You have created a Simulation study of a fairly large assembly.  It has a mixed mesh – Solid elements, also Shell elements, maybe also some Beams.  You are relying on the GLOBAL CONTACT – BONDED to tie together all the mesh bodies of similar type.  You spent maybe 40 minutes, at least, creating the BONDED contact conditions to tie the dissimilar mesh elements together.
 And now when you try to run the study, you get the failure message, “Zero or indefinite stiffness matrix”.
 Ugh.  That means that at least one component is not correctly bonding to a neighbor, and so is free-floating.  But which one?  Since the study won’t run, you can’t display the displacement plot, and see which part(s) are flying off the screen.  How to diagnose this – without wasting the entire morning?  There are two ways I approach solving the problem, and they both use the study Property option, “Use Soft Springs to Stabilize the Model”.
1:  Moving
Sometimes you can simply turn ‘soft springs’ on, and run the study again.  About half the time, I find that the study will now run, (and produce a warning message, that very large displacements were detected).  Tell it to continue running the study anyway, as-is, without turning on the large-displacement flag.  Then, when you plot the Displacement results, you might very well be looking at an empty screen, or you might see that one (or more) parts seem to be missing from the plot.  They are not missing – they simply went flying off the screen.  Set your Displacement scale factor to be wicked small, (think many leading zeros), and then animate the displacement plot – and you’ll see at the start of the animation, which parts are un-tethered, and in which directions they are free to wander.
A cool and easy trick.  Except that, I did say above, I use this trick “about half the time”.  Because the other half of the time, the applied loads might act directly on the under-restrained parts, in the same direction that they are free to move, and then the “Soft Springs” do not supply enough stabilization.  What to do then?
2:  Shaking
The second approach is to Shake the assembly, instead of moving it.  Create a new Modal Study, (this will require the Simulation license to be Simulation Professional or higher).  Set the Modal Study Properties to only ask for the lowest 2 mode-shapes, or maybe 3, but no more.  (this is a diagnosis, not a true study, so you want it to run fast).  Then you copy all your Restraints and Bonded conditions into the new study.
Wait – I did NOT say to re-create all your Restraints and Bonded contacts.  I  said “COPY” them.  This is a “tip within the tip”.  A lot of people I encounter still don’t know you can do this.  If you activate your first, “real” study, you can then select the folder for your Restraints, and DRAG that folder down to the row of tabs at the bottom of your screen, and DROP it into your new Modal study.  Done.  So creating this Modal study to diagnose your problem should take, like, 30 seconds tops.   You do NOT have to copy over your system of loads – just the restraints and bonding, because we are not going to ‘push’ your assembly in any particular direction, we are instead going to vibrate it in-place.
But, you DO still have to turn on the “Soft Springs” option to stabilize the model.
Ok, now you RUN the Modal study.   Because the underlying equations of the Modal study are being solved differently than a Static study, the problem will generally solve O.K. even with under-supported parts.  And, when you activate the plot for the lowest 2 or 3 mode-shapes, each one should show you some rigid-body motion, (and remember, it could also be a rotation!), of the part or parts that are un-tethered.   Animate the model-shape plot, and it will be patently obvious which parts are walking away, and in which directions.
Important caveat: The Modal Analysis method of diagnosing your problem, will NOT allow you to create sliding or intermittent Contact via the “No Penetration” gap elements.  If your problem also has that kind of Contact in it, you need to either ignore that contact for the purpose of diagnosis, (which will too sloppy, and you’ll have to ask for more Mode Shapes –  ignore the responses you know are not ‘real’);   Or, you could replace the Bonded contact, with either an Elastic Support, or a Spring Connector.  With either of these, make the Normal spring stiffness pretty high, but the tangential spring stiffness very low – and you WILL have over-supported the selected faces a little bit.  But again, you’re just looking for which parts want to walk away, not running the ‘real’ study yet, so we can throw a few darts.  And if the problem produces NO rigid-body modes when you’ve replaced the Gap elements with springs, then you know for sure that the Contact conditions are the real problem.
One of these two methods should solve the “Zero or Indefinite Matrix” problem for you, just about all the time.