You are on page 1of 2

http://www.colorado.edu/engineering/CAS/courses.d/IFEM.d/Home.

html Sixth Degree of Freedom of Shell element Quad elements have five degrees of freedom per node in the element local (EL) coordinate system, namely three translational and two rotational degrees of freedom. The third rotational degree of freedom, the rotation around the normal of the element plane (often called the "drill") has no stiffness. So, when the rotations are defined in the E-L system, one can define five degrees of freedom per node point and define these degrees of freedom as global computational (G-C) degrees of freedom. However, when two flat shell elements are not in the same plane, each of the nodes on the intersection between the two elements has an own E-L systems with 5 degrees of freedom. For the definition of the G- C degrees of freedom in these nodes one must choose a specific coordinate system. This means that one of the axes of the computational coordinate system will not coincide with both drill axes in these nodes. A consequence is that the two rotational local degrees of freedom have to be decomposed in three rotational G-C degrees of freedom (see figure below). Thus, six instead of five degrees of freedom per node have to be defined in order be able to carry out the matrix decomposition. In practice this means that in the element routine the two E-L rotational degrees of freedom are decomposed in three rotational degrees of freedom defined in the Branch Local (B-L) coordinate system (the same system in which the node point coordinates are defined). The disadvantage is that one of the three components will depend linearly on the other two components and that, if the angle between the E-L-system and the GC-system is very small, one of the components is small with respect to the other two components. Thus, the stiffness matrix will be singular or almost singular (ill conditioned). This problem can be solved in different ways in B2000. For flat structures the method described in the section called Definition of the geometry in branches is suitable. For structures with cylindrical parts the methods described in the section called Removing a small stiffness components in element stiffness matrix and the section called Defining large drill stiffness can be applied. For folded structures (for instance a hat stiffener) a combination of the methods described in the section called Definition of the geometry in branches and the section called Defining large drill stiffness can be applied. PARAM, K6ROT in SOL 106

The shell elements CQUAD4 and CTRIA3 do not have any rotational stiffness defined in the direction normal to the plate, sometimes called the drilling degree of freedom. In geometric nonlinear analysis, this degree of freedom cannot safely be constrained (as it would be in linear analysis with AUTOSPC for instance) without the possibility of over-constraint. In order to remove the singularity which remains, a small fictitious rotational stiffness is added to these drilling freedoms which is factored by the K6ROT value specified by the user. The small fictitious stiffness is added to the relative rotation of the element to the GRID point; this prevents over constraining rigid body motion. The calculated small fictitious value is then multiplied by the value specified on the K6ROT parameter.

In linear solutions, the default for K6ROT is zero; that is no fictitious drilling stiffness is added. But if K6ROT > 0.0 in linear (or nonlinear) analysis, then a small stiffness is added to the normal rotation for the CQUAD4 and CTRIA3 elements. Although intended primarily for geometric nonlinear analysis, this is an alternate way to suppress grid point singularities in linear analysis and can be useful if two adjacent plates are slightly non-coplanar. In this case, each of the adjacent plates offers a small, but significant stiffness to the other in the drilling freedom, and this small stiffness may be large enough for AUTOSPC to ignore it, but may be small enough to result in numerical ill-conditioning. This situation can always be remedied by remeshing so that the adjacent stiffness goes either lower, to be trapped by AUTOSPC (in which case it may be a potential over-constraint), or goes higher to remove the ill-conditioning; a value of 1.0 for K6ROT can be used to alleviate this problem.

For nonlinear analysis, the default value of K6ROT is 100. In most instances, the default value can be used, but for highly nonlinear problems the value may have to be increased. Too small a value will cause convergence difficulties in these cases, but too large a value may cause the solution to diverge, or even worse, the solution may succeed, but the answers may be inaccurate. Generally speaking the smallest value to achieve convergence should be used, and experience has shown that values in the range 100 to 10000 are acceptable.

You might also like