A two phase circular regression algorithm for quantifying wear in CV joint ball race tracks

Abstract

In this paper, a two-phase circular regression algorithm is presented for extracting wear profiles from Rzeppa-type constant velocity (CV) joints and for quantifying race track wear. In ball races operating under harsh cyclic loading conditions the predominant brinelling and “false brinelling” wear mechanism result in small indentations or grooves in the race track. These are particularly difficult to measure as the reference or nominal surface is curved in two planes in three-dimensional space. The presented algorithm achieves absolute measures of ball race track wear by a two-phase surface fitting regression methodology. The wear grooves are identified in the first phase and extracted to obtain sections of “unworn” surface; in the second regression phase, these sections are used to construct the datum surface and to establish absolute measures of wear groove depth. An important property of this algorithm is that it does not require the user to input parameters specific to the particular type or size of joint. This is particularly important in automotive aftermarket applications such as in drive-line servicing and in CV joint rebuilding/recycling. Results presented for a range of different CV joints demonstrate that race profiles can be obtained quickly and efficiently and that wear groove depth can be measured accurately to a ±3σ repeatability of 2.4 μm (±0.000 093 in).