Textile IR: A Bidirectional Intermediate Representation for Physics-Aware Fashion CAD

TL;DR

Textile IR addresses the fragmentation of CAD, physics simulation, and lifecycle assessment in fashion design by introducing a formal intermediate representation and a seven-layer Verification Ladder that enables bidirectional, uncertainty-aware feedback along the design pipeline. By reframing fashion engineering as a constraint-satisfaction/program-synthesis problem, the approach enables topological design decisions to be validated before costly prototyping, while propagating uncertainty through the chain to produce honest sustainability claims. The paper details architectural principles, a worked example of bidirectional flow, and a forward-looking research agenda aimed at industry SMEs, regulatory compliance, and cross-domain integration. If adopted, Textile IR could shorten design cycles, improve manufacturability and sustainability tradeoffs, and provide a verifiable digital product passport from concept to production.

Abstract

We introduce Textile IR, a bidirectional intermediate representation that connects manufacturing-valid CAD, physics-based simulation, and lifecycle assessment for fashion design. Unlike existing siloed tools where pattern software guarantees sewable outputs but understands nothing about drape, and physics simulation predicts behaviour but cannot automatically fix patterns, Textile IR provides the semantic glue for integration through a seven-layer Verification Ladder -- from cheap syntactic checks (pattern closure, seam compatibility) to expensive physics validation (drape simulation, stress analysis). The architecture enables bidirectional feedback: simulation failures suggest pattern modifications; material substitutions update sustainability estimates in real time; uncertainty propagates across the pipeline with explicit confidence bounds. We formalise fashion engineering as constraint satisfaction over three domains and demonstrate how Textile IR's scene-graph representation enables AI systems to manipulate garments as structured programs rather than pixel arrays. The framework addresses the compound uncertainty problem: when measurement errors in material testing, simulation approximations, and LCA database gaps combine, sustainability claims become unreliable without explicit uncertainty tracking. We propose six research priorities and discuss deployment considerations for fashion SMEs where integrated workflows reduce specialised engineering requirements. Key contribution: a formal representation that makes engineering constraints perceptible, manipulable, and immediately consequential -- enabling designers to navigate sustainability, manufacturability, and aesthetic tradeoffs simultaneously rather than discovering conflicts after costly physical prototyping.

Figures (8)

• Figure 1: The three pillars of AI-assisted fashion engineering operate as disconnected islands despite individual maturity. CAD guarantees geometric validity, physics simulation predicts drape, and LCA quantifies environmental impact. Each pillar's strength creates a blind spot: CAD produces patterns without physics data, simulation loses pattern semantics, LCA calculates impact without geometric context. No shared semantic layer enables bidirectional flow.
• Figure 2: Spatial intelligence converts fashion imagery into immersive experiences. Neural radiance fields mildenhall_nerf_2020 pioneered neural scene representation (2020), evolving through 3D Gaussian Splatting kerbl_3d_2023 (2023) to hybrid neural-geometric representations. Modern world models accept diverse inputs from single photographs to multi-view capture. Fashion applications include virtual try-on, digital twin stores, and immersive fashion shows.
• Figure 3: Multi-criteria decision making exposes tradeoffs rather than hiding them. (A) Pareto frontier showing leather material options. (B) Material options with LCA data: animal leather (110 kg CO$_2$e/m$^2$), PU synthetic (15.8 kg CO$_2$e/m$^2$), cactus leather (1.4 kg CO$_2$e/m$^2$). (C) Radar profiles showing unique performance across criteria. (D) FAHP-TOPSIS weighting interface.
• Figure 4: Eight data representations spanning the fashion product lifecycle, connected through a central Textile IR / DPP Data Hub. Design representations (Vision, Sketch, Tech Pack, Digital Twin) prioritise creative intent. Simulation representations (FEM) prioritise physical accuracy. Lifecycle representations (LCA, Circular, Supply Chain) prioritise environmental metrics. Only the central hub can bridge these modalities.
• Figure 5: The non-differentiability barrier. (A) Continuous parameters form smooth optimisation landscapes where gradient descent converges. (B) Discrete topology decisions create discontinuities---there is no "half a dart." (C) Hybrid solution: search over topology, optimisation over parameters, then physics verification korosteleva_garmentcode_2023kodnongbua_design_2025.