Condition Monitoring of Industrial Machines

Condition monitoring is a discipline within industrial maintenance that tracks the physical and operational state of machinery in real time or at scheduled intervals to detect deterioration before it causes failure. This page covers the definition, technical mechanisms, common deployment scenarios, and decision frameworks that govern when and how condition monitoring is applied in US industrial environments. The subject matters because unplanned equipment downtime in manufacturing carries costs that dwarf the investment required to detect problems early, and because the practice forms the foundation of predictive maintenance for automated machines.

Definition and scope

Condition monitoring refers to the continuous or periodic measurement of parameters that reflect the health of a machine or component — vibration, temperature, acoustic emission, oil particle count, current draw, and pressure among them. The International Organization for Standardization addresses this practice primarily through ISO 13373 (rotating machinery vibration) and ISO 17359 (general condition monitoring guidelines), both of which define it as a subset of the broader maintenance management system rather than a standalone technology.

Scope encompasses stationary plant assets (pumps, compressors, gearboxes, motors), industrial robots, CNC machine tools, conveyors, and any driven equipment where degradation produces a measurable precursor signal. The discipline explicitly excludes scheduled-interval replacement (a time-based maintenance approach) and corrective repair after failure. Its operational boundary is the detection-to-decision window: the elapsed time between a detectable anomaly and the point at which machinery damage becomes irreversible or safety-critical.

Condition monitoring is distinct from simple alarm systems. A high-temperature cutout reacts to a threshold crossing; condition monitoring tracks the trend toward that threshold over time, enabling intervention days or weeks before the cutout would trigger.

How it works

The condition monitoring process follows a structured sequence regardless of which sensing technology is applied:

  1. Parameter selection — Engineers identify the failure modes relevant to the asset class (e.g., bearing raceway fatigue, impeller erosion, insulation breakdown) and select the physical signal that manifests earliest for each mode.
  2. Baseline establishment — The asset is measured under known-good conditions to produce a reference spectrum, temperature profile, or particle count. This baseline must be collected at comparable load and speed each time to remain valid.
  3. Periodic or continuous data acquisition — Sensors transmit data to a local data concentrator or to a plant historian. The IIoT infrastructure and SCADA layer typically serve as the transport and storage backbone.
  4. Signal processing and feature extraction — Raw waveforms are transformed (Fast Fourier Transform for vibration, envelope detection for acoustic emission) to isolate fault-related frequency components.
  5. Comparison against limits — Measured features are compared against baseline values, ISO severity zones (ISO 10816 / ISO 20816 for vibration), or statistically derived control limits.
  6. Diagnosis and prognosis — Detected deviations are classified by fault type and severity. Prognosis estimates remaining useful life based on degradation rate, informing maintenance scheduling windows.
  7. Maintenance action and feedback — Findings drive a work order. Post-repair measurement confirms return to baseline and refines future fault models.

The accuracy of steps 4 through 6 improves substantially when machine learning algorithms are applied to large historical datasets, particularly for distinguishing normal process variation from genuine degradation signals.

Sensing technologies compared: online vs. offline

Online (continuous) monitoring uses permanently installed sensors wired or wirelessly connected to always-on data systems. It captures transient events and enables sub-second response but carries higher per-point hardware cost and demands robust data infrastructure.

Offline (periodic) monitoring uses portable instruments — handheld vibration analyzers, infrared cameras, ultrasonic probes — applied by technicians on a route-based schedule, typically monthly or quarterly. Capital cost per asset is lower, but intermittent measurement risks missing rapidly developing faults in the interval between rounds.

The crossover point favoring online monitoring is generally assets with a fault-development time shorter than the offline measurement interval, or assets whose failure mode carries a safety or production-loss consequence that justifies the infrastructure investment.

Common scenarios

Rotating machinery in process industries — Centrifugal pumps, fans, and compressors in chemical, refining, and food processing plants are the most common targets. Vibration and temperature are the primary parameters; bearing defect frequencies and imbalance signatures account for the majority of detectable faults.

Electric motor health — Motor current signature analysis (MCSA) detects rotor bar breakage, air-gap eccentricity, and winding degradation through current spectrum analysis at the motor drive panel without physical contact with rotating parts. This technique is particularly applicable to motors served by servo drives and variable-frequency drives.

Gearbox wear tracking — Oil debris monitoring and acoustic emission sensing detect early-stage micropitting and spalling. Oil particle counters quantify wear debris in parts per million, with ISO 4406 cleanliness codes providing the classification framework.

Aerospace and defense component monitoring — High-cycle fatigue in structural components is tracked through strain gauges and acoustic emission sensors per guidelines from the Federal Aviation Administration (FAA) for certified aircraft systems.

Collaborative and industrial robot joint monitoring — Torque ripple and joint temperature trends in robot axes signal degradation in harmonic drives or encoder systems, enabling predictive replacement before positional accuracy degrades below tolerance. This connects directly to maintenance planning for collaborative robots.

Decision boundaries

Selecting and scoping a condition monitoring program requires resolving four distinct boundary questions:

1. Which assets qualify?
Assets are prioritized using a criticality ranking that weighs production impact (lost output per hour of downtime), safety exposure (injury or environmental consequence), redundancy (whether a standby unit exists), and repair lead time. Only assets above a defined criticality threshold typically justify continuous online monitoring; lower-criticality assets default to periodic routes or run-to-failure.

2. Which parameters are actionable?
A parameter must produce a detectable signal advance of the failure event — ideally at least 2 weeks before the P-F interval (the point-to-failure interval defined in reliability-centered maintenance methodology). If a failure mode produces no measurable precursor, condition monitoring adds no value for that mode; a time-based replacement interval is appropriate instead.

3. At what severity level does maintenance trigger?
ISO 20816-3 defines four vibration severity zones for industrial machines rated above 15 kW. Zone A represents new machinery in good condition; Zone D indicates levels likely to cause damage in short operating periods and requires immediate action. Maintenance planning should define which zone crossing triggers a work order versus an urgent shutdown, and this threshold should be documented in the equipment maintenance plan.

4. How does condition monitoring integrate with digital twin and automation systems?
When condition monitoring feeds a digital twin model, the twin can simulate remaining component life under projected load profiles, refining the prognosis beyond what raw sensor data alone supports. Integration also requires cybersecurity consideration (machine automation cybersecurity) because condition monitoring data flows connect OT sensor networks to IT analytics platforms.

The boundary between condition monitoring and full predictive maintenance programs lies in the prognosis step: monitoring detects and classifies; predictive maintenance additionally schedules, and may automatically trigger work orders through a CMMS integration. Organizations implementing both disciplines must define ownership of the alert-to-action workflow explicitly, or detection findings accumulate without driving maintenance action.

References

Explore This Site

Regulations & Safety Regulatory References Tools & Calculators Website Performance Impact Calculator