基于EMD与神经网络的机械故障诊断技术

经验模式分解 (EMD)是分析非线性、非平稳信号的有力工具 ,它将信号分解为突出了原信号的不同时间尺度的局部特征信息的内在模函数 (IMF)分量。本文通过将各 IMF分量输入到 BP网络中进行训练学习和故障诊断 ,比直接输入原信号可以提高 BP网络对故障诊断的准确率 ,而且减少了训练时间。     

ARTIFICIAL NEURAL NETWORK CORRECTION OF ROUGH ERRORS OF OBSERVATIONS*

In the context of tower measured radiation datasets.following the correction principle meeting a diagnostic equation in data quality control and in terms of a technique for model construction on data and ANN (artificial neural network) retrieval for BP correction of radiation measurements with rough errors available,a BP model is presented.Evidence suggests that the developed model works well and is superior to a convenient multivariate linear regression model,indicating its wide applications.     

统计降尺度方法在青海省冬季最低温度预测中的应用

利用2003-2007年国家气象中心T213L31全球中期数值预报模式逐日输出产品与青海地区25个气象站的观测数据作为试验资料, 利用相关系数和逐步回归进行因子选择, 并以单隐层神经网络和多元回归作为降尺度方法进行对比研究, 用2003-2006年间的11月1日～次年3月1日的资料作为训练样本, 以数值预报产品和前一日观测的最低温度作为因子, 建立青海省25个气候站的冬季最低温度的24, 48, 72 h预报模型, 并且以2006年12月和2007年的1、 2月作为24, 48, 72 h逐日最低温度预报试验时段。试验表明, 对于青海地区来说, 青海北部地区的预报命中率总体好于南部高原地区; 在4种对比方案中, 以选择数值预报资料结合前一日地面观测的最低温度作为主要因子的方法相对较优, 随着预报时效的延长, 24 h历史实况的作用逐渐减弱; 对于所有台站来说, 这4种方案各有优缺点, 没有一种方案可以完全代替其他所有方案; 在实际业务运行中, 对不同的台站应采用不同的预报方案进行实际业务预报。     

Using advanced soft computing techniques for regional shoreline geoid model estimation and evaluation

This study aims at evaluating the global geoid model for a regional shoreline fitting using advanced soft computing techniques and global navigation satellite system/leveling measurements. Artificial neural networks, fuzzy logic, and least square support vector machine models are developed and used to fit the global geoid model for the north coastal Egyptian line. In addition, a novel estimation geoid model is designed and evaluated based on the latest global geoid models. The results of the three estimation models show that they can be used to correct the shoreline geoid model, in terms of root mean square error that ranges from 1.7 to 8.5?cm. Moreover, it is found that the least square vector machine model is a competitive approach with certain advantage in solving complex problems represented by missing data.     

人工神经网络在爆破块度预测中的应用研究

利用人工神经网络模型对爆破块度进行预测,实验结果表明,该方法是完全可行的。通过对实验样本数据进行归一化处理后再对人工神经网络模型进行训练和预测,其预测精度会得到大大提高。     

Landslide susceptibility mapping using geological data, a DEM from ASTER images and an Artificial Neural Network (ANN)

An efficient and accurate method of generating landslide susceptibility maps is very important to mitigate the loss of properties and lives caused by this type of geological hazard. This study focuses on the development of an accurate and efficient method of data integration, processing and generation of a landslide susceptibility map using an ANN and data from ASTER images. The method contains two major phases. The first phase is the data integration and analysis, and the second is the Artificial Neural Network training and mapping. The data integration and analysis phase involve GIS based statistical analysis relating landslide occurrence to geological and DEM (digital elevation model) derived geomorphological parameters. The parameters include slope, aspect, elevation, geology, density of geological boundaries and distance to the boundaries. This phase determines the geological and geomorphological factors that are significantly correlated with landslide occurrence. The second phase further relates the landslide susceptibility index to the important geological and geomorphological parameters identified in the first phase through ANN training. The trained ANN is then used to generate a landslide susceptibility map. Landslide data from the 2004 Niigata earthquake and a DEM derived from ASTER images were used. The area provided enough landslide data to check the efficiency and accuracy of the developed method. Based on the initial results of the experiment, the developed method is more than 90% accurate in determining the probability of landslide occurrence in a particular area.     

天津港秋冬季低能见度数值释用预报研究

本文利用近5年(2009—2013年)天津港资料,分析了该地区大气能见度的分级特征。采用7年秋、冬季NCEP(2006—2012年)和地面资料,通过相关分析给出了对港口低能见度天气有高影响的高、低空物理量因子;排除沙尘和降水天气,针对不同区间的能见度样本,利用BP神经网络方法分类训练了3个统计模型;并与WRF天气模式产品对接,采用分步筛选法,研发了天津港秋、冬季72 h时效的逐时能见度BP释用预报产品。经过3年业务运行,检验结果表明:对逐时能见度而言,BP释用预报对10 km以下低能见度比WRF模式的预报技巧显著提高,达到10.5%~35.4%;其中对0.5 km大雾的预报技巧总体相当,但当WRF预报有降水时,WRF模式预报结果略优;对0.5~1 km的大雾预报,WRF模式的预报技巧1%,BP释用预报提高到了14%~21%。日最低能见度的检验表明:对小于1 km的大雾过程,BP释用预报的TS评分平均达到75%,比WRF预报技巧提高了24%;对1~10 km的低能见度过程,比WRF的预报技巧平均提高了60%。     

Identifying future climate change and drought detection using CanESM2 in the upper Siem Reap River,Cambodia

Cambodia is one of the most vulnerable countries to climate change impacts such as floods and droughts. Study of future climate change and drought conditions in the upper Siem Reap River catchment is vital because this river plays a crucial role in maintaining the Angkor Temple Complex and livelihood of the local population since 12th century. The resolution of climate data from Global Circulation Models (GCM) is too coarse to employ effectively at the watershed scale, and therefore downscaling of the dataset is required. Artificial neural network (ANN) and Statistical Downscaling Model (SDSM) models were applied in this study to downscale precipitation and temperatures from three Representative Concentration Pathways (RCP 2.6, RCP 4.5 and RCP 8.5 scenarios) from Global Climate Model data of the Canadian Earth System Model (CanESM2) on a daily and monthly basis. The Standardized Precipitation Index (SPI) and Standardized Precipitation Evapotranspiration Index (SPEI) were adopted to develop criteria for dry and wet conditions in the catchment. Trend detection of climate parameters and drought indices were assessed using the Mann-Kendall test. It was observed that the ANN and SDSM models performed well in downscaling monthly precipitation and temperature, as well as daily temperature, but not daily precipitation. Every scenario indicated that there would be significant warming and decreasing precipitation which contribute to mild drought. The results of this study provide valuable information for decision makers since climate change may potentially impact future water supply of the Angkor Temple Complex (a World Heritage Site).     

ARTIFICIAL NEURAL NETWORK CORRECTION OF ROUGH ERRORS OF OBSERVATIONS

In the context of tower measured radiation datasets.following the correction principle meetinga diagnostic equation in data quality control and in terms of a technique for model construction ondata and ANN(artificial neural network)retrieval for BP correction of radiation measurementswith rough errors available,a BP model is presented.Evidence suggests that the developed modelworks well and is superior to a convenient multivariate linear regression model,indicating its wideapplications.     

Evaluation of current statistical approaches for predictive geomorphological mapping

Predictive models are increasingly used in geomorphology, but systematic evaluations of novel statistical techniques are still limited. The aim of this study was to compare the accuracy of generalized linear models (GLM), generalized additive models (GAM), classification tree analysis (CTA), neural networks (ANN) and multiple adaptive regression splines (MARS) in predictive geomorphological modelling. Five different distribution models both for non-sorted and sorted patterned ground were constructed on the basis of four terrain parameters and four soil variables. To evaluate the models, the original data set of 9997 squares of 1 ha in size was randomly divided into model training (70%, n=6998) and model evaluation sets (30%, n=2999).In general, active sorted patterned ground is clearly defined in upper fell areas with high slope angle and till soils. Active non-sorted patterned ground is more common in valleys with higher soil moisture and fine-scale concave topography. The predictive performance of each model was evaluated using the area under the receiver operating characteristic curve (AUC) and the Kappa value. The relatively high discrimination capacity of all models, AUC=0.85–0.88 and Kappa=0.49–0.56, implies that the model's predictions provide an acceptable index of sorted and non-sorted patterned ground occurrence. The best performance for model calibration data for both data sets was achieved by the CTA. However, when the predictive mapping ability was explored through the evaluation data set, the model accuracies of CTA decreased clearly compared to the other modelling techniques. For model evaluation data MARS performed marginally best.Our results show that the digital elevation model and soil data can be used to predict relatively robustly the activity of patterned ground in fine scale in a subarctic landscape. This indicates that predictive geomorphological modelling has the advantage of providing relevant and useful information on earth surface processes over extensive areas, such data being unavailable through more conventional survey methods.