时延动态神经网络的稳定性分析

基于时延动态网络模型的优化计算易于收敛到非法解或局部极小解,以及其算法对模型参数和初始条件具有很强的依赖性等缺点,对时延动态神经网络的稳定性进行了深入地研究讨论,通过Razumikhin-type定理对时延动态神经网络的稳定性进行了分析研究.     

行为NA的随机变量阵列的收敛性

利用NA序列的一个矩不等式,研究了行为NA的随机变量阵列,证明了1/n1p∑ni=1Xni平均收敛于0、完全收敛于0、几乎必然收敛于0、依概率收敛于0的等价性.     

一种信息点多样性的粒子群优化算法

针对PSO在寻优后期尤其在高维搜索空间中无法得到满意结果,提出了一种信息点多样性的改进粒子群优化算法。粒子个体最优位置及全局粒子最优位置是两个有用的精确的信息点,而PSO的信息交互方式正依赖于这两个信息点,从多样性方面考虑,将该有用的信息点增加为粒子个体最优位置附近随机的一点。实验仿真结果表明,新算法的全局搜索能力、收敛速度、精度和稳定性均有了显著提高。     

量纲分析法在土工模型试验中的应用

量纲分析对试验设计和试验结果的简明表达十分有用,有效的量纲分析可以产生一组无量纲成果.以地基承载力为例说明了Buckingham的π定理及Butterfield对其计算方法的改进.通过数值计算的方法对岸坡虹吸排水模型试验进行了分析,初步验证了量纲分析的成果,对模型试验的关键参数选取提出了一些建议.     

Stability analysis of shallow tunnels subjected to seepage with strength reduction theory

Based on strength reduction theory, the stability numbers of shallow tunnels were investigated within the framework of upper and lower bound theorems of limit analysis. Stability solutions taking into account of water seepage were presented and compared with those without considering seepage. The comparisons indicate that the maximum difference does not exceed 3.7%, which proves the present method credible. The results show that stability numbers of shallow tunnels considering seepage are much less than those without considering seepage, and that the difference of stability numbers between considering seepage and without considering seepage increase with increasing the depth ratio. The stability numbers decrease with increasing permeability coefficient and groundwater depth. Seepage has significant effects on the stability numbers of shallow tunnels.     

Effective prediction of DEA model by neural network

In this paper,a fast neural network model for the forecasting of effective points by DEA model is proposed,which is based on the SPDS training algorithm.The SPDS training algorithm overcomes the drawbacks of slow convergent speed and partially minimum result for BP algorithm.Its training speed is much faster and its forecasting precision is much better than those of BP algorithm.By numeric examples,it is showed that adopting the neural network model in the forecasting of effective points by DEA model is valid.     

Comprehensive assessment on dynamic roof instability under fractured rock mass conditions in the excavation disturbed zone

The damage process of fractured rock mass showed that the fracture in rocks induced roof collapse in Yangchangwan Coal Mine,China.The rock mass was particularly weak and fractured.There occurred 6 large-scale dynamical roof falls in the excavation disturbed zone (EDZ) with the collapsing volume of 216 m3.First,the field detailed geological environment,regional seismic dy-namics,and dynamic instability of roadways were generally investigated.Second,the field multiple-index monitoring measurements for detecting the deep delamination of the roof,convergence deformation,bolt-cable load,acoustic emission (AE) characteristic pa-rameters,total AE events,AE energy-releasing rate,rock mass fracture,and damage were arranged.Finally,according to the time-space-strength relations,a quantitative assessment of the influence of rock-mass damage on the dynamic roof instability was accomplished.     

不同分布φ^~混合序列的弱大数定理

通过利用矿混合序列的矩不等式推导出φ^~混合阵列行和的一个弱大数定理,并由φ^~混合阵列行和弱大数定理得出不同分布矿混合序列的一个弱大数定理。     

不对称级联型多电平变频器的分析及改进

不对称级联型多电平变频器拓扑结构各单元直流母线的电压不尽相同,采用混合控制策略,可以用较少的功率器件输出更多的电平数,达到更好的输出性能．引入最大延伸定理,为不对称级联型多电平变频器各单元选取电压比值提供理论依据,同时给出了一种改进型不对称级联型多电平变频器拓扑结构．     

偶数阶中立型差分方程最终正解的存在性

研究一类具连续变量偶数阶中立型时滞差分方程,利用Lebesgue控制收敛定理给出这类方程存在最终有界正解的一个充分必要条件,得到相应新的比较定理.