Analysis and optimization of high temperature proton exchange membrane (HT-PEM) fuel cell based on surrogate model

The stoichiometric ratio and flow channel geometry play a vital role in the performance of high temperature proton exchange membrane (HT-PEM) fuel cells. Because of the high cost of experiments or simulations, most analyses and optimization of the stoichiometric ratio and flow channel geometry are limited to several points in the entire design domain. In this study, an analysis and optimization method for HT-PEM fuel cells based on the surrogate model was proposed. Surrogate models were constructed using some of the available budgets of samples to analyze and optimize the entire design domain. With this method, it was indicated that the effect of the cathode stoichiometric ratio is more significant to the cell performance than the anode stoichiometric ratio and there are significant nonlinear interactions among the flow channel geometry parameters. At the fixed operating voltage, the flow channel geometry with the maximum current density and that with the maximum real power were obtained. Compared with the base design, the designs obtained by the surrogate model improve the current density and real power by 10.54% and 3.93%, respectively. Thus, this analysis and optimization method is demonstrated to be helpful and deserves attention in future research.     

Simulation and sensor data fusion for machine learning application

The performance of machine learning algorithms depends to a large extent on the amount and the quality of data available for training. Simulations are most often used as test-beds for assessing the performance of trained models on simulated environment before deployment in real-world. They can also be used for data annotation, i.e, assigning labels to observed data, providing thus background knowledge for domain experts. We want to integrate this knowledge into the machine learning process and, at the same time, use the simulation as an additional data source. Therefore, we present a framework that allows for the combination of real-world observations and simulation data at two levels, namely the data or the model level. At the data level, observations and simulation data are integrated to form an enriched data set for learning. At the model level, the models learned from observed and simulated data separately are combined using an ensemble technique. Based on the trade-off between model bias and variance, an automatic selection of the appropriate fusion level is proposed. Our framework is validated using two case studies of very different types. The first is an industry 4.0 use case consisting of monitoring a milling process in real-time. The second is an application in astroparticle physics for background suppression.     

Relationship between chatter marks and rolling force fluctuation for twenty-high roll mill

Chatter marks on the surface of a rolled steel strip may seriously affect the final product quality. It is still difficult to discover the generation mechanism of chatter marks and develop an effective method to identify chatter marks. In this paper a rolling force fluctuation model is proposed based on the variation of the rolling force between the forward and backward skid area. A dynamic model coupling vertical and horizontal vibrations is also established to analyze the vibration characteristics of a Sendzimir twenty-high cold roll mill based on the rolling force fluctuation model. Good agreement between the simulated results and the experimental results verifies the effectiveness of the established models. Furthermore, various rolling forces and rolling speeds are taken into consideration in analyzing the vibration characteristics. The proposed models provide clear theoretical support for better comprehension of the mechanism of chatter mark generation.     

The stabilization of a methane–air edge flame within a mixing layer in a narrow channel

The flame stabilization mechanism of a methane–air edge flame formulated in a narrow channel was experimentally investigated and compared with a simple analytical model. Non-premixed flames were classified into premixed flame modes and edge flame modes. The correlation between the propagation velocity and the fuel concentration gradient in a narrow channel was investigated and the applicability of ordinary edge-flame theory was appraised.     

Effect of interfacial heat transfer on the onset of oscillatory convection in liquid bridge

In present study, effect of interfacial heat transfer with ambient gas on the onset of oscillatory convection in a liquid bridge of large Prandtl number on the ground is systematically investigated by the method of linear stability analyses. With both the constant and linear ambient air temperature distributions, the numerical results show that the interfacial heat transfer modifies the free-surface temperature distribution directly and then induces a steeper temperature gradient on the middle part of the free surface, which may destabilize the convection. On the other hand, the interfacial heat transfer restrains the temperature disturbances on the free surface, which may stabilize the convection. The two coupling effects result in a complex dependence of the stability property on the Biot number. Effects of melt free-surface deformation on the critical conditions of the oscillatory convection were also investigated. Moreover, to better understand the mechanism of the instabilities, rates of kinetic energy change and “thermal” energy change of the critical disturbances were investigated     

Enhancing ductile removal of Cf/SiC composites during abrasive belt grinding using low-hardness rubber contact wheels

C_f/SiC composites are used as advanced thermal protection and friction materials. However, machining these materials is difficult because of their hard, brittle, anisotropic, and heterogeneous characteristics. This study investigated the removal behavior and surface integrity of C_f/SiC composites during abrasive belt grinding using rubber contact wheels of various hardness. Additionally, detailed analysis was performed on their thermal-mechanical coupling characteristics, surface integrity (that is, surface roughness, surface micro morphology, and subsurface damages), and the grinding chips produced. Results revealed that with decreasing hardness of the contact wheel, the surface roughness in all directions, grinding force, and temperature decreased significantly. Moreover, the surface removal morphology of the C_f/SiC composites changed from macro-fracture to micro-fracture, and the subsurface morphology changed from SiC matrix cracking and carbon fibers pull-out to matrix plastic flow and fiber micro-fracture, respectively. Furthermore, strip chips with plastically squeezed and cut surfaces were visible in the grinding chips obtained under the 40-HA contact wheel. Therefore, the ductile removal behavior of the C_f/SiC composites was enhanced, and the surface quality in abrasive belt grinding with low-hardness contact wheels was markedly improved.     

钛合金表面缓进给砂带磨削变参数优化方法及其试验研究

目的 改善钛合金砂带全生命周期中磨削的表面质量。方法 提出了钛合金缓进给砂带磨削变参数优化方法。首先,采集磨削过程中的加工参数、砂带磨损、表面粗糙度等数据。其次,采用SVM算法构建以磨削参数和磨损数据为输入、以表面粗糙度为输出的粗糙度预测模型,并且以预测的粗糙度和砂带磨损为约束应用NSGA-Ⅱ算法,针对缓进给砂带磨削过程中的全生命周期的加工参数进行优化。最后,通过对比分析变参数和固定参数磨削方法下的砂带磨损特点和钛合金表面粗糙度、形貌特征、微观特征、表面氧化的特点,对砂带全生命周期变参数磨削方法进行验证。结果 SVM预测的精度可达0.95以上,MAE低至0.064。采用NSGA-Ⅱ算法优化后的加工参数能够有效地改善表面质量,优化前的全生命周期中的粗糙度从0.787μm逐渐降低至0.509μm,优化后的粗糙度从0.934μm降低至0.457μm;并且优化后的钛合金形貌要优于传统的加工方式,变参数磨削的钛合金表面氧化程度明显小于固定参数磨削方法。此外,提出的变参数优化方法能够有效地改善砂带的磨损,降低缓进给磨削所带来的砂带快速磨损现象。结论 本文所提出的SVM-NSGA-Ⅱ磨削参数优化算法...     

A novel combined molecular dynamics–micromechanics method for modeling of stiffness of graphene/epoxy nanocomposites with randomly distributed graphene

In this paper, by combining molecular dynamics and micromechanics methods, a new approach for prediction of the stiffness of the nanocomposites with randomly distributed nanoparticles in the macro level is presented. The molecular dynamics method is used to model the stiffness of the graphene/epoxy nanocomposites containing one layer of an aligned nano graphene embedded in epoxy resin. By considering the large sizes of the length and width of the nano graphene in comparison with its thickness and the shortcomings of the available hardware and software for simulation purposes, a new approach for modeling is also developed. This new approach, by using the moduli of different graphene sheets with different sizes embedded in a representative volume element, can predict the moduli of a real size graphene embedded in the matrix along the longitudinal, transverse and normal directions in the nano-scale. In order to consider the effect of the random distribution of graphene sheets in epoxy resin, a micromechanical approach is used. The results obtained by the molecular dynamics method are used by the micromechanics approach and the stiffness of graphene/epoxy nanocomposites with randomly distributed graphene in the macro-scale is predicted. An experimental program is conducted to evaluate the capability of the model. The result of the modeling is in a very good agreement with the experimental data.