# Trim balancing and saved influence coefficients: how to skip repeated trial runs

> After the calculated weight is installed, vibration usually drops several times over, but it doesn't always land on target the first time. That isn't an instrument error, and it isn't a reason to start from scratch. The software already knows how your machine responds to added mass, so it suggests adding a small weight to what's already on the rotor. And if the influence coefficients were saved from a previous visit, you get by with one run instead of three.

**In short:** Trim balancing (add-on correction) is a refining correction based on already-calculated influence coefficients, with no new trial weight. You install the calculated weight, run a check, and the software shows an additional mass and angle to add to what's already installed. One or two such steps is normal practice. If it takes four or more, look for an error in the entered radius or mass, a loosened fastening, resonance, or nonlinearity in the system.

Source: https://axiline.pt/en/articles/trim-balancing-saved-influence-coefficients/  
Publisher: AXILINE · Vila Nova de Gaia, Portugal · +351 931 831 229 · axilinegeral@gmail.com

## What trim balancing is, and when the software suggests it

The usual workflow: Run #0 gives the initial amplitude and phase of the running-speed component 1x — the vibration at the rotor's rotating frequency, which is what unbalance produces. The phase shows the angle at which the heavy spot sits on the rotation. Run #1, with the trial weight, shows how the system responds to a known mass at a known location. From that pair, the software derives the influence coefficient and outputs the mass and angle of the correction weight.

Next you stop the machine, remove the trial weight, install the calculated one, and run the check run, RunTrim. And this is where the Result tab doesn't ask you to recalibrate. It shows the parameters of an additional weight to add to what's already on the rotor. The achieved residual unbalance appears at the bottom of the window.

This is trim balancing, or add-on correction. No new trial weight is needed: the system's response is already described by the coefficients from the calibration runs, and the software solves the same problem again, starting from the new reading. The method is called successive approximation. It also clears out the errors you introduced while installing or removing metal.

> Add-on correction isn't guesswork touch-up. It's the same vector calculation using the same coefficients — the only difference is that the starting point is now a rotor that's already partly balanced.

## Where the leftover after the first correction comes from

### The weight didn't land exactly where it was calculated

On blades and in bolt holes, a weight can only go in fixed positions — you can't place it in between. Even on a free rim, marking it out by hand gives you ±3…5°. With a large initial unbalance, five degrees leaves a noticeable remainder.

### The mass is different from the calculated value

The software asks for 17.4 g, and you have a 20 g plate and a 4 g washer. Round up, and you get overcorrection. Weld the weight on, and you've added the mass of the weld too.

### The radius isn't the one you entered

The correction is calculated at the stated radius. Put the weight 20 mm closer to the axis, and its effect drops proportionally, while the phase stays the same.

### Part of the remainder isn't from unbalance at all

The instrument measures 1x across 0.02–80 mm/s with a phase accuracy of ±1°. At small residual amplitudes, measurement noise and speed jitter are already comparable to the remainder itself.

### The system isn't perfectly linear

The linear model is accurate for a rigid rotor on securely mounted supports. Any compliance in the joints slightly distorts the mass-to-response ratio.

## One trim run is normal. Four is already a diagnosis

One run after the first correction is normal practice. Two happens too, especially when the weight can only go on the blades. Starting from the third or fourth, stop adding metal and start looking for the cause — beyond that point you're just running the machine in circles.

- The machine isn't bolted to its base. Under noticeable vibration, the unbalance force lifts the unit, the system's stiffness changes from run to run, and the calculation comes out different every time.
- Bearings are damaged. Play and impacts produce shaft displacements the linear model doesn't describe.
- The rotor is behaving like a flexible one. As the mass at the heavy spot grows, so does the bow, which means so does its radius. Vibration increases faster than the mass, and the ordinary calculation stops converging.
- The operating speed is close to resonance. In the resonance zone, a 100 rpm change in speed can raise vibration tenfold, and the phase shifts by roughly 180°. No amount of trim correction will beat that.

| What you observe | What it means | What to check |
| --- | --- | --- |
| 0–1 trim run, vibration within tolerance | Normal. The system is linear, data was entered correctly | Nothing. Close out the report |
| 2 runs on a rotor with fixed positions | Acceptable: the position spacing doesn't allow the exact mass at the exact angle | Number of positions, actual split masses |
| 3 or more, vibration dropping more and more slowly | The calculation is systematically falling short. Usually an input error | Installation radius, trial-weight mass, units (grams or percent), direction of angle reference |
| Added a weight, vibration went up | The weight landed mirrored, or you're drilling instead of adding | Angle-reference direction, the Add/Delete flag |
| Readings differ from run to run | Speed is unstable, or the mounting is loose | Support tightness, soft foot, speed stability |
| Phase of 1x is different every time | Resonance or system nonlinearity | Natural frequencies from coast-down and a bump test (striking the stopped machine and recording the response), bearing condition, foundation attachment |

> Balancing a rigid rotor and working with a rotor in a flexible state are different jobs, covered by different parts of ISO 21940. Treat them as a working reference, and pick the applicable part and edition for your rotor first.

Sources: [ISO 21940-12:2016](https://www.iso.org/standard/50429.html) · [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/)

## Saved influence coefficients: one run instead of three

The influence coefficient links a trial weight of known mass to how the amplitude and phase of 1x changed. It describes not the rotor alone, but the whole system: rotor, supports, foundation and operating mode. Calculate it once, save it in the instrument's archive, and on the next visit no trial run is needed.

In Saved coeff. mode, you select the right record in the coefficient archive, confirm the choice, and the rest of the settings fill in on their own: trial-weight mass, radii, number of fixed positions. Then one Run #0, and the software immediately outputs the correction mass and angle.

### Series rotors

You calibrate the first unit fully. From the second one on, every rotor is one run, install the weight, check run. Three stops become one.

### Routine maintenance

An exhaust fan, a grinding spindle, a pump impeller. Once a quarter you place the sensors at the marked points, do one run, and immediately know where and how much to add.

### Where the savings actually are

Not in calculation minutes. Every trial run is a coast-down, a cool-down, removing the cover, fastening the weight, restarting. That's exactly what you're cutting out.

| Mode | One plane | Two planes |
| --- | --- | --- |
| New rotor, calibration runs | 2 (initial + trial) | 3 (initial + two trials) |
| Saved coefficients (Saved coeff.) | 1 (initial) | 1 (initial) |
| Check run on top | at least one | at least one |

## What has to match for the coefficients to work

The condition about the sensors gets its own separate warning in the manual, and that isn't a formality. The influence coefficient is measured for a specific pair — 'measurement point and correction plane.' Move the sensor from horizontal to vertical, or shift it from the bearing housing to the foot, and you're now measuring a different 'weight-to-vibration' relationship, one the old coefficients don't fit.

- [x] The same machine and the same rotor. Not 'the same type' — this exact one.
- [x] The same vibration-sensor mounting points on the bearing housings, and the same measurement direction.
- [x] The same phase sensor and reflective marker in the same place on the shaft.
- [x] The same correction planes and the same weight-installation radii.
- [x] The same operating mode: speed, load, damper or vane position, warmed-up state.
- [x] The same weight-fastening scheme: on the blades, on bolts, or welded; the same number of fixed positions and the same zero at Z1.
- [x] The trial-weight mass from the first balancing run was entered in grams, not as a percentage.

> If even one item has changed, the influence coefficient has changed with it. The software won't see that — it will honestly calculate from the old numbers. The error will show up as an inexplicably large correction, after which the vibration either doesn't drop or goes up. In that situation, don't try to fix it with trim weights. Go back into New rotor mode and recalibrate — it's cheaper than three wasted runs.

## The first balancing, done 'for the future': six steps

1. **Enter the trial-weight mass in grams** — Percent mode calculates the correction as a percentage of the trial mass and doesn't give you numbers you can reuse later. Weigh the weight and enter the actual grams. Scales are included in the kit.
2. **Install the trial weight exactly at the tachometer marker** — The key condition for reuse: the angular position of the trial weight must match the position of the reflective marker. That way the zero reference is tied to the marker, not to some random point you won't be able to find a year later.
3. **Record the actual installation radius** — You need the radius both now and later: the software uses it to calculate the initial and residual unbalance and to check against tolerance. Enter the radius the weight is actually sitting at, not the wheel's nominal diameter.
4. **Physically mark the sensor points** — A center punch, paint mark or sticker on the bearing housing, plus an arrow for the measurement direction. Photograph both sensors and the phase sensor with the shaft marker.
5. **Save the coefficients under a clear name** — On the Result tab, open the influence-coefficient view and save it to the archive, entering a meaningful name in the Rotor column. 'Fan VDN-9 #3, drive-side bearing' works; 'fan1' won't, a year from now.
6. **Save the report and the run history** — The archive holds the results of every run with timestamps and lets you generate a report in the built-in editor. Calculate the residual-unbalance tolerance by grade G, stating the applicable part and edition of ISO 21940 right in the report.

Sources: [ISO 21940-11:2016](https://www.iso.org/standard/54074.html) · [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/)

## Nowhere to put the weight: four workarounds

### Recalculating for other correction planes

The calculated plane is covered by a housing, falls on a mounting surface, or simply has no metal to work with. The plane-switching function gives you four typical layouts of original-to-new plane positions you can move the mass onto. Enter the distances between the planes and the new installation radii, and the software recalculates the masses and angles. Especially useful on complex-shaped rotors, such as crankshafts.

### Splitting between fixed positions

Set the number of positions (12 holes means a 30° step), and the software splits the calculated mass across two neighboring positions and gives you their numbers, say Z10 and Z11. The count runs in the direction of rotation from Z1, where the trial weight sat. No protractor needed, and a mirrored angle error is ruled out.

### Drilling instead of adding

Switch the mode from adding to removing material. The angle automatically flips 180°: metal is removed from the opposite side. From there, watch the depth, wall thickness and strength — what's removed can't be put back.

### The trial weight can't be removed

Welded on, trapped under a cover, access closed off after reassembly. Turn on the option that leaves the trial weight in the plane, and the software recalculates the correction mass and angle accounting for its mass.

> Any of these paths changes the radius or the plane, which means it changes the influence coefficient going forward too. If you plan to work from saved coefficients, record in the archive the plane and radius where the weight actually sits, not the ones the first calculation used.

## An archive you can come back to a year later

What's in the instrument's archive is enough for the calculation, but not always enough to reproduce the measurement conditions. Those conditions live in your own form and in photographs. Set up one page per machine and keep the following in it.

- Carrying coefficients over to an identical but different machine. Two identical fans on neighboring foundations respond differently to the same weight: different anchors, different frame stiffness, different wear. The result is a correction that looks right but leaves the vibration exactly where it was.
- Changing the radius without recalculating. The weight moved to a different radius, but the settings still show the old one. The correction's effect scales with the ratio of the radii, and you get a systematic under- or over-correction.
- A change of direction or sensor point. Horizontal-radial instead of vertical, a foot instead of a bearing housing, a magnet on paint instead of a stud on a clean surface. The phase shifts, and the weight lands in the wrong place.
- A change of speed. Balanced at 1480 rpm, came back at 990 rpm via a variable-frequency drive. The coefficients don't carry over; a new calibration is needed.
- Trial-weight mass entered as a percentage. The calculation goes through and you get a correction, but the coefficients in the archive will be useless next time.

- [x] The balancing speed and the machine's operating mode: load, damper position, warm-up state.
- [x] The vibration-sensor mounting points and measurement direction, with a photo and an arrow.
- [x] The location of the phase sensor and the reflective marker on the shaft.
- [x] The correction planes and the installation radii in millimeters.
- [x] The trial-weight mass in grams and its angular position relative to the marker.
- [x] The correction-weight fastening scheme: welded, bolted, on a blade; the number of fixed positions and where Z1 is.
- [x] The target residual 1x value and the actually achieved amplitude and phase.
- [x] The achieved residual unbalance in g·mm, and the grade G if you calculated it.
- [x] Photographs of the installed weights tied to their position numbers.

## What you can hand off to AXILINE's engineers

We design and build Balanset instruments and use them to balance on site ourselves. That's why we set up the first visit so the second one can build on it: we mark the sensor points, install the trial weight at the tachometer marker, record the radii and planes, and save the influence coefficients under a clear machine name. After that, a routine visit to the same machine is one run, installing the weight, and a check.

The second path is to get a Balanset-1A kit of your own. Two vibration sensors, a laser phase and speed sensor, a two-channel USB module, and Windows software that handles balancing in one and two planes, trim add-on correction, an influence-coefficient archive, splitting the weight across fixed positions, a drilling calculation, recalculation for other correction planes, an FFT spectrum, and reports. Experienced engineers provide consulting support for your machines in this case too.

Whichever you choose, keep the two results separate. The target value in the software is about the residual running-speed component. Assess the machine's overall condition from broadband vibration at the bearing housings, comparing it against the ISO 20816 criteria for the applicable part and edition, and recording the measurement points, frequency band, operating mode and type of supports.

Sources: [Balanset-1A manufacturer specification](https://vibromera.eu/product/balanset-1/) · [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/) · [ISO 20816-1:2016](https://www.iso.org/standard/63180.html)

## Frequently asked questions

**How many trim runs after a correction count as normal?**

One, sometimes two. One run clears out installation errors on the weight; a second is needed where mass can only go on blades or in coarsely spaced holes. If you're doing a third and fourth and the vibration keeps dropping more slowly each time, stop and check the entered radius and mass, support tightness, speed stability and how close you are to resonance.

**Can saved coefficients be used on a second, identical machine?**

No. The coefficient describes not the rotor, but the system: 'rotor — supports — foundation — operating mode.' Two identical units on different foundations respond differently to the same trial weight. Run a full calibration for the second machine and save its own record in the archive.

**After a trim weight, vibration went up instead of down. What now?**

First check the angle-reference direction and the add-or-remove-metal mode: with drilling, the angle flips 180°, and a mixed-up flag produces exactly that effect. If the angle is correct, remove the added mass and look for a loosened fastening, damaged bearings or resonance. Beyond that, more trim weights won't help.

**Does the trial weight need to come off before installing the correction weight?**

By default, yes: stop the rotor, remove the trial weight, and only then install the calculated one or remove material. If the trial weight can't be removed, turn on the option in the settings to leave it in the plane, and the software recalculates the correction mass and angle to account for it.

**Speed changed by three percent. Are the coefficients still good?**

Away from resonance, a small drift is usually tolerable, and the first trim run will pick it up. But if you're operating near a natural frequency, even a hundred rpm can change the amplitude several times over and shift the phase by tens of degrees. Check where the resonance sits from a coast-down, and balance at a stable speed outside that zone.

**Does trim balancing replace the check run?**

The opposite — it grows out of one. A check run after installing the weights is mandatory: it gives you the residual 1x, the phase, and the residual unbalance. Trim is what the software suggests based on that run's result, if you haven't reached the target value yet.