First, allow me to summarize my understanding of the question. There is a high vibration in some steel structure connected to two fans. This vibration did not exist in the past, but does now. You believe that the fans have been properly balanced.
Both of the previous replies are absolutely correct and appropriate. But, based on my interpretation of the problem, I feel I can make a few comments.
If a vibration has increased, then something changed. The change could be an increase in the excitation or a change in the response of the steel structure. The excitation could come from imbalance, misalignment, surging, speed change, mechanical damage or a number of other causes. If there are known changes in any of these, then that should suggest that they may be the cause. If the blower was recently rebalanced, then TPL is absolutely correct. The balance may not have been performed correctly or the balance specification may not have been adequate. The same point could be made regarding alignment. If the operation of the system has changed (flow rate, louver operation or speed), then the new operating conditions could be causing a new excitation. We have some fans that experience very high vibration at low flow. For one particular fan, we were able to establish that when the louvers closed below 20%, the fan would experience surge condition.
If there was no change in the excitation, then there could be a change in the response of the steel structure. Any change that could have altered the mass or stiffness could have changed the natural frequency of the structure so that it is now resonant at the same frequency as some vibration produced by the fan (fan running speed, two times fan running speed, vane pass frequency, gear mesh frequency, driver running speed, etc.) This could be checked by changing the mass or stiffness of the structure. Depending on how it's made, that can be very easy. If you place a bag of sand on top of the structure and the vibration stops, then it would suggest it was resonating and the extra mass de-tuned it so it does not resonate at that frequency any more.
But, as noted in the first reply by Tmoose, you need to know that the frequency content of the vibration is. That would eliminate at lot of the guess work.