stages of parameter estimation
- Set search space ranges
- Provide initial guesses
- Form Smart groups (reduce computational demand)
Pros and cons of Modelling
Pros:
- Fast and cheap
- Gives info at “grey” areas
Cons:
- Not intelligent
- Can be biased
- Over-promising
Pros and Cons of experimental work
Pros:
- Real time information
- System understanding
Cons:
- Expensive
- Labour Intensive
- Time-consuming
Reasons to perform a model simulation
- Predict the behaviour of a system under various conditions
- Study the system interactions where little knowledge is available
Define Optimisation
Define the best set of conditions that will allow a system to achieve improved performance
Reason to perform model-based control
Maintain the system at a state where it operates at best capacity
Define sensitivity analysis
Tool that enables the identification of hte impact of the parameter uncertainty on the model
Three problems associated with parameter estimation
- Large models may run into scaling issues
- Initial guesses may be biased
- grouping may be biased
Why do we group parameters when performing sensitivity analysis?
- May be a large number of not very significant parameters so grouping may reduce computational demand
- Some may be related to eachother
- Combination of some parameters may be infeasible so group to remove unnecessary computations
