*Noise Reduction*

AE signals may be generated from a number of possible sources including background noise, micro-crack generation, or plastic deformation. In order to reduce uncertainties and determine the AE signals corresponding to crack growth, applying noise reduction techniques on the captured data was required. Various de-noising techniques have been proposed to filter AE signals due to extraneous events during crack growth [10,21,22].

To filter out noise from AE signals associated with fatigue crack propagation, the recorded AE data was filtered using the acquisition system's band pass filter (200 kHz–3 MHz). In addition to the band pass filter that restricts the captured signals based on frequency, an amplitude threshold was determined and filtered out low-amplitude background noise. It should be noted that the lack of significant AE activity in the initial stages of fatigue loading makes it more difficult to distinguish background noise from crack-related acoustic events. For each set of materials, a dummy specimen was tested beforehand under the same conditions as the main experiments to determine the background noise. The dummy specimen was simply the first sample from the test specimen batch and had the same geometry and characteristics. The dummy specimen was installed in the testing machine, and the background noise was captured as a cyclic load was applied for a few cycles. Based on this test, the AE detection threshold was set to 45 dB for Al7075-T6 samples to eliminate the background noises. This value was found to be 35 dB for the Ti-6-4 samples.

It has also been observed that AE events occurring during the loading portion of a cycle are related to crack growth [6,17,21]. Therefore, the AE data taken only during the loading portion of each cycle were used for data analysis, while AE events during the unloading portion of the cycle were ignored. In addition, many researchers have assumed that only events occurring close to the maximum or peak load are associated directly with crack growth [12,21,22]. So, the filtered AE events were separated for different percentages of the applied load range, and it was determined that the AE counts occurring within the top 20% of peak load showed the closest correlation with crack propagation rates [23].
