In a data center, cooling is only as reliable as the drives that run it. At this 6 MW Tier III colocation facility, every one of the 84 variable frequency drives sits in the cooling path, from 48 CRAH fan motors in the data halls to the 400–500 hp drives on the chiller compressors. When one of them fails, the problem becomes a thermal one within minutes.
The facility runs a chilled-water plant with N+1 redundancy on most equipment and N+2 on the CRAH fans. Most of the drives were installed between 2015 and 2017, putting the fleet at 9 to 11 years of service, right where capacitors, fans and connections start to wear out.
Before the program, the site ran its drives to failure. VFDs were serviced only when they faulted, and redundancy was treated as the maintenance plan. That worked for years, until two problems lined up on the same pump pair.
In August, secondary chilled water pump CHWP-2A was locked out for a scheduled seal replacement. While it was down, the drive on its partner pump, CHWP-2B, tripped on a DC bus overvoltage fault and would not reset. With both pumps offline, chilled water flow to Data Hall 2 dropped. Cold-aisle temperatures climbed past 27 °C (80.6 °F), the upper limit of the ASHRAE recommended range, and stayed there for 38 minutes until a technician put CHWP-2B in bypass.
The root cause was a failed DC bus capacitor bank and a cooling fan with a seized bearing. Both are wear items with known service lives, and neither had been checked since the drive was commissioned.
The August excursion was the most visible failure, but not the only one. In the 24 months before the program, the site logged seven unplanned VFD failures, four of them after hours, at an average repair cost of $9,500 each once expedited parts and overtime were added in. The single customer-impacting event cost about $310,000 in SLA credits, incident response and customer remediation.
This facility isn’t an outlier. Uptime Institute’s 2024 analysis found power is the leading cause of impactful outages at 54%, with cooling second at 13%, and more than half of operators said their latest significant outage cost over $100,000.
ICS built the program around the parts of a VFD that wear out on a predictable schedule: capacitors, cooling fans, power connections and filters. Just as important was the sequencing. Every task was scheduled so that no redundant pair was ever out of service at the same time, closing the exact gap behind the August incident.
Most VFD failures start with heat. Our technicians perform infrared scans of line, load and DC bus connections with the drives under load, annually across the fleet and twice a year on the chiller and chilled water pump drives. Power terminals are re-torqued to spec each year, since thermal cycling and vibration loosen connections over time. We also measure DC bus ripple and capacitance annually, and any capacitor bank below about 85% of its rated value is flagged for replacement.
Dust acts as insulation, and every 10 °C rise in operating temperature roughly halves capacitor life. Twice a year, ICS cleans heatsinks, replaces cabinet filters and vacuums enclosures. Cooling fans are checked annually for noise, speed and bearing wear, because a failed fan is the most common path to an overheat trip.
Drives usually report trouble long before they fail. Each quarter, we download fault and trip logs and review the trends. Repeated overvoltage or undervoltage trips point to power quality or capacitor problems while there is still time to act.
When a drive does fail, recovery depends on a fast swap and a working bypass. ICS backs up drive parameters and confirms firmware every year, and function-tests every bypass contactor transfer. A bypass that has never been exercised can’t be counted on in an emergency.
Capacitors and fans are consumables. Following manufacturer guidance such as Hitachi’s, the program replaces DC bus capacitors at 5 to 7 years and cooling fans at 3 to 4 years, so the owner can plan and budget
The first PM cycle found 56 problems on 41 of the 84 drives, nearly half the fleet. Most were the same heat-related wear that caused the August trip.
The most serious findings were three power terminations running more than 20 °C hotter than comparable connections; the worst measured 41 °C on the line side of a chiller drive. Six drives had capacitance below the 85% replacement threshold, all from the 2015–2016 installations. Three bypass contactors failed to transfer, two of them on chilled water pumps, the same kind of fallback that ended the August excursion. ICS corrected every critical finding within 30 days.
In the first 12 months of the program, the facility had zero cooling-related customer impacts. It had one unplanned VFD failure, on a CRAH fan with N+2 coverage, and redundancy absorbed it with no effect on the data hall.
The financial case was just as clear. Annual cost of ownership for the drive fleet fell by about 73%, for a net annual benefit of about $138,000. Against $45,900 in PM spend, that is a return of roughly 3 to 1, with payback in about four months.
| Cost item (annual) | Run to failure | With PM |
|---|---|---|
| Emergency repairs | $33,250 (3.5 failures × $9,500) | $4,200 (1 failure) |
| Outage exposure | $155,000 (one $310,000 event per 24 months) | $0 observed |
| PM inspections and thermography | $0 | $27,300 (84 drives × $325) |
| Corrective parts and labor from findings | $0 | $18,600 |
| Total | $188,250 | $50,100 |
The avoided outage is what makes the numbers work. On repair costs alone, year one cost about $17,000 more with PM, because the program cleared a backlog of worn parts. A single prevented excursion more than covers that, and the outage estimate here is conservative: the Ponemon Institute put the average cost of a complete data center outage at $740,357 in 2016, or $8,851 per minute.
The owner also gained benefits that don’t show up on a cost sheet: a parameter backup for every drive, a tested bypass on every critical pump, and a long-range plan for capacitor and fan replacements.
Redundancy kept this facility running for years, but it was never a maintenance plan. By treating capacitors, fans and connections as the consumables they are, Integrity Control Services turned an aging, run-to-failure drive fleet into a predictable, budgetable asset, and gave the operator confidence that its standby equipment will work when it is needed.
If your facility relies on VFDs approaching 10 years of service, the next fault could be the one that takes a critical system down. Integrity Control Services can assess your drive fleet, build a maintenance schedule around your redundancy, and correct problems before they turn into downtime.
Talk with our team about:
Whether you’re planning ahead or responding to an unexpected failure, call Integrity Control Services today or send us a message to get started.
