摘要：
Multi-target regression has always been a challenging task in engineering applications. Nevertheless, it is easy to encounter problems such as low accuracy and inadequate robustness in some scenarios. To address these issues, an ensemble strategy considering correlations is proposed, named Ensemble-Adaptive Tree-based Correlation Chains. Specifically, a Follow-up Correlation Chaining strategy that quantifies the relationships among targets by arranging the L1 norms of correlations is suggested. Compared with other related strategies, it allows for the representation of these relationships through a single regressor chain. Under the proposed framework, the ensemble strategy integrates ten chains, wherein each chain adaptively updates the sample weights during training. This process involves employing the out-of-sample observations with new convergence criteria. Furthermore, the eXtreme Gradient Boosting is introduced as the base regressor to enhance the overall accuracy of the entire method. Finally, the proposed method is validated based on 25 multi-target datasets and a lightweight design of a high-speed rail bogie. The results demonstrate the superior accuracy and robustness compared to other state-of-the-art methods. In general, this study provides reliable predictions for specific scenarios and delivers practical significance in addressing relevant problems.
作者机构：
[Wang, Hu; Deng, Xinjian] Hunan Univ, State Key Lab Adv Design & Mfg Vehicle Body, Changsha 410082, Peoples R China.;[Li, Enying] Cent South Univ Forestry & Technol, Coll Mech & Elect Engn, Changsha 410004, Peoples R China.
通讯机构：
[Enying Li] C;College of Mechanical & Electrical Engineering, Central South University of Forestry and Technology, Changsha 410004, China<&wdkj&>Author to whom correspondence should be addressed.
摘要：
For the multi-objective design of heat sinks, several evolutionary algorithms usually require many iterations to converge, which is computationally expensive. Variable-fidelity multi-objective (VFO) methods were suggested to improve the efficiency of evolutionary algorithms. However, multi-objective problems are seldom optimized using VFO. Therefore, a variable-fidelity evolutionary method (VFMEM) was suggested. Similar to other variable-fidelity algorithms, VFMEM solves a high-fidelity model using a low-fidelity model. Compared with other algorithms, the distinctive characteristic of VFMEM is its application in multi-objective optimization. First, the suggested method uses a low-fidelity model to locate the region where the global optimal solution might be found. Sequentially, both high- and low-fidelity models can be integrated to find the real global optimal solution. Circulation distance elimination (CDE) was suggested to uniformly obtain the PF. To evaluate the feasibility of VFMEM, two classical benchmark functions were tested. Compared with the widely used multi-objective particle swarm optimization (MOPSO), the efficiency of VFMEM was significantly improved and the Pareto frontier (PFs) could also be obtained. To evaluate the algorithm's feasibility, a polygonal pin fin heat sink (PFHS) design was carried out using VFMEM. Compared with the initial design, the results showed that the mass, base temperature, and temperature difference of the designed optimum heat sink were decreased 5.5%, 18.5%, and 62.0%, respectively. More importantly, if the design was completed directly by MOPSO, the computational cost of the entire optimization procedure would be significantly increased.
摘要：
Inverse heat conduction problems (IHCPs) are problems of estimating unknown quantities of interest (QoIs) of the heat conduction with given temperature observations. The challenge of IHCPs is that it is usually ill-posed since the observations are noisy, and the estimations of QoIs are generally not unique or unstable, especially when there are unknown spatially varying QoIs. In this study, an ensemble physics-informed neural network (E-PINN) is proposed to handle function estimation and uncertainty quantification of space-dependent IHCPs. The distinctive characteristics of E-PINN are ensemble learning and adversarial training (AT). Compared with other data-driven UQ approaches, the suggested method is more than straightforward to implement and also achieves high-quality uncertainty estimates of the QoI. Furthermore, an adaptive active sampling (AS) strategy based on the uncertainty estimates from E-PINNs is also proposed to improve the accuracy of material field inversion problems. Finally, the proposed method is validated through several numerical experiments of IHCPs.
作者机构：
[Jiang, Xinchao; Li, Jinyang; Wang, Hu] Hunan Univ, State Key Lab Adv Design & Mfg Vehicle Body, Changsha 410082, Peoples R China.;[Wang, Hu] Beijing Inst Technol Shenzhen Automot Res Inst, Shenzhen 518000, Peoples R China.;[Luo, Hong; Li, Enying] Cent South Univ Forestry & Technol, Coll Mech & Elect Engn, Changsha, Peoples R China.
通讯机构：
[Wang, H ] H;Hunan Univ, State Key Lab Adv Design & Mfg Vehicle Body, Changsha 410082, Peoples R China.
关键词：
Meta-learning;Surrogate model selection;Automated model selection;Auto machine learning
摘要：
With the development of booming AutoML systems, modeling processes have become more automatic for researchers. However, AutoML systems may struggle to identify the optimal surrogate type, find the best combination of the hyper-parameters or establish a high-fidelity ensembled surrogate model for certain datasets. To address these issues and further improve the warm-start procedure of AutoML, a Ranking Prediction Strategy assisted Automatic Model Selection (RPS-AMS) method is proposed. In the suggested method, an integration of evolutionary algorithms (EA-based) and feature-based driven model selection strategy selects the best or the best combination models for prediction. Based on the proposed criteria, an XGBoost regression model is trained to determine the rankings from the candidate surrogate models and then build an ensembled surrogate model to further enhance accuracy. We evaluate RPS-AMS using 13 mathematical functions, 14 public datasets, and a real engineering problem. Compared with the popular modeling tools, such as Auto-Sklearn and EvalML, the RPS-AMS outperforms in term of accuracy while maintaining the performances of ergodic methods of all surrogate models. The accuracies of RPS-AMS rival EvalML in most tested datasets, although RPS-AMS may be slightly less efficient. Given that EvalML is a masterpiece of the AutoML systems, the performances of RPS-AMS are promising. This code is available at: https://github.com/HnuAiSimOpt/RPS-AMS.
期刊：
International Journal of Computational Methods,2021年18(02):2050033 ISSN：0219-8762
通讯作者：
Wang, Hu
作者机构：
[Wang, Hu; Hu, Wei] Hunan Univ, State Key Lab Adv Design & Mfg Vehicle Body, Changsha 410082, Peoples R China.;[Wang, Hu; Hu, Wei] Joint Ctr Intelligent New Energy Vehicle, Shanghai 201804, Peoples R China.;[Li, Enying] Cent South Univ Forestry & Technol, Coll Mech & Elect Engn, Changsha 41004, Peoples R China.
通讯机构：
[Wang, Hu] H;[Wang, Hu] J;Hunan Univ, State Key Lab Adv Design & Mfg Vehicle Body, Changsha 410082, Peoples R China.;Joint Ctr Intelligent New Energy Vehicle, Shanghai 201804, Peoples R China.
关键词：
Least squares support vector regression (LSSVR);efficient global optimization (EGO);surrogate-assisted optimization (SAO);constrained optimization;infeasible sample points
摘要：
Although the Efficient Global Optimization (EGO) algorithm has been widely used in multi-disciplinary optimization, it is still difficult to handle multiple constraint problems. In this study, to increase the accuracy of approximation, the Least Squares Support Vector Regression (LSSVR) is suggested to replace the kriging model for approximating both objective and constrained functions while the variances of these surrogate models are still obtained by kriging. To enhance the ability to search the feasible region, two criteria are suggested. First, a Maximize Probability of Feasibility (MPF) strategy to handle the infeasible initial sample points is suggested to generate feasible points. Second, a Multi-Constraint Parallel (MCP) criterion is suggested for multiple constraints handling, parallel computation and validation, respectively. To illustrate the efficiency of the suggested EGO-based method, several deterministic benchmarks are tested and the suggested methods demonstrate a superior performance compared with two other constrained algorithms. Finally, the suggested algorithm is successfully utilized to optimize the fiber path of variable-stiffness beam and lightweight B-pillar to demonstrate the performance for engineering applications.
作者机构：
[Wang, Hu; Chen, Boxuan; Zeng, Yang] Hunan Univ, State Key Lab Adv Design & Mfg Vehicle Body, Changsha 410082, Peoples R China.;[Wang, Hu; Chen, Boxuan; Zeng, Yang] Joint Ctr Intelligent New Energy Vehicle, Shanghai 201804, Peoples R China.;[Li, Enying] Cent South Univ Forestry & Teleol, Coll Mech & Elect Engn, Changsha 41004, Peoples R China.
通讯机构：
[Hu Wang] S;State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body, Hunan University, Changsha 410082, China<&wdkj&>Joint Center for Intelligent New Energy Vehicle, Shanghai 201804, China
摘要：
A new equivalent circuit model (ECM) of a Li-ion battery is developed in this study. The developed model is utilized to obtain the dynamic electrical response of the battery when it is deformed under external force. Compared with other models, this model is developed based on a modified Thevenin model, and the parameters of the developed model are relevant to state of charge, the battery surface temperature, and the deformation. In this study, to obtain the real electrical response of the battery when it deformed under external force, batteries that are compressed by different deformations from 0 to 5 mm are studied with pulse discharging tests. Then, the parameters of the circuit elements are identified by a differential evolution algorithm based on the data obtained from these tests. Moreover, the data from the pulse discharging tests of batteries compressed by 3.5, 4.25, and 4.5 mm and the data from the pulse charging tests of batteries compressed by 0 and 1 mm are used to verify the parameters. The results illustrate that the battery capacity should drop significantly when the battery is severely deformed, but the battery still can be charged and discharged. Most importantly, the discharging curves of these tested deformed batteries are similar to those of undeformed ones. Moreover, the developed new ECM can predict the dynamic electrical response of a deformed battery accurately.
期刊：
Structural and Multidisciplinary Optimization,2020年62(2):937-955 ISSN：1615-147X
通讯作者：
Li, Enying
作者机构：
[Shuai, Wenquan; Li, Enying] Cent South Univ Forestry & Technol, Coll Mech & Elect Engn, Changsha 41004, Peoples R China.;[Li, Yu; Wang, Hu; Shuai, Wenquan] Hunan Univ, State Key Lab Adv Design & Mfg Vehicle Body, Changsha 410082, Hunan, Peoples R China.;[Wang, Hu] Joint Ctr Intelligent New Energy Vehicle, Shanghai, Peoples R China.
通讯机构：
[Li, Enying] C;Cent South Univ Forestry & Technol, Coll Mech & Elect Engn, Changsha 41004, Peoples R China.
关键词：
Space mapping (SM);Thin-walled honeycomb structure;Li-ion battery package;Pseudo-plane-strain model
摘要：
A new thin-walled honeycomb structure for Li-ion battery packaging is designed and optimized in this study. Compared with other battery packaging structures, the designed honeycomb structure described here uses a grid to reinforce its strength. At the same time, the weight is reduced to improve the energy density of the entire package. Moreover, the new thin-walled structure can better protect the internal battery and improve the safety of an electric vehicle (EV). A space mapping (SM) algorithm is used to efficiently optimize the thin-walled honeycomb structure due to the expensive computational cost of each evaluation of a fine FE model. Compared with other SM algorithms, the coarse model of SM is based on a pseudo-plane-strain model. The result shows that the magnitude of stress and the distribution of stress are significantly improved compared with the initial structure. Moreover, the computational cost of optimization for the problem is also decreased significantly due to importing the coarse model.
关键词：
Inverse heat conduction problem (IHCP);Approximate Bayesian computation (ABC);Reanalysis;Non-parametric population Monte Carlo (NPMC)
摘要：
Bayesian approach has been widely used in inverse heat conduction problem (IHCP). However, due to either computationally prohibitive or analytically unavailable, its likelihood function is always intractable. In this study, to circumvent the intractable likelihood function, an approximate Bayesian computation (ABC) is extended to IHCP. However, massive expensive forward simulations are needed. It might lead to prohibited computational cost. In order to improve the efficiency of the ABC-IHCP, two strategies are proposed in this study. At first, in order to improve the convergence rate of ABC and reduce the number of samples, a none-parametric population Monte Carlo (NPMC) is proposed to determine the decreasing tolerance value adaptively. Secondly, in order to save the expensive computational cost of heat conduction simulation, the fast computational techniques are utilized. Based on the characteristics of the linear and nonlinear heat transfer problems, two heat conduction solvers are developed, respectively. The linear solver is based on superposition principle. As for the nonlinear problem, the fast and accurate reanalysis solver is suggested. Finally, the accuracy and efficiency of the suggested methods are verified with two numerical examples. (C) 2019 Elsevier Ltd. All rights reserved.
期刊：
International Journal of Material Forming,2018年11(2):279-295 ISSN：1960-6206
通讯作者：
Wang, Hu;Li, Enying
作者机构：
[Wang, Hu; Chen, Lei] Hunan Univ, Coll Mech & Automot Engn, State Key Lab Adv Technol Vehicle Design & Manufa, Changsha 410082, Hunan, Peoples R China.;[Wang, Hu; Chen, Lei] Joint Ctr Intelligent New Energy Vehicle, Shanghai 201804, Peoples R China.;[Li, Enying] Cent South Univ Forestry & Teleol, Sch Mech & Elect Engn, Changsha 41004, Hunan, Peoples R China.
通讯机构：
[Wang, Hu] H;[Wang, Hu] J;[Li, Enying] C;Hunan Univ, Coll Mech & Automot Engn, State Key Lab Adv Technol Vehicle Design & Manufa, Changsha 410082, Hunan, Peoples R China.;Joint Ctr Intelligent New Energy Vehicle, Shanghai 201804, Peoples R China.
关键词：
Surrogate;Time dependent;Sheet metal forming;Firefly algorithm (FA);Expected improvement (EI)
摘要：
For a sheet metal forming optimization problem, time related design variables are seldom considered in practice. The purpose of this work is to handle time dependent sheet metal forming problems. Because it is difficult to investigate all time points during the entire forming procedure, some key time points should be extracted. Therefore, the number of design variables should be significantly increased due to introduce auxiliary time design variables. However, curse of dimensionality is a formidable difficult issue to be solved. To solve such medium-scale problems, Gaussian Process Assisted Firefly Algorithm (GPFA) is suggested. The main idea of the suggested method is to construct a surrogate model-aware search mechanism with Firefly Algorithm (FA) for simulation-based optimization efficiently. Compared with other FAs, the distinctive characteristic of GPFA is to generate new sample points adaptively based on maximum Expected Improvement (EI) criterion, so that the local and global search can be well balanced, and a small promising area can be quickly focused on. Numerical studies on benchmark problems with 20 variables and a real-world application of time dependent sheet metal forming optimization reveal that the GPFA is capable to solve such similar problems.
