Toward the Threshold of Radiation Hazards of U in Chinese Coal through the CART Algorithm

Abstract

The high volume of coal used for combustion usually leads to a large amount of coal combustion residues (CCRs), which contain the naturally occurring radioactive materials (NORMs) decayed from U and Th in coals. The high radioactivity of NORMs can cause potential harm to humans if the CCRs are used as building materials. The activities of CCRs not only depend on the concentrations of radionuclides but also largely depend on the variations of ash yields of coal. On the other hand, ash yields significantly vary in coal from less than 1−50%. This indicates that similar concentrations of radionuclides in coal with different ash yields generally do not result in similar activities in CCRs. Therefore, it is significant to build a threshold of U in coals with different ash yield levels. In this research, based on the data of 945 coal samples from China and the selected optimal model using the classification and regression tree algorithm, the threshold of U for the radiation hazard is determined to be 7.98 mg/kg for coals with ash yields higher than 20%, while the threshold of U for the radiation hazard is 5.28 mg/kg for coals with ash yields lower than 20%.

Collaboration Tripartite Elements Accident Reporting and Potential Hazards Management Coal Mining, Indonesia: Systematic Review

“…The physical hazard factors in coal mining not only cause work accidents but also occupational accidents. For example, For example, the content of radioactive materials in coal potential harm to not only workers but the general public 6 . Coal radiation exposure is associated with silicosis and pneumoconiosis in miners.…”

Application of random forest algorithm in the detection of foreign objects in wine

“…When predicting, the samples are judged from the root node, and are input to the next level of nodes according to the judgment rules of each node, until they reach the leaf node and output the value or category shown in the leaf node. While decision trees alone avoid overfitting by "pruning" the relatively unimportant rules, random forests have been mathematically proven to avoid overfitting by averaging a large number of decision trees [22][23][24]. The operation principle of the random forest algorithm is shown in Figure 1.…”