设计了基于CatBoost算法的多传感器信息融合可燃气体燃爆状态监测系统,其中甲烷采集使用响应更快的催化燃烧传感器和测量范围更大的红外甲烷传感器进行配合采集,使其能在高浓度气体环境中快速检测甲烷浓度。首先将获取的传感器数据通过梯度提升树算法筛选出最能表征燃爆状态的特征;然后使用SMOTEENN算法对特征集进行类不平衡处理;最后通过CatBoost算法对数据进行分类训练得到预测模型。该模型对可燃气体燃爆状态的分类具有较高的准确率。
Abstract
In the paper,a multi-sensor information fusion monitoring system based on CatBoost algorithm is designed for combustible gas ignition and explosion state.In this system,a catalytic combustion sensor with faster response time and an infrared methane sensor with a larger measurement range are used for methane acquisition,making it fast and able to detect methane concentration in a high-concentration gas environment.The sensor data obtained are selected by gradient lifting tree algorithm to select the characteristics that can best characterize the state of ignition and detonation.Then SMOTEENN algorithm is used to deal with class unbalance of feature set.Finally,CatBoost algorithm is used to classify and train the data to get the prediction model.The model has high accuracy for classification of combustible gas explosion state.
关键词
多传感器信息融合 /
SMOTEENN /
CatBoost
Key words
multi-sensor information fusion /
SMOTEENN /
CatBoost
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 魏武,许期聪,邓虎,等.气体钻井技术在七北101井的应用与研究[C]//中国石油学会石油工程专业委员会钻井工作部2005年学术研讨会暨第五届石油钻井院所长会议,2005.
[2] 吴宇尘.基于长短时记忆神经网络硬件加速的气体燃爆状态监测应用[D].成都:西南交通大学,2021.
[3] 刘文鹏,陈向东,吴宇尘.基于脉冲供电的催化燃烧式气体传感系统[J].信息技术,2020,44(8):1-6,11.
[4] Prokhorenkova L,Gusev G,Vorobev A,et al.CatBoost:unbiased boosting with categorical features[J].Advances in neural information processing systems,2018,31(10).
[5] 何鑫.采用RBF神经网络的可燃气体报警器开发[D].哈尔滨:哈尔滨理工大学,2013.
[6] 郭朋飞.CAN总线结合ZigBee的空气钻井燃爆模拟监测系统的设计与实现[D].成都:西南交通大学,2017.
[7] Friedman J H.Greedy function approximation:a gradient boosting machine[J].Annals of statistics,2001,29(5):1189-1232.
[8] Guha S,Mishra N,Roy G,et al.Robust random cut forest based anomaly detection on streams[C]//International conference on machine learning.PMLR,2016:2712-2721.
[9] Batista G E,Prati R C,Monard M C.A study of the behavior of several methods for balancing machine learning training data[J].ACM SIGKDD explorations newsletter,2004,6(1):20-29.
[10] 王洪伟,孟园.在线评论质量有用特征识别:基于GBDT特征贡献度方法[J].中文信息学报,2017,31(3):109-117.
[11] 胡翔.基于SMOTE随机森林算法的信息技术企业财务预警[D].上海:上海师范大学,2022.
基金
*国家自然科学基金重点资助项目(61731016);中央高校基本科研费(2682022ZTPY001)。
PDF(1253 KB)
