This study proposes an intelligent automotive roof frame design method based on the middle layer and component technology on CATIA. It aims to solve core roof modeling issues: determining geometric input quantity but uncertain attributes (tangent vectors, normal vectors, number of curve segments, number of surface patches, and boundaries), high manual interaction dependence, and poor knowledge reuse, to realize efficient design knowledge reuse. Methodologically, it builds a feature-driven parametric template, develops a knowledge rule-embedded componentized UDF library (reducing repeated modeling and geometric reconstruction needs), and integrates knowledge engineering for geometric input verification and operation direction control, eliminating curve/surface attribute uncertainty impacts. Verification shows the template stably generates roof crossbeams under simple/complex inputs, improving model robustness and reuse rate, reducing design workload, shortening verification cycles, and providing an extensible solution for white body design.