EN TR
Generative Facades Research
Generative Systems04 - BUILDINGSimulation Study2023-2025

Generative Facades

Every panel optimized for its exact position. 38-45% less cooling load in our simulation scenarios.

Scale 04 Building
Panels 3.5k+ optimized
Reduction 38-45% modeled
Tools GH Wallacei
Type Simulation Study
Updated 2026-08
01

In our cumulative Radiance runs, a west-facing glass facade on our Istanbul scenario massing collects about 1,420 kilowatt-hours per square meter a year. That's enough energy to cook an office.

Traditional solutions make you choose. Fixed louvers block sun but kill the view. Tinted glass reduces glare but darkens the interior. You can have comfort or daylight, not both.

The idea we test here is older than the tools. In 1981 Mike Davies sketched the polyvalent wall: a facade that filters heat, light, air, and view independently. Frei Otto found structural forms by letting physics vote. What has changed is that an evolutionary solver can now run that vote thousands of times per facade. So we asked: what if every panel was optimized for its exact position? South-facing panels deep, north-facing panels shallow, west-facing panels aggressive, east-facing panels gentle.

Method, in one line: Radiance maps a year of sun onto the massing; six parameters span 2.3 million possible panel configurations; Wallacei's evolutionary solver breeds 50 generations of 100 candidates against two objectives, cooling load and daylight; the Pareto front is where the trade-off lives; clustering compresses 3,472 winning panels into 14 buildable families.

The result, in our full-year simulations: 38-45% reduction in cooling load while keeping daylight above 50%. That's a combination fixed uniform shading cannot achieve. The numbers on this page come from those simulation runs; none of these facades has been built.

Generative facade optimization across building surface

Position-Specific: Each panel responds to its unique solar exposure. The facade is a map of the sun's path.

02

Theoretical Framework

01

Solar Performance

38-45% heat gain reduction compared to clear glazing in the model. Each panel does its exact job.

02

Daylight Balance

Maintaining 50%+ daylight availability even with aggressive shading. Comfort and light coexist.

03

Panel Rationalization

3,472 unique panels clustered into 14 fabrication families. Custom performance, manageable production.

04

Fabrication Logic

Clustering cuts estimated CNC toolpath time by roughly 60% versus the unclustered approach. Fabrication itself is the untested step.

03

Research Process

01

Map Solar Exposure

8,760-hour simulation in Radiance, cumulative kWh/m² per zone

02

Define Parameters

6 variables, 2.3M configurations in the design space

03

Evolve Solutions

50 generations in Wallacei, Pareto front yields 15-20 optimal options

04

Rationalize

Cluster 3,472 panels into 14 families for efficient CNC fabrication

04

Research Phases

01

Environmental Mapping

8,760 hours of solar simulation using Radiance. Cumulative radiation mapped per facade zone.

02

Design Space

6 parametric variables generating 2.3 million valid panel configurations. The space is vast.

03

Evolution

50 generations, 100 panels per generation. Wallacei breeds toward Pareto-optimal solutions.

04

Fabrication

Clustering algorithm groups similar panels. CNC toolpath generation for robotic fabrication.

Pipeline diagram of the generative facade optimization loop

The Loop: One facade generation cycle: map a year of sun, span the design space, breed against two objectives, keep the Pareto front, cluster for fabrication. Numbers from our Ankara scenario run.

05

Key Metrics

8,760
Hours Simulated
Full year, every hour
2.3M
Configurations
Design space size
3,472
Unique Panels
Before clustering
14
Families
After clustering
06

Key Thinkers

01

Mike Davies

Architect, Polyvalent Wall Concept

In 1981, Davies imagined a wall that could filter heat, light, air, and view independently. Per-panel optimization is a fixed-geometry step toward it, forty years later.

02

Klaus Daniels

Building Systems Engineer

Daniels wrote the book on climate-responsive building technology. Our simulation methods extend his performance modeling.

03

Frei Otto

German Architect, 1925-2015

Otto minimized material through form, letting physics vote on shape. We minimize energy through geometry. Same philosophy, different medium.

04

Jan Knippers

ITKE Stuttgart

Knippers leads biomimetic facade research. Our panel articulation draws from his work on adaptive structures.

07

Case Studies

Scenario: Office Tower, Ankara

Design scenario, not built

West-facing curtain wall with optimized aluminum louvers. Our largest simulation scenario, run on the same massing as our Data-Driven Tower concept study. Modeled cooling reduction: 38%.

3,472 Panels
38% Modeled Reduction

Scenario: Mediterranean Pavilion, Izmir

Design scenario, not built

ETFE cushion concept with variable pneumatic fill. A testbed scenario for lightweight materials. Modeled heat reduction: 42%.

612 Panels
42% Modeled Reduction

Scenario: Heritage Retrofit, Athens

Design scenario, not built

Heritage-sensitive secondary skin over an existing stone facade. The hardest constraint set we've modeled: performance without touching the historic exterior. Modeled cooling reduction: 45%.

1,890 Panels
45% Modeled Reduction

Comparative Analysis

Illustrative mockup model: uniform louvers at one fixed angle

Fixed Louvers

One Angle Fits All

Static aluminum blades at fixed angle. Works on one orientation, fails on others.

StaticLow CostLimited
Illustrative mockup model: folding kinetic panels with hinges and actuators

Kinetic Facades

Moving Parts

Panels that track the sun. The Al Bahr Towers report roughly 50 percent solar gain reduction from their kinetic screen: high performance, real maintenance cost.

DynamicExpensiveComplex
Illustrative mockup model: smooth decorative gradient of panels

Parametric Panels

Smooth Variation

Gradual geometric change across the surface. Visually striking, not necessarily optimized.

AestheticSmoothDesign-Led
Illustrative mockup model: individually optimized panels in clustered families

Our Approach

Per-Panel Optimization

Each panel evolved for its specific conditions, no moving parts. Best performance-to-cost ratio in our simulation runs.

OptimizedData-DrivenFabrication-Aware
05

Optimization Results

100% 75% 50% 25% 0%
42%
35%
28%
22%
0%
Generative (Ours)
Reflective Glass
Fixed Louvers
Tinted Glass
Clear Glass

Percentage reduction compared to clear glazing

Scenario model: values from our own simulation runs

08

Key Findings

01

Self-shading cuts cooling 38-45% in Mediterranean climates. All three of our simulation scenarios (Ankara, Izmir, Athens) agree.

38-45% in the model
02

Optimal panel depth varies from 120mm (north) to 480mm (west) on a single building. Uniform shading wastes material.

4× depth variation
03

Voronoi perforation achieves 23% higher daylight uniformity than regular grids at the same openness ratio, in our comparison runs.

+23% uniformity
04

Topology optimization of panel geometry reduces aluminum usage by 23% in the model without compromising computed strength.

23% material saved
09

Honest Limitations

Behavioral Assumption

Variable angles collect dust unevenly. Maintenance is harder than uniform systems.

Data Dependency

Glare risk. Some configurations redirect light into eyes rather than blocking it.

Data Dependency

Cost premium. 12-18% higher fabrication cost versus standard curtain wall, by our estimate.

Temporal Limitation

Winter penalty. Aggressive shading reduces beneficial solar gain in cold climates.

Computational Cost

Everything here is simulation. No facade from this study has been fabricated at full scale, and simulated performance always loses something on site.

11

Conclusion

Every panel can be different and still be rationalized for fabrication. With per-position optimization, our simulations show 38-45% cooling reduction while maintaining daylight above 50%; a performance level fixed uniform shading cannot reach. The method works in the model. Building a bay of it is the honest next test.

Limitations

  • Maintenance complexity
  • 12-18% cost premium

Future Directions

  • Fabricate and instrument a full-scale bay
  • Glare analysis per configuration
  • Winter-balance objectives for cold climates