Search

Molecular representation learning plays a crucial role in AI-assisted drug discovery research. Encoding 3D molecular structures through Euclidean neural networks has become the prevailing method in the geometric deep learning community. However, the equivariance constraints and message passing in Euclidean space may limit the network expressive power. In this work, we propose a Harmonic Molecular Representation learning (HMR) framework, which represents a molecule using the Laplace-Beltrami eigenfunctions of its molecular surface. HMR offers a multi-resolution representation of molecular geometric and chemical features on 2D Riemannian manifold. We also introduce a harmonic message passing method to realize efficient spectral message passing over the surface manifold for better molecular encoding. Our proposed method shows comparable predictive power to current models in small molecule property prediction, and outperforms the state-of-the-art deep learning models for ligand-binding protein pocket classification and the rigid protein docking challenge, demonstrating its versatility in molecular representation learning., 25 pages including Appendix

The reactions of sulfur dioxide (SO2) with surface-bound compounds on atmospheric aerosols lead to the formation of organic sulfur (OS) compounds, thereby affecting the air quality and climate. Here, we show that the heterogeneous reaction of SO2 with authentic urban grime under near-ultraviolet sunlight irradiation leads to a large suite of various organic compounds including OS released in the gas phase. Calculations indicate that at the core area of Guangzhou, building surface uptake of SO2 is 15 times larger than uptake of SO2 on aerosol surfaces, yielding ~20 ng m-3 of OS that represents an important fraction of the observed OS compounds (60 to 200 ng m-3) in ambient aerosols of Chinese megacities. This chemical pathway occurring during daytime can contribute to the observed fraction of OS compounds in aerosols and improve the understanding of haze formation and urban air pollution.

A very recent trend in generative modeling is building 3D-aware generators from 2D image collections. To induce the 3D bias, such models typically rely on volumetric rendering, which is expensive to employ at high resolutions. During the past months, there appeared more than 10 works that address this scaling issue by training a separate 2D decoder to upsample a low-resolution image (or a feature tensor) produced from a pure 3D generator. But this solution comes at a cost: not only does it break multi-view consistency (i.e. shape and texture change when the camera moves), but it also learns the geometry in a low fidelity. In this work, we show that it is possible to obtain a high-resolution 3D generator with SotA image quality by following a completely different route of simply training the model patch-wise. We revisit and improve this optimization scheme in two ways. First, we design a location- and scale-aware discriminator to work on patches of different proportions and spatial positions. Second, we modify the patch sampling strategy based on an annealed beta distribution to stabilize training and accelerate the convergence. The resulted model, named EpiGRAF, is an efficient, high-resolution, pure 3D generator, and we test it on four datasets (two introduced in this work) at $256^2$ and $512^2$ resolutions. It obtains state-of-the-art image quality, high-fidelity geometry and trains ${\approx} 2.5 \times$ faster than the upsampler-based counterparts. Project website: https://universome.github.io/epigraf., Comment: NeurIPS 2022

Limited by the computational efficiency and accuracy, generating complex 3D scenes remains a challenging problem for existing generation networks. In this work, we propose DepthGAN, a novel method of generating depth maps with only semantic layouts as input. First, we introduce a well-designed cascade of transformer blocks as our generator to capture the structural correlations in depth maps, which makes a balance between global feature aggregation and local attention. Meanwhile, we propose a cross-attention fusion module to guide edge preservation efficiently in depth generation, which exploits additional appearance supervision information. Finally, we conduct extensive experiments on the perspective views of the Structured3d panorama dataset and demonstrate that our DepthGAN achieves superior performance both on quantitative results and visual effects in the depth generation task.Furthermore, 3D indoor scenes can be reconstructed by our generated depth maps with reasonable structure and spatial coherency., Comment: 7 pages, 5 figures

Neural rendering can be used to reconstruct implicit representations of shapes without 3D supervision. However, current neural surface reconstruction methods have difficulty learning high-frequency geometry details, so the reconstructed shapes are often over-smoothed. We develop HF-NeuS, a novel method to improve the quality of surface reconstruction in neural rendering. We follow recent work to model surfaces as signed distance functions (SDFs). First, we offer a derivation to analyze the relationship between the SDF, the volume density, the transparency function, and the weighting function used in the volume rendering equation and propose to model transparency as transformed SDF. Second, we observe that attempting to jointly encode high-frequency and low-frequency components in a single SDF leads to unstable optimization. We propose to decompose the SDF into a base function and a displacement function with a coarse-to-fine strategy to gradually increase the high-frequency details. Finally, we design an adaptive optimization strategy that makes the training process focus on improving those regions near the surface where the SDFs have artifacts. Our qualitative and quantitative results show that our method can reconstruct fine-grained surface details and obtain better surface reconstruction quality than the current state of the art. Code available at https://github.com/yiqun-wang/HFS., Comment: To appear in NeurIPS 2022. Project page: https://github.com/yiqun-wang/HFS

With the gradual increase of power grid scale and the gradual complexity of power sources and electric loads, the frequency characteristics of power system are becoming more and more complex. The frequency characteristics are closely related to the electric power grid structure and electrical distances between power sources and electric loads. This paper proposed a transmission planning method concerning power system frequency characteristics. Firstly, mechanism analysis of the space-time distribution of power system frequency is conducted, and the measurement indicators of power system frequency space-time distribution are given. Then the transmission planning method considering frequency space-time distribution characteristics is formulated based on the given measurement indicators. Finally, case studies on a provincial power grid of China are conducted, and the numerical analysis illustrated the effectiveness of the proposed method.

A novel method to make a round end rectangular joint whose total length is much larger than the diameter of the laser spot, named as incremental laser shock clinching, was proposed in this paper. Experimental and numerical investigation was conducted to verify the feasibility for joining three sheets of copper/aluminum/stainless steel. The transverse section and the appearance of the joint were explored to determine the spacer height and the forming strategy. The shape evolution of interlock was discussed by observing sectional images. To describe the material flow characteristics in the forming process, a series of typical points were traced. The mechanical properties of the joints relating to the load directions and clamping ways were studied by tensile shear tests. The results show that fine joints with smooth surface morphology and uniform thickness can be obtained under the larger spacer height and multi-pass forming strategy. Simulated results agree well with the experimental results, which implies that the finite element model can be effectively used to analyze the process. During deformation, most material flows along the transverse direction to form undercuts, meanwhile flows slightly towards the laser spot moving direction. What’s more, the maximum load in the perpendicular shear test is larger than that in the parallel shear test. Two failure modes are observed when the copper foil and pre-slotted sheet are clamped on one side and only one failure mode is found when the copper foil is clamped alone.

Nonvolatile processor (NVP) attracts more and more attentions for its immunity to power loss in energy harvesting scenarios. The overall performance of an NVP is determined by its sleep and wake-up speeds, which refer to how fast the NVP can turn itself off and on when a sudden power failure occurs. To the best of our knowledge, this paper is the first work to improve the system-level sleep/wake-up speed of an energy harvesting NVP. A hybrid CMOS/ferroelectric nonvolatile flip-flop (nvFF) and a high-speed voltage detector are designed and integrated in the proposed NVP to jointly minimize its sleep and wake-up time. Measurement results demonstrate $46~\mu \text {s}$ wake-up time and $14~\mu \text {s}$ sleep time, which is up to 18 and 24 times speedup over existing works. This approach not only improves the robustness of the NVP to power fluctuations but also brings significant advantages for better utilization of harvested energy.

The interconnected power grid emerges because of the temporal and spatial difference of power generation and power utilization as well as the precondition of considering the complementary resources, safety and reliability. The economic dispatch of multi-area interconnected power grids can realize the effect in scale economies and mutual complementary cooperation in supporting capacities of frequency and voltage, to achieve a wider range of the optimal allocation of resources. Considering the increasing power grid interconnection in the new situation and the trend of resource allocation in the wide range, this paper proposes the economic dispatch method for interconnected power grid considering transmission flexibility. The bus tearing method is applied to split the interconnected power grid, and the decomposition-coordination model is built based on the alternating direction multiplier method. In addition, the alternating direction multiplier idea is introduced to realize the parallel computation of the economic dispatch of each regional power grid. The calculation results of each area are coordinated by the central coordination unit, and the iterative calculation is carried out alternately until the convergence solution is obtained. Numerical example analysis illustrates the effectiveness and economic benefits of the proposed dispatch method.

Large-scale wind power development is an important strategic choice in China's new era. In this paper, firstly, we give a brief introduction of offshore wind farm and the specificity of its electrical structure. Then, we elaborate the necessity for conducting reactive power compensation from two aspects, including operation modes, power grid regulations in china and abroad. At last, we discuss about the reactive power compensation strategies one by one in combination with the actual experience of overseas engineering.

As a by-product of nickel production, the ferronickel slag (FNS) puts a lot of pressure on the environment. It is becoming more and more urgent to deal with the increasing FNS. The aim of this study is to explore the durability of concrete with FNS fine aggregate. Two kinds of FNS with different storage time were selected. The radioactivity detection, XRD test and stability detection of FNS were conducted to ensure FNS can be used as construction materials. Then the durability of concrete with 13%, 27%, 40% and 50% FNS (by weight of fine aggregate) was investigated, respectively. It was found that the properties of concrete prepared from FNS with different storage time had little difference. The results indicated that 27% FNS replacement showed improvement in resistance to sulfate attack by 22% but the resistance to chloride ion penetration was not significantly influenced. Moreover, 40% FNS addition brought a 33% abrasion reduction than that of original concrete. SEM analysis showed that FNS produced more C-S-H gels and improved the microstructure of concrete. This study indicated that proper content of FNS can be used as fine aggregate and it was beneficial to the durability of concrete, especially to the abrasion resistance.

Epitaxial graphene (EG) on semi-insulating SiC prepared by a thermal decomposition method is the most promising strategy for graphene application in large-scale integrated circuits due to compatibility with current semiconductor processes. In this study, high-quality few-layer graphene (FLG) was epitaxially grown on the semi-insulating on-axis 4H-SiC by microwave plasma chemical vapor deposition (MPCVD). Both sides of the SiC substrate were etched with H2 plasma at ∼1000 °C to promote the nucleation and the growth of graphite before epitaxial growth. The surface morphology and properties of EG on SiC (0001) were measured by energy-dispersive x-ray (EDX), x-ray photoelectron spectrogram (XPS), and atomic force microscope (AFM). The qualities of EG grown on the two surfaces of SiC at various temperatures were checked by Raman spectroscopy. Furthermore, the EG growth was controlled by the effect of Ar plasma in the MPCVD and the few-layer (1–3 layers) high-quality EG films were formed on SiC (000 1 ¯ ). The room temperature Hall mobility of EGs up to 2790 cm2 V−1 s−1 on SiC(000 1 ¯ ) was realized.

DC power supply is usually obtained by rectification, filtering, voltage stabilization and other processes of municipal electricity. Due to the incomplete rectification or the existence of switching power supply, the output DC stability is superimposed with AC components, called ripple. The generation of ripple will bring noise, reduce the efficiency of power supply and affect the normal operation of the system. Strong ripple will cause the generation of surge voltage or current, and burn down electrical appliances and other equipment. This paper designs a detection device which can quickly and efficiently detect the ripple of carrier power supply. The experimental results show that the design system can help testers quickly judge whether the ripple of carrier module power supply meets the ripple voltage limit stipulated by Southern Power Grid Corporation, and is used to detect the ripple of power supply. Testing provides a new method of implementation.

With the increasing penetration of distribution generations, distribution system is characterized by complex injection uncertainties, which brings formidable challenges for distribution network operation. As distributed renewable energy penetrating into power network, the operation mechanism and control strategy of distribution network become more complex, which makes it necessary to adjust the control measures to improve the operation quality and efficiency. In order to deal with the increasing uncertainties caused by renewable generations, a voltage aware synergetic dispatch method of active distribution network is proposed. By accounting for the voltage regulation characteristics of all kinds of generators and loads, flexible network topology and participation mechanism of regulation reserves of synchronous generators and connected transmission network in response to uncertainties, an optimization model of voltage aware synergetic dispatch of active distribution network is constructed. The proposed model is formulated as a mixed-integer nonlinear optimization problem and is resolved by the combination of firefly algorithm and nonlinear primal-dual interior-point method. Finally, case studies on the modified IEEE 33-bus distribution system are conducted to illustrate the effectiveness of the proposed method.

In order to deal with diversity of massive data structures and the variety of information formats, a novel mechanism is designed for unified management of power grid planning data. By integrating many business systems including production management system (PMS), geographic information system (GIS), energy management system (EMS), distribution network information acquisition system in the power supply company's system and adopting various technical means such as data warehouse technology (e.g. ETL, Extract-Transform-Load) and incremental capture, data structure that supports the whole process management of power grid business is designed, data correlation analysis and integration & migration are carried out, and efficient access and deep fusion of massive relational data, file-type data, distributed data and spatial data are realized. Besides, through computing modes such as diagnostic analysis, load analysis and spatial analysis, the integrated database for power grid planning that integrates data fusion, storage, mining, modeling, computing, analysis and intelligent perception is finally constructed based on the designed data management mechanism, which could provide the comprehensive model and data support for power grid development. Field application shows the engineering benefit of the designed data management mechanism.

To deal with the increasing penetration of uncertainties caused by renewable generations and uncertain loads, this paper proposes a novel robust optimization model for the optimal power flow (OPF) problem of power systems. In the model proposed, the generation and the network topology are co-optimized, and then the proposed robust optimal power flow with transmission switching (ROPF_TS) model is converted to be a two-stage mixed-integer optimization model based on the duality theory. Based on Benders decomposition, the model is decomposed into a master problem and a slave problem, the master problem corresponds to the resolution of the deterministic optimal power flow with transmission switching problem under base-point mode, the slave problem corresponds to the robust optimal power flow with the given network topology, and then the feasible domain of the master problem is restrained by the Benders cuts developed by the slave problem. Finally, Example analyses are conducted to verify the validity of the proposed model and its solution methodology.

Based on the integration of power production management system, electricity information management system, supervisory control and data acquisition, the power grid planning platform support the power planning by big data technology. Load forecasting and analysis play an important role in power planning. By collecting real-time load data and putting history load data into database, power grid planners can implement load data calculation and analysis in different time scales. Load forecasting reveals the multidimensional and multilevel feature. By exploiting multi-level load forecasting system based on power grid planning platform, “point-line-plane-solid” fore-level coordinated load forecasting can be realized. With the popularization and application to provincial company and citywide companies, the punctuality and accuracy of this system has been proved.

Different power grid operation mode corresponds with different short-circuit capacity distribution, while different distribution of short-circuit (current) capacity is contradictory to power grid operation performance and equipment tolerance influence consequence. This paper establishes functional relationship between short circuit (current) capacity of power grid and effective switch set and feasible power grid operation mode on this basis, deploys analysis on relationship between short circuit (current) capacity of power grid and feasible power grid operation mode on the basis that an equivalent topology analysis model is built, in order to take equipment tolerance into consideration in a compromised way during making of decisions concerning with power grid operation mode and extend power grid operation scheduling.

The Long Baseline (LBL) system is very important to precisely localize autonomous underwater vehicles (AUVs) close to the sea bottom in deep water. But the calibration error of the LBL beacons and the pseudo-range measurement caused by the motivation of the AUV, both them dilute the accuracy of LBL acoustic position. This paper proposes a quantitative description of the calibration error of beacons and pseudo-range how they reduce the LBL acoustic position precise. And then this paper proposes an integrated navigation algorithm for AUV, which integrates inertial navigation and LBL acoustic location to reduce the influence of the calibration errors of beacons and pseudo-range by online calibrating the beacon location errors and compensating the motivation of AUV to pseudo-range model. We name this algorithm as Pseudo-range and SLAM of beacons EKF (P-SLAM EKF). Through the sea trials of Qianlong-1, our algorithm proves very reliable and effective.

By analyzing the various factors affecting the current carrying capacity of power cable in actual operating environment, the mathematical model of temperature change of cable in the actual operation can be obtained. On this basis, establishing a dynamic relationship between its real-time current carrying capacity and operating temperature, and solving by algebraic equations and differential equation alternately, to study the dynamic changing process between its current carrying capacity and operating temperature of cable in real-time operating conditions. Finally, the example proves that the study of current carrying capacity of cable in real-time operating environment can be as far as possible to play its maximum transmission capacity and improve operating efficiency.

In the steel strip cold rolling process, the radial deformation of rollers is important for the control precision of AGC system, and difficult to be measured. In allusion to this problem, a method of data analysis is put forward. In this method, the sequence of radial deformation is calculated from acquired data in the rolling process, and the smoothed curve of deforming process is got using wavelet analysis. At last, the analyzed result is used in rollers deform compensating test in 300 cold rolling mill, the effect of compensation show that the deform information analyzed from history data is reasonable, and could used for compensation in the next rolling process.

Thousands of columns with special shape are analyzed by nonlinear numerical methods. The ductility is calculated to investigate the limit of the axial force ratio and circumstantial requirement for stirrups of an reinforced concrete (RC) column with special shape, in the point of view of the characteristic value for providing stirrup. The limit of the axial force ratio of columns with special shape in relation to the characteristic value of the stirrup is obtained. Then, the effect of stirrup arrangement on the ductility of the RC column is discussed in case of buckling of the longitudinal reinforcement and constraint concrete columns. The complete requirement for stirrups of RC column with special shape is given.

An approach for distribution network planning considering distributed generators(DGs) is proposed in this paper. It determines the optimal placement and capacity of distributed generators in power distribution network without considering the feeder changes and substation upgrades. The corresponding planning model is built and the particle swarm optimization with inertia weight (PSO-w) is used to solve it. An efficient coding scheme is presented. Moreover, a boundary mutation strategy is given to improve global searching ability of PSO-w. Finally the proposed method is tested in an example system and its effectiveness is validated.

The control problem under the condition of uncertainty is an important subject of modern control field. In most control problems, the actual control object and mathematical model for the controller design are always present differences. These differences come from the object parameter uncertainty and time-varying, etc. Mill hydraulic position servo system is the complex nonlinear control system with parameter uncertainty and lots disturbance factors. There is biggish error of the model and the actual object. To eliminate the steady-state error caused by this kind of model error, the paper builds parallel compensation, gain integral compensation, online identification self-calibration and dynamic feedback control strategies on the basis of the steady state error resource and model correction. The simulation results show that the dynamic feedback algorithm has the advantages of both dynamic and static performance.

This paper shows that a comprehensive multivariable approach of autopilot design that mixes sliding mode and PID leads to robustly satisfactory results for underwater mobile robot. The 6-DOF linear equations of motion can be divided into three subsystems for speed control, steering, and diving. The sliding mode algorithm is designed for steering control and PID is used for diving and speed control. The mixed design of autopilot is proposed here for the combined control of underwater robot steering, depth and speed during complex flight maneuvers. A series of simulations based on the dynamics of an underwater robot illustrate the validity of the concept.

How to get rid of the cumulative errors of the cross seal system is one of the key problems for packaging quality. Through the theoretical deduction, the movement regularity of double-crank slide composite horizontal seal mechanism has been ascertained. According to the periodic property of the horizontal seal, the repetitive control strategy has been proposed to increase the system steady-state accuracy. In addition, the design scheme of the repetitive controller has been adopted to control the horizontal seal of the BZA600 packaging machine, and all-digital AC servo electric machine system combined with control have been used as the servo control system of the horizontal seal, which can get rid of the cumulative errors of the repetitive localization control of the horizontal seal. It has been shown from the practical BZA600 system that the control strategy is appropriate and feasible.

Hydraulic bending roll system has become one of key equipments to control the exit strip shape in strip rolling mills. The hydraulic bending roll system of a hot rolling mill was researched and analyzed, and its mathematics models were founded in whole based on the action principle of some representative hydraulic components. At last, the hydraulic bending roll system was simulated with MATLAB by computer. It is obvious that the simulation curves are similar to the real response curves very much.

Because the request to the quality of cold rolling sheet strips is higher and higher, the hydraulic AGC (Automatic Gauge Control) has been a requisite means for improving the quality of cold rolling strips' finished products. But the PID control strategy, being comprehensively used in the locale, is difficult to satisfy the request of improving the control precision of cold rolling mill, because robustness is not very strong. The dynamic matrix control, brought forward by Cutler, is one of predictive control theories. It adopts the non-parameter model based the step response of the controlled object, and adopts strategies of multi-step prediction, receding-horizon optimization and feedback correction, which has the good tracking characteristic. The dynamic matrix control has succeeded to be applied in the industry field, so it is indispensable and feasible to explore its application in the hydraulic AGC system of the cold rolling mill. But the step response of the whole hydraulic AGC can't be gained easily in the practical project, so it restricts the application of the dynamic matrix control in the hydraulic AGC. In this article, using for reference the thought of the cascade control, the predictive control is combined in the cascade structure. In the thickness loop of the hydraulic AGC, it adopts PID control. And in the inner loop of the screw-down servo system, it adopts the dynamic matrix control. The cascade predictive control in this article solves the problem of gaining the step response of the whole hydraulic AGC difficultly. When adopting the dynamic matrix control, the state observer is constructed, which makes the control system have the state feedback form with the observer, strengthens robustness of the system by correcting the predictive model, and makes the system have the intelligence. It synthesizes merits of PID control, which possesses of the easy algorithm and adjusting the parameter conveniently, the characteristic of the dynamic matrix control including of good tracking characteristic and strong robustness. The result of the experiment shows that the adopted control strategy makes the dynamic quality of the system improved largely. The response time and the control precision of the screw-down servo system are improved. And the thickness precision of the strip is improved too.

Hydraulic bend control system, a kind of non-linear, close coupling complex electromechanical system, is an important ingredient of rolling shape control. Conventional PID process control is not good enough to meet the increasing precise, intelligent requests of complex electromechanical system. The combination of intelligent and conventional control methods can solve this problem effectively. Applying virtual instrument to hydraulic bend control system of cold mill, this paper designs conventional PID and fuzzy self-adaptive PID controller, presents the solution of integral saturation problem on conventional PID, uses the fuzzy control rules to online adjust the parameters and realizes the optimal parameter modification. Through the practical test comparison of two control methods, it shows fuzzy self-adaptive PID control greatly improves the response speed and control precision of hydraulic bend control system.

The leading features of complex man-machine systems (CMMS) and an important role of human factors in the system safety an described in this paper. The authors discuss the significance of human error factor identification (HEFI) from different respects, such as human reliability analysis, probabilistic safety assessment, human error reduction and prediction, and system reliability design and redesign. In addition, some of the disadvantages of existing HEFI and the basic criteria needed for a good HEFI technique are also proposed. The main feature of the paper is that it gives not only the definition of HEFI, but also a structured method and frame of HEFI: a multi-view method which consists of HEFI table and view factor trees.

Flatness is one of the most important criterion factors to evaluate the quality of the steel strip. To improve the strip’s flatness quality, the most frequently used methodology is to employ the closed-loop automatic shape control system. However, in the shape control system, the shape-meter is always installed at the down way of the exit of the cold rolling mill and can not sense the changes of the strip flatness in the rolling gap directly. This kind of installation results in the delay of the feedback in the control system. Therefore, the stability and response performance of the system are strongly affected by the delay. At present, there is still no mature way to design controllers for systems with time delay. Although the conventional PID controller used in most practical applications has the capability to compensate the delay, the effect of the compensation is limited, especially for the systems with long time delay. Smith predictor, as a compensator for solving this problem, is now widely used in industry systems. However, the request of highly precise model of the system and the poor adaptive performance to the changes of related parameters limit the application of the Smith predictor in practice. In order to overcome the drawbacks of the Smith predictor, a new Smith predictor based on single neural network PID (SNN-PID) is proposed. Because the single neural network is employed into the Smith predictor to improve the controller’s self-adaptability, the adaptive capability to the varying parameters of the system is improved. Meanwhile, for the purpose of solving the problems such as time-consuming and complicated calculation of the neural networks in real time, the learning coefficient of neural network is divided into several stages as usually done in expert control system. Therefore, the control system can obtain fast response due to the improved calculation speed of the neural networks. In order to validate the performance of the proposed controller, the experiment is conducted on the shape control system in a 300 mm four-high reversing cold rolling mill. The experimental results show that the SNN-PID with Smith predictor controller can effectively compensate the delay effects and achieve better control performance than the conventional PID controller.