Surface Profile Characterization by Autoregressive-Moving Average Models

Abstract

The surface texture of a machined part is in general composed of three topographical components: waviness, roughness, and errors of form. A new technique for surface profile characterization is introduced which employs parametric stochastic models of the autoregressive-moving average (ARMA) class. The method for obtaining these models for surface profiles is shown by an example. The ARMA modeling technique for profile description is evaluated in three parts to determine its validity, workability, and descriptive power. This analysis is developed through the criteria of ergodicity, sensitivity, and describability. The ergodicity criterion tests the ability of models for physically identical profiles to convey equivalent information. The sensitivity criterion measures the level of detection of topographical differences among profiles by the ARMA model parameters. The descriptive ability of the models is examined by interpreting their parameters in light of the physical components of the profile. To implement this evaluation, ARMA models for eight different milled surfaces are determined and used.