15.1 Evaluating Bonded Structures

The ASNT NDT Level II ultrasonic general exam lists Evaluation of Bonded Structures as official UT topic 4 on the outline administered prior to 5 February 2027. After that date ASNT's revised UT general exam rolls to four domains. This chapter still teaches the current six-topic outline. Bond items are not asking whether you can spell honeycomb. They are asking which technique interrogates a bond line, what a disbond does to the A-scan, and which joints ultrasound can miss even when the setup looks textbook.

A bonded structure in this topic is any stack whose serviceability depends on an interface: adhesive joints, brazed or soldered joints, explosion-bonded or roll-bonded clad, diffusion bonds, weld overlay treated as a bond, and sandwich construction such as honeycomb core with metallic or composite facings. The acoustic question is the same on every product: does sound cross the interface, or does it reflect from it?

Pulse-echo versus through-transmission

Two families dominate bond work. They answer different questions and fail in different ways.

Pulse-echo is a reflection method. The instrument times and amplitudes echoes that return to the same face. If two metals of similar acoustic impedance are well bonded, the interface is almost invisible. Energy continues into the second layer and returns from the far surface. If an air gap opens at the bond, the metal-to-air (or composite-to-air) impedance mismatch is enormous. Nearly all energy reflects at the gap. The interface echo grows; the far-side echo disappears. That pair — gain of interface, loss of back-wall — is the classic pulse-echo disbond signature.

Through-transmission is an energy-crossing method. A good path from sender to receiver means a high received pulse. A disbond, a crushed core, or a void in the adhesive interrupts the path, so received amplitude falls. Through-transmission does not give a useful depth from time-of-flight the way pulse-echo does. It tells you the path is blocked, not which interface failed if several exist. Its strength is area coverage and sensitivity to anything that stops transmission, including cores that pulse-echo from one face may not present as a clean back-wall.

Do not treat the two methods as interchangeable. Pulse-echo needs a resolvable interface or a back-wall to interpret. Through-transmission needs two-side access and a receiver that stays aligned. A honeycomb panel with only one face reachable is a pulse-echo or pitch-catch problem, not a through-transmission problem. Pitch-catch (separate send and receive on the same face) is a useful middle case on sandwich when the opposite face is closed; it is still not a depth-accurate substitute for a calibrated pulse-echo range.

Honeycomb and sandwich

Honeycomb is a facing–core–facing stack. The bond that usually matters is facing-to-core — the adhesive fillet at the cell walls. From the facing, a pulse-echo probe sees a thin plate sitting on a periodic core.

• Bonded facing. Part of the pulse energy leaves the facing into the adhesive and core. The facing does not ring for long. You may see a modest first interface and a damped multiple.
• Disbonded facing. The facing becomes a free plate over air. The same pulse reverberates in the facing thickness. The A-scan shows a train of multiple echoes spaced at the facing's thickness time-of-flight. Amplitude of those multiples is high because little energy leaks into the core.
• Through-transmission. A good panel delivers a received pulse through facing, core, and opposite facing. A facing disbond, a water-filled cell field, or a crushed core drops that pulse.

Water in the core, crushed cells, and adhesive voids are related but not identical calls. Water can increase through-transmission in some cells and change pulse-echo ringing. The procedure, not a shop slogan, says how those signatures are classified. Do not write "six laminations" because you counted six facing multiples.

Clad and similar-metal bonds

Clad — roll-bond, explosion-bond, or a weld overlay treated as a bond for this topic — is a thin layer on a thicker substrate. Pulse-echo from the clad face is the usual shop method.

• Good metallurgical bond. The interface echo is small because impedances are close and the interface is not an air wall. The substrate back-wall is present.
• Disbond or lamination at the clad interface. The interface echo is large. The substrate back-wall is lost or sharply reduced under that footprint.
• Thin clad. The interface and the clad's own multiples sit close in time. If the pulse is too long, they merge. That is a frequency and damping problem, not a reason to invent a bond.

A weld overlay inspected as a bond follows the same interface-versus-back-wall logic. Do not confuse a geometric reflection from a buttering toe or an overlay stop with a planar disbond under the overlay. Scan it. Geometry stays put relative to the toe; a disbond has area and kills the back-wall under that area.

Loss of back-wall or interface echo

Topic 4 stems recycle a short list of A-scan changes. Learn them as pairs, not as isolated "the signal went away."

1. Loss of back-wall with a new or stronger mid-path echo. Energy never reached the far surface. The new echo is at the bond depth. Classic disbond or lamination at that depth.
2. Loss of back-wall with no mid-path echo. Energy was absorbed or scattered (coarse grain, a thick attenuative adhesive, a geometry that steers the beam away), or the back-wall left the beam because of a thickness change. Do not call a disbond until you have ruled out attenuation, coupling, and geometry.
3. Interface echo present, back-wall still present. Possible partial bond, a weak impedance step (different metals or a thin adhesive that still transmits), or a small reflector under part of the beam. Scan it. Look at amplitude maps, not one frozen gate.
4. Through-transmission amplitude drop with no pulse-echo interface story. The path is blocked somewhere. Use pulse-echo or a second view to locate the interface if the procedure requires location.

Coupling loss on the near surface also kills the back-wall. A dry spot, a lift-off, or a wrinkled facing looks like a "disbond" until you reseat the probe. Confirm coupling — first-material echo, multiples in the delay line, or the known entry echo — before you write a disbond. A Level II who reports unbonds from a dry glove print has not evaluated a bonded structure.

Multiple echoes

Multiple echoes (reverberations) are a teaching target because they are easy to misread.

• In a thin facing over a disbond, multiples at the facing thickness are expected. They are evidence that the facing is free, not evidence of six separate defects.
• In a solid plate with a good back-wall, multiples of the full thickness mean energy is bouncing plate-to-plate. A lamination cuts those multiples short — you get multiples of the lamination depth instead of the full thickness.
• In clad, multiples of the clad thickness with a lost substrate back-wall support a clad disbond.
• Mode-converted multiples and wedge reverberations are not bond evidence. If the time does not match a real thickness in the stack, using the correct velocity, it is not a facing ring.

A candidate who counts every peak as a new discontinuity fails this topic. Count spacing, convert spacing to thickness with the correct velocity, and ask which layer that thickness belongs to.

Air-gap disbond versus kissing bond

This is the limitation the general exam expects you to state out loud.

An air-gap disbond is a separation with a gas film. Reflection coefficients at metal-to-air or composite-to-air interfaces are close to 1. Pulse-echo lights up. Through-transmission dies. Conventional UT is good at this.

A kissing bond (intimate contact without strength) is two surfaces pressed together — or an adhesive that never wetted — with no resolvable air gap. Acoustic impedance across the contact can look almost like a good bond. Energy transmits. The A-scan can look acceptable while the joint has little or no load-carrying capability. Conventional pulse-echo and through-transmission under-call kissing bonds. Higher-order methods (nonlinear ultrasound, some resonance techniques, laser UT) or a mechanical/proof test may be required. Those are beyond a "turn the gain up" fix.

Related hard cases:

• Very thin residual adhesive that still couples sound but has no peel strength.
• Crush of honeycomb that still touches the facing in spots.
• Contamination (oil, release agent) that leaves a weak interface with little air.

If the stem says the surfaces are in intimate contact and the A-scan looks bonded, the honest Level II answer is limitation, not "therefore the bond is structurally good." Ultrasound reports acoustic continuity. It does not certify peel strength.

When to switch frequencies

Frequency is a resolution-versus-penetration trade, the same trade as in weld work, except the layers are thinner and the adhesives are more attenuative.

Raise frequency when you must separate a thin clad or facing echo from the entry pulse or from the next interface, when you need sensitivity to a small air-gap island, and when grain noise is not the limit and the path is short. A highly damped or broadband probe helps range resolution on thin bonds.

Lower frequency when the stack is attenuative — thick adhesive, composite facing, rubber, coarse-core filler — when you cannot obtain a stable through-transmission pulse or a substrate back-wall at the higher frequency, and when honeycomb or sandwich penetration is the goal and facing resolution is already adequate.

Dual-element (TR) probes shrink the dead zone on near-surface bonds and thin remaining ligaments. A delay-line or immersion setup does the same for thin clad. Typical shop practice — not an ASNT-published secret — is a damped higher-frequency probe for thin clad resolution and a lower-frequency, less-damped probe when the job is "get through the panel."

Do not switch frequency as a way to make a kissing bond appear. Changing frequency changes resolution and attenuation; it does not create an air gap that is not there. After a frequency change, recalibrate range, sensitivity, and any DAC or reference echo the procedure uses. A bond call made at 10 MHz is not automatically valid on a 2.25 MHz setup you have not restandardized.

Realistic exam scenarios

A roll-bonded clad shows a large interface echo at the clad thickness and no substrate back-wall under a 40 mm island. Adjacent material has a small interface and a clear back-wall. Coupling is confirmed. Classification: disbond (or lamination) at the clad interface.

A honeycomb facing rings with multiples equal to facing thickness over one cell field; through-transmission from the same field is lost. Classification: facing-to-core disbond (or equivalent unbond), not six defects.

Through-transmission on a sandwich is dead; pulse-echo from both faces shows normal facing rings everywhere and no obvious interface. Next action: check alignment, coupling, and instrument, then consider core damage or a mid-core block. Do not invent a facing disbond you cannot see.

Two plates in a press fixture show a perfect back-wall and almost no interface echo, but the traveler says the adhesive was never applied. That is a kissing-contact limitation, not a passed bond examination.

Topic 4 language is short: which technique, which echo appeared or vanished, which layer the multiples belong to, and what ultrasound cannot guarantee.