基于CatBoost算法的矿井风速传感器布设优化
张化进 , 宋方家 , 孙妍 , 李兵磊 , 蔡和 , 龙翼
矿冶 ›› 2026, Vol. 35 ›› Issue (1) : 122 -129.
基于CatBoost算法的矿井风速传感器布设优化
Optimization of mine wind speed sensors layout based on CatBoost algorithm
传感器可为矿井通风系统的安全运行和管理提供有效参考数据。为解决矿井通风故障诊断中的传感器布设问题,基于集成学习理论,将CatBoost 算法引入至矿井通风故障诊断中,提出基于CatBoost 算法的矿井通风故障诊断与传感器布设优化方法。以矿井通风网络分支风量和故障位置分别作为输入和目标标签,利用基于决策树的集成学习算法CatBoost 进行通风网络故障诊断和特征选择,在保证模型准确率的基础上,合理选择重要网络分支布设风速传感器,减少传感器的布设数量以降低通风系统监测成本。分别以经典T 型角联(算例1)和基于矿井实际布局的室内实验(算例2)通风网络为例,对比分析了3 种常见机器学习算法(逻辑斯蒂回归、支持向量机、人工神经网络),验证基于CatBoost 集成学习方法的可行性和优越性。两个算例的研究结果表明:随着训练分支数量的增多,通风网络故障准确率先增大后保持不变,由此可见选择重要分支布设风速传感器便可有效监控整个通风系统。对比常见机器学习模型,CatBoost 模型预测效果明显优于其它模型,可采用最少的传感器数量获得最高的故障诊断准确率。因此,基于CatBoost 算法的风速传感器布设优化方法,能够合理确定传感器布设数量及其位置,有效节约矿井通风网络监测成本,符合矿井绿色集约式发展趋势。
Sensors can provide effective reference data for the safe operation and management of mine ventilation systems. In order to solve the problem of sensor layout in mine ventilation fault diagnosis, based on the ensemble learning theory, the CatBoost algorithm is introduced into the mine ventilation fault diagnosis, and a mine ventilation fault diagnosis and sensor layout optimization method based on CatBoost algorithm is proposed. Taking the air volume of the branch of the mine ventilation network and the location of the fault as the input and target labels, respectively, the fault diagnosis and feature selection of the ventilation network are carried out by using the ensemble learning algorithm CatBoost based on the decision tree, and on the basis of ensuring the accuracy of the model, the important network branches are reasonably selected to deploy wind speed sensors, and the number of sensors is reduced to lower the monitoring cost of the ventilation system. Taking the classical angular (Case 1) and the indoor experiment (Case 2) based on the actual layout of the mine as examples, three common machine learning algorithms (Logistic Regression, Support Vector Machine, and Artificial Neural Network) are compared and analyzed to verify the feasibility and superiority of the CatBoost ensemble learning method. The results of the two examples show that with the increase of the number of training branches, the accuracy of the ventilation network fault increases at first and then remains constant, which shows that the wind speed sensor can be effectively monitored by selecting important branches. Compared with common machine learning models, the CatBoost model has a significantly better prediction performance than other models, and can use the minimum number of sensors to obtain the highest fault diagnosis accuracy. Therefore, the optimization method of wind speed sensors deployment based on CatBoost algorithm can reasonably determine the number and location of sensor deployment, effectively save the cost of mine ventilation network monitoring, and conform to the development trend of green intensive mine development.
| [1] |
王翔飞, 谢贤平, 许旭. 某铁矿通风系统监测点选址布局研究[J]. 有色金属(矿山部分), 2021, 73(1): 47-52. |
| [2] |
|
| [3] |
|
| [4] |
程纬华, 乔登攀, 张磊, |
| [5] |
|
| [6] |
张庆华, 姚亚虎, 赵吉玉. 我国矿井通风技术现状及智能化发展展望[J]. 煤炭科学技术, 2020, 48(2): 97-103. |
| [7] |
|
| [8] |
李军, 王海军, 赵旭林, |
| [9] |
|
| [10] |
孙继平, 唐亮, 李春生, |
| [11] |
|
| [12] |
|
| [13] |
|
| [14] |
刘剑, 蒋清华, 刘丽, |
| [15] |
|
| [16] |
潘竟涛, 赵丹, 李宗翔, |
| [17] |
|
| [18] |
倪景峰, 乐晓瑞, 常立峰, |
| [19] |
|
| [20] |
黄德, 刘剑, 刘永, |
| [21] |
|
| [22] |
苗丰顺, 李岩, 高岑, |
| [23] |
|
| [24] |
|
| [25] |
赵锴, 叶丹. 基于机器学习的矿床描述文本多标签分类[J]. 中国矿业, 2024, 33(10): 153-161. |
| [26] |
|
| [27] |
崔丽珍, 曹坚, 李丹阳, |
| [28] |
|
| [29] |
|
| [30] |
|
| [31] |
谭勇, 陈记, 杨忠民, |
| [32] |
|
| [33] |
HUANG D, LIU J, DENG L J. A hybrid-encoding adaptive evolutionary strategy algorithm for windage alteration fault diagnosis[J]. Process Safety and Environmental Protection, 2020, 136: 242-252. |
| [34] |
张浪, 张迎辉, 张逸斌, |
| [35] |
|
福建省自然科学基金资助项目(2022J01567)
/
| 〈 |
|
〉 |