A linear step followed by an NLGEOM step: beams and shells with rotational SPCs stop with "*ERROR in meanrotmpc: MPC has wrong number of terms" (exit 201)

Analysis issues
1

Summary

A B31 cantilever clamped with 1, 1, 6, a linear *STATIC step with a tip force, then a second *STATIC step with *STEP, NLGEOM: the second step stops at once with

 *ERROR in meanrotmpc: MPC has wrong number of terms

and exit 201. The same deck with NLGEOM on both steps runs (tip u_y = 1.421378E-01 in both steps). An S4 strip whose clamped edge is held in DOF 1-5 does the same. A B32R cantilever stops the same way when a PERTURBATION *BUCKLE step sits between a linear and an NLGEOM step.

The 2.23 manual allows NLGEOM to be switched on in a later step: section 7.125 (*STEP), “Once the NLGEOM parameter has been selected, it remains active in all subsequent static calculations.” The message is internal; it names neither the node nor the cause, and the user’s input is valid.

Version

ccx 2.23 (macOS arm64) and current master (537aa81, built on macOS arm64): the same message and numbers.

Already reported or fixed?

Not fixed on master (537aa81). I found no GitHub issue or pull request, and no forum topic title, about this message.

Related:

Minimal input

linear_then_nlgeom.inp:

** B31 cantilever, clamped at node 1. Step 1: linear static, tip force.
** Step 2: the same load, now with NLGEOM.
*NODE, NSET=NALL
1, 0., 0., 0.
2, 10., 0., 0.
3, 20., 0., 0.
*ELEMENT, TYPE=B31, ELSET=EB
1, 1, 2
2, 2, 3
*MATERIAL, NAME=STEEL
*ELASTIC
210000., 0.3
*BEAM SECTION, ELSET=EB, MATERIAL=STEEL, SECTION=RECT
1., 1.
0., 0., 1.
*BOUNDARY
1, 1, 6
*STEP
*STATIC
*CLOAD
3, 2, 1.
*NODE PRINT, NSET=NALL
U
*END STEP
*STEP, NLGEOM
*STATIC
*NODE PRINT, NSET=NALL
U
*END STEP

nlgeom_both.inp: the same with *STEP, NLGEOM on the first step too.

linear_then_nlgeom_s4.inp: two S4 elements (nodes 1-6 at x = 0, 10, 20 and y = 0, 10), thickness 1, *BOUNDARY 1, 1, 5 and 4, 1, 5, step 1 linear with *CLOAD 3, 3, 1. and 6, 3, 1., step 2 *STEP, NLGEOM with no new load.

How to run

ccx linear_then_nlgeom
ccx nlgeom_both
ccx linear_then_nlgeom_s4

Expected

The second step runs and gives the NLGEOM response to the load, as nlgeom_both.inp does.

Actual

deck step 1, node 3 step 2 exit
linear_then_nlgeom 1.418136E-16, 1.421627E-01, 5.046600E-12 *ERROR in meanrotmpc: MPC has wrong number of terms 201
nlgeom_both -6.203952E-04, 1.421378E-01, about 1E-16 -6.203952E-04, 1.421378E-01, about 1E-16 0
linear_then_nlgeom_s4 u_z 2.640273E-02 the same *ERROR 201

Where it goes wrong

Line numbers are master 537aa81 (2.23 in brackets).

  1. The rotational SPCs of a beam or shell node without a knot become mean-rotation MPCs with label MEANROTBS (2.23 manual section 10.3.5; gen3dboun.f), and the translational SPCs become linear MPCs on the same expanded nodes.
  2. In a step without NLGEOM cascade.c:109 [109] treats every MPC as linear ((*iperturb<2)) jmpc[i]=0;), and CalculiX.c:1178 [1163] decascades in the first step. So the linear MPCs are substituted into the MEANROTBS MPC in place, which changes its number of terms, while its label stays. (Read from the source, not checked in a debugger.)
  3. In the NLGEOM step nonlinmpc.f:847-849 passes the MEANROTBS MPC to umpc_mean_rot.f, which expects 3 terms per node plus the rotation term and stops at lines 58-64 [same] (nkn=(n-1)/3, if(3*nkn.ne.n-1)).

When the first step is nonlinear, cascade.c treats MEANROTBS as a nonlinear MPC and, called from the main program, does not substitute into it (lines 162-176 [same]: “no expansion if there is a dependence of a nonlinear MPC on another linear or nonlinear MPC and the call is from main”), which is why nlgeom_both.inp runs.

Fix

Not built or tested here. CalculiX.c already stores the undecascaded MPCs in the first step (lines 1160-1172 [1145-1157], nodempcref, coefmpcref: “storing the undecascaded MPC’s if needed … 1) in the first step always”); *EQUATION, REMOVE in a later step already restores it (equations.f lines 110-127). Restoring it (and decascading anew) at the start of the first NLGEOM step that follows a linear step would cover this case, as pull request 252 does after a linear step that follows a nonlinear one.

Workaround: put NLGEOM on the first step as well.

2

Confirmed using official ccx 2.23 build on Windows, needs a GitHub ticket.

linear_then_nlgeom.out:


 Static analysis was selected

 Newton-Raphson iterative procedure is active

 Nonlinear geometric effects are taken into account

 Determining the structure of the matrix:
 Using up to 1 cpu(s) for setting up the structure of the matrix.
 number of equations
 48
 number of nonzero lower triangular matrix elements
 813

 increment 1 attempt 1
 increment size= 1.000000e+00
 sum of previous increments=0.000000e+00
 actual step time=1.000000e+00
 actual total time=2.000000e+00

 iteration 1

 *ERROR in meanrotmpc: MPC has wrong number of terms
1 Like
3

Posted: A linear step followed by an NLGEOM step: beams and shells with rotational SPCs stop with "*ERROR in meanrotmpc: MPC has wrong number of terms" (exit 201) · Issue #325 · Dhondtguido/CalculiX · GitHub

1 Like