Excessive Iterations Research Articles

Human-in-the-loop optimization for vehicle body lightweight design

Automatic optimization algorithms are crucial for vehicle body lightweight design; however, existing methods remain inefficient leading to excessive iterations that increase both time and costs. Current interactive optimization strategies partially mitigate this issue but lack a broad range of manipulation points and auxiliary information models. As such, we introduce a novel approach, “Human-in-the-Loop based method for Vehicle Body Lightweight Design” (HIL-VBLD). This method integrates human decision-making with optimization algorithms to reduce unproductive iterations. HIL-VBLD comprises two key components: (1) an innovative interaction mode that provides multiple manipulation points including constraint modification, algorithm switching, and selection of solutions of interest (SOI); (2) A comprehensive auxiliary information model that supports decision-making for designers. Our analysis demonstrates HIL-VBLD’s efficacy, showing a 54.5 % reduction in iteration cycles for genetic algorithm using SOI selection. Algorithm switching led to a 4.5 % mass reduction, mitigating local optimum pitfalls associated with gradient algorithms. Additionally, the auxiliary information model achieved a further 1.25 % mass reduction, enhancing optimization robustness. Compared to conventional automatic algorithm switching strategies, HIL-VBLD maintains equivalent accuracy with 23.9 % fewer iterations.

Efficient object detector via dynamic prior and dynamic feature fusion

Abstract Sparse R-CNN is a new paradigm of object detection, which predicts objects in a sparse way. However, there are some limitations in Sparse R-CNN. One is the presence of weak prior information caused by fixed learnable proposal boxes and features across different images, necessitating excessive iterations for the model to refine its predictions; the other is the inadequate exploitation of multi-scale information, leading to the sub-optimal detection performance. Thus, building upon Sparse R-CNN, we propose an efficient detector that incorporates dynamic prior and dynamic feature fusion, called $D^{2}$-Det. In particular, for the dynamic prior part, a prior information generator module dynamically generates proposal features and boxes as the dynamic prior for different images to alleviate the inference-inefficient iterative refinement process of predictions, and we further propose the class scores decoupling method to reduce the computation overhead. Furthermore, for the dynamic feature fusion part, we develop a novel lightweight multi-scale feature fusion module, which dynamically aggregates features from all layers for each proposal box, enabling adaptive feature fusion and improving detection precision by nearly 2 AP. Experiments show that $D^{2}$-Det can achieve 46.6 AP on COCO 2017 with fewer computations for the backbone ResNet50, surpassing most of the state-of-the-art detectors.

Topology optimization with graph neural network enabled regularized thresholding

Topology optimization algorithms often employ a smooth density function to implicitly represent geometries in a discretized domain. While this implicit representation offers great flexibility to parametrize the optimized geometry, it also leads to a transition region. Previous approaches, such as the Solid Isotropic Material Penalty (SIMP) method, have been proposed to modify the objective function aiming to converge toward integer density values and eliminate this non-physical transition region. However, the iterative nature of topology optimization renders this process computationally demanding, emphasizing the importance of achieving fast convergence. Accelerating convergence without significantly compromising the final solution can be challenging. In this work, we introduce a machine learning approach that leverages the message-passing Graph Neural Network (GNN) to eliminate the non-physical transition zone for the topology optimization problems. By representing the optimized structures as weighted graphs, we introduce a generalized filtering algorithm based on the topology of the spatial discretization. As such, the resultant algorithm can be applied to two- and three-dimensional space for both Cartesian (structured grid) and non-Cartesian discretizations (e.g. polygon finite element). The numerical experiments indicate that applying this filter throughout the optimization process may avoid excessive iterations and enable a more efficient optimization procedure.

Bionic 3D Path Planning for Plant Protection UAVs Based on Swarm Intelligence Algorithms and Krill Swarm Behavior.

The protection of plants in mountainous and hilly areas differs from that in plain areas due to the complex terrain, which divides the work plot into many narrow plots. When designing the path planning method for plant protection UAVs, it is important to consider the generality in different working environments. To address issues such as poor path optimization, long operation time, and excessive iterations required by traditional swarm intelligence algorithms, this paper proposes a bionic three-dimensional path planning algorithm for plant protection UAVs. This algorithm aims to plan safe and optimal flight paths between work plots obstructed by multiple obstacle areas. Inspired by krill group behavior and based on group intelligence algorithm theory, the bionic three-dimensional path planning algorithm consists of three states: "foraging behavior", "avoiding enemy behavior", and "cruising behavior". The current position information of the UAV in the working environment is used to switch between these states, and the optimal path is found after several iterations, which realizes the adaptive global and local convergence of the track planning, and improves the convergence speed and accuracy of the algorithm. The optimal flight path is obtained by smoothing using a third-order B-spline curve. Three sets of comparative simulation experiments are designed to verify the performance of this proposed algorithm. The results show that the bionic swarm intelligence algorithm based on krill swarm behavior reduces the path length by 1.1~17.5%, the operation time by 27.56~75.15%, the path energy consumption by 13.91~27.35%, and the number of iterations by 46~75% compared with the existing algorithms. The proposed algorithm can shorten the distance of the planned path more effectively, improve the real-time performance, and reduce the energy consumption.

Target temperature-based simulation method for predicting performance and designing aircraft thermal anti-icing systems

The prevention of ice build-up on aircraft surfaces is crucial for safe operations. Thermal ice protection systems, particularly electro-thermal systems, are commonly employed to mitigate ice formation. However, the weight of electric batteries used in these systems significantly impacts aircraft performance, necessitating the minimization of power load during design. Despite this requirement, effective anti-icing operation mandates a sufficient heat flux to maintain surface temperature above the freezing point and evaporate any runback water film. Previous research has utilized numerical simulation analysis to predict the performance of electro-thermal anti-icing systems. The conjugate heat transfer method has been commonly employed to predict temperature distribution on the protected surface. However, non-isothermal surfaces exhibit distinct local convective heat transfer coefficients that influence local thermal balance, unlike isothermal surfaces that adopt the loosely-coupled method for computational efficiency. In this study, we present a novel target temperature-based anti-icing simulation method that predicts the required heat flux for thermal ice protection systems without employing the conjugate heat transfer method. Our objective was to calculate the necessary heat flux to achieve the desired non-isothermal surface temperature for effective anti-icing. By eliminating the excessive iteration process associated with the coupling method for conjugate heat transfer, we have streamlined the simulation procedure. Through validation, target temperature-based anti-icing simulation method, when compared to the tight coupling method, demonstrated the ability to predict the same surface water flow rates while reducing computation time by up to 90 %. Furthermore, when applied to design problems, our method achieved approximately a 10 % reduction in the required heat capacity for anti-icing compared to experimental values. As a result, target temperature-based anti-icing simulation method offers a more efficient alternative to traditional conjugate heat transfer methods, providing valuable insights for both predicting anti-icing system performance and optimizing aircraft design.

Cluster-based multicast optimized routing in VANETs using elite knowledge-based genetic algorithm

The advancement of autonomous and driver-assistance technology has greatly improved the experience of driving. With the growing use of vehicles, there is a greater need for effective communication among them to ensure both applications in Vehicular Ad-hoc Networks (VANETs) for safety and non-safety. VANETs commonly prioritize multicast communication to efficiently use computational resources and achieve a desirable Quality of Service (QoS). However, the dynamic nature of VANETs, marked by high mobility and frequent topology changes, presents significant challenges in establishing and sustaining multicast communication. To ensure that sensing data from all target points is reliably sent to the base station, it is crucial to address the important factors of energy balancing, load balancing, connectivity, and coverage. These factors play a vital role in achieving a balanced and efficient transmission system. In this research paper, clustering-based multicast routing is proposed that utilizes an enhanced Genetic Algorithm Elite Knowledge Sharing Genetic Algorithm (EKSGA) to choose cluster heads. The EKSGA identifies the top 5 % performers with the highest fitness score and generates a new population by exchanging genetic information, aiming to establish a population with the greatest potential for optimal fitness. The performance assessment of the suggested protocol demonstrates significant enhancements in network lifetime and throughput compared to various existing protocols, including LEACH, DEEC, p-WOA, DDEEE, ECBLTR, and GA with improvements ranging from 2.4 % to 22 % in network lifetime and from 3.2 % to 23.2 % in throughput for a system consisting of 500 nodes with energy level set at 0.1 J for rounds 300. Additionally, the proposed approach utilizes the existing population to generate a more fit population, thereby reducing the need for excessive iterations and computational overhead.

XMKR: Explainable manufacturing knowledge recommendation for collaborative design with graph embedding learning

Design for manufacturability is crucial to ensure high-quality products. It fully considers manufacturing-related factors in a collaborative design manner. However, since manufacturing knowledge is unstructured and hard to reuse, designers mainly depend on frequent design iterations to meet complex manufacturing constraints. Such insufficient design-manufacturing collaboration is not conducive to design efficiency improvement. To alleviate the problem, this paper proposes a novel graph embedding-based approach of Explainable Manufacturing Knowledge Recommendation (XMKR) for collaborative product design, which could achieve designer-oriented manufacturing knowledge reuse to avoid design errors proactively and minimize excessive iterations. Firstly, a graph embedding model based on the graph neural network (MKGE-GNN) is proposed to learn implicit semantic information of manufacturing knowledge. Secondly, with MKGE-GNN learning results, the potential manufacturing knowledge preferences of designers are identified in the design context graph by Personalized PageRank to achieve context-aware explainable manufacturing knowledge recommendation. Finally, a car body collaborative design case is conducted to validate the efficacy of the proposed approach, which not only provides accurate manufacturing knowledge in designers’ decision-making, but also enhances designers’ cognitions for manufacturability in the dynamic design context.

Research on incentive mechanisms for anti-heterogeneous federated learning based on reputation and contribution

<abstract> <p>An optimization algorithm for federated learning, equipped with an incentive mechanism, is introduced to tackle the challenges of excessive iterations, prolonged training durations, and suboptimal efficiency encountered during model training within the federated learning framework. Initially, the algorithm establishes reputation values that are tied to both time and model loss metrics. This foundation enables the creation of incentive mechanisms aimed at rewarding honest nodes while penalizing malicious ones. Subsequently, a bidirectional selection mechanism anchored in blockchain technology is developed, allowing smart contracts to enroll nodes with high reputations in training sessions, thus filtering out malicious clients and enhancing local training efficiency. Furthermore, the integration of the Earth Mover's Distance (EMD) mechanism serves to lessen the impact of non-IID (non-Independent and Identically Distributed) data on the global model, leading to a reduction in the frequency of model training cycles and an improvement in model accuracy. Experimental results confirm that this approach maintains high model accuracy in non-IID data settings, outperforming traditional federated learning algorithms.</p> </abstract>

Iterative singular value decomposition-based in-band denoising approach with envelope order analysis for sun gear fault diagnosis of planetary system under varying speed

The vibration signal of the planetary gear train is easily disturbed by the background noise, so the measured signal is complicated. In order to accurately extract the fault features, the resonance method has been widely used. However, even if the optimal resonance frequency band is found, the in-band noise still exists, so it is necessary to study an effective in-band denoising method. In this paper, an in-band denoising method based on Iterative Singular Value Decomposition (ISVD) is proposed for the fault diagnosis of the secondary sun gear of a planetary gear train, combined with the envelope order spectrum analysis. This method uses the enhanced Wavelet Packet Transform Spectral Kurtosis (WPTSK) to determine the best frequency band for the signal, and uses the ISVD method to realize the signal denoising. It sets a threshold to avoid the destruction of the useful information caused by the excessive iteration, and uses the relationship between the singular values and frequency components to reconstruct the denoised signal. Finally, the signal is converted to the fault characteristic order domain by resampling to identify the fault type of the sun gear from the envelope order spectrum. The simulation and experimental results show that compared with the Empirical Mode Decomposition (EMD) and Variational Mode Decomposition (VMD), the ISVD can effectively suppress the in-band noise and more accurately extract the fault characteristic order of the secondary sun gear under varying speed.

The efficient conversion between linear wave function expansions and nonlinear graphically contracted function expansions

Several algorithms are examined for the transformation of a graphically contracted function (GCF) to a configuration state function (CSF) vector, denoted , and for the reverse transformation from CSF vector to GCF, denoted . The initial implementation of these algorithms is applied to a sequence of full-CI wave functions of similar molecules. The most efficient transformation is based on a depth-first search (DFS) and scales as where is an average facet count for the GCF wave function, is the dimension of the vector x, and is the dimension of the GCF basis. Two different transformations are proposed, one based on a pairwise recursive merge algorithm and the other based on a least-squares fitting algorithm. The recursive merge is a reliable algorithm, but our initial implementation does not display the expected scaling behaviour with respect to . The iterative least-squares fitting algorithm is based on an efficient DFS procedure to compute the error ε and its gradient g. However, the resulting iterative least-squares optimisation is slowly convergent, requiring excessive iterations to converge to a target error tolerance. For most of the calculations in this case study, the recursive merge procedure is found to be more efficient than the least-squares optimisation procedure.

Fast warm-start of F-MPC strategy for automotive cruise control with mode switching

A fast iterative algorithm of nonlinear model predictive control is proposed to solve the warm-start problem of the fast model predictive control (F-MPC) method for real-time application with mode switching. Because an incorrect initial guess of F-MPC creates problems such as excessive iterations or even solution failures, the proposed iterative algorithm aims to provide the initial value of F-MPC quickly and efficiently. The idea is to decompose the original nonlinear system into a main linear part and a nonlinear part, which is regarded as measurable disturbance. Then the explicit optimal solution of the “disturbed” system is derived, the obtained series of control inputs are applied to the system, the system state and consequently the “disturbance” are updated, and finally the process is repeated until convergence. The calculated iterative result can be used as the initial solution for F-MPC especially for maneuver mode switching. Automotive cruise control is used as an example for validation, and it is shown that the control strategy has superior adaptability for mode switching, guaranteeing the given safety constraints as well as significantly reducing the computational load.

Fast 3D transient electromagnetic forward modeling using BEDS-FDTD algorithm and GPU parallelization

The time-step sizes of the conventional finite-difference time-domain (FDTD) method are severely restricted by the Courant-Friedrichs-Lewy stability condition. This may result in excessive iterations, high cost and large runtime, and increasing late-stage cumulative errors in 3D transient electromagnetic (TEM) forward modeling. At present, forward modeling speed is a major obstacle in the development of 3D TEM inversion. To address this issue, a new 3D TEM forward modeling algorithm is presented. In the new algorithm, the backward Euler (BE) scheme is used with a view toward relaxing the stability constraint. However, after using the BE approach, a huge sparse matrix needs to be inverted in each iteration. Directly solving such matrices by Gaussian elimination or an iterative method is still time-consuming and requires large amounts of computer memory. To improve computational efficiency, the direct-splitting (DS) strategy is adopted to transform the large sparse matrices into a series of small diagonally dominant tridiagonal matrices, which are easier to solve. Then, the complex-frequency-shifted perfectly matched layer boundary condition is implemented using the bilinear transform method to reduce the size of the model. To further speed up the calculation process, the codes are then programmed to run on graphics processing unit (GPU) devices. Three models are used for calculations, and the results are compared with other numerical methods to verify the accuracy of our algorithm. After that, to demonstrate the BEDS-FDTD algorithm’s ability to model complex earth structures, a model with a very complex shape is used for the calculation via the conformal mesh technique. In addition, the modeling speed is tested using different GPU devices. We find that an NVIDIA Tesla A100 is able to calculate the 50 × 50 × 50 cell model in only 8 s.

Digital twin model of gas turbine and its application in warning of performance fault

The digital twin-driven performance model provides an attractive option for the warn gas-path faults of the gas turbines. However, three technical difficulties need to be solved: (1) low modeling precision caused by individual differences between gas turbines, (2) poor solution efficiency due to excessive iterations, and (3) the false alarm and missing alarm brought by the traditional fixed threshold method. This paper proposes a digital twin model-based early warning method for gas-path faults that breaks through the above obstacles from three aspects. Firstly, a novel performance modeling strategy is proposed to make the simulation effect close to the actual gas turbine by fusing the mechanism model and measurement data. Secondly, the idea of controlling the relative accuracy of model parameters is developed. The introduction of an error module to the existing model can greatly shorten the modeling cycle. The third solution focuses on the early warning based on the digital twin model, which self-learns the alarm threshold of the warning feature of gas-path parameters using the kernel density estimation. The proposed method is utilized to analyze actual measured data of LM2500 +, and the results verify that the new-built digital model has higher accuracy and better efficiency. The comparisons show that the proposed method shows evident superiority in early warning of performance faults for gas turbines over other methods.

A regularized fast multipole method of moments for rapid calculation of three-dimensional time-harmonic electromagnetic scattering from complex targets

The article applies the recent research results of the singular boundary method to rapidly calculate time-harmonic electromagnetic scattering from complex targets. The study focuses on how to efficiently solve the large-scale and completely dense linear equations obtained by the regularized method of moments. The core innovation point is to propose a regularized fast multipole method of moments by combining the regularized method of moments and the fast multipole method. Several key mechanical issues encountered in simulation of electromagnetic scattering are discussed. The Rao-Wilton-Glisson vector basis function is used as the basis function. A modified fundamental solution is employed in place of the fundamental solution of the Helmholtz equation to avoid non-uniqueness at interior resonances. The fast multipole method is used to expedite the solution process. A fast near-field approximation preconditioner is introduced to greatly reduce excessive iterations. The origin intensity factor is applied to calculate the singular terms of interpolation matrix. The radar cross section of the NASA almond and a scaled stealth fighter has been accurately calculated by the proposed method.

A Two-Step Projected Iterative Algorithm for Tropospheric Water Vapor Tomography

The tropospheric tomography is an ill-posed inversion problem due to the sparsity of Global Navigation Satellite Systems (GNSS) stations and the limitation on the projection angles of GNSS signals, which in turn affects the stability and robustness of the tomographic solution. To address this, a new tomographic algorithm, named two-step projected iterative algorithm (TSPIA), is proposed. The wet refractivity (WR) field was constructed in two steps: first, an iterative preprocessing for the initial input values was performed, then its resultant solution was input into the projected iterative method, in which a hypothesis convex set was constructed to constrain the reconstruction based on the classical algebraic iterative reconstruction (AIR) methods. In addition, a two-dimensional normalized cumulative periodogram (2D-NCP) termination criterion was investigated since the traditional criteria for judging the convergence of iterations use pre-fixed empirical thresholds, which may lead to excessive iterations and need complicated work. The TSPIA was tested using GNSS data in Hong Kong over a wet period and a dry period. Statistical results showed that, compared to the classical AIR methods, the accuracy of the reconstructed WR field of the TSPIA were improved by about 10% and 15% when radiosonde and ECMWF data were used as the reference, respectively. Moreover, experiments for the proposed 2D-NCP criterion demonstrated noticeable computational efficiency. These results suggest that the new approaches proposed in this study can improve the performance of the iterative methods for GNSS tropospheric tomography.

A parameter-optimized variational mode decomposition method using salp swarm algorithm and its application to acoustic-based detection for internal defects of arc magnets

The acoustic-based detection is regarded as an effective way to detect the internal defects of arc magnets. Variational mode decomposition (VMD) has a significant potential to provide a favorable acoustic signal analysis for such detection. However, the performance of VMD heavily depends on the proper parameter setting. The existing optimization methods for determining the optimal VMD parameter setting still expose shortcomings, including slow convergences, excessive iterations, and local optimum traps. Therefore, a parameter-optimized VMD method using the salp swarm algorithm (SSA) is proposed. In this method, the relationship between the VMD parameters and their decomposition performance is quantified as a fitness function, the minimum value of which indicates the optimal parameter setting. SSA is used to search for such a minimum value from the parameter space. With the optimized parameters, each signal can be decomposed accurately into a series of modes representing signal components. The center frequencies are extracted from the selected modes as feature data, and their identification is performed by random forest. The experimental results demonstrated that the detection accuracy is above 98%. The proposed method has superior performance in the VMD parameter optimization as well as the acoustic-based internal defect detection of arc magnets.

A secondary selection-based orthogonal matching pursuit method for rolling element bearing diagnosis

Sparse representation based on the matching pursuit (MP) algorithm is an effective method for fault feature extraction involving rolling element bearings. However, in the sparse decomposition stage, the MP algorithm is extremely susceptible to both noise in the residual signal and excessive iterations selecting some atoms that are not related to the fault impulses, thus causing interference components in the reconstructed signal. Considering these problems, a secondary selection-based orthogonal MP (SS-OMP) algorithm is proposed in this paper. In the proposed method, all the atoms selected to decompose the residual signal are selected for a second time, and only those atoms directed by the fault impulses are retained. After the decomposition of the residual signal, these retained atoms are used to redecompose the vibration signal, and then the fault impulses in the vibration signal are extracted. The superiority of the proposed method is verified by a numerical simulation study and experimental analysis.

A Novel Meshing Method Based on Adaptive Size Function and Moving Mesh for Electromagnetic Finite Element Analysis

A moving meshing algorithm with mesh adaptive size function was proposed in this paper with regard to the modeling speed and solution accuracy of electromagnetic equipment in the optimization design process. In the proposed method, a mesh size function that considers curvature, feature size, and distance gradient restrictions is constructed, which can obtain high quality meshes and avoid excessive iteration; when the finite element mesh domain is deformed, only the mesh nodes close to the moving boundary are allowed to move, and the theory of force-balance is used combined with the second-order boundary projection algorithm to perform iterative optimization of the mesh node positions. The proposed method has the advantages of keeping the original mesh structure and minimum mesh deformation as well as speed up the convergence, save time in the finite element meshing, and ensure the quality of the generated mesh. Then, the proposed method was applied to a 37 kw motor for electromagnetic analysis, and the results obtained proved the accuracy of the algorithm; finally, the effectiveness of the mesh movement algorithm in three-dimensional space was tested by moving the sphere inside the cylinder.

Review of Practices to Integrate Aircraft Mass Properties Management and Development Processes

The mass properties of aircraft directly influence their performance and costs, and are particularly subject to high uncertainties in the early phases of the development process. As aircraft systems become more detailed, their mass properties are iteratively updated. Those updates, in turn, may lead to rework on aircraft systems. To avoid excessive iterations, aircraft manufacturers employ the mass properties management (MPM) process during their development processes. However, even when this approach is adopted, the continuous cycle of increasing weight and redesigning aircraft structures represents a significant challenge, which may lead to the cancellation of programs. One of the causes of this problem is inefficient integration between MPM and the aircraft development process. We propose a concept to significantly enhance the integration between MPM and aircraft development processes, suggesting feasible practices to support its implementation. The research methodology combines a review of the literature, an exploratory case study, a three-year longitudinal case study, and verification by experts. The results describe a concept for integrating aircraft MPM and development, supported by 16 practices. They also include a characterization of the MPM process based on literature and practices, which lists 17 characteristics divided into four categories: Goals/Strategy, Activities/Information, Resources/Tools, and Organization/Roles and Responsibilities.

Genetic algorithm artificial neural network in near infrared spectroscopic quantification

The implantation of a genetic algorithm (GA) in quantitating components of interest in near infrared spectroscopic analysis could improve the predictive ability of a regression model. Thus, this study investigates the feasibility of a single layer Artificial Neuron Network (ANN) that trained with Levenberg-Marquardt (SLM) coupled with GA in predicting the boiling point of diesel fuel and the blood hemoglobin using near infrared spectral data. The proposed model was compared with a well-known model of Partial Least Squares (PLS) with and without Genetic Algorithm. Results show that the proposed model achieved the best results with root mean square error of prediction (RMSEP) of 3.6734 and correlation coefficient of 0.9903 for the boiling point, and RMSEP of 0.2349 and correlation coefficient of 0.9874 among PLS with and without GA, and SLM without GA. Findings suggest that the proposed SLMGA is insusceptible to the number of iterations when the SLM was trained with excessive iteration after the optimal iteration number. This indicates that the proposed model is capable of avoiding overfitting issue that due to excessive training iteration.