Automated detection and classification of concealed objects using infrared thermography and convolutional neural networks

This paper presents a study on the effectiveness of a convolutional neural network (CNN) in classifying infrared images for security scanning. Infrared thermography was explored as a non-invasive security scanner for stand-off and walk-through concealed object detection. Heat generated by human subjects radiates off the clothing surface, allowing detection by an infrared camera. However, infrared lacks in penetration capability compared to longer electromagnetic waves, leading to less obvious visuals on the clothing surface. ResNet-50 was used as the CNN model to automate the classification process of thermal images. The ImageNet database was used to pre-train the model, which was further fine-tuned using infrared images obtained from experiments. Four image pre-processing approaches were explored, i.e., raw infrared image, subject cropped region-of-interest (ROI) image, K-means, and Fuzzy-c clustered images. All these approaches were evaluated using the receiver operating characteristic curve on an internal holdout set, with an area-under-the-curve of 0.8923, 0.9256, 0.9485, and 0.9669 for the raw image, ROI cropped, K-means, and Fuzzy-c models, respectively. The CNN models trained using various image pre-processing approaches suggest that the prediction performance can be improved by the removal of non-decision relevant information and the visual highlighting of features.

Nonlinear elastic multi-path reciprocal method for damage localisation in composite materials

Nonlinear ultrasonic techniques rely on the measurement of nonlinear elastic effects caused by the interaction of ultrasonic waves with the material damage, and have shown high sensitivity to detect micro-cracks and defects in the early stages. This paper presents a nonlinear ultrasonic technique, here named nonlinear elastic multi-path reciprocal method, for the identification and localisation of micro-damage in composite laminates. In the proposed methodology, a sparse array of surface bonded ultrasonic transducers is used to measure the second harmonic elastic response associated with the material flaw. A reciprocal relationship of nonlinear elastic parameters evaluated from multiple transmitter-receiver pairs is then applied to locate the micro-damage. Experimental results on a damaged composite panel revealed that an accurate damage localisation was obtained using the normalised second order nonlinear parameter with a high signal-to-noise-ratio (∼11.2dB), whilst the use of bicoherence coefficient provided high localisation accuracy with a lower signal-to-noise-ratio (∼1.8dB). The maximum error between the calculated and the real damage location was nearly 13mm. Unlike traditional linear ultrasonic techniques, the proposed nonlinear elastic multi-path reciprocal method allows detecting material damage on composite materials without a priori knowledge of the ultrasonic wave velocity nor a baseline with the undamaged component.

Phononic Crystal Waveguide Transducers for Nonlinear Elastic Wave Sensing

Second harmonic generation is one of the most sensitive and reliable nonlinear elastic signatures for micro-damage assessment. However, its detection requires powerful amplification systems generating fictitious harmonics that are difficult to discern from pure nonlinear elastic effects. Current state-of-the-art nonlinear ultrasonic methods still involve impractical solutions such as cumbersome signal calibration processes and substantial modifications of the test component in order to create material-based tunable harmonic filters. Here we propose and demonstrate a valid and sensible alternative strategy involving the development of an ultrasonic phononic crystal waveguide transducer that exhibits both single and multiple frequency stop-bands filtering out fictitious second harmonic frequencies. Remarkably, such a sensing device can be easily fabricated and integrated on the surface of the test structure without altering its mechanical and geometrical properties. The design of the phononic crystal structure is supported by a perturbative theoretical model predicting the frequency band-gaps of periodic plates with sinusoidal corrugation. We find our theoretical findings in excellent agreement with experimental testing revealing that the proposed phononic crystal waveguide transducer successfully attenuates second harmonics caused by the ultrasonic equipment, thus demonstrating its wide range of potential applications for acousto/ultrasonic material damage inspection.

On the generation of nonlinear damage resonance intermodulation for elastic wave spectroscopy

Recent nonlinear elastic wave spectroscopy experiments have shown that the nonlinear ultrasonic response of damaged composite materials can be enhanced by higher vibrations at the local damage resonance. In this paper, the mathematical formulation for the generation of nonlinear wave effects associated with continuous periodic excitation and the concept of local defect resonance is provided. Under the assumption of both quadratic and cubic approximation, the existence of higher harmonics of the excitation frequency, superharmonics of the damage resonance frequency and nonlinear wave effects, here named as nonlinear damage resonance intermodulation, which correspond to the nonlinear intermodulation between the driving and the damage resonance frequencies, is proved. All these nonlinear elastic effects are caused by the interaction of propagating ultrasonic waves with the local damage resonance and can be measured at locations different from the material defect one. The proposed analytical model is confirmed and validated through experimental transducer-based measurements of the steady-state nonlinear resonance response on a damaged composite sample. These results will provide opportunities for early detection and imaging of material flaws.

Nonlinear elastic wave tomography for the imaging of corrosion damage

This paper presents a nonlinear elastic wave tomography method, based on ultrasonic guided waves, for the image of nonlinear signatures in the dynamic response of a damaged isotropic structure. The proposed technique relies on a combination of high order statistics and a radial basis function approach. The bicoherence of ultrasonic waveforms originated by a harmonic excitation was used to characterise the second order nonlinear signature contained in the measured signals due to the presence of surface corrosion. Then, a radial basis function interpolation was employed to achieve an effective visualisation of the damage over the panel using only a limited number of receiver sensors. The robustness of the proposed nonlinear imaging method was experimentally demonstrated on a damaged 2024 aluminium panel, and the nonlinear source location was detected with a high level of accuracy, even with few receiving elements. Compared to five standard ultrasonic imaging methods, this nonlinear tomography technique does not require any baseline with the undamaged structure for the evaluation of the corrosion damage, nor a priori knowledge of the mechanical properties of the specimen.

Modelling of multiscale nonlinear interaction of elastic waves with three-dimensional cracks

This paper presents a nonlinear elastic material model able to simulate the nonlinear effects generated by the interaction of acoustic/ultrasonic waves with damage precursors and micro-cracks in a variety of materials. Such a constitutive model is implemented in an in-house finite element code and exhibits a multiscale nature where the macroscopic behavior of damaged structures can be represented through a contribution of a number of mesoscopic elements, which are composed by a statistical collection of microscopic units. By means of the semi-analytical Landau formulation and Preisach-Mayergoyz space representation, this multiscale model allows the description of the structural response under continuous harmonic excitation of micro-damaged materials showing both anharmonic and dissipative hysteretic effects. In this manner, nonlinear effects observed experimentally, such as the generation of both even and odd harmonics, can be reproduced. In addition, by using Kelvin eigentensors and eigenelastic constants, the wave propagation problem in both isotropic and orthotropic solids was extended to the three-dimensional Cartesian space. The developed model has been verified for a number of different geometrical and material configurations. Particularly, the influence of a small region with classical and non-classical elasticity and the variations of the input amplitudes on the harmonics generation were analyzed.

Nonlinear elastic imaging using reciprocal time reversal and third order symmetry analysis

This paper presents a nonlinear imaging method for the detection of the nonlinear signature due to impact damage in complex anisotropic solids with diffuse field conditions. The proposed technique, based on a combination of an inverse filtering approach with phase symmetry analysis and frequency modulated excitation signals, is applied to a number of waveforms containing the nonlinear impulse responses of the medium. Phase symmetry analysis was used to characterize the third order nonlinearity of the structure by exploiting its invariant properties with the phase angle of the input waveforms. Then, a "virtual" reciprocal time reversal imaging process, using only one broadcasting transducer and one receiving transducer, was used to insonify the defect taking advantage of multiple linear scattering as mode conversion and boundary reflections. The robustness of this technique was experimentally demonstrated on a damaged sandwich panel, and the nonlinear source, induced by low-velocity impact loading, was retrieved with a high level of accuracy. Its minimal processing requirements make this method a valid alternative to the traditional nonlinear elastic wave spectroscopy techniques for materials showing either classical or non-classical nonlinear behavior.

Acoustic emission localization in complex dissipative anisotropic structures using a one-channel reciprocal time reversal method

This paper presents an imaging method for the localization of the impact point in complex anisotropic structures with diffuse field conditions, using only one passive transducer. The proposed technique is based on the reciprocal time reversal approach (inverse filtering) applied to a number of waveforms stored into a database containing the experimental Green's function of the structure. Unlike most acoustic emission monitoring systems, the present method exploits the benefits of multiple scattering, mode conversion, and boundaries reflections to achieve the focusing of the source with high resolution. Compared to a standard time reversal approach, the optimal refocusing of the back propagated wave field at the impact point is accomplished through a "virtual" imaging process. The robustness of the inverse filtering technique is experimentally demonstrated on a dissipative stiffened composite panel and the source position can be retrieved with a high level of accuracy in any position of the structure. Its very simple configuration and minimal processing requirements make this method a valid alternative to the conventional imaging Structural Health Monitoring systems for the acoustic emission source localization.

MR in the evaluation of new anterior cruciate ligament and tibial tunnel position: correlation with clinical and functional features

PURPOSE

This study aimed to evaluate correlations between the position of the tibial tunnel, its alignment with the ligament-screw system, presence of intratunnel fluid, position of the tibial tunnel with respect to the Blumensaat line and clinical knee stability in patients who underwent arthroscopic reconstruction of the anterior cruciate ligament (ACL), by using magnetic resonance (MR) imaging.

MATERIALS AND METHODS

Forty-eight patients (40 men, eight women; mean age, 31 years) underwent arthroscopic reconstruction of the ACL using double-strand semitendinosus and gracilis tendons. The new ACL was fixed to the tibial tunnel using Bio-Intrafix (Mitek). All patients underwent MR imaging 12 months after surgery and clinical evaluation at 6 and 12 months using the International Knee Documentation Committee (IKDC) scoring system. MR imaging and clinical features were correlated using the Mann-Whitney U test for continuous variables and Fisher's exact test for categorical variables.

RESULTS

Forty-one patients were clinically stable (groups A and B according to the IKDC test) and seven were unstable (group C). Mean values of tibial tunnel position in clinically unstable vs stable patients were, respectively, -3.6 ±3.8 mm vs. -2.8±3.8 mm in relation to the Blumensaat line (p=0.5712) and 77.3°±11.3 vs. 72.5°±5.5 as concerned the angle measured on the coronal view of the new ACL (p=0.3248); fluid was present in the tibial tunnel in 42.9% and 9.8% of cases, respectively (p=0.2104). MR imaging showed misalignment of ligament screw and tibial tunnel in 57.1% of patients in group C and in 12.2% in groups A and B (p=0.017).

CONCLUSIONS

Misalignment of the ligament-screw system and the tibial tunnel and the presence of fluid in the tibial tunnel appear to be directly correlated with clinical instability.