摘要
为了准确预测和评估地质工程中存在的风险,设计一种基于多源数据融合分析的地质风险巡检系统。该系统能够综合利用探地雷达、红外传感器等设备采集到的数据信息,通过智能化检测算法快速、准确地识别出地质风险。在预测模型的设计中,提出了一种基于SCINet和LSTM的地质风险预测算法。该算法通过前馈神经网络(FFN)来增强多源数据的非线性表示;并且能够利用SCINet在提取多尺度特征方面的优势,引入LSTM使得模型具有捕获长期依赖性的能力,从而提高整体算法的预测精度。实验结果表明,所提出的预测算法能够有效提高对地质风险的识别性能。与GRU和Bi-LSTM等多种识别算法进行的对比实验验证了该算法的优越性,其准确率相比主流的Bi-LSTM提高了15.95%。
In order to accurately predict and evaluate the risks in geological engineering,a geological risk inspection system based on multi-source data fusion analysis is designed.This system can comprehensively utilize devices such as ground penetrating radar and infrared sensors to collect data information,and quickly and accurately identify geological risks by means of intelligent detection algorithms.In the design of the prediction model,a geological risk prediction algorithm based on SCINet(sample convolution and interaction networks) and LSTM(long short-term memory) is proposed,which can enhance the nonlinear representation of multi-source data by FFN(feed-forward network) and utilize the advantages of SCINet in extracting multi-scale features.Meanwhile,the introduction of LSTM enables the model to capture long-term dependencies,thereby improving the overall prediction accuracy of the algorithm.The experimental results show that the proposed prediction algorithm model can effectively improve the ability to identify geological risks.The comparative experiments with various recognition algorithms such as GRU and BiLSTM can verify the superiority of this algorithm,with an accuracy improvement of 15.95% compared with mainstream BiLSTM.
作者
张玮
刘岢
吴志学
董洁
郭昊
姜鹏浩
ZHANG Wei;LIU Ke;WU Zhixue;DONG Jie;GUO Hao;JIANG Penghao(School of Civil Engineering,Shandong Jianzhu University,Jinan 250101,China;Beijing Daxing Urban Construction Comprehensive Development Group Co.,Ltd.,Beijing 102600,China;State Grid Beijing Shunyi Power Supply Company,Beijing 101300,China;Beijing Jianyetong Engineering Testing Technology Co.,Ltd.,Beijing 102627,China)
出处
《现代电子技术》
北大核心
2024年第22期160-164,共5页
Modern Electronics Technique
关键词
地质风险预测
风险巡检系统
多源数据融合
前馈神经网络
SCINet
长短期记忆网络
时间序列特征
geological risk prediction
risk inspection system
multi-source data fusion
feed-forward neural network
SCINet
long short-term memory network
time series characteristic
