基于结合自适应步长布谷鸟搜查算法的模糊神经网络的软件可靠性增长模型

针对现有的软件可靠性增长模型(SRGM)适用性较差、预测精度波动大的问题,使用自适应步长布谷鸟搜查(ASCS)算法对模糊神经网络（FNN）的权重和阈值进行寻优,利用得到了最优权重和阈值的FNN建立SRGM。在使用缺陷数据对FNN训练的过程中,利用ASCS来调整FNN的权重和阈值,以此提高在预测过程中的精度,同时采用多次预测结果取均值的方式来减小FNN预测的波动性,以此建立基于结合自适应步长布谷鸟搜查算法的模糊神经网络（ASCS-FNN）的软件可靠性增长模型。利用3组软件缺陷数据,以误差比均值和误差平方和作为衡量标准,对基于ASCS-FNN、结合模拟退火算法的动态模糊神经网络（SA-DFNN）、FNN、BP网络（BPN）建立的SRGM的一步向前预测能力进行比较。预测结果表明,在四组模型中,基于ASCS-FNN建立的SRGM相对于SA-DFNN、FNN、BPN建立的SRGM的平均预测精度相对提高率RI(AE)和RI(SSE)分别为-1.48%、54.8%、33.8%和14.4%、76%、35.9%,并且该模型比FNN、BPN建立的SRGM在相同缺陷数据下的预测波动性小,而且网络结构比SA-DFNN的网络结构简单。因此该模型具有预测精度较高、预测稳定和具有一定的适用性等优点。     

1553B总线通信终端知识产权核的设计

为满足航天飞行器地面仿真设备使用的需求,设计了一种基于可编程逻辑门阵列(FPGA)的1553B总线通信终端知识产权(IP)核。在保证总线系统可靠性的前提下,采用自顶向下的设计方法与“双进程”编码方式,利用超高速硬件描述语言(VHDL)生成目标代码,使用ModelSim软件进行仿真,最后在实际设备中验证并应用。该IP核可配置在总线控制器、远程终端或总线监控器3种不同的工作模式下运行,易于集成入片上系统(SoC),对进一步应用1553B总线提供了更多的选择。     

集中有序集值信息系统

集值信息系统是完备信息系统的一种推广,按照语义可划分为合取集值信息系统和析取集值信息系统。属性偏好关系也有两种：属性递增偏好有序和属性递减偏好有序。提出一种新的属性偏好关系,建立了一种新的优势关系。这种优势关系能够表示一类属性偏好既不是递增有序也不是递减有序,而是趋近于某个标准值的情形,称这样的优势关系为属性集中有序,它可应用于某些集值信息系统。     

面向不完全维修的数控机床可靠性评估

为实现不完全维修条件下的数控机床可靠性评估,提出基于对数线性比例强度模型(Log-linear proportional intensity model, LPIM)的多台数控机床可靠性评估方法,建立故障时间的对数线性比例强度函数式。用极大似然估计法和Fisher信息矩阵法给出了模型参数的点估计与区间估计,给出机床可靠性指标的点估计和基于Delta法的区间估计,采用似然比检验方法对时间趋势和修复功效的显著性进行检验。实例分析表明,考虑维修活动影响时,LPIM优于Weibull分布法和非齐次泊松过程模型,能定量反映出维修对机床可靠性的作用,得出的瞬时和累积可靠性特性更符合实际。     

Thermal-Cyclic Fatigue Life Analysis and Reliability Estimation of a FCCSP based on Probabilistic Design Concept

To study the fatigue reliability of a flip-chip chip scale package (FCCSP) subject to thermal cyclic loading, a Monte Carlo simulation-based parametric study is carried out in the present study. A refined procedure as compared with the recently released Probabilistic Design System (PDS) of ANSYS is proposed and employed in particular. The thermal-cyclic fatigue life of the package is discussed in detail since it is related directly to the reliability of the package. In consideration of the analytical procedure as well as real manufacturing processes, a few geometric dimensions and material properties of the package are assumed random. The empirical parameters used in the fatigue life prediction formula are also assumed random to account for their uncertainties. Numerical calculation is performed following the standard finite element analysis procedure. The result indicates that PDS can indeed be employed to find the cumulative thermal-cyclic fatigue life distribution of the electronic package owing to various uncertainties. The proposed refined design procedure can further improve the accuracy of the quantitative reliability estimation.     

蒙特卡罗法在露天矿边坡可靠性分析中的应用

基于现场地质勘察和室内试验得到的岩体力学性能参数,利用蒙特卡罗法对某露天矿长期开采形成的边坡进行了6个剖面可靠性分析,获取某矿山边坡的破坏概率和可靠指标,进而为该矿山边坡稳定性评价和后续边坡的管理工作提供可靠的依据。     

基于UGF和Semi-Markov方法的反应堆泵机组多状态可靠性分析

将通用发生函数(UGF)与半马尔可夫过程(Semi-Markov Process)相结合,对反应堆泵机组进行多状态可靠性分析。给出多状态系统可靠性分析的UGF算法模型,推导多状态设备性能状态的Semi-Markov过程概率表达式,定义设备、系统性能值的定量描述方法。以性能参数是否满足需求值做为系统成功与失效的判据,对比分析反应堆泵机组在需求性能条件下的多状态可用度与2状态可用度结果,并给出系统在任务周期内的平均性能值。结果表明,该方法能够定量分析部分失效对系统可靠性的影响,降低传统的2态可靠性分析方法产生的不必要的保守程度。     

基于常数对角优势补偿阵的多变量控制系统逆Nyquist阵列设计

化工过程中的多变量系统变量之间往往存在耦合作用,在控制系统设计时一般将传递函数阵对角优势化,对角优势化后按多个单变量系统进行设计。由于传递函数逆阵的对角优势化更容易实现,本文采用伪对角化方法设计常数对角优势补偿阵实现传递函数逆阵的对角优势化,在一个或多个频率点上通过使开环传递函数逆阵每行的非对角项元素模平方之和最小,实现对角优势。然后,利用逆Nyquist阵列设计法对补偿后的系统设计控制器,基于逆Nyquist稳定判据,通过绘制优势度曲线和Gershgorin带判断系统的优势程度,根据Gershgorin带选定反馈矩阵的参数范围,按照单变量控制系统的方法设计动态补偿器,使系统满足动态控制品质的要求。最后通过3个示例说明本文的设计方法简便,且具有良好的控制性能。     

Design optimization for the hybrid power system of a green building

This paper proposes an optimal design procedure for a green building equipped with renewable energy, energy storages, and proton exchange membrane fuel cells (PEMFCs). First, we introduce the hybrid power system of the green building and construct a simulation model using Matlab/SimPowerSystem?. The model parameters are tuned so that the system responses can be estimated without extensive experiments in the optimization processes. Second, we define the cost and reliability indexes to optimize the system design using three steps: component selection, component sizing, and power management (PM) adjustment. We further define the safety index to evaluate the system's sustainability under extreme conditions when no renewable energy is available. Last, we apply the proposed procedures to the green building and demonstrate the benefits of the optimal design. The proposed method can be directly applied to develop customized hybrid power systems in the future.     

Level of circulating concentrations of progesterone during ovulatory follicle development affects timing of pregnancy loss in lactating dairy cows

The objective of this experiment was to determine the effect of high versus low progesterone (P4) during the pre-dominance or dominance phase (or both) of ovulatory follicle development on follicular dynamics and fertility of lactating dairy cows. Progesterone (P4) was manipulated to reach high (H) or low (L) serum concentrations during the pre-dominance phase (d 0 to 4 of the wave) and dominance phase (d 5 to 7 of the wave) of a second follicular wave ovulatory follicle, creating 4 treatments: H/H, H/L, L/H, and L/L. Luteolysis was induced with PGF_2α on d 7 of the wave and ovulation was induced with GnRH 56 h after PGF_2α. Cows (n = 558) received artificial insemination (AI) 16 h following GnRH. Pregnancy was determined at 6 intervals during gestation and at calving to quantify pregnancy loss beginning at d 23 post-AI utilizing pregnancy-specific protein B (PSPB) in novel within-cow comparisons. Cows with single ovulations assigned to the L/L treatment had greater pre-ovulatory follicle diameter compared with cows assigned to the L/H or H/L treatments. Cows with single ovulations had greater pre-ovulatory follicle diameter compared with cows with double ovulations. Low P4 in H/L, L/H, and L/L increased double ovulation rate compared with H/H. Cows with double ovulations had greater pregnancies per AI (P/AI) on d 23 post-AI compared with cows with single ovulations but had greater losses if ovulations were unilateral. Cows with low P4 during the entire period of the ovulatory follicle development also had greater P/AI on d 23 post-AI compared with cows with high P4 during both phases. However, full-term P/AI was not different between treatments. This was a result of the greater incidence of pregnancy losses between d 35 and 56 of gestation for cows with unilateral double ovulations compared with bilateral double ovulations and single ovulatory cows. Cows with single ovulation and low circulating P4 during the dominance period of follicle development had increased pregnancy losses between d 35 and 56 of gestation compared with cows with single ovulations and high P4. The PSPB measurements on d 16 and 23 post-AI were highly accurate in the prediction of pregnancy at d 28. The PSPB differed on d 23 and 28 between cows that had versus cows that did not have pregnancy losses between d 28 and 35 of gestation. In summary, circulating concentrations of P4 during ovulatory follicle development affected numbers of follicles ovulated and timing of subsequent pregnancy losses.