Neural Network Model Research Articles

Bio-inspired algorithms for feature engineering: analysis, applications and future research directions

Purpose Nature’s evolution has shaped intelligent behaviors in creatures like insects and birds, inspiring the field of Swarm Intelligence. Researchers have developed bio-inspired algorithms to address complex optimization problems efficiently. These algorithms strike a balance between computational efficiency and solution optimality, attracting significant attention across domains. Design/methodology/approach Bio-inspired optimization techniques for feature engineering and its applications are systematically reviewed with chief objective of assessing statistical influence and significance of “Bio-inspired optimization”-based computational models by referring to vast research literature published between year 2015 and 2022. Findings The Scopus and Web of Science databases were explored for review with focus on parameters such as country-wise publications, keyword occurrences and citations per year. Springer and IEEE emerge as the most creative publishers, with indicative prominent and superior journals, namely, PLoS ONE, Neural Computing and Applications, Lecture Notes in Computer Science and IEEE Transactions. The “National Natural Science Foundation” of China and the “Ministry of Electronics and Information Technology” of India lead in funding projects in this area. China, India and Germany stand out as leaders in publications related to bio-inspired algorithms for feature engineering research. Originality/value The review findings integrate various bio-inspired algorithm selection techniques over a diverse spectrum of optimization techniques. Anti colony optimization contributes to decentralized and cooperative search strategies, bee colony optimization (BCO) improves collaborative decision-making, particle swarm optimization leads to exploration-exploitation balance and bio-inspired algorithms offer a range of nature-inspired heuristics.

Fast Detection and Classification of Microplastics below 10 μm Using CNN with Raman Spectroscopy.

In light of the growing awareness regarding the ubiquitous presence of microplastics (MPs) in our environment, recent efforts have been made to integrate Artificial Intelligence (AI) technology into MP detection. Among spectroscopic techniques, Raman spectroscopy is preferred for the detection of MP particles measuring less than 10 μm, as it overcomes the diffraction limitations encountered in Fourier transform infrared (FTIR). However, Raman spectroscopy's inherent limitation is its low scattering cross section, which often results in prolonged data collection times during practical sample measurements. In this study, we implemented a convolutional neural network (CNN) model alongside a tailored data interpolation strategy to expedite data collection for MP particles within the 1-10 μm range. Remarkably, we achieved the classification of plastic types for individual particles with a mere 0.4 s of exposure time, reaching an approximate confidence level of 85.47(±5.00)%. We postulate that the result significantly accelerates the aggregation of microplastic distribution data in diverse scenarios, contributing to the development of a comprehensive global microplastic map.

Convective drying of golden delicious apple enhancement: drying characteristics, artificial neural network modeling, chemical and ATR-FTIR analysis of quality parameters

AbstractIn recent years, many innovative methods have been investigated to provide alternative approaches to the food drying industry, but currently the most widely used method is convective drying. There are difficulties in integrating innovative methods into the food industry due to cost, inapplicability to every food material, or product quality, etc. In addition, it is possible to improve the convective drying method by pre-treating of foods. Thanks to the convective drying method with increased efficiency, shorter drying processes can be achieved. This study investigates the effects of ethanol and citric acid pretreatments on the convective drying process of apple slices and the drying rate, diameter and thickness shrinkage, color properties, total phenolic content (TPC), antioxidant activity (AA), ATR-FTIR spectra, and principal component analysis (PCA) of the dried samples. The results indicate that both ethanol and citric acid pretreatments significantly enhance the drying rate and decrease drying time, with the most favorable outcomes observed for apple slices immersed in an ethanol solution for 20 min. The study employs thin-layer and artificial neural network (ANN) modeling, revealing that ANN modeling outperforms thin-layer models in predicting moisture ratio. Shrinkage ratios in diameter and thickness were observed, but no significant statistical differences are found among the sample groups. The color properties of dried apple slices are influenced by pretreatments. L* values decreased in the ethanol-pretreated samples, whereas a* and b* values increased in all samples. On the other hand, drying process leads to a decrease in TPC and AA. Ethanol pretreatments caused higher losses; lower losses were observed in the citric acid–pretreated and untreated apples slices. ATR-FTIR analysis suggests distinct spectral changes in dried samples, particularly influenced by ethanol and citric acid pretreatments. The ATR-FTIR spectra highlighted shifts in water and carbohydrate levels, proteins, fibers, organic acids, and the occurrence of Maillard reactions throughout the drying process. PCA reveals that samples dried with ethanol and citric acid share a similar plane, while fresh samples and those dried at 60 °C exhibit different arrangements.

Ultrasound-based deep learning radiomics model for differentiating benign, borderline, and malignant ovarian tumours: a multi-class classification exploratory study.

Accurate preoperative identification of ovarian tumour subtypes is imperative for patients as it enables physicians to custom-tailor precise and individualized management strategies. So, we have developed an ultrasound (US)-based multiclass prediction algorithm for differentiating between benign, borderline, and malignant ovarian tumours. We randomised data from 849 patients with ovarian tumours into training and testing sets in a ratio of 8:2. The regions of interest on the US images were segmented and handcrafted radiomics features were extracted and screened. We applied the one-versus-rest method in multiclass classification. We inputted the best features into machine learning (ML) models and constructed a radiomic signature (Rad_Sig). US images of the maximum trimmed ovarian tumour sections were inputted into a pre-trained convolutional neural network (CNN) model. After internal enhancement and complex algorithms, each sample's predicted probability, known as the deep transfer learning signature (DTL_Sig), was generated. Clinical baseline data were analysed. Statistically significant clinical parameters and US semantic features in the training set were used to construct clinical signatures (Clinic_Sig). The prediction results of Rad_Sig, DTL_Sig, and Clinic_Sig for each sample were fused as new feature sets, to build the combined model, namely, the deep learning radiomic signature (DLR_Sig). We used the receiver operating characteristic (ROC) curve and the area under the ROC curve (AUC) to estimate the performance of the multiclass classification model. The training set included 440 benign, 44 borderline, and 196 malignant ovarian tumours. The testing set included 109 benign, 11 borderline, and 49 malignant ovarian tumours. DLR_Sig three-class prediction model had the best overall and class-specific classification performance, with micro- and macro-average AUC of 0.90 and 0.84, respectively, on the testing set. Categories of identification AUC were 0.84, 0.85, and 0.83 for benign, borderline, and malignant ovarian tumours, respectively. In the confusion matrix, the classifier models of Clinic_Sig and Rad_Sig could not recognise borderline ovarian tumours. However, the proportions of borderline and malignant ovarian tumours identified by DLR_Sig were the highest at 54.55% and 63.27%, respectively. The three-class prediction model of US-based DLR_Sig can discriminate between benign, borderline, and malignant ovarian tumours. Therefore, it may guide clinicians in determining the differential management of patients with ovarian tumours.

Exploring the Influence of Dynamic Indicators in Urban Spaces on Residents’ Environmental Behavior: A Case Study in Shanghai Utilizing Mixed-Methods Approach and Artificial Neural Network (ANN) Modeling

The main aim of this article is to evaluate the impact of dynamic indicators associated with urban spaces on the environmental behavior of residents in Shanghai, China. With the city experiencing rapid urbanization and increasing environmental concerns, it is crucial to understand how the design and management of urban spaces can encourage pro-environmental attitudes and actions among the population. The study specifically focuses on dynamic indicators, namely vitality, accessibility, quality, and walkability, to develop a comprehensive understanding of the utilization and experience of urban spaces. The research outcomes will make valuable contributions towards Shanghai’s objectives of achieving environmental sustainability, while also providing insights that have broader relevance to sustainable urban development globally. As the built environment significantly influences energy consumption, resource utilization, pollution generation, and overall human wellbeing, attaining urban sustainability is of paramount importance. The findings of this study will aid in informing strategies and policies that promote sustainable practices, not only benefiting Shanghai but also serving as a valuable resource for urban development initiatives worldwide. Sustainable urban design principles, including compactness, density, mixed land use, greening, and walkability, have been associated with pro-environmental behaviors, including reduced reliance on automobiles, increased walking and cycling, and heightened environmental consciousness. Nevertheless, the relationship between the built environment and sustainability behaviors is intricate and is influenced by multiple factors. Consequently, further research is necessary to comprehend how specific spatial and temporal dynamics impact environmental behaviors within urban settings. In this study, an artificial neural network (ANN) was developed to estimate the quality and walkability of an area and environmental behaviors by considering the augmented vitality and accessibility factors. The ANN’s predictions demonstrate that higher levels of vitality and accessibility positively contribute to improved walkability and environmental behaviors. The accuracy of the ANN’s predictions was assessed using linear regression, which yielded acceptable error rates when compared with experimental results.

Laser processing characteristics of PCD tool and modeling analysis

Through in-depth analysis of the experimental details and forming patterns of the nanosecond laser ablation polycrystalline diamond (PCD) textured tool processing system, this study explores the microscopic morphology and dimensions of micro-pits texture on the surface of PCD tools influenced by defocus amount, laser power, and pulse frequency. Experimental results indicate that the micro-pit textures generated under different parameters exhibit diversity, including rounded structure, fragments, recast layers, and heat-affected zones. The diameter and depth of micro-pits are comprehensively affected by defocus amount, laser power, and pulse frequency, showing complex patterns. After a thorough analysis of the effects of each parameter on the texture morphology, an artificial neural network (ANN) model is introduced for the prediction of micro-pit dimensions. Through model training and optimization, accurate predictions of micro-pit diameter and depth are obtained. In comparison to traditional regression models, the ANN model demonstrates outstanding predictive performance, validating its applicability in complex machining processes. This study not only provides a profound understanding of the processing patterns of PCD textured tools but also offers an effective predictive model for the optimization and control of similar future machining processes.

Positioning error estimation models for horizontal-distributed PUU parallel mechanism

Abstract In this paper, two models were proposed to estimate the positioning error of a 3-translational prismatic-universal-universal parallel kinematic mechanism. The two models were a kinematic error model (KEM) and a backpropagation neural network (BPNN) model, respectively. The KEM was constructed by incorporating three translational joint errors into the ideal kinematic model to describe the errors that occur during machining or assembly. Additionally, a sensitivity analysis was presented for each error parameter. The BPNN model was constructed to establish the relationship between the position of the end effector, the posture of each link, and the positioning error of the end effector using a neural network approach. Moreover, a hybrid method was proposed to decrease the final estimated residual error. The average errors of the KEM and BPNN model were 35% and 15% of the original error, respectively. The hybrid model reduced the final average error to less than 10% of its original value.

Operation and Productivity Monitoring from Sound Signal of Legacy Pipe Bending Machine via Convolutional Neural Network (CNN)

AbstractThis study introduces a non-invasive approach to monitor operation and productivity of a legacy pipe bending machine in real-time based on a lightweight convolutional neural network (CNN) model and internal sound as input data. Various sensors were deployed to determine the optimal sensor type and placement, and labels for training and testing the CNN model were generated through the meticulous collection of sound data in conjunction with webcam videos. The CNN model, which was optimized through hyperparameter tuning via grid search and utilized feature extraction using Log-Mel spectrogram, demonstrated notable prediction accuracies in the test. However, when applied in a real-world manufacturing scenario, the model encountered a significant number of errors in predicting productivity. To navigate through this challenge and enhance the predictive accuracy of the system, a buffer algorithm using the inferences of CNN models was proposed. This algorithm employs a queuing method for continuous sound monitoring securing robust predictions, refines the interpretation of the CNN model inferences, and enhances prediction outcomes in actual implementation where accuracy of monitoring productivity information is crucial. The proposed lightweight CNN model alongside the buffer algorithm was successfully deployed on an edge computer, enabling real-time remote monitoring.

Spatial Downscaling of ESA CCI Soil Moisture Data Based on Deep Learning with an Attention Mechanism

Soil moisture (SM) is a critical variable affecting ecosystem carbon and water cycles and their feedback to climate change. In this study, we proposed a convolutional neural network (CNN) model embedded with a residual block and attention module, named SMNet, to spatially downscale the European Space Agency (ESA) Climate Change Initiative (CCI) SM product. In the SMNet model, a lightweight Convolutional Block Attention Module (CBAM) dual-attention mechanism was integrated to comprehensively extract the spatial and channel information from the high-resolution input remote sensing products, the reanalysis meteorological dataset, and the topographic data. The model was employed to downscale the ESA CCI SM from its original spatial resolution of 25 km to 1 km in California, USA, in the annual growing season (1 May to 30 September) from 2003 to 2021. The original ESA CCI SM data and in situ SM measurements (0–5 cm depth) from the International Soil Moisture Network were used to validate the model’s performance. The results show that compared with the original ESA CCI SM data, the downscaled SM data have comparable accuracy with a mean correlation (R) and root mean square error (RMSE) of 0.82 and 0.052 m3/m3, respectively. Moreover, the model generates reasonable spatiotemporal SM patterns with higher accuracy in the western region and relatively lower accuracy in the eastern Nevada mountainous area. In situ site validation results in the SCAN, the SNOTEL network, and the USCRN reveal that the R and RMSE are 0.62, 0.63, and 0.77, and 0.077 m3/m3, 0.093 m3/m3, and 0.078 m3/m3, respectively. The results are slightly lower than the validation results from the original ESA CCI SM data. Overall, the validation results suggest that the SMNet downscaling model proposed in this study has satisfactory performance in handling the task of soil moisture downscaling. The downscaled SM model not only preserves a high level of spatial consistency with the original ESA CCI SM model but also offers more intricate spatial variations in SM depending on the spatial resolution of model input data.

A novel memristor Hopfield neural network with homogeneous coexisting multi-scroll attractors

Abstract Compared to conventional single-scroll or double-scroll attractors, multi-scroll chaotic attractors possess wide potential for application due to their adjustability and complex topology. However, neural network models for generating multiple scrolls are often implemented using memristors with piecewise nonlinear functions. To further explore multi-scroll attractors with different working mechanisms,a unique memristor containing a group of hyperbolic tangent functions is designed and then applied in a three-dimensional Hopfield neural network (HNN). The proposed memristive Hopfield neural network (MHNN) has multi-scroll chaotic attractors, where the number and parity of the scrolls be changed by adjusting the control parameters of the memristor. The complex dynamical behaviors of MHNN are studied by utilizing diverse numerical modeling approaches like bifurcation diagrams, Lyapunov exponents and phase plot. In addition, the proposed MHNN also has a complicated offset boosting coexisting behavior. By selecting suitable parameters, multiple coexisting chaotic attractors could be obtained. Homogeneous coexisting multi-scroll attractors can be shifted in multiple directions including unidirectional, planar and spatial ones. Moreover, theoretically speaking, there could be an infinite number of coexisting attractors. Finally, experimental results are validated through numerical simulations and circuit experiments to confirm the feasibility of the proposed MHNN model.

Performance Degradation Trend Prediction of Proton Exchange Membrane Fuel Cell Based on GA-TCN

Abstract To improve the prediction accuracy of the performance degradation trend of proton exchange membrane fuel cell (PEMFC), this paper proposes a temporal convolutional network (TCN) model based on genetic algorithm (GA) optimization to predict the performance degradation trend of PEMFC. Firstly, variational mode decomposition (VMD) and wavelet threshold denoising (WTD) algorithms are used to denoise the original data. Then the hyperparameters of the TCN model are optimized by GA, and the GA-TCN model for predicting the performance degradation trend of PEMFC is constructed. Finally, this paper uses the PEMFC stack degradation experimental dataset disclosed in the IEEE PHM 2014 Data Challenge to verify, and compares the proposed model with the backpropagation neural networks model (BP), the long short-term memory model (LSTM) and the classical temporal convolutional network model. The results show that the proposed method has the highest performance degradation trend prediction accuracy. In particular, when the training dataset accounts for 30%, i.e. the training samples are small, the root mean square error (RMSE) and mean absolute error (MAE) of the GA-TCN model are 0.004726 and 0.003119, respectively, which are 14.48% and 20.05% lower than that of the classical TCN model. Consequently, this methodology can forecast the degradation trend of PEMFC with high accuracy.

Optimisation and Calibration of Bayesian Neural Network for Probabilistic Prediction of Biogas Performance in an Anaerobic Lagoon

This study aims to enhance diagnostic capabilities for optimising the performance of the anaerobic sewage treatment lagoon at Melbourne Water’s Western Treatment Plant (WTP) through a novel machine learning (ML)-based monitoring strategy. This strategy employs ML to make accurate probabilistic predictions of biogas performance by leveraging diverse real-life operational and inspection sensor and other measurement data for asset management, decision making, and structural health monitoring (SHM). The paper commences with data analysis and preprocessing of complex irregular datasets to facilitate efficient learning in an artificial neural network. Subsequently, a Bayesian mixture density neural network model incorporating an attention-based mechanism in bidirectional long short-term memory (BiLSTM) was developed. This probabilistic approach uses a distribution output layer based on the Gaussian mixture model and Monte Carlo (MC) dropout technique in estimating data and model uncertainties, respectively. Furthermore, systematic hyperparameter optimisation revealed that the optimised model achieved a negative log-likelihood (NLL) of 0.074, significantly outperforming other configurations. It achieved an accuracy approximately 9 times greater than the average model performance (NLL = 0.753) and 22 times greater than the worst performing model (NLL = 1.677). Key factors influencing the model’s accuracy, such as the input window size and the number of hidden units in the BiLSTM layer, were identified, while the number of neurons in the fully connected layer was found to have no significant impact on accuracy. Moreover, model calibration using the expected calibration error was performed to correct the model’s predictive uncertainty. The findings suggest that the inherent data significantly contribute to the overall uncertainty of the model, highlighting the need for more high-quality data to enhance learning. This study lays the groundwork for applying ML in transforming high-value assets into intelligent structures and has broader implications for ML in asset management, SHM applications, and renewable energy sectors.

Efficient customer behaviour prediction in Indian metropolitan cities for E‐commerce applications

AbstractThe primary objective of this manuscript is to find the service quality factors, which influence customer satisfaction, promotional efforts taken by the retailer, and the perception of customers towards the organized retail sector. This investigation uses both offline and online retail sales datasets. Particularly, in the case of the offline retail sale dataset, an energy‐efficient analysis is conducted to collect scalable, accurate, and real‐time data. The collected data is used in enhancing several aspects of e‐commerce operations, including customer experience optimization and inventory management, which results in better decision‐making with increased efficiency. After data collection, the outliers are eliminated by implementing the z‐score technique. The removal of outliers increases the variability of collected data which reduces statistical power. The statistically sufficient data are given to the Randomized Grasshopper Optimization Algorithm (RGOA) for optimal instance selection. Finally, the selected optimal instances are given to the deep neural network (DNN) model for future customer behaviour prediction. The efficacy of the RGOA‐DNN model is analysed by using evaluation measures like mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), mean percentage error (MPE) and mean absolute percentage error (MAPE). The numerical analysis states that the RGOA‐DNN model obtained a minimal error rate with MSE of 0.09 and 0.10, RMSE of 0.13 and 0.14, MAE of 0.12 and 0.16, MPE of 0.10 and 0.11 and MAPE of 0.08 and 0.10 on the offline and online retail sales datasets. The RGOA‐DNN model has a minimal error rate in future customer behaviour prediction related to the conventional regression models. Furthermore, the elimination of inactive instances or selection of optimal instances reduces the model complexity to linear and computational time to 22.10 and 23.12 s on the offline and online retail sales datasets.

Sensitivity Analysis of Influencing Factors and Two-Stage Prediction of Frost Resistance of Active-Admixture Recycled Concrete Based on Grey Theory–BPNN

Sensitivity analysis of influencing factors on frost resistance is carried out in this paper, and a two-stage neural network model based on grey theory and Back Propagation Neural Networks (BPNNs) is established for the sake of predicting the frost resistance of active-admixture recycled concrete quickly and accurately. Firstly, the influence degree of cement, water, sand, natural aggregate, recycled aggregate, mineral powder, fly ash, fiber and air-entraining agent on the frost resistance of active-admixture recycled-aggregate concrete was analyzed based on the grey system theory, and the primary and secondary relationships of various factors were effectively distinguished. Then, the input layer of the model was determined as cement, water, sand, recycled aggregate and air-entraining agent, and the output layer was the relative dynamic elastic modulus. A total of 120 datasets were collected from the experimental data of another author, and the relative dynamic elastic modulus was predicted using the two-stage BPNN prediction model proposed in this paper and compared with the BPNN prediction results. The results show that the proposed two-stage BPNN model, after removing less-sensitive parameters from the input layer, has better prediction accuracy and shorter run time than the BPNN model.

Practical application of quantum neural network to materials informatics

Quantum neural network (QNN) models have received increasing attention owing to their strong expressibility and resistance to overfitting. It is particularly useful when the size of the training data is small, making it a good fit for materials informatics (MI) problems. However, there are only a few examples of the application of QNN to multivariate regression models, and little is known about how these models are constructed. This study aims to construct a QNN model to predict the melting points of metal oxides as an example of a multivariate regression task for the MI problem. Different architectures (encoding methods and entangler arrangements) are explored to create an effective QNN model. Shallow-depth ansatzs could achieve sufficient expressibility using sufficiently entangled circuits. The “linear” entangler was adequate for providing the necessary entanglement. The expressibility of the QNN model could be further improved by increasing the circuit width. The generalization performance could also be improved, outperforming the classical NN model. No overfitting was observed in the QNN models with a well-designed encoder. These findings suggest that QNN can be a useful tool for MI.

SCDC-Net: Detection and Classification of Skin Cancer using Deep q Neural Network with Hybrid Features

The increment in radiation in atmosphere causing wide spread of skin cancer disease around the globe. Cancer of the skin is one of the most lethal forms of the illness, and it is responsible for millions of deaths annually all over the world. Because of this, the early identification of skin cancer using computer-aided diagnostic (CAD) technology might potentially save a person's life.  However, skin cancer classification of performance is affected by hair artifacts and improper segmentation of lesion. The conventional methods were failed to achieve higher performance due to improper feature analysis. Therefore, to achieve the robust performance, this work was developed the transfer learning-based skin cancer detection and classification (SCDC-Net). Initially, UG-Net model was introduced to eliminate hair from skin lesions by enhancing the region of interest, which is developed by combining U-shaped deep learning network (U-Net) with generative adversarial networks (GAN). In addition, Hybrid U-Net (HU-Net) perform the segmentation operation for identifying the disease effected region of interest. Further, the features are extracted using gray level cooccurrence matrix (GLCM) matrix and discrete wavelet transform (DWT), which extracts the color, texture, statistical features. Finally, these combined characteristics are used to train a deep q neural network (DQNN) model, which then classifies a number of different skin cancer illness categories. The simulations conducted on real-time ISIC-2019 dataset proved that proposed SCDC-Net outperformed in all performance measures as compared to existing methods.

Estimating Surgical Urethral Length on Intraoperative Robot-Assisted Prostatectomy Images using Artificial Intelligence Anatomy Recognition.

Objective To construct a Convolutional Neural Network (CNN) model that can recognize and delineate anatomic structures on intraoperative video frames of robot-assisted radical prostatectomy (RARP) and to use these annotations to predict the surgical urethral length (SUL). Background Urethral dissection during RARP impacts patient urinary incontinence (UI) outcomes, and requires extensive training. Large differences exist between incontinence outcomes of different urologists and hospitals. Also, surgeon experience and education are critical towards optimal outcomes. Therefore new approaches are warranted. SUL is associated with UI. Artificial intelligence (AI) surgical image segmentation using a CNN could automate SUL estimation and contribute towards future AI-assisted RARP and surgeon guidance. Methods Eighty-eight intraoperative RARP videos between June 2009 and September 2014 were collected from a single center. 264 frames were annotated according to: prostate, urethra, ligated plexus and catheter. Thirty annotated images from different RARP videos were used as a test dataset. The Dice coefficient (DSC) and 95th percentile Hausdorff distance (Hd95) were used to determine model performance. SUL was calculated using the catheter as a reference. Results The DSC of the best performing model were 0.735 and 0.755 for the catheter and urethra classes respectively, with a Hd95 of 29.27 and 72.62 respectively. The model performed moderately on the ligated plexus and prostate. The predicted SUL showed a mean difference of 0.64 - 1.86mm difference versus human annotators, but with significant deviation (SD 3.28 - 3.56). Conclusion This study shows that an AI image segmentation model can predict vital structures during RARP urethral dissection with moderate to fair accuracy. SUL estimation derived from it showed large deviations and outliers when compared to human annotators, but with a very small mean difference (<2mm). This is a promising development for further research on AI-assisted RARP. Keywords Prostate cancer, Anatomy recognition, Artificial intelligence, Continence, Urethral length.

Damage Identification of Simple Supported Bridges Under Moving Loads Based on Variational Mode Decomposition and Deep Learning

Aiming at rapid and economical damage detection of a large number of simple supported bridges, a new structural damage identification method under moving load based on variational mode decomposition (VMD) and deep learning is proposed. Firstly, a moving vehicle is used as an exciting load to invoke structural damage feature and enhance the signal-to-noise ratio, and the structural vertical acceleration response is extracted by a finite element simulating analysis under various damage cases. In order to simulate the influence of noise and expand the samples, Gaussian white noise is added to the extracted data, and then the response signal is decomposed into a series of intrinsic mode functions (IMFs) using VMD, and the optimal IMF component is selected as the damage sample of the structure. Then, a one-dimensional convolutional neural network (CNN) model is built and trained by the various samples of damage. The vibration response of the practical bridge is processed and inputted by the trained CNN model to identify the location of the damage and degree of the structure. Finally, the effectiveness and anti-noise performance of the proposed method are verified through numerical analysis and a simply supported beam bridge model experiment. The results show that the average identification accuracy of the numerical simulations and experimental is 93.4% and 86.8% with 20% Gaussian white noise, respectively. Sensors at different locations have almost the same identification effect for various cases of damage, so it is possible to identify structural damage only using a small amount of accelerometer.

Deep learning evaluation of echocardiograms to identify occult atrial fibrillation

Atrial fibrillation (AF) often escapes detection, given its frequent paroxysmal and asymptomatic presentation. Deep learning of transthoracic echocardiograms (TTEs), which have structural information, could help identify occult AF. We created a two-stage deep learning algorithm using a video-based convolutional neural network model that (1) distinguished whether TTEs were in sinus rhythm or AF and then (2) predicted which of the TTEs in sinus rhythm were in patients who had experienced AF within 90 days. Our model, trained on 111,319 TTE videos, distinguished TTEs in AF from those in sinus rhythm with high accuracy in a held-out test cohort (AUC 0.96 (0.95–0.96), AUPRC 0.91 (0.90–0.92)). Among TTEs in sinus rhythm, the model predicted the presence of concurrent paroxysmal AF (AUC 0.74 (0.71–0.77), AUPRC 0.19 (0.16–0.23)). Model discrimination remained similar in an external cohort of 10,203 TTEs (AUC of 0.69 (0.67–0.70), AUPRC 0.34 (0.31–0.36)). Performance held across patients who were women (AUC 0.76 (0.72–0.81)), older than 65 years (0.73 (0.69–0.76)), or had a CHA2DS2VASc ≥2 (0.73 (0.79–0.77)). The model performed better than using clinical risk factors (AUC 0.64 (0.62–0.67)), TTE measurements (0.64 (0.62–0.67)), left atrial size (0.63 (0.62–0.64)), or CHA2DS2VASc (0.61 (0.60–0.62)). An ensemble model in a cohort subset combining the TTE model with an electrocardiogram (ECGs) deep learning model performed better than using the ECG model alone (AUC 0.81 vs. 0.79, p = 0.01). Deep learning using TTEs can predict patients with active or occult AF and could be used for opportunistic AF screening that could lead to earlier treatment.

Quality detection and variety classification of pecan seeds using hyperspectral imaging technology combined with machine learning

Pecan (Carya illinoinensis K.), a well-known dried seed and woody oil tree, faces challenges in its industry due to complex quality assessment methods and confusing varieties. These challenges have seriously hampered the development of a large-scale pecan deep processing industry. This work aimed to apply hyperspectral imaging technology (HSI) combined with machine learning to evaluate the quality of pecan seeds and perform variety classification. The samples of this work were composed of 19 varieties of pecan seeds, with 30 seeds per variety. After spectral preprocessing, spectral features were extracted from the spectral profiles using feature extraction methods. Back-propagation neural network models and partial least squares models were established to predict the contents of crude fat and moisture in pecan seeds. Predictions of the best models gave good results with R2-score of 0.887 for the crude fat model and 0.950 for the moisture model. Additionally, support vector machine models were developed to identify pecan varieties. The model achieved good results in 19 pecan varieties identification with accuracy of 0.965. In conclusion, the combination of HSI and machine learning could be an effective tool in improving the pecan industry and providing sustainable and efficient methods in the production of pecan seeds.