Search

Abstract The lack of labelled data presents a common challenge in many fault diagnosis and machine learning tasks. It requires the model to be able to efficiently capture useful fault features from a smaller amount of labelled data. In this paper, a method to train multiple convolutional auto‐encoders by self‐learning method and integrate them using ensemble learning, called ensemble self‐taught learning convolutional auto‐encoders (STL‐CAEs), is proposed, which can effectively extract features of bearing vibration signals. First, an ensemble learning strategy is proposed to obtain two auto‐encoders that satisfy the strategy by optimizing the model parameters and structure. Then, a self‐taught learning training method is proposed to solve the problem of little label data. Finally, ensemble learning and fault diagnosis is achieved by the SoftMax classifier. Applying the proposed method to the bearing data from Case Western Reserve University, the STL‐CAEs have higher accuracy and generalization than common fault diagnosis methods such as CAE, CNN, SAE and EMD, and also have significant advantages in terms of diagnostic time and training time.

To achieve real-time sensing and deformation reconstruction of flexible plate structure, a 3D reconstruction technique based on a quasi-distributed fiber grating sensor network and order-optimized modal superposition method is proposed. Graphic visualization is introduced to the field of deformation monitoring, then the displacement field reconstruction of the typical structure of spacecraft is realized. Firstly, a finite element model of the four-sided fixed-supported plate structure is constructed, and the conversion matrix of surface strain and displacement are obtained based on modal analysis, and then the modal parameters are optimized through static simulation experiments; Secondly, the influence of the number and position of sensors on the reconstruction accuracy is studied, then a high-density distributed fiber grating sensor network is attached to the surface of the plate to construct an experimental FBG model of the spacecraft; Finally, concentrated loads are applied to the structure at different positions, and structure’s surface strain information is collected by FBG network, then the morphological reconstruction of the deformed structure is realized based on the improved model superposition method, which proves the feasibility of the algorithm. The results show that the reconstruction method’s average absolute error is less than 0.13 mm, so this method can be used to displace field reconstruction problems of flexible structures.

The clutters of rebar in the ground penetrating radar (GPR) images may mask the echoes of the inner defects under the rebars, which adversely affects the identification of the inner structural defects in the reinforced concrete (RC). In this study, a deep learning-based method for rebar clutters removal and defect echoes enhancement in GPR B-scan images was proposed. The residual-inception blocks and attention modules were designed in the network as per the characteristics of the task. Validation of the method has been performed at three levels: first using synthetic data, next by synthetic data superimposed with actual concrete background noise, and finally using a sandbox model test in a realistic scenario. The results indicate that the proposed method had the ability to effectively remove rebar clutters and reconstruct complete defect echoes, with performance superior to those of the traditional deep-learning based methods. A RC structural defect identification experiment was performed to verify the effectiveness of the proposed method. The results indicated that after the rebar signals were removed using the method proposed, the accuracy for identifying the defects under the rebars had been improved from 0.208 to 0.850. The proposed method is promising for applying to practical engineering due to its fast processing and better performance.

To meet the requirements of high accuracy and high efficiency in three-dimensional (3D) measurement, a phase difference-3D coordinate mapping model is proposed based on extreme learning machine (ELM) network. First, the reconstruction model of the ideal measurement system is set following the geometric structure of the system. Subsequently, by generalizing camera and world coordinates, a generalized measurement model is built. Lastly, ELM network is employed to solve the mapping coefficients. During measurement, only one phase difference map is required to complete the 3D reconstruction of the object, which simplifies the data processing process and saves time. The result indicates that the mean square errors (MSEs) of the X, Y and Z of the testing sample are 3.5955×10-4 mm, 9.5113×10-4 mm and 4.4×10-3 mm, respectively. Moreover, the reconstruction experiments of objects with different geometric structures are performed to demonstrate the general application of the proposed method.

Diagnosing tunnel lining structural damages is vital to ensure safe tunnel operations. However, the detection of multiple defect is challenging task due to the size imbalance between cracks, spalling, and backgrounds. Currently, deep-learning-based methods for multiple defect are dependent on multiple-stage networks, which have limited their scalability and complex frame working processes. To accurately recognize the multiple defect at the pixel-level using only one-stage networks, a new method was proposed, which integrated the basic SegNet with a focal loss function, and was referred to as an FL-SegNet method. The focal loss function was adopted to address the problem of the size imbalance by down-weighing the losses assigned to the well-classified samples, and then the training was focused on the hard samples. Furthermore, comparative experiments were performed to evaluate the performances of the different methods. The experimental results demonstrated that FL-SegNet method was capable of accurately predicting the profiles of small-sized cracks and overlapping damages even under various noise conditions, and successfully outperformed the two-stream method and the basic SegNet method in this regard. The performance metrics (MPA and MIoU) of the FL-SegNet method were significantly higher than those of other multiple defect detection approaches in different scenarios (images with small-sized damages attained to 81.53% and 69.86%, increased by 11.99 % and 4.88% compared with two-stream method, and increased by 17.78% and 7.69% compared with basic SegNet). Therefore, this paper provides an effective solution for the future detection of multiple defect in tunnel linings.

Abstract A damage identification system of carbon fiber reinforced plastics (CFRP) structures is investigated using fiber Bragg grating (FBG) sensors and back propagation (BP) neural network. FBG sensors are applied to construct the sensing network to detect the structural dynamic response signals generated by active actuation. The damage identification model is built based on the BP neural network. The dynamic signal characteristics extracted by the Fourier transform are the inputs, and the damage states are the outputs of the model. Besides, damages are simulated by placing lumped masses with different weights instead of inducing real damages, which is confirmed to be feasible by finite element analysis (FEA). At last, the damage identification system is verified on a CFRP plate with 300 mm × 300 mm experimental area, with the accurate identification of varied damage states. The system provides a practical way for CFRP structural damage identification.

Abstract In order to achieve rotation angle measurement, one novel type of miniaturization fiber Bragg grating (FBG) rotation angle sensor with high measurement precision and temperature self-compensation is proposed and studied in this paper. The FBG rotation angle sensor mainly contains two core sensitivity elements (FBG1 and FBG2), triangular cantilever beam, and rotation angle transfer element. In theory, the proposed sensor can achieve temperature self-compensation by complementation of the two core sensitivity elements (FBG1 and FBG2), and it has a boundless angel measurement range with 2π rad period duo to the function of the rotation angle transfer element. Based on introducing the joint working processes, the theory calculation model of the FBG rotation angel sensor is established, and the calibration experiment on one prototype is also carried out to obtain its measurement performance. After experimental data analyses, the measurement precision of the FBG rotation angle sensor prototype is 0.2 ° with excellent linearity, and the temperature sensitivities of FBG1 and FBG2 are 10 pm/℃ and 10.1 pm/℃, correspondingly. All these experimental results confirm that the FBG rotation angle sensor can achieve large-range angle measurement with high precision and temperature self-compensation.

Abstract In some applications in structural health monitoring (SHM), the acoustic emission (AE) detection technology is used in the high temperature environment. In this paper, a high-temperature-resistant AE sensing system is developed based on the fiber Bragg grating (FBG) sensor. A novel high temperature FBG AE sensor is designed with a high signal-to-noise ratio (SNR) compared with the traditional FBG AE sensor. The output responses of the designed sensors with different sensing fiber lengths also are investigated both theoretically and experimentally. Excellent AE detection results are obtained using the proposed FBG AE sensing system over a temperature range from 25 ℃ to 200 ℃. The experimental results indicate that this FBG AE sensing system can well meet the application requirement in AE detecting areas at high temperature.

Abstract An in-line fiber Fabry-Perot interferometer (FPI) based on the hollow-core photonic crystal fiber (HCPCF) for refractive index (RI) measurement is proposed in this paper. The FPI is formed by splicing both ends of a short section of the HCPCF to single mode fibers (SMFs) and cleaving the SMF pigtail to a proper length. The RI response of the sensor is analyzed theoretically and demonstrated experimentally. The results show that the FPI sensor has linear response to external RI and good repeatability. The sensitivity calculated from the maximum fringe contrast is –136 dB/RIU. A new spectrum differential integration (SDI) method for signal processing is also presented in this study. In this method, the RI is obtained from the integrated intensity of the absolute difference between the interference spectrum and its smoothed spectrum. The results show that the sensitivity obtained from the integrated intensity is about –1.34×105 dB/RIU. Compared with the maximum fringe contrast method, the new SDI method can provide the higher sensitivity, better linearity, improved reliability, and accuracy, and it’s also convenient for automatic and fast signal processing in real-time monitoring of RI.

The travel time computation of microseismic waves in different directions (particularly, the diagonal direction) in three-dimensional space has been found to be inaccurate, which seriously affects the localization accuracy of three-dimensional microseismic sources. In order to solve this problem, this research study developed a method of calculating the P-wave travel time based on a 3D high-order fast marching method (3D_H_FMM). This study focused on designing a high-order finite-difference operator in order to realize the accurate calculation of the P-wave travel time in three-dimensional space. The method was validated using homogeneous velocity models and inhomogeneous layered media velocity models of different scales. The results showed that the overall mean absolute error (MAE) of the two homogenous models using 3D_H_FMM had been reduced by 88.335%, and 90.593% compared with the traditional 3D_FMM. On that basis, the three-dimensional localization of microseismic sources was carried out using a particle swarm optimization algorithm. The developed 3D_H_FMM was used to calculate the travel time, then to conduct the localization of the microseismic source in inhomogeneous models. The mean error of the localization results of the different positions in the three-dimensional space was determined to be 1.901 m, and the localization accuracy was found to be superior to that of the traditional 3D_FMM method (mean absolute localization error: 3.447 m) with the small-scaled inhomogeneous model.

The phase-to-height imaging model, as a three-dimensional (3D) measurement technology, has been commonly applied in fringe projection to assist surface profile measurement, where the efficient and accurate calculation of phase plays a critical role in precise imaging. To deal with multiple extra coded patterns and 2π jump error caused to the existing absolute phase demodulation methods, a novel method of phase demodulation is proposed based on dual variable-frequency (VF) coded patterns. In this paper, the frequency of coded fringe is defined as the number of coded fringes within a single sinusoidal fringe period. First, the effective wrapped phase (EWP) as calculated using the four-step phase shifting method was split into the wrapped phase region with complete period and the wrapped phase region without complete period. Second, the fringe orders in wrapped phase region with complete period were decoded according to the frequency of the VF coded fringes and the continuous characteristic of the fringe order. Notably, the sampling frequency of fast Fourier transform (FFT) was determined by the length of the decoding interval and can be adjusted automatically with the variation in height of the object. Third, the fringe orders in wrapped phase region without complete period were decoded depending on the consistency of fringe orders in the connected region of wrapped phase. Last, phase demodulation was performed. The experimental results were obtained to confirm the effectiveness of the proposed method in the phase demodulation of both discontinuous objects and highly abrupt objects.

The in-situ observation is of great significance to the study of the growth mechanism and controllability of two-dimensional transition metal dichalcogenides (TMDCs). Here, the differential reflectance spectroscopy (DRS) was performed to monitor the growth of molybdenum disulfide (MoS2) on a SiO2/Si substrate prepared by chemical vapor deposition (CVD). A home-built in-situ DRS setup was applied to monitor the growth of MoS2 in-situ. The formation and evolution of monolayer MoS2 are revealed by differential reflectance (DR) spectra. The morphology, vibration mode, absorption characteristics and thickness of monolayer MoS2 have been confirmed by optical microscopy, Raman spectroscopy, ex-situ DR spectra, and atomic force microscopy (AFM) respectively. The results demonstrated that DRS was a powerful tool for in-situ observations and has great potential for growth mechanism and controllability of TMDCs prepared by CVD. To the best of the authors’ knowledge, it was the first report in which the CVD growth of two-dimensional TMDCs has been investigated in-situ by reflectance spectroscopy.

The accurate measurement of slope displacement profiles using a fiber Bragg grating flexible sensor is limited due to the influence of accumulative measurement errors. The measurement errors vary with the deformation forms of the sensor, which dramatically affects the measurement accuracy of the slope displacement profiles. To tackle the limitations and improve the measurement precision of displacement profiles, a segmental correction method based on strain increments clustering was proposed. A K-means clustering algorithm was used to automatically identify the deformation segments of a flexible sensor with different bending shapes. Then, the particle swarm optimization method was adopted to determine the correction coefficients corresponding to different deformation segments. Both finite element simulations and experiments were performed to validate the superiority of the proposed method. The experimental results indicated that the mean absolute errors (MAEs) percentages of the reconstructed displacements using the proposed method for six different bending shapes were 1.87%, 5.28%, 6.98%, 7.62%, 4.16% and 8.31%, respectively, which had improved the accuracy by 26.83%, 18.94%, 29.49%, 26.35%, 7.39%, and 19.65%, respectively. Therefore, it was confirmed that the proposed correction method was competent for effectively mitigating the measurement errors and improving the measurement accuracy of slope displacement profiles, and it presented a vital significance and application promotion value.

The measurement accuracy of the intelligent flexible morphological sensor based on fiber Bragg grating (FBG) structure was limited in the application of geotechnical engineering and other fields. In order to improve the precision of intellisense for displacement, an FBG implantable flexible morphological sensor was designed in this study, and the classification morphological correction method based on conjugate gradient method and extreme learning machine (ELM) algorithm was proposed. This study utilized finite element simulations and experiments, in order to analyze the feasibility of the proposed method. Then, following the corrections, the results indicated that the maximum relative error percentages of the displacements at measuring points in different bending shapes were determined to be 6.39% (Type 1), 7.04% (Type 2), and 7.02% (Type 3), respectively. Therefore, it was confirmed that the proposed correction method was feasible, and could effectively improve the abilities of sensors for displacement intellisense. In this paper, the designed intelligent sensor was characterized by temperature self-compensation, bending shape self-classification, and displacement error self-correction, which could be used for real-time monitoring of deformation field in rock, subgrade, bridge, and other geotechnical engineering, presenting the vital significance and application promotion value.

We proposed and studied an impact detection system based on a fiber Bragg grating (FBG) sensor array and multiple signal classification (MUSIC) algorithm to determine the location and the number of low velocity impacts on a carbon fiber-reinforced polymer (CFRP) plate. A FBG linear array, consisting of seven FBG sensors, was used for detecting the ultrasonic signals from impacts. The edge-filter method was employed for signal demodulation. Shannon wavelet transform was used to extract narrow band signals from the impacts. The Gerschgorin disc theorem was used for estimating the number of impacts. We used the MUSIC algorithm to obtain the coordinates of multi-impacts. The impact detection system was tested on a 500 mm × 500 mm × 1.5 mm CFRP plate. The results show that the maximum error and average error of the multi-impacts’ localization are 9.2 mm and 7.4 mm, respectively.

This paper presents a signal correlation-based two-sided detection method to detect delamination defects in widely used carbon fiber reinforced plastics with high accuracy and a convenient process. This method can improve the accuracy of ultrasonic testing and distinguish non-defective signals from defective signals, especially those with small amplitude, and to present the depth and size of defects by images. We combined the echo signals from both sides of ultrasonic detection at the same location to extract the weak defect signals and achieved the detection of materials with high ultrasonic attenuation. First, empirical mode decomposition filtering is performed on the ultrasonic raw signal. Then, the defect echoes are identified by smart thresholding and combined with the results of ultrasonic detection on both sides. Next, the defect depth and size are calculated, and the defect image is drawn. Finally, the ultrasonic phased array C-scan obtained by the ultrasonic phased array equipment was compared with the detection results of carbon fiber laminates with artificial delamination by the algorithm of the present invention, showing that the proposed algorithm defect depth calculation error is less than 4 %, the defect size calculation error is less than 0.5 mm, and it performs well in defect shape presentation and position calculation. [ABSTRACT FROM AUTHOR]

Due to their robust learning and expression ability for complex features, the deep learning (DL) model plays a vital role in bearing fault diagnosis. However, since there are fewer labeled samples in fault diagnosis, the depth of DL models in fault diagnosis is generally shallower than that of DL models in other fields, which limits the diagnostic performance. To solve this problem, a novel transfer residual Swin Transformer (RST) is proposed for rolling bearings in this paper. RST has 24 residual self-attention layers, which use the hierarchical design and the shifted window-based residual self-attention. Combined with transfer learning techniques, the transfer RST model uses pre-trained parameters from ImageNet. A new end-to-end method for fault diagnosis based on deep transfer RST is proposed. Firstly, wavelet transform transforms the vibration signal into a wavelet time-frequency diagram. The signal's time-frequency domain representation can be represented simultaneously. Secondly, the wavelet time-frequency diagram is the input of the RST model to obtain the fault type. Finally, our method is verified on public and self-built datasets. Experimental results show the superior performance of our method by comparing it with a shallow neural network. [ABSTRACT FROM AUTHOR]