Setting up a condition monitoring programme

Operations · 10 min read · Updated 2026-09-01

ISO 17359 sets out the general flow for condition monitoring: audit the assets, select measurements, establish a baseline, set alert criteria, then diagnose, prognose and act. Most failed programmes skip the criticality audit and start at sensor selection.

Key takeaways at a glance

Topic	Key point
Start with criticality, not with sensors	The commonest way to waste a condition-monitoring budget is to begin at "which sensor should we buy".
The flow in practice	Step What you produce Failure if skipped Equipment and criticality audit Ranked asset list with consequence of failure Monitoring spread evenly across
Baselining is where programmes quietly fail	Trending is relative. Every alert rule compares a current reading with a reference, so the reference has to represent a healthy machine under a known operating condition.
Alert criteria need an owner before they need a threshold	Thresholds belong to the relevant standard and to your own baseline — ISO 20816 covers vibration severity evaluation, and the sensible route is to con
Competence and record-keeping	Two supporting standards carry more weight than their profile suggests.
The review loop	ISO 17359 treats condition monitoring as a cycle, not a project.
Standards referenced	Standard Scope ISO 17359 Condition monitoring and diagnostics of machines — general guidelines ISO 13379 Data interpretation and diagnostics technique

Start with criticality, not with sensors

The commonest way to waste a condition-monitoring budget is to begin at "which sensor should we buy". ISO 17359, the general guideline for condition monitoring and diagnostics of machines, begins somewhere else: with an audit of the equipment and its failure modes, and a decision about which assets justify monitoring at all.

That ordering matters because the technique is determined by the failure mode. Choose the sensor first and you have committed to detecting whatever that sensor detects, which may not be what actually takes your plant down.

The flow in practice

Step	What you produce	Failure if skipped
Equipment and criticality audit	Ranked asset list with consequence of failure	Monitoring spread evenly across assets that do not deserve it
Failure mode review	Dominant modes per critical asset	Technique cannot see the mode that actually fails
Measurement selection	Technique, measurement point, interval	Interval longer than the P-F window; findings arrive after failure
Baseline	Known-good reference per point	Existing degradation encoded as normal
Alert criteria	Thresholds and trend rules with an owner	Alarms nobody is accountable for
Diagnosis and prognosis	Fault identification and remaining-life estimate	Data with no interpretation
Action and review	Work order, then feedback on whether the call was right	The programme never improves its own accuracy

Baselining is where programmes quietly fail

Trending is relative. Every alert rule compares a current reading with a reference, so the reference has to represent a healthy machine under a known operating condition.

Two traps recur:

• Baselining a degraded asset. Fitting monitoring to an old machine captures its current fault as "normal", and the programme then reports nothing until the fault worsens considerably.
• Baselining at one duty point. On variable-load equipment a single reference produces alarms whenever the process changes. Either baseline per operating state, or record operating context with every reading so the comparison is like-for-like.

Alert criteria need an owner before they need a threshold

Thresholds belong to the relevant standard and to your own baseline — ISO 20816 covers vibration severity evaluation, and the sensible route is to consult it rather than to copy figures from an article. What no standard can give you is the organisational half.

Before any threshold is set, three things must have names attached: who receives the alert, who decides severity, and who has authority to change the maintenance schedule as a result. A programme without those three produces reports rather than avoided failures — and it is the single most common reason condition monitoring is quietly abandoned two years in.

Competence and record-keeping

Two supporting standards carry more weight than their profile suggests.

ISO 18436 defines qualification and assessment requirements for condition-monitoring personnel, by technique and by category. Data collection can be routed to trained operators; severity calls need qualified interpretation, or the organisation stops trusting the output.

ISO 14224 gives a structure for collecting and exchanging reliability and maintenance data. It matters here because without consistent failure and intervention records you cannot answer the only question that keeps a programme funded: did the findings actually change outcomes?

The review loop

ISO 17359 treats condition monitoring as a cycle, not a project. The step most often dropped is the last one: comparing what the programme predicted with what was found when the machine was opened.

That feedback is what tunes thresholds, retires techniques that never catch anything on a given asset class, and builds the internal evidence that justifies extending the programme. Skip it and the programme's accuracy is frozen at whatever the initial settings happened to be.

Standards referenced

Standard	Scope
ISO 17359	Condition monitoring and diagnostics of machines — general guidelines
ISO 13379	Data interpretation and diagnostics techniques
ISO 13374	Data processing, communication and presentation
ISO 13381	Prognostics
ISO 20816	Vibration measurement and evaluation
ISO 18436	Personnel qualification and assessment
ISO 14224	Reliability and maintenance data collection
EN 13306	Maintenance terminology

Frequently asked questions

Related guides

Software that helps