# How long on-site balancing takes, and how many runs it needs

> You're not asking out of curiosity. You need to give production a window and tell them when the machine goes back under load. The instrument calculates the correction in seconds, while a shift gets eaten up by coast-down (the rotor's free stop by inertia — you can't brake it), hatches, guards and fastening weights. Below is the honest arithmetic of runs and typical time ranges by machine type.

**In short:** Balancing in one plane needs a minimum of three runs: the initial run, one trial run and a check run. In two planes, a minimum of four: the initial run, two trial runs and a check run. In practice, add one or two trim runs, so plan on 4–6 runs. By time, a machine with normal access and bolted weights takes 2–5 hours, while a machine that needs welded weights and a long coast-down easily eats up a whole shift. Time is set by coast-down, access to the correction plane (the location on the rotor where the weight goes) and how the weights are fastened — not by the instrument.

Source: https://axiline.pt/en/articles/how-long-site-balancing-takes-many/  
Publisher: AXILINE · Vila Nova de Gaia, Portugal · +351 931 831 229 · axilinegeral@gmail.com

## The arithmetic of runs: three and four as a minimum

Count runs, not hours. Each run is a full cycle: stopping, coast-down, lockout, access to the correction plane, working with the weight, reassembly, clearance to start, run-up, stabilization, measurement. Multiply the number of runs by how long that cycle takes on your specific machine, and you'll have nearly all your working time.

The method dictates the number of runs. The instrument calculates from influence coefficients — the machine's measured response to an added weight. Until the first trial run, it knows neither your rotor's mass, nor the stiffness of the supports, nor how the belt drive and frame transmit vibration. It learns from a trial weight of known mass. In one plane it learns once; in two planes, twice, one trial run per correction plane.

In the Balanset-1A software, the runs are labeled accordingly: Run #0 is the initial run, Run #1 and Run #2 are trial runs, RunTrim is the refining run. Always do a check run after installing the correction weights, even if the calculation looks clean. Without it you have no 'after' number, and so no report.

| What we're counting | One plane | Two planes |
| --- | --- | --- |
| Initial run (Run #0) | 1 | 1 |
| Trial runs | 1 | 2 |
| Runs before the correction is calculated | 2 | 3 |
| Check run | 1 | 1 |
| Minimum total | 3 | 4 |
| Realistic, with one trim run | 4 | 5 |
| A hard case, two trim runs | 5 | 6 |
| With saved influence coefficients | 2 (correction and check) | 2 (correction and check) |

> After each trial run, the trial weight is removed. In a two-plane setup, you move it from plane 1 to plane 2 rather than leaving two weights on at once. That's an extra stop, and it's the one most often forgotten when planning the window.

Sources: [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/)

## One run is not ten minutes

The engineer presses 'measure' and gets amplitude and phase within a minute. Everything else in the cycle is done by hands, wrenches and the operations crew. Break the cycle into its parts, and you'll immediately see where the time is hiding on your machine.

A reading is only taken at a constant speed. After run-up, let the machine reach its operating point and check that the readings are stable: the amplitude and phase of 1x — the vibration at the running speed, which is what unbalance produces — shouldn't shift by more than 10–15% over the measurement window. If the readings are drifting, you're not saving time, you're collecting noise, and you'll calculate the wrong correction from it.

| Stage of the cycle | What's happening | Typical |
| --- | --- | --- |
| Stopping and coast-down | The rotor stops on its own; it can't be braked | 1–20 min |
| Lockout and clearance | Disconnecting power, tagging, locking, sign-off from the responsible person | 5–20 min |
| Access to the correction plane | Hatch, guard, cover, platform, lighting | 5–40 min |
| Working with the weight | Weighing, marking the position, attaching or removing material | 10–40 min |
| Reassembly | Cover and guard back on, tools out of the machine | 5–20 min |
| Clearance and run-up | Coordination with the operator, starting, reaching operating mode | 2–10 min |
| Stabilization and measurement | Constant speed, readings stable within 10–15% | 3–10 min |

> The ranges in the table are typical and depend on the machine, the access, and how things are run at your plant. Take your own worst and best estimate for each row, add them up, and multiply by the number of runs from the first section. That will be your window — not a figure from a price list.

Sources: [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/)

## What actually eats up the time

The calculation eats up nothing. The software gives you a mass and an angle for each plane right after the trial run. Time goes into physics and organization.

- Run-up and coast-down. A heavy impeller takes a long time to speed up and even longer to stop. On an exhaust fan or a large crusher rotor, a single revolution cycle can eat up more time than the entire measurement.
- Access to the correction plane. An inspection hatch with two bolts is one thing. A removed cover, a dismantled guard mesh, climbing inside a housing and working in a mask is something else entirely. On an unfamiliar machine, the first access is always the slowest.
- Fastening the weights. A weight on a bolt into a ready-made hole takes minutes. Welding takes tens of minutes, plus prep, electrodes and cooling. Removing material by drilling needs marking out, depth, and checking the mass of the shavings.
- Weighing. The mass of the trial weight and the correction weight goes in as the actual measured value, not 'about five grams.' Scales need to be at hand, or every weight means a trip to the workshop.
- Cooling down. After welding, and on a hot machine, you can't measure right away: thermal bow changes the picture. The wait takes as long as the metal needs, not as long as the schedule wants.
- Coordinating runs with production. Every run and every stop goes through the operator and the dispatcher. On a continuous process, this is the most unpredictable item of all.
- Paperwork. Before-and-after readings, the spectrum, the polar diagram, weight position numbers and masses, signatures. The report is assembled from the instrument's archive, but filling it in and getting it signed off still takes time.

> Which gives a simple takeaway for planning. If you can't shorten the coast-down, shorten the number of stops and the time each one takes. Ready access and a prepared mechanic buy you more than any amount of calculation speed.

## Benchmarks by machine type

Below are typical ranges for a machine in sound condition, where the vibration genuinely comes from unbalance. These are planning benchmarks, not a promised deadline. Resonance, worn bearings, a loose wheel fit or a wandering phase will throw you out of any row in this table.

| Machine | Planes | Runs, typically | Time on site |
| --- | --- | --- | --- |
| Small fan, impeller reachable through a hatch, bolted weights | 1 | 3–4 | 1.5–3 h |
| Medium-size fan or exhaust fan, two correction planes | 2 | 4–6 | 3–6 h |
| Medium-size electric motor, correction on the coupling half or cooling fan | 1–2 | 3–5 | 2–4 h |
| Mulcher, crusher, shredder, weights welded on | 1–2 | 4–6 | 4–8 h |
| Long shaft, drum, screw conveyor, access only through disassembly | 2 | 4–6 | a shift, sometimes two |

> A first visit to an unfamiliar machine always runs longer. It includes diagnostics: an overall-vibration and 1x reading at both bearing housings, a spectrum, a fastener check and a resonance check. Sometimes that part alone ends with a decision to repair rather than balance. A repeat visit to the same machine goes faster, because the influence coefficients are already saved.

## Why a long coast-down doubles your time

Here's a typical example. An exhaust fan with a heavy wheel and a large moment of inertia coasts down for 10–15 minutes, then takes another 5 minutes to run up and reach its operating point. Rotation alone eats up about 20 minutes per cycle. Five runs give you an hour and a half during which you're standing there watching the machine. No instrument will shorten those minutes.

You can't brake the coast-down. You'll damage the bearings, and on some machines the seals too. Use a spring or shoe brake, if the machine has one, only as intended.

A long coast-down does have a benefit. While the rotor is slowing down, you get a free frequency experiment: the amplitude and phase of 1x, captured across the speed range, show whether there's a natural frequency near the operating speed. The Balanset software has a dedicated coast-down recording function for this. If the phase swings by roughly 180° as speed drops, and the amplitude gives a sharp peak, you've found a resonance and saved yourself an entire day of pointless trial weights.

> Plan the work around the coast-down. While the rotor is spinning down on its own, prepare the weight, the scales and the tools. While it's stopped, don't go running for electrodes. The difference between organized and disorganized work on a machine with a long coast-down adds up to several hours in a shift.

## What speeds the work up

Everything listed below is something you can prepare before the engineer arrives. Each item shaves minutes off every run, and there will be three to six of them.

- [x] Access is opened in advance: the hatch is unbolted, the guard is off, the platform is lit, the ladder is in place.
- [x] The impeller is clean. Build-up, ash and product create unbalance all over again, and the readings will need repeating after cleaning.
- [x] Speed is known and stable. On a variable-frequency drive, fix the setpoint and don't change it between runs.
- [x] Your own mechanic is on site, and if needed, a welder with electrodes and a hot-work permit.
- [x] Weight positions are marked out and numbered. The instrument can split the correction mass across fixed positions, by blade or by hole, and then instead of an angle you get a position number Z1…Zn and a mass.
- [x] A set of plates, bolts and washers, plus scales with 0.1 g resolution, are kept right by the machine.
- [x] The work permit and clearance for the runs are arranged before work starts, not on the fly.
- [x] Influence coefficients from the last visit are saved in the archive. On the same machine, or an identical one, you skip the trial runs and get by with a correction run plus a check.

> Saved influence coefficients save the most of all. Across a fleet of identical fans, you pay with three runs once, and after that you go through two runs per machine. Just check that the speed, the type of supports, and the sensor mounting points are all the same.

Sources: [Balanset-1A manufacturer specification](https://vibromera.eu/product/balanset-1/) · [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/)

## What stretches the work out, and what it looks like

The five scenarios below turn a three-hour job into a full shift, or into a decision not to balance at all. The engineer sees them from the very first readings, and it's more honest to say so up front than to promise a window and miss it.

### 1x doesn't dominate

Overall vibration is several times higher than the running-speed component. Balancing will only remove a small part of the level. Instead of trial weights, the search for the cause begins: spectrum, fasteners, soft foot (a support that doesn't sit flush against the foundation), the fit of the wheel on the shaft, bearings, misalignment. That's an extra two to four hours, and sometimes a decision to postpone balancing until after a repair or a shaft alignment.

### Resonance

The operating speed sits near a natural frequency of the rotor, the supports or the frame. The amplitude gives a sharp peak, the phase of 1x is unstable, and the influence coefficients don't repeat from run to run. Balancing in this zone is pointless; the result won't hold. Support stiffness, fasteners, or a change in operating speed have to come first.

### Unstable speed

Speed wanders between runs, a damper or valve sits in a different position, the load changes. Amplitude and phase drift by more than 10–15%, and every reading has to be repeated. Every repeat is another coast-down-and-run-up cycle.

### Nonlinear response

You installed a trial weight, and the vibration changed in a way it shouldn't have. The cause is usually mechanical: a loose wheel fit, a crack, play, rubbing. A calculation from that run will send the weight to the wrong place, and instead of a reduction you'll get an increase. What follows is a mechanical investigation, not more weights.

### A weak trial weight

An amplitude change under 20–30%, or a phase change under 20–30°, means the influence coefficient was calculated from noise. The instrument will tell you the weight isn't adequate. You stop the machine, fit a heavier one, enter the actual mass, and run the trial again. One cycle lost.

> A separate category is hot machines. After welding, and on a rotor that's still warm, a reading taken right after stopping isn't valid — thermal bow distorts the picture. Build cooling time into the schedule in advance.

Sources: [ISO 13373-3:2015](https://www.iso.org/standard/40840.html)

## How many machines you can realistically get through in a day

The instrument calculates the correction in seconds, so daily throughput is decided by access and clearances, not by it. Here are honest numbers for planning.

One complicated machine takes the whole shift. Two or three small identical machines are realistic if access to the correction planes is opened in advance, the weights go on with bolts, and clearance for the runs is obtained for the whole group at once. Four or more happens only across a fleet of identical machines, where you're working from saved influence coefficients and doing two runs per machine.

Plan around your bottleneck. If every run at your plant needs twenty minutes of clearance, your ceiling is four or five runs a shift, however fast the engineer is. If there's one welder for the whole shop, your ceiling is his schedule.

- The limiting factor is almost always the same: access, clearance and coast-down.
- Identical machines in a row go faster than mixed ones, because the tooling, speed and weight-mounting logic don't change.
- Diagnostics without balancing goes much faster: 30–60 minutes per machine, if the measurement points are accessible and the speed is known.
- Balancing a removed impeller on a soft-bearing machine goes faster than field work, because there are no hatches, no clearances and no long coast-down. But add time for removal, transport and reinstallation.

> Ask your contractor for a completion criterion, not just a time estimate. 'Machine within tolerance' should mean a specific number. As a working benchmark for stationary-part readings: mm/s RMS (root mean square) in the 10–1000 Hz band, zone A or B. Check the applicable part and edition of ISO 20816 against your own machine separately — there are exceptions by power, speed and machine type. Residual unbalance by grade G is assessed separately, under ISO 21940-11.

Sources: [ISO 20816-1:2016](https://www.iso.org/standard/63180.html) · [ISO 21940-11:2016](https://www.iso.org/standard/54074.html)

## How to plan a window together with AXILINE

We don't quote a deadline over the phone at random. From the machine type, speed, access to the correction planes and how the weights are fastened, we give a range and say exactly what could stretch it. The workflow is straightforward: two vibration sensors on the bearing housings, a laser tachometer on the reflective marker, an initial reading at both bearings, a decision on the number of planes, trial runs, correction, a check run and, if needed, a trim run. Results go into the archive, from which a report with before-and-after numbers is assembled.

If you balance regularly, keeping an instrument on hand is more cost-effective than calling out a visit every time. The Balanset-1A kit fits in a case: two accelerometers, a laser phase sensor, a two-channel USB module with preamplifiers, integrators and an ADC, and Windows software. It handles one and two planes, tells you whether a trial weight is adequate, splits the mass across fixed positions and calculates a drilling correction when removing material, shows a polar diagram, recalculates weights for other correction planes, and stores influence coefficients for next time. The same kit works as the measuring system on a soft-bearing balancing machine.

Balanset instruments are designed and built by the engineers who use them for field balancing themselves. So on choosing the number of planes, mounting the sensors, and troubleshooting a failed run, you get consulting support, not a form response.

- [x] Machine type, drive power, operating speed, and photos of the rotor from both sides.
- [x] Whether the correction planes are accessible without disassembly, and how you'd reach them.
- [x] How weights are fastened on this machine: bolt, welding, material removal.
- [x] How long the coast-down takes, and whether there's a limit on the number of runs per shift.
- [x] Your own readings, if you have any: overall vibration, and the amplitude and phase of 1x at both bearings.
- [x] The window production can give you, and who from the shop floor will be at the machine.

Sources: [Balanset-1A manufacturer specification](https://vibromera.eu/product/balanset-1/) · [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/)

## Frequently asked questions

**What's the minimum number of runs needed?**

Three in one plane: the initial run, one trial run and a check run. Four in two planes: the initial run, two trial runs and a check run. Two runs are needed before the correction is calculated in one plane, three in two planes. In practice, plan on one or two extra trim runs, because the system is rarely perfectly linear.

**Can you skip the trial runs?**

Yes, if you've already balanced this machine, or an identical one, and saved the influence coefficients. Then the instrument calculates the correction from the initial reading alone, and you're left with two runs: the measurement and the check. Conditions: the same speed, the same sensor mounting points, the same type of supports, and the same correction-plane geometry. Change any of that, and the coefficients have to be obtained again.

**Why won't the engineer give an exact time over the phone?**

Because time is set not by the calculation or the instrument, but by three things you can't see over the phone: how long the coast-down takes, access to the correction plane, and how the weights are fastened. Plus a fourth: whether unbalance is even confirmed at all. If overall vibration is several times higher than 1x, the job turns from balancing into a search for the cause, and that's a different scope of work.

**Does balancing in a single run ever happen?**

A complete job, no. One run only gives you the initial picture; you can't calculate a correction from it. The closest thing to 'one run' is the case with saved influence coefficients: you take a reading, install the weight, and confirm the result with a check run. Never skip the check run, or you have no proof of the result.

**How much longer does the job take if the weights have to be welded on?**

Count on an extra 30–60 minutes per weight installation compared with a bolted fastening. That covers prep, marking out, the welding itself, cooling, and a repeat reading once the machine has cooled. On a mulcher or crusher with two planes and one trim run, that easily adds half a shift to the job. Hot work also needs its own permit, which is better arranged before the visit.

**Production is only giving us two hours. What can we realistically get done?**

In two hours you can realistically get diagnostics done and decide whether balancing is even needed: a reading at both bearing housings, a spectrum, a fastener check and a resonance check. Balancing in a single plane, with an open hatch and bolted weights, sometimes gets done too, but with no margin for a trim run. It's more honest to split the work into two windows: diagnostics in the first, balancing in the second, once access and weights are already prepared.