Designing of Commercial Bank Loans Risk Early Warning System Based on BP Neural Networks

According to the index early warning method, a commercial bank loans risk early warning system based on BP neural networks is proposed. The warning signal is mainly involved with the financial situation signal of loaning corporation. Except the structure description of the system structure the demonstration of attemptive designing is also elaborated.

Study of Enterprises Marketing Risk Early Warning System Based on BP Neural Network Model

For effectively early warning the marketing risk caused along with the varied environment, a BP neural network method was introduced on the basis of analyzing the shortcomings of the risk early warning method, and combined with the practical conditions of dairy enterprises, the index system caused by the marketing risk was also studied. The＇principal component method was used for screening the indexes, the grades and critical values of the marketing risk were determined. Through the configuration of BP network, node processing and error analysis, the early warning resuits of the marketing risk were obtained. The results indicate that BP neural network method can be effectively applied through the function approach in the marketing early warning with incomplete information and complex varied conditions.

Deep learning for P-wave arrival picking in earthquake early warning

Fast and accurate P-wave arrival picking significantly affects the performance of earthquake early warning(EEW)systems.Automated P-wave picking algorithms used in EEW have encountered problems of falsely picking up noise,missing P-waves and inaccurate P-wave arrival estimation.To address these issues,an automatic algorithm based on the convolution neural network(DPick)was developed,and trained with a moderate number of data sets of 17,717 accelerograms.Compared to the widely used approach of the short-term average/long-term average of signal characteristic function(STA/LTA),DPick is 1.6 times less likely to detect noise as a P-wave,and 76 times less likely to miss P-waves.In terms of estimating P-wave arrival time,when the detection task is completed within 1 s,DPick′s detection occurrence is 7.4 times that of STA/LTA in the 0.05 s error band,and 1.6 times when the error band is 0.10 s.This verified that the proposed method has the potential for wide applications in EEW.

A Real-time Monitoring and Early Warning System for Landslides in Southwest China

Landslides not only cause property losses,but also kill and injure large numbers of people every year in the mountainous areas. These losses and casualties may be avoided to some extent by early warning systems for landslides. In this paper, a realtime monitoring network and a computer-aided automatic early warning system(EWS) are presented with details of their design and an example of application in the Longjingwan landslide, Kaiyang County, Guizhou Province. Then, according to principle simple method of landslide prediction, the setting of alarm levels and the design of appropriate counter-measures are presented. A four-level early warning system(Zero, Outlook, Attention and Warning) has been adopted, and the velocity threshold was selected as the main warning threshold for the landslide occurrence, but expert judgment is included in the EWS to avoid false alarms. A case study shows the applicability and reliability for landslide risk management, and recommendations are presented for other similar projects.

Coal mine safety production forewarning based on improved BP neural network

Firstly, the early warning index system of coal mine safety production was given from four aspects as per- sonnel, environment, equipment and management. Then, improvement measures which are additional momentum method, adaptive learning rate, particle swarm optimization algorithm, variable weight method and asynchronous learning factor, are used to optimize BP neural network models. Further, the models are applied to a comparative study on coal mine safety warning instance. Results show that the identification precision of MPSO-BP network model is higher than GBP and PSO-BP model, and MPSO- BP model can not only effectively reduce the possibility of the network falling into a local minimum point, but also has fast convergence and high precision, which will provide the scientific basis for the forewarnin~ management of coal mine safetv production.

An Introductory Overview of Earthquake Early Warning

Earthquake early warning(EEW)is discriminated from earthquake prediction by using initial seismic waves to predict the severity of ground motion and issue the warning information to potential affected area.The warning information is useful to mitigate the disaster and decrease the losses of life and economy.We reviewed the development history of EEW worldwide and summarized the methodologies using in different systems.Some new sensors came and are coming into EEW giving more developing potential to future implementation.The success of earthquake disaster mitigation relies on the cooperation of the whole society.

CNN intelligent early warning for apple skin lesion image acquired by infrared video sensors

Video sensors and agricultural IoT(internet of things) have been widely used in the informationalized orchards.In order to realize intelligent-unattended early warning for disease-pest,this paper presents convolutional neural network(CNN) early warning for apple skin lesion image,which is real-time acquired by infrared video sensor.More specifically,as to skin lesion image,a suite of processing methods is devised to simulate the disturbance of variable orientation and light condition which occurs in orchards.It designs a method to recognize apple pathologic images based on CNN,and formulates a self-adaptive momentum rule to update CNN parameters.For example,a series of experiments are carried out on the recognition of fruit lesion image of apple trees for early warning.The results demonstrate that compared with the shallow learning algorithms and other involved,wellknown deep learning methods,the recognition accuracy of the proposal is up to 96.08%,with a fairly quick convergence,and it also presents satisfying smoothness and stableness after convergence.In addition,statistics on different benchmark datasets prove that it is fairly effective to other image patterns concerned.

Remote Monitoring and Early Warning Model of Frozen Soil in Dam Areas

In view of the cumbersome and often untimely process of manual collection and observation of frozen soil data parameters,and the damage caused to dams by frost heaving of frozen soil,a remote monitoring and an early warning model for frozen soil in dam areas was presented.The Pt100 temperature sensors and JM seam gauges were used as measurement tools in the system.The sensor layout was designed,based on the actual situation in the monitoring area.A 4G network was used for wireless transmission to monitor frozen soil data in real time.BP neural network was used to predict the parameters of frozen soil.After analysis,four factors including the average temperature of frozen soil,the type of frozen soil,the artificial upper limit of frozen soil and the building construction time were selected to establish an early warning model using fuzzy reasoning.The experimental results showed that the early warning model could reflect the influence on dam buildings of frost heaving and sinking of frozen soil,and provided technical support for predicting the hazard level.

Pattern changes and early risk warning of Spartina alterniflora invasion:a study of mangrove-dominated wetlands in northeastern Fujian,China

The exotic saltmarsh cordgrass,Spartina alterniflora(Loisel)Peterson&Saarela,is one of the important causes for the extensive destruction of mangroves in China due to its invasive nature.The species has rapidly spread wildly across coastal wetlands,challenging resource managers for control of its further spread.An investigation of S.alterniflora invasion and associated ecological risk is urgent in China's coastal wetlands.In this study,an ecological risk invasive index system was developed based on the Driving Force-Pressure-State-Impact-Response framework.Predictions were made of'warning degrees':zero warning and light,moderate,strong,and extreme warning,by developing a back propagation(BP)artificial neural network model for coastal wetlands in eastern Fujian Province.Our results suggest that S.alterniflora mainly has invaded Kandelia candel beaches and farmlands with clustered distributions.An early warning indicator system assessed the ecological risk of the invasion and showed a ladder-like distribution from high to low extending from the urban area in the central inland region with changes spread to adjacent areas.Areas of light warning and extreme warning accounted for43%and 7%,respectively,suggesting the BP neural network model is reliable prediction of the ecological risk of S.alterniflora invasion.The model predicts that distribution pattern of this invasive species will change little in the next 10 years.However,the invaded patches will become relatively more concentrated without warning predicted.We suggest that human factors such as land use activities may partially determine changes in warning degree.Our results emphasize that an early warning system for S.alterniflora invasion in China's eastern coastal wetlands is significant,and comprehensive control measures are needed,particularly for K.candel beach.

Energy-Efficient Scheduling for a Cognitive IoT-Based Early Warning System

Flash floods are deemed the most fatal and disastrous natural hazards globally due to their prompt onset that requires a short prime time for emergency response.Cognitive Internet of things(CIoT)technologies including inherent characteristics of cognitive radio(CR)are potential candidates to develop a monitoring and early warning system(MEWS)that helps in efficiently utilizing the short response time to save lives during flash floods.However,most CIoT devices are battery-limited and thus,it reduces the lifetime of the MEWS.To tackle these problems,we propose a CIoTbased MEWS to slash the fatalities of flash floods.To extend the lifetime of the MEWS by conserving the limited battery energy of CIoT sensors,we formulate a resource assignment problem for maximizing energy efficiency.To solve the problem,at first,we devise a polynomial-time heuristic energyefficient scheduler(EES-1).However,its performance can be unsatisfactory since it requires an exhaustive search to find local optimum values without consideration of the overall network energy efficiency.To enhance the energy efficiency of the proposed EES-1 scheme,we additionally formulate an optimization problem based on a maximum weight matching bipartite graph.Then,we additionally propose a Hungarian algorithm-based energy-efficient scheduler(EES-2),solvable in polynomial time.The simulation results show that the proposed EES-2 scheme achieves considerably high energy efficiency in the CIoT-based MEWS,leading to the extended lifetime of the MEWS without loss of throughput performance.

Risk monitoring and early-warning technology of coal mine production

This article was written according to the secudty information theory and the secudty cybernetics basic principle, for reducing the accident risk effectively and safeguarding the production safety in coal mine. First, each kind of risk characteristic has carried on the earnest analysis to the coal-mining production process. Then it proposed entire wrap technology system of the risk management and the risk monitoring early warning in the coal-mining production process, and developed the application software-coal mine risk monitoring and the early warning system which runs on the local area network. The coal-mining production risk monitoring and early warning technology system includes risk information gathering, risk identification and management, risk information transmission; saving and analysis, early warning prompt of accident risk, safety dynamic monitoring, and safety control countermeasure and so on. The article specifies implementation method and step of this technology system, and introduces application situations in cooperating mine enterprise, e.g. Xiezhuang coal mine. It may supply the risk management and the accident prevention work of each kind of mine reference.

Interpretability and spatial efficacy of a deep-learning-based on-site early warning framework using explainable artificial intelligence and geographically weighted random forests

Earthquakes pose significant risks globally,necessitating effective seismic risk mitigation strategies like earthquake early warning(EEW)systems.However,developing and optimizing such systems requires thoroughly understanding their internal procedures and coverage limitations.This study examines a deep-learning-based on-site EEW framework known as ROSERS(Real-time On-Site Estimation of Response Spectra)proposed by the authors,which constructs response spectra from early recorded ground motion waveforms at a target site.This study has three primary goals:(1)evaluating the effectiveness and applicability of ROSERS to subduction seismic sources;(2)providing a detailed interpretation of the trained deep neural network(DNN)and surrogate latent variables(LVs)implemented in ROSERS;and(3)analyzing the spatial efficacy of the framework to assess the coverage area of on-site EEW stations.ROSERS is retrained and tested on a dataset of around 11,000 unprocessed Japanese subduction ground motions.Goodness-of-fit testing shows that the ROSERS framework achieves good performance on this database,especially given the peculiarities of the subduction seismic environment.The trained DNN and LVs are then interpreted using game theory-based Shapley additive explanations to establish cause-effect relationships.Finally,the study explores the coverage area of ROSERS by training a novel spatial regression model that estimates the LVs using geographically weighted random forest and determining the radius of similarity.The results indicate that on-site predictions can be considered reliable within a 2–9 km radius,varying based on the magnitude and distance from the earthquake source.This information can assist end-users in strategically placing sensors,minimizing blind spots,and reducing errors from regional extrapolation.

A Cellular-Assisted Efficient Handover Algorithm for Wireless Sensor Networks

Heterogeneous network convergence and handover have become very hot in recent years. This paper proposed an efficient handover scheme in Multi-PAN Wireless Sensor Networks (WSNs). A number of edge nodes are set at the edge of each Personal Area Networks (PANs). A user equipment (UE), which has WSN and cellular network interface, acts as sensor node or mobile cluster head in WSN area. Thus, edge early warning can be acquired from edge nodes and neighbor channel information can be acquired with BS-assistance. Simulation results show that low transmission interrupted delay and low energy consumption can be achieved compared with conventional scheme in WSN.

Seismic monitoring network based on MEMS sensors

This paper provides a brief introduction to the application of the sensor monitoring network of micro-electro-mechanical systems(MEMS)to Zhejiang province.In the Wenzhou Shanxi reservoir and other areas,MEMS and traditional intensity-monitoring instruments have been deployed with complementary functions to implement hybrid networking.The low-cost MEMS network can continuously monitor areas at high risk of earthquakes at a high resolution.Moreover,it can quickly collect the parameters of earthquakes and records of the near-field acceleration of strong earthquakes.It can be also used to rapidly generate earthquake intensity reports and provide early warning of earthquakes.We used the MEMS sensors for the first time in 2016,and it has helped promote the development and application of seismic intensity instruments since then.

Application of Neural Networks in Early Warning Systems for Coronary Heart Disease

This paper presents a BP neural network-based algorithm for the iden-tification of coronary heart disease through the clinical data of cardiology for many years and the personal physiological attributes easily obtained in daily life.The goal of this paper is to judge whether it may have coronary heart disease by testing the attribute values of the tester.First,through the training of samples,the net-work model structure is designed,and a relatively good neural network model is obtained.Second,according to the model,the possibility of coronary heart disease was calculated.

基于LSTM网络的单台仪器地震烈度预测模型

烈度是地震预警系统的关键产出.如何实现快速预测目标场址的地震烈度是地震预警方法技术研究中的核心问题.本文提出了一种基于长短时记忆神经网络(Long Short-Term Memory,LSTM)的单台仪器地震烈度的预测模型(LSTM-Ⅰ).该模型以一个台站观测到地震动参数的时间序列特征为输入,实现动态预测该台站可能遭受的最大烈度.选取了日本K-NET台网记录的102次地震的5103条强震加速度记录训练了神经网络,利用89次地震的3781条数据检验了模型的泛化能力.利用准确率、漏报率以及误报率三个评价指标评价了LSTM-Ⅰ模型的性能.结果表明,当采用P波触发后3 s的序列进行预测时,模型出现漏报的概率为46.78%,出现误报的概率为1.25%;当采用P波触发后10 s的序列进行预测时,模型出现漏报的概率大幅降低到17.6%,出现误报的概率降低到1.14%.结果表明LSTM-Ⅰ模型很好把握住了时间序列中蕴含的特征.进一步基于LSTM-Ⅰ模型评估了Ⅵ度下台站所能提供的预警时间.本文模型能够提供的预警时间与P-S波到时差接近,说明LSTM-Ⅰ模型具有较高的时效性.

推动长江经济带高质量发展的水文实践与思考

长江流域经济社会高质量发展离不开水文的强有力支撑。近年来,长江流域水文站网布局与功能日趋完善,水文监测自动化水平不断提高,水文预报能力进一步提升,水文信息处理智慧化水平逐步提高,对水旱灾害防御、水资源配置调度、水生态保护修复等支撑作用更加凸显,水文现代化发展成效较为显著。当前,推动长江经济带高质量发展对水文提出新的更高要求,通过分析研判当前面临的新形势新要求,提出了下一步工作重点。

数字孪生水网建设应着力解决的几个关键问题

数字孪生水网建设涉及多学科交叉、新技术融合,有诸多重点和难点,应着力把握好坚持问题导向、夯实算据基础、优化算法核心、强化算力支撑、着力业务应用、加强网络安全等关键问题。结合数字孪生水网建设先行先试经验,围绕工程安全、供水安全、水质安全,提出了数字孪生水网建设的关键问题及解决思路:围绕建设目标确定建设内容和重点,以问题为导向开展需求分析;打造“天空地”一体化监测感知体系,完善地面监测站网、提升新型监测感知能力、强化数据治理与运用、推进数据资源共享夯实算据基础;完善模型平台,开发知识平台,优化算法核心;建设信创云平台和高性能计算集群,扩充计算资源、升级通信网络强化算力支撑;提升水网智慧调控能力,着力实现业务“四预”;构建网络安全体系,保障网络和数据安全;理清数字孪生流域、数字孪生水网、数字孪生工程的关系,提升整体效能。

基于卷积神经网络的预警震级分段估算方法

针对地震预警震级估算问题,提出一种基于卷积神经网络(convolutional neural network,CNN)的震级分段估算方法,该方法以单台站的P波初至后3 s时间的波形作为输入,输出结果为地震波形所属的震级区段(大地震,近震震级M_(L)≥5.0;小地震,M_(L)<5.0).如果波形属于大地震区段,直接发出警报;如果波形属于小地震区段,再进行具体震级的估算.对于震级区段估算,CNN模型的准确率可达98.04%.根据震级估算参数τ_(c)和P_(d)估算的小地震震级平均绝对误差(mean absolute error,MAE)分别为0.20和0.31.结果表明,预警震级分段估算方法可以准确预警大地震,减少大地震漏报率;同时使得小地震震级估算结果更为准确.

基于LSTM神经网络的现地烈度实时估算模型——以JMA烈度为例

如何快速并且准确估计目标场点烈度是地震预警中的关键问题。常用基于衰减关系的场点烈度估计和基于P波信息的现地烈度估计往往存在大震烈度低估的问题。本文提出了一种基于长短时记忆神经网络(logn short-term memery,LSTM)的现地JMA烈度持续估计模型。该模型以现地观测地震动的能量、能量增长率、地震动卓越周期和震源距作为输入,以该点的最大仪器地震烈度为预测目标。选取了日本K-NET台网记录101次地震数据作为训练集,94次地震数据作为测试集,训练了现地烈度估算LSTM神经网络模型。结果表明:在采用3 s时窗长度的序列进行预测时,高估的比例为1.51%,低估的比例为4.00%;并且,随着时窗长度的增加,高估和低估的比例也在不断降低。模型对高烈度(大于等于4.5度)样本的预测时效性随震源距的增加而增加,对大震远场高烈度区域能提供20 s以上的预警时间。