Data-Driven Material Modeling

Data-Driven Material Modeling video

Computationally-tuned material fabrication

Research team (Lazarus, Vespers I & II): Christoph Bader, Dominik Kolb, James C. Weaver. Prof. Neri Oxman

Research team (Vespers III): Rachel Soo Hoo Smith, Christoph Bader, Sunanda Sharma, Dominik Kolb, Tzu-Chieh Tang, Ahmed Hosny, Felix Moser, James C. Weaver, Christopher A. Voigt. Prof. Neri Oxman

Research team (Totems): Sunanda Sharma, Christoph Bader, Rachel Soo Hoo Smith, Felix Kraemer, João Costa, Joseph H. Kennedy, Jr. Undergraduate researchers: Joseph Faraguna, Sangita Vasikaran, Sara Laura Wilson. Prof. Neri Oxman

Year: 2015 - present

Projects: Lazarus, Vespers I, Vespers II, Vespers III, Totems


Current advancements in additive manufacturing enable the fabrication of geometrically complex and materially heterogeneous objects with high spatial resolution in manufacturing. Such advancements challenge designers, architects and engineers alike to move beyond shells designed with pre-determined shape, and material composition; and to consider an expanded design space encompassing internal material properties. Color and opacity, stiffness, softness, shape memory, swellability, expansion, wettability and refractive index can be seamlessly tuned, fabricated and leveraged in design applications.

We propose that the “anatomy” of objects can be designed through generative methods. However, current off-the-shelf software tools do not typically take these recent advancements into consideration, thereby missing out on significant design opportunities that lie at the intersection of digital modeling, analysis, and fabrication.

Sphere Aluminum 1 2 D
DDMM sphere on aluminum block
Sphere Aluminum 4 1 D
DDMM sphere on aluminum block
Sphere Aluminum 2 2 D
DDMM sphere on aluminum block

DDMM (Data-Driven Material Modeling) combines generative modelling methods with high-resolution multi-material 3D printing. This generative approach is controlled using existing or dynamic data-sets to drive the simultaneous generation of geometry and material information. For the generation of continuous material distributions, the generative methods are evaluated during slice generation, thereby enabling the production of voxel-matrices describing material property variations at the 3D printer’s native resolution.

However, to leverage advantages of synthetic hybrid material systems, designers and engineers alike have to think beyond geometry-driven CAD tools and consider implementing hybrid design workflows. Combining parametric geometric modeling, parametric volumetric modeling, high resolution material dithering, common design evaluation methods and multi-material 3D printing we propose an integrated design workflow and explore associated design space of 3D printable functionally graded material systems.

DDMM explores this hybrid approach in a five-part framework.

Volume Hand
Transparent hand showing bones
Image Based Elevation
Artificially-colored elevation
Point Cloud Monkey
Point cloud data of monkey
Image Based Airway
Imaged-based airway
01 E
Sectional view of neural links in a brain
Image Based Brainbow
Closeup of neural links
Volume Mixing
Simulated turbulent flow
Tetrahedral Volume
Tetrahedral volume
Point Cloud Moon
Point cloud of the moon


First, data-driven generative geometric and material modelling methods are implemented to model geometry in conjunction with material distributions.

Data-driven generative geometric modelling refers to the ability to describe the generation of a geometric object through a set of generators and operators, and the control of this generation through data-sets. This is advantageous because a change in data results in re-evaluation of the generation description and a new geometric object according to its specification. Similarly, data-driven generative material modelling is the generation of a volumetric region and associated property fields in space through data-sets that describe material property variations.

The volumetric descriptions can be given in form of a voxel representation describing material distributions as a set of material vectors of mixing ratios and additional data organized in a spatial data structure. This material vector, associated additional data and the property-to-material lookup tables (described below) are then used to compute the final multi-material droplet deposition instruction at the resolution of the printer. Parametric geometric and parametric volumetric modelling methods are combined by using geometric properties as parameters in the volumetric modelling workflow and vice versa.

Mapping of data to opacity to material deposition
Workflow for modeling and rendering of high resolution data
Demonstration of data-driven evaluation of geometry
A wide range of variations in geometry can be quickly generated
Sphere 1 3
Topography and flow lines in 3D printed orb

In computational fabrication, material distributions are mapped to experimentally characterized material properties such as opacity, stiffness, or others. This mapping is enabled through two processes. First, the characterization of material mixing ratios through which associated property-to-material map can be generated. Secondly, the continuous material tunability through high resolution material dithering which allows to use the property-to-material maps to translate a desired material behavior to volumetric descriptions of material mixing ratios. The material tunability and the volumetric material description allow to perform design evaluation of material behaviors by either rendering or simulation. This can be in the form of the simulation of the non-linear dynamics of an actuation system or the evaluation of optical properties through volumetric rendering.

Significantly, geometry and material representations are kept distinct and are only combined if necessary for example in the simulation or slicing process to achieve scalability. Geometry, volumetric material descriptions, and property-to-material maps are then processed through a slicing process to generate the above described layer-based instructions for the multi-material fabrication system. This slicing process uses the given data-sets to perform additional computational processes during slice time to achieve high-resolution composite generation. All of these contributing towards a holistic design approach for programmable materiality.

Demonstration of materials produced by DDMM
A polished hemisphere showing simulated turbulent flow


With the DDMM framework, a wide variety of hybrid systems found in nature can be directly fabricated to nature-mimetic entities. By utilizing methods described here, we show that the barrier between the digital description of material behaviors and their physical reality can be obviated more easily and can open a wider space for design.

Enabled through computational design strategies such as parametric geometric, parametric volumetric modeling and design evaluation in combination with high-resolution slicing and multi-material 3D printing, we facilitate an integrated workflow providing means to work with geometric and material structures efficiently and creatively. Such design workflows, will in the future, enable designers to not only precisely mimic systems and mechanics found throughout nature, but also give rise to new mediating systems between synthetic and biological entities.

Sphere 2 1 D
Simulated turbulent flow in 3D printed orb
Sphere 4 1 D
Simulated turbulent flow in 3D printed orb
Sphere 1 4 D
Topography and flow lines in 3D printed orb
Sphere 3 2 D
Neurovascular structures in 3D printed orb

Privacy Policy

We use cookies to ensure the basic functioning of the site, and to analyze its traffic (details).