Research on the Construction of Standardized Test Datasets Based on Characterization Parameters

The fitting of geometric elements is a critical step in evaluating errors in measurement evaluation software. Different approximation methods, evaluation strategies, and the rounding of significant figures can all influence the evaluation results, leading to inconsistent outcomes from different measurement analysis software for the same measurement data. To address the difficulty in certifying the evaluation algorithms for geometric element fitting, this paper analyzes and discusses the basis and rules for generating standardized input test datasets. Characterization parameters for different geometric elements are determined, enabling the dynamic construction of standardized input test datasets based on these parameters. Using the standardized input test datasets, the least squares double fitting algorithm and geometric tolerance evaluation algorithm are studied to generate standardized output test datasets. The evaluation results are then compared and verified with Zeiss' CALYPSO measurement analysis software. By comparing the dynamically constructed standardized output test datasets with the evaluation results from the measurement evaluation software, the certification of the least squares fitting-based evaluation algorithms for different geometric elements in the measurement evaluation software is completed.