The insulation system of transformers degrades over time due to harsh environmental and operational conditions, thus reducing their designated useful life. In oil, insulation degradation compounds are released due to transformer insulation degradation. An oil-immersed transformer can be assessed based on the concentration of these compounds in the oil. Transformer paper insulation conditions can be assessed by the level of 2-furfuraldehyde (2-FAL) and moisture in the oil. The dielectric constant of 2-FAL and moisture is very high in comparison to that of fresh oil, making capacitive sensor-based detection of these compounds in oil a promising and economical option. This paper investigates the feasibility of a simple interdigital transducer (IDT) fringing field capacitive sensor in the measurement of the concentration of 2-FAL and humidity in transformer oil. Firstly, an interdigital capacitive (IDC) sensor is simulated using COMSOL Multiphysics software to design and analyse a 1-r-1 IDC sensor where polyvinyl alcohol (PVA) and polydimethylsiloxane (PDMS) are used as sensing materials. Moreover, three different arrangements of electrode structures, which have been designed, are investigated: single-sided (SS), single-sided with shielding (SSWS) and double-sided (DS). Furthermore, the IDC sensors are fabricated and characterised. The fabricated sensor is tested and characterised using oil samples with different concentrations of humidity and 2-FAL (in the 0-60 ppm range). The sensor exhibits the maximum sensitivity for a double-sided 1-1-1 configuration. In the 0-60 ppm range, the sensor exhibits approximately linear response characteristics and very small characteristics drift. The sensor facilitates real-time on-site condition monitoring of in-service transformers and circumvents all the issues of offline monitoring.

The strong non-linearity and low prediction accuracy of defect depth inversion often lead to challenges in magnetic flux leakage (MFL) detection and evaluation. If a one-dimensional convolutional neural network (1D-CNN) model could automatically extract the features of the original MFL signal and its simple and compact configuration, then real-time and low-cost hardware implementation should be achievable in the future. Therefore, in this work, a pipeline defect size inversion method based on a 1D-CNN model is proposed and the optimal network hierarchy is explored. In order to fully fuse the three-axis MFL signal and input it into the 1D-CNN for defect size inversion, three signal input strategies suitable for MFL defect detection based on the 1D-CNN are proposed and the quantitative results of the different three-axis MFL signal input strategies for an artificial defect dataset are compared and analysed. Experimental results indicate that using a lateral connection between the three-axis MFL signals can effectively achieve high-precision quantification of defect size and especially the quantification of defect depth. In addition, the comprehensive comparison and analysis with other network structures and traditional algorithms also verify that the 1D-CNN method still has advantages in accuracy and time complexity.

Deep learning methods are widely used in the field of rolling bearing fault diagnosis and produce good results when faced with datasets with roughly equal numbers of normal and faulty samples. However, real-world data often has a serious imbalance, with the number of fault samples being significantly less than the number of normal samples. This dataset imbalance challenges the performance of traditional deep learning methods. To address this problem, this paper proposes an efficient imbalanced data rolling bearing fault diagnosis method. The method consists of two parts: a deep learning network based on a multi-scale self-attention mechanism and a novel loss function. In terms of the deep learning network, firstly, the one-dimensional vibration signal is converted into a two-dimensional image through the Gramian angular field. This conversion maximises the inherent feature extraction capability of the network. Subsequently, the multi-scale learning capability of the network is enhanced by implementing different expansion rates for the head of the multi-scale self-attention mechanism. This nuanced approach allows the network to capture the underlying information more efficiently. Finally, the inclusion of Ghost bottlenecks and feature pyramid networks (FPNs) helps to optimise network efficiency and improve generalisation performance. A novel loss function is also proposed to make the method more suitable for imbalanced data. During the training process, the classification of samples in each class is analysed using the recall metric of imbalanced classification and the real-time recall is used as a weight to weaken the dominance of the majority class. This weighting ensures the adaptability of the method to imbalanced datasets. The proposed method is evaluated using rolling bearing datasets from Case Western Reserve University, USA, and Southeast University, China. Comparison results with other state-of-the-art deep learning methods show that the proposed method has a robust performance when dealing with imbalanced data.

This study employs advanced techniques, namely the improved multi-synchrosqueezing transform (IMSST) and the Fourier synchrosqueezed transform (FSST), to extract the instantaneous frequency of the Humber Bridge, in the UK, in operational conditions. First, a brief overview of the relevant theories behind the techniques used is provided. These techniques are then applied to track the instantaneous frequencies within the monitored vertical acceleration data of the bridge. The instantaneous frequencies extracted by both techniques are verified against each other. The investigation further encompasses the development of three distinct regression models: a simple regression model, an expanded regression model and a regression model based on principal component analysis (PCA). The efficacy of the PCA-based model is scrutinised by generating and testing datasets that mirror the monitored data, with the Kolmogorov-Smirnov (KS) test confirming their validity. The findings of this research indicate a statistically significant inverse relationship between modal frequency and acceleration intensity. Furthermore, by incorporating wind effect considerations into the regression analysis, the R-squared values in both the expanded and PCA-based models exhibit substantial improvements. This demonstrates the performance of the enhanced model in capturing the dynamics of the bridge. The robustness of the PCA-based model is further corroborated by its performance during testing, proving its potential as a reliable tool in structural health monitoring (SHM).

To simulate cracks, a series of slits with depths ranging from 17.2% to 40% of 301 stainless steel (SS) plate thickness are fabricated for this study. The examination includes a microfocus X-ray tube and a digital detector array (DDA) with 75 ??m pixel pitch. The basis of this study is to determine the effect of detector scintillator blur, noise and pixelation on the contrast of the resultant radiographic image and, ultimately, on the detectability of a crack. Many crack detectability studies are performed on notched plates rather than true cracks. As the notch width becomes smaller than a pixel, the grey value of the image it generates will be averaged over the entire footprint of the pixel. Therefore, the contrast with neighbouring (no notch) pixels is reduced. This study develops a relationship between the material property, detector parameters, notch geometry and expected contrast that would result from the radiograph image. This relationship can be used as a transfer function towards evaluating the resultant contrast from a notch and extrapolating the results towards a hairline crack.

Arc magnetic field analysis is a valuable approach for assessing the stability of the arc welding process, yet existing methods lack the ability to effectively quantify the disorder within the process. Through an investigation into the characteristics of the arc magnetic field signal, it was observed that the occurrence of low-frequency random fluctuations in arc magnetic field power, induced by unstable factors such as bubbles or short circuits, contributed to increased complexity and randomness in the arc magnetic field signals. To visualise the arc magnetic field signals in a time-frequency domain, a spectrogram was employed, revealing a strong correlation between the distribution of maximum power spectral density (PSD) in the spectrogram and the stability of the arc welding process. Furthermore, a novel method based on sample entropy was introduced to provide a quantitative measure of this relationship. A comprehensive quantitative assessment indicator called arc magnetic field sample entropy (AMFSE) was proposed. This indicator effectively mitigates the influence of varying parameters on the quantitative results, enabling a more accurate and consistent representation of the stability of the arc welding process. The proposed method was validated through testing, yielding an accuracy rate exceeding 90%.

Due to their high strength-to-weight ratio, composite materials are now in use in many high-stress applications, particularly where light weight is also a requirement. In these situations, the detrimental knock-down in mechanical strength due to an out-of-plane wrinkle defect can have serious consequences and is the reason for a requirement to rapidly detect any such wrinkles at manufacture. Unfortunately, current ultrasonic inspection techniques used for quality control at manufacture are not sensitive enough to detect these wrinkles above coherent structural noise variations. This paper exploits the ply resonance that is a characteristic of multi-layer structures to generate two new metrics for both detection and classification of out-of-plane wrinkles, due to their perturbations of the ply spacing. These can be measured at every location on a structure using the instantaneous frequency, which is the rate of change of phase in the pulse-echo ultrasonic response. The proposed two new metrics for detection and classification of wrinkles are mean spacing and spacing difference and they can be applied to each waveform in real time, as it is acquired. Use of an analytical model to predict the ultrasonic response of the structure has allowed an understanding of how these metrics will be affected by various wrinkle types and how they can not only detect wrinkles but also classify the type of wrinkle and provide an approximate indication of severity. Three main types of wrinkle are considered: classic wrinkles near the mid-plane of a structure, back-surface wrinkles formed from a resin bulge near the back of a structure and folded wrinkles where several plies can be folded over completely in the bulk of the structure. Both simulations and experimental results demonstrate the effectiveness of these metrics on various types of structure, including carbon-fibre and hybrid carbon/glass-fibre composites with a range of ply thicknesses and wrinkle types.

Steel is widely used in the aerospace, machinery and automotive industries. Surface defects not only have a negative impact on the appearance of steel but also significantly reduce its wear resistance, high temperature resistance, corrosion resistance and fatigue strength. Therefore, the detection of steel surface defects is very important to improve the quality of steel production. The limited availability of surface defect samples in the industrial sector poses significant challenges for the accurate detection of defects in high-quality materials. In addition, the existing defect detection model is highly complex and not easy to deploy. To solve this problem, a lightweight defect detection network suitable for steel defects is proposed. The cross-layer feature fusion (CFF) in the design enables effective utilisation of multi-layer semantic features, facilitating the detection of small defects in steel. Secondly, a new loss function is designed to make up for the problems of small data volume and uneven data distribution. The experimental results demonstrate that the steel surface defect detection method proposed in this paper achieves the highest detection performance on widely used public datasets such as RSDDS, NEUS and NRSD-CR(test), while maintaining the lowest model complexity.

Metallic samples are widely applied in modern industrial production. Due to non-uniformities in the stress load, such samples may become damaged and produce defects, which can cause unnecessary economic losses. In this paper, an online defect detection method is proposed for the quality monitoring of metallic plates. The research involves the design and optimisation of an electromagnetic tomography (EMT) sensor and the development of a fast tomography algorithm. Specifically, a planar array eddy current sensor is designed for in-situ structural health monitoring of metallic specimens. The parameters of the sensor are optimised using an orthogonal methodology and a response surface methodology to improve the uniformity of the sensitivity field. In addition, a second-order iterative Bregman reconstruction algorithm is investigated to reconstruct the defect image, which can improve the reconstruction speed for this ill-posed problem. Simulation and experimental results indicate that the proposed method can be applied to effectively evaluate the locations and sizes of defects in metallic specimens. The correlation coefficients of the reconstructed images using the proposed method are larger than 0.8. Compared with traditional reconstruction algorithms, the method proposed in this paper shows fast convergence speed and smaller estimation errors.

Intelligent machining requires the online adaptation of the machining parameters to improve tool life and product quality and to reduce machining costs. This article presents a novel hybrid adaptive control (HAC) system for a drilling process. The HAC system is a combination of two adaptive controls: geometric adaptive control (GAC) and adaptive control by optimisation (ACO). It keeps the roughness of the holes within tolerance without compromising tool life. A response surface model (RSM) is used for modelling the drill wear and surface roughness with speed, feed, acceleration and force signals as inputs. The model predicts the wear and roughness with prediction accuracies of 97.1% and 93.6%, respectively. The roughness control is achieved through a Massachusetts Institute of Technology rule and tool wear is minimised by genetic algorithm optimisation. The adaptive algorithms are simulated and validated for the machining conditions given by the adaptive algorithms. The results show an improved tool life of 7% and surface roughness of 11%.