Model of rough surfaces with Gaussian processes

TL;DR

This paper introduces a novel Gaussian process-based method for simulating and modeling rough surfaces, enabling more flexible and accurate representations of complex surface textures for fluid dynamics and contact mechanics studies.

Contribution

It presents a new approach using Gaussian processes to simulate and fit rough surfaces, overcoming limitations of traditional deterministic or structured methods.

Findings

GP-based simulations can model a wider range of surface textures.

The method effectively fits measurement data of real rough surfaces.

Examples include simulated ground and honed surfaces.

Abstract

Surface roughness plays a critical role and has effects in, e.g. fluid dynamics or contact mechanics. For example, to evaluate fluid behavior at different roughness properties, real-world or numerical experiments are performed. Numerical simulations of rough surfaces can speed up these studies because they can help collect more relevant information. However, it is hard to simulate rough surfaces with deterministic or structured components in current methods. In this work, we present a novel approach to simulate rough surfaces with a Gaussian process (GP) and a noise model because GPs can model structured and periodic elements. GPs generalize traditional methods and are not restricted to stationarity so they can simulate a wider range of rough surfaces. In this paper, we summarize the theoretical similarities of GPs with auto-regressive moving-average processes and introduce a linear process view of GPs. We also show examples of ground and honed surfaces simulated by a predefined model. The proposed method can also be used to fit a model to measurement data of a rough surface. In particular, we demonstrate this to model turned profiles and surfaces that are inherently periodic.