tailieunhanh - Gear Noise and Vibration P02

can be done on a test rig out in the main works or sometimes even on the equipment while running normally. However, the basic idea is that pitch, profile and helix errors may combine with tooth bending, gear body distortions and whole gear body deflections to give an overall relative deflection (from smooth running) at the meshpoint between the gears. It is also difficult to convince gear engineers that there is a very big difference between roll (double flank) checking, which is extremely cheap and easy, and . (single flank) checking since they give rather similar looking results. Unfortunately, there. | 10 Chapter 1 can be done on a test rig out in the main works or sometimes even on the equipment while running normally. However the basic idea is that pitch profile and helix errors may combine with tooth bending gear body distortions and whole gear body deflections to give an overall relative deflection from smooth running at the meshpoint between the gears. It is also difficult to convince gear engineers that there is a very big difference between roll double flank checking which is extremely cheap and easy and . single flank checking since they give rather similar looking results. Unfortunately there are a large number of important gear errors which are missed completely by roll checking so this method should be discouraged except for routine control of backlash. The problems with double flank measurement arise from the basic averaging effect that occurs. Any production process or axis error in transfer from machine to machine may produce errors which give ve errors on one flank which effectively cancel -ve errors on the feeing flank. The resulting centre distance variation is negligible but there may be large cancelling errors on the drive and overrun flanks. Shavers and certain types of gear grinders are prone to this type of fault which is worse with high helix angle gears. The question then arises as to the connection between . and final noise. Few practising engineers initially believe the academics claim that noise is proportional to . although the system normally behaves except under light load as a linear system. For any linear system the output should be proportional to input. Doubling the . should give 6dB increase in noise level or with a target reduction of lOdB on noise the . should be reduced by 710 . roughly 3. This only applies at a single frequency and different frequencies encounter high or low responses en route so a major visible frequency component in the . may be minor in the final noise because it could not find a .