Preventive vs Condition-Based Maintenance in Buildings: Choosing per Asset Class

Decide at failure-mode level, not equipment name alone. A chiller may need time-based statutory inspection, usage-based consumables, vibration or oil condition monitoring, and run-to-failure treatment for a non-critical indicator lamp. Score consequence, detectability, warning time, failure pattern, monitoring cost, intervention lead time, redundancy and spare availability. Condition-based maintenance is justified only when a measurable change appears early enough to plan work and the organisation can respond. Otherwise sensors create alerts without reducing failure. Review the policy after every material failure and annually against work history, condition, downtime and total cost.

Age is evidence, but not a verdict

An asset installed in 1998 is not automatically in worse condition than one installed in 2018. The older asset may have operated below rating in a dry, accessible site with disciplined maintenance. The newer one may sit in a saline coastal environment, cycle heavily, suffer poor power quality or have been installed with a latent defect.

Replacing the oldest first is attractive because age is easy to retrieve and explain. It is also a weak capital rule. It confuses elapsed time with failure probability, ignores consequence, and directs money towards assets whose condition may still be acceptable while a younger high-consequence asset deteriorates unseen.

This matters in buildings because the infrastructure decision is being pulled in two directions. Existing portfolios require rehabilitation, while demand and system configuration continue to change. Nama Water Services' 2024 annual report describes planned investment of OMR 128.5 million in extensive portfolio rehabilitation aligned with non-revenue-water reduction, alongside major transmission and storage investment. Nama Power and Water Procurement's 2025–2031 statement projects Main Interconnected System peak demand rising from 7,503 MW in 2024 to 12,198 MW in 2031. A replacement programme therefore cannot be a like-for-like retirement list. It must decide what to retain, what to restore, what to uprate and what no longer fits the future portfolio.

The Authority for Public Services Regulation makes the capital discipline explicit. Its technical regulation overview says the Authority reviews capital programmes through price-control allowances, applies pre-investment appraisal, and undertakes ex-ante and ex-post reviews to drive technical and procurement efficiency. “Old” is not an investment case. Evidence of risk and value is.

The replacement decision in one model

Use four linked measures, not one composite score assembled in a spreadsheet and never challenged:

1. Condition: how degraded is the asset now?
2. Consequence: what happens if it fails at the wrong time?
3. Confidence: how reliable and current is the evidence behind the first two answers?
4. Intervention value: which option reduces whole-life risk at the lowest justified cost?

A practical priority statement is:

Priority = current risk + near-term risk growth + evidence uncertainty, adjusted for intervention value and deliverability.

Current risk is normally expressed as probability of failure multiplied by consequence. But do not let the multiplication hide the inputs. Two assets may both score 16: one from moderate likelihood and catastrophic consequence, the other from high likelihood and moderate consequence. They require different controls while capital is being approved.

Step 1: Define the decision unit

The asset register determines whether condition data can support a replacement programme. “Muscat water portfolio” and “33 kV feeder” are too broad; every defect becomes an attachment against an undifferentiated parent. The opposite extreme—registering every nut and gasket—creates records that nobody maintains.

Choose the lowest unit at which one of these decisions can be made independently:

• replace or refurbish;
• isolate without replacing its parent;
• assign a distinct failure consequence;
• inspect using a distinct method;
• identify cost and work history.

For a water portfolio, that may be a pipe segment between nodes, a valve, pump, motor, reservoir cell or household connection cohort. For electricity, it may be a transformer, circuit breaker, cable section, overhead-line span group, protection relay or battery bank. Keep functional location separate from the asset itself so history survives a move or reconfiguration.

Every priority record needs a stable asset identifier, parent, location, asset class, manufacturer and model where relevant, commissioning date or age band, duty, rating, material, environment and criticality. Unknown values must remain visibly unknown. A guessed installation year is not the same evidence as a commissioning record.

Step 2: Build condition indices by asset class

An asset health index is useful only when its ingredients are physically meaningful. One generic inspection checklist for pumps, pipelines and transformers produces a universal number with no universal meaning.

Asset class	Useful condition evidence	Weak proxy to avoid using alone
Buried water main	Burst frequency by length, wall-thickness or coupon data, pressure transients, leakage, soil and groundwater exposure, repair observations	Installation year
Pump	Vibration trend, bearing temperature, motor current, flow against head, seal leakage, efficiency, oil condition	Number of work orders
Power transformer	Dissolved gas analysis, oil quality, moisture, bushing condition, thermography, load and fault history	Calendar age
Switchgear or breaker	Operation count, timing and contact resistance, partial discharge, mechanism condition, fault duty, obsolescence	Visual grade alone
Underground cable	Partial-discharge or diagnostic tests, joint failures, loading, route conditions, sheath integrity	Feeder age
Overhead line	Pole or tower condition, insulator contamination, thermography, conductor and joint condition, exposure history	Patrol completion percentage

Score each input against an asset-class rule with units, thresholds and a source. Retain the raw measurement. If 78°C becomes “Condition 4”, the system must preserve 78°C, the measuring instrument, the load at the time and the threshold version. Otherwise the score cannot be audited or recalculated when engineering rules change.

Avoid averaging away a critical defect. If a transformer has acceptable paint, foundation and general appearance but a diagnostic result indicates an active internal fault, an arithmetic mean should not return “fair”. Use override rules for red conditions, statutory defects and safety-critical protection failures.

Step 3: Separate condition from consequence

Condition estimates likelihood. Criticality describes consequence. Combining them during inspection encourages surveyors to grade prestigious or visible assets more harshly even when their physical states are identical.

Assess consequence across named dimensions:

• safety: credible harm to workers or the public;
• service: customers interrupted, water volume or electrical load affected, expected duration and redundancy;
• water quality or environmental impact: contamination, uncontrolled discharge, oil release or ecological impact;
• financial impact: repair, emergency procurement, lost water or energy, penalties and collateral damage;
• strategic impact: hospitals, airports, industrial zones, desalination, pumping or national infrastructure;
• reputation and regulatory impact: reportable events, repeat interruptions and commitments to customers.

Use the portfolio model, not asset size alone. A small valve that prevents isolation may turn a local pipe repair into a wide shutdown. A modest protection battery can defeat the operation of high-value switchgear. Conversely, a large transformer with firm N-1 redundancy and a proven transfer arrangement may present less immediate service risk than its replacement cost suggests.

Consequence also changes by season and operating state. Electricity risk during summer peak is different from winter. A water pumping asset can become more critical when storage is constrained or an alternative source is unavailable. Store the assumptions and effective dates behind the criticality result.

Step 4: Make confidence visible

Missing data must not masquerade as good condition. If an asset has no recent inspection, no linked failures and no sensor feed, a blank health record does not mean healthy. It means uncertain.

Assign evidence confidence separately from health:

Confidence	Evidence state	Decision response
High	Recent class-appropriate test, traceable instrument and complete history	Use in the investment model
Medium	Recent visual inspection or partial tests; some history gaps	Plan targeted confirmation
Low	Old, inferred, migrated or unverified data	Inspect before committing replacement, unless consequence demands precautionary action
Unknown	Asset identity, location or condition cannot be confirmed	Resolve data and physical identity urgently

This prevents two bad outcomes: replacing assets merely because their records are poor, and leaving high-consequence assets unexamined because no failure has been recorded. For a high-consequence, low-confidence asset, the first funded intervention may be a diagnostic test, excavation, protection test or planned outage—not a new asset.

Confidence belongs at field level as well as asset level. A verified serial number can coexist with an inferred commissioning date and an outdated condition grade.

Step 5: Convert health into probability of failure

A health score ranks degradation; it does not automatically predict failure. Calibrate it against the utility's own event history by asset class, failure mode and operating environment.

Start simply:

• define condition bands with engineering meaning;
• count service-affecting and non-service-affecting failures consistently;
• calculate exposure, such as asset-years, circuit-kilometres or pipe-kilometres;
• compare observed failure rates across health bands;
• review whether the bands separate risk in practice.

Do not import a vendor's global failure curve without checking local exposure. Heat, salinity, coastal contamination, sand, loading patterns, pressure management, installation practice and maintenance regimes alter degradation. Equally, do not claim a statistically precise probability where the population is small. A transparent ordinal likelihood supported by engineering judgement is better than a decimal with false precision.

Work-order closure quality determines whether calibration is possible. Failure event, failed component, failure mode, cause where known, consequence, downtime and action must be structured and linked to the correct asset. “Repaired” is not reliability data.

Step 6: Compare interventions, not just assets

The highest-risk asset is not automatically the next replacement project. Test plausible options:

Option	Suitable when	Evidence required
Continue and monitor	Condition is stable, consequence tolerable and detection lead time adequate	Inspection interval and trigger limits
Maintain or refurbish	A known failure mechanism can be arrested economically	Scope, restored performance and expected life extension
Add redundancy or protection	Consequence can be reduced more efficiently than likelihood	Network study and tested operating procedure
Replace like for like	Duty remains valid and replacement reduces whole-life risk	Demand, capacity and lifecycle-cost case
Replace and uprate or redesign	Future demand, resilience or standards make like-for-like unsuitable	Planning scenario and option appraisal
Run to failure	Failure is safe, local, detectable and cheaper than planned intervention	Spares, response plan and consequence acceptance
Retire or rationalise	The asset's function is redundant after portfolio change	Approved decommissioning and data updates

Calculate risk reduction per rial, not just project cost. Include planned outage cost, temporary supply, customer impact, procurement lead time, land and access, design, construction risk, residual value, energy or water efficiency, maintenance savings and disposal.

For buried water assets, a cohort or zone intervention may outperform replacing individual segments with the highest burst count. Pressure management, district metering, leak detection, valve restoration and targeted connection replacement can reduce loss and consequence while building evidence for later mains rehabilitation. Nama Water Services' 2024 plan explicitly links portfolio rehabilitation to non-revenue-water reduction; the business case should show that linkage at project level rather than assume every old-pipe replacement produces the same benefit.

For electricity portfolios, coordinate replacement with load growth, protection changes, renewable integration, interconnection and planned outages. Installing today's rating for yesterday's portfolio may lock in the next constraint.

Step 7: Build a funded, deliverable portfolio

Once projects are ranked, apply delivery constraints openly: outage windows, long-lead equipment, engineering capacity, contractor availability, land access, permits, customer coordination and yearly allowances. Do not quietly reorder the list in meetings and leave the model showing something else.

Use three horizons:

1. Committed: justified, designed and deliverable in the current control period or budget year.
2. Develop: likely investment, but evidence, design or approvals are incomplete.
3. Monitor: below intervention threshold, with a defined inspection or trigger.

Every deferred high-risk asset needs an interim control and named risk owner. Examples include enhanced monitoring, operating restrictions, mobile spares, revised switching plans, leak patrols or temporary redundancy. “Deferred due to budget” is not a treatment.

Track the forecast risk trajectory. A project ranked twentieth today may become urgent faster than the ten above it. Remaining-life ranges and next inspection dates are more useful than a single replacement year presented as certainty.

The data model behind a defensible decision

At minimum, preserve this chain:

Asset → condition observation → health rule and version → failure likelihood → consequence model → risk → intervention options → approved project → completed work → post-investment result.

That lineage answers the questions capital reviewers will ask: Which observation changed the priority? Was the inspection current? Which failure mode does the project address? What risk reduction was forecast? Did the project deliver it?

Keep source records, not just dashboard scores. Geospatial portfolio data, SCADA or sensor trends, inspection measurements, laboratory results, protection tests, work orders, outage events, customer impact and project cost each have different owners and update cycles. The asset-management platform should link them without pretending they are one database.

Common failure modes in replacement programmes

• Age becomes the score. Condition fields are added, but the ranking remains a disguised age list.
• Unknown becomes average. Missing inspection data lowers urgency instead of lowering confidence.
• Criticality never changes. Network reconfiguration, new customers and redundancy changes do not update consequence.
• One health formula covers every class. Physically unrelated assets receive comparable-looking but meaningless scores.
• Inspections produce PDFs. The result is attached but the measurements cannot be queried or trended.
• Failures are counted without exposure. A large population appears worse simply because it contains more assets.
• Projects are approved, then disappear. No post-investment review tests whether failures, losses or risk actually fell.
• Data cleansing becomes a precondition. The programme waits for a perfect register instead of prioritising data improvement by consequence.

FAQ

At what age should a utility asset be replaced? There is no defensible universal age. Use age to set inspection and review expectations, then decide from condition, failure mechanism, consequence, future duty and intervention economics.

Is an asset health index enough? No. Health describes condition. It must be paired with consequence, evidence confidence and an option appraisal. A poor-health low-consequence asset may be monitored; a fair-health high-consequence asset may justify diagnostic work or risk reduction.

How should missing condition data affect priority? Through a separate confidence score. High-consequence assets with weak evidence should move up the inspection programme, not automatically into replacement and not silently down the risk list.

Should water and electricity assets use the same scoring model? They can share governance, consequence categories and portfolio reporting. Their condition inputs, failure modes, thresholds and intervention options must remain asset-class specific.

How often should rankings be refreshed? After a material condition result, failure, portfolio reconfiguration or demand change, and on a defined portfolio cycle. Recalculate from retained source data so changes remain traceable.

Where a system helps

Condition-led replacement depends less on a colourful risk matrix than on lineage: the observation behind the health score, the portfolio consequence behind criticality, the version of the rule, and the work and project that followed. An EAM platform should preserve that chain across inspections, failures, GIS locations, planned interventions and capital review, while showing uncertainty rather than filling it with defaults. See eAMS for asset management.

Related reading: What Is a CMMS? (KB-129) and The Work Order Lifecycle: Nine Statuses, and the Two Where Data Quality Dies (KB-500).

Sources

• US DOE, Connecting Buildings: Data and Performance
• ISO 55000 Asset Management