Alignment, Balance & Vibration

Every rotating machine vibrates. The question is never whether, it is at what frequency and how much — and the first of those is the one that names the fault.

Frequency identifies, amplitude ranks

This is the idea the chapter is built on. Mechanical faults excite vibration at frequencies that are characteristic of the fault, and almost all of them are tied to the machine's own running speed. So if you know the running speed, you know where to look, and the peak that stands out tells you what is wrong.

Amplitude tells you how bad it is, not what it is. A large 1× peak and a large 2× peak are the same severity and completely different jobs.

That is why "it vibrates a lot" is not a diagnosis and never becomes one by being measured more precisely.

Why this chapter follows Chapter 12

Chapter 12 kept arriving at the same two answers — misalignment and unbalance — and deferring them here. They deserve their own chapter because between them they cause most of the bearing deaths on a plant, they are both invisible to the eye at the magnitudes that matter, and they are both entirely preventable with a measurement that takes half an hour.

Reading a spectrum

The distinction worth carrying out of the diagram is between synchronous and non-synchronous content:

• Peaks on whole multiples of running speed come from the shaft and what is attached to it — unbalance, misalignment, looseness, blade pass, gear mesh.
• Peaks that are not whole multiples come from inside a bearing, because the rolling elements orbit at a rate that has no tidy relationship to shaft speed.

That single observation separates "something is wrong with how this machine is assembled" from "a bearing is failing", and you can often make it by eye.

The axial reading is the discriminator

Unbalance and misalignment both produce strong 1× vibration, so a radial measurement alone will not always separate them. What does separate them is axial vibration — along the shaft.

Unbalance is a rotating force acting outwards; it produces little axial vibration. Misalignment is a coupling being forced to accommodate a geometric error on every revolution, and a large part of that ends up along the shaft. So high axial vibration, particularly at 2×, is the misalignment signature.

Measure at the bearing housings, in three directions — horizontal, vertical and axial — and the axial one is the reading people skip and the one that decides the diagnosis.

Alignment

Three things decide whether an alignment job holds:

Soft foot, before anything else. If one foot does not make contact, tightening its bolt pulls the frame down and distorts the machine. Align it first and you have aligned a shape that does not exist once it is bolted down. This is why "we aligned it and it is still out" is such a common report.

Both kinds of misalignment, together. Offset and angular almost always occur at the same time, and correcting one while ignoring the other leaves the coupling working just as hard.

Thermal growth. A machine aligned perfectly cold will be out when it reaches operating temperature, because the components grow by different amounts. Manufacturers publish target offsets to align to when cold, so the machine is right when hot. Aligning a hot machine to zero is aligning it wrong for every cold start.

Methods run from a straightedge across the coupling faces — which catches gross errors in seconds and is worth doing every time — through dial indicators, to laser alignment, which is faster and removes most of the arithmetic.

Balance

Unbalance means the mass is not distributed evenly about the axis of rotation. The resulting force grows with the square of speed, which is why a fan that was acceptable at 50 Hz can be unusable at 60 Hz, and why unbalance faults appear when a machine is sped up.

Causes are usually acquired rather than built in: product build-up on a fan impeller, erosion or corrosion, a lost balance weight, a repair that added material unevenly. Cleaning a fan is a balancing operation whether anybody intends it or not.

Resonance, and why drives made it common

Every structure has natural frequencies. Excite one with a forcing frequency that matches, and a small input produces a large motion.

Fixed-speed machines either had this problem from commissioning or never had it. Variable-speed drives changed that, because a machine that now runs at any frequency between 10 and 60 Hz will pass through — or sit on — resonances that nobody found during commissioning.

The signature is unmistakable once you look for it: vibration that is severe at one particular speed and acceptable above and below it. This is exactly what Chapter 6's skip frequencies are for, and it is the correct fix when the resonance cannot be designed out.

Measuring, and what the numbers mean

General machinery vibration is usually measured as velocity in in/s RMS, because over the usual frequency range it correlates reasonably with damage. Standards such as ISO 10816 give severity zones by machine size and mounting, which is useful as a sanity check.

But the number that matters most is your own machine's history. A reading of 0.18 in/s means little on its own; the same point reading 0.07 in/s last quarter and 0.18 in/s today is a machine deteriorating, and that comparison is available to anyone who wrote the first number down.

1. Establish whether this is new, or has grown

   The most valuable question and often the hardest to answer, because it depends on somebody having written a number down previously. A machine that has always run at 0.16 in/s and still does is a different problem from one that has doubled since spring.

2. Record the conditions with the reading

   Speed, load, temperature, what the machine was doing. Vibration readings are comparable only against readings taken under the same conditions, and a measurement without its context cannot be compared with the next one — the same principle as the thermal survey in Chapter 4.

3. Measure at the bearing housings in all three directions

   Safety

   Horizontal, vertical and axial, at each bearing. The axial reading is the one routinely skipped and the one that discriminates misalignment from unbalance, so it is worth the extra thirty seconds every time.

4. Identify the dominant frequency relative to running speed

   Work out the running speed first, then locate the biggest peak in relation to it. 1× dominant and radial suggests unbalance; 2× with high axial suggests misalignment; a run of harmonics suggests looseness; non-synchronous content suggests a bearing.

5. Check for soft foot before touching the alignment

   Loosen each foot in turn and watch for movement. Correcting alignment on a machine that rocks produces a job that will not hold, and the report that comes back — "it is out again" — costs the second visit as well as the first.

6. If it is speed-dependent, suspect resonance

   Severe at one speed and acceptable either side is the signature. On a drive-fed machine, a coast-down or a slow speed sweep will show it clearly, and skip frequencies are the standard remedy.

7. Fix the cause, then measure again and write it down

   The post-repair reading is the new baseline and it is worth more than the repair itself, because every future diagnosis on that machine is a comparison against it. A machine with a history is diagnosable; a machine without one is guesswork every time.

• Balancing a machine that is misaligned

  Why

  The 1× peak is large, unbalance produces a large 1× peak, and balancing is a satisfying job with a measurable result. But misalignment also produces 1×, and weights added to compensate for a misalignment force make the machine worse in every other condition. The axial reading is what would have separated them, and it takes thirty seconds.

• Aligning without checking soft foot

  Why

  Soft foot is invisible, checking it takes ten minutes, and the alignment readings look perfectly convincing without it. The result is a job that measures correct on the day and is out again within weeks, which then gets attributed to the coupling or the base rather than to the step that was skipped.

• Aligning cold to zero

  Why

  Zero is obviously the right answer and any deliberate offset looks like an error. But components grow at different rates as the machine warms, so the manufacturer's cold target exists precisely so the machine is aligned when it is hot — which is when it spends its life.

• Judging a reading against a standard instead of against history

  Why

  A published severity zone is genuinely useful and it is a statement about machines in general. Your machine's own trend is a statement about your machine, and a value inside the acceptable zone that has tripled since spring is far more informative than one that has always sat where it is.

• Skipping the axial reading

  Why

  It is the awkward one to take — often the least accessible face on the bearing housing — and two readings feel like enough. It is also the measurement that decides between the two commonest faults in the chapter, which means skipping it turns a diagnosis into a coin toss.

• Tightening the bolts to stop the vibration

  Why

  It works, briefly, and it is free. But looseness is usually a symptom rather than a cause: something is applying a load the fixings were not sized for, and a bolt that keeps coming loose is reporting that load faithfully. Re-torquing it silences the report.

Five questions. Each is answerable by reasoning about which frequency a fault lives at, rather than by recalling a severity table.