The material solution failed for element "x" with material "x".

Hello, members!

I am modeling a concrete-filled steel tube. I have represented two components only: steel tube and concrete core.

1. In order to represent the behavior of steel, I am using the multilinear isotropic hardening plasticity. The relationship stress x strain was formulated by Tao (2014).

2. In order to represent the behavior of confined concrete, it was modeled the linear behavior with a modulus of elasticity and Poisson ratio. 

3. The concrete nonlinearity also was modeled through the relationship stress x strain by Tao (2014) using multilinear kinematic hardening plasticity (This was the only procedure that I found for input a relationship stress x strain with descending stage).

4. The failure stress of concrete was modeled through Drucker Prager concrete, in order to represent the softening behavior through the Fracture Energy HSD Model. I would like to represent exponential softening behavior. However, the answer in the softening stage is linear, instead of exponential behavior that I have imputed. What could be my error?

5. The solution failed even before the plateau stage after the softening stage. What could I improve in order to achieve this stage?

6. I am using four load steps to process the model. First load step to dead load, second load step to the linear stage, third load step to transition from hardening to softening behavior, and finally, the last load step for the softening behavior. This was because I would like to have more points in the curve in the transition stage from hardening to softening behavior. Besides that, I think that I need minimal increments of load in this stage to help the convergence. Also, I am using automatic time-stepping, in order to positively affect the convergence. However, the program has chosen the number of substeps. Is this combination of NSUBST and AUTOTS is appropriate? Is there another way more interesting?

7. Is recommended to add the reinforcement steel bars in the concrete core to improve the model? If yes, what is the better approach for my model?

8. The material properties that I have cited are available in this link:

9. I am not sure if members of Ansys Learning will be able to access attachments, if necessary, I could share my code in this principal box here.

 I wonder if any member could help to respond and resolve these questions.

Thanks in advance.



  • AndreytestonAndreyteston BrazilMember

    Hello, Ashish Khemka!

    Thank you for your attention and important help, have contributed a lot.

    Regarding the first link, I summarized the principal changes that Peter has suggested. 

    1. Change the fixed support to three displacements supports for x=0, y=0, and z=0 on those three planes, in order to avoid Poisson's Ratio horizontal spreading, so no tensile stress developed. In my model has constraints x=0, y=0, and z=0 applied using multipoint constraint elements. Then, I could believe that this change is not applicable to my model.

    2. Change the value for cohesion. In my model, was necessary to input a Drucker-Prager Classic in order to implement the Cohesion value (N/m²), also with him, the value of Internal friction and Dilatancy was necessary using Drucker-Prager Classic. 

    However, I couldn't remark differences in the softening behavior response of the model. I would like to make a question regarding these: I have been using the NUMMERG command to share nodes between the internal area of the steel tube and the external area of the concrete core. In this kind of technology, is there influênce of parameters like cohesion and friction? It is very important for my research that friction is present in the model because the relationship between vertical and horizontal stress is influenced by that.

    3. The gravity load was applied alone in only one load step and after this, the displacement into different sub-steps was applied. In order to make this, is only necessary define the density using the command DENS in the properties of the material?

    4. In the step controls, was changed for automatic time stepping;

    5. In solver controls, was changed to Direct Solver through the command EQSLV,SPARSE. In the previous analyses, I have omitted this choice from default and at the output, was written SPARSE MATRIX DIRECT SOLVER. Is this that I need to do in order to use Direct Solver? 

    6. I have been perceived in the Solver output soon on the second load step, sub-step number 02, that the solver inverts the matrix. Using the explanation by Peter, "That means when the solver takes the nodal displacements and runs them through the material model, the material model cannot cope with those displacements.". I could believe embased on this explanation the necessity for change any parameters on my material model.

    Regarding the second link, I summarized the principal changes that Sandeep has suggested. 

    1. Change to enhanced strain to eliminate reduced integration issues in lower-order elements. 

    2. In the same discussion, Peter suggested: "If you are using a material model that has an equation that becomes undefined if tensile stress is applied to the element, then there is no solution after that point." I am defining properties about tensile stress in my material model, however, the stress tensile doesn't exceed the set limits. 

    Any insight provides would be greatly appreciated.

    Thanks in advance.


  • akhemkaakhemka Forum Coordinator

    Hi @Andreyteston ,

    Are the queries indicated above resolved or are you still seeking answers to them?


    Ashish Khemka

Sign In or Register to comment.