Your findings

Congratulations on your investigation using sinusoidal
vibration. If you were to dwell at each of these
frequencies, first attaching many accelerometers, you'd
find nodes (lines of zero vibration) and antinodes (lines
of maximum vibration) between the nodes. Phase reverses
when you cross a nodal line.

Of course you'll not attach many accelerometers (even if
there were space to do so), as their presence would
modify the vibration response. But you might scan your
card surface with a laser and get a picture of the modal
response at each frequency.

Please see Animation 10-2 on printed wiring modes on the
CD that accompanies my 2005 text. Animation 10-2 and the
accompanying Figures 10-19, -20 and -21 were provided by
John Starr of Cirvibe, Inc., along with assistance in
writing the accompanying Section 10.6, titled Printed
Wiring Cards. The prior nine chapters help readers
understand Chapter 10, which is entitled Continuous Systems (such
as printed wiring boards).

Realize, please, that all four of your resonances are
simultaneously and continuously excited by the repetitive shock excitation
you are probably using for your HALT and HASS activities.

If I had your e-mail address I could send you Figures
10-18A, -18B and -18C, which had not yet been given me in
early 2005, when I was finishing my text. They show the
kind of damage that you are experiencing. John and I use
them in our lecture courses (mine are less focused on
PWBs than are John's and he has more PWB illustrations
than I do).

Please see
http://www.equipment-reliability.com/course10.htm and
http://www.equipment-reliability.com/course4.htm. These
describe John's upcoming events at Minneapolis and at Las
Vegas. Either of these events will greatly aid your
understanding of your current PWB resonance problems and
will suggest some specific remedies.

I asked John to comment on my draft (above). He agreed with my response, but added:

Can't modify what you have given to tell him how to resolve
the problem. Acceleration measurements at the point of failure are
meaningless. Components fail due to board flexure. He needs my
course to understand his problem and solution. A message
board answer will not address all critical issues and solve
this problem or avoid similar problems in other designs.

Wayne