Design of an underground water leakage defect detection system using a lightweight image segmentation algorithm

ABSTRACT Magnetic Flux Leakage (MFL) testing is an important nondestructive testing technique. Safety performance assessment of pipelines is a key aspect in practice. The inversion of defect dimensions corresponding to the characteristics of pipeline leakage signals is a hot and difficult problem in the field of pipeline defect detection. In this paper, a novel algorithm for quantitative identification of defect leakage magnetic signals is proposed. Firstly, the original leakage data is denoised and the signal is processed to better reflect the characteristics of the defect signal. Secondly, the efficiency and accuracy of defect size quantification are improved by training SSA_BP neural network to estimate defect length and width. It can handle both quantitative and qualitative information, coordinate various input information relationships, and has strong robustness and fault tolerance. In addition, two neural networks are combined with the output of the length and width to estimate the depth indirectly. The feasibility of the method is demonstrated by a large amount of simulation and experimental data to serve as support. The results show that the quantification method proposed in this paper is more accurate in estimating the defect size. This will provide accurate defect size data support for pipeline structural integrity evaluation.

DevOps is a paradigm shift in software development today, aimed at the rapid delivery of software through a task-oriented approach to work and the integration of development and operations teams. Software testing is one of the most important and challenging steps in the DevOps pipeline because it should be effective and stable without being slow. To improve the process of software testing among DevOps professionals, this paper suggests a semi-automatic testing approach as a more viable solution between the extremes. The semi-automated methodology involves automation of processes that are tedious and time-consuming (such as unit testing, regression testing, and continuous integration), but human inspection in processes that require critical thinking, business domain knowledge, and exploratory insight (such as UI validation, business logic, and end-user behavior simulation). The combination creates the ability to do faster feedback loops, reduced test maintaining burden, improved test coverage, and quality assurance (QA) in general. The study utilizes a combination of case study evaluation, the implementation of the tools, and the reaction of the practitioner to the survey to gauge the performance of semi-automation in the DevOps context. The core performance indicators assessment of performance includes reduced test cycle time, defect leakage rate, productivity of the team, and the rate of deployment. The findings indicate that semi-automatic solutions result in radical improvements of test efficiency, scalability, and flexibility, especially in groups with dynamic codebases and limited automation budgets. The proposed study will integrate an affordable yet quality testing model that is dynamic enough to keep up with the DevOps and the advancement of the ideal balance between human judgment and automation. The method described here does not only enhance the effect of testing as an undertaking but also does the more universal DevOps objectives of quicker delivery, quality, and better teamwork.

Relevance. Given the rapid development of software engineering practices and the need for cost-effective quality assurance in competitive environments, the relevance of developing strategic planning approaches for combined testing is growing steadily. The object of research is the strategic planning process for combined software testing that integrates multiple testing methodologies through systematic framework implementation, risk assessment, and resource optimization algorithms. Purpose of the article. This study explores strategic planning approaches for combined software testing and assesses their effectiveness across various application domains. The article aims to provide a structured framework for testing strategy integration and evaluate optimization mechanisms for resource allocation in complex testing environments. Research results. A comprehensive Strategic Planning Framework for Combined Software Testing (SPF-CST) was developed, consisting of six interconnected components: context analysis, multi-dimensional risk assessment, AI-driven prioritization, strategy selection, resource optimization, and monitoring systems. Empirical validation across eight industry case studies demonstrates a 35% reduction in defect leakage rates, 28% improvement in testing efficiency, and 45% decrease in regression testing costs. The study revealed that strategic planning significantly enhances testing effectiveness through systematic methodology integration and adaptive resource management. Conclusions. The study demonstrates the effectiveness of risk-based prioritization and mathematical optimization in testing strategy selection. The proposed framework provides practical tools for organizations to implement comprehensive testing strategies while managing resource constraints and project timelines.

Solid oxide fuel cells (SOFCs) have the highest power-generation efficiency among generators, making them key to achieving a carbon-neutral society. To reduce the cost of an SOFC system, and thus expand its range of applications, we developed a thin-film SOFC suitable for high power density and low-temperature operation. Since the porous nanostructured anode/cathode layers and dense solid-electrolyte layer are formed by sputtering, the internal resistance of the thin-film SOFC drastically reduces. Thus, both high power density and low-temperature operation can be achieved [1, 2].To reduce the thickness of the electrolyte layer, a major problem specific to SOFCs should be solved. A porous electrode is required for SOFCs, because hydrogen (H2) and air should be supplied through the electrode. Therefore, a thin electrolyte layer should be deposited on a porous bottom electrode, which makes it difficult to reduce the electrolyte thickness without leakage failure. Process technologies to reduce the unnecessary roughness of the bottom layer of the electrolyte can improve the limit of electrolyte thinning [3], which enables the reduction of the electrolyte layer to less than 100 nm [4, 5]. However, a small fraction of leakage failure can degrade an SOFC system, which is the practical hurdle to drastically reduce the electrolyte thickness.Cell partitioning and partition management, which are derived from similar technologies in semiconductor products, are suitable solutions for eliminating low-density defects and preventing system failure. An array of small cells on an anodic aluminum oxide (AAO) substrate forms a cell layer, in which only cells without leakage defects are electrically connected; such a selective connection is possible by screening of failed cells prior to stack integration. Figure 1 shows the partitioned cells formed on an AAO substrate. While the bottom electrode (anode layer) and yttria-stabilized zirconia (YSZ: electrolyte layer) are deposited on the entire AAO surface, multiple electrodes (cathode layers) are formed on the top surface, creating partitioned cells. A probe test prior to the stack integration enables us to remove the partitioned cells with large leakage currents.Figure 2 shows the cell yields and normalized power densities as a function of the normalized YSZ thickness. Without leakage failure, the normalized power increases with reduced YSZ thickness and levels off due to the polarization resistance independent of the YSZ thickness. With a defect-induced failure model [6], the yield is derived as exp(-AD), where A and D are the surface area and defect density, respectively. The yield of a conventional cell without partition management (10 cm × 10 cm) can be decreased compared with that of each partitioned cell (3 mm × 3 mm) because of its 1000-time larger surface area. The yield of conventional cells decreases sharply by reducing the YSZ thickness, so the thickness should be large enough to maintain high yield (for example 90%). However, cell partitioning and partition management can improve the yield of each partitioned cell by virtue of the reduced cell size. The failure of the entire cell array on an AAO substrate can be avoided by removing the failed partitioned cell, thus solving the most significant problem of leakage failure in thin film SOFCs. The normalized power of the entire cell array increases as YSZ thickness becomes thinner but finally decreases because most of the partitioned cells are removed due to leakage failure. Therefore, the normalized power of the entire cell array reaches maximum (0.51) at a certain normalized electrolyte thickness. The normalized power improves by over 10 times compared with that of the conventional cell with a yield of 90% (0.044) as shown in Fig. 2.At this conference, we plan to discuss the fabrication process and performance of thin-film SOFCs.

Magnetic Flux Leakage (MFL) is a widely used non-destructive evaluation (NDE) technique for pipeline inspection. However, its signals are highly sensitive to noise and geometric distortions, causing small defects with limited spatial coverage and subtle defects with low-contrast patterns to be embedded in noise, resulting in indistinct boundaries and irregular shapes that complicate segmentation. To address these challenges, we propose PixTransNet, a hybrid CNN–Transformer model built on a UNet encoder–decoder architecture with a ResNet18 backbone, designed to improve the segmentation and boundary localization of small and subtle defects in MFL signals. We embed pixel-aware transformer blocks into the deeper encoder stages to capture long-range dependencies and enhance the modeling of subtle and fragmented defect patterns. To further enhance the interpretation of MFL signals, we introduce a cross-attention module that selectively emphasizes signal regions with strong structural relevance, leading to more continuous and accurate defect boundaries, particularly for small defects. Extensive experiments on a large-scale dataset of 33,000 MFL images demonstrate that PixTransNet achieves notable improvements in segmentation quality, particularly in detecting small, weak, and low-contrast defects compared to existing baselines. PixTransNet achieves 48.30% IoU, representing a 1.97% improvement, and 70.73% Recall, representing a 13.06% improvement over the best-performing baseline

This study presents an experimental analysis of high-pressure liquid and gas gate valve leakage under multiple operating conditions, based on the variation patterns of ultrasonic signals. Focusing on a multi-physics field analysis of gate valve internal leakage and corresponding experiments, this research illustrates the acoustic wave characteristics of gate valves across diverse working media, pressures, internal leakage defect sizes, and valve diameters. By drawing upon both fluid mechanics and acoustics theory, an analytical approach suited to high-pressure gate valve leakage issues is devised. Separate high-pressure gate valve leakage test platforms for liquid and gas environments were designed and constructed, enabling 126 groups of tests under varying conditions, which include one measurement per condition of the valve size, defect size, and pressure value. These experiments examine the quantitative correlation of internal leakage flow rates and ultrasonic signal measurements under different situations. In addition, the distinct behaviors and principles exhibited by high-pressure liquid gate valves and gas gate valves are compared. The findings provide theoretical and technical support for quantifying high-pressure gate valve leakage. The study analyzes the theoretical basis for the generation of ultrasonic signals from valve internal leakage, providing specific experimental data under various operating conditions. It explains the various observations during the experiments and their principles. The conclusions of this research have practical engineering value and provide important references for future studies.

ABSTRACT Si3N4 ceramic ball in ceramic bearings is a key component for the stability of aerospace equipment, and its surface scratches, pits, wear, spalling, and snowflake defects may cause accidents, so it is necessary to detect its quality. However, the existing detection methods mostly use top-down layer-by-layer extraction for feature learning and ignore the texture and location of small defects in multi-scale, resulting in insufficient learning of small features, which leads to defect leakage detection. In order to improve the detection ability of multi-type defects and the generalisation of multi-type defects detection, a small target detection method for multi-type ceramic ball surface texture defects (STC-YOLO) is proposed in this paper. Firstly, a shallow feature extraction module was designed to retain the semantics of the shallow feature maps and be used for feature fusion. Secondly, a deep and shallow parallel feature fusion module was designed, focusing on both local and global features, to address the issue of losing the local details and position information of small targets during the feature fusion process. The method was evaluated on the SOD-SNCB dataset, and the mean average precision (mAP50) of small target defect detection reached 0.992. The results show that STC-YOLO has excellent detection performance.

This article examines the role of Quality Assurance (QA) as a product function in multiplatform development, including medical software subject to HIPAA and FDA requirements. The goal is to demonstrate how QA approaches and metrics (defect leakage, stability, SLOs, regression speed) can be translated into manageable business outcomes: user retention, conversion growth, reduced support costs, and lower compliance risks. The methodology is based on an analytical review of risk-based testing practices, test strategy design, quality monitoring implementation, and the synthesis of case studies from large U.S. projects. The results are presented as a “Quality ? Trust ? Revenue” model, where quality is described as a system of decisions: risk prioritization, the test pyramid, observability, provable compliance, and continuous regression control. The practical value of the study lies in proposing a set of KPIs and management rituals for product teams that enable quality to scale without disproportionate cost growth.

Abstract This study presents a lightweight multi-scale self-calibrating feature fusion network (YSCANet) to address critical challenges in wafer defect detection, including low recognition accuracy for morphologically similar defects, large inter-class scale variations, and severe category imbalance. The proposed framework integrates three key innovations: First, the self-calibrating feature fusion block (SCFFB) to enhance multiscale defect feature representation through cross-scale channel recalibration. Second, the Context Anchor Attention (CAA) mechanism optimizes spatial-semantic correlations via anchor point weighting, effectively resolving feature ambiguity in tiny defects. Third, dynamic sample reweighting via Focal Loss to mitigate class imbalance effects. Experimental validation on the WM-811K dataset demonstrates that YSCANet achieves a state-of-the-art average classification accuracy of 97.72% (0.76% higher than existing state-of-the-art methods), while reducing defect leakage rates by approximately 50%. The network exhibits exceptional computational efficiency with an inference speed of 151.28 FPS and a compact parameter size of 2.07M. Comparative analysis shows that YSCANet has superior performance among similar defect detection solutions. These advances make YSCANet a powerful solution for real-time, high-precision wafer detection in semiconductor manufacturing, while balancing accuracy, efficiency and hardware deployment constraints.

In the OLED manufacturing process, the Cell Repair process plays a critical role in preventing defect leakage by identifying defects using an optical scope and then either repairing or rejecting them. The causes of Pixel‐Off defects addressed during this process can be categorized into two types: floating foreign substances on the top of the cell and foreign substances between the layered structures. However, in Rule‐Based Auto Repair Systems that rely on image processing, distinguishing between these two types of defects is not feasible. As a result, the affected panels are subjected to final judgment by a separate inspector, leading to a combination of issues such as decreased equipment efficiency, reduced productivity, and the need for additional inspectors. Therefore, the company has introduced Image Deep Learning (DL) techniques to enable the automatic classification of defect types, achieving significant automation in the process. However, during the initial production phase of new products, the lack of sufficient defect data sets for DL training results in very low model accuracy, making its application impractical. This proposal addresses the DL accuracy issue in the early stages of new product production by adopting a defect generation method using diffusion‐based generative AI, significantly enhancing the performance of the DL classification system. This technique requires only a single defect sample image in the early stages of new product development, eliminating the need for months of defect data collection. At the same time, it enables the immediate deployment of the defect classification DL system, which was previously unusable during the initial mass production phase of new products. The proposed method has been successfully implemented in the production line process.

ABSTRACT The elbow portion of a pipeline plays a critical role in pipeline systems, and defects can lead to potential gas and oil leakage accidents. Magnetic flux leakage (MFL) detection is an efficient method for identifying pipeline defects. To address the issue of image distortion in defect MFL signals, the accuracy of defect recognition must be improved. This paper proposes an intelligent identification method for MFL defects in small-diameter pipe elbows based on a deep learning target detection algorithm. The MSRCR algorithm is enhanced using bilateral filtering and gamma correction to improve the image features of defect MFL signals. Additionally, the YOLOv5 network is augmented with CBAM, Soft-NMS and Focal-EIOU loss to enhance the feature extraction capabilities. The results show that the proposed improved MSRCR algorithm effectively solves the elbow MFL defect distortion problem. The improved YOLOv5 network achieves average defect recognition accuracies of 86.61% and 94.27% on the original and enhanced datasets, respectively. After experimental verification, the network effectively improves the recognition accuracy of MFL defects in small-diameter pipe elbows, and provides useful technical support for the intelligent detection and safety evaluation of elbows.

Modelling and simulation of TSV considering void and leakage defects

Composite phase change materials (CPCMs) have promising applications as passive cooling technologies in the energy sector. However, the low thermal conductivity and obvious leakage defects of CPCMs limit their large-scale development. Herein, a highly thermally conductive CPCM with polyethylene glycol (PEG), metal–organic gel (MOG), and carbon microspheres (CMS) (PMEC) has been proposed and prepared via a hydrothermal method with noncovalent bonding of metal ions and organic acids. The scanning electron microscopy (SEM) and thermal conductivity of different PMECs indicate that the thermal conductivity of PMEC2 is significantly anisotropic, with 3.79 W·m–1·K–1 when the CMS content is 5 wt %. Besides, the experimental results indicate that PMEC2 can reach a latent heat value of 118 J/g, exhibiting excellent durability after 20 heating and cooling cycles. Additionally, a PMEC2-based battery module incorporating CMS and PEG with epoxy resin (ER) as the supporting skeleton has been assembled for battery modules. The charge and discharge tests are performed on the battery modules at different discharge rates. PMEC2 can control the temperature and temperature difference within 63.03 and 5.01 °C, respectively, which will control the temperature of the battery module and balance the temperature distribution uniformly. This indicates that the designed CPCM can provide an effective approach to explore high thermal conductivity composite materials for thermal management and other energy storage fields.

Automation testing using AI is replacing the conventional testing procedures by optimizing efficiency, accuracy and defect detection. Traditional automation testing is based on the scripts but the prediction of analytics and machine learning is used in AI powered framework to optimize the test execution. The focus of this research is the effects that AI driven automation has for defect leakage reduction, test maintenance cost reduction, and adaptability. The leakage rate defect reduction is found to be 66%, while the cost reduction is 60%. Integration of AI into testing frameworks help organizations speed up the releases, provide higher test coverage and enhanced software reliability. The study also highlights the importance of AI in the future of the software testing.

Abstract B-type sleeve in-service welding repair technology is a permanent repair method for pipeline and leakage defects, and automatic welding technology has the characteristics of high efficiency, stable quality, and low requirements for operators. How to effectively use the advantages of both and improve the quality of in-service welding for pipeline maintenance and emergency repair has become an urgent problem that needs to be solved. The near-surface crack defect analysis of a type B sleeve fillet weld was started. The magnetic particle test confirmed a near-surface crack with a length of 10 mm, and the metallography test confirmed a depth of 2 mm. The cause was related to welding quality. Moreover, three unfused defects with a maximum length of 129 mm were found by phased array inspection. In addition, the welding quality of type B sleeve automatic welding was evaluated. It was found that the results of the guided bending and groove hammer-breaking test of the longitudinal weld were all wrong. The maximum hardness of the weld was higher than the requirements of general technical conditions.

Autonomous defect detection is critical for intelligent industrial production lines. Pipeline leakage as a typical defect presents an irregular or inconspicuous appearance, resulting in the failure of general visual inspection approaches in actual applications. To this end, we propose a contactless pipeline leak detection system with a novel semantic segmentation network, namely, WDA-Net, that can work in various illumination conditions. A novel convolutional neural network (CNN)–Transformer hybrid architecture in the encoder is designed to capture defect features, especially for features with weak appearance, while an efficient fusion strategy that integrates adjacent feature maps and gradually recovers the spatial details by skip connections is proposed in the decoder. Except for the categories and intrinsic characteristics of defects, complex illumination conditions have been considered in the constructed dataset. Comprehensive experiments demonstrate that the proposed method outperforms other methods, achieving 70.0% mean intersection over union (mIoU) on the pipeline leakage defect (PLD) dataset and 84.6% mIoU on the NEU-Seg dataset. The code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/zbl929/WDA-Net</uri>.

Computer vision technology provides an intelligent means for detecting tunnel water leakage areas. However, the accuracy of defect feature extraction and segmentation is limited by factors such as insufficient lighting and environmental interference inside tunnels. To address the problem, this paper proposes a tunnel water leakage area segmentation network model called Customized Side Guided-Unet (CSG-Unet), using Unet as the baseline model. The main contributions are: (1) To improve the accuracy of water leakage area extraction, a customized side guided term is introduced to direct the net’s attention to the changes in light and shade within the image. A parallel attention network module is designed to extract internal information from the guided term. Subsequently, a strengthened channel attention module aggregates the guided term and the original information to achieve accurate segmentation of water leakage areas; (2) To address the scarcity of tunnel water leakage area datasets, a basic dataset is constructed by collecting data from open-source datasets and manually gathered data in tunnels. On this basis, perspective transformation is used to change the camera viewpoint, gaussian noise is randomly added to the images in the dataset to simulate images taken in dimly lit scenes, thereby expanding the dataset and enhancing the network’s generalization. The CSG-Unet network was trained using the constructed training set, achieving a mean Intersection over Union (mi IoU) of 85.54%, a mean Dice coefficient (mi Dice) of 85.26%, and a mean Pixel Accuracy (mi PA) of 90.85%. Compared to its baseline network, U-Net (tiny), these metrics show an improvement of over 3.2% in each indicator. Finally, a visual comparison between the improved network and the baseline network further confirms that the proposed model can effectively adapt to the segmentation of water leakage areas in complex environments.

Abstract To address the challenges of difficult detection of minute magnetic flux leakage (MFL) defects, insufficient inspection data, and low detection accuracy, the denoising diffusion probabilistic model (DDPM) gate dilated parallel convolution swin transformer (DGPST) is proposed. First, we introduce a DDPM-based data generation model, successfully generating a large quantity of diverse and rich MFL defect samples. Second, a gated parallel convolution layer is introduced into the backbone network. This strategy uses the characteristics of dilated convolution to broaden the receptive field of the model, thus enhancing the integration ability of global information. The addition of gating mechanism enables the model to adjust the calculation of attention weight based on broader context information in advance, which not only complicates the shortcomings of window self-attention in global dependence understanding, but also effectively suppress irrelevant calculation. Finally, the loss function of H Intersection over Union is introduced to improve the mean average precision. Following these enhancements, DGPST attains a satisfactory outcome in detecting tiny defects within the MFL problem. Experimental data indicates the accuracy of the algorithm reaches 95.6% and the delay is reduced to 7.6 ms.

Water leakage defects often occur in underground structures, leading to accelerated structural aging and threatening structural safety. Leakage identification can detect early diseases of underground structures and provide important guidance for reinforcement and maintenance. Deep learning-based computer vision methods have been rapidly developed and widely used in many fields. However, establishing a deep learning model for underground structure leakage identification usually requires a lot of training data on leakage defects, which is very expensive. To overcome the data shortage, a deep neural network method for leakage identification is developed based on transfer learning in this paper. For comparison, four famous classification models, including VGG16, AlexNet, SqueezeNet, and ResNet18, are constructed. To train the classification models, a transfer learning strategy is developed, and a dataset of underground structure leakage is created. Finally, the classification performance on the leakage dataset of different deep learning models is comparatively studied under different sizes of training data. The results showed that the VGG16, AlexNet, and SqueezeNet models with transfer learning can overall provide higher and more stable classification performance on the leakage dataset than those without transfer learning. The ResNet18 model with transfer learning can overall provide a similar value of classification performance on the leakage dataset than that without transfer learning, but its classification performance is more stable than that without transfer learning. In addition, the SqueezeNet model obtains an overall higher and more stable performance than the comparative models on the leakage dataset for all classification metrics.