Spatial Stiffness Programming in Single‐Filament Material Extrusion Through Data‐Driven Thermal Foaming

Material‐extrusion printing of variable‐stiffness structures typically relies on multi‐material architectures, which often suffer from poor interfacial bonding, increased fabrication time, material waste, and complex tool‐change operations. Here, we develop a data‐driven manufacturing framework that exploits the established thermal‐foaming response of a commercial TPU to program stiffness both within and across printed layers. By locally varying extrusion temperature through pre‐programmed commands, the expansion of the physical blowing agent is modulated, enabling programmable porosity from 15.7% to 40.7% and corresponding hardness modulation from 90 to 63 A using a single‐filament system. The mechanical response shows a substantial reduction in secant modulus at 200% strain from 7.75 to 4.70 MPa with increasing foaming level. SEM and micro‐CT analyses reveal a transition from a layer‐confined anisotropic morphology to an interconnected foam‐dominated architecture governing stiffness evolution. To achieve spatially resolved property control, a machine‐learning‐assisted G‐code framework is developed based on weighted k‐nearest neighbors for automated local modulation of nozzle temperature and extrusion flow. Compared with literature data for comparable multi‐material TPU interfaces, the single‐feedstock approach provides up to 4.6‐fold higher interfacial strength. It is validated through bio‐inspired structures and soft‐robotic actuators capable of distinct programmed motions from identical geometries, establishing a route toward property‐driven additive manufacturing.