To whom it may concern,
I am modelling a single storey single bay steel frame. The connections between the beam and column in four corners of this steel frame are made of several pins. I have considered release moment for the beams.
I attached the spf file in the dropbox: https://www.dropbox.com/s/h08d8gzmbht3lek/no.spf?dl=0
After the eigenvalue analysis for this model, I can get the good natural frequency. However, when I try to run the static time-history analysis, the error "unable to apply the entire permanent load" appears. I understand there are a lot of topics in the forum discussing the "unable to apply", I have read these topics and try to fix the problem by others' suggestions such as change the elements settings, convergence criteria, or increase the steps. However, it still does not work. I understand that apply the moment releases might lead to convergence difficulties, but if I do not add the moment release, the natural frequency from the eigenvalue analysis is not correct.
I also try to delete the moment release, the model can run for static time-history analysis. However, when I tried to compare the hysteretic curves from modelling with the experimental results, they are quite different. The peak load for modelling can reach 340KN, but the peak load for experiment is about 40KN. I do not know where I made the mistakes for the model. Could anyone please give me some suggestions?
Best regards,
Summer
Static time-history analysis
Re: Static time-history analysis
zhwang, I took a quick look at your model. I wonder what is the purpose of the links, 'joint_1' and 'joint_2'? With the connectivity you have for the beam elements 'b_1_5', 'b_5_2', and 'b_2_6', the links appear to do nothing.
With node 'N_5' restrained but node 'N_1' unrestrained, the applied permanent load would place a large demand on member 'B_1_5', similarly the case for member 'B_2_6'.
The connectivity looks odd, but perhaps it is indeed your intention. I wonder if you could sketch it out on paper and make certain that the member connectivity represents your intended structure?
In particular, the connectivity of beams 'B_1_5' and 'B_5_2' seems very unusual and I wonder if you have mistakenly specified the end nodes for one or both of these?
Your member releases are also a bit unusual. You have specified moment releases about both axes at one end only of the pinned members. I wonder if your intention was to specify M3 release at both ends instead? This is a plane frame model with restraints out-of-plane (global y), and you previously stated the members were pinned, so I don't see what the purpose of M2 releases serves and your members are pinned at only one end as defined in the model.
Keep pressing forward, zhwang. You'll learn a lot in the process.
With node 'N_5' restrained but node 'N_1' unrestrained, the applied permanent load would place a large demand on member 'B_1_5', similarly the case for member 'B_2_6'.
The connectivity looks odd, but perhaps it is indeed your intention. I wonder if you could sketch it out on paper and make certain that the member connectivity represents your intended structure?
In particular, the connectivity of beams 'B_1_5' and 'B_5_2' seems very unusual and I wonder if you have mistakenly specified the end nodes for one or both of these?
Your member releases are also a bit unusual. You have specified moment releases about both axes at one end only of the pinned members. I wonder if your intention was to specify M3 release at both ends instead? This is a plane frame model with restraints out-of-plane (global y), and you previously stated the members were pinned, so I don't see what the purpose of M2 releases serves and your members are pinned at only one end as defined in the model.
Keep pressing forward, zhwang. You'll learn a lot in the process.
Tim Huff
Re: Static time-history analysis
Hi huffte,
Thank you very much for your reply.
My experimental frame is attached in https://www.dropbox.com/s/icn8wuf3f04hr ... e.jpg?dl=0
The bottom beam is made of I section steel, and below it is a flat plate which is connected to the ground. The bottom beam is 2850mm, which is a little larger than the distance between N_1 and N2, so I define N_5 and N_6. Since there are some pins in N_1 and N_2, so I put the links joint_1 and joint_2 at the location of N_1 and N_2. I am not sure whether this is correct. Do you think the model including the bottom beam does not make any sense? Should I delete the bottom beam and restrain N_1 and N_2 instead?
Actually, I am not sure which place I should put member releases. Based on my picture, could you please give me some suggestion about the way for moment release.
I am also confused with my current model, but I am not sure where are the mistakes.
I really learn a lot during this process. Thank you very much for your help.
Thank you very much for your reply.
My experimental frame is attached in https://www.dropbox.com/s/icn8wuf3f04hr ... e.jpg?dl=0
The bottom beam is made of I section steel, and below it is a flat plate which is connected to the ground. The bottom beam is 2850mm, which is a little larger than the distance between N_1 and N2, so I define N_5 and N_6. Since there are some pins in N_1 and N_2, so I put the links joint_1 and joint_2 at the location of N_1 and N_2. I am not sure whether this is correct. Do you think the model including the bottom beam does not make any sense? Should I delete the bottom beam and restrain N_1 and N_2 instead?
Actually, I am not sure which place I should put member releases. Based on my picture, could you please give me some suggestion about the way for moment release.
I am also confused with my current model, but I am not sure where are the mistakes.
I really learn a lot during this process. Thank you very much for your help.
Re: Static time-history analysis
It appears that your experimental setup is a displacement history imposed at the top of the frame with the base remaining fixed. If so, then it seems to me that your model could be greatly simplified.
I do not see the point of nodes N_6 and N_7. I suppose you could keep the base beam, but connect it to nodes N_1 and N_2, using restraints at these nodes for x, y, z, rx, and rz.
I do not see the point in the joint pins either. Nodes N_1L, N_2L, N_3L, and N_4L are thus unnecessary.
The middle beam on top could have M2a and M2b releases at both ends. Then you would need moment releases nowhere else.
Try to make sense of the above suggestions. It is my interpretation of your model based on the picture. See if this makes sense and if it does, in fact, represent your physical structure.
Best of luck zhwang.
I do not see the point of nodes N_6 and N_7. I suppose you could keep the base beam, but connect it to nodes N_1 and N_2, using restraints at these nodes for x, y, z, rx, and rz.
I do not see the point in the joint pins either. Nodes N_1L, N_2L, N_3L, and N_4L are thus unnecessary.
The middle beam on top could have M2a and M2b releases at both ends. Then you would need moment releases nowhere else.
Try to make sense of the above suggestions. It is my interpretation of your model based on the picture. See if this makes sense and if it does, in fact, represent your physical structure.
Best of luck zhwang.
Tim Huff
Re: Static time-history analysis
Hi huffte,
Thank you very much for your kind suggestions. Following your suggestion, I made some change to my model, and attached it
https://www.dropbox.com/s/h08d8gzmbht3lek/no.spf?dl=0
For the eigenvalue analysis, the natural frequency is quite good. However, the hysteretic curves for static time-history analysis are quite different from my experiment. The peak load for modelling hysteretic curves is about 40KN which is similar with my experimental results. However, the modelling hysteretic curves look like straight lines, but the experimental results are much fuller curves. I also attached these curves here.
https://www.dropbox.com/s/i77ottnz817oi ... g.jpg?dl=0
https://www.dropbox.com/s/q1l4nin76g2qm ... t.jpg?dl=0
Could anyone suggest me why this difference happens and how to solve this problem? Thanks again for your help.
Best regards,
Summer
Thank you very much for your kind suggestions. Following your suggestion, I made some change to my model, and attached it
https://www.dropbox.com/s/h08d8gzmbht3lek/no.spf?dl=0
For the eigenvalue analysis, the natural frequency is quite good. However, the hysteretic curves for static time-history analysis are quite different from my experiment. The peak load for modelling hysteretic curves is about 40KN which is similar with my experimental results. However, the modelling hysteretic curves look like straight lines, but the experimental results are much fuller curves. I also attached these curves here.
https://www.dropbox.com/s/i77ottnz817oi ... g.jpg?dl=0
https://www.dropbox.com/s/q1l4nin76g2qm ... t.jpg?dl=0
Could anyone suggest me why this difference happens and how to solve this problem? Thanks again for your help.
Best regards,
Summer
Re: Static time-history analysis
I think you may need to re-examine your loading, zhwang.
You have a time step of 120 seconds. But your first peak occurs at 60 seconds, with every peak thereafter occurring at 120 seconds after the previous one. So, unless I miss something, you miss every peak in the loading history with your analysis time step.
Secondly, it appears that you are applying a 20mm displacement at the top of the column since your multiplier is 1 in the applied loading. This is a very small displacement and I am not sure you will get anything into the inelastic range with such a small displacement.
It appears that the experiment is simply testing the moment at the base of the column since the top beam is all pinned. Is that correct? Where exactly is inelastic response expected and what response parameter are you most interested in tracking.
So double-check your time step, double-check your loading magnitude, and happy modeling.
You have a time step of 120 seconds. But your first peak occurs at 60 seconds, with every peak thereafter occurring at 120 seconds after the previous one. So, unless I miss something, you miss every peak in the loading history with your analysis time step.
Secondly, it appears that you are applying a 20mm displacement at the top of the column since your multiplier is 1 in the applied loading. This is a very small displacement and I am not sure you will get anything into the inelastic range with such a small displacement.
It appears that the experiment is simply testing the moment at the base of the column since the top beam is all pinned. Is that correct? Where exactly is inelastic response expected and what response parameter are you most interested in tracking.
So double-check your time step, double-check your loading magnitude, and happy modeling.
Tim Huff
Re: Static time-history analysis
Dear huffte,
Thank you very much for your kind suggestions. I have made a change for the time step of 60 seconds. This time I can get every peak in the loading history with my analysis time step.
The spf file in attached:
https://www.dropbox.com/s/h08d8gzmbht3lek/no.spf?dl=0
However, I still cannot get a good hysteretic curves, actually the modeled curves look much thinner than the experimental hysteretic curves. I have attached both curves here.
https://www.dropbox.com/s/ta1z3jttuyh2e ... l.jpg?dl=0
https://www.dropbox.com/s/q1l4nin76g2qm ... t.jpg?dl=0
You are correct, I applied a 20mm displacement at the top of the column in the applied loading. However, the experiment tested the moment also at the top of the column, which in node N4 in my spf. model. Now, I am trying to compare the hysteretic curves of node N4 in my model with the experimental testing hysteretic curves in node N4. If the comparison is good, it means the model can represent my physical steel frame.
However, it appears that my current model should have some mistakes, but I cannot know where it is. Do you know where the mistakes happen in my model? I really learn a lot from you. Thanks again for your help.
Best regards,
Summer
Thank you very much for your kind suggestions. I have made a change for the time step of 60 seconds. This time I can get every peak in the loading history with my analysis time step.
The spf file in attached:
https://www.dropbox.com/s/h08d8gzmbht3lek/no.spf?dl=0
However, I still cannot get a good hysteretic curves, actually the modeled curves look much thinner than the experimental hysteretic curves. I have attached both curves here.
https://www.dropbox.com/s/ta1z3jttuyh2e ... l.jpg?dl=0
https://www.dropbox.com/s/q1l4nin76g2qm ... t.jpg?dl=0
You are correct, I applied a 20mm displacement at the top of the column in the applied loading. However, the experiment tested the moment also at the top of the column, which in node N4 in my spf. model. Now, I am trying to compare the hysteretic curves of node N4 in my model with the experimental testing hysteretic curves in node N4. If the comparison is good, it means the model can represent my physical steel frame.
However, it appears that my current model should have some mistakes, but I cannot know where it is. Do you know where the mistakes happen in my model? I really learn a lot from you. Thanks again for your help.
Best regards,
Summer
Re: Static time-history analysis
I believe that the problem is twofold:
1. The application of moment releases to the T-beam, an inelastic force-based element, as noted in the warning when you begin the analysis, can create significant convergence problems.
2. The large time step relative to the loading description.
If I eliminate the moment releases on the T-beam and increase the number of steps to 3024, I begin to see the type of hysteretic behavior you expect.
However, the elimination of the moment releases may not be appropriate, depending upon your physical connection details. It the T-beam connections truly are non-moment-resistant, then the T-beam should participate very little in the lateral response, acting merely as a strut between nodes 7 and 8, in which case perhaps an elastic frame element with the moment releases or a truss element without the moment releases might suffice. Or, you could try to employ link elements with very small rotational stiffness to connect the T-beam to the other beams.
So, in summary, (1) increase the number of steps in the analysis and (2) develop a modeling strategy which maintains the physical attributes of the test while eliminating the application of moment releases to inelastic elements.
Best of luck, zhwang.
1. The application of moment releases to the T-beam, an inelastic force-based element, as noted in the warning when you begin the analysis, can create significant convergence problems.
2. The large time step relative to the loading description.
If I eliminate the moment releases on the T-beam and increase the number of steps to 3024, I begin to see the type of hysteretic behavior you expect.
However, the elimination of the moment releases may not be appropriate, depending upon your physical connection details. It the T-beam connections truly are non-moment-resistant, then the T-beam should participate very little in the lateral response, acting merely as a strut between nodes 7 and 8, in which case perhaps an elastic frame element with the moment releases or a truss element without the moment releases might suffice. Or, you could try to employ link elements with very small rotational stiffness to connect the T-beam to the other beams.
So, in summary, (1) increase the number of steps in the analysis and (2) develop a modeling strategy which maintains the physical attributes of the test while eliminating the application of moment releases to inelastic elements.
Best of luck, zhwang.
Tim Huff
Re: Static time-history analysis
Dear huffte,
Thank you very much for your kind suggestions. I have tried all your suggestions, including elastic frame element with moment releases for T-beam, truss element without moment releases for T-beam, and link element connect T-beam to the other beams. Unfortunately, all these do not have a good result, and they changed the natural frequency as well.
My experimental frame is attached in https://www.dropbox.com/s/icn8wuf3f04hr ... e.jpg?dl=0
For this frame, there are bolt connections on N1, N2, N3, and N4. I think these bolt connections may made major effect on the hysteretic behaviour of this bare steel frame.
In my previous model, I did not consider these bolt connections. Now, I would like to add some link element to N1, N2, N3 and N4.
Do you have any other suggestions for me? Thank you very much.
Best regards,
Summer
Thank you very much for your kind suggestions. I have tried all your suggestions, including elastic frame element with moment releases for T-beam, truss element without moment releases for T-beam, and link element connect T-beam to the other beams. Unfortunately, all these do not have a good result, and they changed the natural frequency as well.
My experimental frame is attached in https://www.dropbox.com/s/icn8wuf3f04hr ... e.jpg?dl=0
For this frame, there are bolt connections on N1, N2, N3, and N4. I think these bolt connections may made major effect on the hysteretic behaviour of this bare steel frame.
In my previous model, I did not consider these bolt connections. Now, I would like to add some link element to N1, N2, N3 and N4.
Do you have any other suggestions for me? Thank you very much.
Best regards,
Summer