Acoustic BLISTER DETECTION in ALUMINUM 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.
-
Blisters, also known as pinhole blowouts, are localized weak spots in extruded aluminum caused by trapped gas within the billet. These defects can compromise structural integrity and quality, making them especially problematic in high-precision, high-volume metal forming processes.
-
Traditionally optical sensors are used to detect blisters, however pinhole blowouts may be covered in coating making them hard to detect, or originating from sides and angles that are invisible for an optical sensor. Also the pinhole size has a big effect on the sensitivity of an optical detection system.
-
Squarehead’s solution uses a microphone array with an integrated video camera to form a directional “listening beam” focused on a specific area, like the mouth of the extrusion press. This beam isolate and capture the unique popping sound generated by a blister, while removing surrounding noise, significantly improving detection even in noisy industrial environments.
-
When gas trapped in the aluminum escapes during extrusion, it produces a sharp, impulse-like sound—similar to a pop or crack. This distinct acoustic signature, which can be both loud or subtle, depending on the blister size, occurs precisely at the moment of blister formation and can be captured with focused microphone arrays.
-
The acoustic data from the microphone array is analyzed by a machine learning classifier trained to recognize blister signatures. It provides real-time confidence scores that are streamed over an industrial protocol to a control system or PLC and enables operators to respond immediately, reduce waste, and enhance production quality. Over time, logged data also supports process optimization and predictive maintenance.
