Problem Of Poor Convergence Research Articles

Digital twin method for intelligent lifting of cable structures under multi-collaborative mode

In order to ensure the high efficiency and quality consistency of cable structure in each construction stage, this study proposes a digital twin method for intelligent lifting of cable structure under multi-collaborative mode. Firstly, a construction and operation methods of the digital twin logic model are formed. In the logical model, a relationship between the transmission and iteration of construction information in the collaborative process is defined. In order to realize the adaptability of working conditions, the coupling relationship between construction step and structural elevation is established. A cable structure lifting analysis model driven by an improved long short-term memory neural network is proposed, which effectively solves the problem of poor convergence of the traditional finite element method. An automatic time window selection method is formed for the neural network. An adaptive moment estimation method is used for neural network parameter optimization, which improves the efficiency and accuracy of the analysis. Taking the lifting process of a cable truss structure test model as an example, a digital twin collaborative construction platform is established to verify the effectiveness and feasibility of the research method. The results show that the digital twin logic model realizes the process collaboration, scene collaboration and data collaboration, and significantly improves the stability of construction quality and the linkage of construction process. Through the improved long short-term memory neural network, the coupling relationship between the construction step and the structural elevation is established, which accurately guides the whole construction process.

Free energies at QM accuracy from force fields via multimap targeted estimation.

Accurate predictions of ligand binding affinities would greatly accelerate the first stages of drug discovery campaigns. However, using highly accurate interatomic potentials based on quantum mechanics (QM) in free energy methods has been so far largely unfeasible due to their prohibitive computational cost. Here, we present an efficient method to compute QM free energies from simulations using cheap reference potentials, such as force fields (FFs). This task has traditionally been out of reach due to the slow convergence of computing the correction from the FF to the QM potential. To overcome this bottleneck, we generalize targeted free energy methods to employ multiple maps-implemented with normalizing flow neural networks (NNs)-that maximize the overlap between the distributions. Critically, the method requires neither a separate expensive training phase for the NNs nor samples from the QM potential. We further propose a one-epoch learning policy to efficiently avoid overfitting, and we combine our approach with enhanced sampling strategies to overcome the pervasive problem of poor convergence due to slow degrees of freedom. On the drug-like molecules in the HiPen dataset, the method accelerates the calculation of the free energy difference of switching from an FF to a DFTB3 potential by three orders of magnitude compared to standard free energy perturbation and by a factor of eight compared to previously published nonequilibrium calculations. Our results suggest that our method, in combination with efficient QM/MM calculations, may be used in lead optimization campaigns in drug discovery and to study protein-ligand molecular recognition processes.

Actor-Critic for Multi-Agent Reinforcement Learning with Self-Attention

The rapid development of deep reinforcement learning makes it widely used in multi-agent environments to solve the multi-agent cooperation problem. However, due to the instability of multi-agent environments, the performance is insufficient when using deep reinforcement learning algorithms to train each agent independently. In this work, we use the framework of centralized training with decentralized execution to extend the maximum entropy deep reinforcement learning algorithm Soft Actor-Critic (SAC) and proposes the multi-agent deep reinforcement learning algorithm MASAC based on the maximum entropy framework. Proposed model treats all the agents as part of the environment, it can effectively solve the problem of poor convergence of algorithms due to environmental instability. At the same time, we have noticed the shortcoming of centralized training, using all the information of the agents as input of critics, and it is easy to lose the information related to the current agent. Inspired by the application of self-attention mechanism in machine translation, we use the self-attention mechanism to improve the critic and propose the ATT-MASAC algorithm. Each agent can discover their relationship with other agents through encoder operation and attention calculation as part of the critic networks. Compared with the recent multi-agent deep reinforcement learning algorithms, ATT-MASAC has better convergence effect. Also, it has better stability when the number of agents in the environment increases.

Fault data expansion method of permanent magnet synchronous motor based on Wasserstein-generative adversarial network

Aiming at the characteristics of non-smooth, non-linear, multi-source heterogeneity, low density of value and unevenness of fault data collected by the online monitoring equipment of permanent magnet synchronous motor (PMSM), and the difficulty of fault mechanism analysis, this paper proposes a method of PMSM data expansion based on the improved generative adversarial network. First, use the real fault data of the motor to train the model to obtain a mature and stable generative countermeasure network. Secondly, use the generative countermeasure network model to test the remaining data and generate pseudo samples. Finally, use the two-dimensional data analysis method and the time-domain analysis method to generate validity analysis of samples. Aiming at the characteristics of unbalanced motor data, the data expansion method of inter-turn short-circuit faults is carried out based on the data expansion method of the improved generative countermeasure network, and the two-dimensional data analysis method and the time-domain analysis method are used for analysis. The experimental results show that the improved Wasserstein-Generative Adversarial Network (W-GAN) has a better ability to generate fake data, which provides a data basis for the mechanism analysis and machine fault diagnosis of PMSMs. Data analysis results show that the improved W-GAN effectively solves the problem of poor convergence of GAN.

Energy management strategy via maximum entropy reinforcement learning for an extended range logistics vehicle

The modern energy management strategy (EMS) plays a vital role in the energy efficiency of the extended range electric vehicle. However, some modern strategies such as model predictive control (MPC) and dynamic programming (DP) have limited practical potential because they are subject to the pre-known environment information and noise interference. The reinforcement learning (RL)control strategy can be adopted as online control to interact with the vehicle and the environment. In this study, a novel auxiliary power unit (APU) charging strategy with multi-object optimization is proposed to achieve high fuel conversion efficiency while maintaining battery charging health. The state-of-the-art algorithm, Soft Actor-Critic (SAC), is applied to achieve better exploration of the possible APU behaviour and solve the sensitivity and poor convergence problems from the current RL studies. Its performance is further verified by the results of the Deep Deterministic Policy Gradient (DDPG) algorithm and DP. Three innovative targets are selected as the RL rewards for optimization: the engine fuel rate, SOC charging trajectory, and the battery charging rate (C-rate). The first adoption of the battery C-rate monitoring in RL-based energy management strategy helps extend the battery lifespan from excessive discharge. The comparative results show that the SAC had a 36% faster convergence speed than DDPG while providing a smoother and more stable action space. The fuel consumption with SAC also outplays that of DDPG by around 3%, which achieves almost 95% of the global optimization result. The successful deployment of the SAC algorithm as an EMS indicates its standout ability in dealing with wide-range actions and states with high randomness, revealing the practical potential compared with the existing RL strategies.

Structural Damage Identification Based on Improved Fruit Fly Optimization Algorithm

To locate the damage of the structure efficiently and judge the damage degree, this paper proposes an improved fruit fly optimization algorithm (IFOA). Aiming at the problem of poor convergence of the standard fruit fly optimization algorithm in the face of complex structure damage identification, the IFOA introduces the concept of collaborative search of two subpopulations. The IFOA divides the entire population into positive subgroups and negative subgroups based on the individual taste concentration results. Among them, the positive subgroup uses the improved dynamic adaptive search step size to perform a fine search locally to improve its local search ability. Negative subgroups continue to use the standard fruit fly optimization algorithm for optimization, taking advantage of the powerful global search capabilities of the standard fruit fly optimization algorithm. It enables the algorithm to balance global and local search capabilities, prevents the algorithm from falling into local optimum, and speeds up the convergence speed and accuracy of the algorithm. Simulation results show that IFOA can effectively identify the damage location and damage degree of the structure, and it still performs well when facing the complex steel truss damage identification.

Research on robot optimal path planning method based on improved ant colony algorithm

To overcome the problem of poor convergence and obstacle avoidance when traditional methods are used to plan the optimal path of robot, a new optimal path planning method based on improved ant colony algorithm is proposed. Firstly, the odometer model, ultrasonic sensor model and robot motion model are constructed to obtain the environmental information and robot motion state information. Then, according to the adaptive transformation heuristic function of the target point and the principle of wolf swarm assignment, the pheromone is refreshed. On this basis, the core parameters of the improved ant colony algorithm are optimised by particle swarm algorithm, so as to complete the optimal path planning of the robot. The experimental results show that the overall mean value of collision avoidance of the proposed method is 0.97, and the planning performance is significantly better than that of similar planning methods, with considerable application value.

Wavelet features embedded convolutional neural network for multiscale ear recognition

The ear has emerged as a new biometric trait to recognize humans from their profile faces. Stability over the years, noninvasive capturing process, expressionless images, and significant variation in shape among individuals make the ear a suitable choice when compared with other biometrics. Convolutional neural network (CNN)’s capability to learn and discriminate specific features irrespective of image variation makes it the most suitable choice for ear recognition in comparison to local geometrical feature extraction methods. The recognition performance affecting factors such as occlusion, image resolution, and head rotation during ear acquisition in an uncontrolled environment makes it a challenging task, and the use of spatial features only in CNN limits the recognition rate. We explored the various CNN configurations and propose the spectral–spatial feature-based CNN. An embedding algorithm is proposed to fuse multilevel spectral information from the image domain with spatial features extracted at each deep layer of the CNN network. The poor convergence problem of CNN is tackled by introducing batch normalization after every convolution. The analysis of the result shows that the proposed framework is superior among all CNN achieving 98.15% and 78.88% accuracy on the Annotated Web Ears with 1000 images and earVN1.0 with 28,412 images, respectively. In addition to fine-tuning and configuring the layers of CNN, the fusion of missing spectral features and spatial features can strengthen the CNN.

Spectral-Spatial Features Integrated Convolution Neural Network for Breast Cancer Classification.

Cancer identification and classification from histopathological images of the breast depends greatly on experts, and computer-aided diagnosis can play an important role in disagreement of experts. This automatic process has increased the accuracy of the classification at a reduced cost. The advancement in Convolution Neural Network (CNN) structure has outperformed the traditional approaches in biomedical imaging applications. One of the limiting factors of CNN is it uses spatial image features only for classification. The spectral features from the transform domain have equivalent importance in the complex image classification algorithm. This paper proposes a new CNN structure to classify the histopathological cancer images based on integrating the spectral features obtained using a multi-resolution wavelet transform with the spatial features of CNN. In addition, batch normalization process is used after every layer in the convolution network to improve the poor convergence problem of CNN and the deep layers of CNN are trained with spectral–spatial features. The proposed structure is tested on malignant histology images of the breast for both binary and multi-class classification of tissue using the BreaKHis Dataset and the Breast Cancer Classification Challenge 2015 Datasest. Experimental results show that the combination of spectral–spatial features improves classification accuracy of the CNN network and requires less training parameters in comparison with the well known models (i.e., VGG16 and ALEXNET). The proposed structure achieves an average accuracy of 97.58% and 97.45% with 7.6 million training parameters on both datasets, respectively.

Dynamic Soft Sensor Development for Time-Varying and Multirate Data Processes Based on Discount and Weighted ARMA Models

To solve the soft sensor modeling (SSMI) problem in a nonlinear chemical process with dynamic time variation and multi-rate data, this paper proposes a dynamic SSMI method based on an autoregressive moving average (ARMA) model of weighted process data with discount (DSSMI-AMWPDD) and optimization methods. For the sustained influence of auxiliary variable data on the dominant variables, the ARMA model structure is adopted. To reduce the complexity of the model, the dynamic weighting model is combined with the ARMA model. To address the weights of auxiliary variable data with different sampling frequencies, a calculation method for AMWPDD is proposed using assumptions that are suitable for most sequential chemical processes. The proposed method can obtain a discount factor value (DFV) of auxiliary variable data, realizing the dynamic fusion of chemical process data. Particle swarm optimization (PSO) is employed to optimize the soft sensor model parameters. To address the poor convergence problem of PSO, ω-dynamic PSO (ωDPSO) is used to improve the PSO convergence via the dynamic fluctuation of the inertia weight. A continuous stirred tank reactor (CSTR) simulation experiment was performed. The results show that the proposed DSSMI-AMWPDD method can effectively improve the SSM prediction accuracy for a nonlinear time-varying chemical process. The AMWPDD proposed in this paper can reflect the dynamic change of chemical process and improve the accuracy of SSM data prediction. The ω dynamic PSO method proposed in this paper has faster convergence speed and higher convergence accuracy, thus, these models correlate with the concept of symmetry.

Multi-sensor cross cueing technology based on grey wolf algorithm

In order to solve the problem of poor convergence and low precision of the current Multi-sensor Detection Network algorithm, a multi-sensor cross cueing technology based on Grey Wolf algorithm is proposed. By endowing the grey wolf with autonomous decision-making ability, the algorithm can jump out of the local optimal solution as soon as possible, and the convergence degree of the algorithm is improved. The simulation results show that the algorithm can find the optimal solution in a short time and maintain good stability, and the optimization ability is further strengthened. By comparing with other algorithms, it is more effective to prove that the algorithm has better solution quality.

Multidisciplinary hybrid hierarchical collaborative optimization of electric wheel vehicle chassis integrated system based on driver’s feel

The optimization design of chassis integrated system mainly involves steering, suspension and brake subsystems, which is essentially a multidisciplinary design optimization. This paper mainly researches the multidisciplinary optimization of the chassis integrated system for the electric wheel vehicle, from the view of ensuring a favorable feel for the driver. The dynamic models of differential steering system, brake system, active suspension system and vehicle are established. Then, taking the coupling relationship of the chassis subsystems into account, this paper proposes an evaluating index of driver’s ride comfort (Drc), which is composed of the steering road feel, brake feel and suspension ride comfort. In order to determine the weight coefficient in the quantization formula of Drc, the technique for order preference by similarity to ideal solution (TOPSIS) method is used to overcome the subjectivity in the selection. Based on these, a multidisciplinary hybrid hierarchical collaborative optimization (HHCO) method is proposed on the basis of the collaborative optimization (CO), which consists of a system level coordinator and a coupling analyzer to solve the problem of poor convergence and the low efficiency of CO method. The optimization results show that the proposed HHCO method has excellent computational efficiency and better convergence compared with the CO method, which can further improve the steering road feel and the drive ride comfort, on the premise of ensuring the brake feel and suspension ride comfort.

A Hybrid MoM/FDTD Method for Dosimetry of Small Animal in Reverberation Chamber

A hybrid approach combining the method of moment (MoM) and the finite-difference time-domain (FDTD) method is applied to a problem of evaluating specific absorption rates (SARs) in a small rat inside a reverberation chamber (RC). The hybrid method is an alternative technique used to overcome the problem of poor convergence for a solution of the FDTD method in analyzing the RC. First, the RC with a dipole or a helical antenna is designed numerically to function properly at 2 GHz. Then, the MoM/FDTD method is used to calculate the whole-body average SAR (WBA-SAR) for a small animal inside an open cavity. The results are compared with those obtained with the FDTD method in order to verify accuracy. The SAR distributions when stirrers are rotating are also determined. Finally, it is numerically found that the ratio of the WBA-SAR of the 118 g rat to the mean squared electric field is 0.149 (mW/kg)/(V/m) <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> and an input antenna power of 2.2 W is required to produce a WBA-SAR of 4 W/kg for our designed chamber.

Research on Multi-Objective Reservoir Optimal Operation Based on Adaptive Genetic Algorithms

The comprehensive reservoir scheduling is based on multi-objective reservoir operation, should try to coordinate between each target scheduling of progression. In this paper, based on the characteristics of scheduling period, this paper proposes a reservoir optimal operation based on adaptive genetic algorithm solution of the problem. Because of the adaptive genetic algorithm can during evolution according to individual fitness and dispersion degree of genetic control parameters are adjusted automatically, can better solve encountered in the application of the standard genetic algorithm in the problem of poor convergence and prematurity.

An Improved Road Roughness Simulation Method Based on Trigonometric Series Method

Accurate road data is critical to vehicle ride comfort research. Trigonometric series (TS) method is widely used in road roughness construction, but the accuracy of the road data in time domain is not high and the power spectrum density (PSD) exits poor convergence problem. To solve the above problems, this paper has made two improvements: generate random frequency in each band using the local PSD function, optimize phase angle using the principle of minimum energy. The simulation results show that the improved method has higher precise and better convergence of PSD.

Local-gauge finite-element method for electron waves in magnetic fields

The finite-element method (FEM) has already been extended to analyze transport properties of electron waves of two-dimensional electron systems in magnetic fields. Although many researchers have created new formulations or improvements to this method, few have analyzed how this method is applied to realistic systems. The present paper suggests that conventional formulations of the FEM do not give accurate results for large systems or for strong magnetic fields; in addition, it suggests that the selected gauge significantly influences the numerical results. Furthermore, this paper proposes a conceptually different formulation of the FEM that solves the poor convergence problem. This formulation is simple: matrix elements are multiplied by the Peierls phase in the absence of a magnetic field. To show the advantages of this formulation, numerical examples are presented.

Group Decision Making Coordinating Characteristic Analysis and Modeling of Water Resource Allocation and Scheduling

Group decision-making management is an important issue in water management reform and development. The lacking of effective communication and cooperation is the major defects of the existing group decision-making models. Based on the in-depth analysis of the coordinating characteristic in group decision making, this study proposed a multi-layer dynamic model of water resource allocation and scheduling. This model focuses on effective communication and coordination. In order to solve the problem of poor convergence of multi-round decision-making process in water resource allocation and scheduling, the scheme-recognized cooperative satisfaction index and scheme-adjusted rationality index are introduced. An optimization algorithm based on the effective distance of group utility is proposed, which can solve the problem about coordination of limited resources-based group decision-making process. The simulation results show that the proposed model has better convergence than the existing models.

Analysis of multi-layer optical waveguides using a pseudospectral technique

A pseudospectral method is presented for the determination of propagation characteristicsin multi-layer optical waveguides. The method is based on transforming the Helmholtzeigenvalue equation from the spatial domain into the spectral domain and solving thetransformed equation in the spectral domain as an unbounded interval problem employinga point matching technique. Appropriate basis functions which take into account thephysical properties of the electromagnetic fields inside multi-layer optical waveguides areintroduced. Representing the constitutive relation with a convolution in the spectraldomain, we overcome the poor convergence problem of conventional pseudospectralmethods, which is due to abrupt changes in refractive-index profiles. The presented methodimplicitly includes all the boundary conditions of the problem. Its application to severalmulti-layer optical waveguides is demonstrated to verify the accuracy and efficiency of themethod. All the results are found to be in excellent agreement with those of analyticalsolutions.

An Approach for Short-Term Prediction on Time Series from Parameter-Varying Systems

Time series prediction is a very important problem in many applications and the current prediction techniques are nearly all based on the Takens' embedding theorem. Many realistic systems are parameter-varying systems, and the embedding theorems are invalid, predicting the behavior of parameter-varying systems is more difficult. This paper proposes the novel prediction techniques for parameter-varying systems reconstruction, which are based on wavelet neural network (WNN) and multiwavelets neural network (MWNN). These techniques absorb the advantages of high resolution of wavelet and learning of neural networks. The significant improvement is that the error's functions of both networks are convex, and the problem of poor convergence and undesired local minimum can be solved remarkably. Ikeda time series generated by the parameter-varying systems is adopted to check the prediction performance of the proposed models. The numerical experiments show that the three proposed models are feasible, MWNN has the top performance, and WNN could lead the better results than NN in the prediction of the parameter-varying systems.

SOURCE LOCALIZATION WITH MINIMAL NUMBER OF NETWORKED SENSORS

Based on concentration measurements from spatially distributed electronic noses, the location of a point source is to be determined. It is assumed that the emitted substance is transported by advection caused by a known homogeneous wind field and by diffusion. A new two-step approach for solving the source localization problem is presented. The new approach overcomes the problem of poor convergence of iterative algorithms, which try to minimize the least squares output error. Finally the problem of minimal sensor number and optimal sensor placement is investigated.