How a Variable Speed Drive Changes the Mechanical Picture

Why this matters

Putting a variable speed drive on a belt-driven machine feels like a controls job, and mechanically it is treated as if nothing happened: same belt, same sheaves, same tension, same everything. What actually changed is that every mechanical check that used to be true at one operating point now has to be true across a range, and the point where the drive is most stressed is usually not the point anyone commissioned it at. On top of that, the drive adds one wear mechanism the fixed-speed machine did not have, and it arrives at the bearing rather than at the belt.

Isolate properly: a stopped drive is not an isolated drive

Pressing stop on a variable speed drive leaves the motor circuit live and the machine available to a remote command or an automatic restart after a fault clears. Mechanical work goes behind a locked and tagged disconnect, not a stop button, under 29 CFR 1910.147, and the wheel still holds rotational energy afterwards, so verify zero rotation by eye and block the wheel or close and secure the damper before reaching past the guard. A high-inertia load can also be turning long after the drive has stopped commanding it.

Opening the drive enclosure is electrical work under 29 CFR 1910.333(b)(2), and it carries a hazard the contactor world does not: the DC bus capacitors hold charge after the incoming disconnect opens. Wait the manufacturer's stated discharge time, then verify the DC bus voltage at the labelled test points is below the manufacturer's stated safe threshold, using the live-dead-live proving sequence in NFPA 70E-2021, 120.5. A drive whose display is dark is not a drive whose bus is discharged. The guard goes back on before restart under 29 CFR 1910.219 in general industry or 29 CFR 1926.300(b) on a construction site.

What does not change at all

Worth stating plainly, because the mistake is usually an assumption of the opposite. Net pull in the belt is still torque divided by the driver's effective radius. The belt still needs a measured tension with a reserve above the peak torque it will see. Sheave alignment still matters at every speed. The bearings either side still carry the belt pull continuously, and that pull does not fall when the machine is turned down, because it is set by the installed tension and not by the load.

One thing genuinely can improve: a ramped start can remove the breakaway torque peak that an across-the-line start produces, which is often the peak the tension reserve was sized against. But that is a property of the programmed acceleration ramp and torque limit, not of the drive being a drive. Be careful how you state the risk, though: a drive's output torque is capped by its current limit, typically 110 to 150 percent of rated, while an across-the-line start produces locked-rotor torque well above that. A drive does not normally out-torque a direct start. What it does is HOLD near its limit far longer than a direct start does, and an aggressive ramp or low-speed boost can hold enough torque for long enough to exceed the reserve the belt tension was set against. Read the parameters before you claim the benefit.

The binding case moved to the ends of the range

A fixed-speed drive is checked once. A variable speed drive has to pass at the lowest commanded speed, at the highest, and at whatever the load law does in between, and the check that fails is rarely at the design point.

• Belt speed becomes a range. Published horsepower ratings per belt rise through a broad mid range of belt speed and fall off at both ends, so a drive turned down to half speed is asking its belts to work at a rating you never looked up. Read the rating table at the lowest commanded speed, not at the design speed.
• The lowest speed at full torque is the belt's worst case, because torque decides net pull and low belt speed decides the rating available to carry it.
• The highest commanded speed is the motor's worst case on a centrifugal load, since absorbed power climbs with the cube.

Whether turndown is benign depends entirely on the load law

This is the single fact that separates a variable speed retrofit that is mechanically free from one that is not.

Centrifugal loads (fans, centrifugal pumps and blowers) absorb power with the cube of speed along a fixed system curve, so torque falls with the square of speed. Turn one down and the belt's burden collapses. The condition is real and it is often violated: a system with a large fixed pressure component that does not scale with the square of flow, such as a building pressurization requirement or a stubborn fixed loss, does not follow the cube law down, and neither does a system whose dampers moved between the two readings.

Constant-torque loads (positive-displacement pumps, screw and reciprocating compressors, conveyors) demand roughly the same torque at any speed. Turn one down and the net pull in the belt does not change at all, while belt speed halves and motor cooling falls. Nothing gets easier and two things get harder.

Motor cooling falls faster than the heat does

On a totally enclosed fan-cooled motor with the cooling fan mounted on its own shaft, airflow over the frame falls roughly in proportion to speed. Meanwhile the dominant resistive loss in the winding tracks current, and current tracks torque, so on a constant-torque load at half speed the heat is close to unchanged while the cooling air has halved.

That is why continuous low-speed operation at high torque needs either a separately powered blower, an inverter-duty machine, or a derate read off the motor maker's speed-torque curve. The relationship above is stated for a shaft-mounted-fan enclosed motor; a blower-cooled machine has cooling that is independent of speed and does not behave this way at all, which is exactly why the two get specified differently.

The mechanism that is genuinely new: current through the bearing

Inverter output switches fast, and the switching produces a common-mode voltage that appears between the motor shaft and the frame. When that voltage exceeds what the lubricant film between rolling element and raceway can insulate, it discharges through the film. Each discharge removes a tiny amount of metal, in the same way a spark-erosion machining process does, and over months the accumulated pitting organises into the washboard pattern called fluting.

Two things matter for a field tech. First, this is a bearing failure with no mechanical cause, so a drive that is aligned, tensioned and loaded correctly can still eat bearings, and replacing them without addressing the current does exactly nothing. Second, the risk is not universal: it rises with frame size, with motor cable length, and with switching frequency, so an existing installation that has never shown it may be fine and a new larger machine on a long cable run may not be. The mitigations are a shaft grounding ring, an insulated bearing at one end, an output filter, and correct cable shielding and bonding, and which combination applies is the motor and drive makers' call rather than a field judgment.

The commanded speed can park on a resonance

A fixed-speed machine either sits on a resonant band or does not, and you find out once. A variable speed machine sweeps through every band between its limits and can be commanded to sit in one indefinitely. The control action is a skip or lockout band programmed into the drive, and it has to be set from measurement during a commissioning sweep, never from a calculation. Telling a resonance apart from a genuine mechanical fault is a separate skill with its own article, and this one does not re-derive it.

Worked example: one drive, two loads, two frequencies

Motor 5.0 hp, base speed 1750 rpm at 60 Hz. Driver sheave pitch diameter 4.0 in, driven 10.0 in, so the driven shaft runs 700 rpm at 60 Hz. Belt speed is pi x 4.0 x 1750 / 12 = 1,833 ft/min at 60 Hz.

As a fan, at 60 Hz. Absorbed shaft power measured at 4.2 hp, or 84 percent of nameplate. Motor torque is 63,025 x 4.2 / 1750 = 151 lb-in, and the net pull at the 2.0 in driver radius is 151 / 2.0 = 75.6 lb.

As a fan, turned down to 30 Hz. Motor 875 rpm, fan 350 rpm, belt speed 916 ft/min. On a fixed system curve with no significant fixed pressure component, absorbed power is 4.2 x 0.5 cubed = 0.525 hp. Torque is 63,025 x 0.525 / 875 = 37.8 lb-in, and net pull is 18.9 lb, a quarter of what it was. The belt is loafing. Motor current is low, so winding heat is low, and the halved cooling airflow does not matter. Turndown on this machine is mechanically free, and that is the case everyone generalises from.

Same drive, a positive-displacement load, at 30 Hz. Torque stays at about 151 lb-in because the load is constant-torque, so net pull stays at 75.6 lb. Absorbed power is 151 x 875 / 63,025 = 2.10 hp, exactly half of the 4.2 hp at 60 Hz, which is what constant torque means. Now look at what changed around that unchanged net pull:

• Belt speed halved to 916 ft/min, so the per-belt horsepower rating has fallen and the drive is asked to transmit the same pull on a lower-rated belt. Read the rating at 916 ft/min before you accept the range.
• Winding heat is roughly unchanged because current tracks torque, while the shaft-mounted cooling fan is moving about half the air. This is where a continuous 30 Hz command needs the maker's derate curve or separate cooling.
• Tension reserve is unchanged in absolute terms, which is the one piece of good news, and it means the tension you set for 60 Hz still covers 30 Hz.

The failure this produces. Nobody commissions at 30 Hz. The sequence is a retrofit signed off at full speed, a control strategy that later parks the machine at its minimum for most of the day, and a motor that fails thermally in a season with no mechanical symptom anywhere and a belt drive that inspects perfectly. The record shows a motor failure, so a motor gets fitted, and it fails on the same schedule.

How to commission one properly

• Sweep the commanded range with the guard closed, standing out of the plane of rotation, and log driven speed and motor current at intervals rather than at the endpoints only. The endpoints are the two speeds least likely to hide anything.
• Note any band where vibration rises during the sweep and program a skip band around it. Interpreting what that vibration is belongs to the vibration articles, not to this decision.
• Check the belt rating at the lowest commanded speed and the motor's thermal capability at the lowest speed you intend to hold continuously, at the torque the load actually demands there.
• Record the minimum and maximum commanded frequencies on the equipment, because the next tech will size a replacement belt or sheave from the nameplate and the nameplate does not know about the drive.
• Ask what the load law is before any of the above. Fan or positive displacement is the first question, and everything in this article resolves differently on the two answers.

References

• 29 CFR 1910.147 for mechanical isolation and stored rotational energy, and for why a stop command is not isolation; 29 CFR 1910.333(b)(2) for work in the drive enclosure and NFPA 70E-2021, 120.5 for the live-dead-live proving sequence
• 29 CFR 1910.219 (general industry) and 29 CFR 1926.300(b) (construction) for guarding of belts, pulleys and sheaves
• Drive manufacturer documentation for DC bus discharge time and safe-voltage threshold, and for skip-band programming
• Motor manufacturer data for speed-torque derating on shaft-mounted-fan enclosed machines, inverter-duty ratings, and shaft grounding or insulated-bearing recommendations
• Belt manufacturer engineering data for horsepower rating per belt as a function of belt speed
• See related: What a Direct Drive Changes About the Whole Problem; The Difference Between a Resonance and a Genuine Mechanical Fault; How Sheave Diameter Sets Speed and Torque