Ultrasonic detector test for centrifugal impeller of air compressor

Last spring, a maintenance lead at a gas processing plant called us in a hurry. Their three-stage integrally geared air compressor had tripped on high shaft vibration. When the rotor came out, the second-stage centrifugal impeller had a thumb-sized chunk missing from the back shroud. The fracture surface showed beach marks — a classic fatigue crack that had grown from a subsurface porosity cluster nobody caught during receiving inspection. The plant had relied on a certificate that said “UT passed.” What the certificate didn’t reveal was that the supplier had scanned only the bore area with a straight-beam probe and never touched the blade roots, where the real stress lives.

That failure turned into a six-figure repair bill and two weeks of downtime. It also changed how the plant’s quality, procurement, and maintenance teams thought about ultrasonic testing. If you buy, inspect, or keep centrifugal impellers spinning for a living, this is the article you’ll want your team to read before the next shipment arrives or the next overhaul starts.

The flaws you can’t see are the ones that ground a compressor

Centrifugal impellers for air compressors are not simple discs. They are highly stressed rotating components that operate at tip speeds often exceeding 400 m/s. A closed impeller might be milled from a single forging, cast in 17-4PH stainless steel, or fabricated by welding blades to a hub and shroud. Each manufacturing route comes with its own internal defect profile: shrinkage and gas porosity in castings, non-metallic inclusions or bursts in forgings, lack of fusion or toe cracks in welded assemblies. Once in service, corrosion pitting, erosion, and high-cycle fatigue add a second layer of risk.

Visual inspection cannot peer beneath the surface. Radiography on a thick-section impeller is slow, requires access to both sides, and often misses tight cracks unless perfectly oriented. Magnetic particle and dye penetrant are surface-only. Ultrasound remains the only practical volumetric method that you can use at incoming inspection, in the machine shop, and even on an assembled rotor during a maintenance window — provided your team knows how to do it right.

Why an ultrasonic detector makes sense for your impeller

An ultrasonic flaw detector sends a short pulse of high-frequency sound into the metal and listens for echoes. A clean back-wall reflection tells you the material is solid. An echo arriving before the back wall tells you something is inside that shouldn’t be. For centrifugal impellers, you typically work with frequencies between 2 MHz and 5 MHz. Lower frequencies penetrate attenuative or coarse-grained materials better; higher frequencies give you the resolution to spot millimetre-sized flaws near the surface.

The real value lies in how you apply the technique. On a forged 17-4PH impeller, a normal-beam dual-element probe can pick up near-surface laminations or stringers that would concentrate stress right where the blade root meets the shroud. On a welded impeller, angle-beam shear-wave probes let you hunt for lack of fusion along the weld prep faces — defects that a quick “UT passed” scan along the easy-to-reach hub face would completely overlook. A single missed signal is all it takes for a crack to grow under cyclic loading. Over the years, our team has learned to treat every impeller geometry as a separate puzzle that demands its own scan plan.

Setting up a test procedure your QC team can trust

Good ultrasonic testing starts long before you put the probe on the metal. For a centrifugal impeller, you need to answer four questions: where to scan, what probe to use, what reference standard to calibrate against, and what acceptance level to enforce.

Begin by identifying the highly stressed zones: blade leading and trailing edges, the fillet radii where blades meet the hub and shroud, the bore area around the shaft fit, and any weld seams if the impeller is fabricated. On an open-face impeller without a shroud, you can often place a straight-beam probe on the blade tips and scan through the blade length toward the hub. For closed impellers, you may need a combination of normal-beam scans from the inlet eye and shroud surfaces, plus low-angle longitudinal or shear-wave probes that sneak into the tight radius transitions.

Calibration is where many supplier reports fall apart. A 0.5-inch diameter, 2.25 MHz probe calibrated on a generic carbon steel V1 block will not give you reliable sensitivity on a stainless steel impeller. You need a reference block made from the same material — or at least one with the same acoustic velocity and attenuation characteristics — containing flat-bottom holes or side-drilled holes at depths covering the sound path distances you intend to inspect. For a typical impeller hub thickness of 40 to 80 mm, a set of 1.5 mm and 3 mm flat-bottom holes works well. Build a distance-amplitude correction (DAC) curve on the flaw detector screen, and you instantly have a repeatable sensitivity baseline. If a signal breaks that curve during the scan, it gets logged, sketched, and evaluated.

What procurement managers should write into the purchase order

If you are responsible for sourcing centrifugal impellers, your leverage is largest before the contract is signed. A vague note saying “UT required” on the drawing means almost nothing. A detailed specification, on the other hand, forces the supplier to assign competent technicians and document the results.

Here’s a clause we have seen work in the real world:

*“Each finished impeller shall be 100% ultrasonically inspected per ASTM A609 (for castings) or ASTM A388 (for forgings) with supplementary zone scans at all blade-to-hub and blade-to-shroud fillet radii. The supplier shall use a calibration block of identical material grade and heat treatment condition, containing 1.5 mm and 3.0 mm diameter flat-bottom holes. Scan sensitivity shall be set so that the smallest reference hole produces a signal of 80% full-screen height, and the DAC curve shall be recorded. Indications exceeding 50% of the reference level shall be reported with coordinates, amplitude, and estimated size. The inspection report must include probe frequency, diameter, wedge details, couplant, scan coverage map, and the technician’s ISO 9712 or ASNT Level II certificate number.”*

You don’t need to be an NDT expert to spot a thin report. If the supplier hands you a one-page document listing only “no defects found” without scan directions, reference hole sizes, or probe data, treat it the same way you’d treat a material certificate without heat numbers. Ask for the complete data package. If they can’t provide it, commission a third-party inspection before accepting delivery. The cost of a day’s UT work is trivial compared with an impeller burst inside a high-speed air compressor.

Inspection during maintenance — keeping compressors running

Maintenance crews inherit impellers that may have run for 20,000 hours in damp, acidic, or dusty environments. The question shifts from “is the part free of manufacturing defects?” to “has service caused any cracks or wall loss?” A portable digital ultrasonic flaw detector with a good data logger becomes your go-to tool here.

Before pulling the rotor, you can sometimes access the first-stage impeller through the inlet piping. If you have line-of-sight, a small dual-element probe on an extension handle lets you check the blade leading edges for erosion thinning and crack indications without a full teardown. Once the rotor bundle is on the bench, clean the impellers thoroughly — oil, scale, and old coating must be removed to Sa 2.5 in the areas to be scanned, because even a 0.1 mm air gap under the probe can kill the sound transmission.

Pay special attention to areas where service conditions promote cracking: the blade root on the back shroud side of high-pressure stages, the impeller eye near the shaft fit where fretting may initiate a crack, and any areas that show pitting or discoloration. On closed impellers, corrosion can eat away the internal passages you cannot see; scanning from the outside surface with a normal-beam probe looking for double echoes or loss of back-wall reflection often reveals internal wall loss long before a leak appears.

When a signal lights up the screen during a maintenance scan, don’t immediately call it a defect. Impeller geometry creates reflections that look alarming. A blade fillet, a machined step, or the edge of the bore will return a clean, position-predictable signal that rises and falls smoothly as you move the probe. A crack or a void tends to produce a sharp, jittery echo that peaks over a short scan distance and then disappears. If you sweep the probe in an arc and the echo remains locked to a local spot on the part rather than following the geometry, you are almost certainly looking at a real flaw. When in doubt, reproduce the indication from another surface. A crack that appears from both the shroud side and the blade side is not a geometry trick.

Acceptance criteria that keep you out of trouble

What do you do with the indication once you find it? For new impellers, we default to the relevant ASTM standard: ASTM A388 for steel forgings generally rejects any indication that exceeds the back-reflection amplitude or causes complete loss of back reflection, along with any signal larger than the DAC reference hole in a critical zone. For castings per ASTM A609, the acceptance level depends on the quality class specified — Class 1 (highest) allows only isolated indications smaller than the reference reflector, while lower classes permit some distributed porosity.

For in-service impellers, the criteria shift. A tiny subsurface inclusion that was acceptable when new may become a stress raiser after years of vibration. Many maintenance teams adopt a pragmatic approach: record the indication amplitude, depth, and coordinates, measure the wall thickness at that location, and track it during the next planned outage. If the signal grows by 3 dB or more, plan a repair or replacement. Surface-connected cracks detected by shear-wave or surface-wave probes generally call for immediate action — either blend out the crack by machining if allowable, or scrap the impeller if the remaining wall thickness drops below the minimum design value.

One practical note: always involve the compressor OEM or a qualified engineer when applying acceptance criteria. The critical areas defined by FEA stress analysis may demand tighter rejection limits than the generic standard. It is not unusual to require a “no recordable indication” zone extending 10 mm either side of the blade-to-shroud fillet on the trailing edge of a high-speed centrifugal stage. Write that into your procurement spec and your maintenance procedure alike.

Building ultrasonic testing into your impeller life cycle

The ultrasonic detector test is not a one-time checkbox. It is a thread that connects the supplier’s foundry, your incoming QC bench, and the maintenance shop floor. When procurement specifies a detailed UT report, the QC team verifies it, and the maintenance crew repeats the scan during every overhaul, you create a timeline of the impeller’s internal health. A small casting flaw that stays stable for five years is one thing; the same flaw growing into a crack is a warning you can act on.

We have seen sites that file a printed A-scan envelope alongside the vibration and lube analysis reports in the compressor history folder. When something changes — a new frequency peak in the vibration spectrum, a drop in efficiency — the UT record gives the troubleshooter an immediate clue. It answers the question “has the impeller changed internally?” without guesswork.

If you are evaluating a new supplier, ask to witness a UT scan on one of their impellers. Watch where they place the probe and whether they cover the blade transition zones. A shop that truly understands centrifugal compressor impellers will have a scan plan mapped onto the drawing, not just a generic work instruction. That attention to detail carries over into the machining, balancing, and mechanical integrity of the entire rotor.

Take the ultrasonic flaw detector out of its case, calibrate on the right block, and trace the geometry of each impeller as if a crack is already hiding somewhere. That mindset — shared across quality, procurement, and maintenance — keeps air compressors turning and prevents the kind of phone call nobody wants to receive on a Saturday morning.