Often times the dwells will do just fine without changing them, however if data quality is not what you had hoped, or you otherwise suspect noise on the hose, adjusting them to more appropriately match your roadway speeds will definitely help. After changing them, if the data was noisy (lots of bouncing air pulses -- which can cause double hits) you should notice it "clean up" quite a bit after you change those.

Sometimes the extra noise on hoses cannot be avoided, playing with your physical setup can help some, hose length and where the vehicles are hitting the hose can be factors with "echos" (air pulses bouncing back and forth causing multiple triggers of the channel for the same hose hit). Some of the factors cannot be controlled (such as the vehicle tires themselves -- some cause larger secondary pulses than others). Generally you will want to coil excess hose on the counter side of the road, that is one simple rule that will help keep noise down. This is for two reasons: A little extra length to the counter can sometimes help quiet things down a bit just due to the air pulses having to travel further to the counter, and even more importantly, if there are echos the closer the vehicles are to the end of the hose means the "echos" (bounce back from that far side of the hose) will be closer to the original hit and will be filtered out more easily by the software. When there is a lot of hose on the far side, the "echos" that make it back will be further from the legitimate hit, so it can be more difficult for the software to tell (or it takes higher dwells to filter them).

Here is some detailed information on dwells that you can use to decide when to adjust your dwells and what may be affected by them...

What we call "dwell" or "dwell time" is the amount of time that additional "hits" (whether legitimate or "noise") on the same channel (hose) will be ignored after each hit on that channel (hose). All hose-based counters have this in one way or another (there are many terms for it, I think another manufacturer calls it "deadband" or "dead zone"). This is used to get rid of noise or echo on the hoses... after the hose is struck by a tire air pulses in the hose are bouncing around for a short period afterwards, the dwell will help avoid re-triggering.

When adjusting the dwell what we are trying to avoid is a dwell so high that it will cause legitimate hits to get missed. Since adjusting the dwells needs to take speed into account, we will figure out how fast a car travels per foot, here are some quick values (just so you can see how we arrive at the dwell numbers later):
At 20MPH a vehicle travels at about 34ms per foot.
At 40MPH a vehicle travels at about 17ms per foot.
At 80MPH a vehicle travels at about 8.5ms per foot.

Typically the main thing to worry about is dual axles on heavy trucks, since heavy trucks are typically an important part of a classification study. We want to affect those at little as possible even though a truck with a missed dual axle will frequently still end up classifying as a heavy truck of some sort. Those axles can be down around 4 feet apart, so typically that will be what to avoid when adjusting the dwell. If trucks are going excessively fast, then echo or loss of dual axles may be unavoidable.

A dwell of 55ms, which is our default, works pretty well up to about 50MPH. If you are having noise issues you will want to increase this value to improve the quality of your data. And for any normal circumstances you should always set both A and B dwells to the same values. Here are some values for dwells that I would pick if I suspected any sort of noise (or perhaps even if I didn't)...

If we calculate some dwells (rounding down) so that we don't miss a 4 foot axle spacing, our dwell settings would be:
20MPH - 136ms dwell
30MPH - 90ms dwell
40MPH - 68ms dwell
50MPH - 54ms dwell
60MPH - 45ms dwell
70MPH - 38ms dwell

If you are not trying to classify vehicles, then you don't need to worry as much about setting the dwells higher. In very low speed applications (parking lots, drive-thrus, etc) the dwells should be set extremely high (several seconds), just attempting to get a single hit for each vehicle can work best in those situations (you cannot get classifications out of super slow vehicles). Slower vehicles can tend to cause more echo than faster vehicles (since the tire is on the hose longer), below 10MPH or so things can get very noisy.

Related Articles
Setting up for low speeds
Setting Your Counter Up To Collect Data