**5. Sensing and Signal Processing**

Conventional piezoceramic sensors still dominate AE applications to structures, although PWAS has started to become smaller and more functional for potential aerospace uses [14]. Recently, Ono [41,157,158] reexamined sensor calibration methods for different types of wave propagation. These new approaches incorporated laser interferometry as the basis for calibration and Vallen [159] took an initial step for their standardization. It was also demonstrated that the so-called reciprocity calibration methods [160–162] have fundamental flaws. It was also found that most AE sensors fail to satisfy required reciprocity conditions [157,158].

Attempts to develop practical AE sensors based on optical fibers have continued [163,164], but cost and directionality problems still are obstacles. Yu et al. [165] presented a novel use of a fiber Bragg grating (FBG) sensor, a version that relies on phase shifts. This FBG sensor, connected by an optical fiber to a CFRP plate, has a broad bandwidth and allowed Lamb wave mode separation. Using FBG sensors for multiple channel operation needed in practical AE monitoring remains a challenge. Source location with arrays of FBG sensors are reported; e.g., Shrestha et al. [166] and Innes et al. [167].

On AE specific signal processing, Barat et al. [168] and Elizarov et al. [169] successfully developed new techniques for AE hit detection without relying on threshold crossing. They used an optimized combination of filters to detect the arrival of AE hit. This is possible due to the differences between noise processes and AE signals, which produce short-time perturbations. Their algorithms have been implemented into hardware. This approach traces back to high-order statistics and the use of short-time average and long-time average as used by Lokajícek [170]. Another method of hit time of arrival based on wavelet transform was given by Pomponi [171]. This method is conceptually easier to understand than those of high-order filtering and is based on a constraint imposed on wavelet decomposition. This comes from the rise time limit arising from the frequency characteristics of AE sensor used and an efficient signal denoising was achieved. Other methods were also reported [172,173]. Sagista et al. [174] applied a slightly different use of wavelet transform to amplitude distribution analysis defining b-values in terms of bandpass-filtered signal energy. Another innovative approach for pattern recognition analysis was explored by Godin and coworkers [175]. This method is based on genetic algorithm, which is a relatively new optimization method, pioneered by Goldberg [176]. See also [128] for its earlier application in source location. This approach can provide an alternate scheme is data clustering, which has relied primarily on the k-means method [177,178].

Another front of signal processing was AE tomography. Katsuyama et al. [179] presented the principle of this method in 1992, which was independently developed in 2004 by Schubert [180]. This utilizes AE hits (natural or artificial) received by multiple sensors to interrogate the conditions of wave paths. This imaging strategy can use attenuation, wave speed, or impedance mismatch. Shiotani and coworkers have applied it for concrete inspection and reported successful outcomes relying on wave speed variation [7,181,182]. Since concrete quality affects its sound velocity, special consideration was needed, including Kalman filtering [183]. Nishida et al. [184,185] applied this

method to concrete bridge decks subjected to fatigue loading in laboratory and field use. A good correlation was demonstrated between fatigue cracking and wave speed loss. Takamine et al. [186] used a variation of this method in deck inspection.

Section 5 discussed AE signal detection or sensing and signal processing other than signal location, which was covered in Section 3. In sensing part, piezoelectric ceramics are still dominant, even for PWAS devices. Yet, their calibration methods lagged behind till recently. Despite high hopes held earlier, optical fiber-based detectors have not reached a breaking point, mainly from high component cost and from the lack of revolutionary concepts. Still, FBG sensors are advancing as discussed. Several new approaches of signal processing appeared in the last ten years as noted. AE tomography, developed originally in 1992, has made a significant progress of late.
