深度置信网络的代价敏感多粒度三支决策模型研究

最优粒度选择是自编码网络构造多粒度特征的关键环节。针对自编码网络粒度选择方法不合理导致特征提取效果差以及错误分类成本和测试成本高的问题,提出一种基于小批量梯度下降(mini-batch gradient descent, MBGD)的粒度层选取策略。该方法通过改变粒度选择方式重新构建多粒度空间,设计一个新的基于深度置信网络(deep belief network, DBN)的代价敏感多粒度三支决策模型。更优的粒度选择方法提升网络的特征提取能力,促使多粒度空间的构造朝着最快到达最细粒度空间的方向发展,降低图像重构误差以达到更小的错误分类代价和测试代价。实验结果表明,提供合理的粒度选取策略提高了代价敏感多粒度三支决策模型的决策准确性,并在给定代价情况下更快地获得总代价最小的最优粒层。     

Tool positioning error (TPE) characterisation in milling

Where the geometrical features so permit, the {workpiece–work-holding fixture} assembly is generally considered to be infinitely rigid. The {tool–tool-holder–spindle} assembly and the machine axes are then deformed under the action of the cutting forces. This deformation leads to a positioning error of the tool in relation to the theoretical position. With the aim of taking this positioning error into account, the inaccuracies obtained during end milling and side milling were experimentally modelled from the cutting conditions used for a given machine/mill/material triplet (TriM). Our “Virtual Worker” then used these models to predict machining errors according to the type of machining and to compensate for them.     

A virtual age model based on a bathtub shaped initial intensity

This paper presents a new reliability model for complex repairable systems, which combines a bathtub shaped ageing and imperfect maintenance. A bathtub shaped initial intensity function allows to take into account the burn-in period, the useful life and wear out of the systems. Repair effect is expressed by a reduction of the system virtual age, which depends on the ageing of the system. The main characteristics of the model are derived. The most important one is that the maintenance efficiency allows an extension of the system useful life duration. A statistical analysis of the model and an application to real failure data are presented.     

Lossy compression of volumetric medical images with 3D dead-zone lattice vector quantization

This paper presents a new lossy coding scheme based on 3D wavelet transform and lattice vector quantization for volumetric medical images. The main contribution of this work is the design of a new codebook enclosing a multidimensional dead zone during the quantization step which enables to better account correlations between neighbor voxels. Furthermore, we present an efficient rate–distortion model to simplify the bit allocation procedure for our intra-band scheme. Our algorithm has been evaluated on several CT- and MR-image volumes. At high compression ratios, we show that it can outperform the best existing methods in terms of rate–distortion trade-off. In addition, our method better preserves details and produces thus reconstructed images less blurred than the well-known 3D SPIHT algorithm which stands as a reference.     

Kinetic Selectivity of Olefin Metathesis Catalysts Bearing Cyclic (Alkyl)(Amino)Carbenes

The evaluation of ruthenium olefin metathesis catalysts 4-6 bearing cyclic (alkyl)(amino)carbenes (CAACs) in the cross-metathesis of cis-1,4-diacetoxy-2-butene (7) with allylbenzene (8) and the ethenolysis of methyl oleate (11) is reported. Relative to most NHC-substituted complexes, CAAC-substituted catalysts exhibit lower E/Z ratios (3:1 at 70% conversion) in the cross-metathesis of 7 and 8. Additionally, complexes 4-6 demonstrate good selectivity for the formation of terminal olefins versus internal olefins in the ethenolysis of 11. Indeed, complex 6 achieved 35 000 TONs, the highest recorded to date. CAAC-substituted complexes exhibit markedly different kinetic selectivity than most NHC-substituted complexes.     

A roadmap to understand battery performance in electric and hybrid vehicle operation

This work attempts to bridge laboratory and real-life battery testing data with a comprehensive analysis to provide a coherent approach for a realistic model to simulate battery performance, including life prediction. From electric vehicle field-testing results, we explain how to handle real-life data through driving cycle analysis to establish a scheme of “building blocks” that can be validated by test results obtained in the laboratory. We also show that a simple battery model can be built upon laboratory test data and validated by real-life duty cycles, therefore deriving a more realistic understanding and prediction of battery performance.     

Effect of discharge rate on charging a lead-acid battery simulated by mathematical model

To simulate lead-acid battery (LAB) charging has never been an easy task due to the influences of: (1) secondary reactions that involve gas evolution and recombination and grid corrosion, (2) prior end-of-discharge (EOD) and rest conditions; and (3) complexity caused by charging algorithm. In this work, successful results have been obtained with considerations of internal oxygen cycle and gas phase in the valve-regulated lead-acid (VRLA) cells. The success is first attributed to the satisfactory validation of a mathematical model that has been able to simulate discharge regimes with various rates consistently. The model has been subsequently used to simulate a galvanostatic charge regime performed at C/10. The results give a better understanding of the role each electrode played in the polarization, the nature of the polarization (constituted by reaction kinetics and mass transport), and the charging efficiency. We were able to extrapolate the simulation results to rates beyond what the model has been validated for, and the results are still consistent, confirming some experimental observations, notably the maximum charging rate specified by most LAB manufacturers.