Search

Considering the weakness of the conventional physical model of power diodes in the rectifier design, an improved lumped-charge model is presented in this paper to achieve both high accuracy and high efficiency in the simulation of power electronic equipment. To improve the convergence of the physical model simulation in the rectifier circuit, the transmission-line-theory is implemented in the model. Furthermore, a dynamic partition method is implemented in the model to achieve high accuracy of the transient characteristics of the diode during the level transition of the rectifier, with reducing the number of numerical solutions through reallocation of computing resources. Compared with the conventional model, the simulation error is decreased. This successfully adopts the physical model to realize both equipment-level simulation and high accuracy of device simulation in the design of the rectifier and RC snubber circuit. The experimental results show that the error of the proposed model is less than 10% and 5% the minimum, which is a maximum 30% decreasing compared with that of the conventional model. The half-cycle (500μs) simulation of the rectifier circuit with 10 and 20 power diodes cost 18s and 33s.

Aiming to efficiently capture the formaldehyde (HCHO) with low content in the air exceeding the standard, 31,399 hydrophobic metal–organic frameworks (MOFs) were first selected from 137,953 hypothetical MOFs to calculate their formaldehyde adsorption performance, namely, adsorption capacity (NHCHO) and selectivity ( S HCHO / N 2 + O 2 ) by molecular simulation and machine learning (ML). To combine the S HCHO / N 2 + O 2 and NHCHO, a new performance metric, the tradeoff between selectivity and capacity (TSC) was proposed to identify more reasonably the top-performing MOFs. The MOFs were divided into three datasets (i.e., all of the MOFs (AM), MOFs with top 5% of S HCHO / N 2 + O 2 (PS) and MOFs with top 5% of NHCHO (PN)) to scrutinize and explore the characteristics of different materials capturing formaldehyde from the air (N2 and O2). Furthermore, after four ML algorithms (the back propagation neural network (BPNN), support vector machine (SVM), extreme learning machine (ELM), and random forest (RF)) are applied to quantitatively assess the prediction effects of performance indexes in different datasets, RF algorithm with the most accurate prediction revealed that the TSC has strong correlations with the MOF descriptors in PS dataset. In view of 14.10% of the promising MOFs occupied PN, the design paths of excellent adsorbents for six MOF descriptors were quantitatively determined, especially for the Henry's coefficient (KHCHO) and heat of adsorption of formaldehyde (Q0st). Their probabilities of obtaining excellent MOFs could reach 100% and 77.42%, respectively, and both the relative importance and the trends of univariate analysis coherently confirm the important positions of KHCHO and Q0st. Finally, 20 best MOFs were identified for the single-step separation of formaldehyde with low concentration. The microscopic insights and structure-performance relationship predictions from this computational and ML study are useful toward the development of new MOFs for the capture of formaldehyde from air.

Image forgery detection focuses more on tampering regions than image content of semantic segmentation, it is revealed that wealthier features need to be learned. Moreover, insufficient semantic information causes low efficiency of forgery detection. To address these issues, we propose a hybrid features and semantic reinforcement network (HFSRNet) for image forgery detection, which is an encoding and decoding based network. Specifically, long-short term memory with resampling features has been applied to capture traces from the image patches for finding manipulating artifacts. Consolidated features extracted from rotating residual units are further leveraged to amplify the discrepancy between un-tampered and tampered regions. We then hybridize features from them through a concatenation to further incorporate spatial co-occurrence of these two modalities. In addition, for achieving the semantic consistency between two same level features associated by across layers, semantic reinforcement is implemented on the decoding stage. HFSRNet is an end-to-end architecture that handles multiple types of image forgery including copy-move, splicing, removal. Experiments on three standard image manipulation datasets (NIST16, COVERAGE and CASIA) demonstrate that HFSRNet obtains state-of-the-art performance compared to existing models and baselines.

Hybrid dc circuit breakers (HCBs) are vital equipment in dc grid systems. However, HCBs are usually costly since massive full-controlled power semiconductors are required to withstand high surge voltage and current stresses during current interrupting. Thyristors are good alternatives in realizing HCBs in view of their low cost and high current carrying capability; however, turning off the thyristor fast and reliably is a challenging task. In this article, a novel HCB topology is proposed, of which the solid-state part is realized by hybrid connection of thyristors and insulated gate bipolar transistor (IGBT) half-bridge submodules. The use of IGBT half-bridges provides a negative voltage across the thyristors to turn them off ; accordingly, thyristors withstand a major of turn- off surge voltage instead of IGBTs. By this means, a low-cost HCB can be established while maintaining a high breaking speed. Besides, fast reclose and rebreak function is provided by the proposed topology. Operation principle and design consideration of the novel HCB topology are analyzed in detail, and the effectiveness of the proposed HCB is verified by both software simulation and downscaled experimental results.

The key to achieving high efficiencies, high performance, and low costs of adsorption heat pumps/chillers (AHPs/ACs) is to choose a suitable adsorbent. A computational screening of 6013 computation-ready experimental metal–organic frameworks (CoRE-MOFs) is performed for methanol–MOF pairs in AHPs/ACs, and 137 953 hypothetical MOFs (hMOFs) are further directionally screened based on the range of optimal CoRE-MOF descriptors. Compared with the blind screening of CoRE-MOFs (∼13%), ∼80% of hMOFs obtained by the directional screening exhibited high performance in AHPs/ACs, proving the effectiveness of directional screening. Then, the techno-economic analysis (Ctotal) of AHPs/ACs for each MOF was performed based on the equipment cost (Cequipment), cycle cost (Ccycle), and material cost (CMOF). Cequipment accounted for the greatest proportion of Ctotal, but the proportion of Cequipment decreased and the proportion of Ccycle gradually increased as the MOF data set tended to be the materials with better performance. This confirmed the reduction of Ctotal by an effective enhancement of the Ccycle proportion, and that the high manufacturing cost of MOFs would not restrict their widespread use in AHPs/ACs. Among 6 machine learning (ML) algorithms, the random forest could yield the best prediction effect with little effect by the metal type of MOFs, whereas the heat of adsorption and MOF density were two key descriptors to determine Ctotal. 12 lowest-cost CoRE-MOFs and hMOFs were identified for each application, and their average costs in state 2 were only ∼1 USD kJ−1 in AHPs/ACs. Finally, a variant of the well-known MOF (Cu3BTC2) was predicted to possess superior techno-economy, which was confirmed by a parallel methanol adsorption experiment. These comprehensive insights from directional computational screening, ML algorithms to experiment can guide the development of low-cost and high-performance MOFs for AHPs/ACs in a variety of energy conservation and industrial applications.

Over the past two decades, the number of works on metal–organic frameworks (MOFs) in the fields of gas adsorption and separation has experienced explosive growth due to their high void fraction and ultra-high specific surface area. With the rapid increase of MOF databases, high-throughput computational screening (HTCS) has become the main method for selecting high-performance target materials from the large quantity of MOFs. Traditional HTCS methods, e.g. grand canonical Monte Carlo (GCMC) and density functional theory (DFT), could accelerate the discovery of materials; however, there are some shortcomings in these methods such as high computational cost and slow speed, considering the vast and almost infinite MOF database as well as different separation systems and diverse operating conditions. Machine learning (ML) is a potential screening method with the ability to accurately predict the high-performance materials through the training of data, which were obtained by HTCS, and the ML model that fits accurately the complex system can improve the screening speed by 2–3 orders of magnitude. In this work, in view of ML-assisted HTCS of MOFs in recent years, the relevant research progress including CH4 storage, H2 storage, CO2 separation, etc. is summarized, aiming to clarify the potential problems and challenges about ML-assisted HTCS by categorizing the application and development of ML in this field. Then, a series of ML algorithms were designed and developed to adapt to different MOF systems, and to search key descriptors based on ML to reverse design new MOFs with excellent performance. Therefore, the ML-assisted HTCS method could accelerate the development of MOFs and promote their applications in various fields.

This letter focuses on image manipulation detection which aims to recognize the manipulated regions under the contextual semantic information. Existing approaches usually overlook the semantic discrepancy between different levels of feature maps, and directly fuse (e.g., addition, or concatenation) them for detection. In this letter, we argue that the semantic gap is the main reason for the low effectiveness of feature fusion in manipulation predictions. To address this problem, we propose a Global Semantic Consistency Network (GSCNet) for image manipulation detection, which is based on an encoder-decoder structure. Specifically, to make GSCNet include more global texture information which has been empirically confirmed to be beneficial to manipulation detection, gram block is first deployed on each level of feature maps in the encoding stage. Based on that, bi-directional convolutional LSTM is further implemented on the decoding stage, such that feature maps of the same level have semantic consistency. Experimental results on NIST16, and CASIA v1.0 declare that GSCNet can accurately locate the manipulated regions. Furthermore, compared to the existing models, GSCNet can achieve new state-of-the-art results.

Gallium nitride (GaN) HEMT have become popular in high voltage direct current (HVDC) transmission systems due to its fast switching speed and the ability to operate at high frequencies. However, it is generally known that high-speed switching actions will increase the electromagnetic interference (EMI) of power electronic devices and affect the stability of the HVDC system. Hence, this paper proposes an analytical method to predict EMI generated by GaN HEMT during switching process, and quantifies the impact of reverse conduction characteristics on EMI. The results show that the reverse conduction characteristics of GaN HEMTs mainly affect EMI in the high frequency range, and as the reverse conduction voltage increase, EMI becomes more severe. Experimental tests are also performed to verify the analysis results.

The rising level of CO2 in the atmosphere has attracted attention in recent years. The technique of capturing CO2 from higher CO2 concentrations, such as power plants, has been widely studied, but capturing lower concentrations of CO2 directly from the air remains a challenge. This study uses high-throughput computer (Monte Carlo and molecular dynamics simulation) and machine learning (ML) to study 6013 computation-ready, experimental metal-organic frameworks (CoRE-MOFs) for CO2 adsorption and diffusion properties in the air with very low concentrations of CO2. First, the law influencing CO2 adsorption and diffusion in air is obtained as a structure-performance relationship, and then the law influencing the performance of CO2 adsorption and diffusion in air is further explored by four ML algorithms. Random forest (RF) was considered the optimal algorithm for prediction of CO2 selectivity, with an R value of 0.981, and this algorithm was further applied to analyze the relative importance of each metal-organic framework (MOF) descriptor quantitatively. Finally, 14 MOFs with the best properties were successfully screened out, and it was found that a key to capturing a low concentration CO2 from the air was the diffusion performance of CO2 in MOFs. When the pore-limiting diameter (PLD) of a MOF was closer to the CO2 dynamic diameter, this MOF could possess higher CO2 diffusion separation selectivity. This study could provide valuable guidance for the synthesis of new MOFs in experiments that capture directly low concentration CO2 from the air.

It remains a great challenge to separate O2 from N2 at room temperature. Pressure swing adsorption (PSA) technology is a potential candidate, and the development of high-efficiency adsorbents for O2/N2 separation at room temperature has attracted a great deal of interest. In this work, machine learning (ML)-assisted high-throughput computational screening (HTCS) techniques were performed to screen the dynamic adsorption of O2 and N2 in 6,013 computation-ready experimental metal–organic frameworks (CoRE-MOFs), including the competitive adsorption of O2 and the diffusion of pure N2 and O2, to identify the best materials for O2/N2 separation. First, based on HTCS, we established the relationships between the structural/energetic descriptors with the performance indicators. Three machine learning (ML) algorithms were then applied to predict the performance indicators of MOFs. In addition, the relative importance of the structural/energetic descriptors and metal center type in MOFs toward the separation performance was evaluated, indicating that the metal center type of MOFs is a key factor for the separation of O2/N2. Transition metal elements were determined to have highest importance by ML. Moreover, the 13 best MOFs were identified for the dynamic adsorption of O2 from the air. Finally, three types of design strategies could significantly improve the performance of MOFs, such as regulating the topology and alternating the metal node and organic linker. The combination of HTCS, ML, and design strategies from bottom to top provide powerful microscopic insights for the development of MOF adsorbents for the separation of O2 at room temperature.

To improve the poor robustness and low accuracy of the existing algorithms of image splicing detection, a novel passive image forgery detection method is proposed in this study, which is based on DOCT (discrete octonion cosine transform) and Markov. By introducing the octonion and DOCT, the colour information of six image channels (the RGB model and the HSI model) can be exhaustively extracted, which enhances the robustness of the algorithm. On the issue of improving the detection accuracy, the standard deviation is used to characterise the relationship of the colour information between the parts of DOCT coefficient matrix, and the K -fold cross-validation is introduced to improve the identification performance of the classifier. The steps of the algorithm are as follows: Firstly, the 8 × 8 block DOCT transform is used to the original image to obtain parts of block DOCT coefficient. Secondly, the standard deviation is used to process the corresponding parts of all blocks of the image. Finally, the Markov feature vector of the DOCT coefficient is extracted and feds to the LIBSVM (a library for support vector machines). When using LIBSVM for classification, K -fold cross-validation is executed to select the best parameter pairs. The experiment results demonstrate that the algorithm is superior to the other state-of-the-art splicing detection methods.

In multimedia forensics, many efforts have been made to detect whether an image is pristine or manipulated with high enough accuracies based on specially designed features and classifiers in the past decade. However, the important task for localizing the tampering regions in a fake image still faces more challenges compared with the manipulation detection and relatively a few algorithms attempt to tackle it. With this in mind, a technique that utilizes the dual-domain-based convolutional neural networks (D-CNNs) taking different kinds of input into consideration is proposed in this paper. In the proposed framework, two sub-networks, named the spatial-domain CNN model (Sub-SCNN) and the frequency-domain-based CNN model (Sub-FCNN), are designed and trained, respectively. With the well-trained parameters, a transfer policy is applied to the training process of the D-CNN. While CNNs are capable of learning classification features directly from data, in their standard form they tend to learn features related to the image’s content. To overcome this issue in image forensics tasks, a new image pre-processing layer is proposed to jointly suppress image’s content and adaptively learn manipulation detection and localization features. After investigating the properties of datasets, two post-processing operations are finally proposed and compared to obtain the final results of the pixel-wise manipulation region localization. The D-CNNs is trained and validated using 75 percent of images in the CASIA v2.0 and tested using the remaining images in the CASIA v2.0, all images in Columbia Uncompressed and Carvalho datasets. The extensive experiments show that the proposed post-processing operations optimize the final tamper probability map, and our framework with the combination of Sub-SCNN and Sub-FCNN significantly outperforms the state-of-art techniques with the best F1 scores on the datasets.

To further improve the detection rate with relatively low dimension feature vector, a novel passive splicing detection method using textural features based on the grey level co-occurrence matrices, namely TF-GLCM, is proposed in this study. In the TF-GLCM, the GLCM are calculated based on the difference block discrete cosine transform arrays to capture the textural information and the spatial relationship between image pixels sufficiently. The discriminable properties contained in the GLCM are described by six textural features, which include two new introduced ones and four independent ones. In addition, the statistical moments mean Me and standard deviation SD of textural features are used instead of themselves as elements in feature vector to reduce the dimensionality of feature vector and computational complexity. A support vector machine is employed for classification purpose. Experimental results show that the TF-GLCM achieves the detection rates of 98% on CASIA v1.0, and 97% on CASIA v2.0 with 96-D feature vector. The detection rates benefit from the two new textural features. Meanwhile, the TF-GLCM is superior to some state-of-the-art methods with lower dimension feature vector.

With the increasingly serious energy-consumption, the adsorption-driven heat pumps/chillers were paid more attention. To select the promising adsorbents and suitable working fluids in this technology, a high-throughput computational screening of 6013 computation-ready experimental metal organic frameworks (MOFs) is employed to identity the best candidates for methanol-MOF pairs in adsorption-driven heat pumps and chillers. The highest working capacity (ΔW) and coefficient of performance (COP) under operating condition of heat pump are 512.86 mg/g and 1.83, respectively. The structure-performance relationships are derived for ΔW and COP with 6 MOF descriptors, the saturated ΔW is, for the first time, defined for the improvement of heat pump efficiency, and 10 best MOFs are identified. Then four machine learning algorithms are applied to predict the relative importance of each MOF descriptor. The microscopic insights obtained from our bottom-up approach provide the guidelines for the design of new MOFs in adsorption-driven heat pumps and chillers.

Recently, image splicing tamper detection has become an increasingly significant challenge, because of which all color information of color images and low detection rate of existing algorithms cannot be exploited. To overcome the shortcomings of this method, we propose a model, which employs difference matrix in quaternion domain (Q-DIFF) and Markov in quaternion domain (Q-Markov) in the quaternion discrete cosine transform domain (QDCT) for encoding tampering traces and quaternion back propagation neural network (QBPNN) for decision making. Furthermore, by introducing Q-DIFF and Q-Markov in the proposed model, the entire architecture of the algorithm is accumulated in the four-dimensional frequency domain (i.e., all color channels of color images are utilized). Moreover, the experimental results on public domain benchmark datasets demonstrate that the proposed model is superior to the other state-of-the-art splicing detection methods. Based on the experimental results, we suggest the direction that designs image tamper detection model, which invite all the processing in the model to operate in four-dimensional space (i.e. quaternion space).

Aiming at the problem of low accuracy, high computational complexity and incomplete edge information of most image splicing localization algorithm, this paper proposes a new image splicing localization algorithm. First, the SLIC image segmentation algorithm is used to segment the image. Secondly, the noise estimation value of each super-pixel block is calculated by the FAST noise estimation algorithm. Then, weight of each image block is calculated through noise and image features. Finally, the noise value sequence is processed by clustering and statistical processing to determine the pixels of the background area and the splicing area, thus the splicing area is located. In this paper, the algorithm is tested on the color image database of Columbia, and compared with the existing image splicing localization algorithms based on block-segmentation and based on pixel. The experiment shows that the proposed algorithm can preserve the connection between image features, hold the edge of the splicing area, and effectively improve the efficiency of localization detection under the premise of ensuring the accuracy of image splicing localization.

To improve the detection accuracy of spliced images, a new blind detection method based on fuzzy run-length is proposed in this study. Firstly, the edge gradient matrix of a digital image is calculated to obtain the gradient direction of each pixel and then quantified them to the vertical, horizontal, main diagonal, and minor diagonal directions. Then, according to the gradient direction of each pixel, histograms, such as the fuzzy run-length histogram, the fluctuation degree histogram and the fluctuation count histogram, are calculated. After that, the first three moments of the characteristic function of histograms are extracted as features for splicing detection. To further improve the detection accuracy, the fuzzy run-length is applied on original images, predict-error images and reconstructed images to extract a total of 45-D feature vector. Finally, the above features are trained and classified using SVM, by which the spliced images can be identified from the natural ones. The experimental results showed that, when testing on the Columbia image splicing detection dataset, the detection accuracy of the proposed method has improved substantially, which shows good growth prospects.