A tier-two automotive components plant in southern Vietnam brought unplanned line stoppages down from an average of eleven per month to fewer than one, after retiring an ageing commercial-grade UPS and a scattering of point-of-use stabilisers in favour of a single 120 kVA online industrial UPS with a lithium iron phosphate string. Across the first twelve months of operation the site logged a 92 percent reduction in power-related stops, recovered close to 240 hours of saleable production time, and cut scrap and unplanned overtime by an amount its finance team put at roughly 180,000 US dollars. Published with the customer's agreement; the company name is withheld under the supply contract.
The plant runs progressive stamping presses, four injection-moulding machines and an automated inspection cell across three shifts, six days a week. Its connected load sits at 340 kW, of which 62 kW belongs to equipment that cannot lose supply without consequence: the moulding-machine controllers, the servo drives on the inspection cell, the plant SCADA server and the quality laboratory.
Supply arrives on two feeders from a regional substation shared with a galvanising works and a cement terminal. Neither neighbour is doing anything wrong, but together they place a great deal of switched inductive load on one network.
Nobody at the plant described the problem as a power-quality problem. They described it as a controller problem. Moulding machines dropped mid-cycle and had to be purged and re-primed. The inspection cell lost its reference position and demanded a manual re-home. Twice in eighteen months the SCADA server corrupted a database after an abrupt shutdown, costing a day of traceability records.
The maintenance team responded the way most teams do: consumer-grade standby units under individual cabinets, two replaced controller boards, and the worst-affected mould changes pushed to the night shift. All of it treated the symptom.
Our application engineers installed a class A power analyser on the main incomer and on the sub-board feeding the moulding hall, and left it running for fourteen days. The record was unambiguous.
The sag duration is the number that explains everything else. Most industrial controller power supplies ride through somewhere between eight and twenty milliseconds. A 200 millisecond sag is an order of magnitude beyond that, so the controllers were behaving exactly as designed.
The standby units were rated for office equipment, yet sat inside sealed enclosures in a hall running at 38 degrees Celsius; their sealed lead-acid batteries had lost most of their capacity within a year and several were failing self-tests with the buzzer muted. The point-of-use stabilisers corrected slow voltage drift but could do nothing about a sag lasting a fifth of a second, because a tapped autotransformer cannot manufacture energy it does not hold. Neither device touched harmonic distortion.
The design brief was deliberately narrow: protect the 62 kW of critical load properly rather than the 340 kW poorly. The installed configuration was a 120 kVA / 108 kW online double-conversion industrial UPS in a filtered, forced-ventilated cabinet rated for continuous operation to 45 degrees Celsius, fed from a dedicated sub-board and feeding a new segregated critical distribution board.
Three decisions did most of the work. Continuous double conversion keeps the incoming waveform away from the controllers, so sags, distortion and frequency drift stop at the rectifier. Lithium iron phosphate was chosen over sealed lead-acid because the room could not be economically air-conditioned. An input harmonic filter held reflected current distortion below 5 percent, so the UPS did not become a fresh nuisance on the shared feeder.
Runtime was deliberately set at eight minutes. That covers every event the survey recorded plus a controlled shutdown of the SCADA server, whereas thirty minutes would have tripled the battery cost to guard against an event this network has never produced.
The plant could offer no shutdown window longer than one Sunday. The critical distribution board was therefore built and cabled alongside the existing one while production continued, each circuit fitted with a changeover so it could migrate individually. Over that Sunday the load moved circuit by circuit with a functional test after every transfer. Downtime attributable to the installation totalled two hours and ten minutes.
The figures come from the plant's own SCADA event log and downtime register, not from our instrumentation.
The one stop that still occurs monthly is a genuine long outage on the regional network, during which the UPS carries the load for its rated eight minutes and then hands over to an orderly shutdown. A generator is now under evaluation to close that last gap.
Three lessons generalise. Instrument before specifying, because a fortnight of logging costs almost nothing and prevents an expensive guess. Protect a defined critical bus rather than the whole plant, since the critical fraction is usually under a quarter of connected load. And treat electrical-room temperature as part of the specification, because ambient heat is what quietly destroys the batteries being relied on.
If your site shows the same signature of controller dropouts, mystery resets and rising scrap, our engineers can arrange a power survey and return a costed configuration. Equipment built for plant-floor conditions is described at https://www.upsboss.com/industrial-ups/, and the full catalogue including ratings and single-line options is at https://www.upsboss.com/products/.
Key takeaway: the plant did not have a controller problem, it had a 200 millisecond sag problem, and no amount of board replacement was ever going to fix it. Measuring the disturbance first, then protecting a properly defined critical bus with an industrial-rated online UPS, removed 92 percent of the stoppages within a year.
How long does a power survey like this take?
Two weeks of continuous logging is the practical minimum: it must capture a full production cycle, weekend behaviour and the pattern of neighbouring loads. Shorter surveys miss the events that matter.
Why protect only 62 kW out of 340 kW of connected load?
Because presses and heaters restart without incident after a sag, while controllers, servo drives and servers do not. Segregating the sensitive fraction cut the UPS rating, the battery cost and the cooling load by roughly a factor of five without weakening the protection.
Was lithium worth the higher purchase price here?
In this room, yes. At a sustained 38 to 41 degrees Celsius a lead-acid string would need replacing every eighteen to twenty-four months, so over an eight-year horizon lithium was cheaper in total.
Contact: Frank Zhang
Phone: +86-135 5688 8641
Email: frank@upsboss.com
Add: Jufeng Road, Guangming Street, Guangming District, Shenzhen City, Guangdong Province, China
