Why it matters
Without AI, misidentifications can lead to irreversible treatment choices, unsafe solvent use, delayed condition reports, and contested insurance claims due to uncertain material provenance.
From design to safety
Design
AI-assisted treatment planning
The AI reads existing sample images, dye references, and preliminary microscopy notes to propose methods, materials, and reversibility considerations, while flagging ethical constraints and conservation priorities rooted in PACR guidelines.
- polarized light microscope
- FTIR spectrometer
- Munsell color chart
Build
Analytical bench integration
The system configures data streams from microscopy, spectroscopy, and imaging to support reproducible analytics, standard operating procedures, and secure project documentation within the studio information system.
- polarized light microscope
- Raman spectrometer
- UV-Vis spectrophotometer
Operation
Real-time identification workflow
During examination, AI suggests dye identifications and confidence scores, cross-checks with reference libraries, generates preliminary analysis reports, and logs observations into the condition report workflow.
- digital imaging station
- Raman spectrometer
- UV-Vis spectrophotometer
Maintenance
Calibration and data governance
Regular calibration, versioned reference libraries, audit trails, data schemas, and long-term archival plans preserve data integrity, traceability, and re-analysis capability across projects.
- digital imaging station
- FTIR spectrometer
- Munsell color chart
Safety
Ethics and regulatory compliance
All AI-driven decisions are signed off by the accredited conservator or PACR-accredited supervisor, with appropriate COSHH adherence, ethics review, and comprehensive documentation for insurance.
- polarized light microscope
- digital imaging station
- Munsell color chart
Part by part
- 1
polarized light microscope
- Watch for
- Small misidentification due to overlapping fibre features; fading fluorescence under certain illumination.
- AI action
- AI flags uncertain results and prompts confirmatory tests; triggers order for reference standards.
- 2
FTIR spectrometer
- Watch for
- Spectral overlap between natural and synthetic components; low signal intensity due to sample thickness.
- AI action
- AI assesses spectra against library; suggests potential dye/fibre classes and required sample prep.
- 3
Raman spectrometer
- Watch for
- Fluorescence masking Raman bands; weak signals in dark dyes.
- AI action
- AI routes high-confidence peaks to analyst; proposes alternative laser wavelength or micro-sampling.
- 4
UV-Vis spectrophotometer
- Watch for
- Low absorbance signals; baseline drift across runs.
- AI action
- AI evaluates spectra, flags suspect baselines, and recommends re-measurement with calibration check.
- 5
digital imaging station
- Watch for
- Colorimetric mismatch across lighting conditions; inconsistent metadata.
- AI action
- AI records standardized lighting, captures color references, and creates reproducible condition-logging templates.
- 6
Munsell color chart
- Watch for
- Color drift due to aging or light exposure; non-uniform colour patches.
- AI action
- AI aligns sample color with standardized references and logs color coordinates for reports.
Worked example
Fibre and dye screening for a mid-19th century dress
Illustrative scenario with sample data.
A conservator seeks to identify silk fibre type and natural vs synthetic dyes to inform a reversible cleaning plan, ensuring compatibility with prior repairs and future loan requirements.
Signal in
sample_id: 5678, fibre: silk, threads: 20, dye_candidates: logwood, madder, aniline, spectra: FTIR-1, Raman-3, imaging: macro, micro, prep: sectioned, baseline: 0.02 AU
System out
analysis_report: fibre_type=sericin-rich silk; dye_class=natural/madder; colourfastness rating=moderate; confirmation_needed=solvent test scope; confidence=high; recommended treatment_pathway=selective solvent testing with reversibility notes; archival entry created.
Human sign-off: Head conservator signs off the treatment proposal, approves the analytical strategy, and authorizes archival tagging for the conservator’s file and loan dossier.
How to roll it out
Phase 1
Weeks 1 to 4
Baseline data capture, equipment validation, and staff training on data entry, terminology, reporting templates, and governance for dye databases.
Phase 2
Weeks 5 to 12
Pilot runs with real casework, library alignment, SOP finalization for dye identification workflows, and integration with condition-report templates.
Phase 3
Weeks 13 to 20
Full integration with treatment proposals, archival records, and preventive conservation plans; ongoing audits, updates to reference libraries, and cross-department reporting.
Where people stay in charge
- AI-derived decisions must be reviewed by an accredited conservator before treatment proposals are issued, and any solvent strategy requires direct sign-off by the lead conservator.
- All data and analytics must be stored in the studio's audit-trail with versioned references for future re-analysis, including provenance notes for each dye class.
- No irreversible treatment decisions should be based solely on AI without manual corroboration by a conservator, with a documented rationale and required approvals.
Questions owners and crews ask
What is AI fibre identification in a textile lab?
AI fibre identification assists analysts by interpreting microscopy and spectroscopy data, guiding confirmatory tests, and documenting identifications within the conservation report, while maintaining traceability in the condition and treatment records.
How does AI handle natural vs synthetic dye identification?
AI compares spectral fingerprints against reference libraries, flags ambiguities, and suggests additional tests or literature reviews before reporting, with an emphasis on reversibility and archival accuracy.
Who signs off AI-generated treatment recommendations?
The accredited conservator or PACR-accredited supervisor signs off to ensure ethical, reversible, and well-documented practices, with final approval recorded in the treatment proposal.
What records are created for AI-driven analysis?
Audit trails, condition records, treatment proposals, and final reports include AI-derived notes, data sources, decision rationales, and linked digital images for provenance.