On the relation of moderate-short term anomaly of earth resistivity to earthquake

1139 moderate-short term anomalies of earth resistivity before 196 earthquakes with magnitude M_s=3.2-7.9 (the Ms≥4.0 event accounting for 94%) are studied in this paper, the results are concluded as following: ①There is a nonlinear function between anomaly time and magnitude of earthquake. For earthquakes Ms≤5.0 or so anomaly time linearly increases quickly with magnitude increasing; for earthquakes 5.0<M_s<6.5 the increasing rate of the time with magnitude increasing gradually become small; for earthquakes M_≥6.5 the rate is quite small.②There is a nonlinear exponential function between anomaly amplitude and magnitude. For earthquakes Ms≤5.0 or so the amplitude increases slowly with the increasing of magnitude, for earthquakes 5.0<M_s<6.5 the increasing of the amplitude is gradually accelerated with magnitude increasing; for earthquakes M_s≥6.5 the increasing is accelerated quickly with magnitude increasing. The two non-linear functions mentioned above are interpreted qualitatively, and the mechanism of this phenomenon is discussed based on the model of rheomorphic medium.

ANALYSIS ON SHORT-TERM PRECURSORY ANOMALIES AND SEQUENCE CHARACTERISTIC OF NINGLANG EARTHQUAKE 1998

From Octobet 1998 to January 1999,5 earthquakes ( M s≥5) occurred between Ninglang and Yanyuan counties (27°07′～27°12′N,100°40′～101°00′E area).They were situated in 140km southwest of the Xichang.Among them,the largest one is M s 6 2 on November 19,1998.Based on small seismic data by the seismic remote sensing station of Xichang and the seismological station of Muli,and regional observation data,passing through careful observation and scientific analyses,we had made better forecasts before the earthquakes.That results obvious social benefits.By processing data of precursory earthquakes,such as,original observation data of total geomagnetic intensity from the station of Xichang,pressure capacitance stressometer and quartz horizaontal pendulum tiltmeter from the Xiaomiao station of Xichang,we summarized the sequence characteristics of the series earthquakes.The information about short\|term anomaly of gruond strain,total geomagnetic intensity and ground tilt before the earthquake is emphatically explained.

NUMERICAL STUDY OF SHORT TERM EFFECT OF SST ANOMALIES

Using a 5-layer P- o mixed coordinates primitive equations model, a process of heavy rain is simulated that occurred over the middle-and lower-reaches of the Changjiang River on July 1- 2, 1991 and numerical experiments are done of the effects of sea surface temperature (SST) anomalies over different waters on the precipitation. The result has shown that the appearance of SST anomaly is followed in a short term (2 or 3 days) by. A change in the pattern of circulation as well as in precipitation to some extent.

Short-term and imminent geomagnetic anomalies of the Wenchuan M_S8.0 earthquake and exploration on earthquake forecast

The diurnal variation of the geomagnetic vertical component is exhibited mainly by changes of phase and amplitude before strong earthquakes. Based on data recorded by the network of geomagnetic observatories in China for many years, the anomalous features of the appearance time of the minima of diurnal variations （i.e, low-point time） of the geo- magnetic vertical components and the variation of their spatial distribution （i.e, phenomena of low-point displacement） have been studied before the Wenchuan Ms8.0 earthquake. The strong aftershocks after two months＇ quiescence of M6 aftershocks of the Ms8.0 event were forecasted based on these studies. There are good correlativities between these geomagnetic anoma- lies and occurrences of earthquakes. It has been found that most earthquakes occur near the boundary line of sudden changes of the low-point time and generally within four days before or after the 27th or 41st day counting from the day of the appearance of the anomaly. In addition, the imminent anomalies in diurnal-variation amplitudes near the epicentral areas have also been studied before the Wenchuan earthquake.

Study on medium-short term earthquake forecast in Yunnan Province by precursory events

The medium-short term forecast for a certain kinds of main earthquake events might be possible with the time-to-failure method presented by Varnes (1989), Bufe and Varnes (1993), which is to simulate an accelerative releasing model of precursory earthquake energy. By fitting the observed data with the theoretical formula, a medium-short term forecast technique for the main shock events could be established, by which the location, time and magnitude of the main shock could be determined. The data used in the paper are obtained from the earthquake catalogue recorded by Yunnan Regional Seismological Network with a time coverage of 1965~2002. The statistical analyses for the past 37 years show that the data of M2.5 earthquakes were fairly complete. In the present paper, 30 main shocks occurred in Yunnan region were simulated. For 25 of them, the forecasting time and magnitude from the simulation of precursory sequence are very close to the actual values with the precision of about 0.57 (magnitude unit). Suppose that the last event of the precursory sequence is known, then the time error for the forecasting main shock is about 0.64 year. For the other 5 main shocks, the simulation cannot be made due to the insufficient precursory events for the full determination of energy accelerating curve or disturbance to the energy-release curve. The results in the paper indicate that there is no obviously linear relation in the optimal searching radius for the main shock and the precursory events because Yunnan is an active region with damage earthquakes and moderate and small earthquakes. However, there is a strong correlation between the main shock moment and the coefficient k/m. The optimal fitting range for the forecasting time and magnitude can be further reduced using the relation between the main shock moment lgM0 and the coefficient lgk/m and the value range of the restricting index m, by which the forecast precision of the simulated main shock can be improved. The time-to-failure method is used to fit 30 main shocks in the paper and more than 80% of them have acquired better results, indicating that the method is prospective for its ability to forecast the known main shock sequence. Therefore, the prospect is cheerful to make medium-short term forecast for the forthcoming main shocks by the precursory events.

POSSIBLE CONTRIBUTION OF A TROPICAL CYCLONE TO SHORT-TERM CLIMATE ANOMALIES IN EAST ASIA VIA SNOW COVER ON THE TIBETAN PLATEAU

Snow cover on the Tibetan Plateau(TP) has been shown to be essential for the East Asian summer monsoon.In this paper, we demonstrate that tropical cyclone(TC) 04B(1999) in the northern Indian Ocean, which made landfall during the autumn of 1999, may have contributed to climate anomalies over East Asia during the following spring and summer by increasing snow cover on the TP. Observations indicate that snow cover on the TP increased markedly after TC 04B(1999) made landfall in October of 1999. Sensitivity experiments, in which the TC was removed from a numerical model simulation of the initial field, verified that TC 04B(1999) affected the distribution as well as increased the amount of snow on the TP. In addition, the short-term numerical modeling of the climate over the region showed that the positive snow cover anomaly induced negative surface temperature, negative sensible heat flux, positive latent heat flux, and positive soil temperature anomalies over the central and southern TP during the following spring and summer. These climate anomalies over the TP were associated with positive(negative) summer precipitation anomalies over the Yangtze River valley(along the southeastern coast of China).

基于ResNet-LSTM的航空发动机性能异常检测方法

为了实现数据驱动的航空发动机性能异常的智能检测,提出了一种基于残差网络(ResNet)-长短期记忆网络(LSTM)的发动机性能异常检测方法。采用发动机性能数据图像化方法,在数据降维的同时,完备保留数据的关联特征和时序特征;以残差单元构建发动机性能异常检测模型,在加深网络结构的同时,消除深层网络梯度消失问题,提高发动机性能图像空间关联特征的提取能力。同时,引入LSTM,提出基于ResNet-LSTM的发动机性能异常检测模型,通过ResNet与LSTM的融合,强化异常检测模型对时序特征的提取,提升发动机性能异常检测的准确率;通过发动机运行数据进行验证。结果表明:在训练集上,该方法的异常检测准确率为94.95%,比基于ResNet18、ResNet34、ResNet50异常检测模型的分别提高10.87%、8.00%、3.23%;在测试集上,该方法的异常检测准确率为92.15%,比基于ResNet18、ResNet34、ResNet50异常检测模型的分别提高11.81%、9.45%、3.78%。

基于LSTM的多维度进程异常行为检测

恶意进程是威胁系统安全的一个重要因素,通过检测系统调用序列能有效发现恶意进程。不同的进程类型产生的调用序列会出现长度等方面的差异,因此单一化的检测方法无法在多种进程类型的检测上保持较高的准确率。针对这个问题,提出了基于长短时记忆(Long Short-Term Memory,LSTM)神经网络的多维度进程异常行为检测方法,在时间维度的基础上通过N-gram算法增加数据的信息维度,在多维数据形式中选取特征表现更好的维度进行异常判决。在UNM和ADFA-LD数据集上的实验结果表明,多维度的方法可以丰富调用序列的特征表现,以此减小不同进程类型的特征差异,且在多种进程类型下均有较好检测效果,使检测泛化能力得到提升。该方法在ADFA-LD数据集上与常用机器学习模型相比,准确率比支持向量机(Support Vector Machine,SVM)高11个百分点且误报率降低了70%,且相对于其他模型,检测效果均有所提升。

基于LSTM-WGAN的时间序列数据异常检测

在时序数据中发现隐藏的异常行为或事件,可以保障生产安全,具有重要意义。目前的异常检测模型存在训练不稳定、容易产生梯度消失的问题,影响异常检测效果,针对该问题,提出一种LSTM-WGAN模型,WGAN负责捕获变量之间的潜在关联,进一步提升了LSTM的检测能力。同时,以Wasserstein距离代替交叉熵损失训练判别器和生成器,结合重构损失以及判别损失实现异常检测。在NAB公开数据集上的实验结果表明LSTM-WGAN相较于基准模型在准确率、召回率以及F1得分上都有较大幅度的提升。

基于动态规整与改进变分自编码器的异常电池在线检测方法

针对电池生产成组过程中,传统异常检测方法对混入的容量及压差异常电池检测精度低及生产结束后离线异常检测方法效率低等问题,该文提出一种集合长短期记忆变分自编码器与动态时间规整评价的锂电池异常在线检测方法(VAE-LSTM-DTW),实现了异常电池的在线检测,避免了离线异常检测所造成的时间和能源的浪费。该方法首先将长短期记忆网络(LSTM)引入变分自编码器(VAE)模型,训练电池时序数据重构模型;其次,在电池异常检测的度量标准中引入动态时间规整值(DTW),并基于贝叶斯寻优获得最优检测阈值,对每个单体电池重构数据的动态规整值进行异常辨别。实验结果表明,相较该领域传统异常检测方法,VAE-LSTM-DTW模型性能优越,查准率和F1值都得到了较大的提升,具有较高的有效性和实用性。

LSTM-GAN:融合GAN和Bi-LSTM 的无监督时间序列异常检测

多元时间序列数据的异常检测旨在发现对应时序特征中不符合一般规律的特异性模式,识别特定时间步长中的异常状态.针对多元时序数据时间依赖性建模难以及数据维度不断增加导致难以有效进行异常检测等问题,本文以自编码器为基础,融合生成对抗网络(Generative Adversarial Network,GAN)和双向长短期记忆神经网络(Bi-directional Long Short-Term Memory,Bi-LSTM),提出了一种无监督异常检测模型LSTM-GAN,该模型在每一轮训练中,以迭代的方式重构正常数据,通过GAN来放大异常,Bi-LSTM来捕获时间特性,训练完成后的模型用于时序数据的异常检测.本文在4个公开数据集上和几种先进同类方法进行了对比实验,实验结果表明,LSTM-GAN的检测性能提升了4.4%~16.6%,在IT数据集SMD中的模型检测F1分数达到0.9672,实现了高效的时序数据异常检测.

基于Transformer_LSTM编解码器模型的船舶轨迹异常检测方法

[目的]为提升船舶轨迹异常检测的精度和效率,解决传统异常检测方法存在的特征表征能力有限、补偿精度不足、容易出现梯度消失、过拟合等问题,提出一种基于Transformer_LSTM编解码器模型的无监督船舶轨迹异常检测方法。[方法]该方法基于编码器解码器架构,由Transformer_LSTM模块替代传统神经网络实现轨迹特征提取和轨迹重构;将Transformer嵌入LSTM的递归机制,结合循环单元和注意力机制,利用自注意力和交叉注意力实现对循环单元状态向量的计算,实现对长序列模型的有效构建;通过最小化重构输出和原始输入之间的差异,使模型学习一般轨迹的特征和运动模式,将重构误差大于异常阈值的轨迹判定为异常轨迹。[结果]采用2021年1月的船舶AIS数据进行实验,结果表明,模型在准确率、精确率以及召回率上相较于LOF,DBSCAN,VAE,LSTM等经典模型有着明显提升;F1分数相较于VAE_LSTM模型提升约8.11%。[结论]该方法的异常检测性能在各项指标上显著优于传统算法,可有效、可靠地运用于海上船舶轨迹异常检测。

基于VAE-LSTM模型的无人机飞行数据异常检测

无人机飞行数据是反映其自身飞行安全的重要状态参数,通过对飞行数据进行异常检测,是提高无人机整体飞行安全性的关键举措。尽管基于数据驱动方法不需专家先验知识和精确的物理模型,但缺乏参数选择且检测网络结构模型单一,使得检测模型由于参数过多导致过拟合以及无法有效捕捉数据异常模式的问题。文中结合变分自编码器和长短期记忆网络的优势,提出了一种基于VAE-LSTM的无人机飞行数据异常检测模型方法。首先,引入肯德尔相关性分析方法用于选择相关依赖的飞行数据参数集;其次,将具有相关性的参数集对所设计的VAE-LSTM深度混合模型进行训练,学习不同数据特征之间的关系映射;最后,以无监督异常检测方式在真实多维无人机飞行数据进行验证。实验结果表明,VAE-LSTM的精密度、检测率、准确率、F1分数及误检率的各项平均性能指标分别达到95.24%、98.71%、98.8%、96.82%、1.31%,相比于KNN、OC-SVM、VAE、LSTM模型,整体上展现出较好异常检测性能。

基于LSTM预测模型的应用性能异常检测

目前高性能计算系统规模和复杂性不断增加,应用软件作业性能异常的原因变得更加复杂多样,传统的针对基于监控数据进行人工分析的方法存在效率低下和过分依赖分析人员经验的问题。提出一种基于长短期记忆网络(LSTM)的性能异常检测方法。以天气预报模式WRF为研究对象,首先从历史作业数据中学习出正常性能数据的变化情况,然后通过引入boxplot方法对LSTM模型预测值与实际观测值之间的残差进行统计分析,并将大于下四分位的数据判定为异常,从而实现应用软件作业性能异常的检测。实验结果表明,上述方法不仅可以较好地检测出性能的异常,而且能适用于多种不同类型的数据集。

基于LC-GAN的电力碳排放数据异常检测方法

针对目前电力碳排放数据存在的坏数据虚假注入问题,提出一种基于生成对抗网络的电力碳排放数据异常检测方法。首先构建面向时序数据的改进生成对抗网络,引入双层长短期记忆网络,深入挖掘时序数据的内在联系;再通过生成对抗网络改进模型的生成器和判别器的共同打分,判别电力碳排放数据异常值;最后在华东某省碳排放监测平台进行电力碳排放数据测试。结果表明,基于生成对抗网络改进模型的电力碳排放数据异常检测方法对抗训练稳定、损失函数收敛速度快,检出率为87.5%,针对电力碳排放时序异常数据检测的准确度较高。

基于多维指标关联的物联网装备异常预测方法

针对高动态物联网装备多维指标相互作用使得装备状态时变性极强,难以实现装备状态精准的评估与异常检测的问题,提出一种面向多维指标关联的装备状态异常预测方法。该方法通过计算物联网装备状态海量指标的斯皮尔曼相关系数,得到多维指标间的相关性,利用主成分分析对与目的检测指标强相关的其他指标进行特征提取,将提取结果和目的指标本身历史数据作为基于长短期记忆神经网络的装备状态感知模型的输入,进而对目的指标未来状态趋势进行精准预测;在此基础上,利用无监督的DBSCAN算法对装备状态感知模型的输出结果进行分析,定位目的指标未来可能出现的异常,实现了装备状态的评估。实验研究结果表明:该方案能够高精度预测物联网装备未来异常的发生,保护物联网装备免受潜在异常的影响,增强物联网装备的稳定性。

基于注意机制LSTM-CNN的准周期时间序列异常检测框架

为提升时间序列异常检测方法的通用性与精度,提出一种基于注意机制LSTM-CNN的准周期时间序列异常检测框架。该文通过QTS分割算法将准周期时间序列分割成多个连续的高质量准周期子序列,提升抗噪声能力;基于LSTM-CNN模型同时捕捉准周期的总体变化趋势和局部特征,精确地模拟准周期的波动模式。在4个公共数据集上的实验结果表明,提出的方法能够有效提升序列行为异常检测的效果。

基于超前探测技术的煤矿瓦斯异常预警方法

考虑到瓦斯浓度数据为非连续性序列,如果直接将原始的非连续性序列输入模型中进行瓦斯浓度预测,会导致模型对序列内部潜在的周期性、趋势和异常特征的捕捉能力较弱,影响模型的预测准确性和稳定性,为此,提出基于超前探测技术的煤矿瓦斯异常预警方法。通过瑞利波超前探测技术获取煤矿地质结构信息,充分考虑煤层、岩层、断层等地质特征,将可能产生高浓度瓦斯的区域作为重点监测区域,并结合安全距离和通风情况,均匀布置瓦斯传感器。通过瓦斯传感器采集煤矿内瓦斯浓度数据,使用离散小波变换分解获取的瓦斯浓度数据,获取其高频分量、低频分量后,输入至LSTM-注意力机制模型中,利用LSTM充分捕捉瓦斯浓度数据的时序关系,并将注意力机制融入LSTM模型中,使用tanh函数处理LSTM模型输出的特征信息,增强关键信息的提取能力,完成浓度预测,最后通过设置阈值完成异常预警。实验结果表明,所提方法的瓦斯浓度异常预测准确率最大值为97.0%,明显高于现有方法,说明其煤矿瓦斯异常预警效果更好,更适用于实际场景。

2011年腾冲M_(S)5.2地震前近场地球物理异常分析

2011年6月20日和8月9日,云南腾冲地区先后发生M_(S)5.2地震,构成腾冲M_(S)5.2双震。2次腾冲M_(S)5.2地震前,地震地球物理异常集中分布在震中100 km范围的近场区,对2次地震近场区内水位、形变(水平摆观测)、水氡、水温、流量等地球物理观测资料进行综合分析,分析发现:①大部分异常持续时间在3个月以上,部分异常始于2010年初,显著中期异常测项为腾冲滇19井水位和腾冲水平摆倾斜观测;②2次地震发震前2个月,龙陵邦腊掌水氡测项出现高值异常;③临震阶段,6月20日腾冲M_(S)5.2地震前,腾冲荷花塘、热海大滚锅出现明显宏观异常,震后异常强度和频率明显增强,对8月9日腾冲M_(S)5.2地震有明显指示意义。值得注意的是,在震中85 km范围内,2011年3月10日发生盈江M_(S)5.8地震,2次腾冲地震的中期异常成为此次盈江地震“共同的地震地球物理异常”,且震后部分异常持续,并出现新的异常,表现出异常时空演化过程与地震活动迁移相匹配的特征。

基于BFGO-CNN-LSTM的EV充电数据异常诊断研究

为应对各种扰动对电动汽车(EV)在充电过程中因故障识别精度而造成的负面影响,提出一种创新的BFGOCNN-LSTM模型,对EV充电数据中的异常部分进行精确诊断。通过该方式,模型能充分利用EV充电数据的时空特征,极大提升EV充电数据异常检测的准确性和运算效率。