Study guide

12.2 Rate-of-Change and Multi-Inspection Trending

Certified Infrared Thermographer Level II Exam

Track your progress as you read

Finished sections get a check mark.

1 Introduction & Exam Overview

1.1 About the Level II Exam 1.2 Level I vs Level II Scope and Prerequisites

2 Chapter 2: Thermometry Fundamentals

2.1 Temperature Scales and Conversions 2.2 Contact vs Non-Contact Thermometers 2.3 Identifying and Reducing Measurement Errors

3 Chapter 3: Thermal Science and Heat Transfer

3.1 Thermal Science: Temperature, Energy Units, and Heat Capacity 3.2 Heat Transfer Modes and Conduction Variables

4 Chapter 4: Advanced Infrared Theory

4.1 Radiant Power and Stefan-Boltzmann Law 4.2 Planck's Curves and Wien's Displacement Law 4.3 Real Surfaces: Emissivity, Reflectivity, Transmissivity

5 Chapter 5: Emissivity Correction Methods

5.1 Why Emissivity Correction Matters at Level II 5.2 Reference Emitter, Tape, and Paint Methods 5.3 Reflected Apparent Temperature (RAT) Methods 5.4 Emissivity Tables: Metals and Non-Metals

6 Chapter 6: Camera Parameters and Measurement Setup

6.1 The Five Required Camera Parameters 6.2 Atmospheric Temperature, RH, and Distance 6.3 IR Window Transmittance Corrections 6.4 Target Size, IFOV, and Distance-to-Spot Ratio 6.5 Detectors, Optics, and Imager Specifications

7 Chapter 7: Calibration and Advanced Error Sources

7.1 Equipment Calibration and Accuracy Specs 7.2 Cross-Verification with Contact Thermometers 7.3 Filters, Windows, and Spectral Considerations

8 Chapter 8: Severity Classification Standards

8.1 NETA ATS Table 100.18 and Repair-Priority Conventions 8.2 Absolute Temperature Limits (ANSI/IEEE/NEMA) 8.3 ISO 18434 and Machinery Severity Criteria 8.4 Building Envelope Criteria (ASTM C1060 / ISO 6781) 8.5 Load and Ambient Correction of Absolute Limits

9 Chapter 9: Electrical Pattern Analysis

9.1 Three-Phase Electrical Fault Patterns 9.2 Transformer and Switchgear Thermography 9.3 Capacitor, Bus, and Connection Anomalies 9.4 High-Voltage Inspection Conditions and Load Requirements

10 Chapter 10: Mechanical and Steam System Analysis

10.1 Bearing and Rotating Equipment Signatures 10.2 Steam Trap and Steam System Inspection 10.3 Mechanical Drive and Alignment Indicators

11 Chapter 11: Building Envelope Analysis

11.1 Insulation Defects and Thermal Bridging 11.2 Air Infiltration and Blower-Door IR 11.3 Moisture Signatures and Flat Roof Protocols 11.4 Inspection Conditions: ΔT, Wind, and Timing

12 Chapter 12: Trending and Baseline Comparison

12.1 Establishing Baselines and Documentation 12.2 Rate-of-Change and Multi-Inspection Trending 12.3 Escalation Criteria from Trend Data

13 Chapter 13: QA/QC of Level I Work

13.1 Parameter Verification and Coverage Review 13.2 Equipment Identification and Finding Reclassification 13.3 Supervisory Sign-Off Responsibilities

14 Chapter 14: Reporting Standards and Professional Ethics

14.1 Quantitative Report Content Requirements 14.2 Paired Image Documentation and Presentation 14.3 Professional Ethics and Client Communication 14.4 Image Processing and Report Generation Software

12.2 Rate-of-Change and Multi-Inspection Trending

Key Takeaways Explain This Topic AI

Rate of change of ΔT (or same-point temperature under comparable conditions) is often more predictive than a single snapshot severity band alone

Normalize or qualify trends for load and ambient whenever possible; raw temperature plots without condition context produce false trends

Multi-year route plots should use consistent ROI, geometry, and metadata so each point is comparable

Condition mismatch (different load, ε, path, solar, process state) is a leading cause of false upward or downward trends

Level II interprets trend slope, step changes, and scatter—not just the latest number—before escalating or clearing findings

Last updated: August 2026

A single survey answers a static question. Rate-of-change analysis answers a dynamic one: Is the asset stable, slowly degrading, or racing toward failure? Level II programs treat multi-inspection trending as a core analytical product—especially for assets that sit for years in mild Priority 4 bands while quietly worsening, or for unique equipment without peers.

Why ΔT Over Time Can Beat a Single Snapshot

  • Insight Example
  • Early faults grow Lug ΔT_peer: 3 °C → 6 °C → 11 °C over three annual surveys at similar load
  • Snapshot band lags risk Still “Priority 4” at 9 °C, but tripling in two years warrants earlier action
  • Post-event steps Bearing housing jumps +15 °C after a lubrication miss, then stays high
  • Process cycles hide in one visit Seasonal load changes look like “faults” without a multi-point view

Exam theme: The slope and acceleration of thermal abnormality matter. Standards-style priority bands remain mandatory for snapshot classification, but Level II adds trend intelligence when history exists.

Define quantities clearly:

  • ΔT_peer(t) = T_suspect(t) − T_similar(t) at survey t (NETA-style)
  • ΔT_base(t) = T_ROI(t) − T_ROI(baseline) under comparable conditions
  • ΔT_amb(t) = T_ROI(t) − T_ambient(t) (context; not a full substitute for peer/baseline when those exist)

Rate of change might be expressed as °C per month, per year, or per 1,000 operating hours—whatever matches the inspection interval and runtime data available.

Normalization: Making Points Comparable

Raw temperature is a function of fault resistance/friction , load/process , ambient/cooling , and measurement error . Plotting raw °C without context is how false trends are born.

  • Factor Effect on temperature Level II practice
  • Electrical load (I) Connection heating ~ I²R Record amps/%; prefer surveys at similar % rated; note if not
  • Mechanical load / RPM / process rate Bearing and drive heat rise with work Same production state when possible
  • Ambient / cooling air Shifts absolute surface T Record ambient; consider rise above ambient or peer ΔT
  • Wind / enclosure ventilation Changes convection Outdoor and enclosure notes
  • Emissivity / RAT / window τ Apparent T bias Lock methods; document changes
  • Geometry / focus / distance Spot-size errors Standardize stand-off and ROI
  • Practical normalization approaches
  • Approach When useful Caveat
  • Survey only at similar load bands (e.g., 50–70% rated) Best simple control Plant may not cooperate every visit

Peer ΔT (same survey) Cancels much common-mode ambient/load on three-phase gear Fails if all phases degrade together or loads differ

Temperature rise vs ambient Unique assets Still load-sensitive

Engineering load correction (estimate full-load T from partial load) Advanced programs Document assumptions; do not fake precision

Qualifying notes (“↑ load this visit”) Always available Not a full numeric normalize but prevents false alarm

Rule: If you cannot normalize, you must qualify . A report that says “ΔT rose 8 °C but load rose from 30% to 90%” is honest; a chart that screams “failure” without load is not.

Multi-Year Route Plotting

Predictive routes (switchgear rooms, motor lists, steam trap walks, roof grids) generate time series per asset or per ROI.

  • Plot element Good practice
  • X-axis Date or operating hours
  • Y-axis Primary metric: ΔT_peer or ΔT_base (label which!)
  • Secondary series Load %, ambient (on second axis or companion table)
  • Events Repair, outage, process change markers
  • Limits Snapshot priority thresholds as horizontal reference lines
  • Missing data Gaps shown; never interpolate a fake “smooth failure” without basis

Consistency beats camera megapixels. The same bolt head, same phase, same window, same ε method, year after year, is worth more than a new camera model with wandering ROIs.

Example multi-year table (teaching model)

  • Survey Load (% rated) Ambient (°C) T_PhaseB (°C) T_peer avg (°C) ΔT_peer (°C) Notes
  • 2023-04 60 22 38 35 3 Baseline year
  • 2024-04 58 23 42 36 6 Rising
  • 2025-04 61 21 48 37 11 Crosses into P3 band
  • 2025-10 60 24 49 37 12 Mid-cycle check; still climbing slowly

Interpretation: rate of change ~ +4 °C/year in peer ΔT, then a slower rise. Snapshot classification in 2025 is Priority 3, but the trend justified mid-cycle inspection before the next annual only survey.

False Trends from Condition Mismatch

  • Mismatch False story Correction
  • Load much higher this year “Fault doubled” Compare % load; use peer ΔT; reschedule at matched load
  • Load much lower “Repair miracle” without work Under-loaded survey hides I²R faults
  • ε set 0.95 then 0.60 on metal Huge fake ΔT Standardize surface preparation / tape
  • Different IR window or open door vs closed Path bias Same path + τ
  • Solar-loaded outdoor bus Afternoon “hotspot” Time-of-day control; shade; peer check
  • Cold start vs steady Transient rise Wait for thermal equilibrium
  • ROI moved from lug to cable jacket Different object Template ROIs
  • Different camera range/calibration status Bias Calibration program (Chapter 7)
  • Process fluid temperature change Whole machine warmer Normalize to process T or peer unit

Level II skepticism: Any large year-to-year jump without a matching work order, failure mode story, or confirmed matched conditions is suspect measurement or condition change until proven otherwise.

Reading Trend Shapes

  • Shape Typical meaning Response thinking
  • Flat low ΔT Stable healthy or stable minor anomaly Maintain interval

Linear slow rise Progressive degradation (loosening, wear, fouling) Tighten interval; plan repair before next band jump

Accelerating rise Approaching failure Escalate priority earlier than snapshot alone (Section 12.3)

Step change up Event (torque loss, lube failure, blocked cooling) Investigate cause now; do not wait for “next year”

  • Step change down after work Successful repair Re-baseline
  • High scatter Uncontrolled conditions or intermittent load Fix method before trusting slope
  • Linking Trending to Other Severity Systems
  • Domain Snapshot tool Trend add-on
  • Electrical NETA-style peer ΔT bands Same-connection ΔT_base and ΔT_peer history
  • Rotating machinery ISO 18434-style zones vs reference Bearing ROI vs baseline under same load
  • Steam traps Failed vs live patterns Repeat failures on same trap ID
  • Buildings Pattern + envelope ΔT validity Seasonal series only under valid weather

Do not force electrical Priority numbers onto unrelated domains, but do carry the shared idea: history changes urgency .

Worked Scenarios

Scenario A — Normalized rise. ΔT_peer goes 4 → 9 → 14 °C at 55–60% load each year. Real degradation; schedule repair (P3) and consider shorter interval.

Scenario B — False rise. Absolute lug T up 18 °C versus last year, but load was 25% then and 95% now; peer ΔT unchanged at 2 °C. Not a failure trend—load mismatch.

Scenario C — False improvement. Peer ΔT falls from 16 °C to 5 °C, yet load collapsed from 80% to 15%. Do not close the work order as “healed”; re-inspect at load.

Scenario D — Multi-year route. Motor DE bearing ROI rise vs ambient climbs 2 °C/year for four years with stable production. Even if still in an “advisory” ISO-style band, plan bearing service before an exponential end-stage.

Scenario E — Step event. Gearbox surface ROI jumps +20 °C between monthly routes after oil leak. Treat as event-driven investigation, not a gentle annual slope problem.

Common Traps

  • Trap Better practice
  • Charting absolute T only across seasons outdoors Use peers, rise, or matched conditions
  • Declaring trends from two points with mismatched load Need comparable conditions or heavy qualification
  • Ignoring accelerating curvature Late-stage risk
  • Changing ROI freely Kills multi-year meaning
  • Assuming “Priority 4 forever” if still < 10 °C Rising 2 → 9 °C may still need action planning
  • Summary for Recall

Rate-of-change of peer or baseline ΔT often reveals risk that a single snapshot understates. Normalize or qualify for load and ambient (and lock radiometric method). Build multi-year route plots with consistent ROIs, event markers, and clear metric definitions. Condition mismatch is the chief source of false trends—Level II challenges dramatic slopes that lack matched conditions or plausible failure physics before alarming the plant.