971 has run belts in their drivetrains for 3 years now so we have seen both the failures and successes of belt drivetrains. Below are some thoughts from what the team has learned. Sorry for the long post, but we've learned a lot, so I think it's best to get the information out there.
Tensioning belts "correctly" is not necessarily obvious. I see people saying that you should always tension belts coreectly, but to many teams the correct tension is difficult to attain. Some teams might take "correct" tensioning to be a tight belt, but this is not the case. An overtensioned belt can be just as bad as an undertensioned one. Tensioning belts correctly often invovles finding some sort of sweet spot that is not too loose or too tight. 971 has started to tension based on the tone created when strumming the belt. We have found that depending on the tension, the frequency of the tone is different. We are using 45 Hz right now. A smartphone app such as
this one can be used to analyze the tone. The frequency can be calculated given certain equations; I think you can find one on the Gates website.
A rule of thumb is also if teeth are coming off, the belt is undertensioned. The reason this happens is that the teeth ride up out of the pulleys on the slack side. This puts a bending force on them, and the root of the belt tooth is a good stress riser, so the tensile elements break there as opposed to somewhere else.
If tensile elements are tearing out of both sides, the belt is wearing out, possibly prematurely because of overtensioning. If you have wear or some sort of discoloration in the recesses of the belt and noticeable wear on the outer parts of the pulleys, then the belt is probably overtensioned.
For many FRC applications, running belts a little overtensioned is at times necessary, but it's still best to know the amount you're overtensioning your belts. Overall, knowing the warning signs and being consistent with how you tension your belts is going to help your team in the long run.
Another thing to consider when analyzing belt failure is where your robot's center of gravity is. When your cg is higher, you'll get more load on your outer wheels. If you are rocking on the outer wheels of a 6wd while applying full power, you can assume that all torque is being put through your drive belts. We think this is one reason why we we have had belts break under the same conditions last year. As many people can attest, 971's 2014 robot has a much higher cg than 2013's. This could be the case for your team. Be wary of how high your cg is and remember that it can affect your drivetrain too.
As for junk getting caught in belt drivetrains, I have seen this too. Junk seems to get caught in many FRC drivetrains, so its not an isolated problem. With belts, its important to inspect the drivetrain and clean in before a match. Cleaning debris from a drivetrain early is going to be a lot lower maintenance than removing gobs of carpet off a drive pulley.
Another thing: In terms of belt quality, we prefer the Gates GT2 over the VEX HTD belts. We like the manufacturing quality of the GT2 belts so far.
We've found documents like
this and
this are extremeley useful. It's important that before a team jumps from chain to belt that they understand how to use them properly. Belts aren't a magic elixir that will make your drivetrain better immediately. A major change in the drivetrain requires thorough research and testing. Before running belts in our drivetrain, we ran a timing belt elevator, and before we did that we did stress tests to make sure we could do it.
That being said, we still really like belts, and will continue to use belts in many applications, including but not limited to our drivetrains.