Advances in Manufacturing

Improving RSW nugget diameter prediction method: unleashing the power of multi-fidelity neural networks and transfer learning

Zhong-Jie Yue, Qiu-Ren Chen, Zu-Guo Bao, Li Huang, Guo-Bi Tan, Ze-Hong Hou, Mu-Shi Li, Shi-Yao Huang, Hai-Long Zhao, Jing-Yu Kong, Jia Wang, Qing Liu

2024, 12(3): 409-427. doi:10.1007/s40436-024-00503-2

摘要 ( 92 )

PDF (191KB) ( 36 )

参考文献 | 相关文章 | 多维度评价

This research presents an innovative approach to accurately predict the nugget diameter in resistance spot welding (RSW) by leveraging machine learning and transfer learning methods. Initially, low-fidelity (LF) data were obtained through finite element numerical simulation and design of experiments (DOEs) to train the LF machine learning model. Subsequently, high-fidelity (HF) data were collected from RSW process experiments and used to fine-tune the LF model by transfer learning techniques. The accuracy and generalization performance of the models were thoroughly validated. The results demonstrated that combining different fidelity datasets and employing transfer learning could significantly improve the prediction accuracy while minimize the costs associated with experimental trials, and provide an effective and valuable method for predicting critical process parameters in RSW.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00503-2

A multi-objective optimization based on machine learning for dimension precision of wax pattern in turbine blade manufacturing

Jing Dai, Song-Zhe Xu, Chao-Yue Chen, Tao Hu, San-San Shuai, Wei-Dong Xuan, Jiang Wang, Zhong-Ming Ren

2024, 12(3): 428-446. doi:10.1007/s40436-024-00492-2

摘要 ( 110 )

PDF (191KB) ( 41 )

参考文献 | 相关文章 | 多维度评价

Wax pattern fabrication in the investment casting of hollow turbine blades directly determines the dimension accuracy of subsequent casting, and therefore significantly affects the quality of final product. In this work, we develop a machine learning-based multi-objective optimization framework for improving dimension accuracy of wax pattern by optimizing its process parameters. We consider two optimization objectives on the dimension of wax pattern, i.e., the surface warpage and core offset. An active learning of Bayesian optimization is employed in data sampling to determine process parameters, and a validated numerical model of injection molding is used to compute objective results of dimension under varied process parameters. The collected dataset is then leveraged to train different machine learning models, and it turns out that the Gaussian process regression model performs best in prediction accuracy, which is then used as the surrogate model in the optimization framework. A genetic algorithm is employed to produce a non-dominated Pareto front using the surrogate model in searching, followed by an entropy weight method to select the most optimal solution from the Pareto front. The optimized set of process parameters is then compared to empirical parameters obtained from previous trial-and-error experiments, and it turns out that the maximum and average warpage results of the optimized solution decrease 26.0% and 20.2%, and the maximum and average errors of wall thickness compared to standard part decrease from 0.22 mm and 0.051 7 mm using empirical parameters to 0.10 mm and 0.035 6 mm using optimized parameters, respectively. This framework is demonstrated capable of addressing the challenge of dimension control arising in the wax pattern production, and it can be reliably deployed in varied types of turbine blades to significantly reduce the manufacturing cost of turbine blades.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00492-2

Data-driven approach to predict the fatigue properties of ferrous metal materials using the cGAN and machine-learning algorithms

Si-Geng Li, Qiu-Ren Chen, Li Huang, Min Chen, Chen-Di Wei, Zhong-Jie Yue, Ru-Xue Liu, Chao Tong, Qing Liu

2024, 12(3): 447-464. doi:10.1007/s40436-024-00491-3

摘要 ( 90 )

PDF (349KB) ( 57 )

参考文献 | 相关文章 | 多维度评价

The stress-life curve (S-N) and low-cycle strain-life curve (E-N) are the two primary representations used to characterize the fatigue behavior of a material. These material fatigue curves are essential for structural fatigue analysis. However, conducting material fatigue tests is expensive and time-intensive. To address the challenge of data limitations on ferrous metal materials, we propose a novel method that utilizes the Random Forest Algorithm and transfer learning to predict the S-N and E-N curves of ferrous materials. In addition, a data-augmentation framework is introduced using a conditional generative adversarial network (cGAN) to overcome data deficiencies. By incorporating the cGAN-generated data, the accuracy (R²) of the Random Forest Algorithm-trained model is improved by 0.3-0.6. It is proven that the cGAN can significantly enhance the prediction accuracy of the machine-learning model and balance the cost of obtaining fatigue data from the experiment.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00491-3

A machine learning-based calibration method for strength simulation of self-piercing riveted joints

Yu-Xiang Ji, Li Huang, Qiu-Ren Chen, Charles K. S. Moy, Jing-Yi Zhang, Xiao-Ya Hu, Jian Wang, Guo-Bi Tan, Qing Liu

2024, 12(3): 465-483. doi:10.1007/s40436-024-00502-3

摘要 ( 117 )

PDF (245KB) ( 48 )

参考文献 | 相关文章 | 多维度评价

This paper presents a new machine learning-based calibration framework for strength simulation models of self-piercing riveted (SPR) joints. Strength simulations were conducted through the integrated modeling of SPR joints from process to performance, while physical quasi-static tensile tests were performed on combinations of DP600 high-strength steel and 5754 aluminum alloy sheets under lap-shear loading conditions. A sensitivity study of the critical simulation parameters (e.g., friction coefficient and scaling factor) was conducted using the controlled variables method and Sobol sensitivity analysis for feature selection. Subsequently, machine-learning-based surrogate models were used to train and accurately represent the mapping between the detailed joint profile and its load-displacement curve. Calibration of the simulation model is defined as a dual-objective optimization task to minimize errors in key load displacement features between simulations and experiments. A multi-objective genetic algorithm (MOGA) was chosen for optimization. The three combinations of SPR joints illustrated the effectiveness of the proposed framework, and good agreement was achieved between the calibrated models and experiments.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00502-3

Exoskeleton active assistance strategy for human muscle activation reduction during linear and circular walking

Wen-Tao Sheng, Ke-Yao Liang, Hai-Bin Tang

2024, 12(3): 484-496. doi:10.1007/s40436-024-00504-1

摘要 ( 128 )

PDF (225KB) ( 17 )

参考文献 | 相关文章 | 多维度评价

The exoskeleton is employed to assist humans in various domains including military missions, rehabilitation, industrial operation, and activities of daily living (ADLs).Walking is a fundamental ADL, and exoskeletons are capable of reducing the activation and metabolism of lower extremity muscles through active assistance during walking. To improve the performance of active assistance strategy, this article proposes a framework using an active hip exoskeleton. Subsequently, it correlates to an already established Bayesian-based human gait recognition algorithm, with a particular focus on linear and circular walking within industrial and ADL contexts. In theorizing this strategy for exoskeletons, this study further reveals, in part, the activation characteristics of human hip muscles for the instruction and regulation of active assistance duration and onset timing. This proposed active assistance strategy provides new insights for enhancing the performance of assistive robots and facilitating human robot interaction within the context of ADLs.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00504-1

Understanding microstructure-property relationships of HPDC Al-Si alloy based on machine learning and crystal plasticity simulation

Qiang-Qiang Zhai, Zhao Liu, Ping Zhu

2024, 12(3): 497-511. doi:10.1007/s40436-024-00488-y

摘要 ( 88 )

PDF (337KB) ( 15 )

参考文献 | 相关文章 | 多维度评价

Al-Si alloys manufactured via high-pressure die casting (HPDC) are suitable for a wide range of applications. However, the heterogeneous microstructure and unpredictable pore distribution of Al-Si high-pressure die castings result in significant variations in the mechanical properties, thus leading to a complicated microstructure-property relationship that is difficult to capture. Hence, a computational framework incorporating machine learning and crystal plasticity method is proposed. This framework aims to provide a systematic and comprehensive understanding of this relationship and enable the rapid prediction of macroscopic mechanical properties based on the microstructure. Firstly, we select eight variables that can effectively characterize the microstructural features and then obtain their statistical information. Subsequently, based on 160 samples obtained via the Latin hypercube sampling method, representative volume elements are constructed, and the crystal plasticity fast Fourier transformation method is executed to obtain the macroscopic mechanical properties. Next, the yield strength, elastic modulus, strength coefficient, and strain-hardening exponent are used to characterize the stress-strain curve, and Gaussian process regression models and microstructural variables are developed. Finally, sensitivity and univariate analyses based on these machine-learning models are performed to obtain insights into the microstructure-property relationships of the HPDC Al-Si alloy. The results show that the Gaussian process regression models exhibit high accuracy (R² greater than 0.84), thus confirming the viability of the proposed method. The results of sensitivity analysis indicate that the pore size exerts the most significant effect on the mechanical properties. Furthermore, the proposed framework can not only be transferred to other alloys but also be employed for material design.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00488-y

Accelerating the solving of mechanical equilibrium caused by lattice misfit through deep learning method

Chen-Xi Guo, Hui-Ying Yang, Rui-Jie Zhang

2024, 12(3): 512-521. doi:10.1007/s40436-024-00494-0

摘要 ( 84 )

PDF (223KB) ( 10 )

参考文献 | 相关文章 | 多维度评价

Precipitation is a common phenomenon that occurs during heat treatments. There is internal stress around the precipitate owing to the lattice misfit between the precipitate and matrix. This internal stress has a significant influence not only on the precipitation kinetics but also on the material properties. The misfit stress can be obtained by numerically solving the mechanical equilibrium equations. However, this process is complex and time-consuming. We developed a new approach based on deep learning to accelerate the solution process. The training data were first generated by a phase-field model coupled with elastic mechanical equilibrium equations, which were solved using the finite difference method. The obtained precipitate morphologies and corresponding stress distributions were input data for training the physics-informed (PI) UNet model. The well-trained PI-UNet model can then be applied to predicting stress distributions with the precipitate morphology as the input. Prediction accuracy and efficiency are discussed in this study. The results showed that the PI-UNet model was an appropriate approach for quickly predicting the misfit stress between the precipitate and matrix.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00494-0

Predicting fatigue life of automotive adhesive bonded joints: a data-driven approach using combined experimental and numerical datasets

Chen-Di Wei, Qiu-Ren Chen, Min Chen, Li Huang, Zhong-Jie Yue, Si-Geng Li, Jian Wang, Li Chen, Chao Tong, Qing Liu

2024, 12(3): 522-537. doi:10.1007/s40436-024-00500-5

摘要 ( 102 )

PDF (246KB) ( 37 )

参考文献 | 相关文章 | 多维度评价

The majority of vehicle structural failures originate from joint areas. Cyclic loading is one of the primary factors in joint failures, making the fatigue performance of joints a critical consideration in vehicle structure design. The use of traditional fatigue analysis methods is constrained by the absence of adhesive life data and the wide variety of joint geometries. Therefore, there is a pressing need for an accurate fatigue life estimation method for the joints in the automotive industry. In this work, we proposed a data-driven approach embedding physical knowledge-guided parameters based on experimental data and finite element analysis (FEA) results. Different machine learning (ML) algorithms are adopted to investigate the fatigue life of three typical adhesive joints, namely lap shear, coach peel and KSII joints. After the feature engineering and tuned process of the ML models, the preferable model using the Gaussian process regression algorithm is established, fed with eight input parameters, namely thicknesses of the substrates, line forces and bending moments of the adhesive bonded joints obtained from FEA. The proposed method is validated with the test data set and part-level physical tests with complex loading states for an unbiased evaluation. It demonstrates that for life prediction of adhesive joints, the data-driven solutions can constitute an improvement over conventional solutions.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00500-5

A data-driven approach for predicting the fatigue life and failure mode of self-piercing rivet joints

Jian Wang, Qiu-Ren Chen, Li Huang, Chen-Di Wei, Chao Tong, Xian-Hui Wang, Qing Liu

2024, 12(3): 538-555. doi:10.1007/s40436-024-00498-w

摘要 ( 100 )

PDF (219KB) ( 25 )

参考文献 | 相关文章 | 多维度评价

In lightweight automotive vehicles, the application of self-piercing rivet (SPR) joints is becoming increasingly widespread. Considering the importance of automotive service performance, the fatigue performance of SPR joints has received considerable attention. Therefore, this study proposes a data-driven approach to predict the fatigue life and failure modes of SPR joints. The dataset comprises three specimen types: cross-tensile, cross-peel, and tensile-shear. To ensure data consistency, a finite element analysis was employed to convert the external loads of the different specimens. Feature selection was implemented using various machine-learning algorithms to determine the model input. The Gaussian process regression algorithm was used to predict fatigue life, and its performance was compared with different kernel functions commonly used in the field. The results revealed that the Matern kernel exhibited an exceptional predictive capability for fatigue life. Among the data points, 95.9% fell within the 3-fold error band, and the remaining 4.1% exceeded the 3-fold error band owing to inherent dispersion in the fatigue data. To predict the failure location, various tree and artificial neural network (ANN) models were compared. The findings indicated that the ANN models slightly outperformed the tree models. The ANN model accurately predicts the failure of joints with varying dimensions and materials. However, minor deviations were observed for the joints with the same sheet. Overall, this data-driven approach provided a reliable predictive model for estimating the fatigue life and failure location of SPR joints.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00498-w

Improved genetic algorithm based on reinforcement learning for electric vehicle front-end structure optimization design

Feng-Yao Lyu, Zhen-Fei Zhan, Gui-Lin Zhou, Ju Wang, Jie Li, Xin He

2024, 12(3): 556-575. doi:10.1007/s40436-024-00495-z

摘要 ( 113 )

PDF (348KB) ( 18 )

参考文献 | 相关文章 | 多维度评价

The structural optimization of electric vehicles involves numerous design variables and constraints, making it a complex engineering optimization task over the past decades. Many population-based evolutionary algorithms encounter issues such as converging to local optima and lacking population diversity when tackling such optimization problems. Consequently, the solutions obtained for the optimization may be flawed or suboptimal. To address these problems, an improved genetic algorithm (GA) based on reinforcement learning is proposed in this paper. The proposed method introduces a population delimitation method based on individual fitness ranking. The population is divided into two parts: the excellent population and the ordinary population, and different selection and cross-mutation methods are applied to each part separately. More efficient crossover and mutation methods are then applied to the ordinary population to enhance the generation of excellent individuals. Furthermore, the proposed approach replaces the traditional fixed crossover and mutation rates with a dynamic selection method based on reinforcement learning to enhance optimization efficiency. A markov decision process model is constructed based on GA environment in this context. The population state determination method and reward method are designed for reinforcement learning in the GA environment, dynamically selecting the most appropriate genetic parameters based on the current state of the population. Finally, the uncertainty in the manufacturing process is introduced into the optimization problem and the case study results demonstrate that the proposed reinforcement learning-based GA significantly outperforms other evolutionary algorithms when applied to solving the structural optimization of electric vehicles.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00495-z

Deep learning methods for roping defect analysis in aluminum alloy sheets: prediction and grading

Yuan-Zhe Hu, Ru-Xue Liu, Jia-Peng He, Guo-Wei Zhou, Da-Yong Li

2024, 12(3): 576-590. doi:10.1007/s40436-024-00499-9

摘要 ( 118 )

PDF (215KB) ( 13 )

参考文献 | 相关文章 | 多维度评价

Roping is a severe band-like surface defect that occurs in deformed aluminum alloy sheets. Accurate roping prediction and rating are essential for industrial applications. Recently, the authors introduced an artificial neural network (ANN) model to efficiently forecast roping behavior across the thickness of large regions with texture gradients. In this study, the previously proposed ANN model for roping prediction is briefly reviewed, and a few-shot learning (FSL)-based method is developed for roping grading with limited samples. To consider the directionality of the roping patterns, the roping dataset constructed from experimental observations is transformed into the frequency domain for more compact characterization. A transfer-based FSL method is further presented for grade roping with manifold mixup regularization and the Sinkhorn mapping algorithm. A new component-focused representation is also implemented for data-processing, exploiting the close correlation between roping and power distribution in the frequency domain. The ultimate FSL method achieved an optimal accuracy of 95.65% in roping classification with only five training samples per class, outperforming four typical FSL methods. This FSL approach can be applied to grade the roping morphologies predicted by the ANN model. Consequently, the combination of prediction and grading using deep learning provides a new paradigm for roping analysis and control.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00499-9

Al-enabled properties distribution prediction for high-pressure die casting Al-Si alloy

Yu-Tong Yang, Zhong-Yuan Qiu, Zhen Zheng, Liang-Xi Pu, Ding-Ding Chen, Jiang Zheng, Rui-Jie Zhang, Bo Zhang, Shi-Yao Huang

2024, 12(3): 591-602. doi:10.1007/s40436-024-00485-1

摘要 ( 112 )

PDF (213KB) ( 36 )

参考文献 | 相关文章 | 多维度评价

High-pressure die casting (HPDC) is one of the most popular mass production processes in the automotive industry owing to its capability for part consolidation. However, the nonuniform distribution of mechanical properties in large-sized HPDC products adds complexity to part property evaluation. Therefore, a methodology for property prediction must be developed. Material characterization, simulation technologies, and artificial intelligence (AI) algorithms were employed. Firstly, an image recognition technique was employed to construct a temperature-microstructure characteristic model for a typical HPDC Al7Si0.2Mg alloy. Moreover, a porosity/microstructure-mechanical property model was established using a machine learning method based on the finite element method and representative volume element model results. Additionally, the computational results of the casting simulation software were mapped with the porosity/microstructure-mechanical property model, allowing accurate prediction of the property distribution of the HPDC Al-Si alloy. The AI-enabled property distribution model developed in this study is expected to serve as a foundation for intelligent HPDC part design platforms in the automotive industry.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00485-1

CMAGAN: classifier-aided minority augmentation generative adversarial networks for industrial imbalanced data and its application to fault prediction

Wen-Jie Wang, Zhao Liu, Ping Zhu

2024, 12(3): 603-618. doi:10.1007/s40436-024-00496-y

摘要 ( 148 )

PDF (215KB) ( 20 )

参考文献 | 相关文章 | 多维度评价

Class imbalance is a common characteristic of industrial data that adversely affects industrial data mining because it leads to the biased training of machine learning models. To address this issue, the augmentation of samples in minority classes based on generative adversarial networks (GANs) has been demonstrated as an effective approach. This study proposes a novel GAN-based minority class augmentation approach named classifier-aided minority augmentation generative adversarial network (CMAGAN). In the CMAGAN framework, an outlier elimination strategy is first applied to each class to minimize the negative impacts of outliers. Subsequently, a newly designed boundary-strengthening learning GAN (BSLGAN) is employed to generate additional samples for minority classes. By incorporating a supplementary classifier and innovative training mechanisms, the BSLGAN focuses on learning the distribution of samples near classification boundaries. Consequently, it can fully capture the characteristics of the target class and generate highly realistic samples with clear boundaries. Finally, the new samples are filtered based on the Mahalanobis distance to ensure that they are within the desired distribution. To evaluate the effectiveness of the proposed approach, CMAGAN was used to solve the class imbalance problem in eight real-world fault-prediction applications. The performance of CMAGAN was compared with that of seven other algorithms, including state-of-the-art GAN-based methods, and the results indicated that CMAGAN could provide higher-quality augmented results.

The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-024-00496-y

当期目录