基于LSTM-AE-OCSVM的带式输送机火灾监测隐患识别技术  

Hidden Danger Identification Technology of Belt Conveyor Fire Monitoring Based on LSTM-AE-OCSVM

作　　者：邓军 王志强 王伟峰 张宝宝 杨博 任浩 DENG Jun;WANG Zhiqiang;WANG Weifeng;ZHANG Baobao;YANG Bo;REN Hao(School of Safety Science and Engineering,Xi'an University of Science and Technology,Xi'an 710054,China;School of Computer Science and Technology,Xi'an University of Science and Technology,Xi'an 710054,China;School of Electrical and Control Engineering,Xi'an University of Science and Technology,Xi'an 710054,China)

机构地区：[1]西安科技大学安全科学与工程学院,西安710054 [2]西安科技大学计算机科学与技术学院,西安710054 [3]西安科技大学电气与控制工程学院,西安710054

出　　处：《煤炭技术》2023年第1期225-229,共5页Coal Technology

基　　金：国家重点研发计划(2019YFE0131400);国家自然科学基金(52074213);陕西省重点研发计划(2021SF-472);榆林市科技计划(CXY-2020-036)。

摘　　要：针对传统带式输送机火灾隐患识别方法的漏报率和误报率高的问题,通过挖掘带式输送机火灾监测中多元时间序列(MTS)数据,提出了一种长短时记忆-自编码的一类支持向量机神经网络(LSTM-AE-OCSVM)火灾隐患识别算法。首先,改进自动编码器(AE)将隐藏层中的神经元替换为LSTM神经元;然后,提取带式输送机火灾无异常监测数据的时序特征并重构输入数据;其次,改进LSTM-AE将重构值与实际值的差值序列经OCSVM训练得到包含无隐患异常样本的超平面;最后,通过计算测试集与超平面距离函数值来划分隐患异常。仿真结果表明,实验中所提出的改进方法与传统的LSTM和OCSVM等隐患异常检测方法相比准确率更高,达到了90.1%。该方法在识别矿井带式输送机火灾隐患上具有重要的应用价值。Aiming at the high false alarm rate and false alarm rate of traditional tape lane fire hazard identification methods,by mining the multivariate time series(MTS)data in the tape lane fire monitoring,a long and short-term memory-self-encoding type of support vector is proposed.Machine neural network(LSTM-AE-OCSVM)fire hazard identification algorithm.First,the improved autoencoder(AE)replaces the neurons in the hidden layer with LSTM neurons.Then,extracts the time series features of the tape lane fire monitoring data and reconstructs the input data;secondly,the improved LSTM-AE will reconstruct.The difference sequence between the actual value and the actual value is trained by OCSVM to obtain a hyperplane containing no hidden abnormal samples.Finally,the hidden abnormal abnormalities are classified by calculating the distance function value between the test set and the hyperplane.The simulation results show that the improved method proposed in the experiment has a higher accuracy rate than the traditional hidden anomaly detection methods such as LSTM and OCSVM,reaching 90.1%.This method has important application value in identifying hidden fire hazards in belt lanes in mines.

关 键 词：矿井火灾 一类支持向量机 长短时记忆神经网络 自编码器 隐患识别 

分 类 号：TD75[矿业工程—矿井通风与安全] TP391.9[自动化与计算机技术—计算机应用技术]