PBSVM: Partitioning and biased support vector machine for vocal fold pathology assessment using labeled and unlabeled data sets

Most of the existing classification methods, used for voice pathology assessment, are built based on labeled pathological and normal voice signals. This paper studies the problem of building a classifier using labeled and unlabeled data. We propose a novel learning technique, called Partitioning and Biased Support Vector Machine Classification (PBSVM), which tries to utilize all the available data in two steps: (1) a new heuristically partition-based algorithm, which extracts high quality pathological and normal samples from an unlabeled set, and (2) a more principle approach based on biased formulation of support vector machine, which is fairly robust to mislabeling and unbalance data problem. Experiments with wavelet-based energy features extracted from sustained vowels show that the new recognition scheme is highly feasible and significantly outperform the baseline classical SVM classifier, especially in the situation where the labeled training data is small.     

Fatigue crack initiation and damage characterization in Brazilian test specimens for adhesive joints

The present study is focused on the fatigue failure initiation at bimaterial corners by means of a configuration based on the Brazilian disc specimens. These specimens were previously used for the generalized fracture toughness determination and prediction of failure in adhesive joints, carried out under static compressive loading. Under static loading, local yielding effects might affect the asymptotic two-dimensional linear elastic stress representation under consideration. Fatigue loading avoids this fact due to the lower load levels used. The present tests were performed using load control; video microscopy and still cameras were used for monitoring initiation and crack growth. The fatigue tests were halted periodically and images of the corner were taken where fatigue damage was anticipated. Damage initiation and subsequent crack growth were observed in some specimens, especially in those which presented brittle failure under static and fatigue tests. These analyses allowed the characterization of damage initiation for a typical bimaterial corner that can be found in composite to aluminium adhesive lap joints.     

采用“GA+LM”优化BP神经网络的电液伺服阀故障诊断

针对标准BP神经网络用于故障诊断时学习效率低、收敛速度慢、易陷入局部极小点及对初始参数较为敏感等不足,提出了一种组合优化的方法,即采用遗传算法（GA）确定BP神经网络的最佳初始权值矩阵,以规避BP神经网络对初始参数较为敏感的不足;应用LM（Levenberg-Marquardt）算法在局部解空间里对BP神经网络进行精确训练,搜索全局最优解。该方法在保留BP神经网络的广泛映射能力的前提下,提升了网络的学习速度和精确搜索能力,进而大幅提高了基于BP神经网络的电液伺服阀故障诊断的效率和精度。通过对MOOG D761-2716A机械反馈伺服阀进行故障诊断,进一步说明了该方法的实用性和高效性。     

Experimental studies on the effects of sheet resistance and wettability of catalyst layer on electro-catalytic activities for oxygen evolution reaction in proton exchange membrane electrolysis cells

As the most important part of electrochemical reaction in proton exchange membrane electrolysis cells (PEMECs) for water splitting, oxygen evolution reaction (OER) occurs at the anode catalyst layer (CL). The distribution of the OER site is affected by many factors, such as properties of CL, operation parameters, procedures, etc. To study the effects of properties of CLs on the distribution of OER site on the CL, and consequently affect the performance of PEMECs, CLs with different sheet resistances are tested under different operation conditions. The phenomena of OER on CLs are captured by a high-speed and micro-scale visualization system in-situ and analysed coupled with electrochemical results. The results show that both sheet resistance and wettability of CLs have significant impact on the distribution of the OER site, which can help optimize the design of membrane electrode assembly and improve the operating parameters for electrochemical devices.     

爆炸载荷下泡沫铝夹芯板变形与破坏模式的实验研究

系统地开展了爆炸载荷作用下泡沫铝夹芯板变形与破坏的实验研究,获得了冲量45.6 N·s、 76.2 N·s、104.6 N·s、131.7 N·s、183.6 N·s 5种不同爆炸载荷作用下泡沫铝夹芯板背面板中心点的变形挠度,给出了泡沫铝夹芯板前面板、泡沫铝芯体和背面板在不同爆炸载荷作用下的变形与破坏模式,分析了泡沫铝芯体产生的剪切断裂和拉伸断裂两种不同机理。研究结果表明,泡沫铝芯体呈现“渐进式”压缩变形,泡沫铝夹芯板背面板中心点的变形挠度与爆炸冲量之间近似满足二次关系。     

Acoustic emission source location considering refraction in layered media with cylindrical surface

To solve the problem that the existing acoustic emission (AE) source location algorithms cannot always obtain accurate results for multilayer cylindrical media, a new acoustic emission source location method considering refraction was proposed. AE source coordinates were solved by the complex method. Pencil-lead-break experiments were used to verify this method. The absolute distance errors of location results are less than 3 mm, much less than those by the traditional method. The numerical experiments were used to further analyze factors that affect location accuracy. The results of numerical experiments show that the location accuracy of the proposed method is not affected by the ratio of wave velocities but affected by the measurement accuracy of wave velocity. These results show that new method can obtain accurate AE source location in the two-layered cylindrical surface media such as the triaxial compression test.     

Alcohol-assisted rapid growth of vertically aligned carbon nanotube arrays

Millimeter-to-centimeter scale vertically aligned carbon nanotube (VACNT) arrays are widely studied because of their immense potential in a range of applications. Catalyst control during chemical vapor deposition (CVD) is key to maintain the sustained growth of VACNT arrays. Herein, we achieved ultrafast growth of VACNT arrays using Fe/Al₂O₃ catalysts by ethanol-assisted two-zone CVD. One zone was set at temperatures above 850 °C to pyrolyze the carbon source and the other zone was set at 760 °C for VACNT deposition. By tuning synthesis parameters, up to 7 mm long VACNT arrays could be grown within 45 min, with a maximal growth rate of ∼280 μm/min. Our study indicates that the introduction of alcohol vapor and separation of growth zones from the carbon decomposition zone help reduce catalyst particle deactivation and accelerate the carbon source pyrolysis, leading to the promotion of VACNT array growth. We also observed that the catalyst film thickness did not significantly affect the CNT growth rate and microstructures under the conditions of our study. Additionally, the ultralong CNTs showed better processability with less structural deformation when exposed to solvent and polymer solutions. Our results demonstrate significant progress towards commercial production and application of VACNT arrays.     

Modelling of emission of volatile organic compounds from building materials—estimation of gas-phase mass transfer coefficient

A mathematical model is developed to predict Volatile Organic Compound (VOC) emission rates from homogeneous materials. The model considers both mass diffusion and mass convection processes in the boundary layer between the material surface and the air flow. Establishing the relationship between the surface air flow and emission rate; the model, therefore can predict the material emission rate under different environmental conditions. The other feature of the model is that all the parameters have clear physical meaning and can be either found in literature or calculated using known theories and/or equations.The prediction of the mathematical model was validated at three different levels; with experimental results from the CBS specially designed test chamber, with experimental results from the EPA which were carried out in an ASTM chamber, and finally with the predictions made by other models. The results indicate that there is, in general, good agreement between the model predictions and the experimental results. The main advantage of this model is that the model does not require any experimental data as input.     

Modularized design-oriented systematic inventive thinking approach supporting collaborative service innovations

The rapid evolution of new service systems raises crucial challenges for service design and requires effective methods. This study depicts a conceptual service design framework, called design-oriented systematic inventive thinking (DSIT) approach, which can be applied in different problem contexts. DSIT is presented as a new systematic and collaborative intelligence approach for creating and evaluating complex service systems using multi-criteria data analytics. DSIT synthesizes the current field of TRIZ service-design knowledge system and the emerging area of non-TRIZ service-design knowledge system. DSIT enables integrated development of service offerings at four dimensions and provides the matching integrated service design approach for each dimension. Four types of service design approaches are conceptualized as “human-independent service engineering,” “problem-clarified service engineering,” “solution-converged service engineering,” and “designing for service.” A new service computer-aided design system (service CAD) named DSIT explorer is developed consisting of customization, compatibility, and extensiveness of DSIT modules. A pervasive and smart collaborative service system (i.e., the smart MOS burger service solution) designed using DSIT explorer is illustrated. DSIT is a holistic, interdisciplinary, and collaborative service design concept, which is incorporated into a collaborative and intelligent service CAD framework to enable systematic inventive thinking throughout phases of service design lifecycle from problem definition, problem resolution, to solution evaluation.