Condition Monitoring and Inspection Technologies

Introduction

Condition monitoring is the systematic observation, measurement and interpretation of equipment condition so that developing faults can be detected before they cause functional failure. It is a central element of predictive maintenance, but it is not synonymous with simply collecting sensor data. Effective condition monitoring connects a physical failure mechanism to a measurable symptom, establishes a reliable decision threshold, and converts evidence into a maintenance action.

In production systems, inspection technologies support three related decisions:

1. Is the equipment operating within an acceptable condition range?
2. If deterioration is present, what failure mode is most likely?
3. When and how should maintenance be planned without creating unnecessary risk or cost?

No single technology detects every failure mode. Vibration analysis is highly effective for rotating machinery, thermography reveals abnormal heat patterns, oil analysis identifies lubricant and internal wear conditions, ultrasound detects high-frequency emissions and leaks, motor-current analysis evaluates electromechanical faults, process-parameter monitoring observes functional performance, and non-destructive testing examines material integrity without destroying the component. Professional practice therefore uses technology selection based on equipment criticality, failure physics, accessibility, operating conditions and the consequence of failure.

Learning objectives

By the end of this chapter, the reader should be able to:

• explain the relationship between failure mechanisms, symptoms and condition indicators;
• select appropriate monitoring and inspection technologies for different equipment;
• interpret basic vibration, thermal, oil, ultrasound, motor-current and process data;
• distinguish screening, diagnosis, prognosis and acceptance inspection;
• evaluate measurement uncertainty, false alarms and limitations;
• design an integrated condition-monitoring programme; and
• assess the managerial, ethical and governance implications of data-driven maintenance.

1. The logic of condition monitoring

A failure mechanism is the physical process that produces deterioration—for example fatigue, abrasion, corrosion, looseness, insulation breakdown or lubricant degradation. A failure mode describes how the asset loses its required function, such as a bearing seizure, shaft fracture or pump inability to deliver flow. A failure symptom is the observable evidence, such as increased vibration, rising temperature or falling discharge pressure.

A useful diagnostic chain is:

Failure mechanism → physical change → measurable signal → interpretation → maintenance decision

The chain is not deterministic in every case. The same symptom may have several causes. For example, high vibration can result from imbalance, misalignment, looseness, resonance, bearing damage or hydraulic excitation. Diagnosis therefore requires operating context, historical trends and, where necessary, more than one measurement technology.

1.1 Screening, diagnosis and prognosis

Screening identifies equipment that may require attention. It should be rapid, repeatable and economical. A handheld infrared scan or overall vibration reading may be suitable.

Diagnosis identifies the probable fault and its severity. Spectrum analysis, phase measurements, oil particle morphology or ultrasonic waveform analysis may be required.

Prognosis estimates how deterioration may progress and supports the remaining useful life decision. Prognosis is inherently uncertain because load, environment, operating regime and failure propagation may change.

A condition indicator should be evaluated against a baseline, not interpreted in isolation. The baseline must represent healthy operation under comparable speed, load, temperature and process conditions. Standards and manufacturer limits can guide interpretation, but plant-specific history is often more informative.

2. Vibration analysis

2.1 Physical basis

Rotating components generate dynamic forces. These forces are transmitted through bearings, housings and foundations and can be measured using accelerometers, velocity transducers or displacement probes. The signal may be examined in the time domain, frequency domain or order domain.

Important quantities include:

• Displacement: useful at low frequencies and for shaft motion;
• Velocity: commonly useful for machine-severity evaluation over a broad frequency range;
• Acceleration: sensitive to high-frequency impacts such as rolling-element bearing damage;
• Frequency: the rate at which a periodic event occurs; and
• Phase: the angular relationship between vibration signals or a vibration signal and a shaft reference.

For sinusoidal motion, displacement, velocity and acceleration are related by frequency. If displacement is represented as x(t)=X sin(2πft), then the peak velocity and acceleration are:

Here, X is peak displacement in metres, f is frequency in hertz, v_peak is peak velocity in metres per second, and a_peak is peak acceleration in metres per second squared. These relationships assume sinusoidal motion and consistent units. They should not be applied directly to complex multi-frequency signals without considering each frequency component.

For a set of sampled values, root-mean-square vibration is commonly calculated as:

where x_i is the ith measured value and N is the number of samples. RMS expresses the effective magnitude of a varying signal. Peak and crest factor provide complementary information: a high crest factor may indicate impacts even when overall RMS remains moderate.

2.2 Frequency-domain interpretation

A fast Fourier transform converts a time record into frequency components. Typical diagnostic relationships include:

• 1× rotational frequency: imbalance, although other faults can also produce it;
• 2× rotational frequency and axial vibration: possible angular misalignment;
• multiple harmonics of rotational speed: mechanical looseness or distortion;
• gear-mesh frequency: gear tooth problems, eccentricity or loading issues;
• bearing characteristic frequencies: localised rolling-element or race defects; and
• sidebands: modulation caused by load variation, gear faults or electrical phenomena.

For a rolling-element bearing, characteristic frequencies depend on rolling-element diameter, pitch diameter, contact angle, number of rolling elements and shaft speed. Because geometry conventions and slip assumptions vary, frequencies should be calculated using the bearing manufacturer’s data or a validated diagnostic database rather than an unexplained generic formula.

2.3 Measurement practice and limitations

Sensor location, mounting stiffness, orientation and sampling frequency strongly affect results. Magnetic mounting is convenient but may attenuate high-frequency content; stud mounting generally provides better repeatability. Measurements should be taken at consistent points and operating states.

Vibration analysis is particularly effective for motors, pumps, fans, compressors, gearboxes, turbines and centrifuges. It is less decisive for slowly rotating equipment, machines with changing speed, structurally isolated components and faults that produce little dynamic energy. Resonance can amplify a small excitation, while poor sensor placement can conceal a serious fault.

3. Infrared thermography

Thermography measures infrared radiation and displays apparent surface temperature. It is a non-contact technique valuable for electrical connections, switchgear, bearings, couplings, furnaces, insulation, steam systems and process equipment.

3.1 Interpretation principles

A thermal anomaly is meaningful only when compared with a suitable reference. Useful comparisons include:

• the same component on similar phases or parallel machines;
• the same location at different times;
• a loaded component versus its expected load-temperature relationship; and
• the temperature difference between a suspect component and a healthy reference.

Temperature readings depend on emissivity, reflected radiation, viewing angle, atmospheric attenuation, distance and load. Shiny metals can produce misleading apparent temperatures because they reflect surrounding radiation. Correct emissivity settings, suitable viewing angles and, where possible, a high-emissivity reference patch improve reliability.

For electrical systems, a loose or resistive connection often produces localised heating. If current is I and resistance is R, the electrical power dissipated as heat is:

where P is in watts, I in amperes and R in ohms. The relationship assumes resistive heating and helps explain why a small increase in contact resistance can become significant at high current. It does not by itself predict surface temperature, which also depends on heat transfer, enclosure geometry and ambient conditions.

Thermography detects temperature effects, not necessarily the underlying cause. A hot bearing may reflect inadequate lubrication, overload, misalignment, contamination or imminent seizure. A cold region may indicate blocked flow, failed heating, poor electrical continuity or simply a change in emissivity.

4. Oil and wear-debris analysis

Lubricant is both a functional medium and a source of diagnostic evidence. Laboratory or online analysis can assess viscosity, water, oxidation, additive depletion, contamination, ferrous density and particle morphology.

4.1 Main analytical categories

Fluid properties: Viscosity changes may indicate wrong oil, contamination, thermal degradation or shear. Water can reduce film strength and promote corrosion. Acid number and oxidation indicators reveal chemical deterioration.

Contaminants: Particles, dirt, fibres, coolant, water and process chemicals can accelerate wear or alter lubricant performance. Particle-count coding systems should be reported using the relevant standard and particle-size bands; codes must not be compared without confirming the same standard and calibration basis.

Wear debris: Ferrography, microscopy and elemental analysis can distinguish cutting wear, sliding wear, fatigue particles, corrosion products and abnormal large particles. Shape, size, material and concentration matter more than particle count alone.

A simplified wear-rate indicator may be expressed as:

where ∆M is the change in measured wear mass or a validated wear proxy and ∆t is the sampling interval. This is a trend metric, not a universal physical wear law. It is meaningful only when sampling location, volume, operating hours and analytical method remain comparable.

4.2 Sampling and decision quality

A poor sample can invalidate excellent laboratory analysis. Samples should be taken from a live, well-mixed zone upstream of filters where possible, using clean containers and a consistent procedure. Sampling after an oil change may produce misleading results. The sample record should include machine identity, lubricant grade, operating hours, oil temperature, load and recent maintenance.

Oil analysis is especially useful for gearboxes, hydraulic systems, engines, compressors, turbines and circulating lubrication systems. It is less useful where lubricant volume is very small, contamination is not well mixed or the component has no oil circuit.

5. Ultrasound inspection

Ultrasound instruments detect high-frequency acoustic energy, usually by converting it into an audible signal and a quantitative indication. They are useful for compressed-air and steam leaks, vacuum leaks, electrical arcing, lubrication verification and early bearing impacts.

Leak detection is attractive because turbulent flow through a small orifice generates ultrasound. However, sensitivity depends on background noise, distance, pressure differential, instrument frequency response and accessibility. A leak survey should distinguish an audible indication from a quantified loss estimate; estimating financial loss requires validated flow relationships and pressure data.

For lubrication, ultrasound can help identify changes in bearing friction. Grease should not be added merely because a high reading occurs: a high reading may indicate over-lubrication, damaged surfaces or contamination. Lubrication should follow the correct grease specification, quantity and relubrication interval.

Ultrasound is also valuable for detecting corona, tracking and arcing in electrical equipment, but safety boundaries and arc-flash procedures remain mandatory. The instrument does not make an energised inspection safe.

6. Motor-current and electrical signature analysis

Motor-current analysis evaluates current magnitude, phase balance, harmonics and sidebands. It can identify electrical faults such as broken rotor bars, stator winding imbalance, phase problems and eccentricity. It may also reveal mechanically induced load variation.

The technique is attractive because current can be measured from a control panel without direct access to the motor. Nevertheless, supply quality, variable-frequency drives, load variation and power-electronic switching can complicate interpretation. A current anomaly does not automatically identify a motor fault.

For a balanced three-phase system, approximate active power is:

where V_L is line-to-line RMS voltage, I_L is line RMS current and cosφ is power factor. The formula assumes a balanced sinusoidal three-phase system. With significant harmonics or an unbalanced supply, true-power instruments and phase-specific analysis are preferable.

Motor-current analysis should be integrated with vibration, insulation testing, load information and electrical-history data. It is most useful for inaccessible motors, critical drives and applications where process interruption is costly.

7. Process-parameter monitoring

Process monitoring observes whether equipment continues to deliver its required function. Typical variables include flow, pressure, differential pressure, temperature, speed, power, level, quality and energy consumption.

A pump may be mechanically healthy yet fail to meet production requirements because of a blocked suction line, changed system resistance or incorrect valve position. Conversely, a machine may show acceptable process output while developing a bearing fault. Process parameters therefore provide functional evidence but may not provide early mechanical diagnosis.

A pump or fan operating point is determined by the intersection of the equipment performance curve and the system curve. Monitoring flow, pressure and power together is more informative than observing any one variable. Trends should be normalised for product, speed, ambient conditions and control mode.

Statistical process-control methods can support monitoring. For a stable process with approximately normal observations, a standardised score may be written as:

where x is the observation, μ is the baseline mean and σ is the baseline standard deviation. This interpretation assumes a stable baseline and approximately constant variance. A high z-score is an investigation signal, not proof of equipment failure.

8. Non-destructive testing

Non-destructive testing (NDT) evaluates material or component integrity without rendering the item unusable. Method selection depends on defect type, orientation, material, geometry, access and required sensitivity.

NDT is generally an inspection technology rather than continuous condition monitoring. It is essential for pressure vessels, welds, shafts, rails, aircraft structures, pipelines and critical fabricated components. Inspection results require qualified personnel, calibrated equipment, documented procedures and acceptance criteria from applicable codes or engineering specifications.

The distinction between an indication, a discontinuity and a rejectable defect is important. An indication is a response from the test; it becomes a relevant discontinuity after evaluation, and it is rejectable only if it exceeds an applicable acceptance criterion. This prevents both unsafe acceptance and unnecessary repair.

9. Integrating technologies into a monitoring strategy

Technology selection should begin with a failure-mode analysis such as FMEA or reliability-centred maintenance. For each failure mode, identify the physical degradation, the earliest detectable symptom, the detection interval, the consequence and the feasible intervention.

A mature programme contains four layers:

1. Operator care: visual, auditory, smell, leakage and process checks.
2. Routine inspection: handheld vibration, infrared, ultrasound and basic electrical checks.
3. Specialist diagnosis: spectrum analysis, oil laboratory work, advanced NDT and motor testing.
4. Online monitoring: permanently installed sensors and automated alarms for high-criticality assets.

Alarm design should use at least two levels: an alert that prompts investigation and an action alarm that requires a defined response. Limits should account for measurement uncertainty, operating state and consequence. Alarm flooding is a management failure: excessive alarms reduce attention and encourage overrides.

Data quality is governed by repeatability, reproducibility, calibration, sensor health, metadata and traceability. A trend without consistent sampling conditions can be less valuable than a smaller but reliable dataset.

10. Illustrative case: centrifugal process-water pump

Background

A process plant operates a critical centrifugal pump continuously. The pump has rolling-element bearings, a flexible coupling and a mechanical seal. An online system records vibration and discharge pressure; monthly routes collect infrared, ultrasound and oil samples.

Problem

Over six weeks, overall radial vibration rises from 2.4 to 4.1 mm/s RMS. Discharge pressure falls slightly, bearing temperature increases by 7 °C, and an oil sample shows increasing fine ferrous debris. The operator also reports intermittent ultrasonic noise near the suction line.

Analysis

The vibration trend alone does not distinguish imbalance, misalignment, bearing damage or cavitation. Spectrum analysis shows increased energy near a bearing-related frequency and a broadband high-frequency rise. Thermography confirms that the drive-end bearing is warmer than the non-drive-end bearing. Oil microscopy finds fatigue-like ferrous particles rather than only harmless environmental dust. Suction pressure is lower than normal, and ultrasound is consistent with turbulent flow or incipient cavitation.

The evidence suggests two interacting conditions: deteriorating drive-end bearing condition and inadequate suction conditions. Continued operation risks bearing failure and seal damage, but immediate replacement without checking suction conditions could lead to recurrence.

Approach and solution

Maintenance schedules a controlled outage. The bearing, lubricant, coupling alignment and foundation are inspected. The suction strainer is partially blocked and the bearing has evidence of surface fatigue. The strainer is cleaned, the bearing is replaced, alignment is verified using a laser system, and the correct lubricant quantity is applied. Baseline measurements are then repeated at the normal operating point.

Outcome

After restart, vibration falls to 1.9 mm/s RMS, bearing temperature returns near its historical value, and oil debris declines over subsequent samples. Discharge pressure is restored. The monitoring plan is revised to include a suction differential-pressure indicator and a shorter initial follow-up interval after bearing replacement.

Learning points

• No single measurement established the diagnosis.
• Trend, spectrum, temperature, oil and process evidence were mutually reinforcing.
• The apparent bearing problem had a contributing hydraulic condition.
• Corrective action addressed both the failed component and the causal environment.
• Post-maintenance baselining is essential for verifying the result.

11. Limitations, governance and contemporary developments

Condition monitoring can generate false positives, missed detections and ambiguous evidence. Common causes include poor sampling, inappropriate alarm limits, sensor drift, changing operating conditions, untrained interpretation and confirmation bias. The cost of an inspection programme should be compared with avoided failure cost, safety benefit, production protection and decision quality—not merely the number of sensors installed.

Safety is non-negotiable. Thermography and electrical ultrasound may involve energised equipment; vibration routes may require access near rotating parts; NDT may involve radiation, chemicals, high pressure or confined spaces. Procedures must comply with applicable occupational-safety, electrical-safety, pressure-equipment and NDT qualification requirements.

Digital systems increasingly combine wireless sensors, edge computing, cloud historians, digital twins and machine-learning models. Artificial intelligence can classify spectra, detect multivariate anomalies and prioritise work orders, but it does not eliminate engineering responsibility. Models may fail when the operating regime changes, labels are biased, sensors drift or rare failures are underrepresented. Every automated recommendation should retain data provenance, confidence information, human review and an auditable decision record.

A responsible governance framework should define data ownership, cybersecurity, access control, retention, model validation, alarm changes and override authority. Maintenance data can affect employee performance assessment and contractor accountability; therefore, it should not be used carelessly or without transparent rules.

Best-practice principles

1. Start with criticality and failure modes, not available technology.
2. Measure under known and repeatable operating conditions.
3. Establish a healthy baseline and document metadata.
4. Use complementary technologies for ambiguous or high-consequence faults.
5. Separate screening from diagnosis and diagnosis from prognosis.
6. Set alarm limits with engineering rationale and review them after experience.
7. Link every alarm to an owner, response time and work process.
8. Verify corrective action through post-maintenance measurements.
9. Maintain calibration, competency and traceability.
10. Treat automated analytics as decision support, not unquestionable authority.

Chapter Summary

Condition monitoring converts physical evidence of deterioration into planned maintenance decisions. Vibration analysis is central for rotating machinery and uses time, frequency and phase information to identify dynamic faults. Thermography detects abnormal heat but requires careful control of emissivity, load and reference conditions. Oil and wear-debris analysis reveals lubricant condition, contamination and internal wear when sampling is representative. Ultrasound supports leak detection, lubrication checks and high-frequency fault detection. Motor-current analysis evaluates electrical and electromechanical conditions, while process-parameter monitoring determines whether equipment is fulfilling its functional requirement. NDT examines material integrity without destroying components and must be interpreted against qualified acceptance criteria.

Effective programmes integrate technologies through failure-mode analysis, criticality ranking, reliable baselines, controlled alarms and defined responses. Measurement uncertainty, operating context, safety, cybersecurity and human judgement remain essential. Digital and AI-enabled systems can improve detection and prioritisation, but they require validation, transparency and governance.

Key Terms

Condition monitoring: Systematic measurement and interpretation of equipment condition.

Failure mechanism: Physical process that causes deterioration.

Failure mode: Specific manner in which an asset loses its required function.

Condition indicator: Measured variable used to represent equipment health.

Baseline: Reference condition established during known healthy operation.

Vibration RMS: Effective magnitude of a varying vibration signal.

Crest factor: Ratio of signal peak to RMS value, often sensitive to impacts.

Frequency spectrum: Distribution of signal energy across frequencies.

Thermography: Infrared measurement and imaging of apparent surface temperature.

Emissivity: Efficiency with which a surface emits infrared radiation.

Wear debris: Material released from a component through wear.

Ferrography: Analysis of ferrous particles separated from lubricant.

Ultrasound inspection: Detection and interpretation of high-frequency acoustic energy.

Motor-current signature analysis: Evaluation of motor-current patterns for fault evidence.

Process parameter: Variable such as flow, pressure, temperature or power that represents function.

Non-destructive testing: Examination of material or component integrity without damaging usability.

Indication: Response produced by an inspection method.

Alarm threshold: Defined value or condition that initiates investigation or action.

Remaining useful life: Estimated time or usage before functional failure under stated assumptions.

Review Questions

Short Answer Questions

1. Distinguish screening, diagnosis and prognosis in condition monitoring.
2. Why is a machine-specific baseline important when interpreting vibration data?
3. State two factors that can make an infrared temperature reading unreliable.
4. What information can wear-debris analysis provide that viscosity measurement may not?
5. Why must an NDT indication not automatically be classified as a rejectable defect?

Long Answer Questions

1. Explain the physical basis, measurement methods and limitations of vibration analysis for rotating machinery.
2. Compare thermography, oil analysis and ultrasound for detecting lubrication-related problems.
3. Discuss how process-parameter monitoring complements equipment-focused condition monitoring.
4. Design an integrated monitoring strategy for a critical gearbox and justify each selected technology.
5. Explain the governance, safety and competency requirements for an AI-enabled condition-monitoring programme.

Analytical / Application Questions

1. A motor-driven pump shows rising vibration, a warmer drive-end bearing, falling discharge pressure and increasing ferrous debris. Develop a fault hypothesis, identify alternative explanations and propose an inspection sequence.
2. A thermal image shows one electrical terminal 25 °C hotter than the corresponding terminals. Explain what additional information is required before deciding whether to shut down the circuit.
3. A predictive model generates frequent alarms after production changes. Analyse possible causes and propose a validation and alarm-governance plan.

Multiple Choice Questions with Answers

Question 1

Which technology is generally most directly suited to identifying imbalance and misalignment in rotating machinery?

A. Vibration analysis

B. Liquid penetrant testing

C. Oil acidity testing

D. Visual inspection only

Correct Answer: A. Vibration analysis

Question 2

For sinusoidal displacement with peak amplitude X and frequency f, which expression gives peak acceleration?

A. X ÷ 2πf

B. 2πfX

C. (2πf)^2 X

D. X + 2πf

Correct Answer: C. (2πf)^2 X

Question 3

A major source of error when measuring the temperature of a shiny metal surface with infrared thermography is:

A. Excessive lubricant viscosity

B. Reflected infrared radiation and incorrect emissivity

C. Bearing characteristic frequency

D. Insufficient oil sample volume

Correct Answer: B. Reflected infrared radiation and incorrect emissivity

Question 4

Which finding is most directly associated with wear-debris analysis?

A. Particle morphology and material composition

B. Shaft rotational phase only

C. Surface emissivity

D. Electrical power factor only

Correct Answer: A. Particle morphology and material composition

Question 5

Ultrasound is particularly effective for detecting compressed-air leaks because leaks:

A. Always reduce lubricant viscosity

B. Generate high-frequency turbulent acoustic energy

C. Create magnetic fields in ferromagnetic parts

D. Produce bearing characteristic frequencies

Correct Answer: B. Generate high-frequency turbulent acoustic energy

Question 6

The balanced three-phase active-power relationship P = √3 V_L I_L cosφ assumes:

A. A balanced sinusoidal three-phase system

B. A single-phase resistive circuit only

C. Zero current in every phase

D. A system with no voltage

Correct Answer: A. A balanced sinusoidal three-phase system

Question 7

Which NDT method is generally suitable for surface-breaking cracks in non-porous materials?

A. Liquid penetrant testing

B. Radiographic testing only

C. Oil particle counting

D. Motor-current analysis

Correct Answer: A. Liquid penetrant testing

Question 8

If a stable baseline has μ = 100 and σ = 5, what is z when x = 115?

A. 1

B. 2

C. 3

D. 5

Correct Answer: C. 3

Question 9

Why should a condition-monitoring alarm not be interpreted without operating context?

A. All alarms are necessarily false

B. Signals can change with load, speed and process state

C. Sensors never require calibration

D. Failure modes are independent of operating conditions

Correct Answer: B. Signals can change with load, speed and process state

Question 10

In the illustrative pump case, which combination best supported the diagnosis?

A. Only a visual inspection of the coupling

B. Vibration, temperature, oil debris and process evidence

C. Only the discharge-pressure reading

D. Only the ultrasonic instrument reading

Correct Answer: B. Vibration, temperature, oil debris and process evidence