Abstract:This paper considers the formation problem for a group of autonomous underwater vehicles (AUVs). In order to reduce the energy consumption of communications among AUVs, a distributed dynamic event-triggered leader-follower consensus control strategy is proposed. First, we design an auxiliary variable that includes general? ized position and carrier speed, which simplifies the AUV model. Based on sliding mode variable structure control, consistency theory, and dynamic event-triggered strategy, a distributed formation controller is designed. The control? ler ensures that the AUV system can achieve the formation goal in the presence of external disturbances. Second, no Zeno behavior is exhibited under the proposed control algorithm. Last, numerical simulation results are provided to verify the correctness of the presented theorem and the effectiveness of the proposed control algorithm.

Abstract:Aiming at the negative effect of a hub motor on the ride comfort of an electric vehicle, a four-degree of freedom vibration plane model of hub motor electric vehicle is established, in order to study the influence of pas? sive suspension and active suspension on the pulse road ride comfort of the electric vehicle. An active suspension control strategy for hub motor electric vehicles is designed, and MATLAB/Simulink control simulation model is developed with the constrained state H∞ control method. The effects of no eccentricity, 10% eccentricity, and 20% eccentricity on the excitation of the hub motor are analyzed. The time history of pulse road vibration response of pas? sive suspension and active suspension of hub motor electric vehicle, as well as the ride comfort evaluation indexes of four cases, are compared. The results show that the eccentricity of the hub motor can produce vibration excitation on the electric vehicle, which not only affects the pulse road ride comfort but also affects the state feedback H∞ control.

Abstract:In order to achieve the effect of vehicle lightweight, based on the steel anti-collision beam of a light passenger car, the structure optimization design of the anti-collision beam made of carbon fiber was carried out ac? cording to the low-speed collision standard. Firstly, the cross-section parameters of the anti-collision beam are deter? mined by the full factor test. Considering the material discontinuity caused by different thicknesses of each area, the ply principle based on ply compatibility is proposed, and the thickness space of the anti-collision beam and the corre? sponding ply sequence are determined. In order to optimize the thickness of the anti-collision beam, the multiobjective particle swarm optimization algorithm based on the kriging model is adopted. Based on the traditional par? ticle swarm optimization algorithm, the multi-objective adding point strategy is introduced, which can effectively solve the repeated test design problem caused by the insufficient accuracy of the approximate model, and improve the optimization efficiency. The simulation and sled test of the optimized anti-collision beam show that the low-speed im? pact performance of the carbon fiber anti-collision beam meets the requirements.

Abstract:Aiming at the difficulty in determining the motion parameters such as effective drive wheels and track slip rate of a polymetallic nodule mining vehicle, based on the modeling of traction force and load characteristics of tracks, a higher-order nonlinear system for estimating the motion parameters of a polymetallic nodule mining vehicle is proposed. To address the problem that the UKF based on Gaussian model cannot achieve high estimation accuracy in high-order nonlinear systems, an adaptive traceless particle filtering algorithm (AUPF) based on Monte Carlo sam? pling principle is proposed，and the adaptive traceless Kalman filter (AUKF) is used to improve the Particle Filter (PF) by refining the probability density function, which solves the shortcomings that the PF is easy to diverge and the estimation accuracy of UKF is low. The experimental results show that the AUPF algorithm can achieve an accurate estimation of the motion parameters of polymetallic nodule mining vehicles, and there is an important engineering ap? plication value.

Abstract:In order to explore the influence of heavy-duty tire performance on multi-axle heavy vehicle driving characteristics, the model parameters of the GL073A heavy-duty meridian tire were identified based on a sixcomponent tire test. Aiming at the characteristics of strong nonlinear changes caused by multiple parameters of MF (Magic Formula) model and large vertical load range of heavy radial tire, a hybrid optimization parameter identifica? tion method based on Particle Swarm Optimization (PSO) and Levenberg-Marquardt algorithm was proposed. The pa? rameter identification and result verification of the tire model under longitudinal slip and sidesway conditions of the heavy-load radial tire were carried out. The results show that the parameter identification accuracy of the tire model can be improved based on the hybrid optimization algorithm, and the residual of the identification results can be controlled in a range of 5%. Based on the Sobol sensitivity analysis method，the influence of multi-characteristic param? eters on the tire MF model was studied. The first-order sensitivity and total order sensitivity of each characteristic pa? rameter were used as evaluation criteria to screen out the leading parameters affecting tire mechanical properties. The results show that based on Sobol sensitivity analysis, 13 leading parameters are selected from 58 characteristic param? eters of the magic formula tire model. Compared with the results of direct hybrid optimization, the maximum increase of residual error is 0.138%, and the maximum increase of model convergence rate is 30.4%.

Abstract:As the loader in the process of steering produces pressure shock and pressure fluctuation due to the position of the cylinder hinge point arrangement, in this paper, the minimum stroke difference, the minimum arm dif? ference, and the minimum steering power are set as objective functions, and Genetic algorithms are used to optimize it. The AMESim simulation and experiment are combined to verify the feasibility of the optimization results. After the optimization, the average stroke difference is reduced by 89.23%. The average moment arm difference is reduced by 88.40%. The average power consumed by the engine at idle and full speed is reduced by 32.56% and 24.03%, respec? tively. Through an in-depth study of stroke difference and moment arm difference curves, moment arm difference is identified as the dominant factor causing pressure fluctuation. Genetic algorithm is used to optimize the cylinder hinge coordinates quadratic. The optimization results show that the maximum stroke difference and moment arm dif? ference are reduced by 14.29% and 19.44%, respectively, compared with the first optimization. A full load and full speed fast rotation experiment are carried out after the cylinder is modified by the hinge point. There is no obvious pressure anomaly in its pressure curve.

Abstract:To better simulate the real driver behavior, a longitudinal driver regression model based on quadratic regression and the Lasso regression method is proposed. Firstly, by collecting longitudinal driving behavior data, state parameters that may affect driving behavior are extracted, and a quadratic regression driver model was estab? lished. Then, the Lasso regression method is applied to screen the state parameters for the multicollinearity problem in the multi-parameter regression model. Finally, a quadratic regression driver model is developed based on the screened data. In order to verify the effectiveness of the model, a simulation comparison is made between the pro? posed model and PI driver and Lasso regression driver models. The simulation results show the driver model devel? oped here not only has a better effect on driving cycle traction when compared with the other two models but can also better reflect the characteristics of actual driving behavior.

Abstract:To solve the problems of complex structure and difficulty in tuning the exceptional points in the exist? ing PT-symmetric structures, a PT-symmetric beam for flexural waves is designed based on piezoelectric shunting technology. Firstly, the PT-symmetric condition is derived. Then, based on the effective medium method and finite el? ement simulation, it is verified that the effective parameters of the gain and loss unit meet the PT-symmetric condi? tion, and the tunability of exception points is studied by changing the resonant frequency and the shunting resistance. Finally, the scattering property of the PT-symmetric beam is derived by the transfer matrix method and finite element simulation, and the relationship between exceptional points and unidirectional non-reflection is illustrated. The cal? culated and simulated results show that the PT-symmetric beam has several exceptional points including 511 Hz and 520.5 Hz. When the incident flexural wave of 511 Hz is applied at the right side of the PT-symmetric beam, the re? flection coefficient is close to zero. However, when the frequency of the incident flexural waves changes to 520.5 Hz， it should be applied on the left side of the PT-symmetric beam to gain an entire transmission without reflection.

Abstract:By introducing an intermediate-mass into the nonlinear three-parameter isolator with an X-shaped mechanism, an improved three-parameter isolator is proposed, and its theoretical model is established. The Har? monic Balance Method is used to obtain the steady-state analytical solution of the vibration isolation system, and the correctness is verified via the fourth order Runge-Kutta method and ADMAS. By using the force transmissibility as the indicator for evaluating the isolating performance, some comparisons are carried out among the proposed isolators. Meanwhile, according to the engineering application, the time-domain displacement responses of three types of isolat? ing systems under multi-frequency steady excitation are solved, and a comparative study is carried out, respectively. The vibration power flow and maximum kinetic energy are investigated, and some typical design parameters are se? lected and analyzed. At last, from the perspective of the vibration dynamic absorber, the effects of the nonlinear connection on the vibration suppressing performance are further discussed. The calculation results show that the natural frequency of the vibration isolator can be decreased after the intermediate mass is introduced into the system, the ef? fective vibration isolation frequency band of the vibration isolation system becomes wider, and an anti-resonance fre? quency is introduced at the original resonant frequency at the same time. Compared with the traditional threeparameter isolator and the nonlinear vibration isolator with an X-shaped mechanism, the vibration isolation perfor? mance of the developed isolator at both low and high frequencies is improved accordingly. In addition, the abovementioned all design parameters exist the optimal values when the force transmissibility of the nonlinear isolator is used as an evaluating index. Besides, the vibration suppressing performance of the vibration dynamic absorber with a grounded X-shaped mechanism is enhanced.

Abstract:In order to obtain the necessary and sufficient conditions for the strongly modulated response (SMR) of a nonlinear energy sink (NES) system under harmonic excitation, the SMR study of the NES system is carried out. Firstly, the equation of systematic slow variation flow is derived by using the complex-averaging method. Secondly, the multi-scale method is used to realize the separation of the rapid variable manifold and the slow variable manifold. And the systematic slow invariant manifold under the different parameters and the global bifurcation property are ob? tained. Then, by building the one-dimensional mapping function of the rapid variable manifold, the sufficient and necessary conditions for the NES system in the SMR state are revealed. Finally, the simulation circuit of the NES sys? tem is constructed, and the response detection circuit test is carried out. The simulation and experimental results show that the SMR is caused by the saddle-node bifurcation of limit cycles in the slowly varying power flow of the coupling system, and it is an actual phenomenon in engineering. The system in which SMR state can appear must sat? isfy the following two conditions: the response of the NES system exceeds the amplitude of the extremum point on the slow invariant manifold, but it does not attract to a branch of the slow invariant manifold, and a continuous jump loop without falling into a local cycle is formed.

Abstract:Aiming at the problem of cooling water flow distribution for multi-cavity hot stamping molds, a cool? ing water flow distribution method is proposed. A simulation model of the hot stamping process of U-shaped parts with variable channel parameters is designed. Based on the joint simulation of FLUENT and LS-DYNA, the variation law of the temperature of the formed parts with the thickness of the sheet, the holding time, and the velocity of the pro? file is studied, respectively. And the flow distribution of the mathematical model is established. When the target tem? perature of the part is determined, the specific flow rate required for cooling the thin plate can be quickly determined according to the constructed flow distribution mathematical model, and the reasonable distribution of cooling water in each cavity can be realized. Taking a one-mold three-cavity hot stamping mold as the research object, the valve opening of its external waterway is adjusted according to the established flow distribution mathematical model, and the results before and after the adjustment are compared. After the adjustment, the temperature of the thick plate threshold drops from160.4 ℃ to 150.1 ℃ with a drop of 6.42%. The research results show that the adjustment of the cooling water flow has a significant effect on the final temperature of the formed part. The method of cooling water flow distribution of the multi-cavity hot stamping mold can shorten the overall pressure holding time.

Abstract:In order to improve the machining accuracy and efficiency of harmonic gear, this paper proposes a skiving machining and the tool design method for double-arc harmonic rigid wheels. Based on the envelop theory, the conjugate tooth profile of the harmonic rigid wheels is solved, and the least squares method is used to fit it. The conju? gate surface and rake surface are solved by constructing the skiving machining coordinate system and rake face coor? dinate system, and NURBS surface is used to fit. The rake and conjugate surface are intersected by Newton interative method to obtain the cutting edge data, and the mathematical model of skiving tool is established by importing the data to CAD software. On this basis, the influence of different rake edge angles and back edge angles on tool profile error is analyzed. The results show that the skiving tool has a certain deviation between the actual tooth profile and the theoretical tooth profile due to the rake edge angle and back edge angle, and this deviation increases gradually from the tooth tip to the tooth root with the increase of rake angle and back angle.

Abstract:Aiming at the problems of the traditional Fuzzy Neural Network (FNN) evaluation model in the ser? vice quality evaluation of the National Quality Infrastructure (NQI) comprehensive service information platform, such as slow convergence speed and likely falling into the local optimal solution, a fuzzy neural network intelligent evalua? tion method based on Optimized Principal Component Analysis (OPCA) and Improved Genetic Algorithm (IGA) was proposed. In order to improve the network convergence speed of FNN, OPCA was used to delete redundant index fac? tors reduce the amount of network input, and realize the dimensionality reduction of network input, according to the correlation between evaluation indexes. Then, IGA is combined with FNN, and the coefficients of the membership function of FNN are searched globally by using adaptive crossover and mutation probability, so as to overcome the problem that FNN is easy to fall into local extremum in intelligent evaluation effectively. Based on the actual service quality survey data of the NQI platform in China, the experimental analysis shows that the OPCA-IGAFNN evalua? tion model has a more efficient and accurate evaluation effect.

Abstract:Machine learning methods have been widely used in the field of short-term load forecasting of power systems. However，it is difficult for load operators to obtain high-performance forecasting models due to insufficient data samples，poor model generalization ability, and high data privacy protection requirements in the application pro? cess. In this paper, meteorological, date, and historical load are used as input features to construct a short-term load forecasting model based on Long Short-Term Memory(LSTM). A federated learning(FL) based coordinated training method of a short-term load forecasting model is proposed. The proposed method mainly iteratively updates model pa? rameters through decentralized training and aggregation of centers, so as to realize cooperative construction of the pre? diction model by all load operators under the condition of data privacy. The simulation results based on the GEF?Com2012 dataset show that the proposed method not only ensures the data privacy of operators but also effectively im? proves the forecasting accuracy of the load forecasting model, and the trained model has satisfied generalization abil? ity in multiple scenarios.

Abstract:To improve the accuracy of wind power prediction, an ultra-short-term combination forecasting method based on Frequency Run Length Discriminant and Variational Modal Decomposition (VMD) residual error correction is proposed. Firstly, the original wind power sequence is decomposed by VMD to obtain a series of subsequences with different center frequencies, and then the residual sequence is extracted from the difference in the se? quences. The residual sequence has the characteristics of large fluctuation, nonlinear complexity, and unsteadiness, which inherits the original sequence noise component and the masked information during decomposition, and the adaptive t-distribution Sparrow Search Algorithm (t-SSA-LSTM) combined with the weather features is used for the prediction. The sub-sequences are divided into two kinds of signals class, namely high and low- frequency se? quences, by using the Frequency Run Length Discriminant method. The low-frequency sequences are linear stability and the adaptive t-distribution Sparrow Search Algorithm (t-SSA) is used to optimize the autoregressive integrated moving average (ARIMA) model prediction. The characteristics of high-frequency sequences are volatile and com? plex, and the t-SSA is used to optimize the Long Short-Term Memory (LSTM) neural network for the prediction of high-frequency sequences. Finally, the wind power prediction results are achieved by linearly superimposing the pre? diction results of different sequences. The proposed model is finally applied to a wind farm in China, and the results show that the model can effectively improve the prediction accuracy.

Abstract:Focusing on the problems of low accuracy, poor real-time performance, and no feature extraction in the current grid single learner power-theft detection method, a power-theft behavior detection method based on the Convolutional Neural Network-Light Gradient Boosting Machine (CNN-LG) model is proposed. First, the power fea? tures of user electricity data are extracted through the Convolutional Neural Network (CNN), and the extracted fea? tures are input into the Light Gradient Boosting Machine (LightGBM, LG) classifier based on the decision tree in or? der to train the data. On this basis, a detection method of electricity theft based on the CNN-LG model is estab? lished. Finally, the State Grid Corporation of China and Irish Smart Energy Trail（ISET）datasets are used to conduct experiments to verify the accuracy and effectiveness of the method proposed in this paper. The experimental results show that the method proposed in this paper can quickly and accurately realize the detection of various power theft behaviors in the power grid. Compared with the existing detection methods, it has higher accuracy, better generaliza? tion performance, and real-time performance.

Abstract:In order to assist differentiated prevention of bird-related faults in power grid, this paper proposes a method for the identification of bird species related to power grid faults based on combined features and a Convolu? tional Neural Network (CNN). Firstly, based on the information from historical bird-related faults in the power grid and the investigation results of bird species around transmission lines, 13 high-risk, 8 low-risk, and 2 harmless bird species were selected to build a sound sample set. Then, the Mel-frequency Cepstrum Coefficients (MFCC), Gamma? tone Frequency Cepstrum Coefficients（GFCC），and Short-term Energy (STE) features of bird sounds were extracted after preprocessing such as framing, windowing, noise reduction, and clipping. To solve the problem of insufficient ex? pression ability of a single feature set, a new sound feature set was generated combining MFCC, GFCC, their firstorder differences, and STE features after normalization. Finally, a CNN was built to train and recognize the combined features. The identification accuracy of the test set reaches 91.8%, which is better than those with a single MFCC and GFCC feature set

Abstract:The internal characteristics and system operation of MMC on the fault side are greatly affected when an asymmetric fault occurs in the AC power grid of MMC-HVDC based on a modular multilevel converter. Based on the average value model of the MMC bridge arm, this paper proposes an optimal control strategy for the MMC-HVDC system under asymmetric conditions, which enhances the fault ride-through capability of a flexible HVDC system. The strategy is mainly composed of two parts: AC current measurement control and circulating current suppression.On the AC side, the phase-locked loop based on a double second order generalized integrator (DSOGI-PLL) is used to extract the positive and negative sequence components of voltage and current accurately under asymmetric conditions, and the negative sequence current is suppressed with a double vector controller to control the three-phase current bal? ance on the AC side. The Proportional Integral and Repetitive Control（PI-RC）composed of PI controller and repetitive controller in series is used in MMC to suppress the positive and negative zero sequence components of double frequency in circulating current, so as to realize the constant transmission of DC power. Finally, a simulation model of the MMCHVDC system is built in MATLAB/Simulink software to verify the efficiency of the proposed optimal control strategy.

Abstract:The multipath effect and Doppler effect of the Underwater Acoustic (UWA) channel cause intersymbol interference and inter-carrier interference at the receiver of the orthogonal frequency division multiplexing (OFDM) communication system, which degrades the system performance. A novel Passive Time ReversalConvolutional Neural Network (PTR-CNN) is constructed and applied to the OFDM UWA communication system re? ceiver. The PTR-CNN network consists of two parts. Firstly, it weakens the multipath and enhances the main path in? formation energy based on passive time reversal theory. Secondly, the above-mentioned output result is converted into a two-dimensional matrix, which is input into the CNN for signal detection to simultaneously combat the interfer? ence caused by the multipath and Doppler effect. Finally, the network output directly restores the bit stream. Simula? tion and experimental results demonstrate that when compared with the current mainstream channel estimation and signal detection algorithms, the proposed method can improve the reliability of the system, and it has better robust? ness in different UWA channel environment tests.

Abstract:The samples are insufficient due to the difficulty of obtaining real brain tumor MR images, which se? riously affects the performance of deep learning models. Therefore, a sample generation method based on the Multiple Discriminator Cycle-consistent Generative Adversarial Network (MD-CGAN) is proposed in this paper. Firstly, the MD-CGAN model is used to generate brain tumor pathological region images, and then these pathological region im? ages are overlaid with the normal sub-regions of brain images to synthesize brain tumor MR images. Among them, the double adversarial loss introduced by MD-CGAN avoids the problem of model collapse, and the cycle consistency loss function introduced can ensure that the normal brain sub-region images generate the pathological region images of brain tumors, so that the images generated by MD-CGAN have high quality and diversity. Taking the Fréchet In? ception Distance(FID) as the evaluation index, the MD-CGAN proposed in this paper and the more classic generative networks in recent years are used to generate the images of brain tumor pathological regions and calculate the FID value. The experimental results show that the FID of our MD-CGAN is 26.43%, 21.91%, and 12.78% lower than those of SAGAN, StyleGAN, and StyleGAN2, respectively. To further demonstrate the effectiveness of our proposed method, we use the generated brain tumor images to expand the training set and then train the segmentation models on this expanded dataset. The experimental results show that the performance of segmentation networks trained on the expanded dataset is better. Based on the above experimental results, it can be concluded that the brain tumor MR images generated by our proposed method have high quality and rich diversity. These samples can be used to expand the training set and effectively solve the problem that brain tumor MR images are insufficient.

Abstract:Traditional image denoising methods are difficult to effectively retain detailed features while filtering the speckle noise of sonar images. To overcome this problem, an image denoising method based on density clustering and grayscale transformation in a non-subsampled shearlet domain is proposed. Non-subsampled shearlet transform is used to decompose the noisy image into high-frequency coefficients and low-frequency coefficients. According to the distribution characteristics of speckle noise in the sonar image, the Density-based Spatial Clustering of Applica? tions with Noise(DBSCAN) algorithm is used to process high-frequency coefficients to separate noise interference and retain detailed information. Gray-scale transformation is performed on the low-frequency coefficients to enhance image contrast. Finally, the processed high-frequency coefficients and low-frequency coefficients are reconstructed by non-subsampled shearlet inverse transformation to achieve image denoising. The experimental results show that the method is effective in improving the image’s mean square error, peak signal-to-noise ratio, structural similarity, and so on. After denoising, the visual effect and edge retention ability of the image are greatly improved. With the gradual increase of the noise variance, the superiority of this method is further manifested, and it is suitable for the denoising of sonar images with high-density noise.

Abstract:User interests change over time. If the same level of privacy protection is used for data of all periods in the recommender systems, it is easy to introduce unnecessary noise and reduce data utility. Therefore, a differen? tial privacy protection recommendation algorithm based on the time weight factor is proposed. The algorithm first de? signs a time weight factor to measure the importance of data and then allocates the different privacy budgets to the data according to the time weight factor. That is, different intensity of privacy protection is performed on the data in different periods. Moreover, a probability matrix factorization model based on differential privacy is constructed for a personalized recommendation. Experimental results show that the proposed algorithm can preserve data utility more effectively and improve the accuracy of recommendation results under the condition of privacy protection.

Abstract:The recommender system aims to build a model from the user-item interaction and recommend the content of interest to users, so as to improve the user experience. However, most user-item sequences are not always sequentially related but have more flexible sequences and even noise. In order to solve this problem, a deep reinforce? ment learning sequence recommender algorithm based on strategy memory is proposed. The algorithm stores the user’s historical interaction in the memory network, and then uses a strategy network to divide the user′s current behavior pattern into short-term preference, long-term preference, and global preference, and introduces the attention mecha? nism to generate the corresponding user memory vector. The deep reinforcement learning algorithm is used to identify the projects with great benefits in the future. The strategy of the reinforcement learning network is continuously up? dated in the interaction between users and items to improve the accuracy of the recommender. Experiments on two public data sets show that the proposed algorithm improves the recall index by 8.87% and 11.20%, respectively, com? pared with the most advanced baseline algorithm.