Machine learning at the interface of structural health monitoring and non-destructive evaluation.

While both non-destructive evaluation (NDE) and structural health monitoring (SHM) share the objective of damage detection and identification in structures, they are distinct in many respects. This paper will discuss the differences and commonalities and consider ultrasonic/guided-wave inspection as a technology at the interface of the two methodologies. It will discuss how data-based/machine learning analysis provides a powerful approach to ultrasonic NDE/SHM in terms of the available algorithms, and more generally, how different techniques can accommodate the very substantial quantities of data that are provided by modern monitoring campaigns. Several machine learning methods will be illustrated using case studies of composite structure monitoring and will consider the challenges of high-dimensional feature data available from sensing technologies like autonomous robotic ultrasonic inspection. This article is part of the theme issue 'Advanced electromagnetic non-destructive evaluation and smart monitoring'.

Investigation of synthetic aperture methods in ultrasound surface imaging using elementary surface types.

Synthetic aperture imaging methods have been employed widely in recent research in non-destructive testing (NDT), but uptake has been more limited in medical ultrasound imaging. Typically offering superior focussing power over more traditional phased array methods, these techniques have been employed in NDT applications to locate and characterise small defects within large samples, but have rarely been used to image surfaces. A desire to ultimately employ ultrasonic surface imaging for bone surface geometry measurement prior to surgical intervention motivates this research, and results are presented for initial laboratory trials of a surface reconstruction technique based on global thresholding of ultrasonic 3D point cloud data. In this study, representative geometry artefacts were imaged in the laboratory using two synthetic aperture techniques; the Total Focusing Method (TFM) and the Synthetic Aperture Focusing Technique (SAFT) employing full and narrow synthetic apertures, respectively. Three high precision metallic samples of known geometries (cuboid, sphere and cylinder) which featured a range of elementary surface primitives were imaged using a 5MHz, 128 element 1D phased array employing both SAFT and TFM approaches. The array was manipulated around the samples using a precision robotic positioning system, allowing for repeatable ultrasound derived 3D surface point clouds to be created. A global thresholding technique was then developed that allowed the extraction of the surface profiles, and these were compared with the known geometry samples to provide a quantitative measure of error of 3D surface reconstruction. The mean errors achieved with optimised SAFT imaging for the cuboidal, spherical and cylindrical samples were 1.3mm, 2.9mm and 2.0mm respectively, while those for TFM imaging were 3.7mm, 3.0mm and 3.1mm, respectively. These results were contrary to expectations given the higher information content associated with the TFM images. However, it was established that the reduced error associated with the SAFT technique was associated with significant reductions in side lobe levels of approximately 24dB in comparison to TFM imaging, although this came at the expense of reduced resolution and coverage.

Time-of-flight measurement techniques for airborne ultrasonic ranging.

Airborne ultrasonic ranging is used in a variety of different engineering applications for which other positional metrology techniques cannot be used, for example in closed-cell locations, when optical line of sight is limited, and when multipath effects preclude electromagnetic-based wireless systems. Although subject to fundamental physical limitations, e.g., because of the temperature dependence of acoustic velocity in air, these acoustic techniques often provide a cost-effective solution for applications in mobile robotics, structural inspection, and biomedical imaging. In this article, the different techniques and limitations of a range of airborne ultrasonic ranging approaches are reviewed, with an emphasis on the accuracy and repeatability of the measurements. Simple time-domain approaches are compared with their frequency-domain equivalents, and the use of hybrid models and biologically inspired approaches are discussed.

Wireless recording of the calls of Rousettus aegyptiacus and their reproduction using electrostatic transducers.

Bats are capable of imaging their surroundings in great detail using echolocation. To apply similar methods to human engineering systems requires the capability to measure and recreate the signals used, and to understand the processing applied to returning echoes. In this work, the emitted and reflected echolocation signals of Rousettus aegyptiacus are recorded while the bat is in flight, using a wireless sensor mounted on the bat. The sensor is designed to replicate the acoustic gain control which bats are known to use, applying a gain to returning echoes that is dependent on the incurred time delay. Employing this technique allows emitted and reflected echolocation calls, which have a wide dynamic range, to be recorded. The recorded echoes demonstrate the complexity of environment reconstruction using echolocation. The sensor is also used to make accurate recordings of the emitted calls, and these calls are recreated in the laboratory using custom-built wideband electrostatic transducers, allied with a spectral equalization technique. This technique is further demonstrated by recreating multi-harmonic bioinspired FM chirps. The ability to record and accurately synthesize echolocation calls enables the exploitation of biological signals in human engineering systems for sonar, materials characterization and imaging.

Optical frequency-domain reflectometry for microbend sensor demodulation.

The operation of an incoherent optical frequency-domain reflectometer for monitoring the continuous Rayleigh backscatter in a multimode optical fiber is presented. A simple but effective model to predict the value of beat frequencies arising in the system when excited by a linearly frequency-swept amplitude modulation has been developed. We have verified the model's predictions by experimental measurement of beat frequencies and modulation depth indices of different lengths of standard graded-index multimode optical fiber. Demonstration of the system sensitivity to the detection of microbending loss is then discussed. In particular the detection of loss in a hydrogel-based water-sensing cable allows an alternative interrogation to conventional optical time-domain reflectometry techniques to be implemented. We demonstrate that the incoherent optical frequency-domain reflectometer is capable of detecting and locating sections of increased loss in a multimode optical fiber, and we discuss the fundamental limits on spatial resolution and dynamic range.

Generation and reception of ultrasonic guided waves in composite plates using conformable piezoelectric transmitters and optical-fiber detectors.

A condition monitoring nondestructive evaluation (NDE) system, combining the generation of ultrasonic Lamb waves in thin composite plates and their subsequent detection using an embedded optical fiber system is described. The acoustic source is of low profile with respect to the composite plate thickness, surface conformable, and able to efficiently launch a known Lamb wave mode, at operating frequencies between 100 and 500 kHz, over typical propagation distances of 100 to 500 mm. It incorporates both piezocomposite technology and interdigital design techniques to generate the fundamental symmetrical Lamb wave mode in both metallic and carbon-fiber composite plates. Linear systems and finite element modeling techniques have been used to evaluate the operation of the transducer structure, and this is supplemented by experimental verification of the simulated data. An optical fiber, either bonded to the surface or embedded across the length of the composite plate samples, is used to detect the propagating ultrasonic Lamb waves. Single mode silica fiber has been used in conjunction with a portable 633 nm Mach-Zehnder interferometer for signal demodulation and subsequent data acquisition. This hybrid system is shown to generate and detect the fundamental symmetrical Lamb wave (s(0)) in both carbon-fiber and glass-fiber reinforced composite plates. Importantly, the system signal-to-noise ratio (SNR) associated with the acoustic source compares favorably with s(0) Lamb wave generation using a conventional transducer and angled perspex wedge arrangement.

Surface-bonded and embedded optical fibers as ultrasonic sensors.

The effectiveness of surface-bonded and embedded optical fibers for the detection of ultrasonic Lamb waves in 2-3-mm-thick steel, carbon-fiber-reinforced plastic (CFRP) and glass-reinforced plastic (GRP) plates are compared. A novel integrating ultrasonic sensor was achieved using the signal arm of an actively stabilized 633-nm homodyne Mach-Zehnder fiber interferometer which was either bonded directly to the plate surface or spliced to single-mode fibers embedded within a composite plate during manufacture. An embedded fiber is shown to be about 20 times more sensitive to Lamb wave motions than a surface-bonded fiber. However, the latter may be more practical.

Acoustic and flexural excitation of a floating structure by a single laser pulse.

The acoustic and flexural vibrations of a small-scale floating structure following irradiation by a pulsed Nd:glass laser are compared with a radiated underwater sound field. A single subablative laser pulse of 600-µs duration was used both to bend and shock the floating structure at the irradiation site. The laser pulse caused the structure to flex at a frequency of approximately 1 kHz whereas relaxation oscillations in the laser output simultaneously excited ultrasonic Lamb waves within the material bulk. We present results to illustrate the broad bandwidth provided by this unusual form of excitation.