Network Model Research Articles

The Fault Diagnosis of Rolling Bearings Is Conducted by Employing a Dual-Branch Convolutional Capsule Neural Network

Currently, many fault diagnosis methods for rolling bearings based on deep learning are facing two main challenges. Firstly, the deep learning model exhibits poor diagnostic performance and limited generalization ability in the presence of noise signals and varying loads. Secondly, there is incomplete utilization of fault information and inadequate extraction of fault features, leading to the low diagnostic accuracy of the model. To address these problems, this paper proposes an improved dual-branch convolutional capsule neural network for rolling bearing fault diagnosis. This method converts the collected bearing vibration signals into grayscale images to construct a grayscale image dataset. By fully considering the types of bearing faults and damage diameters, the data are labeled using a dual-label format. A multi-scale convolution module is introduced to extract features from the data and maximize feature information extraction. Additionally, a coordinate attention mechanism is incorporated into this module to better extract useful channel features and enhance feature extraction capability. Based on adaptive fusion between fault type (damage diameter) features and labels, a dual-branch convolutional capsule neural network model for rolling bearing fault diagnosis is established. The model was experimentally validated using both Case Western Reserve University’s bearing dataset and self-made datasets. The experimental results demonstrate that the fault type branch of the model achieves an accuracy rate of 99.88%, while the damage diameter branch attains an accuracy rate of 99.72%. Both branches exhibit excellent classification performance and display robustness against noise interference and variable working conditions. In comparison with other algorithm models cited in the reference literature, the diagnostic capability of the model proposed in this study surpasses them. Furthermore, the generalization ability of the model is validated using a self-constructed laboratory dataset, yielding an average accuracy rate of 94.25% for both branches.

Leakage detection of an acoustic emission pipeline based on an improved transformer network

Pipeline leakage detection is an integral part of pipeline integrity management. Combining AE (Acoustic Emission) with deep learning is currently the most commonly used method for pipeline leakage detection. However, this approach is usually applicable only to specific situations and requires powerful signal analysis and computational capabilities. To address these issues, this paper proposes an improved Transformer network model for diagnosing faults associated with abnormal working conditions in acoustic emission pipelines. First, the method utilizes the temporal properties of the GRU and the positional coding of the Transformer to capture and feature extract the data point sequence position information to suppress redundant information, and introduces the largest pooling layer into the Transformer model to alleviate the overfitting phenomenon. Second, while retaining the original attention learning mechanism and identity path in the original DRSN, a new soft threshold function is introduced to replace the ReLU activation function with a new threshold function, and a new soft threshold module and adaptive slope module are designed to construct the improved residual shrinkage unit (ASB-STRSBU), which is used to adaptively set the optimal threshold. Finally, pipeline leakage is classified. The experimental results show that the NDRSN model is able to make full use of global and local information when considering leakage signals and can automatically learn and acquire the important parameters of the input features in the spatial and channel domains. By optimizing the GRU improved Transformer network recognition model, the method significantly reduces the model training time and computational resource consumption while maintaining high leakage recognition accuracy. The average accuracy reached 93.97%. This indicates that the method has good robustness in acoustic emission pipeline leakage detection.

Multi-focused image fusion algorithm based on multi-scale hybrid attention residual network.

In order to improve the detection performance of image fusion in focus areas and realize end-to-end decision diagram optimization, we design a multi-focus image fusion network based on deep learning. The network is trained using unsupervised learning and a multi-scale hybrid attention residual network model is introduced to enable solving for features at different levels of the image. In the training stage, multi-scale features are extracted from two source images with different focal points using hybrid multi-scale residual blocks (MSRB), and the up-down projection module (UDP) is introduced to obtain multi-scale edge information, then the extracted features are operated to obtain deeper image features. These blocks can effectively utilize multi-scale feature information without increasing the number of parameters. The deep features of the image are extracted in its test phase, input to the spatial frequency domain to calculate and measure the activity level and obtain the initial decision map, and use post-processing techniques to eliminate the edge errors. Finally, the decision map is generated and optimized, and the final fused image is obtained by combining the optimized decision map with the source image. The comparative experiments show that our proposed model achieves better fusion performance in subjective evaluation, and the quality of the obtained fused images is more robust with richer details. The objective evaluation metrics work better and the image fusion quality is higher.

Error Analysis and Correction of FENGYUN-4A GIIRS Temperature Profile Products in Summer over the Qinghai–Tibet Plateau

To understand the applicability of the temperature profile product of the FENGYUN-4A (FY-4A) geostationary interferometric infrared detector (GIIRS) in summer over the Qinghai–Tibet Plateau and to improve product quality, the error of GIIRS temperature products was analyzed based on radiosonde data. A long short-term memory network model (LSTM) was used to correct the GIIRS temperature profile product in summer over the plateau at a high altitude (above 500 hPa), and further evaluation of the corrected product was conducted. The results show that summertime GIIRS temperature retrievals over the Qinghai–Tibet Plateau had a positive bias above 150 hPa and a negative bias below 150 hPa, resulting in an overall negative bias. The root mean square error was between 2 and 2.9 K, and the root mean square error was relatively large at 100 hPa and above. The LSTM established in this study could effectively correct the GIIRS temperature over the plateau. The correlation and root mean square error of the corrected GIIRS temperature and the radiosonde observation temperature were significantly improved. Using a trained LSTM correction model to correct the hourly GIIRS temperature can improve the accuracy and usability of the product. After correction, the average bias of the GIIRS temperature compared to the ERA5 temperature product was reduced from −0.26 K to 0.06 K, and the root mean square error was reduced from 2.25 K to 1.25 K. The correction model can be applied to different seasons, and it can also correct the GIIRS temperature in larger areas based on other high-precision and high-resolution data, achieving good results, thus indicating that the correction model has universal applicability.

Characterization of clinical data for patient stratification in moderate osteoarthritis with support vector machines, regulatory network models, and verification against osteoarthritis Initiative data

Knee osteoarthritis (OA) diagnosis is based on symptoms, assessed through questionnaires such as the WOMAC. However, the inconsistency of pain recording and the discrepancy between joint phenotype and symptoms highlight the need for objective biomarkers in knee OA diagnosis. To this end, we study relationships among clinical and molecular data in a cohort of women (n = 51) with Kellgren–Lawrence grade 2–3 knee OA through a Support Vector Machine (SVM) and a regulation network model. Clinical descriptors (i.e., pain catastrophism, depression, functionality, joint pain, rigidity, sensitization and synovitis) are used to classify patients. A Youden’s test is performed for each classifier to determine optimal binarization thresholds for the descriptors. Thresholds are tested against patient stratification according to baseline WOMAC data from the Osteoarthritis Initiative, and the mean accuracy is 0.97. For our cohort, the data used as SVM inputs are knee OA descriptors, synovial fluid proteomic measurements (n = 25), and transcription factor activation obtained from regulatory network model stimulated with the synovial fluid measurements. The relative weights after classification reflect input importance. The performance of each classifier is evaluated through ROC-AUC analysis. The best classifier with clinical data is pain catastrophism (AUC = 0.9), highly influenced by funcionality and pain sensetization, suggesting that kinesophobia is involved in pain perception. With synovial fluid proteins used as input, leptin strongly influences every classifier, suggesting the importance of low-grade inflammation. When transcription factors are used, the mean AUC is limited to 0.608, which can be related to the pleomorphic behaviour of osteoarthritic chondrocytes. Nevertheless, funcionality has an AUC of 0.7 with a decisive importance of FOXO downregulation. Though larger and longitudinal cohorts are needed, this unique combination of SVM and regulatory network model shall help to stratify knee OA patients more objectively.

MNMST: topology of cell networks leverages identification of spatial domains from spatial transcriptomics data

Advances in spatial transcriptomics provide an unprecedented opportunity to reveal the structure and function of biology systems. However, current algorithms fail to address the heterogeneity and interpretability of spatial transcriptomics data. Here, we present a multi-layer network model for identifying spatial domains in spatial transcriptomics data with joint learning. We demonstrate that spatial domains can be precisely characterized and discriminated by the topological structure of cell networks, facilitating identification and interpretability of spatial domains, which outperforms state-of-the-art baselines. Furthermore, we prove that network model offers an effective and efficient strategy for integrative analysis of spatial transcriptomics data from various platforms.

GMAlignNet: multi-scale lightweight brain tumor image segmentation with enhanced semantic information consistency

Although the U-shaped architecture, represented by UNet, has become a major network model for brain tumor segmentation, the repeated convolution and sampling operations can easily lead to the loss of crucial information. Additionally, directly fusing features from different levels without distinction can easily result in feature misalignment, affecting segmentation accuracy. On the other hand, traditional convolutional blocks used for feature extraction cannot capture the abundant multi-scale information present in brain tumor images. This paper proposes a multi-scale feature-aligned segmentation model called GMAlignNet that fully utilizes Ghost convolution to solve these problems. Ghost hierarchical decoupled fusion unit and Ghost hierarchical decoupled unit are used instead of standard convolutions in the encoding and decoding paths. This transformation replaces the holistic learning of volume structures by traditional convolutional blocks with multi-level learning on a specific view, facilitating the acquisition of abundant multi-scale contextual information through low-cost operations. Furthermore, a feature alignment unit is proposed that can utilize semantic information flow to guide the recovery of upsampled features. It performs pixel-level semantic information correction on misaligned features due to feature fusion. The proposed method is also employed to optimize three classic networks, namely DMFNet, HDCNet, and 3D UNet, demonstrating its effectiveness in automatic brain tumor segmentation. The proposed network model was applied to the BraTS 2018 dataset, and the results indicate that the proposed GMAlignNet achieved Dice coefficients of 81.65%, 90.07%, and 85.16% for enhancing tumor, whole tumor, and tumor core segmentation, respectively. Moreover, with only 0.29 M parameters and 26.88G FLOPs, it demonstrates better potential in terms of computational efficiency and possesses the advantages of lightweight. Extensive experiments on the BraTS 2018, BraTS 2019, and BraTS 2020 datasets suggest that the proposed model exhibits better potential in handling edge details and contour recognition.

Promoting Sustainable Coal Gas Development: Microscopic Seepage Mechanism of Natural Fractured Coal Based on 3D-CT Reconstruction

Green mining is an effective way to achieve sustainable development in the coal industry. Preventing coal and gas outburst dynamic disasters are essential for ensuring sustainable and safe mining. The numerous microscopic pores within the coal serve as the primary storage space for gas, making it critical to explore the structural distribution and seepage characteristics to reveal the disaster mechanism. Under mining stress, gas within the micropores of the coal migrates outward through cracks, with these cracks exerting a significant control effect on gas migration. Therefore, this study focuses on utilizing natural fractured coal bodies as research objects, employing a micro-CT imaging system to conduct scanning tests and digital core technology to reconstruct sample pore and fracture structures in three dimensions, and characterizing the pores, cracks, skeleton structure, and connectivity. A representative elementary volume (REV) containing macro cracks was selected to establish an equivalent model of the pore network, and a seepage simulation analysis was performed using the visualization software. Revealing the seepage characteristics of fractured coal mass from a microscopic perspective. The research results can provide guidance for gas drainage and dynamic disaster early warning in deep coal mines, thus facilitating the sustainable development of coal mining enterprises.

Data-driven traffic sensor location and path flow estimation using Wasserstein metric

This paper introduces link information value obtained by the traffic sensors and presents a traffic sensor location and flow estimation joint optimization model in an urban road network. In contrast to most previous studies, this paper adds new traffic sensors into the existing sensor network and proposes a data-driven path flow measurement method based on Wasserstein metric, which is utilized to measure the distance between the estimated traffic flow distribution and the actual distribution. Furthermore, this paper develops a customized greedy algorithm by combining a search strategy for the link information value to obtain the optimal sensor location scheme and perform traffic flow estimation under different budget conditions. Numerical experiments are conducted on Sioux-Falls test network and Eastern Massachusetts interstate highway subnetwork to verify the accuracy and effectiveness of the proposed model based on Wasserstein metric and the developed solution method. Computational results show that the sensor location scheme generated by the model based on Wasserstein metric can reduce the estimation error of the traffic flow compared with KL divergence model under the same deployment cost. Additionally, the customized greedy algorithm can achieve the better performance than the Brute force algorithm in terms of computing time and solution quality.

Synergizing Transfer Learning and Multi-Agent Systems for Thermal Parametrization in Induction Traction Motors

Maintaining optimal temperatures in the critical parts of an induction traction motor is crucial for railway propulsion systems. A reduced-order lumped-parameter thermal network (LPTN) model enables computably inexpensive, accurate temperature estimation; however, it requires empirically based parameter estimation exercises. The calibration process is typically performed in labs in a controlled experimental setting, which is associated with a lot of supervised human efforts. However, the exploration of machine learning (ML) techniques in varied domains has enabled the model parameterization in the drive system outside the laboratory settings. This paper presents an innovative use of a multi-agent reinforcement learning (MARL) approach for the parametrization of an LPTN model. First, a set of reinforcement learning agents are trained to estimate the optimized thermal parameters using the simulated data in several driving cycles (DCs). The selection of a reinforcement learning agent and the level of neurons in the RL model is made based on variability of the driving cycle data. Furthermore, transfer learning is performed on a new driving cycle data collected on the measurement setup. Statistical analysis and clustering techniques are proposed for the selection of an RL agent that has been pre-trained on the historical data. It is established that by synergizing within reinforcement learning techniques, it is possible to refine and adjust the RL learning models to effectively capture the complexities of thermal dynamics. The proposed MARL framework shows its capability to accurately reflect the motor’s thermal behavior under various driving conditions. The transfer learning usage in the proposed approach could yield significant improvement in the accuracy of temperature prediction in the new driving cycles data. This approach is proposed with the aim of developing more adaptive and efficient thermal management strategies for railway propulsion systems.

Mechanical performance degradation investigation on FRP reinforced concrete based on neural network design method

To predict the effect of chemical-freezing coupling erosion on the properties of four kinds of FRP-reinforced concrete. Rapid freeze–thaw tests were conducted. The mass loss rate, relative dynamic elastic modulus, compressive strength, and flexural capacity were tested to investigate the Mechanical Performance of specimens. The compression specimens are cylindrical specimens wrapped with FRP, and the flexural specimens are pasted with FRP prismatic specimens on the pre-cracked side. A database was built based on 45 groups of experimental test results, and the prediction effect of the BP neural network and CNN model on compressive strength and flexural capacity was compared, respectively. The results showed that CNN did a better job. Finally, the maximum number of freeze–thaw cycles of different FRP-reinforced specimens was predicted based on the CNN model with mass loss rate and relative dynamic modulus as the evaluation criteria. This method can provide a new perspective for predicting the durability of FRP-reinforced concrete.

Enhancing oral squamous cell carcinoma detection: a novel approach using improved EfficientNet architecture

ProblemOral squamous cell carcinoma (OSCC) is the eighth most prevalent cancer globally, leading to the loss of structural integrity within the oral cavity layers and membranes. Despite its high prevalence, early diagnosis is crucial for effective treatment.AimThis study aimed to utilize recent advancements in deep learning for medical image classification to automate the early diagnosis of oral histopathology images, thereby facilitating prompt and accurate detection of oral cancer.MethodsA deep learning convolutional neural network (CNN) model categorizes benign and malignant oral biopsy histopathological images. By leveraging 17 pretrained DL-CNN models, a two-step statistical analysis identified the pretrained EfficientNetB0 model as the most superior. Further enhancement of EfficientNetB0 was achieved by incorporating a dual attention network (DAN) into the model architecture.ResultsThe improved EfficientNetB0 model demonstrated impressive performance metrics, including an accuracy of 91.1%, sensitivity of 92.2%, specificity of 91.0%, precision of 91.3%, false-positive rate (FPR) of 1.12%, F1 score of 92.3%, Matthews correlation coefficient (MCC) of 90.1%, kappa of 88.8%, and computational time of 66.41%. Notably, this model surpasses the performance of state-of-the-art approaches in the field.ConclusionIntegrating deep learning techniques, specifically the enhanced EfficientNetB0 model with DAN, shows promising results for the automated early diagnosis of oral cancer through oral histopathology image analysis. This advancement has significant potential for improving the efficacy of oral cancer treatment strategies.

Maize leaf disease recognition based on improved MSRCR and OSCRNet

Accurate identification of maize pests and diseases plays a crucial role in the growth and yield of maize. The recognition method of maize leaf diseases based on deep learning is currently a research hotspot. In practical applications, it is difficult to ensure a high recognition accuracy due to the impact of environmental factors (such as light intensity, background noise) and insufficient training data. To address this challenge, this paper proposes a maize image recognition method based on image enhancement and OSCRNet. Firstly, a maize leaf image enhancement framework and algorithm based on improved MSRCR were designed to enhance the characteristics of the maize leaf in a complex environment and to solve the problems of high noise and blur of maize images. Subsequently, the OSCRNet maize leaf recognition network model was established, which uses octave convolution with characteristics to accelerate network training and reduce unnecessary redundant spatial information in maize leaf images. Additionally, a self-calibrated convolution with multi-scale features was employed to realize the interactions of different feature information in the maize leaf images, enhance feature extraction, and solve the problems of similarity of maize disease features and easy learning disorders. Concurrently, batch normalization was employed to prevent network overfitting and enhance the robustness of the model. The experimental results showed that the highest identification accuracy of the proposed method for rust, gray leaf disease, northern wilt disease, and healthy maize are 96.67%, 94.86%, 92.81%, and 95.63%, respectively. Our methods were beneficial in solving the problems of slow efficiency, low accuracy and insufficient training data, and also outperformed other comparison models. The proposed method demonstrates strong robustness for maize disease images collected in the natural environment, providing a reference for the intelligent diagnosis of other plant leaf diseases.

Construction of a Multicellular Communication Network Model for Cell Co-Culture Technology and Evaluation of Its Simulation Capability

Abstract Objective: Cell co-culture technology has been widely used to analyze the effects of drugs on cell proliferation and the expression of some proteins in cells, especially in the field of traditional Chinese medicine (TCM); however, the interactions between cells and the transmission of TCM effects between cells have not been adequately studied. Materials and Methods: Using data on gene transcription regulation, biological response, signal channel, and cell-specific expression protein, we built a network for cell types based on entity grammar. Through the correspondence and location information of signal molecules and receptors, type-specific networks of single cells were connected and a multicellular network of smooth muscle cells, neurons, and vascular endothelial cells was constructed. The mechanism of action of nimodipine was analyzed based on the multicellular communication network and its simulation capability was evaluated. Results: The outputs generated by the model developed in this study showed that nimodipine inhibited smooth muscle contraction, due to the overload of Ca2+ and the toxicity of excitatory amino acids, and protected neurons and vascular endothelial cells by supporting cell proliferation and inhibiting cell apoptosis. These results were consistent with the known mechanism of nimodipine action, thus confirming that the multicellular network can be used to study the transmission of drug effects among cells. Conclusions: This study lays a foundation for the analysis of the transmission of drug effects in multi-cells, tissues, organs, and other spatial scales through multicellular co-culture experiments, based on a multicellular communication network. In addition, it provides a biological network model for the analysis of TCM action mechanisms.

MinActionPath2: path generation between different conformations of large macromolecular assemblies by action minimization.

Recent progress in solving macromolecular structures and assemblies by cryogenic electron microscopy techniques enables sampling of their conformations in different states that are relevant to their biological function. Knowing the transition path between these conformations would provide new avenues for drug discovery. While the experimental study of transition paths is intrinsically difficult, in-silico methods can be used to generate an initial guess for those paths. The Elastic Network Model (ENM), along with a coarse-grained representation (CG) of the structures are among the most popular models to explore such possible paths. Here we propose an update to our software platform MinActionPath that generates non-linear transition paths based on ENM and CG models, using action minimization to solve the equations of motion. The new website enables the study of large structures such as ribosomes or entire virus envelopes. It provides direct visualization of the trajectories along with quantitative analyses of their behaviors at http://dynstr.pasteur.fr/servers/minactionpath/ minactionpath2_submission.

Artificial intelligence based smoke flow mechanism analysis and prediction patterns of fire for large space building

The deposition and dynamics of smoke layers in large-space building fires are governed by a complex interplay of factors, making the prediction of such dynamics using traditional mathematical models challenging. In response, this study introduces a Back Propagation (BP) neural network model, devised from field simulation data, to efficiently forecast the temporal progression of smoke layers. It was observed that the model exhibits minimal training errors and high processing speeds, thereby fulfilling the stringent accuracy demands of fire engineering. Specifically, the model achieved a minimum relative error of 0.0005 and a maximum of 0.0845 across various prediction points, underscoring its reliability and precision. The ability of this BP neural network model to predict smoke layer changes significantly enhances the design optimization of smoke control systems swiftly and accurately in large buildings and supports rapid, informed decision-making during fire emergencies. Moreover, the model facilitates the development of engineering calculation models tailored for the quick prediction of fire smoke dynamics, which are essential for both theoretical research and practical applications. This approach not only conserves experimental resources but also advances the implementation of scientific, effective rescue operations in the event of large space building fires.

Peri-device leakage and delayed endothelialization of the Watchman device: a computed tomography study.

This study aimed to explore the endothelialization process and assess the potential association between endothelialization and peri-device leak (PDL) following Watchman implantation via a quantitative method. This is a single-center retrospective study of consecutive patients undergoing LAAO between December 2015 and November 2021. Device endothelialization, compared between PDL and non-PDL group, were quantitatively analyzed based on hypoattenuated thickening in cardiac computed tomography angiography (CCTA). Advancement in endothelialization over time were explored using the Cochran-Armitage test and generalized estimating equation approach. Potential risk factors of delayed endothelialization were analyzed using the Cox proportional-hazards model. A total of 172 patients (mean age, 68 years ± 10 [standard deviation], 114 men) were finally included. The average endothelialization ratio of the study population was 89.8 ± 7.2 percent. In the follow-up period of postprocedural 3 months to more than 12 months, an incremental trend of endothelialization over time was observed with the ratio of 85.8 ± 8.0, 89.6 ± 7.6, 92.2 ± 4.5, 94.3 ± 2.9 percent, respectively (p < 0.0001). Notably, patients without PDL exhibited a swifter advancement in endothelialization compared to those with PDL, irrespective of device size. The multivariable Cox regression model showed that PDL (HR = 2.113, 95%CI: 1.300-3.435, p = 0.003), DSP (HR = 1.717, 95%CI: 1.113-2.647, p = 0.014) were independent risk factors of delayed endothelialization. CCTA holds promise as an effective means of quantitatively assessing device endothelialization. Endothelialization advanced gradually over time, with PDL potentially impeding device endothelialization. A comprehensive understanding of the correlation between endothelialization ratio, time, and residual shunt can establish a more dependable foundation for determining the appropriate anticoagulation treatment following left atrial appendage closure. Current recommendations for postleft atrial appendage occlusion anti-platelet and anticoagulation therapy are based on animal studies. Cardiac computed tomography angiography (CCTA) combined with the UNet neural network model enables the quantitative assessment of device endothelialization. This technique will allow for additional studies to better understand device endothelialization to optimize treatments in this population.

High-temperature tensile behavior and constitutive model in Co-free Fe40Mn20Cr20Ni20 high-entropy alloy

High-entropy alloys (HEAs) have attracted widespread attention from scholars as a new type of material employed in extreme environments. However, as a main kind of HEAs, many face-centered cubic single-phase HEAs are restricted in industrial applications due to their lower yield strength and high cost when containing expensive elements such as Co. In this study, dispersion strengthening by heat treatment was introduced in low-cost Co-free Fe40Mn20Cr20Ni20 HEA to improve its strength, and its high-temperature tensile behavior and constitutive model were studied to explore its potential application at high temperatures. It is found that when subjected to quasi-static room-temperature stretching, the heat-treated sample exhibits a yield strength of 534 MPa and a tensile plasticity of 26.8%. In addition, the tensile behavior of samples after heat treatment was investigated at high temperature (573–873 K) and low strain rate (10−3–10−1 s−1). The results suggest that the yield strength decreases with increasing temperature and decreasing strain rate. Moreover, at 873 K and 10−3 s−1, the electron backscatter diffraction system and x-ray diffraction results of the deformed sample indicate that the softening curve is driven from the recovery of materials. Finally, the flow stress was predicted using the Arrhenius equation and Artificial Neural Network model (ANN), and the two models were assessed using the average absolute relative error and coefficient of correlation (R). The results showed that the ANN had higher accuracy.

Intelligent Stroke Disease Prediction Model Using Deep Learning Approaches

Stroke is a high morbidity and mortality disease that poses a serious threat to people’s health. Early recognition of the various warning signs of stroke is necessary so that timely clinical intervention can help reduce the severity of stroke. Deep neural networks have powerful feature representation capabilities and can automatically learn discriminant features from large amounts of data. This paper uses a range of physiological characteristic parameters and collaborates with deep neural networks, such as the Wasserstein generative adversarial networks with gradient penalty and regression network, to construct a stroke prediction model. Firstly, to address the problem of imbalance between positive and negative samples in the stroke public data set, we performed positive sample data augmentation and utilized WGAN-GP to generate stroke data with high fidelity and used it for the training of the prediction network model. Then, the relationship between observable physiological characteristic parameters and the predicted risk of suffering a stroke was modeled as a nonlinear mapping transformation, and a stroke prediction model based on a deep regression network was designed. Finally, the proposed method is compared with commonly used machine learning-based classification algorithms such as decision tree, random forest, support vector machine, and artificial neural networks. The prediction results of the proposed method are optimal in the comprehensive measurement index F. Further ablation experiments also show that the designed prediction model has certain robustness and can effectively predict stroke diseases.

Hot rolled steel surface defect detection and classification using an automatic ensemble approach

This study introduces an ensemble-based Deep Neural Network (DNN) model for detecting defects on steel surfaces. The method suggested in this study classifies steel surface conditions into six possible fault categories, namely, crazing, inclusion, rolled in, pitted surface, scratches, and patches. The images undergo preprocessing and extraction of features in spatial and frequency domains using image segmentation techniques such as grey level difference method (GLDM), fast Fourier Transform (FFT), grey level co-occurrence matrix (GLCM), texture analysis and discrete wavelet transform (DWT). The ensembling of image features into a fused feature pool is carried out after the preprocessing of input images that are provided as input to a light-weight neural network model for training and testing. The performance of the model is comprehensively evaluated via an ablation study both before and after ensembling. In addition, the model capability is effectively analyzed using receiver operating characteristics (ROC) curve, confusion matrix from which classification accuracy of the model could be obtained and other parameters including precision and f1-score. It was observed that the proposed deep learning network presents phenomenally high accuracy of 99.72% for detection and classification of steel surface faults. This result was found to be superior when compared with the performance of the same neural network over each feature type individually. This study also compares the classification results of the model built based on the ensembled feature set with the results of various other classification approaches available in literature. The ensemble-based model could potentially be integrated into existing inspection systems for real-time, efficient and robust condition monitoring of steel surfaces.