Time-based maintenance (TBM) persists in electric utility distribution programs for understandable reasons. Fixed intervals are auditable, defensible to regulators, and easy to budget. A five-year inspection cycle for distribution transformers and a 50,000-operation refurbishment trigger for reclosers are simple rules that field supervisors can communicate to crews and that asset managers can plan capital expenditures around. The cognitive burden is low and the compliance story is clean.
The problem is that time and operation counts are poor proxies for actual asset condition. Equipment does not degrade on a clock or a counter — it degrades in response to loading, temperature, fault exposure, environment, and installation quality. Two identical 25-year-old transformers on the same utility system can have radically different remaining life depending on whether one has been operating at 60% of nameplate on a mild feeder and the other has been pushed to 110% during summer peaks on a high-growth corridor. A maintenance program that treats them identically is wasting resources on the low-stress unit while potentially missing the one that needs intervention.
Why TBM Survived So Long
It is worth understanding why time-based maintenance has been the dominant paradigm for distribution assets before arguing for something different. The primary reason is data scarcity. Before continuous telemetry from field assets was economically and practically feasible, there was no alternative to time-based triggers — you could not base maintenance on condition data you did not have. Periodic field inspection was the only condition assessment available, and its cost per visit made high-frequency inspection uneconomical.
Fixed intervals represent a rational optimization under data constraints: if you cannot monitor continuously, set a replacement interval that is short enough to catch most failures before they cause unplanned outages, and accept that some assets will be replaced early and some late. This is not irrational — it is the best policy available without continuous condition data.
Condition-based maintenance (CBM) becomes superior to TBM only when the cost of the condition monitoring data is less than the value of the maintenance optimization it enables. That crossover point has shifted dramatically as cellular connectivity and low-cost IoT sensor hardware have made continuous distribution asset monitoring economically viable at scale. The economics of CBM are now materially different from what they were 10–15 years ago.
The CBM Economic Case: Three Cost Buckets
The economic case for switching from TBM to CBM for distribution transformers and reclosers rests on three cost savings that are partially offsetting against the monitoring investment:
Savings 1: Avoided Premature Replacements
In a fixed-interval program, some assets are replaced or refurbished before they have exhausted their useful life. The exact fraction depends on the interval length and the distribution of actual asset lifespans, but for distribution transformers with a 35-year replacement policy, a meaningful share of units still have significant remaining insulation life at the replacement trigger age. The economic value of extending those units' service life by 5–10 years — either deferring capital replacement or redeploying spare inventory — is straightforward to quantify.
The catch is that you cannot defer replacement on unknown-condition assets without accepting higher failure risk. CBM changes the equation by providing condition knowledge that allows confident deferral of well-conditioned units while flagging the deteriorating ones for earlier intervention. Deferral with monitoring is a different risk proposition than deferral without it.
Savings 2: Avoided Unplanned Failure Costs
An unplanned transformer failure costs 2x to 4x more than a planned replacement, depending on the time-of-failure, spare unit availability, and the secondary damage (cable termination damage, switchgear damage) that sometimes accompanies a failure event. For a utility that experiences 40–80 emergency transformer replacements per year, converting a meaningful fraction of those to planned replacements through early anomaly detection produces direct, measurable cost savings — before counting the SAIDI impact and associated regulatory exposure.
The conversion rate is not 100% — some failures occur too suddenly for any monitoring-based intervention, particularly assets that fail due to sudden external events (lightning, through-fault from an external source). But for the 60–75% of transformer failures that are preceded by a degradation period of weeks to months, continuous monitoring provides the advance warning needed for planned intervention.
Savings 3: Optimized Inspection Efficiency
CBM does not eliminate field inspection — it makes inspection more targeted. Instead of a uniform 5-year inspection schedule across all assets, a CBM program concentrates inspection resources on condition-flagged assets while extending the inspection interval on demonstrably healthy ones. The total inspection cost per unit of risk-reduction is lower under a targeted inspection regime than under a uniform one.
The risk of eliminating field inspection entirely for healthy-monitoring assets should not be underestimated. Condition monitoring detects certain failure modes well — mechanical loosening, thermal aging, contact wear. It does not detect external damage (vandalism, wildlife contact, physical damage from vehicles or storms) or certain external contamination conditions that are only visible on a field visit. A hybrid program — frequent CBM flagging with extended but not eliminated field inspection on healthy assets — is more defensible than eliminating field inspection entirely on monitoring grounds.
The Data Requirement Is Specific
CBM does not work without continuous, reliable data. This is the constraint that has prevented wider CBM adoption in distribution maintenance programs and it is worth being direct about: if your monitoring data has significant gaps, if your baseline models are not calibrated, or if your anomaly flags are generating false positive rates above 20–30%, the CBM program will not produce the cost savings described above. Crew confidence in the anomaly flags is a prerequisite for the program to change dispatch behavior.
The data requirements for a functional distribution CBM program are: per-asset telemetry at sufficient resolution (temperature, load, and vibration at minimum 15-minute intervals), continuous connectivity with store-and-forward buffering for connectivity gaps, and a calibrated anomaly model that has been validated against the specific asset classes in the fleet. The calibration period — typically 4–8 weeks of baseline data per asset subtype — is a real program startup cost that TBM does not require.
We are not arguing that CBM is automatically superior to TBM for every utility and every asset class. For small fleets below a certain size, the monitoring infrastructure cost may not be justified by the maintenance savings. For asset classes with very long, predictable lifespans and low failure variability, TBM may perform nearly as well as CBM at lower cost. The economic case for CBM is strongest for large, heterogeneous fleets with significant loading variability and significant consequences from unplanned failures — which describes the distribution transformer and line recloser fleets of most medium-to-large IOUs and cooperatives operating in climate-stressed territories.
The Transition Path
Utilities that move from pure TBM toward CBM typically do not switch overnight — they run a hybrid program where CBM data informs TBM decisions rather than replacing them entirely. In practice this looks like: maintain the existing scheduled replacement and inspection cycles, but use CBM data to trigger early interventions on flagged assets and to justify deferral decisions on well-conditioned assets that would otherwise cycle through on schedule.
The hybrid approach has an important data value: it generates labeled outcomes (planned interventions triggered by CBM flags, inspections that confirm or deny the flag) that allow the monitoring program to be validated and calibrated against actual fleet behavior rather than just against the model's internal validation datasets. This validation data is what builds the organizational confidence to shift more decision weight toward CBM triggers over time.
