An Optimization Approach for Convolutional Neural Network Using Non-Dominated Sorted Genetic Algorithm-Ⅱ

In computer vision,convolutional neural networks have a wide range of uses.Images representmost of today’s data,so it’s important to know how to handle these large amounts of data efficiently.Convolutional neural networks have been shown to solve image processing problems effectively.However,when designing the network structure for a particular problem,you need to adjust the hyperparameters for higher accuracy.This technique is time consuming and requires a lot of work and domain knowledge.Designing a convolutional neural network architecture is a classic NP-hard optimization challenge.On the other hand,different datasets require different combinations of models or hyperparameters,which can be time consuming and inconvenient.Various approaches have been proposed to overcome this problem,such as grid search limited to low-dimensional space and queuing by random selection.To address this issue,we propose an evolutionary algorithm-based approach that dynamically enhances the structure of Convolution Neural Networks(CNNs)using optimized hyperparameters.This study proposes a method using Non-dominated sorted genetic algorithms(NSGA)to improve the hyperparameters of the CNN model.In addition,different types and parameter ranges of existing genetic algorithms are used.Acomparative study was conducted with various state-of-the-art methodologies and algorithms.Experiments have shown that our proposed approach is superior to previous methods in terms of classification accuracy,and the results are published in modern computing literature.

Suspended sediment load prediction using non-dominated sorting genetic algorithm Ⅱ

Awareness of suspended sediment load (SSL) and its continuous monitoring plays an important role in soil erosion studies and watershed management.Despite the common use of the conventional model of the sediment rating curve (SRC) and the methods proposed to correct it,the results of this model are still not sufficiently accurate.In this study,in order to increase the efficiency of SRC model,a multi-objective optimization approach is proposed using the Non-dominated Sorting Genetic Algorithm Ⅱ (NSGA-Ⅱ) algorithm.The instantaneous flow discharge and SSL data from the Ramian hydrometric station on the Ghorichay River,Iran are used as a case study.In the first part of the study,using self-organizing map (SOM),an unsupervised artificial neural network,the data were clustered and classified as two homogeneous groups as 70% and 30% for use in calibration and evaluation of SRC models,respectively.In the second part of the study,two different groups of SRC model comprised of conventional SRC models and optimized models (single and multi-objective optimization algorithms) were extracted from calibration data set and their performance was evaluated.The comparative analysis of the results revealed that the optimal SRC model achieved through NSGA-Ⅱ algorithm was superior to the SRC models in the daily SSL estimation for the data used in this study.Given that the use of the SRC model is common,the proposed model in this study can increase the efficiency of this regression model.

GREEDY NON-DOMINATED SORTING IN GENETIC ALGORITHM-ⅡFOR VEHICLE ROUTING PROBLEM IN DISTRIBUTION

Vehicle routing problem in distribution （VRPD） is a widely used type of vehicle routing problem （VRP）, which has been proved as NP-Hard, and it is usually modeled as single objective optimization problem when modeling. For multi-objective optimization model, most researches consider two objectives. A multi-objective mathematical model for VRP is proposed, which considers the number of vehicles used, the length of route and the time arrived at each client. Genetic algorithm is one of the most widely used algorithms to solve VRP. As a type of genetic algorithm （GA）, non-dominated sorting in genetic algorithm-Ⅱ （NSGA-Ⅱ） also suffers from premature convergence and enclosure competition. In order to avoid these kinds of shortage, a greedy NSGA-Ⅱ （GNSGA-Ⅱ） is proposed for VRP problem. Greedy algorithm is implemented in generating the initial population, cross-over and mutation. All these procedures ensure that NSGA-Ⅱ is prevented from premature convergence and refine the performance of NSGA-Ⅱ at each step. In the distribution problem of a distribution center in Michigan, US, the GNSGA-Ⅱ is compared with NSGA-Ⅱ. As a result, the GNSGA-Ⅱ is the most efficient one and can get the most optimized solution to VRP problem. Also, in GNSGA-Ⅱ, premature convergence is better avoided and search efficiency has been improved sharply.

Improved non-dominated sorting genetic algorithm (NSGA)-II in multi-objective optimization studies of wind turbine blades

The non-dominated sorting genetic algorithm （NSGA） is improved with the controlled elitism and dynamic crowding distance. A novel multi-objective optimization algorithm is obtained for wind turbine blades. As an example, a 5 MW wind turbine blade design is presented by taking the maximum power coefficient and the minimum blade mass as the optimization objectives. The optimal results show that this algorithm has good performance in handling the multi-objective optimization of wind turbines, and it gives a Pareto-optimal solution set rather than the optimum solutions to the conventional multi objective optimization problems. The wind turbine blade optimization method presented in this paper provides a new and general algorithm for the multi-objective optimization of wind turbines.

Planning of DC Electric Spring with Particle Swarm Optimization and Elitist Non-dominated Sorting Genetic Algorithm

This paper addresses the planning problem of parallel DC electric springs (DCESs). DCES, a demand-side management method, realizes automatic matching of power consumption and power generation by adjusting non-critical load (NCL) and internal storage. It can offer higher power quality to critical load (CL), reduce power imbalance and relieve pressure on energy storage systems (RESs). In this paper, a planning method for parallel DCESs is proposed to maximize stability gain, economic benefits, and penetration of RESs. The planning model is a master optimization with sub-optimization to highlight the priority of objectives. Master optimization is used to improve stability of the network, and sub-optimization aims to improve economic benefit and allowable penetration of RESs. This issue is a multivariable nonlinear mixed integer problem, requiring huge calculations by using common solvers. Therefore, particle Swarm optimization (PSO) and Elitist non-dominated sorting genetic algorithm (NSGA-II) were used to solve this model. Considering uncertainty of RESs, this paper verifies effectiveness of the proposed planning method on IEEE 33-bus system based on deterministic scenarios obtained by scenario analysis.

Optimization of dynamic aperture by using non-dominated sorting genetic algorithm-Ⅱ in a diffraction-limited storage ring with solenoids for generating round beam

Purpose Round beam,i.e.,with equal horizontal and vertical emittance,is preferable than a horizontally flat one for some beamline applications in Diffraction-limited storage rings(DLSRs),for the purposes of reducing the number of photons getting discarded and better phase space match between photon and electron beam.Conventional methods of obtaining round beam inescapably results in a reduction of dynamic aperture(DA).In order to recover the DA as much as possible for improving the injection efficiency,the DA optimization by using Non-dominated sorting genetic algorithm-Ⅱ(NSGA-Ⅱ)to generate round beam,particularly to one of the designed lattice of the High Energy Photon Source(HEPS)storage ring,are presented.Method According to the general unconstrained model of NSGA-Ⅱ,we modified the standard model by using parallel computing to optimize round beam lattices with errors,especially for a strong coupling,such as solenoid scheme.Results and conclusion The results of numerical tracking verify the correction of the theory framework of solenoids with fringe fields and demonstrates the feasibility on the HEPS storage ring with errors to operate in round beam mode after optimizing DA.

基于NSGA-Ⅱ的滑油泵叶轮结构优化设计

滑油泵常需要在高空、低压工况下稳定运转,常会出现供油不足、效率降低等问题。为了得到满足设计要求且具有最佳性能的滑油泵,以某直升机用滑油泵叶轮为研究对象,对其结构进行优化设计。选择高空两个典型工况的效率与扬程作为优化目标,利用NSGA-Ⅱ算法对滑油泵叶轮几何参数进行寻优,对优化前后的滑油泵效率、扬程进行对比分析。采用CFD流体仿真及实验方法对优化结果进行对比验证。结果表明:所选优化参数对滑油泵性能有较大影响,优化后的滑油泵叶片位置附近流动更加平稳,高低压区域过渡平缓,能量损失更小,且降低了汽蚀发生的可能性;优化后的滑油泵设计点扬程提高2.6 m,效率提高2.86%。

基于代理模型和NSGA-Ⅱ的超高强钢电阻点焊工艺参数多目标优化

为寻求超高强钢电阻点焊时最佳的焊接工艺参数,开展正交试验法设计三因素五水平的平板搭接点焊试验,以焊接时间、焊接电流和电极压力为可调的工艺参数,将熔核直径、压痕深度、拉剪强度及飞溅情况作为焊接接头质量评价指标.基于高斯过程回归和BP神经网络建立起焊接工艺参数与焊接接头质量评价指标之间关系的代理模型,训练的结果显示模型精度很高.最后利用带精英策略的非支配排序的遗传算法NSGA-Ⅱ实现多目标优化,得到各评价指标之间的最优pareto解集.经验证,各评价模型的相对误差值都很小.结果表明,该优化方法有较好的预测效果和稳定性.通过使用较少的试验数据,建立优化模型的方法对电阻点焊及其它焊接领域最佳焊接工艺参数的选取具有重要的指导价值.

Strengthened Dominance Relation NSGA-Ⅲ Algorithm Based on Differential Evolution to Solve Job Shop Scheduling Problem

The job shop scheduling problem is a classical combinatorial optimization challenge frequently encountered in manufacturing systems.It involves determining the optimal execution sequences for a set of jobs on various machines to maximize production efficiency and meet multiple objectives.The Non-dominated Sorting Genetic Algorithm Ⅲ(NSGA-Ⅲ)is an effective approach for solving the multi-objective job shop scheduling problem.Nevertheless,it has some limitations in solving scheduling problems,including inadequate global search capability,susceptibility to premature convergence,and challenges in balancing convergence and diversity.To enhance its performance,this paper introduces a strengthened dominance relation NSGA-Ⅲ algorithm based on differential evolution(NSGA-Ⅲ-SD).By incorporating constrained differential evolution and simulated binary crossover genetic operators,this algorithm effectively improves NSGA-Ⅲ’s global search capability while mitigating pre-mature convergence issues.Furthermore,it introduces a reinforced dominance relation to address the trade-off between convergence and diversity in NSGA-Ⅲ.Additionally,effective encoding and decoding methods for discrete job shop scheduling are proposed,which can improve the overall performance of the algorithm without complex computation.To validate the algorithm’s effectiveness,NSGA-Ⅲ-SD is extensively compared with other advanced multi-objective optimization algorithms using 20 job shop scheduling test instances.The experimental results demonstrate that NSGA-Ⅲ-SD achieves better solution quality and diversity,proving its effectiveness in solving the multi-objective job shop scheduling problem.

基于NSGA-Ⅱ的智能化电铲多目标最优挖掘轨迹规划

为实现智能化电铲实时节能的挖掘,提出了一种基于非支配排序遗传算法(Non-dominated Sorting Genetic Algorithm-II,NSGA-Ⅱ)的智能化电铲多目标最优挖掘轨迹规划方法。首先,通过拉格朗日方程建立智能化电铲工作装置动力学模型;然后,使用高次多项式对挖掘轨迹进行插值,将挖掘轨迹寻优问题转化为多项式系数寻优问题,最后,以挖掘时间最短及单位体积物料的挖掘能耗最小作为优化目标,以电机性能与挖掘过程中几何条件等作为约束,利用多目标优化平台PlatEMO,将NSGA-Ⅱ作为多目标优化算法,指定待优化问题的目标函数及约束函数,获取到多目标优化Pareto最优解集,基于决策偏好设置权重并根据TOPSIS法获取最优解,得到多目标最优挖掘轨迹规划结果。结果表明,优化后挖掘轨迹满足实时节能的挖掘要求。

基于改进NSGA-Ⅱ算法的RV减速器参数多目标优化研究

旋转矢量(RV)减速器是工业机器人核心部件,对于机器人的性能起到关键作用。针对提升RV减速器综合性能的问题,从优化传动压力角的相关参数出发,对其结构参数(摆线轮齿数、短幅系数、针径系数、摆线轮宽度等)的多目标优化设计进行了研究。首先,研究了摆线轮平均压力角、传动效率和传动机构体积三者的相关参数之间的关系;然后,以此为优化目标,在摆线轮标准齿廓方程的基础上建立了多目标优化数学模型(该模型采用了基于非支配占优排序遗传学算法(NSGA-Ⅱ)改进了交叉算子系数生成的改进NSGA-Ⅱ算法);通过模型求解得到了帕累托最优解集,根据模糊集合理论的相关方法选取了最优解;最后,以某公司220-BX型RV减速器为例,进行了优化设计,建立了3D模型后进行了有限元分析,并加工出实验样机,进行了传动效率对比实验。实验结果表明:摆线轮平均压力角减小了7.19%,体积减小了11.1%,传动效率提高了4.9%。研究结果表明:该模型交互性强,能提高设计效率并节省设计开销,可为实际RV减速器工程优化设计提供参考。

Satellite constellation design with genetic algorithms based on system performance

Satellite constellation design for space optical systems is essentially a multiple-objective optimization problem. In this work, to tackle this challenge, we first categorize the performance metrics of the space optical system by taking into account the system tasks（i.e., target detection and tracking）. We then propose a new non-dominated sorting genetic algorithm（NSGA） to maximize the system surveillance performance. Pareto optimal sets are employed to deal with the conflicts due to the presence of multiple cost functions. Simulation results verify the validity and the improved performance of the proposed technique over benchmark methods.

基于NSGA-Ⅱ的自适应多尺度特征通道分组优化算法

针对轻量型卷积神经网络(LCNN)的精确度和复杂度均衡优化问题,提出基于快速非支配排序遗传算法(NSGA-Ⅱ)的自适应多尺度特征通道分组优化算法对LCNN特征通道分组结构进行优化。首先,将LCNN中的特征融合层结构的复杂度最小化和精确度最大化作为两个优化目标,进行双目标函数建模及理论分析;然后,设计基于NSGA-Ⅱ的LCNN结构优化框架,并在原始LCNN结构的深度卷积层之上增加基于NSGA-Ⅱ的自适应分组层,构建基于NSGA-Ⅱ的自适应多尺度的特征融合网络NSGA2-AMFFNetwork。在图像分类数据集上的实验结果显示,与手工设计的网络结构M_blockNet_v1相比,NSGA2-AMFFNetwork的平均精确度提升了1.2202个百分点,运行时间降低了41.07%。这表明所提优化算法能较好平衡LCNN的复杂度和精确度,同时还可为领域知识不足的普通用户提供更多性能表现均衡的网络结构选择方案。

基于强化学习的改进NSGA-Ⅱ算法的城市快速路入口匝道控制

为了缓解城市快速路拥堵和尾气排放问题,提出了基于竞争结构和深度循环Q网络的改进非支配排序遗传算法(non-dominated sorting genetic algorithm Ⅱ based on dueling deep recurrent Q network, DRQN-NSGA-Ⅱ).该算法结合了基于竞争结构的深度Q网络(dueling deep Q network, Dueling DQN)、深度循环Q网络(deep recurrent Q network, DRQN)和NSGA-Ⅱ算法,将Dueling DRQN-NSGA-Ⅱ算法用于匝道控制问题.除了考虑匝道车辆汇入以提高快速路通行效率外,还考虑了环境和能源指标,将尾气排放和燃油消耗作为评价指标.除了与无控制情况及其他算法进行比较之外, Dueling DRQN-NSGA-Ⅱ还与NSGA-Ⅱ算法进行了比较.实验结果表明:与无控制情况相比,本算法能有效改善路网通行效率、缓解环境污染、减少能源损耗;相对于无控制情况,总花费时间(total time spent, TTS)减少了16.14%,总尾气排放(total emissions, TE)减少了9.56%,总燃油消耗(total fuel consumption, TF)得到了43.49%的改善.

基于改进NSGA-Ⅱ的铁路项目进度计划多目标优化

以总工期最短和总费用最低为目标,针对包含线性活动、条状活动、块状活动等多种施工场景的铁路工程项目,基于RSM方法构建铁路项目多目标优化模型,并提出一种改进的NSGA-Ⅱ算法对模型进行求解.算法设计一种分层次选取种群个体的均匀进化精英选择策略,以提高种群多样性和收敛性;同时引入差分进化算法的变异、交叉算子,构造分层多策略自适应变异、交叉算子,以平衡整个种群的局部搜索能力和全局搜索能力.结果表明,增加对特殊活动和施工方向的考虑,可增强模型对铁路项目的适用性;改进后的算法收敛速度快,运行稳定,得到的结果更优,能够满足较大规模铁路项目进度计划优化.

Models for Location Inventory Routing Problem of Cold Chain Logistics with NSGA-Ⅱ Algorithm

In this paper,a novel location inventory routing(LIR)model is proposed to solve cold chain logistics network problem under uncertain demand environment. The goal of the developed model is to optimize costs of location,inventory and transportation.Due to the complex of LIR problem( LIRP), a multi-objective genetic algorithm(GA), non-dominated sorting in genetic algorithm Ⅱ( NSGA-Ⅱ) has been introduced. Its performance is tested over a real case for the proposed problems. Results indicate that NSGA-Ⅱ provides a competitive performance than GA,which demonstrates that the proposed model and multi-objective GA are considerably efficient to solve the problem.

基于NSGA-Ⅱ与方案优选的机场航站楼大跨度钢结构多目标优化研究

为优选出机场航站楼大跨度钢结构最佳施工方案,实现施工工期短、成本低和质量高的综合优化目标,文章以非支配排序遗传算法(non-dominated sorting genetic algorithm Ⅱ,NSGA-Ⅱ)与方案优选为基础,结合建筑信息模型(building information modeling,BIM)技术和工作分解结构(work breakdown structure,WBS)技术,构建兼具优化与施工仿真模拟的大跨度钢结构多目标优化体系;以某军民合用机场为例,应用该体系确定该工程航站楼大跨度网架结构安装采用分块安装法,优选出的施工方案较优化前不仅质量水平保持在较高的0.95,而且工期缩短22 d、成本减少57625元,进一步验证了该体系具有很好的可行性与有效性。研究结果可为科学合理地确定大跨度钢结构施工方案提供依据,并有助于提高机场航站楼施工管理水平,为类似工程提供参考。

基于超平面NSGA-Ⅱ的双输入双降压逆变器系统参数优化设计

针对第二代非支配排序遗传算法(NSGA-Ⅱ)计算过程中存在种群分布不均匀、收敛性速度较慢的问题,提出超平面NSGA-Ⅱ(HP-NSGA-Ⅱ).该算法通过连接反映种群边缘分布的极值点构造超平面,以其法向量为进化趋势,对临界层个体在超平面进行投影,促使种群朝着分布均匀且收敛良好的最优解进化.以双输入双降压型逆变器(DIDBI)为多目标优化对象,开关损耗、输出电压总谐波失真和滤波元件体积为优化目标,依据谐振频率、电感电流纹波和功率因数的要求,推导出滤波电容、滤波电感和开关频率的约束条件,比较分析HP-NSGA-Ⅱ与NSGA-Ⅱ、考虑各目标重要度的γ-NSGA-Ⅱ的应用场合和价值.以某型逆变器样机为例,开展参数优化设计实验研究,结果表明了设计的有效性与正确性.

基于NSGA-Ⅱ的电力信息物理系统骨干网络辨识

针对实际存在一一对应的电力信息物理系统,辨识其抗灾型骨干网架并进行加固,可提高电力信息物理系统在面对自然灾害或网络攻击下的可靠性。文中提出骨干网络辨识多目标优化模型,所提模型综合考虑整个系统的经济性、抗毁性和恢复性,并满足重要负荷约束、网络连通性约束、网络规模约束和潮流约束。首先,使用改进的非支配排序遗传算法(NSGA-Ⅱ)求解多目标优化模型,并利用多目标决策中的熵权法给各个目标函数赋权重。然后,使用逼近理想解排序法筛选出帕累托解集中的最优解。最后,以IEEE 39节点系统和中国某地区500 kV实际电网为例,验证了所提的电力信息物理系统骨干网络辨识算法的有效性。

基于改进NSGA-Ⅱ的考虑自动引导车充电策略的集成调度

针对自动引导车(AGV)在自动化集装箱码头(ACT)执行任务过程中的电量问题,提出基于改进的非支配排序遗传算法-Ⅱ(NSGA-Ⅱ)的考虑AGV充电策略的集成调度。首先,在岸桥、场桥和AGV集成调度模式下,考虑AGV在不同作业状态下的耗电量,并建立以最小化作业完工时间和总耗电量为目标的多目标混合规划模型;其次,为提高传统NSGA-Ⅱ的性能,设计自适应NSGA-Ⅱ,并将所提算法与CPLEX求解器、NSGA-Ⅱ和多目标粒子群优化(MOPSO)算法进行性能对比;最后,设计AGV不同充电策略并对设备数量配比进行实验研究。算法对比实验结果表明:相较于传统NSGA-Ⅱ算法,自适应NSGA-Ⅱ对双目标的优化分别提升了2.8%和2.63%。利用自适应NSGA-Ⅱ进行的充电策略和设备数量配比实验的结果表明:增加AGV充电次数能够减少AGV的充电时间,且调整设备数量配比至3∶3∶9和3∶7∶3时,场桥和AGV的时间利用率分别达到最高。可见,AGV充电策略及设备数量配比对码头多设备集成调度有一定影响。