Many people would argue you could have predicted it’s failure when you saw the “Banebots” sticker on it. The brand has become somewhat notorious for poor design, workmanship, and material quality. Although, I cannot comment on that specific model, so I’ll leave it at that. Perhaps things have improved since those days.
To better answer your question, yes, the failure of a gearbox can absolutely be predicted. The problem can be approached a few different ways:
- All factors and loads known/given, lifetime to be found
- All factors and lifetime given, maximum load to be found
- Lifetime and load given, design/sizing of the gears to be found
After a machine design course spending a few weeks on gear train design, it is not a terribly difficult process to analyze a gear train, but it does take a little while. There’s probably about a dozen pages in my notebook that run through all the calculations.
Also, in gear train design, there are a lot of “fudge factors.” Usually they are referred to as “K factors.” There are about 8 or so of them that deal with temperature, gear quality, mounting configuration, lubricants, etc and can greatly affect the design and calculation of the gear train. Many times, these K factors cannot be calculated and must be estimated. So, while the failure (or success) of a gearbox can be calculated, its lifespan is only estimated to some reasonable degree of certainty. Typically in engineering design, when we talk about lifespan, we’re talking greater than 10^4 cycles. For your application, you’re just concerned with if it will work for your robot.
Honestly, I’ve never done the above calculations for a FIRST Robot. There just isn’t time to fully analyze everything. Over time, you just learn what works, and develop a rule of thumb, and build off past designs. If you’re concerned, you throw it into FEA to give you a rough idea if you’re screwed or way in the clear.
To determine the safe loading for that gearbox, the easiest way would be a destructive test, like you mentioned. Lock one side, put a torque meter (digital torque wrench perhaps, rod and weights?) on the other side, and have at it until something breaks. If you wanted to, the maximum safe load could be calculated, but you’d need all the gear pitches, tooth counts, material, face width, hardness, etc.
There is not any listing I know of that indicates the maximum allowable torque for common gearboxes used in FRC.