Dear colleagues,
I am currently conducting a non-linear Pushover Analysis using a Finite Element Method (FEM) software. To ensure the accuracy of my numerical model, I am seeking to validate it against experimental results. Specifically, I am interested in calculating the error between the obtained numerical results and the experimental data and what is the limit error value.
I kindly request your guidance on the most appropriate metrics for calculating the error in my Pushover Analysis. Any advice, references, or past experiences you can share with me would be greatly appreciated.
Thank you in advance for your assistance.
Best regards,
Ismail
How to validate your numerical model is an important question. I profess that I am not familiar with pushover analysis. But I am very familiar with the validation procedure. Let me share with you two particular situations that I like to cite and examples. The first one is the ship stability in which the ship rolling (like a pendulum) is analyzed. In the 196x to 199x people ignored the added mass properties effects. Of course numerous experimental evidences were provided. Now the analysis is considered simply garbage while it helped develop perturbation methods. Clearly the evidences were either fabrication or simply not related to the problem. Another famous case is the analysis of die casting (a manufacturing process). People used various methods (FE, finite volume, finite difference etc.) and of course their work were supported with numerous experimental ``evidences'' . Again it turns out that these evidences either were fabricated or not related. So my suggestion is that you have to check that your work does not violate known physical laws. Then attempt to build a simplified model and check if you work approaching that model.
I'd like to @Genick-Bar-Meir mention, that you shouldn't forget about the FEM and experimental results are not exact themselves. First you have to estimate errors in FEM analysis and experimental research and after try to find suitable norm - its my opinion. Good luck
First you calculate an experimatal results and absolute errors, and also calculate numerical results and absolute errors saparatly at same interval. Then compare the results of both methods at the same interval. you can understand the accuracy of both methods.
Otherwise, If exact solution of the problem is not calculated then you can calculate relative errors numerecally and experemantly and compare its accuracy.
- Badges
- Science topic
- Similar topics
- Mathematics