AiVibeGeometric Dimensioning and Tolerancing (GD&T)
GD&T is a symbolic language on engineering drawings that defines the allowed variation in form, orientation, location, profile and runout of part features relative to datums, defined by ASME Y14.5 and the ISO GPS standards.
GD&T uses feature control frames that state a geometric characteristic, a tolerance zone, optional material condition modifiers and datum references. Characteristics include form controls such as flatness, straightness, circularity and cylindricity, orientation controls such as perpendicularity, parallelism and angularity, location controls such as position, profile of a line or surface, and circular and total runout. Datums establish the reference frame from which features are toleranced.
GD&T expresses design intent more precisely than plus-or-minus coordinate tolerancing. It ties tolerances to function and assembly, can permit larger tolerance zones where function allows, and gives manufacturing and inspection an unambiguous reference for fixturing and measurement. It is widely used in automotive, aerospace, medical and precision engineering.
Two main standards systems exist: ASME Y14.5 and the ISO geometrical product specification (GPS) standards, including ISO 1101 for geometrical tolerancing and ISO 5459 for datums. They differ in some default principles; for example, ASME applies the envelope principle by default while ISO 8015 sets the independency principle, so the governing standard must be stated on the drawing. Verification usually involves coordinate measuring machines and trained interpretation.
Key points
- Controls form, orientation, location, profile and runout relative to datums
- Feature control frames state characteristic, zone, modifiers and datums
- Defined by ASME Y14.5 and ISO GPS standards such as ISO 1101
- ASME and ISO differ in defaults, so the drawing must name the standard
Where AiVibe comes in
AiVibe designs and manufactures the AiAmbA AI Factory: edge devices plus AI agents that let people talk to CNC, PLC and robot controllers in plain language. Agents only propose changes, and a trained operator confirms each one.