Desirable Performance Research Articles

Retinex Image Enhancement Based on Sequential Decomposition With a Plug-and-Play Framework.

The Retinex model is one of the most representative and effective methods for low-light image enhancement. However, the Retinex model does not explicitly tackle the noise problem and shows unsatisfactory enhancing results. In recent years, due to the excellent performance, deep learning models have been widely used in low-light image enhancement. However, these methods have two limitations. First, the desirable performance can only be achieved by deep learning when a large number of labeled data are available. However, it is not easy to curate massive low-/normal-light paired data. Second, deep learning is notoriously a black-box model. It is difficult to explain their inner working mechanism and understand their behaviors. In this article, using a sequential Retinex decomposition strategy, we design a plug-and-play framework based on the Retinex theory for simultaneous image enhancement and noise removal. Meanwhile, we develop a convolutional neural network-based (CNN-based) denoiser into our proposed plug-and-play framework to generate a reflectance component. The final image is enhanced by integrating the illumination and reflectance with gamma correction. The proposed plug-and-play framework can facilitate both post hoc and ad hoc interpretability. Extensive experiments on different datasets demonstrate that our framework outcompetes the state-of-the-art methods in both image enhancement and denoising.

Distributed Dynamic Event-Triggered Control for Voltage Recovery in Islanded Microgrids by Using Artificial Delays.

This article tackles secondary voltage recovery problem in islanded microgrids with the aim of reducing communication frequency among distributed generation (DG) units, while maintaining desired performance and saving communication network workload. To pursue this objective, a distributed proportional-integral-derivative controller is first introduced, whose sampled-data implementation is enabled by leveraging the finite-difference approximation for the derivative action, which leads to a distributed proportional-integral-retarded (PIR) controller with a small enough sampling period . Then, the resulting fully distributed PIR control law is combined with a dynamic event-triggered mechanism (DETM), which embeds Zeno-freeness property and avoids the requirement of continuous transmission in triggering process. Thus, the communication burden is significantly mitigated and the waste of communication resources is avoided. By exploiting Lyapunov-Krasovkii method, we derive exponential stability conditions expressed as linear matrix inequalities (LMIs), whose solution allows evaluating the maximum sampling period and DETM parameters preserving the stability of the microgrid. A thorough numerical analysis, carried out on the standard IEEE 14-bus test system, confirms the theoretical derivation.

Reweighted Regularized Prototypical Network for Few-Shot Fault Diagnosis.

In this article, we study the challenging few-shot fault diagnosis (FSFD) problem where limited faulty samples are available. Metric-based meta-learning methods have been a prevalent approach toward FSFD; however, most of them rely on learning a generalized distance metric and fall short of leveraging intraclass and interclass distribution information. To this end, we develop a novel reweighted regularized prototypical network to improve the performance of FSFD, where an intraclass reweighting strategy is proposed to reduce the influence of noise and outliers and obtain stable estimations of fault prototypes. In addition, a novel balance-enforcing regularization (BER) is proposed to hedge against the between-class imbalance and improve the discrimination capability. These two remedies help to reduce the intraclass difference and enlarge the interclass difference via episodic training. In this way, an improved metric space and a better diagnostic performance can be attained in a few-shot learning context. Case studies on the Tennessee Eastman benchmark process and a real-world railway turnout dataset demonstrate that the proposed FSFD approach compares favorably against state-of-the-art methodologies with desirable diagnostic performance.

Error Loss Networks.

A novel model called error loss network (ELN) is proposed to build an error loss function for supervised learning. The ELN is similar in structure to a radial basis function (RBF) neural network, but its input is an error sample and output is a loss corresponding to that error sample. That means the nonlinear input-output mapper of the ELN creates an error loss function. The proposed ELN provides a unified model for a large class of error loss functions, which includes some information-theoretic learning (ITL) loss functions as special cases. The activation function, weight parameters, and network size of the ELN can be predetermined or learned from the error samples. On this basis, we propose a new machine learning paradigm where the learning process is divided into two stages: first, learning a loss function using an ELN; second, using the learned loss function to continue to perform the learning. Experimental results are presented to demonstrate the desirable performance of the new method.

BERT4ST:: Fine-tuning pre-trained large language model for wind power forecasting

Accurate forecasting of wind power generation is essential for ensuring power safety, scheduling various energy sources, and improving energy utilization. However, the elusive nature of wind, influenced by various meteorological and geographical factors, greatly complicates the wind power forecasting task. To improve the forecasting accuracy of wind power (WP), we propose a BERT-based model for spatio-temporal forecasting (BERT4ST), which is the first approach to fine-tune a large language model for the spatio-temporal modeling of WP. To deal with the inherent characteristics of WP, BERT4ST exploits the individual spatial and temporal dependency of patches and redesigns a set of spatial and temporal encodings. By well analyzing the connection between bidirectional attention networks and WP spatio-temporal data, BERT4ST employs a pre-trained BERT encoder as the backbone network to learn the individual spatial and temporal dependency of patches of WP data. Additionally, BERT4ST fine-tunes the pre-trained backbone in a multi-stage manner, i.e., first aligning the language model with the spatio-temporal data and then fine-tuning the downstream tasks while maintaining the stability of the backbone network. Experimental results demonstrate that our BERT4ST achieves desirable performance compared to some state-of-the-art methods.

Optimization of the Thermally Conductive Low-k Polymer Dielectrics Based on Multisource Free-Volume Effects.

Polymers/polymer matrix composites possessing low dielectric constants (low-k polymer dielectrics) contribute to the advance of electronics, for instance, microprocessor chips, mobile phone antennas, and data communication terminals. However, the intrinsic long-chain structural characteristic results in poor thermal conductivities, which draw heat accumulation and undermine the outstanding low-k performance of polymers. Herein, multisource free-volume effects that combine two novel kinds of extra free volume with the known in-cage free volume of polyhedral oligomeric silsesquioxanes (POSSs) are discussed to reduce the capacity for dielectric constant reduction. The multisource free-volume effects of POSSs are associated with the thermal conductive network formed by the hexagonal boron nitride (BN) in the polymer matrix. The results show a decent balance between low-k performance (dielectric constant is 2.08 at 1 MHz and 1.98 at 10 GHz) and thermal conductivity (0.555 W m-1 K-1, 4.91 times the matrix). The results provide a new idea to maximize the free-volume effects of POSSs to optimize dielectric properties together with other desired performances for the dielectrics.

Catalytic Activity Prediction of α-Diimino Nickel Precatalysts toward Ethylene Polymerization by Machine Learning

The present study explored machine learning methods to predict the catalytic activities of a dataset of 165 α-diimino nickel complexes in ethylene polymerization. Using 25 descriptors as the inputs, the XGBoost model presented the optimal performance among six different algorithms (R2 = 0.999, Rt2 = 0.921, Q2 = 0.561). The results of the analysis indicate that high activity is related to the presence of polarizable atoms and less bulky substituents within the N-aryl group. This approach offers valuable insights on the variation principle of catalytic activity as a function of complex structure, helping to effectively design and optimize α-diimino Ni catalysts with desirable performance.

Design of a Microwave Quadrature Hybrid Coupler with Harmonic Suppression Using Artificial Neural Networks

In this paper, a compact and simple structure of an elliptic microstrip lowpass filter (LPF) is designed for harmonic suppression in microwave quadrature hybrid coupler (QHC) applications. A radial resonator and a rectangular resonator are used to produce an elliptic LPF. The proposed LPF is used on the outer sides of the branch line coupler, which has improved the coupler harmonic suppression. Furthermore, artificial neural networks (ANNs) are incorporated to improve the LPF design process. The LPF best structure is obtained using the proposed ANN model. The proposed LPF has a compact size, which only occupies 16.4 mm × 7.3 mm equals to 0.164 λg × 0.073 λg, has a cut frequency of 2.2 GHz, and shows a sharp transmission band with a roll-off rate of 158.3 dB/GHz. Finally, the deigned QHC operates correctly at 1 GHz, which shows high harmonic suppression ability. The proposed QHC provides wide suppression band from 2.25 GHz up to more than 14 GHz, which can effectively suppress 3rd, to 14th harmonics. The proposed coupler features desirable parameters of S11, S21, S31, and S41, with magnitude of −21 dB, −3.4 dB, −3.3 dB, and −22.5 dB, at the operating frequency. The proposed approach mitigates the complexity of the circuit fabrication, compared with the previous methods while achieved desirable performances for the proposed QHC.

Stable and ultralong room-temperature phosphorescent copolymers with excellent adhesion, resistance, and toughness.

Developing stable room-temperature phosphorescence (RTP) emission without being affected by moisture and mechanical force remains a great challenge for purely organic systems, due to their triplet states sensitive to the infinitesimal motion of phosphors and the oxygen quencher. We report a kind of highly robust phosphorescent systems, by doping a rigid phosphor into a copolymer (polyvinyl butyral resin) matrix with a balance of mutually exclusive features, including a rigidly hydrophilic hydrogen bond network and elastically hydrophobic constituent. Impressively, these RTP polymeric films have superior adhesive ability on various surfaces and showed reversible photoactivated RTP with lifetimes up to 5.82 seconds, which can be used as in situ modulated anticounterfeit labels. They can maintain a bright afterglow for over 25.0 seconds under various practical conditions, such as storage in refrigerators, soaking in natural water for a month, or even being subjected to strong collisions and impacts. These findings provide deep insights for developing stable ultralong RTP materials with desirable comprehensive performance.

Manipulating CNT Films with Atomic Precision for Absorption Effectiveness–Enhanced Electromagnetic Interference Shielding and Adaptive Infrared Camouflage

AbstractPromoting advanced functional films that integrate protective behaviors for synergetic radar, infrared (IR), and visible light is confronted with extreme challenges. The effective integration of shielding capacity and IR camouflage into single‐component carbon films remains an enormous challenge because of limited conductivity and high IR absorption. In this work, carbon nanotube (CNT) films with excellent electromagnetic interference (EMI) shielding, switchable IR camouflage, and energy conversion are obtained via a (FCCVD) method and subsequent defect engineering strategies induced by temperature (Strategy I) and N/S co‐doping (Strategy II). Wherein the effect on the electronic configuration as well as multi‐spectra performance is deeply explored. Ascribing to the synergistic effects of strategies, defective CNT films present superior EMI shielding effectiveness (SE) with a high absorption effectiveness ratio of 86.9%, and a large span of emissivity (0.479), which provides the necessary condition for various application scenarios. Under the applied electric field excitation, the IR radiation of CNT films can adapt to rapidly changing surroundings, such as low‐high/high‐low temperatures. Besides, the desirable energy conversion performance and de‐icing can be achieved. This work supplies a feasible strategy for designing EM absorption, adaptive IR camouflage, and energy conversion to confront multiband surveillance.

Bioinspired Photoactive Cu-Halide Coordination Polymers for Reduction Activation and Oxygen Conversion.

Natural copper oxygenases provide fundamental principles for catalytic oxidation with kinetically inert molecular oxygen, but it remains a marked challenge to mimic both their structure and function in an entity. Inspired by the CuA enzymatic sites, herein we report two new photoactive binuclear copper-iodine- and bisbenzimidazole-comodified coordination polymers to reproduce the natural oxo-functionalization repertoire in a unique photocatalytic pathway. Under light irradiation, the Cu-halide coordination polymers effectively reduce NHP esters and complete oxygen reduction activation via photoinduced electron transfer for the aerobic oxidative coupling of hydroquinone with terminal alkynes, affording hydroxyl-functionalized ketones with high efficiency and selectivity. This supramolecular approach to developing bioinspired artificial oxygenases that merge transition metal- and photocatalysis supplies a new way to fabricate distinctive photocatalysts with desirable catalytic performances and controllable precise active sites.

From CNN to Transformer: A Review of Medical Image Segmentation Models.

Medical image segmentation is an important step in medical image analysis, especially as a crucial prerequisite for efficient disease diagnosis and treatment. The use of deep learning for image segmentation has become a prevalent trend. The widely adopted approach currently is U-Net and its variants. Moreover, with the remarkable success of pre-trained models in natural language processing tasks, transformer-based models like TransUNet have achieved desirable performance on multiple medical image segmentation datasets. Recently, the Segment Anything Model (SAM) and its variants have also been attempted for medical image segmentation. In this paper, we conduct a survey of the most representative seven medical image segmentation models in recent years. We theoretically analyze the characteristics of these models and quantitatively evaluate their performance on Tuberculosis Chest X-rays, Ovarian Tumors, and Liver Segmentation datasets. Finally, we discuss the main challenges and future trends in medical image segmentation. Our work can assist researchers in the related field to quickly establish medical segmentation models tailored to specific regions.

Recent Advances in Implantable Neural Interfaces for Multimodal Electrical Neuromodulation.

Electrical neuromodulation plays a pivotal role in enhancing patient outcomes among individuals suffering from neurological disorders. Implantable neural interfaces are vital components of the electrical neuromodulation system to ensure desirable performance; However, conventional devices are limited to a single function and are constructed with bulky and rigid materials, which often leads to mechanical incompatibility with soft tissue and an inability to adapt to the dynamic and complex 3D structures of biological systems. In addition, current implantable neural interfaces utilized in clinical settings primarily rely on wire-based techniques, which are associated with complications such as increased risk of infection, limited positioning options, and movement restrictions. Here, the state-of-art applications of electrical neuromodulation are presented. Material schemes and device structures that can be employed to develop robust and multifunctional neural interfaces, including flexibility, stretchability, biodegradability, self-healing, self-rolling, or morphing are discussed. Furthermore, multimodal wireless neuromodulation techniques, including optoelectronics, mechano-electrics, magnetoelectrics, inductive coupling, and electrochemically based self-powered devices are reviewed. In the end, future perspectives are given.

Secure control for remote networked stochastic systems via integral sliding mode

This paper deals with the secure control problem for a class of networked stochastic systems with false data injection attacks via an integral sliding mode control technique. The networked control system is with a hierarchical structure, and the main controller and a remote controller are considered to realize the secure control against false data injection attacks on the network between a main controller and the plant. A mode-shared event-triggering controller is designed as the main controller, by utilizing a time delay approach. An input–output model based on a two-term approximation is applied to cope with the formulated time-varying delay. Then, the scaled small gain theory is employed to analyze the stability of the resulting system. Sufficient conditions on ensuring the desired system performance are derived and then the controller parameters are synthesized. Moreover, an elaborated sliding mode control law is proposed for the desired secure control action. Finally, two simulation examples are permitted to verify the effectiveness of the theoretical derivations.

Efficient and exquisite concurrent optimization of hierarchical structures with non-uniform eccentric body centered cubic lattice

To obtain lightweight structures with desirable mechanical performances and to enlarge the design space, this work presents an innovative concurrent optimization methodology for the precise inverse design of micro lattices and the modeling of hierarchical structures. The inverse design of micro lattices with continuously changing anisotropic properties is achieved by offsetting the center nodes of eccentric body centered cubic lattices (EBCC) to designated locations. Simultaneously, the macro material distribution is adjusted utilizing 3D density-based topology optimization method. To enhance the computation efficiency and significantly amplify the versatility in optimization design, a Parameterized Interpolation for Anisotropic Lattice material (PIAL) is established based on elastic stiffness matrices of micro lattices with 21 independent components. Specifically, a cubic multinomial function is employed to construct a direct correlation between micro topology configurations and equivalent elasticities of the EBCC lattices. Several numerical examples of compliance optimization are carried out to confirm the validity of the proposed approach. Concurrently, Additive Manufacturing (AM) using AlSi10Mg is adopted to manufacture the obtained hierarchical structures, which exhibit well-improved mechanical performance, as confirmed by experimental testing, when compared to the classical density-graded structures.

Meliorative dielectric properties in core@double-shell structured Al@Al2O3@PDA/PVDF nanocomposites via decoupling the intra-particle polarization and inter-particle polarization

Percolating polymeric composites present enormous potential owing to high dielectric constant (ε) which can be realized near the percolation threshold, but the accompanied large loss forbids their extensive use in practice. Great efforts have been devoted to coat conductive particles with an insulating shell for constrained dielectric loss, yet they markedly reduce ε. In this work, we explore poly(vinylidene fluoride) (PVDF) composites with a serial of core@double-shell Al@Al2O3@PDA (polydopamine) nanoparticles with various PDA shell thicknesses. It reveals that the high ε of the nanocomposites results from a fast intra-particle polarization and a slow inter-particle polarization. The formation of double-shell enables the independent control of the two polarizations always coupled in traditional percolating composites. Through facilitating intra-particle polarization and repressing inter-particle polarization, Al@Al2O3@PDA/PVDF can achieve a much higher ε and lower dielectric loss simultaneously, far exceeding the unmodified Al@Al2O3/PVDF. Moreover, the calculated activation energy of carrier migration in Al@Al2O3@PDA/PVDF is obviously higher than that in untreated nanocomposites, indicating enhanced charge-trapping capability in the core@double-shell nanofiller composites. This core@double-shell strategy offers a new paradigm for the design and preparation of percolating composites with desirable dielectric performances.

Analogous Confinement Effect Enables High Stability and High Capacity Ammonium Storage in Polyaniline@Poly(o-fluoroaniline)@Carbon Layer.

Rechargeable aqueous ammonium ion (NH4 +) batteries have attracted much attention due to the unique properties of NH4 +. Polyaniline (PA) with outstanding conductivity is a potential cathode material, but it can be oxidized to pernigraniline (PG) rapidly, resulting in its poor stability. In this study, polyaniline@poly(o-fluoroaniline)@carbon layer (PA@POFA@C) is prepared for excellent and durable NH4 + storage. PA@POFA@C exhibits a high capacity of 208 mAh g-1 at 0.2 A g-1 and maintains 126 mAh g-1 at 10 A g-1. More importantly, an excellent capacity retention rate of 88.24% is achieved after 2000 cycles with ≈100% coulombic efficiency. Spectroscopy studies suggest analogous confinement effect can effectively limit the escape of hydrogen in imine group, and form the hydrogen-restricted region between the PA and POFA layer which can provide H+ for the complete reduction of PG. Meanwhile, the hydrophobic effect of POFA effectively restrains the hydrolysis of PG. Interestingly, the introduction of C layer improves the hydrophilicity of electrode and shortens the activation process, serving as the outermost protective layer of the electrode. Finally, PA@POFA@C achieves desirable electrochemical performances with analogous confinement effect. This research provides ideas for the preparation of advanced polymer electrodes for aqueous NH4 + batteries.

Event-triggered adaptive secure tracking control for nonlinear cyber–physical systems against unknown deception attacks

This article presents an event-triggered adaptive neural networks secure tracking control method for a class of nonlinear cyber–physical systems under unknown sensor and actuator deception attacks. To obtain the desired system performance, dynamic surface technique is applied to design controller and radial basis function neural networks are introduced to deal with unknown nonlinear and actuator attacks. By skillfully combining compensation signals with the attack compensators, the unknown deception attacks are effectively mitigated. To reduce the transmission communication load, a novel event-driven control scheme is developed by applying relative threshold triggered mechanism. On this basis, all the signals of closed-loop system are bounded under deception attacks by using Lyapunov stability analysis. Additionally, the presented secure tracking control strategy can ensure the tracking error converges to a small neighborhood of origin and the Zeno behavior is ruled out. Finally, a practical example is employed to verify the feasibility and effectiveness of the designed control algorithm.

Rapid Fabrication of Electrodes for Symmetrical Solid Oxide Cells by Extreme Heat Treatment

Symmetrical solid oxide cells (SSOCs) are very useful for energy generation and conversion. To fabricate the electrode of SSOC, it is very time‐consuming to use the conventional approach. In this work, we design and develop a novel method, extreme heat treatment (EHT), to rapidly fabricate electrodes for SSOC. We show that by using the EHT method, the electrode can be fabricated in seconds (the fastest method to date), benefiting from enhanced reaction kinetics. The EHT‐fabricated electrode presents a porous structure and good adhesion with the electrolyte. In contrast, tens of hours are needed to prepare the electrode by the conventional approach, and the prepared electrode exhibits a dense structure with a larger particle size due to the lengthy treatment. The EHT‐fabricated electrode shows desirable electrochemical performance. Moreover, we show that the electrocatalytic activity of the perovskite electrode can be tuned by the vigorous approach of fast exsolution, deriving from the increased active sites for enhancing the electrochemical reactions. At 900 °C, a promising peak power density of 966 mW cm−2 is reached. Our work exploits a new territory to fabricate and develop advanced electrodes for SSOCs in a rapid and high‐throughput manner.

Maximum output discrepancy computation for convolutional neural network compression

Network compression methods minimize the number of network parameters and computation costs while maintaining desired network performance. However, the safety assurance of many compression methods is based on a large amount of experimental data, whereas unforeseen incidents beyond the experiment data may result in unsafe consequences. In this work, we developed a discrepancy computation method for two convolutional neural networks by giving a concrete value to characterize the maximum output difference between the two networks after compression. Using Imagestar-based reachability analysis, we propose a novel method to merge the two networks to compute the difference. We illustrate reachability computation for each layer in the merged network, such as the convolution, max pooling, fully connected, and ReLU layers. We apply our method to a numerical example to prove its correctness. Furthermore, we implement our developed methods on the VGG16 model with the Quantization Aware Training (QAT) compression method; the results show that our approach can efficiently compute the accurate maximum output discrepancy between the original neural network and the compressed neural network.