I am trying to validate a small-scale rough tension pile embedded in sand using PLAXIS 3D. I am using a volume pile with the linear elastic constitutive model for concrete and sand with MC constitutive model. The volume pile was modeled with an interface by providing an R inter value of 0.75. The attached image is the comparison of the uplift load-deformation curve of the FE model and experimental model. I have also modeled a similar tension pile using field data from literature resulting in a similar curve with the FE model giving a linear response. I have used different meshes and different constitutive models also but the response of the FE model remains somewhat the same.
Is the numerical model considered validated when we are getting such a response? Is there anything I can do to improve the model further?
Dear Jerin,
Have you tried using a more advanced model to model the sand? Because now you're using a elastic model for sand, which means that also the soil-pile interaction will be pretty much elastic with constant stiffness, which would explain the response you get. Modelling the sand with for instance Hardening Soil model may improve this.
With kind regards,
Dennis Waterman
Hello Mr. Waterman,
I did try to model a similar problem using different constitutive models for instance I am sharing the figure to show another validation example of a field test uplift test of sand.
I had hope adding the elasto-plastic behaviour to the soil would be sufficient, but I see from your graph that it isn't. Hence, the solution is in the modelling of the soil-pile interaction. You have now used the simple method of an Rinter value in the interface, but this may be too simplified. Since the soil around the pile has been remoulded due to the pile installation the behaviour may be more complicated than just reduced strength and stiffness with the same factor. Not to mention that also the stress state around the pile was altered by the installation. Maybe the interface requires its own material set with manually reduced stiffness and strength parameters. So my suggestion would be to give the interface its own material set. Initially just a copy of the material set used for sand and then try to reduce some of the parameters. Since the interaction is mostly shearing and it seems that the pile behaviour is too stiff in the beginning, I would start to reduce E50ref to see what improvement that brings.
Thank you very much for the suggestions. I will definitely try to implement your idea. I do have another query whether the current model can be used to predict an ultimate tension load using the load-deformation curve for example if I am assuming the ultimate load is at 10% of deformation.?
You're currently doing a back calculation, which is not a prediction. So if you find a consistent method that gives good results on multiple back calculations under well defined conditions then that method may be used for a prediction as well under the same conditions.