Why it matters
Without AI, misidentification can lead to irreversible treatment errors, unresolved solvents interactions, and disputed insurance claims due to uncertain spectral matches or missing library coverage.
From design to safety
Design
AI-guided spectroscopy strategy
Define data requirements, select spectral ranges for Raman and FTIR, specify library scope (binders, varnishes, pigments), align with treatment proposals and reversibility goals, and establish templates with identifiers.
- Raman spectrometer
- FTIR spectrometer
- Spectral libraries database
Build
Instrument and workflow setup
Configure spectrometers, calibrations, ATR accessories, and data pipelines; document sampling plans and reporting templates for condition records; implement versioned library schemas and QA checkpoints and traceable audit trails.
- Raman spectrometer
- FTIR spectrometer
- KBr pellet press
Operation
Real-time spectra matching
Acquire spectra, perform baseline correction, peak deconvolution, compare to libraries, assign IDs or confidence levels, and generate annotation-rich reports with reversible options for client review and archival records.
- Raman spectrometer
- FTIR spectrometer
- Spectral libraries database
Maintenance
Calibration and data hygiene
Schedule periodic calibration, validate library updates, monitor instrument drift, and archive all processed spectra for audit and follow-up treatments; enforce data integrity rules and access controls throughout operations.
- Raman spectrometer
- FTIR spectrometer
- ATR crystal accessory
Safety
Ethical and safety governance
Ensure COSHH assessment, reversibility criteria, and PACR accreditation alignment; sign-off by accredited conservator and records for insurance and curator review; establish on-site safety regimes for lasers and powders.
- Diamond knife microtome
- KBr pellet press
- ATR crystal accessory
Part by part
- 1
Raman spectrometer
- Watch for
- Fluorescence interference and fluorescence quenching masking true Raman signals.
- AI action
- Flag weak or conflicting peaks and prompt re-acquisition with adjusted laser power or notch filters.
- 2
FTIR spectrometer
- Watch for
- Baseline drift or atmospheric water peak masking key bands.
- AI action
- Trigger baseline correction and spectral deconvolution, re-run with alternative ATR/PI settings.
- 3
KBr pellet press
- Watch for
- Inconsistent pellet density causing peak broadening or scattering artifacts.
- AI action
- Flag low-density pellets and prompt re-press with standardized press parameters.
- 4
Varnish identification kit
- Watch for
- Aging varnish bands overlapping pigment signals causing misidentification.
- AI action
- Suggest targeted spectra at varnish/pigment interface and compare against updated varnish library entries.
- 5
Spectral libraries database
- Watch for
- Missing reference spectra for certain binding media or varnishes.
- AI action
- Flag gaps and propose library supplementation or targeted reference sample analysis.
- 6
ATR crystal accessory
- Watch for
- Crystal contamination or coating affecting surface-sensitive measurements.
- AI action
- Prompt cleaning or re-measurement with fresh crystal and adjusted contact pressure.
Worked example
Catching a binder ID without invasive sampling
Illustrative scenario with sample data.
A private studio seeks to identify binder type and varnish age in a painted panel with non-destructive Raman and FTIR, using AI-assisted matching against a curated library and targeted surface sampling when needed.
Signal in
spectraRaman: scan_id=R1234, run=1, laser=532nm; spectraFTIR: scan_id=F5678, ATR, resolution=4cm-1; libraryVersion=2026.1; sampleInfo=panel_side_a; history=previous varnish layer present; temperature=22C, humidity=45%, exposure=slow-dry, referenceSpectra=LR2026-B1. Notes: surface retouch risk assessment performed; baseline assay completed; future re-scan recommended after cleaning tests. LibraryVersion confirmed and linked to project file.
System out
AI confidence: BinderID=Apparent Oil-Root Binder; varnishAge=early/oil-aged; confidence=0.86; recommended actions: further targeted spectra at interface, confirm with literature, and document reversible options in the treatment proposal. All data linked to project file and specimen ID with date stamp for audit trails.
Human sign-off: Accredited conservator reviews the AI-generated ID, validates it against binder records, cross-checks with reference samples, and signs off on the treatment plan and documentation package for curator approval and client.
How to roll it out
Phase 1
Weeks 1 to 4
Baseline setup, calibrations, and library alignment across Raman and FTIR workflows.
Phase 2
Weeks 5 to 8
Pilot runs with known reference samples; recordkeeping and initial treatment proposals.
Phase 3
Weeks 9 to 12
Full integration into condition reports, treatment proposals, and client-facing documentation.
Where people stay in charge
- All spectral identifications must be reviewed and signed off by an accredited conservator before inclusion in any treatment proposal.
- The system must enforce auditable approval workflows and cannot finalize reports without curator review.
- Spectral library updates and new reference data require curator and conservator approval prior to deployment.
Questions owners and crews ask
What data do I need before starting AI Raman and FTIR matching?
Collect spectra from areas of interest, ensure instrument calibration, record sample provenance, compile existing reference spectra, and attach these to the project file with timestamps to support traceability and audit trail records.
How does AI handle ambiguous spectral matches?
The system assigns confidence levels, requests targeted re-measurement, and suggests additional library entries; final material ID requires expert review and corroboration with reference material and contextual records from treatment history file docs.
What records are produced for treatment documentation?
A treatment proposal with spectral IDs, method notes, and reversible options is prepared, plus condition reports and post-treatment verification spectra. All data linked to project file, sample IDs, and curator sign-off permanently.
How is safety and accreditation maintained with AI analysis?
Results are signed off by an accredited conservator, COSHH considerations documented, PACR alignment recorded, and ongoing training conducted; the system supports audit trails and insurance-ready documentation for clients and project records continually.