Predicted Class3
Target StatusMeets Class ≥3 target
Main DriverYarn Specifications
Main Trade-offBalanced profile

Guided Demo

Open Sample Report

AI Engineering Assistant

AI explains the model; it does not replace the prediction.
AI Mode

For the live demo, the key is used from this browser. Clear it after presenting, especially on shared computers.

Recommended: GPT-5 mini for clear engineering briefs and lab-trial planning.

Paste an API key and choose an AI action, or switch to Demo fallback for a no-network presentation.
Prompt and structured scenario data sent to AI

Pilling Class

Meets target
Design Typicality
85%
🤲
HandleFair
💰
Relative CostMedium
⏱
Processing TimeMedium
⚡
Energy UseMedium

Baseline vs Scenario

Click Set as Baseline, then change inputs to compare development alternatives.

MetricBaselineScenarioChange

Scope / Boundary Warnings

    Path to Class ≥3

      Riopele Decision Value

      The app is useful before full validation because it structures engineering decisions and reduces low-value trial paths.
      Reduce wasted trialsScreen weak fabric recipes before committing lab time.
      Accelerate Class ≥3 pathsMove from vague improvement ideas to controlled scenario changes.
      Capture expert knowledgeMake pilling know-how visible, configurable, and easier to transfer.
      Improve communicationTurn technical scenarios into briefs for R&D, production, finance, and customers.

      Influence by Process Stage

      Parameter-Level Feature Importance

      Trade-off Analysis

      Validation & Calibration Plan

      AI explains; the transparent model predicts; ISO 12945-2 testing validates.

      Data Riopele can provide

      • Fabric composition, yarn count, twist, spinning process, hairiness, weave, density, finishing route and heat settings.
      • ISO 12945-2 pilling class, rub count, product family, hand-feel notes and production constraints.
      • Known pass/fail examples near the Class ≥3 threshold.

      12-week pilot path

      • Weeks 1-2: UX, explanation and report polish.
      • Weeks 3-5: validation dataset template and data-quality rules.
      • Weeks 6-9: calibrate the transparent model around Class ≥3.
      • Weeks 10-12: user guide, validation report and handoff package.

      Pilot Integration Path

      Start lightweight, then connect to Riopele systems only after the method proves value.
      NowCSV/Excel scenario template, local reports, no production-system dependency.
      PilotVersioned model coefficients, validation dataset and repeatable review workflow.
      LaterAPI connection to lab, PLM, MES or quality records if Riopele wants integration.
      GovernanceModel version, calibration scope, assumptions and limitations visible in every export.

      From Pilling To Textile Engineering Intelligence

      Pilling is the first vertical. The explainable structure can extend to other textile decisions.
      Finishing Recipe OptimizationPredict how finishing variables influence softness, easy-care, durability, drape and pilling.
      Textile Tradeoff SimulatorCompare comfort, appearance retention, durability, sustainability, cost and customer segment fit.
      PillMarkPatentable concept: engineer visible wear as an intentional surface signal.
      ComfortBloomPatentable concept: tune comfort and perceived softness over repeated use.

      Model Basis & Validation Plan

      Model Basis

      PillNot uses a transparent deterministic scoring model that maps fiber composition, yarn structure, woven construction, and finishing choices to an estimated ISO 12945-2 pilling class. Model coefficients are separated in model-config.js so the prototype is calibration-ready without rewriting application code.

      Model version: 1.1

      Validation Plan

      • Collect representative polyester-predominant woven articles.
      • Run ISO 12945-2:2000 pilling tests.
      • Record material, yarn, construction, and finishing metadata.
      • Calibrate model weights against actual class results.
      • Track error bands by product family.