Objective In optical three-dimensional (3D) shape reconstruction, phase unwrapping is a critical step that directly affects reconstruction accuracy and efficiency. Traditional temporal phase unwrapping methods rely on multi-frame input, limiting their adaptability to dynamic scenes due to poor real-time performance. Meanwhile, spatial unwrapping techniques, though capable of single-frame processing, are prone to spectral aliasing, leading to significant phase jump errors at edges. Current deep learning-based methods face three core limitations: accuracy bottlenecks under single-frame input, with noticeable prediction errors in complex texture and jump regions of industrial parts; contradictions between interpretability and robustness, as "black-box" models hinder error tracing while physics-integrated models suffer from reduced convergence efficiency due to complex constraints; and the challenge of balancing efficiency and precision, where lightweight models sacrifice detail extraction and high-precision models fail to meet real-time requirements in industrial online detection. To address these issues, this study proposes a multi-task deep learning framework for high-precision absolute phase unwrapping under single-frame input conditions. Methods A multi-task architecture model (MTU-SEDD) featuring a shared encoder and dual decoders is developed. The model incorporates three key innovative designs: 1) a shared encoder-dual branch decoder structure, where both the encoder and decoder employ multi-branch dilated convolution modules. These modules utilize convolution kernels with different dilation rates (e. g., dilation rate 1 for focusing on local fringe deformation and rate 4 for capturing global order distribution patterns) to extract multi-scale fringe features, enabling simultaneous prediction of wrapped phase and fringe order from a single input frame. 2) A channel-spatial hybrid attention (CSHA) module is specifically designed for fringe projection images. In the frequency domain, this module performs precise channel weight calibration through global average pooling and fully connected layers; in the spatial domain, it introduces anisotropic Gaussian kernels to enhance feature responses in phase jump regions, effectively suppressing noise interference. 3) A dynamic gradient balancing strategy is proposed, which adaptively adjusts task weights by calculating the gradient norms of task losses is proposed. This strategy mitigates gradient competition between the wrapped phase regression task [with mean-square error (MSE) loss] and the fringe order classification task (with cross-entropy loss), ensuring balanced convergence of both tasks. Experiments were conducted on both simulated and real datasets. The simulated dataset was constructed using a random virtual object surface generator, producing height maps with 1-3 convex structures (including rectangles, ellipses, polygons, and noise-disturbed surfaces) with heights ranging from 5 mm to 60 mm and resolutions of 128x128 and 480x640. The real dataset consisted of 1120 groups of deformed fringe images of objects with diverse materials (plaster, metal, and plastic), covering complex surface deformations (grooves, edge mutations) and specular reflection interference. Performance was evaluated using metrics including structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), normalized root mean square error (NRMSE), and root mean square error (RMSE), with comparisons against traditional convolutional neural network (CNN), DL-SPU, MultiResHNet, and U-Net (double) models. Ablation experiments were also conducted to verify the contributions of key modules, including the multi-task balancing strategy, dual-branch structure, and multi-branch dilated convolution. Results and Discussions Experimental results clearly demonstrate the superiority of the proposed method. In simulated dataset tests, MTU-SEDD achieved significantly lower RMSE than the traditional CNN model across all samples-for instance, image 1 showed an RMSE of 0.9260 for MTU-SEDD versus 8.9656 for CNN. The model maintained stable performance across different resolutions, with an RMSE fluctuation of <= 0.15 between 128x128 and 480x640 simulations. In real scenario experiments, MTU-SEDD outperformed other single-frame models, achieving the highest SSIM (0.9885), PSNR (54.64 dB), and the lowest NRMSE (0.0019), with an inference speed of 75.1 ms/frame. Ablation experiments further validated the necessity of each module: removing the multi-task balancing strategy (Exp-1) reduced SSIM by 0.72% and increased edge errors due to the marginalization of the fringe order task; Independent training of dual branches (Exp-2) led to a 131 % increase in runtime and accuracy degradation due to the lack of shared feature learning and potential misalignment between branch outputs; replacing multi-branch dilated convolution with ordinary convolution (Exp-3) caused severe detail loss, dropping SSIM to 0.8847 as the model failed to capture multi-scale features. Specifically, the CSHA module reduced edge region errors by 68%, while the dynamic gradient balancing strategy decreased inference time by 40 %and avoided multi-frame registration errors. Conclusions This study successfully develops a multi-task learning-based deep learning framework for structured light phase unwrapping. The shared encoder-dual branch decoder structure enables joint optimization of wrapped phase regression and fringe order classification tasks, achieving high precision and real-time performance. The channel-spatial hybrid attention module effectively suppresses spectral aliasing by enhancing phase jump region features. The dynamic gradient balancing strategy ensures stable model performance across different resolutions, demonstrating strong generalization. Future work will focus on integrating topology repair algorithms with 3D attention mechanisms and achieving model lightweighting via knowledge distillation to further accelerate training and inference.

Objective As chip technology continues to evolve toward miniaturization and high performance, the ongoing reduction in packaging size imposes increasingly severe constraints on solder joint reliability. Statistical data indicates that approximately 70% of chip failures originate from packaging-related issues, among which solder joint failure is the dominant cause. Among various defect types, void defects are a critical factor affecting the thermal fatigue life of solder joints. Specifically, these voids act as initiation sites for intergranular fatigue cracks, readily leading to solder ball fracture and failure. This issue is particularly pronounced in ball grid array (BGA) packaging. As a key technology for interconnection and mechanical support in high-performance chips, BGA solder balls are prone to the formation of internal voids during ball placement and secondary reflow soldering processes, thereby compromising the long-term reliability of the chip. Traditional defect detection methods, such as threshold-based segmentation, Fourier transform techniques, and low-rank sparse decomposition models, are highly dependent on image quality. Due to the limitations imposed by X-ray penetration mechanisms and material density variations, minute voids within solder balls typically exhibit low contrast in projection images. Consequently, traditional algorithms relying on handcrafted features struggle to construct robust segmentation boundaries and are highly susceptible to interference from internal chip structures, metal artifacts, and imaging noise, leading to high false-positive rates and poor adaptability to complex industrial scenarios. Moreover, existing detection methods often exhibit insufficient accuracy when dealing with voids characterized by small scales, blurred boundaries, and occlusions in BGA X-ray images. To address these challenges, this study proposes an improved YOLOv8n-based BGA solder ball void detection network, termed SGC-YOLO, which demonstrates notable advantages in detecting minute and occluded void defects. The proposed design strategy and experimental results are expected to provide valuable insights and practical references for void defect detection in chip solder balls. Methods The SGC-YOLO network incorporates a semantic-guided cross-layer (SGC) framework into the neck architecture to enable the synergistic fusion of shallow texture features and deep semantic features, thereby enhancing the detection capability for small targets with similar visual patterns. In addition, a semantic-guided local feature fusion (LFF) module is designed to strengthen void feature extraction and improve feature discriminability through the coordinated mechanisms of detail preservation, semantic enhancement, and background suppression. Furthermore, an improved feature alignment and redistribution (FAR) module is introduced to regulate the arrangement and activation distribution of shallow and deep feature channels. This design promotes sufficient inter-channel interaction and balance, effectively alleviating the issue that shallow feature responses are easily overwhelmed by dominant deep semantics under blurred-edge conditions, thus improving the effectiveness of multi-level feature fusion. Finally, a smooth geometric loss (SGL) function is proposed. By exploiting the circular geometric characteristics of void defects, the rectangular prediction boxes are transformed into equivalent circular representations. Meanwhile, continuous gradients are constructed based on the center distance and scale discrepancy between predicted and ground-truth boxes, which effectively mitigates the gradient vanishing problem encountered in the localization of minute and edge-blurred voids, enabling the predicted boxes to better conform to the actual defect contours. Results and Discussions Experimental results demonstrate that SGC-YOLO achieves a mean average precision (mAP) of 91.9 % on the self-built dataset. Compared with Faster R-CNN, YOLOv8, YOLOv11, YOLOv12, RT-DETR, and Gold-YOLO, the proposed method yields performance improvements of 6.8 percentage points, 4.3 percentage points, 4.3 percentage points, 4.6 percentage points, 7.2 percentage points, and 5.0 percentage points, respectively, indicating a clear advantage in detecting minute and occluded void defects. Overall, SGC-YOLO exhibits detection performance that is consistently superior to that of the comparative methods. In particular, for voids that spatially overlap with lead structures-scenarios in which conventional methods frequently suffer from missed detections-SGC-YOLO effectively discriminates void defects from background interference by leveraging the enhanced feature fusion strategy, thereby achieving accurate detection. Moreover, for minute void defects, SGC-YOLO demonstrates higher detection sensitivity than competing approaches, which substantially reduces the occurrence of missed detections. Conclusions Addressing the detection challenges posed by void defects within BGA solder balls-characterized by minute scale, low contrast, blurred boundaries, and susceptibility to interference from complex internal structures-this study develops an improved detection network, termed SGC-YOLO, based on YOLOv8n and oriented toward practical engineering applications. By introducing a semantic-guided cross-layer feature fusion framework, along with local feature enhancement and channel alignment and redistribution strategies, the proposed model effectively coordinates the complementary roles of shallow detailed features and deep semantic representations, thereby substantially enhancing the characterization of minute and occluded voids under complex backgrounds. In addition, a Smooth Geometric Loss function is designed in accordance with the geometric properties of solder voids, which improves gradient propagation and localization stability for small-scale targets during the training process. Experimental evaluations demonstrate that SGC-YOLO achieves a mean average precision of 91.9 %, outperforming representative mainstream detectors-including Faster R-CNN, RT-DETR, and various YOLO-based models-by margins ranging from 4.3 percentage points to 7.2 percentage points, with particularly pronounced advantages in scenarios involving minute voids and structural overlap. Overall, the results confirm the effectiveness and engineering applicability of the proposed method for BGA solder void detection, providing reliable technical support for chip packaging quality assessment and long-term reliability analysis.

With the deceleration of Moore's Law, the improvement of chip performance is transitioning from pure transistor scaling to the parallel advancement of device miniaturization and advanced packaging technologies, aiming to achieve higher levels of system integration and interconnect density. Following this trend, copper pillar bumps, owing to their miniature dimensions, fine pitch, and outstanding electrical and thermal properties, have emerged as essential interconnect structures in 2.5D and 3D packaging technologies. However, copper pillar bump technology continues to encounter issues, including elevated interconnection temperatures, inadequate thermo-electric reliability, and constrained performance under ultra-high-density conditions. This review provides an overview of the research background, current progress, and prospective trends in copper pillar bump technology. Beginning with the basic fabrication processes, this paper elaborates on the evolution of material systems for copper pillar bumps and the most recent advancements in novel fabrication techniques. Next, the underlying mechanisms and suitability of different interconnection approaches are examined. Moreover, advances in reliability testing, failure mechanisms, and optimization strategies for interconnections are summarized, providing theoretical guidance for extending their service life and enhancing long-term reliability. Finally, the prospects and research directions of copper pillar bump technology are discussed.

Induction motors are widely used in the industrial field such as electric drive systems for new energy vehicles and synchronous condenser for improving the power factor of the power grid. The motor health condition often influences the operation of the entire mechanical system, so it is necessary to conduct a health assessment on it. Current induction motor fault diagnosis largely relies on expert knowledge, while many deep learning methods suffer from limited generalization and poor interpretability, leading to unreliable results. To address these issues, a fault mechanism-guided interpretable causal disentanglement domain generalization detection method (ICGN) is proposed for typical fault diagnosis of induction motor. Firstly, a primary feature extractor is constructed based on transformer, which adaptively screens causal and non-causal factors through the self-attention mechanism, and an attention score evaluation mechanism is constructed to visually demonstrate interpretability. Secondly, to further disentangle and refine causal features and non-causal features, the developed causal aggregation loss and causal decoupling loss are combined, ensuring the cross-domain consistency of causal factors and promote the domain generalization ability of the network. Finally, the proposed method is validated using vibration signals collected from two Spectra Quest test benches from University of Ottawa and the private laboratory. The cases of cross device motor fault diagnosis are included, and the ICGN is compared with several advanced domain generalization algorithms. The results demonstrate that the proposed method achieves superior performance both in interpretability and domain generalization capability.

In industrial operational environments, rolling bearing vibration signals not only contain fault related periodic impulse components, but also interference noises. The coupling of these signals attenuates the fault characteristics, particularly in cases of compound faults, significantly affecting the diagnostic accuracy of bearings. This paper proposes a rolling bearing fault separation and compound diagnosis method using a sparse Bayesian framework with adaptive prior knowledge. Firstly, the proposed method segments signals into a finite number of modes, employing filter banks to adaptively select decomposition modes with a new compound fault indicator, which reduces noise interference and eliminates irrelevant components while ensuring the preservation of fault-related information. Secondly, by employing an interpretable strategy enables fault frequency estimation, eliminating errors from empirical parameter settings and avoiding the computational complexity of parameter optimization. The feature frequencies are estimated based on the envelope harmonic product spectrum and used as a prior knowledge of sparse Bayesian learning. Finally, to avoid the restriction of current feature extraction methods that primarily focus on the selection and extraction of a single demodulation frequency band, a sparse Bayesian probabilistic model is particularly designed for compound fault diagnosis. Simulation and experimental results show that the proposed method can effectively extract and separate the fault features of each individual modes and realize the diagnosis of rolling bearings.

To enhance the understanding of blade non-synchronous vibration from the blade tip timing perspective, a single-stage centrifugal compressor representative of industrial architecture has been investigated experimentally. The characteristic of the wall pressure and blade tip timing (BTT) signal is first measured and analyzed experimentally under rotating stall condition. The relationship between the rotor vibration in the stationery and rotor reference frame is then established. A combination of BTT and strain gauge gives a full interpretation of non-synchronous vibration phenomenon and identifies the fluid-structure interaction parameters. To further verify the rotating stall state and corresponding signal characteristics, a numerical simulation of unsteady flow and aerodynamic forcing function is also conducted. The numerical results match the experimental findings well and can interpret the underlying physics. These results contribute to an in-depth understanding of the aeroelastic phenomena in industrial centrifugal compressors.

Flip chips are widely used in electronic systems for defense, aerospace, and other applications where packaging reliability is critical. However, flip chip defect samples present a variety of defect types and few samples with labels in actual industrial applications. The paucity of labeled defect samples indicates that the existing data volume cannot be matched with deep learning detection models. Therefore, flip chip intelligent defect detection faces the problems of poor model adaptability and weak generalization performance. As a solution to these problems, a semi-supervised dual-constraint centroid contrastive prototypical network (SSDCPN) for flip chip defect detection under limited labeled data is proposed in this paper. First, a prototype-based supervised contrastive learning strategy is developed to construct the contrastive prototypical network, which increases the inter-class sparsity and intra-class compactness of features to acquire more discriminative features. Then, to address the susceptibility of the support set prototypes to outliers, dual constraints are imposed on the support set prototypes to calibrate and refine the prototypes. Finally, a pseudo-labeled sample selection mechanism based on epistemic uncertainty and entropy is proposed to obtain rich semi-supervised information to guide the model training. The mechanism can select high-confidence pseudo-labeled samples that can complement the training samples to further strengthen the generalization performance of the model. Defect detection experiments on flip chip vibration signals indicate that the present method is superior to other methods in the case of limited labeled samples.

To achieve effective detection of loose particles within small cavities in microelectronic devices, this paper proposes the optimal detection conditions of loose particles in ceramic packaging for microelectronic devices based on simulation, and presents a method for weak signal reconstruction using weighted sparse representation. To address the challenge of ineffective detection of loose particles under recommended vibration conditions, a particle-cavity collision dynamics model is established. Subsequently, the detection laws of loose particles under different accelerations, frequencies, and cavity heights are studied through simulation, leading to the establishment of optimal detection conditions for microelectronic devices with varying cavity heights. To address the issue of weak impulses being submerged in background noise under optimal detection conditions, sparse representation is utilized for signal reconstruction. First, a Laplace wavelet dictionary is constructed according to the impulse characteristics. Second, the generalized minimax-concave (GMC) penalty function is applied as a sparse regularization term to preserve signal amplitude. Meanwhile, a weight matrix based on kurtosis and singular value is designed to threshold sparse coefficients. The results demonstrate that the proposed optimal detection conditions and signal reconstruction method effectively detect loose particles. Compared with other algorithms, the proposed method reduces background noise in the particle impact noise detection (PIND) signals while maintaining impulse amplitude; it also improves signal reconstruction accuracy and offers valuable insights into effective loose particle detection in microelectronic packaging devices.

The effective extraction of weak defect features is crucial to the reliability analysis of flip-chip microbumps. A novel dual-drive adaptive reweighted sparse (DDARS) framework is proposed, which can skillfully realize the time-varying noise suppression and unsupervised feature stabilization of flip-chip vibration signals under large-scale environmental disturbances. Specifically, the compact representation (CR) of patch operators. and its inverse Lambda(-1) are defined to segment the high-dimensional data into low-scale time-varying noisy patches. By integrating the prior knowledge of noise variance distribution and singular value sparse distribution, an adaptive knowledge prior reweighted matrix(ARM) is constructed to precisely reweight and match sparse regularization terms, thereby effectively mitigating time-varying noise interference. Quadratic sparse projection constraints are introduced under the multi-patch sample condition, utilizing quadratic sparsity to evaluate feature saliency and stabilize the potential representation of signal features. Based on the potential connection between patches and sweep-frequency signals, analyze the sparsity intensity of signals in different time periods to guide the optimal frequency range for ultrasonic excitation of chips. The validity of DDARS is verified through qualitative and quantitative analysis with simulated signals and data from engineering application examples, which can provide technical support for flip-chip microbumps reliability analysis.

This article proposes a sparse denoising method for ultrasonic signals based on the dual memory-enhanced deep unfolding network (DME-DUN), which aims to accurately reconstruct high-frequency ultrasonic signals in flip chip detection. To address information loss during the iterative process in deep unfolding networks, two memory enhancement mechanisms are designed, i.e., adjacent-stage and cross-stage. For adjacent-stage, the high-throughput information from the previous stage is introduced into the current stage, thus facilitating the transmission of information between adjacent stages. Additionally, the cross-stage memory enhancement mechanism is applied to promote information fusion between separated stages. This mechanism is based on Richardson extrapolation and calculates the output error between every two adjacent stages to update the sparse coefficient. Extensive experiments verify that the proposed method effectively captures weak echo features while removing noise.

Flip chip technology has been extensively adopted in integrated circuit (IC) packaging, and the integration of flip chip technology with solder joint interconnection methods has been utilized in the manufacturing of electronic devices universally. As the development of flip chip towards high density and ultra-fine pitch, the inspection of flip chips is confronted with great challenges. Aiming at the characteristics of the chip signal with strong interference noise and weak feature information, the sparse representation model under the reweighting constraint is proposed. We analyzed the low-rank prior of the signal feature matrix and explore the construction of a sparse representation dictionary to establish an adaptive singular value decomposition (SVD) dictionary for the variant features. Subsequently, we investigate the adaptive weight matrix that aligns with the time-varying structure of the noise and the distribution pattern of the feature singular values, and construct a sparse characterization model under the reweighting constraint. Simulation analysis shows that the proposed algorithm can effectively realize the filtering of time-varying noise in the signal, so as to reliably extract the weak defect features.

The flip-chip technology is widely used in aerospace and defense electronic systems because of its high information processing, rapid response and autonomous control. As flip-chip technology advances towards higher density and finer pitch, the mechanical reliability of chip packaging will decrease, with solder bumps defect being the most common and severe concern. To ensure the safety and availability of electronic systems in industrial production applications, it is very necessary to develop the nondestructive solder bumps detection technology with less manual intervention, lightweight architecture and high detection accuracy. In this study, a self-search lightweight framework (SLF), which integrates architecture feature search (AFS) and logic-space decoupling distillation (LDD), is developed to analyze vibration signals and identify different defects of flip-chip solder bumps. AFS makes up for the three sub-vulnerabilities in existing neural architecture search, which provide a homotypic combination AFST-AFSS for SLF. LDD decouples the logic space of traditional knowledge distillation into 3-probability spaces, which learn logical feature representations with class boundary discriminability, and facilitates feature migration in the SLF. The vibration noise signal contains a large amount of state information, and the combination with artificial intelligence technology will further support the high precision, efficiency, and reliability of detection. We collected vibration signals of flip-chips through ultrasonic excitation experiments and used them as data input for the proposed method. Specially, the dual-convergence property of AFS and the semantic mapping capability of LDD in SLF have been investigated respectively with the support of this vibration signals. Furthermore, an in-depth investigation of orthogonal analysis and adaptive performance reveals that SLF, which is a combination of AFS and LDD, delivers the best lightweight detection performance compared with other popular methods, which has shown tremendous potential on deploying edge equipment with limited computing power to complete detection in real industrial environments.

The widespread use of flip-chip technology in the field of microelectronics packaging makes defect detection technology face great challenges, which requires the development of detection technology with less manual intervention, a lightweight network architecture, and high precision. Aiming at the dual-stability of the neural architecture search (NAS), and explicitly associating the network architecture searched by NAS with the architecture features of vibration signals. This study proposes a dual-convergence sparse feature extractor (DSFE) for visualization vibration signals architecture feature searching, which attempts to learn a representation whose architecture feature can uniquely represent the intrinsic information of the target signal. DSFE analyzed and summarized three deficiencies of gradient-based NAS, namely, “huge GPU memory consumption,” “high collapse probability,” and “rigid number of node precursor operations.” The corresponding solution modules are proposed in turn, which are “primary and secondary search spaces,” “skip_connect coefficient modification,” and “dynamic sparse selection of precursor operations.” The effectiveness of DSFE is verified with flip-chip vibration signals excited by the air-coupled ultrasonic wave. By comprehensive comparative experiments with fixed-structure networks and gradient-based NAS methods, it is proved that DSFE cannot only achieve dual-stability of detection precision and architecture searched, but also ensure less resource consumption and multifarious architecture, which can broaden the application scope of DSFE in practical engineering.

Currently,experimental research on variable stiffness design mainly focuses on lami-nates.To ensure adaptability in practical application,it is imperative to conduct a systematic study on stiffened variable stiffness structures,including design,manufacture,experiment,and simula-tion.Based on the minimum curvature radius and process schemes,two types of T-stiffened panels were designed and manufactured.Uniaxial compression tests have been carried out and the results indicate that the buckling load of variable stiffness specimens is increased by 26.0％,while the fail-ure load is decreased by 19.6％.The influence mechanism of variable stiffness design on the buck-ling and failure behavior of T-stiffened panels was explicated by numerical analysis.The primary reason for the reduced strength is the significantly increased load bearing ratio of stiffeners.As experimental investigations of stiffened variable stiffness structures are very rare,this study can be considered a reference for future work.

Surface defect detection is very important to ensure the quality of industrial products. Traditional machine learning cannot be well extended to a non-identically distributed dataset, making surface defect detection a data-limited task. Unsupervised domain adaptation (UDA) can solve this problem by transferring knowledge from a labeled source domain to an unlabeled target domain. Adversarial learning is one of the latest heuristic methods to deal with domain shift in UDA tasks. Although impressive results have been achieved, the adversarial model still suffers from the equilibrium challenge, which may lead to under-transfer or negative transfer. To this end, we utilize joint distribution adaptation to propose a novel UDA model, named significance-augmented-based category-level selective dual-adversarial (CDASA) network, to learn a generalized model. Specifically, to promote positive transfer, we use the selective dual-adversarial (DA) learning strategy to further minimize the feature distribution difference at equilibrium to achieve better domain confusion. Meanwhile, guided by defect recognition performance, the transferability of confusion domain is measured to enhance the distributions of potential domains. In addition, to avoid under-transfer, we consider the relationship between the target data and the decision boundary, and then the significance-augmented (SA) mechanism is proposed to encourage class-level alignment. Thus, the alignment features with domain-invariant and category discrimination can be captured simultaneously. Extensive experiments on collected real industrial datasets and publicly available steel surface defect datasets confirm the effectiveness of our approach.

To achieve effective detection of loose particles within small cavities in microelectronic devices, this paper proposes the optimal detection conditions of loose particles in ceramic packaging for microelectronic devices based on simulation. Through establishing a particle-cavity collision dynamics model in ADAMS, the motion process of particles under sinusoidal excitation is simulated. Subsequently, the paper examines the impact of different accelerations, frequencies, loose particles' properties, and cavity heights on optimal detection conditions. Furthermore, the paper presents the optimal conditions for vibration testing, providing a valuable reference for the effective detection of loose particles in hermetically sealed microelectronic devices.

It is of great significance to detect solder defects of ceramic packaging chips by X-ray detection to improve the quality of electronic products. However, the scale of the solder defects varies violently and there are a lot of tiny, dense, and long narrow defects, making it difficult to identify timely and accurately. To solve the issues mentioned above, an improved object detection model based on the YOLOv5 network, namely YOLO-STPN, is proposed in this paper. We add another prediction head to the original model to detect tiny solder defects effectively. In addition, the swin transformer block and convolutional block attention module are integrated into the path aggregation network, which improve the networks' ability to identify dense and long narrow defects. The positive sample matching mechanism and the complete intersection over union (CIoU) loss are modified to make the model focus on long narrow defects. Experimental results of 10 types of ceramic packaging chips show that the mean average precision at 50 % IoU (mAP50) of YOLO-STPN is 10.05 % higher than the original YOLOv5 on average, while the decrease of frames per second (FPS) is acceptable. The proposed method has high accuracy and speed. Therefore, it is applicable to the real-time detection of chip packaging.

针对气体绝缘开关设备(gas insulated switchgear,简称GIS)断路器声信号易受强背景噪声干扰且特征提取困难的问题,提出一种可以有效诊断其机械故障的方法.首先,使用多通道声传感器阵列采集断路器原始观测信号,利用独立成分分析(independent component analysis,简称ICA)方法将观测信号分离为多维源信号,并选取源信号中模糊熵最小的分量作为特征信号;其次,计算特征信号的多尺度模糊熵(multi-scale fuzzy entropy,简称MFE)生成断路器的声纹特征;最后,利用极限学习机(extreme learning machine,简称ELM)算法识别断路器的故障.实验结果表明,基于声信号的检测方法为GIS断路器的机械故障诊断提供了一种新的解决方案,所提出的算法能够有效提取声纹特征,故障诊断准确率较传统方法有明显提高.

Although deep learning has achieved remarkable results in the industry, to achieve the ultimate prediction accuracy improvement, many existing models with poor performance will be added with other modules, which will become more and more bloated. To achieve the deployment of the lightweight network in an industrial system, knowledge distillation (KD) is currently an effective way to reduce the model size. This study focused on the research of knowledge transfer logic and decoupled the interactive information into three-probability spaces, namely, the target space ( $t$ ), the intraclass class space ( $o\backslash t$ ), and the out-of-class class space ( $c\backslash o$ ). The three-probability spaces method is named TOCKD ( $t$ - $o\backslash t$ - $c\backslash {o}$ spaces KD), which introduces three components corresponding to three-probability spaces, that is, target KD (TKD), intraclass KD (OKD), and out-of-class KD (CKD). Based on the support of the flip-chip vibration signal with defect information obtained by ultrasonic excitation experiment, the research confirms the primacy of TKD in the information transmission process and the auxiliary properties of the other two components by a qualitative analysis. To realize the quantitative analysis of the corresponding coefficients of three components in TOCKD, a tree-structured Parzen estimator (TPE) is used to search the coefficients. On the basis of it, combined with subjective and objective weight analysis methods, the guide value of coefficients is defined. This research verifies the excellent performance and robustness of TOCKD by analyzing the flip-chip vibration signals and hopes that TOCKD will be helpful for future research and the application of lightweight technology in the engineering field.

The defect inspection of flip chips has become a meaningful and challenging task with the decrease of size and distance of solder balls. In this paper, we proposed a novel defect diagnostic method for flip chips based on vibration signals and improved multi-grained cascade forest (gcForest). Firstly, the flip chip is excited by an air-coupled capacitive ultrasonic transducer, and the corresponding vibration signals of flip chip are captured by a laser scanning vibrometer. Then, the feature information of original vibration signals is extracted by muti-grained scanning (MGS) automatically. Finally, the extracted feature information is input into the cascaded forest for defect diagnosis. Considering the low data transmission efficiency and feature redundancy between MGS and cascaded forest, a feature extraction channel strategy based on kernel principal component analysis (KPCA) is introduced into the cascaded forest. Besides, in order to improve the generalization ability of the improved gcForest, the classifiers of each layer in cascade forest are upgraded. The results demonstrate the superiority of the improved gcForest over the traditional methods on experimental vibration data of flip chips, especially under small training sizes, which is very beneficial to practical engineering.