基于FP和MZ法的光纤光栅传感解调技术研究

提出将FP法与非平衡MZ法相融合的解调方法,采用气体吸收(GC)加标准FP腔波长标定技术,对解调系统光电检测电路噪声进行严格控制,实现高精度解调要求,并对光路功率进行计算和合理分配。实验表明,该方法不仅能够实现多点温度信号及缓变应力信号的检测,也能够实现高频动态参数的实时检测,兼有FP法与非平衡MZ法两种解调技术的优点。     

Iron bioavailability from ferric pyrophosphate in rats fed with fructan-containing yacon (Smallanthus sonchifolius) flour

The effects of inulin-type fructans (ITF)-containing yacon flour (YF) on Fe bioavailability from ferric pyrophosphate (FP) were evaluated in Fe-deficient rats using the Hb repletion efficiency (HRE) assay. Weanling male Wistar rats were fed a low-Fe diet (12 mg/kg) for 15 days followed by 2 weeks of Fe repletion with diets providing 35 mg Fe/kg as either ferrous sulphate (FS) or FP, supplemented with 7.5% ITF as either YF or Raftilose (RAF), a purified ITF. ITF increased caecal fermentation, whereas YF was more butyrogenic than RAF. ITF improved HRE in FP-fed rats, and those fed YF had a higher relative biological value compared with those fed FP and RAF. Liver Fe was increased by ITF, but only YF led to values similar to those in the FS group. It is observed that ITF increased caecal fermentation and Fe bioavailability. These effects were more pronounced when YF was the ITF source.     

软件评价的工程方法

本文介绍ＬＯＣ及ＦＰ方法的思想，并结合软件评价方法的通用性和准确性要求，提出了对ＦＰ方法的改进方案，根据模糊数学理论确定了数学模型，提出了进一步工作的有关想法。     

对图形识别具有平移、旋转、伸缩不变性的神经网络

本文讨论如何构造在平面刚体运动和伸缩变换下，具有不变性模式识别能力的神经网络，即建立一个前向多层的神经网络，它具有如下性质，若网络能识别某一样本，同同时也能识别该样本经刚体运动及伸缩之后的任一状态，按文中给出了的构造方法，得到的网络具有如下性质：网络对运动和伸缩变换具有不变性，拓扑结构简单，网络的元件个数为Ｏ（ｎ）连接复杂性低，每个元件的输入端只需Ｏ（ｎ）（其中ｎ是输入向量的维数）学习和识别的速度     

Quality evaluation of transmission devices using the GA with FP

We propose a new method of evaluating the quality of a transmission device according to the acoustic data by using the genetic algorithm (GA) with floating points (FP). We extracted the spectrum average and the frequency variation from the acoustic data of operating transmission devices. Then we used the GA with FP to select the significant frequencies and made a quality evaluation rule to classify the good and defective products. For the frequency variation, we used the learning vector quantization (LVQ) to evaluate the quality of transmission devices. Using either the spectrum average or frequency variation alone might lead to incorrect results. However, we can ensure the final quality of transmission devices by using both frequency variation and spectrum average. The experimental results showed that the proposed method can perform the quality evaluation of transmission devices successfully and has better performance than the results obtained by the conventional GA with binary representation.This work was presented, in part, at the 8th International Symposium on Artificial Life and Robotics, Oita, Japan, January 24–26, 2003     

消爆腔对毫秒导爆管雷管延期时间的影响

采用红外测温仪及高速摄影,分析毫秒延期系列二、三、四、五段导爆管雷管的消爆腔长度对延期时间的影响.研究结果表明:8号二段至五段毫秒延期导爆管雷管在4～ 32mm的有限消爆腔内,三芯延期体均被点燃;二段至五段导爆管雷管都存在最佳消爆距离,此时点火面温度最高、延期偏差最小、延期精度最佳.     

台面结构硅基法珀型光纤MEMS压力传感器的研究

提出了一种新的光纤压力传感器的设计,该传感器敏感膜采用了台面结构而非传统的平面结构.用法布里-珀罗(Fabry-Perot)干涉理论阐述了传感器的工作原理,提出了敏感膜的力学模型.基于Fabry-Perot干涉理论推导出光纤MEMS压力传感器中台面敏感膜受到的压力与干涉光强的关系表达式,通过ANSYS有限元软件分析了台面膜型的力学性能,结果表明台面敏感膜在平行度上优于平面膜.通过数值模拟分析了传感器的关键参数对其性能的影响,为光纤MEMS压力传感器的加工和制作提供了理论依据.     

基于直觉模糊——神经网络的色情图像识别算法

网络中色情图像的传播严重影响了网络信息内容的安全性。为提高色情图像识别的准确度,提出了一种直觉模糊理论和FP（Forward Propagation）神经网络相结合的色情图像识别算法。算法以颜色直方图为底层特征,根据色情图像颜色分布情况,由模糊理论和直觉模糊理论共同构建图像特征矩阵;采用FP网络实现色情图像特征训练过程,其中特征矩阵的权重通过反向传播神经网络训练得到,以加权距离建立球形邻域半径;最后通过球形邻域覆盖情况识别色情图像。实验结果表明,该算法能够在不影响识别速率的前提下,有效的提高识别准确度。     

基于JMS的一种消息过滤改进算法

本文分析了JMS发布／订阅模式下传统的消息过滤机制,针对其不能快速处理大量相同或相似消息的缺陷,提出了一种改进的消息过滤算法。通过对所有订阅条件进行预处理来优化匹配结果,在此过程中并对相同或相似订阅条件属性进行处理,从而提高消息匹配效率。     

FP传感器双峰级次腔长解调算法

在应用双峰法解调光纤法布里-珀罗(Fabry-Perot,FP)传感器时,由于难以准确判断干涉级次,而导致解调腔长误差明显过大。为了应对这种问题,提出一种双峰级次腔长解调算法。该算法通过跟踪传感器反射光谱中相距最远的两个波峰准确判断出所选波峰的干涉级次,然后根据单峰法精确解算出绝对腔长。由于有效消除了双峰法的干涉级次模糊,腔长解算精度显著提高。对腔长30—100 μm的FP传感器进行了仿真分析和验证实验。仿真解调误差小于0.7 nm,实验误差小于1.4 nm,解调误差远小于双峰法。