A measurement-consistent geometric error stitching method for full-volume accuracy enhancement of CMMs
Year: 2026
Type: article
Abstract: In practical coordinate measuring machine (CMM) applications, factors such as structural deformation and load variation can induce significant spatial variations in geometric errors across the workspace. These position-dependent errors undermine measurement consistency and challenge conventional compensation approaches, which typically rely on error data collected only from the central region of the working volume. As a result, measurement accuracy for large or off-center workpieces is often compromised. To overcome these limitations, this study proposes a measurement-consistent geometric error stitching method based on point cloud registration, which unifies multi-region geometric error data obtained under different measurement coordinate systems. First, rigid and non-rigid volumetric error models of the CMM are established using a homogeneous transformation framework to describe geometric error propagation along the kinematic chain. A Truncated least squares Estimation And SEmidefinite Relaxation (TEASER)-Convolutional Siamese Point Net (CSPN) model is established to implement a coarse-to-fine registration approach. This rigid registration framework performs multi-scale geometric feature extraction and attention-based fusion to achieve high-precision spatial alignment of error point clouds. A comprehensive error measurement experiment is conducted to divide the workspace into distinct regions for error registration. Experiments show that the proposed method improves full-volume compensation accuracy by 38.83 % compared with conventional rigid and non-rigid methods.Engineering validation via workpiece feature evaluation further demonstrates a 52.87 % improvement in measurement accuracy. These findings indicate that the proposed method effectively constructs a unified error model covering the entire workspace, offering reliable technical support for high-precision CMM measurement applications.
Language: English
Subfield: Mechanical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7234738083
