Search

We describe a general-purpose end-to-end audio embeddings generator that can be easily adapted to various acoustic scene and event classification applications. In contrast to many other models for audio classification, this end-to-end embeddings generator does not require a separate feature extraction step, but processes audio samples directly which simplifies its porting into hardware platforms. Our approach learns a generic audio embedding representation that is pre-trained on a large audio dataset. It can then be fine-tuned via transfer learning with limited data requirements for an audio classification target application. We compare different transfer learning strategies and evaluate the proposed embeddings generator on three popular publicly available datasets, and report results for different model sizes. The experimental results show that this end-to-end approach presents a scalable, efficient, and competitive alternative to more common spectral-based audio classifiers.

Rain streaks will degrade the visibility of images. To tackle this problem, we propose a novel Adaptive Dilated Network (ADN) to remove rain streaks from a single image while using less parameters and running faster than previous methods. Specifically, an Adaptive Dilated Block (ADB) is constructed as the sub-module of ADN. In ADB, we apply a shared dilated block to extract multi-scale features. Then a dilated selection block is added to leverage the importance of features in different scales. All the multi-scale features are fused together to obtain features with rich rain details. To further model the inter-dependencies of the fused features, a feature selection block is employed in ADB to assign different weights to each feature. Moreover, all the hierarchical features extracted by each ADB are concatenated together and fed into a rainy map generator to estimate rain layer. Experimental results demonstrate that the proposed method is superior to the state-of-the-art methods on performances and running time while using less parameters. The source code is available at https://github.com/nnUyi/ADN.

Recent attempts show that factorizing 3D convolutional filters into separate spatial and temporal components brings impressive improvement in action recognition. However, traditional temporal convolution operating along the temporal dimension will aggregate unrelated features, since the feature maps of fast-moving objects have shifted spatial positions. In this paper, we propose a novel and effective Multi-Directional Convolution (MDConv), which extracts features along different spatial-temporal orientations. Especially, MDConv has the same FLOPs and parameters as the traditional 1D temporal convolution. Also, we propose the Spatial-Temporal Feature Pyramid Module (STFPM) to fuse spatial semantics in different scales in a light-weight way. Our extensive experiments show that the models which integrate with MDConv achieve better accuracy on several large-scale action recognition benchmarks such as Kinetics, AVA and Something-Something V1&V2 datasets.

As a key part of the mechanical automation system, the gear coupling rotor system often occurs faults, and more than 60% of faults are caused by misalignment. In order to further explore the impact of misalignment on the dynamic performance of the rotor system and reduce the occurrence of faults. A signal processing method combining EEMD, median filter as well as FFT is introduced to denoise the original signal, and finite element model is built. A gear coupling-multi-span rotor system test platform is established and results of the finite element analysis is used to verify the experiment. Finally, the parallel misalignment of the gear coupling is simulated by adjusting the distribution of bearing seat gasket. The dynamic response results of the rotor system under different parallel misalignments and different speeds are studied and the impact of parallel misalignment on rotor system is obtained, which has guiding significance for reducing rotor faults.

The existing multi-layer and multi-pass welding process normally adopts teaching-playback method in low-level automation. Also, it is difficult for this method to adapt to the diversity of workpiece by reason of its poor flexibility and universality. To solve this problem, this paper proposes a new robot system for multi-layer and multi-pass welding process based on vision sensor. Firstly, the mathematical model for the welding robot system is established to describe the spatial relationship between the coordinate systems of the tip of the robot, vison sensor and the tip of the welding torch. Then, the groove profile is reconstructed by the detected data of the vision sensor using the proposed data processing method. Based on the spatial model for the welding robot system and the groove profile obtained by the vision sensor, the mathematical model for the welding trajectory is established to realize the multi-layer and multi-pass welding process. Finally, the welding trajectory is verified in simulation and experiment, and the results demonstrate that the welding robot system proposed in this paper has a certain degree of reliability and feasibility.

Most of medium-thick plates are connected by multi-layer and multi pass welding. This paper deals with multi - layer and multi - pass welding of single side V-groove of medium-thick plate. In this paper, the multi-layer and multi-pass welding of medium-thick plate is mainly studied. Firstly, a welding track planning algorithm based on welding parameters is proposed, then the relationship between welding parameters and welding forming coefficient is obtained through experiments on different welding parameters by automatic welding equipment. Finally, the algorithm is verified by automatic welding, after quality inspection, the test plate for welded steel members conforms to the practical engineering standard. The algorithm can guide the welding operation of the automatic welding equipment in the actual construction site.

The linear guide rail has high accuracy requirement for straightness in some mechanical equipment. In order to meet the straightness accuracy of the linear guide rail, straightening machine can be used to straighten the linear guide rail in the industrial manufacturing process. Deformation and bending often occur in different planes, which is a challenge to the traditional straightening control systems. This paper proposes a control system to straighten the linear guide rail precisely. The function of the control system is modularized, which is divided into four main functional modules. Based on the analysis of functional modules of the control system, the straightening control flow is designed, and the workpiece rotation operation is added based on the previous straightening control process, which allows the straightening press head to straighten the linear guide rail at the optimal angle. This also provides a new idea for optimizing the structure of the straightening machine. The hardware and software of the control system are also designed, motion control card is the key part of the hardware, and the function of the software is divided into different functional modules. The control system is used in the straightening machine to straighten the linear guide rail. The experimental data shows that the straightened linear guide rail has a high straightness accuracy, the efficiency of the straightening process has also been improved.

The double-drive feed system with high rigidity and fast response is widely used in high-precision NC machine tools and cartesian coordinate robots. However, the center of gravity of the workbench changes constantly in the process of movement. Therefore, this paper takes the independently-developed double-drive two-dimensional feeding test platform as the research object, analyzes the influence of the change of center of gravity on the double drive feed synchronization performance, takes into account the influence of the transmission chain stiffness of the feed system, and establishes the synchronization error mathematical model of the double-drive feed system. The synchronization performance of the system when different dual-drive synchronization control strategies were applied was analyzed by Simulink. Through experiments, the synchronization error of the dual-drive workbench at different positions and the error based on control strategy of fuzzy PID are obtained, which are compared with the simulation results to verify the accuracy of the simulation model and greatly improve the synchronization error.

The dual-driving feed system (DDFS) based on drive by center of gravity technology has been widely used because of its strong load capacity and better dynamic performance. For the influence of high feed speed and movement of moving parts on gantry-type DDFS, a new cross-coupled control algorithm was proposed. Based on the motion equation of DDFS, the differential equation of DDFS is obtained by considering the mechanical coupling of two axes and the change of friction on both sides of the beam caused by moving parts movement. The dynamic model of DDFS is obtained by Laplace transform. Use this dynamic model to design a single neuron PID cross-coupled controller (SNC) to optimize the synchronous performance and anti-interference ability between two axes. In addition, the fuzzy single neuron PID cross-coupled controller (FSNC) is designed to adjust the single neuron proportional coefficient K. The simulation and experiment show that the FSNC has the best synchronous performance and anti-interference ability when dealing with the movement of moving parts and the high-speed feed of beams.

The dual-driving gantry engraving and milling machine (DGEMM) is taken as the research object. Engraving and milling capabilities are combined on DGEMM with high speed, high precision and high efficiency processing performance. Using Park and Clark conversion methods to establish a mathematical model of the motor; Using Simulink software to perform system simulation, and combine the actual parameters of the motor to tune the three-loop PID parameters; Designing and developing DGEMM software with integrated control function and trajectory planning function through C# and TwinCAT software; Using Mastercam software to verify the accuracy of the servo system model and the functions of the software. The experimental results show the accuracy of the servo control model and the practicality of the software in terms of control functions and trajectory planning.

As the transmission device of the ship, the vibration of the propulsion shafting has a very important influence on the overall working performance of the ship. Based on the transfer matrix method, the lumped parameters-distributed parameter hybrid model of propulsion shafting's whirling vibration is established in this paper. The simulation result shows that the model can reflect the system dynamics well. On the basis of the whirling vibration model, the influence of the rear stern bearing's supporting position and the propeller's entrained water effect are analyzed. Moreover, the natural frequency of the propulsion shafting increases with the increase of the rear stern bearing's equivalent stiffness. The propulsion shafting's variation law of the natural frequency with the above characteristics is revealed.

It is challenging to generate loop invariants for programs automatically in the field of software analysis and verification. Loop invariants are the weakened forms of the postconditions for loops. Therefore, we propose PBLInv, a postcondition-based approach to generate loop invariants for C programs with the machine learning method. First, we generate the postcondition for a loop program automatically. Second, we learn classifiers as the updated candidate loop invariants with the Kernel Support Vector Machine (KSVM) method iteratively. PBLInv is evaluated with 60 benchmark programs collected from the recent papers and the 2019 Software Verification Competitions (SV-Comp 2019). The experimental results show that PBLInv is efficient at learning loop invariants for C programs. Compared with five state-of-the-art methods of generating loop invariants, PBLInv not only generates loop invariants for more benchmarks, but also reduces the number of used samples and iterations for learning loop invariants.

Inspired by the procedure that experts construct loop invariants, we propose a novel data-driven approach to generate verified loop invariants automatically. The approach consists of the prediction phase and the evaluation phase. The prediction phase aims to generate candidate loop invariants automatically by solving polynomial equations and synthesizing the extended loop conditions. The evaluation phase aims to generate verified loop invariants that can be utilized to prove the correctness of program postconditions. The spurious candidate invariants are first pruned out using the method of predicate abstraction. Then, the redundant relations are removed from the candidate invariants using an Satisfiability Modulo Theories (SMT) solver. The proposed approach is compared with state-of-the-art methods on 32 benchmarks collected by the recent papers. The experimental results demonstrate that the novel data-driven approach generates 29 verified loop invariants successfully. The proposed approach is efficient at discovering verified loop invariants automatically for executable loop programs. Meanwhile, compared with other methods of generating loop invariants, the proposed approach not only costs less time to generate invariants, but also generates invariants with better quality.

We demonstrate a 1030 nm 520 fs-pulse-driven ErAs: InGaAs photoconductive receiver suitable for detecting 14 mW average THz power (~ 950 nJ pulse energy). To the knowledge of the authors, this is the highest detected pulsed THz power using photoconductive receivers reported so far. The current (field) responsivity of this receiver is in the range of $110\pm 25\mu A/\sqrt{W}$ yet with slight saturation.

Wind power is a kind of green energy, renewable, pollution-free, inexhaustible, so wind power has a broad market prospects, its technology is also in continuous progress, wind turbine assembly unit capacity is higher and higher, the scale of wind farm is also towards large-scale development. As the capacity of wind turbine assembly increases, the blade diameter of the matching wind turbine needs to be increased correspondingly to obtain more wind energy. Taking offshore wind turbine as a representative, the blade diameter of some wind turbines has reached more than 100 meters. When the blade diameter becomes larger, it will bring a series of problems. The main performance is that the elasticity and flexibility of wind turbine tower, transmission shaft, blades and other main components will increase correspondingly, while the damping value of power electronic components will be smaller. These problems make the fatigue load variable of large wind turbine more complex, which will lead to the low frequency mode of the unit The more complex the modes are, the more complex they are. When the diameter of the wind turbine increases, the pitch angle increases due to tower shadow effect, wind shear and turbulence time-varying factors, which causes the blade to swing and shimmy more significantly, and then causes the fatigue load condition of the wind turbine to be more obvious, which seriously affects the mechanical stress intensity change between various transmission mechanisms of the wind turbine, thus reducing the service life of the wind turbine Limit. With the increase of blade diameter, the frequency of grid flicker will also be caused, which will cause voltage instability and reduce the power quality of grid after the wind turbine is connected to the grid. With the increase of impeller diameter and fan capacity, in recent years, independent variable pitch control technology came into being.Independent pitch control technology not only has the characteristics of centralized pitch control, but also can balance the unbalanced load on variable speed wind turbine, so as to reduce the fatigue load of wind turbine. Therefore, it has been paid attention to in the field of wind power research in the world. There are many research results in this area, but the research on the relationship between independent pitch control strategy and grid flicker is relatively less. The factors causing voltage fluctuation include air density, impeller speed, pitch angle, wind field distribution and power grid condition. This paper analyzes the wind shear and tower shadow effect which cause the vibration of wind turbine, studies the relationship between independent pitch strategy and wind turbine voltage change, proposes a control strategy of independent pitch and designs a controller to control the power of wind turbine during operation The method of voltage variation to make it stable, and then reduce the flicker of power grid, and carry out simulation experiments to verify its effectiveness. The control strategy can reduce the grid flicker at the common point of wind turbine outlet and grid connection, improve the stability of wind turbine grid connection, and improve the overall power quality of the grid.

With the progress and development of socio-economic technology, the contradiction between the strong demand for energy in various industries and the increasing shortage of traditional petrochemicals is deepening. People are increasingly aware of the importance of developing renewable energy, especially the use and development of wind energy. Pitch control is the core technology of wind power generation. At present, most of the mainstream products in the international wind power market apply centralized variable pitch control. The advantage of centralized pitch control is that it has a high wind energy utilization coefficient, which can ensure the maximum power in the low wind speed section, the rated power in the high wind speed section, improve the starting performance and braking performance of the wind turbine, improve the overall flexibility of the wind turbine, Force conditions of small machines and blades. Independent pitch control technology is a control technology method developed in recent years with the continuous increase of fan capacity and the constant increase of impeller diameter.This paper takes the wind energy conversion system of large doubly-fed induction generators as the research object, proposes an independent pitch control strategy, uses FAST and MATLAB / Simulink for dynamic modeling, simulation and comprehensive analysis. The results of simulation experiments show that The unbalanced load of wind turbine units is an effective control method to improve the power supply quality and extend the service life of wind turbine units.

Rainy images contain rain streaks with different sizes, shapes, directions, and densities. To efficiently remove rain streaks from rainy images, it is necessary to capture rich rain details. In this paper, we propose a Residual Dense Dilated Aggregated Network (RDDAN) to focus on different types of rain steaks and efficiently model rain distribution from rainy images. Specifically, a Residual Dense Dilated Aggregated Block (RDDAB) is constructed to fully extract and exploit rain details hierarchically. In RDDAB, dilated aggregated module is applied to capture multi-scale rain details, dense connection is employed to fully exploit hierarchical features extracted by dilated aggregated module, and residual connection is introduced to keep flow of rain details among different blocks. Besides, all the features extracted by each RDDAB are fused progressively which allows the network to adaptively focus on significant hierarchical features inter blocks. Extensive experiments demonstrate that our method outperforms the state-of-the-art methods on synthetic and real-world datasets. The source code is available at https://github.com/nnUyi/RDDAN.

Image rain removal has been widely studied with traditional methods and learning based methods for years. However, traditional methods like Gaussian mixture model and dictionary learning methods are time consuming and fail to well tackle images with heavy rain streaks since image patches are severely contaminated. By considering the line-like property and angle distribution of rain streaks, this problem can be well solved. In this paper, by introducing Directional b radient operator of arbitrary direction, we propose an efficient and robust Constraints based Model (DiG-CoM) for single image rain removal. Moreover, a density metric of rain streaks is applied to generalize the proposed model to light and heavy rain streak occasions. Extensive experiments on synthetic datasets demonstrate that the proposed model outperforms GMM and JCAS while requiring less time. Furthermore, on real-world occasions, the proposed method obtains better generalization ability compared with the stateof-the-art learning based methods. The source code is available at https://github.com/Schizophreni/Set-vanish-to-the-rain.

The dual-driving feed worktable (DDFW) is increasingly used in feed systems with high precision and dynamic performance requirements, while its geometric error, caused by the factors such as manufacturing, assembly errors and the deformation, is the main factor affecting positioning accuracy. The research on geometric error compensation of DDFW is significant. In this paper, geometric error analysis and compensation of DDFW, equipped with the open CNC system of BECKHOFF, are conducted. Firstly, the geometric error model of DDFW is established, and the geometric errors are measured by using dual-frequency laser interferometer. Then the electronic cam combined with virtual axis compensation algorithm is designed. Finally, the compensation experiment is completed, and the results show that the positioning precision of DDFW improves a lot with the proposed compensation strategy.

The accuracy and efficiency of rotor fault diagnosis can be improved by using the local mean decomposition (LMD) in rotor fault features extraction. In this paper, a rotor fault features extraction technique depending on LMD and singular value entropy is proposed. In the first place, the local mean decomposition is implemented to attain multiple amplitude-modulation product (PF) components by decomposing the original vibration signal of the rotor. Then, the PF components are decomposed by singular value decomposition to obtain singular values and the singular value entropy. Finally, input the singular value entropy into the support vector machine (SVM) to identify and classify rotor faults. The results of experiment reveal that this method can extract the characteristics of rotor faults effectively, and identify different rotor typical faults accurately.

As the key technology in the weld seam visual tracking system, the laser strip center extraction is highly valued to the performance of detection system. In this paper, a laser strip center extraction methodology based on geometrical center method, direction template and gray centroid method is proposed. Firstly, the laser strip center is picked up as an approximate skeleton by geometric center method. Then a normal judgment method based on the Lagrange mean theorem is designed to determine the normal direction of the skeleton. At last, accurate laser strip center gray is obtained by gray centroid method. The experimental results indicate that suggested method is both simple and accurate, and meet the speed and precision requirement of the weld seam detection.

The dual-drive gantry structure has been wildly used in industrial machines, especially in high-speed and high-loaded equipment. However, it needs precise synchronization control strategies to guarantee the performance. Aiming to the uncertainties in the industrial environment, a strategy with Artificial Bee Colony Algorithm (ABCA) and adaptive cross-coupled synchronization control strategy are adopted to improve the dual-drive control performance. Based on the integral of the time-weighted absolute error criterion, ABCA is used to optimize the tracking ability of single drive subsystem, and the simplified variable universe fuzzy PID type compensator is used to reduce the synchronization error of two parallel drive subsystems in one particular direction. The simulation results and experiment result show that such kind of gantry feed system using proposed control strategy has better synchronization performance.

Motor shaft is widely used in many industrial fields. As a crucial factor, the motor shaft straightness has significant impact on its working performance, thus the straightening process becomes an essential step in shaft manufacturing. In order to achieve automatic straightening of the motor shaft, a motor shaft straightness control system is designed in this paper. A motor shaft straightening process is developed for this system. The factors affecting the straightness of the motor shaft are analyzed. Three deformation patterns are proposed based on the analysis of motor shaft deformation. A structure of the modular straightness control system is designed based on the motor shaft straightening process. A motor shaft straightening stroke prediction algorithm is developed based on the three-point bending process by FEM and elastic-plastic deformation theory. The validity of the system and the correctness of the straightening stroke prediction algorithm are verified by experiments.

The manipulator is widely applied in bolt disassembly in the field of remanufacturing. The intelligent manipulator with the ability of generating suitable operating path is increasingly required. To improve the path planning ability of manipulator in bolt disassembly, a strategy of dual path planning using a hybrid algorithm which combines an improved genetic algorithm (GA) and A* algorithm is proposed in this paper. The model of bolt disassembly sequence and algorithm are first established to find out the shortest disassembly path. The improvement of selecting and crossing operators are designed to enhance the stability and efficiency of basic GA. Both basic GA and improved GA are used and compared in the simulation of path planning with MATLAB. Then, the A* algorithm is adopted to accomplish the obstacle avoidance based on the path obtained with improved GA. Last, the experiment focusing on the bolt disassembly of a specific reduction gearbox is performed, and the correctness and feasibility of the hybrid algorithm are verified in the simulation and experiment.

The research of voltage sag mitigation method is an important task. A new method using hybrid energy storage (HES) strategy is proposed, which consists of Superconducting Magnetic Energy Storage (SMES), Battery Energy Storage (BES) and Dynamic Voltage Restorer (DVR). In addition, a power allocation optimization strategy for HES-based DVR is proposed and validated using the life cycle cost (LCC) theory. LCC is the objective function, which can be used to establish the optimization model. The constraint conditions of the model include the power of the throughput, the state of charge (SOC) and the power allocation optimization strategy. Finally, the cost analysis and the compensation performance are implemented using the SEMI F47 curve for the single BES, single SMES and SMES-BES device. The compared results show the superiority and effectiveness of the proposed method on the voltage sag mitigation.

The aim of this paper is to address the energy-efficient scheduling problems in the hybrid flow shop problem (HFSP) with machine maintenance. For this aim, the mixed integer programming model (MIPM) considering energy consumption and machine maintenance is proposed firstly. Given the NP-hard nature of the model, a hybrid multi-objective algorithm (HMOA) combining an improved algorithm based on the Non-Dominated Sorting Genetic Algorithm (NSGA-II-imp) with a heuristic algorithm is developed to solve MIPM. The NSGA-II-imp designs a new elite remained strategy for the purpose of balancing two conflict objectives and a heuristic algorithm is mainly designed to add machine maintenance during production scheduling. Finally, the proposed algorithm is compared to GA, SA-GA, NSGA-II separately based on extensive numerical experiments. Results show the effectiveness of the algorithm in this paper.

Single image deraining is an ill-posed inverse problem due to the presence of non-uniform rain shapes, directions, and densities in images. In this paper, we propose a novel progressive single image deraining method named Recurrent Multi-scale Aggregation and Enhancement Network (ReMAEN). Differing from previous methods, ReMAEN contains a symmetric structure where recurrent blocks with shared channel attention are applied to select useful information collaboratively and remove rain streaks stage by stage. In ReMAEN, a Multi-scale Aggregation and Enhancement Block (MAEB) is constructed to detect multi-scale rain details. Moreover, to better leverage the rain details from rainy images, ReMAEN enables a symmetric skipping connection from low level to high level. Extensive experiments on synthetic and real-world datasets demonstrate that our method outperforms the state-of-the-art methods tremendously. The source code is available at https://github.com/nnUyi/ReMAEN.