I am looking for information and practical examples on how to perform and use sensitivity analysis in engineering sciences -- namely in building physics, structural engineering, mechanical engineering, architecture engineering, as well as in urban planning and design.
When (in what cases) do you do sensitivity analysis?
How do you do it? What methods and software programs do you use for statistical data processing, analysis and data visualization?
Thanks
Sensitivity analysis boils down to calculating the relative change of a quantity of interest with respect to a parameter. You can think of sensitivity analysis as computing the partial derivative of a dependent variable with respect to one of many independent variables.
In practice, computing a sensitivity assumes you have a computational model for your problem. Models can be of two types, deterministic and stochastic.
If your model is deterministic (the FDM used in structural analysis is usually implemented in deterministic models) the computation of sensitivity is straightforward, but you must pay careful attention to three issues, a) numerical roundoff (insuring that the impact of a change in the underlying parameter of your sensitivity is not overpowered by numerical inaccuracy), b) insuring the parameter your sensitivity is referred to is properly isolated, and c) non-linearity (insuring your computed sensitivity stays within linear range, if linear sensitivity is what you are after, which is normally the case.)
If your model is stochastic (such as simulations using quasi-random numbers), a naive approach to sensitivity will be problematic and specialized techniques, known a variance reduction techniques, are required to keep simulation error from overwhelming your computations.
hello, in my opinion sensitivity analisis are needed when you carry on parametric studies. at the end of the studies the results are analised and a good thing is determine what is the isolate influence of every parameters on the magnitudes of interest. some time is easy but some time i sa complicate thing because of the influence of each other. for statistical analisis "Star graph (centurion)" is a good option. In graphics results the best software I think is Origin".
I agree with Domingo, a univariate analysis where you vary one parameter at a time and observe output responses
I am a beginner to this topic. I have seen few Master's and PhD thesis with Sensitive Analysis section written under their results chapters...
But, is there any reading Materials for me to read further.. to have more understand of the topic at hand...
To bring it down to the simplest form, look at the below example
x + y + z = a. If you want to find how a variation of 10% in 'x' affects the final outcome of 'a', sensitivity analysis can be used.
This is the simplest of examples. From a engineering science perspective, if you want to find the how much a electricity line heats up when a certain amount of power flows through it, you can form an equations with load as one key variable. You can now perform sensitivity analysis for showing how varying loads affect the temperature of the line.
One of the simplest tools to perform sensitivity analysis is Microsoft excel. It has a built in function specifically designed to conduct sensitivity analysis. Try this if you have the relevant data as in a spread sheet. https://www.youtube.com/watch?v=x0BflbKIIss
Hope this helps:)
Thank you all for your answers.
I think it could be a good topic to write articles about especially for those of you who are well-experienced in the area of sensitivity analysis in engineering sciences.
I would appreciate it if you could share links to the relevant literature if you have in mind as well.
Domingo Tavella Ruben Borrajo I would appreciate it if you could also reply to my new question at:
https://www.researchgate.net/post/How_do_you_define_a_dataset_for_glm_when_one_level_of_a_factor_is_meaningless_for_one_level_of_another_factor
- Badges
- Science topic
- Similar topics
- Engineering
- Mechanical Engineering