Acoustic Blister Detection in Aluminium Extrusion

In high-volume metal forming processes such as extrusion, rolling, and pressing, precision and repeatability are paramount. Yet, even in tightly controlled environments, material defects can emerge. One particularly elusive and damaging defect in aluminum extrusion is the formation of blisters, also known as pinhole blowouts, localized weak spots caused by trapped gas within the metal.

Traditional quality assurance tools, including optical and infrared sensors, often fall short in detecting these faults in real time—especially in the harsh, noisy, and dusty environments typical of heavy industry. However, these defects often produce a telltale sign: sound.

The Sound of a Fault

Blister formation during extrusion generates a distinct acoustic signature, typically heard as a sharp impulse or popping sound. This occurs when trapped gas inside the billet rapidly expands and escapes as the material is forced through the die, breaching the metal’s surface. The result is a brief but unmistakable sound, similar to a crack or pop, emitted precisely at the moment of blister formation.

Unfortunately, this sound is easily lost in the industrial symphony of constant machinery hum, mechanical vibrations, and background noise. Its brief and intermittent nature makes it virtually impossible to detect consistently with the human ear or a single microphone.

Enter Microphone Arrays

To overcome this challenge, Squarehead’s microphone array technology offers a powerful alternative. Unlike traditional microphones, our arrays consist of hundreds of acoustic elements combined with an integrated video camera. This setup allows us to focus "listening beams" in specific directions, filtering out irrelevant noise and honing in on the subtle acoustic footprint of a blister event.

In the context of aluminum extrusion, a listening beam is directed toward the mouth of the extrusion press. This allows us to capture the characteristic popping sound that escapes during a blister event. This directional focus dramatically improves the signal-to-noise ratio, enabling detection of sounds that would otherwise be buried in industrial noise.

From Sound to Insight

Detection, however, is only the first step. The true power of the system lies in real-time audio analysis. The acoustic stream captured by the array is fed into a machine learning classifier trained to recognize the signature of a blister. By analyzing frequency, intensity, duration, and other features, the classifier continuously outputs a confidence score whether a blister event has occurred.

This live detection capability empowers operators to intervene immediately, reducing waste, minimizing defective output, and improving overall production quality. Over time, logged data can also support process optimization and predictive maintenance strategies.