It's not just about selecting a level of vibration. There's a lot more that goes into developing a vibration test program that does what you're talking about. There's probably not any one vibration test that will address everything you're looking to do.
First, you need to understand the vibration environment that your product will see. Are you only interested in the operating environment? What about the vibration during manufacturing, transport and installation. If it's portable, you should consider that as well.
Second, what kind of thing are you designing? A mass produced near-commodity product, say something along the lines of a cheap clock radio? Or is it a unique high cost item, like a satellite that can't be damaged in testing? Or something in the middle of that spectrum? How much are you willing to invest in testing? How much are your customers willing to pay for the resulting reliability? At a certain point, it's better to simply replace/repair a product than it is to design the thing that will last forever.
When developing a vibration test program, intensity is only one component of the input, frequency is another. The rotating frequency of your motor will likely dominate, but there will be other factors as well and if you're serious about this, it will pay off to develop a random vibration profile based on your device's vibration environment. Does the operating vibration of your device excite any of the resonances of your subcomponents/systems? I've seen transformers and large caps pull away from pcb's because the vibration the system experienced was at the component's resonant frequency.
How many degrees of freedom are significant? In other words, is the vibration mainly up/down, right/left, fore/aft? What about pitch, roll yaw?
To deliver a product out of the gate with high reliability, you're talking about developing a HALT/HASS (Highly Accelerated Life Test/ Highly Accelerated Stress Screening) test program for your product. You'll have to break a few test samples, but if that kind of reliability is what you're after, HALT/HASS will get you closer than just picking a few specs from the internet and testing to those.
Basically, the plan goes something like this:
- Identify frequencies of interest (Operating frequencies, component resonances, environmental inputs etc.) You can do this experimentally (sine sweep, resonance search), analytically or in some combination.
- Collect vibration data from the device in operation for each axis you want to test and for each, compute a PSD (Power Spectral Density) which tells you how much energy is distributed over any bandwidth. That level represents one standard deviation away from the mean acceleration of zero and the integral of the PSD plot is the overall RMS acceleration.
- Start shaking the DUT (device under test) while it's operating at increasing levels of vibration until something breaks. Find and fix that failure and repeat until something else fails and so on. Each time something fails, use a time varying stress model, typically some sort of cumulative damage model, to estimate the B1 life (the time at which reliability = 99% at 100% stress) Typically, this data analysis would be done with software like ALTA.
- If the B1 life isn't enough at that point, fix the issue and retest.
Typically, multiple units will be tested at once until they all fail or the B1 life prediction becomes adequate. Fix the failures that showed up during testing and re-test.