Health Monitoring Research Articles

Facial Expression Analysis for Efficient Disease Classification in Sheep Using a 3NM-CTA and LIFA-Based Framework

Facial Expression (FE) in sheep is analysed to distinguish normal and abnormal sheep behaviours that are crucial for health monitoring and welfare assessment. However, the prevailing techniques failed to concentrate on the sheep disease type, which hindered the practical treatment analysis. Therefore, the proposed work used 3 Non-Monotonic Mish Convoluted T-Max Average Neural Network (3NM-CTA) and Lyapunov Integrated Fuzzy Algorithm (LIFA) to efficiently detect the sheep disease type for a faster and more accurate treatment process. First, the sheep images are pre-processed, and the sheep's facial parts are detected using the Ortho-Greedy Viola Jonas Initialization Algorithm (OGVJI). Then, the facial landmarks are detected for each part, and the features of the landmarked eye, ear, nose, and mouth are extracted. Meanwhile, from the eye landmarks, the eye aspect ratio is calculated to predict Blepharospasm in sheep. After that, 3NM-CTA predicts the normal and abnormal sheep. Here, for abnormal sheep, the PH value is estimated. Finally, the estimated PH value and calculated aspect ratio are given to LIFA, which categorizes Acidosis, Mastitis, Listeriosis, and Blepharospasm in sheep using a minimum Fuzzification Time of 3465 ms, respectively. Thus, the proposed work outperformed the traditional methodologies.

Comparison of three point-of-care blood testing instruments for rapid on-site health monitoring of Atlantic salmon Salmo salar.

Biomarkers in blood are useful for assessing health and welfare in animals. This study evaluated the agreement among 3 point-of-care testing (POCT) instruments (Seamaty SMT-120VP, Mnchip Pointcare V2/V3, and Zoetis Vetscan VS2 analyzer) on Atlantic salmon Salmo salar. A repeatability study investigated internal measurement variation. In total, 60 plasma samples from adult fish were analyzed simultaneously using different rotors with multiple biomarkers. A comparison between blood and plasma was conducted on 35 blood samples. Lin's concordance correlation coefficient was <0.9 for all analyte comparisons between the 3 POCT except for bile acids; therefore, the McBride strength of agreement was generally poor and was moderate for bile acids. Internal measurement showed a low coefficient of variation for most analytes, except for aspartate aminotransferase (Pointcare V2/V3), alanine transaminase (Pointcare V2/V3), blood urea nitrogen (Pointcare V2/V3), and creatinine (Pointcare V2/V3, SMT-120VP). There was high concordance between whole blood and plasma samples for most analytes on both SMT-120VP and Pointcare V2/V3 systems, except for sodium, total bilirubin, and total CO2. This study underscores the necessity for system-specific calibration and validation of POCT systems like the Seamaty SMT-120VP and Mnchip Pointcare V2/V3 when used in aquaculture for clinical assessment of Atlantic salmon. The reproducibility study demonstrated that the precision of analysis was acceptable for most analytes. The comparison between whole blood and plasma suggests that whole blood can be used on-site to reduce the complexity of analysis. In summary, these systems offer promising tools for rapid on-site health monitoring in salmonid aquaculture but they require validation against gold-standard methods.

High-Strength Conductive Hydrogel Fiber Prepared Via Microfluidic Technology for Functionalized Strain Sensing.

The rapid advancement of wearable flexible electronics has heightened the demand for hydrogel materials that combine mechanical robustness with electrical conductivity. Herein, the TEMPO-oxidized cellulose nanofibers-Graphene nanosheets/poly(vinyl alcohol)-sodium alginate-tannic acid (TOCN-GN/PVA-SA-TA, TGG) composite hydrogel fibers are prepared by microfluidic spinning technology to solve the bottleneck problems of poor dispersion of GN and imbalance of mechanical-conductive properties of traditional hydrogels. TOCN, acting as a biotemplate, effectively inhibits GN agglomeration via hydrogen bonding and mechanical interlocking, thereby enhancing GN dispersion and facilitating the formation of 3D conductive networks within hydrogel fibers. The optimized TGG fibers achieved a tensile strength of 0.96MPa, 150% elongation at break, and electrical conductivity of 2.66 S m-1, while exhibiting enhanced energy dissipation and fatigue resistance. As strain sensors, TGG fibers demonstrated high sensitivity (gauge factor is 1.81 at 40-100% strain) and rapid response (≈0.3 s), enabling precise monitoring of joint movements, facial micro-expressions, and swallowing actions. Furthermore, PDMS-encapsulated textile sensors enabled encrypted Morse code transmission, demonstrating innovative potential for next-generation flexible electronics in health monitoring and human-machine interfaces.

Detection of Negative Emotions in Short Texts Using Deep Neural Networks.

Emotion detection is crucial in various domains, including psychology, health, social sciences, and marketing. Specifically, in psychology, identifying negative emotions in short Spanish texts, such as tweets, is vital for understanding individuals' emotional states. However, this process is challenging because of factors such as lack of context, cultural nuances, and ambiguous expressions. Although much research on emotion classification in tweets has focused on applications such as crisis analysis, mental health monitoring, and affective computing, most of it has been conducted in English, leaving a significant gap in addressing the emotional needs of Spanish-speaking communities. To address this gap, we used a corpus of 12,000 Spanish tweets tagged with Ekman's negative emotions (sadness, anger, fear, and disgust). Traditional features (n-grams of different types and sizes), syntactic n-grams, and combined features were evaluated. Different deep neural networks, including convolutional neural networks, Bidirectional Encoder Representations of Transformers (BERT), and the robust optimized BERT approach called RoBERTa, were implemented and compared with traditional machine learning methods to identify the most effective method. Extensive testing revealed that BERT achieved the best result, with a macro F1 score of 0.9973. Furthermore, we reported the carbon emissions generated during the training of each implemented method. This study makes a unique contribution by focusing on negative emotions in Spanish, leveraging one of the largest and highest-quality corpora available. It stands out for implementing advanced transformers such as RoBERTa and integrating combined and syntactic n-grams in traditional methods. Furthermore, it highlights how parameters, features, and preprocessing significantly influence performance.

All-Inclusive Sensing Tablet with Integrated Passive Mixer for Ultraviscous Solutions.

Developing low-cost and easy-to-use point-of-care devices is necessary for timely disease diagnosis and health monitoring. Here, we introduce all-inclusive, tablet-based chemo/biosensors with rapid automixing features, capable of mixing in highly viscous solutions with viscosities up to 1700 mPa·s. These tablets are created using a simple powder compression method and contain all necessary reagents to perform assays in a "drop-and-detect" manner, without the need for vigorous shaking or vortex mixing. As proof of concept, we demonstrated the applicability of our Speedy tablets for detecting nitrite in human saliva, a challenging medium due to its viscosity. The strong mixing capability of the proposed tablets ensured consistent and reliable results across range of viscosities, from low to high, while delivering an excellent detection range of 0.03-1.50 mg/dL, covering nitrite levels in human saliva. Additionally, we developed a straightforward method to encapsulate enzymes in trehalose, making them bulkier and more stable using only a mist sprayer, nonstick tray, and spatula, eliminating the need for expensive equipment. This approach allowed us to incorporate small amounts of enzymes into tablet formulations and fabricate the first automixing tablet biosensor. These biosensors were used for the bienzymatic detection of glucose in real human urine within the biologically relevant range of 0.3-2.5 mM, indicating the compatibility of automixing tablets with bioreagents. Each tablet costs less than $0.30 to produce and remains stable for at least one month at room temperature. The affordability and convenience of our tablets make them a valuable addition to the array of diagnostic tools.

In CNC: Cutting Tool Health Monitoring Using Convolutional Neural Networks

In CNC: Cutting Tool Health Monitoring Using Convolutional Neural Networks

Low temperature cross-sensitivity vector torsion sensor based on an in-fiber Mach-Zehnder interferometer with helical side-hole fiber

What we believe to be a novel in-fiber vector torsion sensor based on a Mach-Zehnder interferometer (MZI) constructed using helical side-hole fibers (HSHFs) is proposed. A segment of HSHF is spliced between two single-mode fibers to excite the LP01 and LP11 modes supported by the HSHF, generating intermodal interference and forming the MZI structure. Owing to the helical side-hole configuration, the interference patterns of the LP01 and LP11 modes rotate with the applied torsion angle. A theoretical model for torsion sensing reveals that the light intensity at the resonant dip of the interference patterns varies periodically with the increasing torsion angle. In contrast, the wavelength shift of the resonant dip exhibits a linear relationship with the torsion angle, enabling precise torsion measurement. Experimental results confirm the validation of the model, showing torsion sensitivities of 2.13 dB/(rad/m) and 0.282 nm/(rad/m) for intensity and wavelength responses, respectively. Furthermore, the proposed torsion sensor shows minimal crosstalk to temperature and axial strain thanks to the interference between the two core modes. Additionally, introducing the helical structure during fiber drawing simplifies processing and preserves strength, providing a wide measurement range. The proposed HSHF-MZI torsion sensor exhibits high sensitivity, low temperature and strain crosstalk, and the ability to detect the torsion direction, making it a promising candidate for applications in structural health monitoring and engineering measurements.

Piezoelectric-based impedance monitoring: A critical analysis of indirect electromechanical impedance measurements in non-destructive evaluation

Electromechanical impedance (EMI) measurements have revolutionized non-destructive evaluation (NDE) of manufactured parts. By bonding piezoelectric elements to structures, EMI analyzes dynamic responses to identify manufacturing anomalies. Particularly effective in additive manufacturing (AM), EMI detects defects like mass alterations and internal porosity. In structural health monitoring, it continuously compares impedance signatures to baseline, detecting emerging defects over time. This innovative approach addresses limitations of conventional methods, providing a powerful tool for assessing material integrity. The practical application of EMI however faces challenges due to the need for direct instrumentation of each part with piezoelectric elements, which introduces time, cost, and variability issues. To address these, the concept of indirect EMI (IEMI) has emerged, utilizing a secondary structure or instrumented fixture to temporarily couple with the part under test. This approach allows for testing multiple specimens, reduces labor requirements, and facilitates process automation. This paper investigates the sensitivity of IEMI to defects in the specimen, focusing on fixture design, instrumentation process, and clamping force calibration. Various interface conditions and clamping forces are explored to understand their impact on defect detection capabilities. Experimental results indicate that interface conditions significantly influence IEMI measurements, with metal-on-metal contact providing the best sensitivity. Additionally, the clamping force is found to affect the impedance signature, emphasizing the need for consistent force application during measurements. Overall, this study underscores the potential of IEMI as a viable NDE solution, highlighting its ability to detect defects in manufactured parts while addressing practical implementation challenges. Future work will focus on optimizing fixture design for enhanced sensitivity.

MOF-MoS2 nanosheets doped PEDOT:PSS for organic electrochemical transistors in enhanced glucose sensing and machine learning-based concentration prediction

Abstract Organic electrochemical transistors (OECTs) are regarded as a promising platform for chemical and biological sensing due to their biocompatibility, cost-effectiveness and flexibility. However, maintaining long-term stability of OECTs while achieving high sensitivity remains a challenge for their practical applications. One of the main reasons is the relatively low electronic and ionic conductivity of the channel material. Herein, we present a p-type OECT fabricated by incorporating metal–organic framework (MOF)-MoS2 hybrid nanosheets into the PEDOT:PSS channel via solution-based processes. The strategy significantly improves the sensitivity of OECT, with the transconductance of the device increasing by ∼threefold to 19.34 mS. The higher transconductance is attributed to the hybrid MOF-MoS2 dopant, which not only enhances the electronic conductivity, but also strengthens ion transport and capacitance of the PEDOT:PSS film due to the synergistic effects from high electron mobility of MoS2 and MOF porous structure with large surface area. The fabricated OECT demonstrates high selectivity and sensitivity as a glucose biosensor across a wide concentration range in saliva. Finally, we illustrate the merits of integration machine learning algorithms to construct predictive models using the extensive datasets produced by our sensors for both classification and quantification tasks. These findings highlight the great potential of OECTs incorporating MOF-MoS2 hybrid, as a promising candidate for ultra-sensitive biological detections, and broaden the applications of our OECT biosensors for non-invasive health monitoring and wearable electronics.

Damage Correlation Analysis of Truss Structures Based on Monte Carlo Method and Bayesian Updating

In the field of structural engineering, the degradation of structural performance is an inherent characteristic that is difficult to avoid, making structural damage identification a core research topic for ensuring engineering safety. This study addresses the damage identification problem in truss structures by integrating the Monte Carlo method with Bayesian updating theory. A finite element numerical model of the truss structure was established using OpenSees to conduct research on damage parameter updating, systematically revealing the correlation between monitoring information and the stiffness parameters of members requiring updating. The results demonstrate that highly correlated monitoring information exerts a significant positive effect on parameter updating. Under the premise of ensuring parameter identification accuracy, the quantity of detection information can be reduced to effectively lower monitoring costs. The information quality evaluation framework developed in this study provides theoretical support for data optimization and screening in structural damage identification, offering valuable references for the construction of practical structural health monitoring systems. This research contributes methodological insights for enhancing the efficiency and economy of damage detection in engineering structures.

Intelligent Estimation Method for Sports Injuries Based on RBF Neural Network

Evaluating sports injuries is of great significance in improving the quality of sports. In order to achieve an accurate and effective evaluation of sports injuries, this paper proposes an intelligent estimation method for sports injuries based on the RBF neural network. It includes the following: sampling statistical data for sports evaluation based on a big data network environment; integrating outlier cleaning methods based on mean and standard deviation with weighted moving average filtering methods to achieve data cleaning, smoothing, and normalization processing; analyzing the injury factors that cause sports injuries from the perspectives of internal factors, external factors, and triggering stimuli; and extracting sports data features. On the basis of the quantification level of sports injury risk, an innovative radial basis function (RBF) neural network is used to achieve an intelligent estimation of sports injuries. The experimental results show that the proposed method for estimating sports injuries suggests significant advantages in comparing different methods. Specifically, when estimating sports injuries for test samples, the maximum time consumed by our method is only 0.94[Formula: see text]s. This indicator is particularly important in sports health monitoring systems that require high real-time performance, as it ensures that the system can respond promptly and provide injury risk estimates, thereby helping athletes or coaches take preventive measures quickly. Meanwhile, the performance of this method in estimating errors is also outstanding, with a maximum estimation error of only 0.015. This error value is very small compared to the actual risk value, indicating that this method can accurately quantify the risk of sports injuries. In practical applications, this high-precision estimation is crucial for developing personalized training plans, avoiding potential sports injuries, and helping to improve athletes’ training effectiveness and competitive performance.

Augmented Vision-Enabled Digital Twin System for Real-Time ICU Patient Monitoring and Emergency Response

Abstract: This project introduces a real-time health monitoring system for ICU emergency patients, leveraging Digital Twin (DT) technology integrated with Augmented Vision (AV). The proposed system enables early screening and rapid identification of high-risk individuals using a unique DT-based AV code assigned to each patient. Upon hospital admission for critical conditions such as cardiac or pulmonary emergencies, patients receive wearable biomedical sensors and an AV code printout, which facilitates swift ICU admission. The sensor data is continuously uploaded to a centralized server via IoT. When urgent care is needed, medical personnel can scan a patient's DT AV code using a dedicated AV camera system. The DT Vision Software instantly retrieves the patient's real-time sensor data and displays it in a visual format. This includes color-coded imagery: green Digital Twin images for stable patients and red Medical Reality (MR) images for patients showing signs of severe lung conditions. This system allows healthcare providers to prioritize treatment effectively, potentially saving lives through faster and more informed decision-making

Doctor Booking Management

Abstract - Health management systems play a crucial role in modern healthcare by enabling efficient handling of patient records, appointments, medical histories, billing, and communication between patients and healthcare providers. This project focuses on the development of a web-based Health Management System using PHP and MySQL to streamline healthcare services. The system is designed to cater to clinics, hospitals, and independent healthcare providers by offering an integrated platform for managing patients' information securely and effectively. It provides modules for patient registration, doctor scheduling, appointment booking, diagnosis tracking, medical prescriptions, and billing management. Security measures are incorporated to ensure that sensitive patient data remains protected, complying with data protection regulations. PHP was selected for development due to its flexibility, ease of use, and strong support for database interactions, while MySQL was used for storing structured medical data. Testing revealed that the system significantly reduces paperwork, minimizes errors, speeds up administrative processes, and improves service delivery in healthcare institutions. The user-friendly interface ensures that both patients and healthcare providers can navigate the system effortlessly. Feedback from initial users indicated high levels of satisfaction with the system's speed, reliability, and functionality. Future enhancements may include integration with wearable health monitoring devices and mobile app extensions for remote access. The proposed Health Management System represents a scalable, secure, and cost-effective solution for modern healthcare institutions aiming to digitize their services. Key Words: Health Management System, PHP, MySQL, Patient Records, Appointment Scheduling, Healthcare Digitization, Data Security, Medical Billing, Web Application, Hospital Management.

Validation of a Swine Cough Monitoring System Under Field Conditions

Precision livestock farming technologies support health monitoring on farms, yet few studies have evaluated their effectiveness under field conditions using reliable gold standards. This study evaluated a commercially available technology for detecting cough sounds in pigs on a commercial farm. Audio was recorded over six days using 16 microphones across two pig barns. A total of 1110 cough sounds were labelled by an on-site observer using a cough induction methodology, and 8938 other sounds from farm recordings and open-source datasets (ESC-50, UrbanSound8K, and AudioSet) were labelled. A hybrid deep learning model combining Convolutional Neural Networks and Recurrent Neural Networks was trained and evaluated using these labels. A total of 34 audio features were extracted from 1 s segments, including validated descriptors (e.g., MFCC), unverified external features, and proprietary features. Features were evaluated through 10-fold cross-validation based on classification performance and runtime, resulting in eight final features. The final model showed high performance (recall = 98.6%, specificity = 99.7%, precision = 98.8%, accuracy = 99.6%, F1-score = 98.6%). The technology tested was shown to be efficient for monitoring cough sounds in a commercial swine production facility. It is recommended to test the technology in other environments to evaluate the effectiveness in different farm settings.

Broadband-Responsive Rubbery Stretchable Vertical-Structured Photodetectors Based on Rubbery Stretchable Transparent Conductors.

A rubbery stretchable conductor with high conductivity and transparency is crucial for the development of rubbery stretchable vertical-structured photodetectors. However, the development of such a rubbery conductor is still nascent. Here, we report the scalable manufacturing of rubbery stretchable transparent conductors (RSTCs) and the development of a rubbery stretchable vertical-structured photodetector (RSVPD). The RSTC is fabricated into a specialized micromesh structure by utilizing a close-packed monolayer of polystyrene microspheres as a mask. The micromesh structure not only enhances the conductor's stretchability and transparency but also maintains its conductivity, making it ideal for various applications in stretchable electronics. The RSTCs are used to construct RSVPDs that have high response over a broad spectrum, and their electrical performances can be retained even when subjected to mechanical strains of up to 50%. Furthermore, a stretchable imager based on RSVPD was developed to detect the multipoint light distribution. Lastly, a photoplethysmography (PPG) sensor was also developed for real-time health monitoring.

Broadband Quantum‐Junction Photodiode Achieving Ultralow‐Noise Light Sensing

AbstractLow‐noise photodetectors hold immense promise for advancing weak‐light sensing in cutting‐edge applications such as health monitoring, intelligent driving, and military surveillance. Quantum‐junction photodiode (QJPD), composed solely of light‐absorbing quantum dots, are highly expected to overcome the noise limitation of conventional narrow‐bandgap semiconductor photodetectors due to their inherent advantages of low hot carrier density as well as weak inter‐dot electronic coupling. However, the targeted device engineering of QJPD remains a critical exploration area to fully unlock their low‐noise potential. Here, a low‐noise, broadband QJPD is proposed through the strategic incorporation of an ultrathin ALD SnOx layer. This SnOx modification efficiently suppresses the interfacial barrier, reduces the interface resistance and minimizes dark current of QJPDs. Consequently, SnOx‐modified QJPDs achieve an ultralow noise current of 4.85 × 10−14 A Hz⁻1/2 and a specific detectivity exceeding 1012 Jones across a broad spectral range (350−1050 nm), significantly outperforming unmodified QJPDs (≈1010 Jones) and other n‐i‐p quantum dot photodetectors (≈109 Jones). SnOx‐modified QJPDs enable high‐accuracy blood oxygen saturation measurements using near‐infrared light and real‐time heartbeat monitoring under weak ambient light conditions. This work establishes a foundation for the development of low‐noise QJPDs, underscoring their potential for weak‐light detection in challenging environments with strong hindering and ambient light.

Emerging trends of biomedical nanotechnology in nutrition, health monitoring and disease diagnosis.

The transdisciplinary nature of nanotechnology has facilitated its application across various fields, especially in biological sciences. The primary aim of this review is to consolidate the many facets of nanomedicine, theranostics, and nanotechnology in food preservation into a unified framework and to underscore established research methodologies in the medical domain. Nanoparticles serve a crucial function in improving the bioavailability of orally delivered bioactive substances. This review demonstrated that nanoparticles can enhance the bioavailability of micronutrients, such as vitamin B12, vitamin A, folic acid, and iron. New advances in nanotechnology have made big differences in finding pathogens and killing them specifically, helping peopleto get better health through medication delivery and imaging, improving food packaging better so it lasts longer, and making foods healthier overall. Nanotechnology currently enhances the safety of delivering highly hazardous medicines through the use of nanozymes that exhibit antioxidant and antibacterial characteristics. Moreover, wearable devices can identify significant alterations in vital signs, medical problems, and infections occurring within the body. We anticipate that these technologies will provide physicians with enhanced direct access to crucial information about the causes of changes in vital signs or diseases, as they are directly connected to the source of the problem. This review paper thoroughly examines the latest developments in nanomaterials and nanozymes as antimicrobial agents in food science and nutrition, wound healing, illness diagnostics, imaging, and potential future uses. The paper presents a concise and structured report on nanotechnology, which will be beneficial to researchers and scientists for future research opportunities.

Motion‐Interference Free and Self‐Compensated Multi‐Receptor Skin with all Gel for Sensory Enhancement

AbstractStretchable multimodal electronic skin (e‐skin) has attracted intensive research interest but faces great challenges related to strain interference, crosstalk issues, and integration of multiple sensitive materials. Herein, a stretchable and strain‐isolated multimodal (SSIM) e‐skin capable of concurrently and sensitively monitoring temperature, humidity, UV light, and oxygen, while also possessing self‐compensation capability is developed. The SSIM sensing platform is created by chemically anchoring polyethylene terephthalate onto polydimethylsiloxane through silane treatment to form island‐bridge structures. This method effectively isolates strain and improves interfacial adhesion, achieving a state‐of‐the‐art low strain interference of 0.2% and an adhesion energy exceeding 300 J m−2 (13.4 times that of the untreated material), ensuring the e‐skin's stable operation even under dynamic stretching. To mitigate crosstalk and fabrication complexity, a single hydrogel film is employed to facilitate self‐compensating multimodal sensing through various sensing mechanisms and physical isolations. The SSIM e‐skin can simultaneously monitor several environmental and physiological signals with minimized crosstalk without interference from body movements. It enables remote respiration monitoring with wireless circuitry, highlighting its substantial potential in health monitoring, medical diagnostics, and neurorehabilitation.

Predicting climate change impacts on groundwater aquifer levels in the Henan North China Plain

Monitoring GWL over extended periods is crucial for comprehending the fluctuations of groundwater resources in the present context for ongoing global changes. This study analyzed the effects of climate variations on the GWL in Henan Province North China Plain using two deep-learning models Bidirectional Long Short-Term Memory (BidLSTM) and Gated Recurrent Unit (GRU). These models predicted monthly variations in GWL at 85 monitoring wells across the area using a dataset from 1980 to 2015. For validation and evaluation, both models were quantitatively calibrated using training set (1980–2015) to predict GWL from 2016 to 2100. The dataset was partitioned, with 80% allocated for training and 20% for testing. The result interpreted that in AHP3 well, GWL declined to 120 m in 1980 due to reduced precipitation 57 mm and Et 62 mm, while temperature stayed at 10 °C as of 2070, In the Zhengzhou and Keifing regions GWL declined by 98 m in the 1980 s despite rising precipitation 72 mm and Et 60 mm, due to insufficient recharge by 2100, GWL is expected to reach 140 m, driven by climate changes, including a temperature increase to 17 °C. The results indicated significant changes with the effect of precipitation, significant increase in temperature and surface Et. Anthropogenic activity also impacted GWL in the area. The trained models demonstrated good performance, with a prediction error of 0.0350, 0.0346 m, and the root mean square error (RMSE) was recorded at 0.1870, 0.1860 m. By accurately predicting GWLs, the BidLSTM model can help ensure that groundwater resources are used sustainably and efficiently.

Fault diagnosis of wind turbine blades under wide-weather multi-operating conditions based on multi-modal information fusion and deep learning

Continuous health monitoring of wind turbine blades under all-weather and multi-operating conditions represents a significant challenge in the renewable energy sector. In this article, we present a fault diagnosis approach leveraging multi-modal information fusion and deep learning with continuous state division, thereby overcoming the limitations of traditional methods in complex and noisy environments. The operational conditions of wind turbine blades are categorized into two primary states: the wind operation state and the sudden shutdown state. Additionally, various climate types, including sunny, foggy, windy conditions, differing lighting levels, and others, are considered in the analysis. During wind operation, sound and vibration signals exhibit higher efficacy for fault detection; however, high noise levels may introduce interferences. To address the issue of indistinct fault characteristics after deep convolution due to multiple noise factors, which could result in reduced diagnostic accuracy, we propose a robust fast Fourier transform-ResTransNet model. In the shutdown state, vibration and sound data features become less prominent, making image processing techniques advantageous. Nevertheless, diverse climate types can lead to challenges such as low visibility, high noise, and other interferences. Consequently, we design a Swin-Transformer model that integrates infrared thermal imaging and visible light imagery. This model resolves the problem of non-homologous data representation and ensures accurate information interaction under multi-source data fusion. Simulation results confirm that the proposed fault diagnosis method achieves substantial improvements over existing approaches. To validate the practical applicability of our method, we construct a real-world wind turbine operational environment, simulate several common blade fault scenarios, and collect actual vibration, sound, and image data under varying weather conditions. Based on these simulations, we establish a multi-modal information fusion model tailored for different weather types. Furthermore, to facilitate the integration of our research into real-world wind farm operations, we develop a human–computer interface that enables seamless deployment. The corresponding source code is publicly available at https://github.com/midfigher/Humancomputer-interaction .