NUCLEAR TECHNIQUES, Volume. 47, Issue 12, 120502(2024)
SSA-XGBoost model based high-precision density prediction method for well logging
Figures & Tables(12)
Fig. 3. Density prediction result comparison of SSA-XGBoost parameter optimization with XGBoost (color online)
Fig. 4. Model error comparison of SSA-XGBoost hyperparameter optimization with XGBoost (color online)
Fig. 5. Comparison of simulation density prediction results of different models (color online)
Fig. 6. Taylor plot of model performance comparison (color online)
Table 1. Parameters of the instrument model
View tableView in ArticleTable 1. Parameters of the instrument model
参数Parameters 规格Instructions 伽马源Gamma ray source 137Cs，能量Energy: 0.662 MeV 探测器类型Detector types 近探头Near detector: NaI，远探头Far detector: NaI 源距Sonde-to-source spacing 近源距Near space: 19 cm，远源距Far space: 39 cm 仪器设计Instrument design 直径Diameter: 6 cm，长度Length: 55 cm
Table 2. Statistical error of count rates for different detector
View tableView in ArticleTable 2. Statistical error of count rates for different detector
密度
Density / g∙cm-3
短探头计数率Near detector count rates 短探头统计误差Near detector statistical error / % 长探头计数率Far detector count rates 长探头统计误差Far detector statistical error / % 2.00 3.063×10-8 0.62 3.839×10-10 1.16 2.30 2.801×10-8 0.68 2.155×10-10 1.21 2.65 2.497×10-8 0.75 1.120×10-10 1.27 2.87 2.352×10-8 0.80 7.874×10-11 1.33 3.01 2.256×10-8 0.82 5.274×10-11 1.47
Table 3. Statistical analysis of processed data
View tableView in ArticleTable 3. Statistical analysis of processed data
编号
Number
孔隙度范围
Porosity / P.U.
孔隙流体
Pore fluid
密度范围
Density range / g∙cm-3
K1 0~40 H2O 2.36~2.71 K2 0~40 H2O 2.00~2.68 K3 0~40 H2O 2.11~2.94 K4 0~40 H2O 2.24~2.83 K5 0~40 H2O 2.06~2.88 K6 0~40 H2O 2.15~2.87 K7 0~40 H2O 2.21~2.98 K8 0~40 H2O 2.14~3.01 K9 0~40 H2O 2.25~2.95 K10 0~40 H2O 2.41~2.96
Table 4. Results of parameter optimization
View tableView in ArticleTable 4. Results of parameter optimization
模型
Model
最大迭代次数
Maximum iterations
学习率
Learning rate
树的深度
Tree depth
SSA-XGBoost 90 0.918 2 6 XGBoost 81 0.954 6 11
Table 5. Evaluation metrics for different models
View tableView in ArticleTable 5. Evaluation metrics for different models
模型Model SSA-XGBoost Spine-ribs Plot XGBoost SSA-LSTM SSA-RFR SSA-SVR MAE 0.017 4 0.028 4 0.031 4 0.033 0 0.072 1 0.045 9 RMSE 0.020 6 0.034 6 0.036 5 0.038 2 0.083 4 0.053 0 SD 0.061 3 0.068 2 0.074 4 0.078 9 0.097 5 0.083 5 R2 0.948 5 0.818 2 0.644 8 0.749 9 -0.191 4 0.518 7
Tools
Get Citation
Copy Citation Text
Rui LI, Wensheng WU. SSA-XGBoost model based high-precision density prediction method for well logging[J]. NUCLEAR TECHNIQUES, 2024, 47(12): 120502
Paper Information
Category: NUCLEAR PHYSICS, INTERDISCIPLINARY RESEARCH
Received: Apr. 25, 2024
Accepted: --
Published Online: Jan. 15, 2025
The Author Email: WU Wensheng (WUWensheng)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20