用分形和神经网络算法改善仪表精度和响应时间

研究在仪器仪表中利用分形插值和人工神经网络算法改善测试精度和响应时间的方法。在仪表检测中有时会遇到检测的分辨率与实时性相冲突的情况 ,此时对数据进行插值是一个很好的解决办法。利用分形和人工神经网络算法插值是一种可以进行多点数插值的优越方法 ,因为它可以通过训练学习不断修正网络的权值 ,使检测误差的方差控制在预定的范围。还研究了利用神经网络做谱分析来求取主频的方法 ,它在运算速度和分辨率方面都优于 FFT。文章还提出了一些减少人工神经网络学习训练时间的方法。     

基于G.P算法的关联维计算中参数取值的研究

从理论和实践两个方面对基于G．P算法的关联维计算中参数的取值问题进行了研究。分析了重构相空间维数m的取值原则;提出了标量r的取值范围和取值序列;对数据序列长度n的取值分别从理论和应用两个方面进行了探讨。所做的工作有益于分形理论在水声信号处理、雷达信号处理、地震信号分析、目标识别等领域的应用研究。     

Dewatering of alumino-humic sludge: impacts of hydroxide

The paper draws together information on factors which influence the conditioning and dewatering behaviour of an alum sludge gained from the coagulation of a low-turbidity coloured water. A principal focus is the potential impact of aluminium hydroxide on the sludge character. Background information is provided on the composition of the source floc for the domain pH 6.0-6.5 and Al>2.0mg/l. From this, there were many pointers to the presence of Al(OH)(3)(s) within the floc. A series of comparisons were made between an alum sludge and a hydroxide suspension at a concentration equivalent to the coagulant fraction within the sludge. The parameters studied included floc size, floc density, polymer adsorption and dewatering behaviour at different time-scales. In all cases, there were strong similarities in the behaviour of the two suspensions-indicating the potential impact of the hydroxide. There was also evidence of common features being displayed by both the organic fractions and the hydroxide. It was suggested that some of the behavioural features might emanate from a common fractal structure within the source floc, the fractal dimension (approximately 1) being insensitive to composition.     

分形研究中无标度区的计算机识别

基于相空间重构理论和嵌入定理,提出了一种分形研究中无标度区的计算机识别方法。应用结果表明,该方法便于编程实现,并能准确地识别出系统观测序列的分形无标度区,计算效率高,人为因素对计算结果的影响小。     

Self-affinity of fracture surfaces and implications on a possible size effect on fracture energy

As demonstrated within the last 15 years by numerous experimental studies, tensile fracture surfaces exhibit a self-affine fractal geometry in many different materials and loading conditions. In the last few years, some authors proposed to explain an observed size effect on fracture energy by this fractality. However, because they did not consider a lower bound to this scale invariance (which necessarily exists, at least at the atomic scale), they had to introduce a new definition of fracture energy with unconventional physical dimensions. Moreover, they were unable to reproduce the observed asymptotic behavior of the apparent fracture energy at large specimen sizes. Here, we show that this is because they considered self-similar fracture surfaces (not observed in nature) instead of self-affine. It is demonstrated that the ignorance of the self-affine roughness of fracture surfaces when estimating the fracture energy from the work spent to crack a specimen necessarily leads, if the work of fracture is proportional to the fracture area created, to a size effect on this fracture energy. Because of the self-affine (instead of self-similar) character of fracture surfaces, this size effect follows an asymptotic behavior towards large scales. It is therefore rather limited and not likely detectable for relatively large sample sizes (10^–1 m). Consequently, significant and rapid increases of the apparent fracture energy are more likely to be explained mainly by other sources of size effect.     

热图象场的分形测量方法研究

影像的分形研究是图象分析中的一个新课题,该文通过对分形原理、传统的分形计算方法的研究,结合热图象特征给出了３种不同的分维计算方法－灰度曲面分形计算,剖面曲线分形计算,象元点分形计算,实践表明灰度曲线分形计算可以揭示热图象整体的分形特征;剖面曲线分形计算则可对比研究典型地域的热辐射特征;象元点分形计算可以给出热辐射变化梯度的定量描述。在讨论有关分形性质后,给出了上述方法在上海市热图象场分形测量中的良     

基于分形的DCT域数字水印算法研究

提出一种改进分形数字水印算法,对传统的分形数字水印进行改进,利用混沌序列对水印图像进行置乱处理,增强了水印的鲁棒性.实验结果表明,该算法具有较好的鲁棒性,而且对高斯噪声以及剪切等都有良好的稳健性.     

基于分形理论的DEM数据内插算法研究

本文利用分形理论与分形布朗运动原理,提出了基于子分形维数的随机中点位移内插算法,并采用不同的插值比系数来有效避免插值点附近的"尖峰现象"。应用该算法对研究区内的低分辨率数字高程模型(DEM)格网数据进行了内插计算,结果显示内插精度有一定的提高,同时生成的地形起伏平稳,并且能较逼真的反映地形表面的细节。     

基于DLA模型的植物生长模拟研究

归纳了分形算法中的Diffusion Limited Aggregation模型（即DLA模型）的分形演化算法，提供了基于DLA模型在计算机上进行植物生长模型的实现方法。模型利用Borland Delphi高级语言进行了实现，验证了基于DLA模型的植物生长模拟方法。