Multi-objective genetic algorithms based automated clustering for fuzzy association rules mining

Researchers realized the importance of integrating fuzziness into association rules mining in databases with binary and quantitative attributes. However, most of the earlier algorithms proposed for fuzzy association rules mining either assume that fuzzy sets are given or employ a clustering algorithm, like CURE, to decide on fuzzy sets; for both cases the number of fuzzy sets is pre-specified. In this paper, we propose an automated method to decide on the number of fuzzy sets and for the autonomous mining of both fuzzy sets and fuzzy association rules. We achieve this by developing an automated clustering method based on multi-objective Genetic Algorithms (GA); the aim of the proposed approach is to automatically cluster values of a quantitative attribute in order to obtain large number of large itemsets in less time. We compare the proposed multi-objective GA based approach with two other approaches, namely: 1) CURE-based approach, which is known as one of the most efficient clustering algorithms; 2) Chien et al. clustering approach, which is an automatic interval partition method based on variation of density. Experimental results on 100 K transactions extracted from the adult data of USA census in year 2000 showed that the proposed automated clustering method exhibits good performance over both CURE-based approach and Chien et al.’s work in terms of runtime, number of large itemsets and number of association rules.     

引入信息熵的CURE聚类算法

为了提高传统CURE(Clustering Using REpresentatives) 聚类算法的质量,引入信息熵对其进行改进。该算法使用K-means算法对样本数据集进行预聚类;采用基于信息熵的相似性度量,利用簇中的元素提供的信息度量不同簇之间的相互关系,并描述数据的分布;在高层、低层聚类阶段,采取不同的选取策略,分别选取相应的代表点。在UCI数据集和人造数据集上的实验结果表明,提出的算法在一定程度上提高了聚类的准确率,且在大型数据集上比传统CURE算法有着更高的聚类效率。     

基于马氏距离的双层聚类电力远动异常检测

通过对电力远动监测系统和数据挖掘技术的讨论,提出一种基于马氏距离的双层聚类异常检测算法。针对远动系统数据非球面分布的特点,该算法通过K-means聚类改进算法对数据进行初始分类,然后使用基于马氏距离的Clustering Using Representatives(CURE)聚类改进算法对初始分类结果进行优化,以较少的计算成本去除K值设定的影响,达到预期的检测结果。同时,基于马氏距离的CURE聚类改进算法对球面和非球面分布的数据有非常好的适应能力。     

基于CURE聚类算法改进的原型选择算法

针对传统K近邻分类器在大规模数据集中存在时间和空间复杂度过高的问题,可采取原型选择的方法进行处理,即从原始数据集中挑选出代表原型（样例）进行K近邻分类而不降低其分类准确率.本文在CURE聚类算法的基础上,针对CURE的噪声点不易确定及代表点分散性差的特点,利用共享邻居密度度量给出了一种去噪方法和使用最大最小距离选取代表点进行改进,从而提出了一种新的原型选择算法PSCURE （improved prototype selection algorithm based on CURE algorithm）.基于UCI数据集进行实验,结果表明：提出的PSCURE原型选择算法与相关原型算法相比,不仅能筛选出较少的原型,而且可获得较高的分类准确率.     

大数据环境下超声波焊缝缺陷识别方法的研究

为了解决常规超声波焊缝缺陷识别方法分类模型固定和训练集规模有限而难以体现不同缺陷的差异性和同类缺陷的多态性的问题,结合当今大数据环境下的数据分析策略和基因缺陷识别中匹配的思想,通过主成分分析和CURE聚类算法将缺陷回波信号编码转换成可进行匹配的对象,进而将当前检测缺陷特征与历史检测数据进行匹配,并利用最近邻方法实现了对缺陷历史检测数据集的扩充。通过在R上应用基于基本空位罚分的Smith-Waterman比对算法进行仿真实验验证了该缺陷识别方法是可行的,有效地识别了气孔、夹渣、裂纹、未焊透和未熔合五类常见缺陷,具有较好的识别准确率。     

Effect of process variables on melt temperature profiles in extrusion process using single screw plastics extruder

Abstract

Melt temperature is an important parameter in the melt processing of polymers. However, it is not possible to control melt temperature directly, only to influence it using processing parameters such as processing temperature settings. It is therefore important to know the influence of controllable process parameters on melt temperature. In this work, the relationships between controllable process parameters and melt temperature have been investigated for a 50 mm S + B single screw extruder. The extruder was equipped with a thermocouple mesh at the die inlet to determine melt temperature. It was found that melt pressure, die size, feed section barrel temperature, and compression section barrel temperature had a negligible effect on the melt temperature profiles generated, while increasing the screw speed resulted in higher melt temperatures. The metering section barrel temperature had a significant effect on melt temperature, thermal conduction effects being more important than shear heating effects. Equipment wall temperatures, downstream of the screw, produced changes in the melt temperature in the melt located within 7 mm of the wall. It was found that melt temperatures can be significantly different from those set on the equipment.     

Rheometers in scanning mode for cure of rubbers with low enthalpy: effect of heating rate and rubber–die contact

Abstract

Rheometers are usually run under isothermal conditions for the cure of rubbers, in spite of the following drawbacks: three experiments at least are necessary at different temperatures selected within a narrow temperature window, and it takes some time for thermal equilibrium to establish between the rubber sample initially at room temperature and the dies. In scanning mode, the dies of the rheometer and the sample should be heated from room temperature up to the selected final temperature at a constant rate. The effect of the value given to the heating rate, which is the main factor in this new method, is especially studied, as well as the quality of the contact between the dies and the sample. The rubber used in the present study is a natural rubber vulcanised with 2% sulphur with a low enthalpy. The kinetic parameters of this cure reaction have been previously determined by isothermal rheometry, and the enthalpy measured by calorimetry. The state of cure–temperature curves in scanning mode have been calculated using a numerical model taking into account the conduction heat transfer through the rubber, the coefficient of heat transfer at the die–rubber interface, and the kinetics of the heat evolved from the cure reaction with the enthalpy. The kinetic parameters are determined from the state of cure–temperature curves, which are similar to the torque–temperature curves, using a least squares method. The values of the kinetic parameters obtained by this last calculation, based on the mathematical treatment of these curves, are the same as those introduced for obtaining these curves using the numerical model. Thus, whatever the heating rate, ranging from 2 to 10 K min ^–1 , and the value of the coefficient of heat transfer at the die–rubber interface expressing the quality of the contact, in scanning mode the moving die rheometer (MDR) can be used to determine the kinetics of the cure of rubbers.     

Surface quality prediction of thermoset composite structures using geometric simulation tools

Abstract

The surface quality of polymer composite laminates was examined via geometric modelling techniques and compared to experimental data. TexGen software provided the platform for the development of a surface roughness simulation tool which accounted for textile architecture and specific cure kinetics of the matrix. The study focused on the influence of thermal and chemical shrinkage during cure and the change in localised volume fraction across the surface of a unit cell. A one-dimensional analysis was used to determine proportional dimensional changes to the matrix region, with the results stitched together to form a three-dimensional topological plot. Three demonstrator geometric models were developed to represent a carbon 2 × 2 twill weave fabric with 3000, 6000 or 12 000 tows. These models were analysed with low and high shrink resin properties. Optical microscopy was used to determine accurate tow forms for compacted tows which aided the development of the geometric model. Simulated profiles, topography and surface roughness measures were compared to experimental data which demonstrated the significance of matrix contraction and fabric architecture on the final surface quality. The simulations were shown to represent experimental data typically within 6%.