基于GF-LSTM和GAN网络的小样本集人工水体溶解氧浓度预测

采用推流方式改善人工水体溶解氧分布不均衡以防止富营养化时,需要对其分布进行预测来提高推流效率,为此构建了基于生成式对抗网络(GAN,Generative Adversarial Networks)和长短期记忆神经网络(LSTM,Long-Short Term Memory Network)的溶解氧浓度预测模型。以广西大学镜湖35 m2的一片水体区域为研究对象,采用不同电压直流水泵推流,用无人船搭载在线检测仪在一段时间内定时定点采集水体溶解氧浓度数据作为原始数据样本,并采用GAN扩充数据样本。利用遗传算法和改进的一阶滤波算法进行溶解氧的噪声数据处理,结合LSTM网络构建溶解氧浓度预测模型GF-LSTM(Genetic And Filtering Algorithm-Long Short Term Memory Network)。结果表明:相比常用的BP网络,GF-LSTM网络预测的平均误差降低了62%,均方误差降低了75%;相比传统的LSTM网络,GF-LSTM网络预测的平均误差降低了22%,均方误差降低了50%。     

煤体电荷感应信号评价冲击危险性试验研究

为完善冲击地压矿井的冲击危险性评价方法,提高冲击危险性预测的准确率,应用自主研制的煤岩电荷监测系统,选择典型冲击地压矿井的煤样,开展了煤体单轴压缩冲击危险性测试与电荷感应监测试验研究。基于冲击地压扰动响应失稳理论,将应变软化阶段产生的电荷信号变化作为预测冲击地压发生的前兆信息,得到了煤体冲击危险性指标的临界软化系数Kρ、临界应力系数Kp及其冲击危险的等级分类标准,分析了煤体电荷感应信号的电荷事件数CSJ和电荷事件的平均幅值CFZ参量与冲击危险性指标Kρ和Kp之间的量化关系。结果表明:煤样破裂应力峰后,冲击倾向性K与电荷事件数CSJ呈指数递减关系,与CFZ呈指数递增关系;随着Kρ或Kp的增大,煤样应力峰后CSJ呈幂函数关系递增,应力峰后电荷平均幅值CFZ呈一次函数关系递减,以此可预测煤体的冲击危险程度,最后得到了河南某矿煤层冲击危险性的应力峰后冲击危险等级的CSJ和CFZ判据与划分标准。煤体冲击危险性电荷感应信号评价方法为冲击地压矿井冲击危险性评价提供了一种新的方向,对现场煤层冲击危险性评价具有指导作用,但也还需要开展大量现场试验对其进行不断修正和完善。     

Evaluation on consequences prediction of fire accident in emergency processes for oil-gas storage and transportation by scenario deduction

This paper takes the safety in emergency processes as the starting point, from the perspective of scenario deduction, to study the consequences of fire accidents for oil-gas storage and transportation. Through the statistical analysis of actual accident cases, 19 frequently occurring basic scenarios in emergency processes are summarized. The scenario evolution paths of fire accidents for oil-gas storage and transportation are given by analyzing the evolution law of the accident development. Fuzzy numbers are introduced to express experts' qualitative judgment on accident scenarios. The empirical probabilities of scenario nodes are obtained by defuzzification calculation, and the state probability of each scenario node is calculated by using the dynamic Bayesian network joint probability formula. Under the comprehensive consideration about the probability statistics of actual accident cases, the critical scenario nodes on the evolution path and their final scenario probabilities are jointly determined to realize the optimization of the scenario evolution path. By constructing the correlation between the optimized scenario evolution path and the accident consequences, an accident consequence prediction model is established. The occurrence probability of accident consequences is calculated by the defuzzification method and dynamic Bayesian network. The accuracy of the consequence prediction model is verified by the July 16 Dalian's Xingang Harbor oil pipeline explosion accident. The research results provide scientific basis for helping decision makers to make the effective emergency measures that are most conducive to the rapid elimination of accidents and reducing the severity of accident consequences.     

不同噪声强度对煤矿工人作业失误率的影响研究

为探究噪声强度对煤矿工人作业失误率的影响程度,采用实验室模拟试验和现场调研收集数据的方式对噪声和作业失误率间的关系展开研究,利用模拟试验所得结果和煤矿现场调研收集结果分别构建噪声-作业失误率、噪声-"三违"数量回归预测模型,定量分析作业失误率、"三违"数量和噪声水平之间的关系,研究结果表明:煤矿"三违"数量、实验室模拟人员失误率与环境噪声均呈正相关,随噪声强度增加,失误率及"三违"数量呈上升趋势;当外界噪声强度高于75 dB后,人员失误次数会出现显著性变化;基于事故预防的角度出发,可将75 dB设定为噪声安全临界值,以降低人员作业失误率,达到事故预防的目的。     

A novel hybrid model for water quality prediction based on VMD and IGOA optimized for LSTM

● A novel VMD-IGOA-LSTM model has proposed for the prediction of water quality. ● Improved model quickly converges to the global optimal fitness and remains stable. ● The prediction accuracy of water quality parameters is significantly improved. Water quality prediction is vital for solving water pollution and protecting the water environment. In terms of the characteristics of nonlinearity, instability, and randomness of water quality parameters, a short-term water quality prediction model was proposed based on variational mode decomposition (VMD) and improved grasshopper optimization algorithm (IGOA), so as to optimize long short-term memory neural network (LSTM). First, VMD was adopted to decompose the water quality data into a series of relatively stable components, with the aim to reduce the instability of the original data and increase the predictability, then each component was input into the IGOA-LSTM model for prediction. Finally, each component was added to obtain the predicted values. In this study, the monitoring data from Dayangzhou Station and Shengmi Station of the Ganjiang River was used for training and prediction. The experimental results showed that the prediction accuracy of the VMD-IGOA-LSTM model proposed was higher than that of the integrated model of Ensemble Empirical Mode Decomposition (EEMD), the integrated model of Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Nonlinear Autoregressive Network with Exogenous Inputs (NARX), Recurrent Neural Network (RNN), as well as other models, showing better performance in short-term prediction. The current study will provide a reliable solution for water quality prediction studies in other areas.     

Modeling using dynamic variables – An approach for the design of loss prevention programs

Records from The National Safety Council [National Safety Council, 2007. Safety Intervention Evaluation: A Systematic Approach. <http://www.acgih.org/events/ControlBand/Thomas_SafetyIntervention.pdf> (accessed 10.01.07)] have shown that in 2004 alone, on-the-job injuries to workers constituted 35% of total recorded injuries in the United States. This generated an associated cost of about $142.2 billion. Unfortunately, the safety intervention programs enforced at work places to mitigate such losses are driven mainly by intuition and experience of involved safety personnel. This paper details implementing a computer program to furnish safety personnel with an empirical basis for designing loss prevention programs based on historical safety data. The computer tool is driven by a dynamic mathematical model which adapts itself to variations in data patterns and explains the correlation between historical incident rates and corresponding resources committed to interventions. This study empowers the industry with a tool that is capable of forming the core of optimizing valuable human resource allocation in safety program designs.     

Fisher判别法在煤与瓦斯突出危险程度预测中的应用

为了提高煤与瓦斯突出的预测精度,根据煤与瓦斯突出的综合作用假说,选取开采深度、瓦斯压力、瓦斯放散初速度、煤的普氏系数以及煤体破坏类型作为判别指标。利用国内典型突出矿井20个实测数据作为训练样本,建立煤与瓦斯突出危险程度预测的Fisher判别分析模型,并应用于其他待判样本的预测。结果表明:Fisher判别分析模型能够反映多因素对煤与瓦斯突出的影响,分类性能良好,误判率低,借助SPSS软件实现,具有计算简单的特点,是煤与瓦斯突出预测的一种有效方法。     

BP神经网络法在斜拉桥换索施工控制中的应用研究

运用神经网络理论,结合一座双塔三跨混凝土斜拉桥换索施工过程,对斜拉桥换索过程索力和主梁标高进行预测研究。运用平面杆系有限元理论计算出施工状态的理论索力和理论标高,再通过数值计算软件MATLAB7.1建立神经网络仿真模型得出各施工阶段的索力和标高预测值,为具体的换索工程提供参考。通过对比该斜拉桥换索过程中索力及主梁控制点标高的计算值和实测值可知,预测数据与目标数据相近,变化趋势一致,验证了BP神经网络在换索施工控制中进行参数预测的可行性。     

基于支持向量回归机的煤层瓦斯含量预测研究

为了对煤层瓦斯含量进行准确预测,应用支持向量回归机(SVR)理论建立煤层瓦斯含量预测模型,结合现场实测数据利用支持向量机(SVM)工具箱进行模型的求解及预测,并从均方根误差、希尔不等系数和平均绝对百分误差3个不同误差指标与人工神经网络预测模型进行比较分析。研究结果表明:SVR模型其预测精度及可行性高于神经网络模型,而且运算快,实时性较好,用于煤层瓦斯含量的预测较理想,具有良好的应用前景,可以为煤矿瓦斯防治提供理论依据。     

基于BP神经网络人群流量预测的实现

人群拥挤踩踏突发性的特点决定了现场的事故救援措施效果较差,事前预防是唯一有效的策略。对商业区人群流量进行预测,对于合理控制商业网点人口,预防人群类事故的发生具有重要的意义。本文介绍了基于BP神经网络的人群流量预测方法,利用Matlab建立了相关模型,并结合实际数据对模型进行了调整,分析了隐含层神经元个数、不同输入-输出结构、不同传递函数等因素对网络性能的影响。研究表明利用神经网络的非线性映射能力对人群流量进行预测时可行的。