From computational-design-skills
Provides strategies for panelization, surface rationalization, attractor patterning, double-skin, kinetic, responsive, environmental, and fabrication-aware facade design in AEC.
npx claudepluginhub amanbh997/claude-skills-for-computational-designersThis skill uses the workspace's default tool permissions.
The building facade is not a wrapper. It is the single most consequential architectural element — the mediator between interior environment and exterior climate, the primary determinant of energy consumption, the structural skin that must resist wind, seismic, and thermal loads, and the public expression of a building's identity. Computational facade design treats every square meter of this sur...
Guides building envelope design for walls (masonry, timber, steel, CLT, curtain wall), glazing (single-triple, low-e, gas fills), roofing (flat, green), thermal metrics (U/R-value, SHGC), moisture management, air barriers, and climate targets.
Provides 7 Python CLI calculators for AEC computational design: geometry analysis, structural checks, solar radiation, panel optimization, mesh analysis, material estimation, fabrication costs.
Applies climate-specific urban design strategies for hot-arid, tropical, temperate, and cold climates. Covers building orientation, shading, wind management, vegetation, heat island mitigation, stormwater, and thermal comfort.
Share bugs, ideas, or general feedback.
The building facade is not a wrapper. It is the single most consequential architectural element — the mediator between interior environment and exterior climate, the primary determinant of energy consumption, the structural skin that must resist wind, seismic, and thermal loads, and the public expression of a building's identity. Computational facade design treats every square meter of this surface as a field of optimizable variables rather than a repeating module selected from a catalog.
Every facade simultaneously manages five environmental flows:
| Flow | Inward | Outward |
|---|---|---|
| Solar radiation | Daylight, solar heat gain | Glare, overheating |
| Thermal energy | Heat loss in winter | Heat gain in summer |
| Air | Natural ventilation, infiltration | Exfiltration, stack effect |
| Moisture | Rain penetration, condensation | Vapor diffusion |
| Sound | Exterior noise intrusion | Interior noise escape |
A computationally-driven facade optimizes across all five flows simultaneously, varying panel geometry, material, porosity, and depth point-by-point across the surface based on orientation, local microclimate, interior program, and structural constraints.
Traditional practice separates these into different consultancies — the architect draws the pattern, the structural engineer sizes the mullions, the facade consultant specifies the glass, and the energy modeler checks compliance. Computational facade design collapses these into a single parametric model where every design decision is simultaneously evaluated against structural, thermal, daylight, acoustic, and aesthetic criteria. The parametric model is the single source of truth, and downstream deliverables — shop drawings, energy models, structural calculations, panel schedules — are all derived outputs.
Standard curtain wall systems (stick-built or unitized) impose a regular orthogonal grid with fixed mullion depths and standardized infill panels. This approach optimizes for fabrication simplicity and erection speed at the cost of environmental performance — every panel on the building receives the same glass, the same SHGC, the same U-value, regardless of whether it faces north or south, is at ground level or the 60th floor, or fronts an office or a server room.
Computational facade design replaces this uniformity with gradient variation:
The result is a facade that performs 30-50% better than a uniform curtain wall while often using less material, because material is concentrated where loads demand it rather than uniformly distributed.
Modern computational facades are designed from data, not from intuition:
Each of these data layers becomes an input field that the computational model reads and responds to, producing a facade that is locally optimized everywhere.
Panelization is the process of decomposing a continuous design surface into discrete, fabricable panels. The choice of panelization strategy determines fabrication cost, structural behavior, weatherproofing strategy, and visual character.
Planar panels are the most economical to fabricate because flat glass can be cut from stock sheets without any forming process.
The simplest approach: subdivide the surface along its natural UV parameter lines to produce quadrilateral panels. On a planar or single-curved surface, these quads are inherently planar. On a double-curved surface, they will deviate from planarity.
Planarity tolerance: Industry standard is < 2mm deviation of the fourth corner from the plane defined by the other three corners. For structural silicone glazing, tolerances tighten to < 1mm. For mechanically captured glazing with gaskets, up to 3mm may be acceptable depending on gasket profile.
PQ-mesh (Planar Quadrilateral Mesh) generation methods:
Triangulation guarantees planarity (any three points define a plane) but produces more edges, more mullion intersections, and higher framing cost. Triangulated facades typically cost 15-25% more in framing than quad facades of equivalent area due to the increased total edge length and the complexity of six-way mullion intersections.
Triangulation methods:
Hex panels produce three-way intersections (120-degree angles) which are structurally efficient and visually distinctive. However, hexagonal panels on a double-curved surface cannot all be planar — the Euler characteristic of the sphere requires exactly 12 pentagonal panels in any hexagonal tiling of a closed convex surface (Euler's formula: V - E + F = 2).
Generation methods:
Voronoi tessellations, irregular polygonal meshes, and other non-regular planar decompositions. These maximize design freedom but complicate fabrication scheduling and erection sequencing. Every panel is unique, so panel identification and tracking become critical — each panel requires a unique ID, a fabrication drawing, and a location tag.
Single-curved panels (cylindrical, conical, or general ruled surfaces) can be fabricated by bending flat material along one axis.
A ruled surface is generated by moving a straight line through space. If the facade surface can be decomposed into strips where each strip is a ruled surface, the panels can be fabricated from flat sheet material bent around a single-curved mold.
Ruling analysis: For a given surface, compute the asymptotic directions (directions of zero normal curvature). Along these directions, the surface is locally ruled. On a surface with negative Gaussian curvature, there are two asymptotic directions at every point; on a surface with zero Gaussian curvature (developable), there is one; on a surface with positive Gaussian curvature, there are none (the surface is locally non-ruled).
A developable surface has zero Gaussian curvature everywhere — it can be unrolled flat without stretching. Decomposing a free-form surface into developable strips is a powerful rationalization strategy because each strip can be fabricated from flat sheet material with zero waste from forming.
Strip decomposition algorithms:
Cold bending involves forcing a flat glass panel into a curved frame, inducing residual stress in the glass. This is the most economical way to achieve single-curved panels.
Minimum bend radius by glass thickness and type:
| Glass Thickness | Annealed (min radius) | Heat-Strengthened | Fully Tempered |
|---|---|---|---|
| 4 mm | 2.0 m | 1.5 m | 1.0 m |
| 6 mm | 3.0 m | 2.2 m | 1.5 m |
| 8 mm | 4.0 m | 3.0 m | 2.0 m |
| 10 mm | 5.0 m | 3.8 m | 2.5 m |
| 12 mm | 6.0 m | 4.5 m | 3.0 m |
| 15 mm | 7.5 m | 5.6 m | 3.8 m |
| 19 mm | 9.5 m | 7.1 m | 4.8 m |
Stress limits: Cold-bent annealed glass should not exceed 7 MPa residual bending stress under sustained load. Heat-strengthened glass can tolerate up to 24 MPa, and fully tempered up to 46 MPa. These limits must account for additional wind and thermal stresses during service.
Bending moment calculation: For a rectangular panel of width w, thickness t, and bend radius R, the bending stress sigma = E * t / (2 * R), where E = 70 GPa for soda-lime glass. This determines whether a given curvature is achievable with a given glass type and thickness.
Double-curved panels require forming processes that deform the material in two directions simultaneously.
Glass is heated to approximately 620-680 degrees C (above its softening point) and slumped or pressed over a mold. This allows complex curvatures but requires a unique mold for each panel geometry.
Process constraints:
For non-glass materials (GFRC, FRP, precast concrete), molds can be CNC-milled from foam, 3D-printed, or fabricated from sheet metal. Mold cost dominates when panel count per unique geometry is low.
Mold cost amortization: If a mold costs $2,000 and produces 1 panel, the mold cost per panel is $2,000. If it produces 20 identical panels, the cost drops to $100/panel. This is why panel clustering and repetition are so critical for double-curved facades.
| Panel Type | Relative Cost (per m²) | Typical Application |
|---|---|---|
| Flat (planar) | 1.0x (baseline) | Standard curtain wall |
| Cold-bent single-curved | 1.3-1.8x | Gentle curvature, towers |
| Hot-bent single-curved | 1.5-2.0x | Tighter curves |
| Cold-bent double-curved | 1.8-2.5x | Warped quads, minimal curvature |
| Hot-bent double-curved | 3.0-5.0x | Moderate double curvature |
| Free-form hot-bent | 5.0-8.0x | Complex sculptural forms |
| 3D printed mold + cast | 4.0-10.0x | Unique panels, small runs |
The goal of panel optimization is to minimize cost by reducing the number of unique panel geometries while maintaining design intent and surface quality.
Strategies:
K-means clustering: Represent each panel as a feature vector (e.g., four corner deviation from planarity, edge lengths, diagonal lengths, curvatures). Apply k-means to group panels into k families. Each family shares a single mold or cutting template.
DBSCAN clustering: Density-based clustering that does not require specifying k in advance. Panels that are geometrically similar within a tolerance epsilon are grouped together. Outliers (panels that do not fit any cluster) are flagged for individual fabrication.
Hierarchical clustering: Build a dendrogram of panel similarity. Cut at the desired tolerance level to produce families. Allows interactive exploration of the tradeoff between unique count and geometric deviation.
A panel family is a group of panels that share a common fabrication template. Within a family, panels may differ by:
| Metric | Definition | Target |
|---|---|---|
| Unique panel count | Number of distinct geometries | Minimize (< 20% of total ideal) |
| Total panel count | Total panels on facade | Determined by subdivision |
| Repetition ratio | Total / Unique | Maximize (> 5:1 ideal) |
| Waste ratio | Material wasted in cutting / total material | < 15% |
| Planarity deviation | Max corner deviation from plane (mm) | < 2mm for glass |
| Edge length variation | Std dev of edge lengths within a family | < 5% of mean |
| Curvature deviation | Max deviation from design surface (mm) | < 5mm typically |
Surface rationalization transforms a free-form design surface into a geometry that can be constructed from discrete elements with known fabrication processes. It is distinct from panelization (which subdivides a surface into panels) — rationalization modifies the surface itself to be more constructible.
Any smooth surface can be approximated by a collection of developable strips. The quality of approximation depends on strip width and surface curvature.
Ruling analysis: Compute the Gaussian curvature K at every point. Where K = 0, the surface is already developable. Where K is small (|K| < threshold), the surface can be closely approximated by a developable surface. Where |K| is large, the surface must be split into narrower strips.
Strip decomposition: Segment the surface into strips along one family of curvature lines. Each strip is then approximated by a ruled surface (the simplest developable form). The approximation error is proportional to strip width squared times the Gaussian curvature.
A conical mesh is a polyhedral mesh where, at every interior vertex, the face planes are tangent to a common cone. This geometric property has profound practical consequences:
Generation algorithms:
Reference: Helmut Pottmann, Andreas Wallner, et al., "Freeform surfaces from single curved panels," ACM Transactions on Graphics, 2008.
Dupin cyclides are surfaces where all lines of curvature are circles or straight lines. They include tori, cones, cylinders, and their inversions. A hex mesh derived from a Dupin cyclide decomposition of a surface produces planar hexagonal faces — a non-trivial geometric result since general hex meshes on curved surfaces are not planar.
Method:
An edge-offset mesh is a mesh where every edge has a well-defined offset direction perpendicular to the edge and lying in the bisector plane of the adjacent faces. This property allows beams of constant cross-section to be placed along edges without custom end cuts — the beam profile at each node fits perfectly with its neighbors.
This is critical for steel or aluminum structural facades where mullions and transoms are extruded profiles. Without the edge-offset property, every beam end requires a custom miter cut, dramatically increasing fabrication cost.
Replacing straight mullion segments with circular arcs allows a coarser mesh (fewer panels) to approximate a curved surface. Circular arcs can be fabricated by rolling straight profiles through a three-roll bender — a standard steel fabrication process.
Design parameters:
The principal curvature lines of a surface are curves along which normal curvature is maximized or minimized. They form an orthogonal network on the surface (except at umbilical points where principal curvatures are equal). This network has special properties:
Computation: Principal curvature lines are found by integrating the principal direction field across the surface. Singularities occur at umbilic points, where the direction field is undefined. Special handling (rounding, splitting) is required at these points.
| Tool | Platform | Capabilities | Limitations |
|---|---|---|---|
| Evolute Tools | Rhino/GH | Conical mesh, PQ mesh, edge offset mesh optimization | Commercial, no longer actively developed |
| Kangaroo 2 | Grasshopper | Planarization, developability, mesh relaxation | General purpose — requires custom goal setup |
| LunchBox | Grasshopper | Panel types (diamond, hex, quad, random) | Geometry generation only, no rationalization optimization |
| Paneling Tools | Rhino | UV-based panelization, attractor-based | Limited to surface UV structure |
| ShapeOp | C++/Python | Projective dynamics for geometric optimization | Research code, requires integration |
| Custom scripts | Python/C# | Full control over rationalization algorithms | Development time, no GUI |
| Karamba3D | Grasshopper | Structural analysis of facade meshes | Analysis only, not geometry generation |
Attractor-based patterning uses geometric primitives (points, curves, surfaces) as control inputs to modulate facade properties across the surface. This produces gradient effects that respond to environmental conditions, program, or purely aesthetic intent.
A point attractor P located at coordinates (px, py, pz) influences a panel centered at (cx, cy, cz) based on the distance d = |P - C|.
Distance-based scaling: Panel size S = S_base * f(d), where f is a mapping function:
Rotation: Panel rotation angle theta = theta_max * f(d). Useful for louver facades where louver angle varies with proximity to a design feature (entrance, corner, sightline).
Density variation: Subdivision density increases near the attractor (smaller panels near the point, larger panels far from it). Implemented by adaptive subdivision: refine quads where d < threshold, coarsen where d > threshold.
A curve attractor C(t) influences panels based on the minimum distance from the panel center to the curve. This produces band-like gradient effects along the facade.
Applications:
Implementation: For each panel center, compute the closest point on the curve using iterative projection (Newton-Raphson on the distance function) or by sampling the curve at fine intervals and finding the minimum.
When multiple attractors are active simultaneously, their effects must be combined:
| Method | Formula | Character |
|---|---|---|
| Weighted average | f = sum(w_i * f_i) / sum(w_i) | Smooth blending, values stay in range |
| Nearest | f = f_i where d_i is minimum | Sharp transitions at equidistant boundaries |
| Additive | f = sum(f_i), clamped | Reinforcement where attractors overlap |
| Multiplicative | f = product(f_i) | Rapid falloff, only activates near intersection |
| Maximum | f = max(f_i) | Each attractor dominates in its zone |
| Minimum | f = min(f_i) | Intersection-like behavior |
Use attractors to vary window-to-wall ratio (WWR) or glazing transparency across the facade:
Perforated metal screens can vary their perforation density, hole size, or hole shape based on attractor distance:
Typical open area ratios for facade screens: 20-60%. Below 20%, the screen reads as nearly solid. Above 60%, the screen loses its shading effectiveness and structural integrity.
Horizontal or vertical louvers can vary their tilt angle based on attractor distance:
Typical Grasshopper data flow for attractor-based patterning:
Surface → Subdivide (UV) → Panel Centers (points)
Attractor Point(s) → Distance (panel centers to attractors)
Distance → Remap (to 0-1 domain) → Scale/Rotate/Color panels
Panels → Geometry output
Panels → Data output (panel schedule)
Key components: Surface Divide, Distance, Remap Numbers, Graph Mapper (for custom falloff curves), Scale, Rotate, Extrude.
The most powerful application maps environmental simulation data directly to attractor fields:
A double-skin facade (DSF) consists of an outer skin, an inner skin, and a ventilated cavity between them. The cavity acts as a thermal buffer, an acoustic buffer, a natural ventilation path, and a space for integrating shading devices protected from wind and rain.
| Typology | Cavity Height | Cavity Depth | Ventilation | Best For |
|---|---|---|---|---|
| Box window | 1 story, 1 bay | 200-300 mm | Inlet/outlet per box | Renovation, noise reduction |
| Shaft-box | Multi-story shaft + box | 200-400 mm | Stack effect through shaft | High-rise, natural vent |
| Corridor | 1 story, full width | 400-800 mm | Horizontal flow per floor | Maintenance access, moderate height |
| Multi-story | 3+ stories | 600-1000+ mm | Stack-driven, full height | Landmark buildings, atria |
Cavity depth guidelines:
Ventilation strategies:
Airflow modes:
The DSF cavity enables natural ventilation in high-rise buildings where direct window opening would be impractical due to wind pressure. The outer skin acts as a wind buffer while the inner skin provides operable openings.
Design requirements for natural ventilation via DSF:
The DSF cavity provides significant acoustic attenuation, typically 10-15 dB improvement over a single skin of equivalent mass. This is due to the mass-air-mass resonance system formed by the two skins and the air cavity.
Acoustic design parameters:
The DSF cavity presents fire safety challenges:
DSF energy modeling requires coupled thermal-airflow simulation:
| Project | Location | Type | Cavity Depth | Key Innovation |
|---|---|---|---|---|
| 30 St Mary Axe (The Gherkin) | London | Multi-story (6 floors) | 1200 mm | Spiraling light wells with DSF ventilation |
| GSW Headquarters | Berlin | Corridor | 900 mm | West-facing thermal flue with automated blinds |
| KfW Westarkade | Frankfurt | Box window | 350 mm | Pressure-equalized box modules, 80% natural vent |
| One Angel Square | Manchester | Multi-story | 600 mm | BREEAM Outstanding, passive solar heating |
| Manitoba Hydro Place | Winnipeg | Multi-story | 1000 mm | Solar chimney, 115 m tall, -40°C to +35°C climate |
| Stadttor Düsseldorf | Düsseldorf | Corridor | 1400 mm | Habitable cavity, full walk-in access |
Kinetic facades contain elements that physically move in response to environmental conditions, occupant input, or programmed patterns. They represent the most sophisticated integration of computational design, mechanical engineering, and environmental performance.
| Sensor | Measures | Typical Placement | Range | Application |
|---|---|---|---|---|
| Pyranometer | Global solar irradiance (W/m²) | Roof, facade | 0-1400 W/m² | Solar tracking, shading control |
| Photodiode array | Light level and direction | Facade surface | 0-100,000 lux | Glare detection, daylight optimization |
| Thermocouple/RTD | Temperature | Cavity, interior, exterior | -40 to +80°C | Thermal control, overheating protection |
| Anemometer | Wind speed and direction | Roof | 0-50 m/s | Wind safety, natural vent control |
| Occupancy sensor | Presence/absence | Interior zones | Binary or count | Demand-based facade response |
| Rain sensor | Precipitation | Roof or facade | Binary | Close ventilation openings |
| CO2 sensor | Indoor air quality | Interior | 0-5000 ppm | Ventilation demand control |
Open-loop: Facade elements follow a pre-programmed schedule based on time-of-day and season. No sensors. Simple and reliable but cannot respond to actual conditions (cloud cover, etc.).
Closed-loop: Sensors measure environmental conditions; a controller adjusts facade elements to maintain setpoints (e.g., interior illuminance = 500 lux, interior temperature < 25°C). PID control or rule-based logic.
Predictive: Weather forecast data (cloud cover, temperature, wind) is integrated into the control algorithm. The facade pre-adjusts before conditions change, reducing lag and overshoot. Machine learning models can learn optimal strategies from historical data.
Distributed vs. centralized: Each facade module can have its own controller (distributed — resilient but complex to coordinate) or a central BMS (building management system) can command all modules (centralized — easier to coordinate but single point of failure).
| Mechanism | DOF | Motion Type | Complexity | Example |
|---|---|---|---|---|
| Rotation (single axis) | 1 | Angular | Low | Louver blade |
| Translation (slide) | 1 | Linear | Low | Sliding screen |
| Folding (single hinge) | 1 | Angular | Low | Hinged panel |
| Bi-fold | 1 (coupled) | Double angular | Medium | Folding shutter |
| Scissor mechanism | 1 | Expanding | Medium | Deployable screen |
| Origami fold | 1-3 | Complex folding | High | Miura-ori panel |
| Iris mechanism | 1 | Radial open/close | Medium | Circular aperture |
| Auxetic expansion | 1 | 2D expansion | High | Rotating square pattern |
| Cable-net deformation | Multiple | Surface deformation | High | Tensioned mesh |
Solar position calculation:
Tracking strategies:
The Al Bahar Towers in Abu Dhabi (2012, Aedas Architects) feature a kinetic mashrabiya screen on the exterior of twin 145m towers. Key technical data:
Modeling kinetic facades requires time-step simulation because the facade state changes throughout the day and year:
Environmental performance facades are designed from the outside in — the environmental loads on each square meter of facade surface determine its composition, geometry, and behavior.
For a horizontal overhang to shade a window at a given cut-off solar altitude angle alpha:
Overhang depth D = H / tan(alpha)
Where H is the height from the bottom of the window to the overhang.
Cut-off angle by latitude and orientation (for south-facing facade, northern hemisphere, summer solstice noon):
| Latitude | Solar Altitude (summer solstice) | Recommended Cut-off Angle |
|---|---|---|
| 0° (Equator) | 66.5° (at solstice) / 90° (equinox) | 75° |
| 10° | 76.5° / 80° | 70° |
| 20° | 86.5° / 70° | 65° |
| 30° | 83.5° / 60° | 55° |
| 40° | 73.5° / 50° | 50° |
| 50° | 63.5° / 40° | 45° |
| 60° | 53.5° / 30° | 40° |
Example: At 40°N latitude, south-facing window 1.5m tall, with overhang at window head. Cut-off angle = 50°. Overhang depth D = 1.5 / tan(50°) = 1.5 / 1.19 = 1.26 m.
Vertical fins are most effective on east and west facades where the sun angle is low. Fin depth and spacing determine the shading mask:
Shading mask angle = arctan(fin depth / fin spacing)
For 80% shading of low-angle sun: fin depth / spacing ratio should be approximately 1.0 to 1.5.
Combines horizontal and vertical elements. The resulting shading mask is the intersection of the horizontal overhang mask and the vertical fin mask. Provides omnidirectional shading but reduces daylight admission and views.
| BIPV Type | Efficiency | Transparency | Appearance | Cost ($/Wp) |
|---|---|---|---|---|
| Monocrystalline silicon | 18-22% | Opaque (or spaced cells for semi-transparent) | Dark blue/black cells | 0.30-0.50 |
| Polycrystalline silicon | 15-18% | Opaque | Blue cells | 0.25-0.40 |
| Amorphous silicon (a-Si) | 6-8% | 10-30% VLT achievable | Uniform dark tint | 0.40-0.60 |
| CdTe thin-film | 12-16% | 10-40% VLT achievable | Dark brown/black | 0.30-0.50 |
| CIGS thin-film | 14-18% | Low (typically opaque) | Black | 0.35-0.55 |
| Organic PV (OPV) | 8-12% | Up to 50% VLT | Colored, printable | 0.50-1.00 |
| Perovskite (emerging) | 15-25%+ | Tunable | Tunable color | Research stage |
Aesthetic integration strategies:
Fabrication-aware design integrates manufacturing constraints into the computational model from the outset, preventing designs that are geometrically elegant but unbuildable or prohibitively expensive.
Glass types and properties:
| Type | Thickness Range | Max Size (typical) | Key Property |
|---|---|---|---|
| Float (annealed) | 2-25 mm | 3210 x 6000 mm | Base product, can be cut to shape |
| Heat-strengthened | 4-19 mm | 2440 x 4800 mm | 2x bending strength of annealed |
| Fully tempered | 4-19 mm | 2440 x 4800 mm | 4x bending strength, safety glass |
| Laminated | 2+2 to 19+19 mm | Limited by autoclave | Safety, acoustic, UV blocking |
| Insulated (IGU) | 16-60 mm total | 2800 x 6000 mm | Thermal performance |
| Structural glass | 15-25 mm tempered | 3000 x 6000 mm | Load-bearing fins, beams |
Processing capabilities:
Structural glass systems:
| Material | Thickness Range | Max Panel Size | Key Properties |
|---|---|---|---|
| Aluminum composite (ACM) | 3-6 mm total (0.5 mm skins) | 1500 x 5000 mm | Lightweight, foldable, fire rating concerns |
| Zinc | 0.7-1.5 mm | 1000 x 3000 mm | Self-healing patina, long life |
| Copper | 0.6-1.5 mm | 1000 x 3000 mm | Patina development, premium |
| Stainless steel | 0.5-3.0 mm | 1500 x 6000 mm | Durable, corrosion resistant |
| Perforated metal | 0.5-6.0 mm | 1500 x 3000 mm | Shading, screening, decorative |
| Expanded metal | 1.0-6.0 mm | 1250 x 2500 mm | 3D texture, directional transparency |
| Woven metal mesh | Wire dia 0.5-4.0 mm | Custom widths | Drapeable, large spans |
| Corten steel | 2.0-12.0 mm | 2500 x 12000 mm | Weathering patina, structural |
| System | Description | Speed | Cost | Tolerance Absorption | Thermal Break |
|---|---|---|---|---|---|
| Unitized | Factory-assembled frames with infill, hung on floor edge brackets | Fast | High | Good (stack joint, mullion joint) | Integral |
| Stick-built | Mullions and transoms assembled on site, infill glazed on site | Slow | Medium | Moderate | Add-on |
| Point-fixed | Glass bolted to spider fittings on steel structure | Medium | High | Low (requires precise structure) | At fitting |
| Cable-net | Pre-tensioned cable grid with point-fixed glass | Slow | Very high | Very low | At clamp |
| Rainscreen | Cladding panels on brackets with open or sealed joints | Medium | Low-Med | Good (bracket adjustment) | At bracket |
Facade systems must accommodate tolerances from multiple sources:
| Source | Typical Tolerance | Accumulated at 60m Height |
|---|---|---|
| Structural frame (concrete) | ±20 mm per floor | ±80 mm |
| Structural frame (steel) | ±10 mm per floor | ±40 mm |
| Facade bracket | ±15 mm adjustment range | — |
| Facade mullion | ±3 mm fabrication | — |
| Glass panel | ±1 mm cutting | — |
| Gasket/sealant | ±3 mm compression range | — |
| Total system | Must accommodate ±25 mm | — |
Tolerance absorption strategy: The facade bracket (angle or channel connecting facade to structure) is the primary tolerance absorption point. It must provide adjustment in three axes: ±15-25 mm in-out, ±10-15 mm vertical, ±10-15 mm lateral. Slotted holes and shim packs are standard methods.
| Metric | Unit | Description | Typical Range |
|---|---|---|---|
| U-value | W/(m²·K) | Overall thermal transmittance | 0.7 - 5.8 |
| R-value | (m²·K)/W | Thermal resistance (1/U) | 0.17 - 1.43 |
| SHGC | dimensionless | Solar heat gain coefficient | 0.15 - 0.70 |
| VLT | % | Visible light transmittance | 10% - 80% |
| LSG | dimensionless | Light-to-solar gain ratio (VLT/SHGC) | 0.8 - 2.5 |
| Uf | W/(m²·K) | Frame U-value | 1.0 - 7.0 |
| Ug | W/(m²·K) | Glass center-of-pane U-value | 0.5 - 5.7 |
| Psi | W/(m·K) | Linear thermal transmittance (edge spacer) | 0.03 - 0.10 |
| Metric | Unit | Description | Typical Design Values |
|---|---|---|---|
| Wind load resistance | kPa | Maximum wind pressure | 1.0 - 6.0 kPa |
| Dead load | kg/m² | Self-weight of facade assembly | 25 - 120 |
| Live load (maintenance) | kN | Point load on glass for cleaning cradle | 1.0 kN per pad |
| Seismic drift | mm | Inter-story drift accommodation | ±15 - ±75 mm |
| Impact resistance | J | Soft body / hard body impact | Cat 1-5 (EN 14019) |
| Deflection limit | span/L | Maximum mullion deflection | L/175 to L/250 |
| Metric | Unit | Description | Typical Range |
|---|---|---|---|
| STC | dimensionless | Sound Transmission Class (US/Canada) | 28 - 55 |
| Rw | dB | Weighted sound reduction index (ISO) | 28 - 55 |
| Ctr | dB | Spectrum adaptation term (traffic noise) | -3 to -10 |
| Rw + Ctr | dB | Traffic noise adjusted rating | 25 - 48 |
| OITC | dimensionless | Outdoor-Indoor Transmission Class | 25 - 45 |
| Classification | Standard | Description |
|---|---|---|
| Non-combustible | ASTM E136, EN 13501-1 A1 | Does not contribute to fire (glass, steel, stone) |
| Limited combustible | ASTM E136 alternate | Minimal contribution (some composites) |
| Class A | ASTM E84 Class A | Flame spread index 0-25 |
| B-s1,d0 | EN 13501-1 | Very limited contribution, no smoke, no droplets |
| Fire-rated | BS 476, ASTM E119, EN 1364 | EI 30/60/90/120 rating |
| Cavity barrier | Local codes | Fire stops within rainscreen cavities |
| Metric | Unit | Description | Typical Values |
|---|---|---|---|
| Design life | years | Expected service life | 25 - 60 |
| Maintenance interval | years | Between major maintenance events | 5 - 15 |
| Sealant life | years | Expected sealant replacement cycle | 10 - 25 |
| Embodied carbon | kgCO2e/m² | Carbon footprint of materials and fabrication | 40 - 250 |
| Recyclability | % by mass | Proportion recoverable at end of life | 30 - 95% |
| LCA impact | Various | Full life cycle assessment categories | Per EN 15804 |
| Circular design | Qualitative | Design for disassembly and reuse potential | Score 1-5 |
| Facade Type | U-value (W/m²K) | SHGC | STC | Weight (kg/m²) | Cost ($/m²) |
|---|---|---|---|---|---|
| Single glazed curtain wall | 5.5 | 0.80 | 28 | 25 | 300-500 |
| Double glazed curtain wall | 1.6-2.0 | 0.25-0.40 | 32-36 | 35-45 | 500-900 |
| Triple glazed curtain wall | 0.7-1.0 | 0.20-0.35 | 36-42 | 55-70 | 900-1500 |
| Double-skin facade | 0.8-1.5 | 0.10-0.30 | 42-55 | 80-120 | 1200-2500 |
| Unitized curtain wall | 1.2-1.8 | 0.20-0.40 | 34-40 | 40-60 | 700-1200 |
| Stone rainscreen | 0.25-0.35 | 0 (opaque) | 45-55 | 80-150 | 800-1500 |
| Metal rainscreen | 0.20-0.30 | 0 (opaque) | 38-48 | 25-50 | 400-900 |
| ETFE cushion | 1.5-2.8 | 0.50-0.85 | 15-20 | 3-5 | 400-800 |
| Structural glass | 1.8-2.5 | 0.30-0.65 | 30-38 | 30-50 | 1500-3500 |