Hup
Mechanical
- Jan 13, 2006
- 36
Hello,
I'm trying to find a way, how to build FE models of thick welded plates as accurately as possible. The plates should be represented by shells, because usually we need to simulate large structures and solid modelling of details is simply impossible.
Commonly we have a line of nodes shared by both (all) welded plates - shells. This represents "absolutely stiff" connection with shared DOFs at the interface.
Everybody knows, that real welds don't behave this way. There are many types of them, they can be continuous or not, there are stress concentrators,....
-> they are far to be "absolutely stiff".
If we want to improve the model, there must be a compliant interface added between the shells. As an arbitrary stiffness could be used MATRIX27 element. Thus if we have two disconected shells with pairs of (probably coincident) nodes, this stiffness (adjusted to represent the real weld-section behaviour) introduced "between" each pair could probably do the work.
I see, that the biggest problem is to adjust the MATRIX appropriatelly.
Does anybody have some experience, theory, thought?
Thanks, H-up.
I'm trying to find a way, how to build FE models of thick welded plates as accurately as possible. The plates should be represented by shells, because usually we need to simulate large structures and solid modelling of details is simply impossible.
Commonly we have a line of nodes shared by both (all) welded plates - shells. This represents "absolutely stiff" connection with shared DOFs at the interface.
Everybody knows, that real welds don't behave this way. There are many types of them, they can be continuous or not, there are stress concentrators,....
-> they are far to be "absolutely stiff".
If we want to improve the model, there must be a compliant interface added between the shells. As an arbitrary stiffness could be used MATRIX27 element. Thus if we have two disconected shells with pairs of (probably coincident) nodes, this stiffness (adjusted to represent the real weld-section behaviour) introduced "between" each pair could probably do the work.
I see, that the biggest problem is to adjust the MATRIX appropriatelly.
Does anybody have some experience, theory, thought?
Thanks, H-up.