6. Discussion and conclusion
This paper reconsiders the drawbacks of the existing T-spline models, and proposes a set of new T-spline data models to obtain better data storing and operating efficiencies. The conventional T-mesh is decomposed into the parametric, topological, and Cartesian layers, respectively. The characteristics of the redesigned T-spline models are analyzed and reviewed. Using the proposed T-spline data structures, the STEP-compliant data models are presented for the standardized data exchange between different CAx systems in the future. A prototype system which consists of a TSM-to-STEP convertor, a STEP parser and a T-spline kernel, is proposed to implement the data models and realize the STEP-compliant data exchange. In order to verify the feasibility and efficiency, some T-spline examples have been tested and presented. The testing results prove that the new T-spline data models are much more friendly to both human and computer, hence have a great prospect for the future development of Tspline technologies. Although recent data models work well for the regular T-spline surfaces, other more generalized models have to be considered in the future. For example, unstructured T-splines with extraordinary vertices should be considered [28], T-NURCCS models will be supported for a manifold T-spline surface [1,29], volumetric T-spline will be defined for models with higher dimensions [30,31], and the research and development of newmodel based algorithms should be investigated and discussed.