Multiple-fault detection and identification scheme based on hierarchical self-organizing maps applied to an electric machine

Strategies of condition monitoring applied to electric motors play an important role in the competitiveness of multiple industrial sectors. However, the risk of faults coexistence in an electric motor and the overlapping of their effects in the considered physical magnitudes represent, currently, a critical limitation to provide reliable diagnosis outcomes. In this regard, additional investigation efforts are required towards high-dimensional data fusion schemes, particularly over the features calculation and features reduction, which represent two decisive stages in such data-driven approaches. In this study, a novel multiple-fault detection and identification methodology supported by a feature-level fusion strategy and a Self-Organizing Maps (SOM) hierarchical structure is proposed. The condition diagnosis as well as the corresponding estimated probability are obtained. Moreover, the proposed method allows the visualization of the results while preserving the underlying physical phenomenon of the system under monitoring. The proposed scheme is performed sequentially; first, a set of statistical-time based features is estimated from different physical magnitudes. Second, a hybrid feature reduction method is proposed, composed by an initial soft feature reduction, based on sequential floating forward selection to remove the less informative features, and followed by a hierarchical SOM structure which reveals directly the diagnosis and probability assessment. The effectiveness of the proposed detection and identification scheme is validated with a complete set of experimental data including healthy and five faulty conditions. The accuracy’s results are compared with classical condition monitoring approaches in order to validate the competency of the proposed method.     

Chaotic Flower Pollination and Grey Wolf Algorithms for parameter extraction of bio-impedance models

Precise parameter extraction of the bio-impedance models from the measured data is an important factor to evaluate the physiological changes of plant tissues. Traditional techniques employed in the literature for this problem are not robust which reflects on their accuracy. In this paper, the Flower Pollination Algorithm (FPA), the Grey Wolf Optimizer (GWO) and ten of their chaotic variants are employed to extract the parameters of bio-impedance models. These models are the simplified Hayden and Double-Shell models in their integer and fractional order forms. Experimental data sets of three different vegetables are employed and the performance of the chaotic variants of FPA and GWO are tested mathematically over CEC 2017 benchmark functions. The optimization results are compared based on fitting error and execution time to visualize the trade-off between accuracy, consistency, and speed of convergence. The fitting results are compared with the traditional Non-Linear Least Square (NLLS) fitting and Particle Swarm Optimization (PSO) algorithms. The target is to select the most adequate algorithm for extracting the parameters for each presented model. For the simplified Hayden model, both CGWO and CFPA achieved good results. However, there is a trade off between time and consistency of the results. If the time factor (consistency of the results) is the dominant factor, the CGWO (CFPA) algorithm will perform better. For the double shell model, the GWO and CGWO achieve more accurate and more consistent results and less execution time than the FPA and CFPA algorithms. The fractional order version of the studied models provid a better fitting to the measured data compared with their integer order counterparts.     

Building sustainable futures for post-industrial regeneration: the case of Taranto,Italy

In Italy, planning processes are often introduced by a Program of strategic objectives, drawn up in conjunction with the initial phases of a Strategic environmental assessment. This combination of preliminary processes offers significant space for strategic approaches and even strategic plans, particularly at the local level. This paper deals with Taranto, a socially and environmentally problematic and decaying industrial city in southern Italy. The present research shows and discusses the starting context of a recent, innovative strategic planning process, dealing with the complexity of the Taranto context through a scenario-building approach, hybridized with a cognitive-map-based interaction model.     

Optimum hot forming temperature of AZ61 magnesium alloy

A new concept of stability of materials is introduced by defining the optimum hot forming temperature for any given strain rate. This temperature is obtained through forming maps that are based on Lyapunov concepts and the introduction of a Garofalo equation in the Lyapunov criterion. This new approach is applied to a magnesium alloy AZ61. Torsion tests were carried out in the temperature range 574–734?K and strain rate range 0.7–8.7?s^?1 and the microstructures were determined using optical microscopy. Using the peak stress, optimum workability at 630?K is obtained at 12?s^?1. The results and the maps are compared with data and maps of other authors for AZ61 alloys in various states.     

H桥直流斩波变换器边界碰撞分岔和混沌研究

H桥变换器是一类被广泛应用于电力电子领域的重要拓扑,存在复杂的非线性现象,探索其非线性行为的产生机理具有重要的理论和实际意义。研究电流峰值控制型H桥直流斩波器的分岔和混沌现象,建立电流峰值控制H桥直流斩波器的离散映射模型,分析平衡点的稳定性,给出了系统稳定运行参数域。分析边界碰撞分岔和混沌演化过程及其产生机理,推导出碰撞边界线的解析表达式,指出发生边界碰撞的根本原因在于开关变换器占空比的有界性。数值仿真和实验验证了理论分析的正确性。     

混沌频率调制系统设计

针对混沌调制系统中存在相图信息泄密的问题,文中提出了基于混沌参数调制模型设计混沌频率调制系统的方法。该方法利用预传信号作为调制信号,调制混沌系统中特定的参数,使加载信息的混沌信号仅产生频谱分布上的变化,而信号的吸引子位置特性不随时间发生变化,同时文中也给出了基于误差检测的信号解调方法。通过证明及实验仿真,该系统实现了对连续信号的混沌频率调制,并有效地消除了混沌信号在相图中的分叉现象,以及提高了混沌调制系统的安全性能。     

A granular computing framework for self-organizing maps

When using granular computing for problem solving, one can focus on a specific level of understanding without looking at unwanted details of subsequent (more precise) levels. We present a granular computing framework for growing hierarchical self-organizing maps. This approach is ideal since the maps are arranged in a hierarchical manner and each is a complete abstraction of a pattern within data. The framework allows us to precisely define the connections between map levels. Formulating a neuron as a granule, the actions of granule construction and decomposition correspond to the growth and absorption of neurons in the previous model. In addition, we investigate the effects of updating granules with new information on both coarser and finer granules that have a derived relationship. Called bidirectional update propagation, the method ensures pattern consistency among data abstractions. An algorithm for the construction, decomposition, and updating of the granule-based self-organizing map is introduced. With examples, we demonstrate the effectiveness of this framework for abstracting patterns on many levels.     

基于分组卷积和特征图级联的轻量级目标检测

针对Pelee轻量级目标检测网络中参数量和计算量较多、检测精度较差等缺陷,提出了基于分组卷积和特征图级联的轻量级目标检测网络GCPelee。首先,利用分组卷积替换检测模块中的标准卷积形式以减少模型参数量和计算量;其次,在检测模块上应用特征图级联,将感受野较大的特征图包含的信息传递至感受野较小的特征图,提升后者的感受野大小。实验结果表明,优化后的GCPelee模型参数量和计算量均得到减少,检测精度得到了提升。     

跨阶段结构下的人体姿态估计

目的 基于图像的人体姿态估计是计算机视觉领域中一个非常重要的研究课题,并广泛应用于人机交互、监控以及图像检索等方面。但是,由于人体视觉外观的多样性、遮挡和混杂背景等因素的影响,导致人体姿态估计问题一直是计算机视觉领域的难点和热点。本文主要关注于初始特征对关节点定位的作用,提出一种跨阶段卷积姿态机（CSCPM）。方法 首先,采用VGG （visual geometry group）网络获得初步的图像初始特征,该初始特征既是图像关节点定位的基础,同时,也由于受到自遮挡和混杂背景的干扰难以学习。其次,在初始特征的基础上,构建多层模型学习不同尺度下的结构特征,同时为了解决深度学习中的梯度消失问题,在后续的各层特征中都串联该初始特征。最后,设计了多尺度关节点定位的联合损失,用于学习深度网络参数。结果 本文实验在两大人体姿态数据集MPII （MPII human pose dataset）和LSP （leeds sport pose）上分别与近3年的人体姿态估计方法进行了定性与定量比较,在MPII数据集中,模型的总检测率为89.1%,相比于性能第2的模型高出了0.7%;在LSP数据集中,模型的总检测率为91.0%,相比于性能第2的模型高出了0.5%。结论 实验结果表明,初始特征学习能够有效判断关节点的自遮挡和混杂背景干扰情况,引入跨阶段结构的CSCPM姿态估计模型能够胜出现有人体姿态估计模型。     

Computing area filling contours for surfaces defined by piecewise polynomials

For contour plotting, bivariate polynomials are often used to interpolate between the data points. The polynomials are defined over triangular or rectangular domains and have continuous interfaces resulting from various interpolation schemes. In this paper the principal ideas of the Trip Algorithm are presented, which finds closed polygons for filling the area between contour lines. These polygons consist of points P defining the contours, intersections S with the boundary of the domain, and vertices V of the domain. The points P are computed successively by a nonlinear method that combines extrapolation and the regula falsi in order to adjust the distance between the points to the curvature of the contour. For the extrapolation, derivatives of Lagrange polynomials are used. Empirical parameters for the automatic step size control are given. Once the points S are determined, the method is independent of the type of the interpolation function f(x, y). Two examples for applications of the Trip Algorithm are presented: one from scattered data interpolation, the other from stress analysis by finite element methods.