Journal of Nondestructive Evaluation最新文献

Gearboxes, as critical components, often operate in demanding conditions, enduring constant exposure to variable loads and speeds. In the realm of condition monitoring, the dataset primarily comprises data from normal operating conditions, with significantly fewer instances of faulty conditions, resulting in imbalanced datasets. To address the challenges posed by this data disparity, researchers have proposed various solutions aimed at enhancing the performance of classification models. One such solution involves balancing the dataset before the training phase through oversampling techniques. In this study, we utilized the Sparse Autoencoder technique for data augmentation and employed Support Vector Machine (SVM) and Random Forest (RF) for classification. We conducted four experiments to evaluate the impact of data imbalance on classifier performance: (1) using the original dataset without data augmentation, (2) employing partial data augmentation, (3) applying full data augmentation, and (4) balancing the dataset while using Kernel Principal Component Analysis (KPCA) for dimensionality reduction. Our findings revealed that both algorithms achieved accuracies exceeding 90%, even when employing the original non-augmented data. When partial data augmentation was employed both algorithms were able to achieve accuracies beyond 98%. Full data augmentation yielded slightly better results compared to partial augmentation. After reducing dimensions from 18 to 11 using KPCA, both classifiers maintained robust performance. SVM achieved an overall accuracy of 98.72%, while RF achieved 96.06% accuracy.

In this paper, the effectiveness of the eddy current methodology for crack detection in MAR 247 nickel-based superalloy with aluminide coatings subjected to cyclic loading was investigated. The specimens were subjected to force-controlled fatigue tests under zero mean level, constant stress amplitude from 300 MPa to 600 MPa and a frequency of 20 Hz. During the fatigue, a particular level of damage was introduced into the material leading to the formation of microcracks. Subsequently, a new design of probe with a pot core was developed to limit magnetic flux leakage and directed it towards the surface under examination. The suitability of the new methodology was further confirmed as the specimens containing defects were successfully identified. The changes in probe resistance values registered for damaged specimens ranged approximately from 8 to 14%.

Thermal embrittlement is a key concern for the structural integrity of engineering components. Monitoring thermal embrittlement may indicate susceptibility to crack initiation and growth and therefore act as a damage precursor. In this study the correlation between thermoelectric power (also known as the Seebeck Coefficient) and the hardness of thermally aged 2507 super duplex stainless steel was demonstrated, showing the suitability of using thermoelectric power as a proxy measurement for embrittlement. This article presents a continuous high-temperature thermoelectric power monitoring system that is suitable for installation on large engineering assets. Using temperature gradients in the sample of < 6.5 °C a measurement standard deviation of 5.8 nV/°C was possible, which was sufficient to monitor the ~ 850 nV/°C increase in thermoelectric power that occurred in this study.

Surface defect detection in mechanical processing mainly adopts manual inspection, which has certain issues including strong dependence on manual experience, low efficiency, and difficulty in online detection. A surface texture elimination method based on improved frequency domain filtering in conjunction with morphological sub-pixel edge detection is put forward in order to address the aforementioned issues with machining surface defects. Firstly, ascertain whether textures exist in the image and determine their feature values using the grayscale co-occurrence matrix. The main energy direction of the textured surface in the frequency domain was then obtained by applying the Fourier transform to the processed surface. An elliptical domain narrow stopband was designed to reduce the energy in the band region corresponding to the processed surface texture and eliminate the processed surface texture. Finally, improve morphology and sub-pixel edge fusion to extract surface defect images. Cracks and scratches have a detectable width of 0.01 mm, a detection accuracy of 97.667%, and a detection time of 0.02 s. Therefore, the combination of machine vision and texture removal technology has achieved the detection of surface scratches and cracks in machining, providing a theoretical basis for defect detection in workpiece processing.

Estimation of volume fraction percentage of the multiple phases flowing in pipes with limited access is a challenge in oil, gas, chemical processes, and petrochemical industries. In this research, the gamma backscattered spectra together with the machine learning model were used to predict precise volume fraction percentages in water-gasoil-air three-phase flows and solve the aforementioned challenge. The detection system includes a single energy ¹³⁷Cs source and a NaI(Tl) detector to measure the backscattered rays. The MCNPX code was used to simulate the setup and produce the required data for the artificial neural network. The volume fraction was calculated with mean relative error percentage 13.60% and the root mean square error 2.68, respectively. Then, the results were compared with the acquired results of transmitted gamma-ray spectra. The proposed design is a suitable, safe, and low-cost choice for industries.

The falling weight deflectometer (FWD) test is a prevalent non-destructive testing (NDT) technique in engineering that is essential for evaluating pavement conditions. In this work, the transfer function (TF) theory in frequency domain analysis was applied to address the technical challenges present in FWD research. A pavement system transfer function (PSTF) was proposed as a novel approach for evaluating pavement conditions. The spectral method with fixed-end boundary conditions (B-SEM) was employed to compute the theoretical deflection data for different pavement structures with bedrock during FWD testing. The fast Fourier transform (FFT) technique was used to convert the data into the frequency domain, enabling the construction and calculation of the PSTF. The validity of the PSTF theory was confirmed, and the pavement information contained in the PSTF spectrum was discussed. An analysis and summary are conducted on the impact of variations in pavement attributes on the spectrum. The results indicate that the proposed PSTF contains information regarding pavement system, including the structural layer modulus, structural layer thickness, and bedrock depth. The pavement conditions can be evaluated by directly analyzing the PSTF without considering external factors. The PSTF spectrum is most significantly influenced by bedrock depths between 200 and 500 cm. For every 50 cm variation in bedrock depth, the coefficient of increase and decrease (CIE) of peak frequency ranges from 8.1% to 23.1%. The PSTF spectrum is highly sensitive to variations in the subgrade modulus between 40 and 70 MPa. In this range, the CIE of peak amplitude is greater than 11% for every 10MPa variation in subgrade modulus. The impact of the modulus and thickness of both the surface layer and base layer on the spectrum is noteworthy and should not be disregarded. Spectral analysis is used to summarize the variation in pavement attributes within the PSTF spectrum, serving as a theoretical foundation for further investigations.

A new Forward Ray Tracing Instrument (FRTI) for simulating X-ray CT scanners is presented. The FRTI enables the modelling of various detector geometries to optimise instrument designs. The FRTI is demonstrated by comparing experimentally measured sphere centre-to-centre distances from two material measures with digital clones. The measured length deviations were smaller than the reconstructed grid spacing for both the experimental and simulated acquisitions. As expected the experimentally measured length deviations were larger than the simulated measurements. The results demonstrate the FRII’s capability of simulating an X-ray CT scanner and performing length measurements.

The surface of superalloy precision castings might exhibit defects after forming, posing a significant risk to their service life, necessitating inspection during post-process. Radiographic inspection, with its extensive research in automation, can achieve efficient and accurate detection of defects. However, it is limited in surface defects detection due to limited sensitivity to non-volumetric defects and high cost. In contrast, fluorescent penetrant inspection (FPI) is highly efficient for surface defect inspection due to its low cost, high sensitivity, and speed. However, manual examination introduces variability in the results, impacting the consistency and reliability of the inspection process. Automation is needed to ensure consistency and reliability of inspection. The implementation of an automated defect identification system based on FPI using convolutional neural networks (CNNs) was systematically investigated. Among the CNN models tested, MobileNetV2 exhibited exceptional performance, achieving a remarkable recall rate of 0.992 and an accuracy of 0.992. Additionally, the effect of class imbalance on model performance was carefully examined. Furthermore, the features extracted by the model were visualized using Grad-CAM to reveal the attention of the CNN model to the fluorescent display features of defects. This study underscores the strong capability of deep learning architectures in identifying defects of precision casting components, paving the way for the automation of the entire FPI process.

This paper presents a review of the possible methods for testing the fire performance properties of reused timber through non-destructive techniques, focusing on structural elements. Evaluating the fire performance of old wooden specimen is necessary to facilitate reuse, in the support of the transition to a circular economy. The use of non-destructive methods minimizes damages to the pieces during the evaluation process. Three angles are reviewed: (1) The properties of wood influencing fire performance, (2) the change of wood properties over time, and (3) the known non-destructive tests. Some properties of wood are known to influence the fire performance, e.g., the density. Of these, there is no evidence of irreversible changes due to the passage of time only. The many different non- and semi- destructive techniques that can be applied to wood seldom relate to these properties, but rather to mechanical properties or geometry. Additionally, accurate measurements are often difficult, while some are only done in laboratories. This review concludes that currently there is no known non-destructive method that permits to estimate the fire performance of a reused timber element compared to a new one. There is a gap of knowledge on the evolution of the fire properties of timber during the use phase of the building, and there are no established methods to test for these properties without destroying a significant portion of the element. Development of non-destructive test methodologies to assess fire properties of timber will expand the market for reused timber to include load carrying timber.

Concrete is a versatile construction material, which is often chemically attacked by various environmental agents. Concrete, being porous, allows movement of chemicals within its interior. The transport property of various chemicals depends on hydraulic diffusivity, which in turn depends on the degree of moisture saturation (DoS). Therefore, DoS is an important parameter and its estimation is highly significant with regards to material characterization. In this paper, cement concrete samples of size 75 mm × 75 mm × 300 mm are fabricated with water to cement ratio (w/c) of 0.45, 0.55 and 0.65. These samples are conditioned to various DoS in two methods described as drying and wetting cycles. A set-up for electrical measurements along the length of the sample is proposed, in which a pulse-based electrical input is imposed, which enables simultaneous acquisition of the material response at multiple frequencies, ranging between 100 and 500 kHz. Using a simple circuit model, the real and imaginary parts of impedivity are calculated along the length of the samples and the bulk conductivities and bulk relative permittivities at various DoS are estimated. The conductivity values are found to follow a regular pattern for various DoS and at different excitation frequencies, which facilitates the establishment of an empirical quantitative relationship between conductivity and the DoS of concrete. Further, on evaluation of permittivity it is found that the value of this parameter is much higher than that of its constituents which was seen in the literature.