Ultrasonic Flaw Detection for Forging Parts

• 1
  How ultrasonic flaw detection for forging parts finds a defect
• 2
  What UT can and cannot see in a forged part
• 3
  Turning an indication into an accept or reject decision
• 4
  Where the procedure usually breaks down
• 5
  Fitting UT into a CNC production sequence
• 6
  Alloy behavior that changes the test
• 7
  What to put on the drawing
• 8
  Choosing a test method for forged parts

A piezoelectric probe fires a short pulse of 2–10 MHz sound into the part. The wave travels until it meets a change in acoustic impedance: a void, a crack, a slag stringer, or the back wall. Part of the energy reflects, returns to the probe, and the instrument plots amplitude against time. Time converts to depth once you enter the sound velocity for that alloy. For steel it sits near 5,900 m/s; for aluminum near 6,300 m/s; for titanium near 6,100 m/s.

Forging makes this harder than it sounds. The grain flow that gives a forged connecting rod its fatigue life also scatters sound. Coarse grains act like a fog of small reflectors, raising the noise floor until a 1 mm pore disappears into it. That is why forged and cast material usually needs a lower test frequency and a larger focused probe than a rolled plate of the same alloy.

The classic setup is a dual-element or single-element probe coupled to a machined face with water, gel, or oil. Immersion tanks and squirter systems trade the manual couplant for a repeatable water path. Either way, the surface finish under the probe matters: a band-sawn face at Ra 12.5 μm traps couplant in the valleys and lets it drain from the peaks, so the echo amplitude drifts from pass to pass.

What comes back is not a picture. A-scan gives amplitude versus time for one line. B-scan stacks those lines into a cross-section. Phased-array (PAUT) steers and focuses the beam electronically, so a single probe sweeps a sector and builds a C-scan map of the volume. That map is what lets an engineer say "the indication sits 6 mm below the machined face, 14 mm long, 2 mm from the bore" instead of "there is something in there."

• 1
  Frequency2–5 MHz for coarse forged grain, 5–10 MHz for fine-grain or thin sections.
• 2
  CouplantWater, gel, or oil. Keep the path constant or amplitude drifts.
• 3
  VelocityEnter the alloy value or depth readings shift by several millimeters.
• 4
  Reference blockSame alloy and heat treatment as the part, with flat-bottom holes of known size.

Ultrasonic testing is strong on volumetric defects that lie roughly perpendicular to the beam: internal voids, shrinkage porosity, forging bursts, and large non-metallic inclusions. It is also good at finding planar cracks that face the sound path, such as fatigue cracks growing from a fillet or a quench crack under a hardened case.

It is weak on defects parallel to the beam. A tight lamination or a flat crack oriented along the sound path returns almost nothing, because the reflection goes sideways rather than back to the probe. A single normal-incidence scan can miss a defect that a 45° angle beam would light up. This is why a real procedure uses more than one angle on any part with a critical section.

Geometry sets the limit. A forging with a 200 mm thick hub needs low frequency and high energy to reach the center, which cuts resolution. A thin rib 8 mm thick needs high frequency to separate the back-wall echo from a defect echo, which cuts penetration. There is no single setting that covers both, so a complex forging gets scanned in zones with different probes.

Surface condition and access decide whether the test is even possible. A rough as-forged surface, a deep die draft, or a bore you cannot reach with a probe all block the beam. Machining a flat inspection land, or scanning after the first CNC operation, often makes the difference between a real inspection and a paper one.

• 1
  Good targetsVoids, porosity, bursts, inclusions, cracks facing the beam.
• 2
  Poor targetsTight laminations, cracks parallel to the beam, defects under rough scale.
• 3
  Depth trade-offLow frequency penetrates deeper but resolves less.
• 4
  AccessNo probe contact, no data. Plan inspection lands early.

An amplitude reading alone means little. What matters is the equivalent reflector size, and that comes from comparing the echo against a reference block. A flat-bottom hole of 2 mm diameter in the same alloy gives a known amplitude at a known depth. If the part echo matches or exceeds it, the indication is written as FBH 2 mm equivalent or larger.

Standards give you the acceptance language. ASTM A388 covers heavy steel forgings and uses back-reflection loss plus reference-hole comparison. AMS 2630 and AMS 2631 cover premium aerospace forgings with tighter class limits. ASTM E2375 covers wrought products including forgings and sets practice for the examination itself. Your drawing should name one of these, not just say "UT per standard."

Distance-amplitude correction matters. Sound attenuates as it travels, so an identical defect at 20 mm and at 180 mm returns different amplitudes. The instrument applies DAC or TCG to flatten that curve. Without it, you reject shallow defects that are harmless and accept deep ones that are not. Ask the supplier to show the DAC curve on the report.

Finally, decide what happens to a reject. A 3 mm indication in a non-critical boss may be acceptable under a written deviation. The same indication in a high-cycle fatigue fillet is not. That call belongs to the design engineer, and it should be made before production, not after a failed pressure test.

• 1
  Equivalent sizeCompare against a flat-bottom hole, not against the noise floor.
• 2
  Back reflectionA 50% loss is a red flag even if no discrete echo appears.
• 3
  DAC / TCGRequired to compare defects at different depths fairly.
• 4
  DispositionWrite the accept/reject rule before the run starts.

Most failed UT programs fail on setup, not on physics. A probe with the wrong focal length, a couplant that dries mid-scan, a velocity value left at the default 5,900 m/s for a titanium part: each of these shifts the depth reading and can move a real defect out of the gate. The fix is a written scan plan with the probe, frequency, angle, and gate position stated per zone.

Calibration drift is the second trap. The reference block must be the same alloy and heat treatment as the production part. A 4140 block used to set up a 17-4PH forging will read amplitude differently because attenuation differs. Re-calibrate at the start and end of each shift, and after any probe change.

Operator certification is the third. ASNT SNT-TC-1A Level II is the usual baseline for forging inspection. A Level II can set up, calibrate, and interpret under a written procedure. A Level I can scan but not judge. If a supplier cannot name the certification level of the person who signed the report, treat the report as unverified.

The last gap is documentation. A report that says "no defects found" is not useful. You want the scan plan, the calibration record, the DAC curve, the probe details, and a C-scan image where phased array was used. That package is what lets a second engineer review the call six months later.

• 1
  Scan planProbe, frequency, angle, and gate per zone, in writing.
• 2
  Reference blockSame alloy and heat treatment as the part.
• 3
  CertificationASNT SNT-TC-1A Level II for setup and interpretation.
• 4
  RecordsCalibration, DAC curve, probe data, and C-scan images.

The cheapest place to inspect a forging is before it becomes an expensive machined part. If UT runs on the raw forging, a reject costs you the forging price. If it runs after 12 hours of 5-axis work, a reject costs the forging plus the machining plus the schedule slot. For a part with a 3–5 day lead time, that lost slot hurts more than the metal.

In practice, the sequence that works is: UT the raw forging on accessible faces, machine the part, then run a final UT or magnetic particle check on the finished critical zones. The first pass catches gross internal defects early. The second pass catches cracks introduced by machining, heat treatment, or handling.

Machining also creates better test surfaces. A face milled to Ra 0.8–1.6 μm gives a clean, flat inspection land that a rough forged surface never will. On parts with tight tolerance, that land doubles as a datum for the CNC setup, so the inspection feature costs almost nothing extra.

For phased-array work, the C-scan data can be tied to the part serial number and stored with the inspection report. If a field failure ever occurs, you have a record of what the part looked like internally on the day it shipped. That is the real value of integrated UT: it turns an inspection step into a traceable quality record.

• 1
  Inspect earlyRaw forging reject costs far less than a finished-part reject.
• 2
  Two-passPre-machining UT plus final surface or near-surface check.
• 3
  Machined landsRa 0.8–1.6 μm faces give stable coupling and double as datums.
• 4
  TraceabilityStore C-scan data against the serial number.

Not every forging tests the same way. Aluminum forges at lower temperature and tends to have finer grain, so higher frequencies work and resolution is better. A 6061 or 7075 part can often be tested at 5–10 MHz with good sensitivity to small defects. The trade-off is that aluminum is softer, so probe wear and surface damage need attention.

Stainless steels are more forgiving on corrosion but less forgiving on sound. Austenitic grades such as 304 and 316 have a coarse, sometimes columnar grain structure that scatters sound badly. Testing these with conventional UT is difficult; a lower frequency and a dual-element probe help, but the noise floor stays high. Martensitic grades such as 17-4PH and 440C behave more like carbon steel and test cleanly.

Titanium forges well but attenuates sound more than steel. Ti-6Al-4V is the common aerospace grade, and it is usually tested with a focused probe and careful gain settings. The material itself is not the problem; the alpha-beta structure and the occasional alpha case at the surface are what complicate the reading.

Nickel alloys such as Inconel are the hardest of the common forging materials to inspect. They attenuate heavily, and a small defect deep in a thick section may sit below the noise floor no matter what probe you use. For these parts, radiography or a destructive test on a witness coupon often carries more weight than UT alone.

• 1
  AluminumFine grain, high frequency works, watch probe wear.
• 2
  Austenitic stainlessCoarse grain scatters sound; low frequency and dual element.
• 3
  TitaniumModerate attenuation; focused probe and careful gain.
• 4
  Nickel alloysHeavy attenuation; UT alone may not be enough.

A drawing that says "UT per ASTM A388" leaves too much open. Add the acceptance class, the reference hole size, the scan zones, and the required report contents. For a rotating part, name the critical zone and the maximum allowable indication. For a static bracket, you can be looser, but you still need a number.

State the inspection stage. Pre-machining, post-machining, or both. If the part gets heat treated after machining, note whether UT runs before or after, because heat treatment can open or close defects and can change the surface condition.

Name the reference standard for the examination practice, not just the acceptance criteria. ASTM A388 tells you how to test heavy steel forgings. ASTM E2375 covers the practice for wrought products. AMS 2630 and AMS 2631 give the aerospace class structure. These are different documents and they are not interchangeable.

Finally, require the record package. Scan plan, calibration record, DAC curve, probe details, operator certification level, and C-scan images if phased array was used. If the drawing asks for the package up front, the supplier builds the process around it instead of reconstructing paperwork afterward.

• 1
  Acceptance className the class or the maximum indication size.
• 2
  Inspection stagePre-machining, post-machining, or both.
• 3
  Practice standardA388, E2375, AMS 2630, or AMS 2631 as applicable.
• 4
  Record packageScan plan, calibration, DAC curve, certification, C-scan.