Repeatability of on-machine probing by a five-axis machine tool

Abstract

Assessing the repeatability of on-machine probing is required to estimate the measurement uncertainty. Currently, standards such as ISO 230–2:2014 and ISO 230–6:2002 include approaches only for a single-axis or a diagonal-path evaluation of the machine repeatability. Also, ISO 230–10:2016 is limited to the evaluation of the probing repeatability at a specific position of the machine for single-point surface measurement as well as for measuring the circle and sphere centre location. In this paper, a general model capable of predicting the measurement repeatability, considered as a covariance matrix, for a five-axis machine in any probing situation is sought from replicate measurements on a spherical artifact at various positions and with different probing directions. Polynomial functions fitted to the X, Y and Z recorded coordinates as well as to the normal (radial) and tangential projection values respectively perpendicular and tangent to the target surface estimate the standard deviations, i.e. the square roots of the diagonal elements of the covariance matrix, of on-machine probing in terms of experimental factors. Separate polynomials were fitted to the covariance quantities obtained between the recorded components as well as the normal and the tangential projections to estimate the off-diagonal elements of the covariance matrix. The factors include the positions of all five machine axes at the target point as well as the measurement direction. The repeatability models are validated by comparing measured and predicted values for a separate validation data set, which shows that a quantitative prediction of the randomness of on-machine measurement is achievable. The potential effects of the probe orientation, the hysteretic behavior of the measurement system, the measurement strategy and the time span on the on-machine probing are also studied through different experiments on a ring gauge. The effect of hysteresis on probing repeatability was found negligible but biases were detected.