Intermediate Layer Research Articles

Effect of diffusion barriers on steam oxidation properties and interface evolutions of Cr coating for zirconium alloy at 1200 °C

Currently, Cr-based coatings are considered the most promising protective coatings for zirconium alloy claddings under the ATF design concept. However, the primary factor affecting the oxidation resistance of Cr coatings is the interdiffusion of Cr and Zr. In this study, Cr coatings with three refractory metal interlayers of Nb, Mo, and Ta were deposited on zirconium alloy surfaces using closed-field unbalanced magnetron sputtering. The study investigated the effects of different interlayers on the microstructure, oxidation properties, and interface evolution process of Cr coatings under steam conditions at 1200 ℃, and compared them with Cr coatings without intermediate layers. The results indicate that the refractory metal interlayer influences the orientation and growth rate of Cr coating grains. The formation of a Laves phase mixed layer in the interlayer during oxidation effectively hinders the diffusion of O and Cr to the zirconium alloy matrix, enhancing its oxidation resistance. Furthermore, the diffusion rate and vacancy concentration of Mo and Ta elements to the Zr layer exceed that of diffusion to the Cr layer. During the cooling process, a precipitate phase forms in the Zr-4 matrix, while the Nb coating forms an infinite solid solution in the β-Zr.

Analysis of High and Low Temperature Performance of Titanium Zirconium Molybdenum Alloy and High Temperature Alloy Brazed Joint

Abstract Using niobium as the intermediate layer and gold nickel (Au82-Ni) filler metal, TZM alloy and high-temperature alloy GH3128 were brazed. The brazed specimens were subjected to 400 cycles of high and low temperature cycling tests at temperatures ranging from 650 °C ∼ 680 °C (high temperature) to -196 °C (low temperature) for 10 minutes seperately. Then the microstructure variation before and after the experiment were observed and the shear strength of the specimens were tested at room temperature for 50 cycles of 400 high and low temperature testing.. The results show that the microstructure of the transition zone at the brazing interface after brazing is uniform and dense, with good wettability with the substrate and no defects such as cracks. The solder and substrate undergo diffusion reactions, and the molybdenum alloy, niobium, and niobium, high-temperature alloy had each form diffusion layers, formed intermetallic compounds on the brazing surface. After high and low temperature cycling, cracks exist in the braze seam between molybdenum alloy and niobium alloy. The cracking mode should be thermal fatigue based on the fracture morphology and phase analysis. The reasons of interface cracking are that due to different thermal expansion coefficients, low mutual solid solubility, and the formation of a large number of different types of intermetallic compounds, Mo and Nb materials have poor brazing compatibility and lower interface strength. Under cold and hot cycling conditions, brittle intermetallic compounds are prone to form crack source areas. As the number of cycles increases, the cracks gradually expand, ultimately leading to product leakage. In addition, the shear strength of brazed joints shows a decreasing trend with the increase of high and low temperature cycles in 400 high and low temperature tests.

Direct electrodeposition on conductive polymer structures for microwave device manufacturing

This work introduces a method that merges Additive Manufacturing (AM) with Direct Electrodeposition to efficiently produce microwave (MW) components. Using material extrusion (MEX) with conductive polymers and an improved direct copper deposition approach, this method streamlines the manufacturing process, bypassing the need for intermediate conductive layers. By minimizing the distance electrical currents travel within the conductive polymer, we obtain homogeneous copper deposits directly on the polymer with low roughness and excellent adhesion. The practical application of our technique is showcased through the radiative performance evaluation of a millimeter wave horn antenna manufactured using our approach, demonstrating similar electromagnetic performances with respect to a commercially available reference. The method proposed in this paper results in lightweight, material-efficient components with increased design flexibility, opening up possibilities for rapid prototyping and fabrication of new MW devices.

Modeling forest canopy structure and developing a stand health index using satellite remote sensing

Biotic and abiotic disturbances modify tree structure and degrade stand health. Accurate geospatial data on stand structure is important for monitoring tree growth, forest health, progression and severity of diseases and pests, and estimating resilience to climate stress. The live crown ratio (LCR) of trees serves as a key health indicator but has been understudied at the landscape level using remote sensing data. This study generated the leaf area index (LAI) and a novel spatial layer of LCR at site and landscape scales using a combination of satellite data and ground observations. We conducted field surveys to collect plot-level (10 m × 10 m) data in four eastern white pine (EWP; Pinus strobus L.)-dominated sites in the state of Maine, USA. The plot-level data were used to develop regression models for LAI and LCR estimation using microwave (Sentinel-1) and optical (Sentinel-2) remote sensing data and applying the Random Forest (RF) and Support Vector Machine (SVM) machine learning algorithms. The RF model showed higher prediction accuracy than the SVM model at the site level. Moreover, the prediction accuracy at the site and landscape levels were comparable for LAI (R2 > 0.76) and LCR (R2 > 0.71) using the RF model. Furthermore, the predicted LAI and LCR were integrated with canopy height and stand density to develop a novel health index map for EWP. The resulting health index map successfully delineated patches representing various health categories. Forestry practitioners and decision-makers can use the derived health index map and intermediate spatial data layers (LAI and LCR) to guide stand management. The developed framework can potentially be applied to other coniferous and broadleaved species for remote sensing-based LCR estimation and forest health assessment upon further studies and verification.

Microstructure and mechanical properties of CP780 steel - 7075 aluminum alloy laser welded joint assisted by rotating magnetic field

Abstract This study uses a rotating magnetic field for laser welding on 1 mm thick CP780 high-strength steel and 1.5 mm thick 7075 aluminum alloy. The effects of different welding parameters (B = 0 mT, B = 65 mT with V = 0°/s, B = 65 mT with V = 10°/s) on the morphology, microstructure, and tensile properties of welded joints are analyzed. At B = 0 mT, the weld shape is V-shaped, with the intermetallic compounds primarily consisting of needle-like brittle Al-rich (Fe, Si)Al2 phase and fewer granular ductile Fe-rich (Fe, Si)Al phase, resulting in poor mechanical properties. With the application of the rotating magnetic field, the laser energy becomes more concentrated, forming a ‘T’ shape weld. The rotating magnetic field (B = 65 mT with V = 10°/s) generates a constantly changing Lorentz force, promoting molten pool flow and enhancing Fe diffusion within the weld. This process reduces needle-like brittle Al-rich (Fe, Si)Al2 phase and increases granular ductile Fe-rich (Fe, Si)Al phase. It also accelerates the weld cooling rate and inhibits the reaction time and grain growth of intermetallic compounds, thereby reducing the thickness and content of the intermediate transition layer and significantly improving mechanical properties. A comprehensive comparison shows that the best mechanical properties are achieved at B = 65 mT with V = 10°/s. This study offers new insights and a theoretical foundation for achieving cost-effective, high-performance welded joints in advanced high-strength steel and high-strength aluminum alloy for automobiles, thereby facilitating lightweight vehicle development.

Improving the color gamut of gold nanostructures using vertical tandem nano-disks

Metallic nanoparticles have been used in structural coloration applications since they provide an alternative for the synthesis of color filters, with extended color gamut, enhanced color saturation, and brightness. These optical properties can be further improved when integrating nanoparticles into coupled dimers that give rise to enhanced hybridized plasmon resonances. In this contribution, gold nano-disks are split into Au/SiO2/Au layered structures to enhance its plasmon resonances and evaluate its improved performance as reflective color filters. To do so, the reflectance spectra of the layered disks were obtained with numerical simulations, identifying high amplitude peaks in the visible band and low-amplitude peaks in the near infrared one, related to the hybridized resonances. By choosing the appropriate size of the intermediate dielectric and metallic layers, as well as the near-field coupling, peaks can be tuned along the visible and near infrared bands, respectively. The chromatic coordinates of the addressed nanostructures obtained from the reflectance spectra show that the nanostructures hold color gamut that extends toward magenta, orange, yellowish, and reddish shades of colors.

Optical quality changes of the eye during peak SARS-CoV-2 pandemic in young adults

IntroductionTo determine whether severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection affects corneal morphology and optical quality. MethodsIn this cross-sectional study, ophthalmological indicators were examined during the peak of SARS-CoV-2 infection after adjusting for public health control measures. Participants were divided into control (remained uninfected), A (infected during follow-up), and B (infected prior to the first consultation) groups. Effects of varying SARS-CoV-2 infection levels were determined using reverse transcription polymerase chain reaction. Changes in corneal morphology, backscatter, and aberrations were measured. Corneal parameters, such as flat keratometry, steep keratometry, and surface variance, vertical asymmetry, height asymmetry, and height decentration indices were considered. ResultsOverall, 110 participants (208 eyes, 42.7% male; age 17–31 years) were enrolled. Eighteen (16.3%) were infection-free during the outbreak with unchanged corneal morphology, backscatter, and aberration. In group A, 34.73 ± 9.30 days after infection, the backscatter of the anterior corneal layer and central layer (both p=0.000) decreased. Total low-order aberration, defocus, horizontal coma, and spherical aberration of the cornea increased (p<0.05), while corneal morphology after infection did not change (p>0.05). In group B, a decrease in backscattering in the corneal middle layer and an increase in horizontal coma (p<0.05) were noted. ConclusionBackscattering of the anterior and intermediate layers of the cornea decreased and corneal aberrations increased after SARS-CoV-2 infection, which affected corneal optical quality. However, corneal morphology and thickness remained unchanged. Ophthalmological indicators and optical quality should be monitored during SARS-CoV-2 infection.

Online probabilistic knowledge distillation on cryptocurrency trading using Deep Reinforcement Learning

Leveraging Deep Reinforcement Learning (DRL) for training agents for financial trading has gained significant attention in recent years. However, training these agents in noisy financial environments remains challenging and unstable, significantly impacting their performance as trading agents, as the recent literature has also showcased. This paper introduces a novel distillation method for DRL agents, aiming to improve the training stability of DRL agents. The proposed method transfers knowledge from a teacher ensemble to a student model, incorporating both the action probability distribution knowledge from the output layer, as well as the knowledge from the intermediate layers of the teacher’s network. Furthermore, the proposed method also works in an online fashion, allowing for eliminating the separate teacher training process typically involved in many DRL distillation pipelines, simplifying the distillation process. The proposed method is extensively evaluated on a large-scale cryptocurrency trading setup, demonstrating its ability to both lead to significant improvements in trading accuracy and obtained profit, as well as increase the stability of the training process.

Improved JPEG Lossless Compression for Compression of Intermediate Layers in Neural Networks Based on Compute-In-Memory

With the development of Convolutional Neural Networks (CNNs), there is a growing requirement for their deployment on edge devices. At the same time, Compute-In-Memory (CIM) technology has gained significant attention in edge CNN applications due to its ability to minimize data movement between memory and computing units. However, the deployment of complex deep neural network models on edge devices with restricted hardware resources continues to be challenged by a lack of adequate storage for intermediate layer data. In this article, we propose an optimized JPEG Lossless Compression (JPEG-LS) algorithm that implements serial context parameter updating alongside parallel encoding. This method is designed for the global prediction and efficient compression of intermediate data layers in neural networks employing CIM techniques. The results indicate average compression ratios of 6.44× for VGG16, 3.62× for ResNet34, 1.67× for MobileNetV2, and 2.31× for InceptionV3. Moreover, the implementation achieves a data throughput of 32 bits per cycle at 600 MHz on the TSMC 28 nm, with a hardware cost of 122 K Gate Count.

Integrated terahertz topological valley-locked power divider with arbitrary power ratios

Integrated power dividers (PDs) are essential in terahertz (THz) communication and radar systems, but miniaturization often leads to performance degradation due to fabrication inaccuracies and sharp bends. Topological photonics offers a solution to these issues, yet creating THz power dividers with arbitrary splitting ratios remains challenging. We present a design methodology for on-chip topological THz power dividers with customizable splitting ratios using valley-locked photonic crystals. These crystals feature a tri-layered structure with two distinct valley Chern number layers and an intermediate semimetal layer. Utilizing the Jackiw–Rebbi model, we show that the characteristic impedance of the valley-locked photonic crystals, and thus the power division ratio, can be tuned by adjusting the semimetal layer width. Our approach is validated through simulations and experiments for both equal (1:1) and unequal (4:9) power ratios. This method enables efficient navigation around sharp bends and robust THz on-chip connectivity.

Long-Term Effects of 90Sr and 239+240Pu Contaminated on the Seafloor Sediment on the Transfer through Marine Food Web: Sensitivity to the Physical Property- and Sediment-Related Parameters

This study, as a follow-up to the &lt;sup&gt;137&lt;/sup&gt;Cs study of Kim et al. (2023), performs a set of sensitivity experiments on ocean physical properties and sediment-related parameters based on initial bottom contamination of &lt;sup&gt;90&lt;/sup&gt;Sr and &lt;sup&gt;239+240&lt;/sup&gt;Pu in the upper sediment layer. Both studies utilizes the Extended BURN model proposed by Bezhenar et al. (2016). The initial contamination levels were set to 3 and 0.4 Bq/kg, respectively, based on observations off the coast of Fukushima. The sensitivity experiments were divided into EXP-WOHAD, which examined the effects of the vertical diffusion coefficient and vertical sediment flux, and EXP-WHAD, which investigated the effects of ocean currents. The results of the EXP-WOHAD experiments showed that both radionuclides are more sensitive to sediment fluxes than to vertical diffusion coefficients. For both &lt;sup&gt;90&lt;/sup&gt;Sr and &lt;sup&gt;239+240&lt;/sup&gt;Pu, transport was dominated by transport to the intermediate sediment layer rather than to seawater. Specifically, for &lt;sup&gt;239+240&lt;/sup&gt;Pu, transport to seawater was insignificant due to its strong sorption properties in sediments. In terms of bioaccumulation, &lt;sup&gt;90&lt;/sup&gt;Sr accumulated most in fish and invertebrates, while &lt;sup&gt;239+240&lt;/sup&gt;Pu accumulated most in invertebrates, followed by zooplankton. Notably, &lt;sup&gt;239+240&lt;/sup&gt;Pu showed high bioaccumulation despite its low initial sediment contamination. The changes in radioactivity concentrations in the EXP-WHAD experiments were similar to those in the EXP-WOHAD experiments, and the bioaccumulation trends in biota were also comparable. This study confirms that the accumulation of radionuclides in seawater, intermediate sediments, and biota is sensitive to the marine physical environment and varies by nuclide. When the sediment flux increased by a factor of 100, the remaining radioactivity levels of &lt;sup&gt;90&lt;/sup&gt;Sr and &lt;sup&gt;239+240&lt;/sup&gt;Pu in biota after 100 years showed a significant reduction, decreasing by more than 900 and 30,000 times, respectively; it highlighting the importance of parameterization according to the physical environment of the area of interest.

Assessment of Siltation Impact and Mitigation Strategies for Sustaining Storage Capacity in Lwanyo Dam, Tanzania

The ongoing generation, transportation, and deposition of silt in the Lwanyo Dam has significantly reduced the storage capacity of the Lwanyo Reservoir, originally constructed to support irrigation and the surrounding ecosystem. The objective of this paper was to assess the extent of siltation in Lwanyo Dam, evaluate its impact on the dam's storage capacity, and propose measures to mitigate silt accumulation. The upstream catchment area, approximately 39.6 km², includes around 128,991 m² allocated for rain-fed crop cultivation and 5.89 km² for pastoral activities. Frequent overtopping of the reservoir has been observed, largely due to siltation reducing its live storage capacity. In the reservoir trial pits were excavated and assessed, and they indicate that average silt layers range in thickness from 0.54 m to 0.98 m per rainy season. The deposited material consists of a silt layer from 0 to 540 mm, followed by an intermediate clay layer from 540 mm to 3100 mm. The impounded silt depth was measured at 1270 mm, with an estimated siltation volume of 58,349.4644 m³. The reservoir's original storage capacity of 210,153 m³ has been reduced by 27.765% due to siltation. The reservoir’s structural design inadequately addresses silt management, lacking both silt flushing tunnels and upstream silt check dams. The analysis indicates that storage capacity decreases by 3.085% annually, and if this linear trend continues without any intervention measures, the dam will lose all storage capacity within 24 years. The study recommends urgent measures to mitigate silt accumulation.

Design and investigation of indium tin oxide-based transparent broadband microwave absorbers

This paper presents the design of a novel wideband transparent microwave absorber based on indium tin oxide (ITO). The absorber adopts a multilayer structure, with the top layer composed of a nested square structure made of ITO material, an intermediate dielectric layer of polyvinyl chloride (PVC), and a bottom layer of polyethylene terephthalate (PET) coated with ITO as the reflector. By optimizing structural parameters such as the thickness of the ITO film and the dimensions of the spacer layer, efficient broadband microwave absorption exceeding 90% is achieved within the frequency range of 2.43 to 5.67 GHz, while simultaneously meeting requirements for optical transparency and broadband performance. Simulation experiments and theoretical analyses demonstrate excellent polarization and incident angle stability of the absorber structure. Finally, samples are fabricated using screen printing technology, and free-space testing results are consistent with simulation predictions, validating the feasibility of the design.

Concept of Artificial Intelligence-oriented Public Health Model in Cancer Care

Abstract In recent years, the escalating volume of essential information for oncologists has created a challenge, making it arduous to stay abreast of the latest developments in the multifaceted field of cancer care. Although Artificial Intelligence (AI) is increasingly applied in healthcare, particularly for tasks like image recognition and big data analysis, we advocate for an AI-centric public health model tailored to comprehensive cancer care. This model aims to guide patients from their initial doctor’s visit to the conclusion of treatment, thereby minimizing direct doctor involvement. Results. The proposed AI system comprises distinct units: Regional AI (RAI) for patient management and coordination with healthcare specialists and facilities in specific areas, General AI (GAI) to oversee healthcare processes on a broader scale, and Scientific AI (SAI) for data analysis and hypothesis generation, essential for scientific research and clinical trials. To enhance cost efficiency, we suggest introducing an intermediate layer, Teacher AI (TAI), facilitating the development of AI systems like GAI or RAI based on human needs without necessitating extensive specialist intervention. Conclusions. Implementing this model can simplify oncologists’ daily tasks, reduce errors, improve treatment outcomes, and lower the cost of cancer care while maintaining its high quality. The Human–TAI–AI development model can streamline the system’s development and implementation, making it more cost-effective.

Superior role of V2O5 and yttrium interface layers in enhancing MIS radical photodiode performance

In this study, Cu/n-Si Schottky diodes with V2O5–Y insulating interface layer were designed for the fabrication of metal/insulator/semiconductor (MIS) structures. The effects of different Y–V2O5 interface layers placed between metals and semiconductors on the critical electrical parameters of Schottky diodes were examined. Electrical parameters of the prepared sample such as barrier height (ΦB), ideality factor (n), photosensitivity (Ps), photosensitivity (R), quantum efficiency (QE) and detectivity (D∗) were obtained from the expression. Characteristics of I–V The modification in these parameters can be attributed to the use of Schottky diodes as intermediate layers. Additionally, n, ΦB values were calculated using the thermionic emission process and the photodiode parameters Ps, R, QE and D∗ the results were compared with each other. Experiments have shown that the Schottky structure with V2O5–Y interlayer leads to higher values of Ps = 799.70 % and under the light condition of 6 wt% of Y lower value n = 2.01. This provides evidence of the improved performance of MIS models.

Ionic surfactants critical micelle concentration prediction in water/organic solvent mixtures by artificial neural network

Abstract Critical micellar concentration (CMC) is a key physicochemical property of surfactants used to study their behaviour. This property is affected by factors such as temperature, pressure, pH, the type of organic solvent/water mixture, the chemical structure of the surfactants and the presence of electrolytes. Most of the existing studies in the literature have predicted the CMC under fixed conditions based on the chemical parameters of the surfactant. In this study, a machine learning approach using artificial neural network (ANN) models was used to estimate the CMC of some ionic surfactants. These models considered variables defining both the organic solvent-water mixture (T, molecular weight, molar fraction and log P) and the chemical structure of the surfactant (number of atoms of each element). A database consisting of a total of 258 CMC values for 10 ionic surfactants was collected from the literature. The ANN architecture consisting of an input layer with 12 neurons, an intermediate layer with 25 neurons and one neuron in the output layer is proposed. According to the results, the normalized ANN models provided the best statistical adjustments for the CMC prediction. These ANN models could be a promising method for CMC estimation.

Expression of Programmed Death Ligand 1 and Indoleamine 2,3-Dioxygenase in Oral Lichen Planus and Oral Lichenoid Lesions.

Oral lichen planus (OLP) and oral lichenoid lesions (OLL) are inflammatory T-cell mediated disorders of the oral mucosa (OM). Both are associated with an increased risk of oral squamous cell carcinoma, with OLL possibly having a higher rate of malignant transformation than OLP. Programmed death ligand 1 (PD-L1) and indoleamine 2,3-dioxygenase (IDO) are immunosuppressive molecules possessing inhibitory effect on T-cells and have been implicated in carcinogenesis. The aim of this study was to examine the expression of PD-L1 and IDO in OLP and OLL. Sixty-eight formalin-fixed, paraffin-embedded tissue samples diagnosed as OLP, compatible with OLP, or OLL were divided into OLP (n = 39) or OLL (n = 29) groups based on both clinical and histopathological diagnostic criteria. Samples of healthy OM (n = 9) served as controls. Samples were immunohistochemically stained for PD-L1 and IDO, and staining distribution and intensity were evaluated. Immunohistochemical expression of PD-L1 was increased in the basal and intermediate layers of epithelium in OLP and in lamina propria in both OLP and OLL compared to controls. OLP and OLL showed increased expression of IDO in epithelium and lamina propria compared to controls. PD-L1 staining intensity in the basal epithelial layer, and IDO staining intensity in lamina propria were increased in OLP compared to OLL. The results indicate that the expression of PD-L1 and IDO increases in OLP and OLL, suggesting that these molecules may play a role in the pathogenesis of both disorders.

Visual search and real-image similarity: An empirical assessment through the lens of deep learning.

The ability to predict how efficiently a person finds an object in the environment is a crucial goal of attention research. Central to this issue are the similarity principles initially proposed by Duncan and Humphreys, which outline how the similarity between target and distractor objects (TD) and between distractor objects themselves (DD) affect search efficiency. However, the search principles lack direct quantitative support from an ecological perspective, being a summary approximation of a wide range of lab-based results poorly generalisable to real-world scenarios. This study exploits deep convolutional neural networks to predict human search efficiency from computational estimates of similarity between objects populating, potentially, any visual scene. Our results provide ecological evidence supporting the similarity principles: search performance continuously varies across tasks and conditions and improves with decreasing TD similarity and increasing DD similarity. Furthermore, our results reveal a crucial dissociation: TD and DD similarities mainly operate at two distinct layers of the network: DD similarity at the intermediate layers of coarse object features and TD similarity at the final layers of complex features used for classification. This suggests that these different similarities exert their major effects at two distinct perceptual levels and demonstrates our methodology's potential to offer insights into the depth of visual processing on which the search relies. By combining computational techniques with visual search principles, this approach aligns with modern trends in other research areas and fulfils longstanding demands for more ecologically valid research in the field of visual search.

Enhancement of photovoltaic parameters of thermally stable graphene/LaVO3 semitransparent solar cell by employing interfacial graphene quantum dots

Semitransparent solar cells are attracting attention not only for their visual effects but also for their ability to effectively utilize solar energy. Here, we demonstrate a translucent solar cell composed of bis(trifluoromethane sulfonyl)-amide (TFSA)-doped graphene (Gr), graphene quantum dots (GQDs), and LaVO3. By introducing a GQDs intermediate layer at the TFSA-Gr/LaVO3interface, we can improve efficiency by preventing carrier recombination and promoting charge collection/separation in the device. As a result, the efficiency of the GQDs-based solar cell was 4.35%, which was higher than the 3.52% of the device without GQDs. Furthermore, the average visible transmittance of the device is 28%, making it suitable for translucent solar cells. The Al reflective mirror-based system improved the power conversion efficiency by approximately 7% compared to a device without a mirror. Additionally, the thermal stability of the device remains at 90% even after 2000 h under an environment with a temperature of 60 °C and 40% relative humidity. These results suggest that TFSA-Gr/GQDs/LaVO3-based cells have a high potential for practical use as a next-generation translucent solar energy power source.

A clustered federated learning framework for collaborative fault diagnosis of wind turbines

Data-driven approaches demonstrate significant potential in accurately diagnosing faults in wind turbines. To enhance diagnostic performance and reduce communication costs in federated learning with data heterogeneity among different clients, we introduce a clustered federated learning framework to wind turbine fault diagnosis. Initially, a lightweight multiscale separable residual network (LMSRN) model is proposed for each local client. The LMSRN model integrates a multiscale spatial feature derivation unit and a depthwise separable feature extraction unit. Subsequently, to tackle data heterogeneity among clients, canonical correlation coefficients of representations are extracted from the intermediate layers of local LMSRN models, and a representational canonical correlation clustering (RCCC) method is proposed to assess the similarity of local LMSRN models and group them into clusters. Finally, a global model is trained for each cluster. Real-world wind turbine data experiments showcase the superior performance of the proposed clustered federated learning framework over traditional methods in terms of diagnostic accuracy and computational speed. Additionally, the optimal choice of the number of clusters is also discussed.