基于神经网络规则提取的航空发动机磨损故障诊断知识获取

针对神经网络智能诊断与专家系统中知识难于理解和诊断解释能力差等问题,研究了一种新的基于功能性观点的神经网络规则提取方法,介绍了方法流程及关键算法.并用UCI(加利福尼亚大学埃尔文分校)机器学习数据对方法进行了分析和验证.最后,将方法应用于实际航空发动机磨损故障诊断中,采集了某型航空发动机实测油样光谱数据237个样本,利用神经网络规则提取方法提取了发动机磨损故障诊断知识规则,并对其进行了解释,结果表明了方法的正确有效性.      

冷轧张应力分布模型及其在辊系变形计算中的应用

以往张应力分布计算多以实验为基础,或理论模型建立过程中采用了过多假设,使张应力分布计算精度受到很大限制,针对这些情况,提出利用变分法建立单元出口张应力计算模型,并与影响函数法相结合建立了冷轧前张力横向分布计算模型及辊系弹性变形计算模型,并开发了相应的计算程序.用该程序计算了某公司单机架可逆冷轧机的单位宽轧制力、辊间压力、轧件厚度以及张应力等的横向分布并进行了分析.实例计算表明,计算结果与生产实测数据吻合较好,证明了理论模型是实用可靠的.     

基于分布式专家系统的运载火箭故障诊断技术

针对运载火箭结构庞大,各分系统纵横交错、相互耦合的特点,基于分布式专家系统思想,提出了一种分布式运载火箭故障诊断技术方案.该方案采用产生式知识表达技术,将专家知识通过由测点表、征兆表、规则表和结论表组成的4表结构来表示,具有表述清晰、易于扩展和移植的特点.推理机制采用分布式推理和分级推理相结合的方式.分布式推理将运载火箭的系统级故障与各分系统级故障分别同时进行诊断,可有效提高诊断速度.分级推理将危害级别高的故障优先进行诊断,因此运载火箭一旦有危害程度大的故障发生时,可以及时发现、及时处置,从而使故障对系统的影响降至最低.      

自主虚拟人智能驾驶行为模型的研究和实现

描述了自主虚拟人的分层驾驶行为及对虚拟世界的感知.在对自主虚拟人的驾驶行为进行研究的基础上,结合导向行为与专家经验,建立了基于产生式规则的智能行为模型,并给出了规则的具体实现.通过虚拟现实技术对该智能模型进行了验证,证明了其可行性.      

武汉地区电离层TEC和NmF2及板厚的季节变化

通过利用武汉电离层观测站(114.4°E,30.6°N)1980-1990年对E8T-Ⅱ卫星信标的法拉第旋转测量的TEC(电子浓度总含量)数据,以及由测高仪测量的1980-1990年间的f0F2(F2层临界频率)数据,研究了武汉地区TEC,NmF2(最大电子浓度)和板厚的季节变化,同时比较了IRI和武汉单站模式在预测NmF2季节性方面的有效性.武汉单站模式在预测NmF2季节性变化方面优于IRI模式.      

海南地区电离层Es特性研究

利用中国科学院空间科学与应用研究中心海南电离层观测站DPS-4电离层测高仪2002年3月至2005年2月的观测数据,对太阳活动下降期间海南地区上空发生的偶发E层(Es)的逐年变化和季节变化进行了统计分析.主要结果表明,在这3年中海南地区Es事件出现次数逐年增加;Esr,Esq出现最频繁,其次是Esr,而Esc和Esf出现次数较少,出现次数最少的是Esh;Es事件在夏季出现最为频繁,秋季和冬季次之,春季最为稀少;各季节不同类型Es事件的出现率也不同,其中Esl和Esq一般在各季节占主导地位,Esr其次,Esc,Esf和Esh出现次数较少,尤其是Esf在秋季几乎没有出现.这些结果对于进一步研究Es产生的物理机制提供了探测基础,同时对于电离层空间天气预报模型的建立也具有重要的意义.      

Estimating sea level rise around Australia using a new approach to account for low frequency climate signals

Regional sea level studies help to identify the vulnerable areas to the sea level rise. This study investigates the impact of climate modes on sea level variations and trends around Australia using altimetry data, climate indices, and sea level records from tide gauge stations. Here, we show that the sea level variations are negatively correlated with climate indices to the north and west of Australia. The spectral analyses of the climate indices and tide gauge data suggest that a low frequency signal with a period of 11 years emerges during the mid 1980s. Since the 25-year length of the satellite altimetry record is yet too short to detect low frequency signals, their effect on the estimation of regional sea level trend is unknown. Therefore, we estimate the sea level trend with consideration of this signal and using a two-step method. All signals with periods shorter than 7.5 years are first removed from sea level time series and then the trend is estimated using the parametric model that includes the 11-year signal. The skill of the parametric model in explaining the variations in sea level anomaly validates the presence of the 11-year signal detected in the spectrograms of the tide gauge data and climate indices. The average sea level trend for the study area is estimated as 3.85 ± 0.15 mm/year.     

液体火箭发动机故障的数值型关联规则挖掘

数值型关联规则的算法大多是将多值属性关联规则挖掘问题转化为布尔型关联规则挖掘问题,而连续属性的离散化是数值型关联规则的核心问题。本文基于数值型关联规则的理论,用一种数理统计的方法进行连续属性的离散化。将该方法应用于某大型液体火箭发动机稳态段的热试车数据,然后利用FP-Growth算法对其进行测试,挖掘出了故障数据,进而验证了其可行性。     

Tide variation monitoring based improved GNSS-MR by empirical mode decomposition

Global Navigation Satellite System multipath reflectometry (GNSS-MR) technology has great potential for monitoring tide level changes. GNSS-MR techniques usually extract signal-to-noise ratio (SNR) residual sequences using quadratic polynomials; however, such algorithms are affected considerably by satellite elevation angles. To improve the stability and accuracy of an SNR residual sequence, this study proposed an SNR signal decomposition method based on empirical mode decomposition (EMD). First, the SNR signal is decomposed by EMD, following which the SNR residual sequence is obtained by combining the corresponding intrinsic mode function with the frequency range of the coherent signal. Second, the Lomb–Scargle spectrum is analyzed to obtain the frequency of the SNR residual sequence. Finally, the SNR frequency is converted into the tide height. To verify the validity of the SNR residual sequence obtained by EMD, the algorithm performance was assessed using multigroup satellite elevation angle intervals with measured data from two station, SC02 in the United States and RSBY in Australia. Experimental results demonstrated that the accuracy of the improved algorithm was improved in the low-elevation range. The improved algorithm demonstrated high monitoring accuracy, and the effective number was not less than 80% of the total in SC02, which means it could effectively describe the trend of the tide with accuracy of approximately 10 cm, meanwhile, the RMS error of RSBY could be reduced by 30 cm, to the maximum extent. The EMD method effectively expands the range of available GNSS-MR elevations, avoids the loss of effective information, enhances considerably the utilization rate of GNSS data, and improves the accuracy of GNSS-MR tide level monitoring.     

Effects of sector structure of the interplanetary magnetic field on the upper mesosphere–lower thermosphere dynamics

In this paper we study the influence of the interplanetary magnetic field (IMF) polarity changes caused by the Earth passing through the IMF sector boundary on the dynamic processes taking place in neutral atmosphere within the altitude interval of the upper mesosphere–lower thermosphere (83–101 km). The analysis has revealed the influence of the IMF sector structure on dynamics of the upper mesosphere–lower thermosphere. There has been a significant seasonal variation of the wind reaction to the IMF polarity changes observed. The influence of the IMF polarity changes on neutral atmosphere dynamics within the altitude range of 83–101 km is most pronounced in the zonal component of neural wind when the IMF polarity changes from negative to positive in all the seasons except for spring and when IMF polarity changes from positive to negative – in spring only.