12.1 Establishing Baselines and Documentation
Key Takeaways
- A baseline is a documented reference thermal state of equipment under known operating conditions—not merely the first pretty picture saved on a camera
- Baseline packages must record load, ambient (and related environment), camera parameters (ε, RAT, distance, atmosphere/window), equipment ID/location, date/time, and the thermogram plus visual image
- Establish baselines on new or major-modified equipment after commissioning at representative load; re-baseline after repair, rebuild, or configuration change that alters normal thermal behavior
- Same-component baseline trending complements (does not replace) similar-component NETA-style snapshot severity when peers exist
- Level II owns baseline quality: comparable geometry, correct radiometry, and metadata complete enough that a future inspector can reproduce the comparison
Snapshot severity (Chapter 8) answers “Is this abnormal right now versus a peer or a design limit?” Trending answers “Is this changing in a way that predicts failure or demands earlier action?” Both skills require a trustworthy baseline—a documented reference thermal state against which later inspections are compared. Level II thermographers design baseline programs, not just collect random archives.
What a Baseline Is (and Is Not)
| Concept | Definition | Level II implication |
|---|---|---|
| Baseline image/set | Quantitative thermogram(s) of a specific asset under known, recorded conditions, with full metadata | Enables same-component ΔT over time |
| Pretty archive | Image without load, ambient, parameters, or equipment ID | Nearly useless for defensible trending |
| Peer reference | Similar component at the same survey (NETA-style) | Snapshot severity; different from multi-year baseline |
| OEM “normal” table | Generic design guidance | Complements but does not replace site-specific baselines |
A baseline is a measurement package, not a single JPEG. Future inspectors must recreate enough of the original conditions to interpret rate of change without inventing history.
Minimum Contents of a Baseline Package
Every baseline (and every later trend point) should capture the fields below. Missing fields create false trends or force “cannot compare” conclusions.
| Field | What to record | Why it matters |
|---|---|---|
| Equipment ID / location | Unique asset tag, panel ID, phase, cubicle, bearing position, route stop | Wrong asset = meaningless trend |
| Date and time | Local date/time; note shift if relevant | Timeline for rate-of-change |
| Load / process state | Amps or % rated, production rate, RPM, steam load, HVAC mode | I²R and process heat dominate temperature |
| Ambient / environment | Air temperature; for outdoors wind/sun; for buildings indoor–outdoor ΔT when relevant | Absolute T and some ΔT contexts shift with ambient |
| Camera parameters | Emissivity, RAT, distance, atmospheric temp, RH; window τ if used | Radiometric comparability |
| Geometry | Lens, stand-off, angle, IR window ID | Spot-size and path errors |
| Thermogram + visual | Radiometric file preferred; annotated ΔT regions of interest (ROI) | Pattern + identification |
| Reference description | What was compared (same lug over time; peer phase; baseline ROI) | Clarity for later reviewers |
| Measured values | T_suspect (or ROI stats), optional T_peer, ambient | Quantitative core |
| Inspector / camera ID | Name/cert level, camera serial or model | Traceability and QA |
Worked minimum label (electrical connection)
Asset: MCC-3 / Starter 12 / Load side Phase B lug
2026-03-12 14:20 Load: 48 A (62% FLA) Ambient: 24 °C
ε = 0.95 (oxidized lug + tape reference), RAT = 24 °C, d = 1.2 m, no window
ROI max = 41 °C; Phase A peer = 36 °C; ΔT_peer = 5 °C
Camera: Model X / SN … Inspector: J. Doe, Level II
That package can support both peer severity today and same-lug trend next year.
Same-Component Baseline vs Peer Snapshot
| Comparison | Best for | Weak when |
|---|---|---|
| Similar component, similar load (NETA-style) | Connection faults, phase imbalance signatures | Unique assets; intentional unbalance |
| Same component vs its baseline | Slow degradation, post-repair verification, assets without peers | Baseline missing or conditions not comparable |
| Absolute / OEM limits | Design temperature ceilings | Does not show early peer ΔT faults |
Level II rule: Use both when possible. A lug can sit in Priority 4 versus peers (small ΔT) yet still show a rising same-component trend that justifies earlier re-inspection. Conversely, a one-time peer ΔT of 22 °C needs Priority 2 action even if last year’s baseline is incomplete.
When to Establish Baselines
New and major-modified equipment
Baseline after commissioning when the asset is:
- Mechanically/electrically complete and accepted for service
- At representative operating load (electrical IR commonly ≥ ~40% of rated when safe; prefer higher)
- Thermally steady (not mid-ramp after a cold start unless the program explicitly trends start-up signatures)
- Accessible with the same IR path you will use later (window installed, covers policy defined)
| Situation | Baseline timing |
|---|---|
| New switchgear lineup | After energization and loaded operation; per circuit or representative loads |
| New motor / gearbox | After break-in at normal process load and stable bearing temperatures |
| Building envelope after renovation | Under valid envelope ΔT conditions (Chapter 8 / 10), not on a mild spring day only |
| Steam system after trap replacement | At normal steam pressure/load once the circuit is stable |
Do not treat factory FAT images under unknown load, or construction-phase cold equipment, as the plant’s operating baseline unless the program documents them as a separate “as-built cold” set.
Fleet / route programs
For predictive maintenance routes, baselines are often established on the first valid survey of each stop after the program starts—or when a stop is added. Level II standardizes ROI placement (same bolt head, same bearing housing location) so multi-year plots compare the same physical point.
When to Re-Baseline
A baseline becomes obsolete when the asset’s normal thermal identity changes for reasons other than the fault mode you are monitoring.
| Trigger | Action |
|---|---|
| Repair or replacement of the monitored component (retorque, new lug, new bearing, new trap) | Re-baseline after return to stable service at comparable load |
| Major modification (rewire, different fuse type, coupling change, insulation upgrade) | New baseline; archive old series as closed |
| IR path change (new IR window, different stand-off policy, permanent shield) | Re-baseline with new path parameters |
| Coating / surface change that alters emissivity | Re-establish ε method and re-baseline temperatures |
| Load profile permanent change (upsized motor, different process duty) | New “normal” baseline under the new duty |
| Suspected bad historical baseline (wrong ε, wrong asset ID) | Invalidate; create a corrected baseline rather than trend garbage |
Post-repair verification is both a QA step and a new baseline seed: confirm the anomaly cleared and capture the new normal under load for future trending.
Documentation Discipline and Storage
| Practice | Rationale |
|---|---|
| Radiometric originals + report derivatives | Later re-analysis with corrected ε if needed |
| Consistent file naming (site-asset-date) | Findable multi-year sets |
| ROI templates / overlays | Same measurement point each visit |
| Database fields for load and ambient | Enable normalized trends (Section 12.2) |
| Link work orders to IR finding IDs | Close the loop after repair re-baseline |
Paper thumbnails without parameters fail insurance, ISO-style condition monitoring programs, and Level II QA review.
Quality Checks Before Accepting a Baseline
- Identity: Asset tag matches the physical equipment and single-line / P&ID location.
- Radiometry: ε and RAT defensible; focus sharp; spot size adequate for the target.
- Load: Adequate and recorded; not a no-load energization photo labeled “baseline.”
- Steady state: Temperatures not climbing rapidly after a load step unless noted.
- Completeness: All minimum metadata fields present.
- Peer context: Where peers exist, record peer ΔT as well as absolute ROI temperature—future readers get both lenses.
Worked Scenarios
Scenario A — New feeder. After commissioning, Phase A/B/C lugs are balanced within 2 °C at 55% load. Baseline package stored with parameters and currents. One year later Phase B is +8 °C vs peers and +12 °C vs its own baseline at similar load → early degradation signal even if still Priority 4 by peer band.
Scenario B — Missing load. Archive image shows a warm breaker with no amps recorded. Cannot use as quantitative baseline; schedule a proper baselining survey.
Scenario C — After repair. Priority 2 connection repaired; re-inspect at load shows ΔT_peer = 1 °C. Re-baseline the repaired lug; do not keep trending against the pre-repair hotspot as if it were “normal.”
Scenario D — Window retrofit. IR window installed on a gear door. Old open-door baselines are not path-comparable; establish through-window baselines with measured/known τ.
Scenario E — Wrong asset. Baseline labeled “Pump 4 drive end” was actually non-drive end. Level II invalidates the series and starts a correct baseline rather than inventing a failure trend.
Common Traps
| Trap | Correct Level II view |
|---|---|
| “First image ever” without metadata = baseline | Incomplete; not trendable |
| Baselining at 10% electrical load | Inadequate for connection-sensitive programs; re-baseline at ≥ ~40% |
| Never re-baselining after repair | Trends compare fault to repaired state incorrectly |
| Changing ε every year without notes | False temperature trends |
| Using only palette color memory | Not quantitative |
| Mixing peer ΔT and ambient ΔT in one chart without labels | Confuses rate-of-change |
Summary for Recall
A baseline is a full documentation package: equipment ID, date/time, load, ambient/environment, camera parameters, geometry/path, thermogram + visual, and measured ROI values. Establish baselines on new/commissioned equipment at representative steady load, and re-baseline after repair or modification. Same-component baselines power multi-year trending; they complement peer-based NETA snapshot severity. Level II refuses pretty-but-empty archives and builds comparisons future inspectors can trust.
Which set of elements is most essential for a defensible IR baseline package used in future trending?
When should Level II typically re-baseline a connection that was repaired after a Priority 2 finding?
Why is baselining a loaded electrical connection preferred over a no-load energization photo?
A plant installs a new IR window on a previously open-door surveyed switchgear cubicle. What is the best baseline action?