	US 12,475,392 B2
	Auxiliary decision-making method for urban subway waterlogging risk disposal based on Bayesian network
Chao Zhang, Beijing (CN); Shuai Huang, Beijing (CN); Tingxin Qin, Beijing (CN); Xiangcheng Meng, Beijing (CN); Wan Wang, Beijing (CN); Fengjiao Xu, Beijing (CN); Qian Zhou, Beijing (CN); Ying Qu, Beijing (CN); and Dehua Guo, Beijing (CN)
Assigned to China National Institute of Standardization, Beijing (CN)
Filed by China National Institute of Standardization, Beijing (CN)
Filed on Mar. 27, 2025, as Appl. No. 19/093,219.
Claims priority of application No. 202410476717.7 (CN), filed on Apr. 19, 2024.
Prior Publication US 2025/0328789 A1, Oct. 23, 2025
Int. Cl. G06N 7/01 (2023.01)

CPC G06N 7/01 (2023.01)

4 Claims

1. An auxiliary decision-making method for urban subway waterlogging risk disposal based on a Bayesian network, comprising the following steps performed by one or more processors:

acquiring waterlogging data and basic data of a target city subway stored in a memory, and preprocessing the waterlogging data in the memory, wherein the waterlogging data comprises historical waterlogging data, meteorological data, and monitoring data;

in the memory, obtaining first evaluation data and second evaluation data from the preprocessed waterlogging data, wherein the first evaluation data characterizes a predicted water flow value of the target city subway, the second evaluation data characterizes an available drainage capacity of the target city subway, the predicted water flow value is a water flow rate predicted based on the monitoring data and the meteorological data, and the drainage capacity is an available water discharge volume obtained based on the basic data and the monitoring data;

in the memory, performing risk assessment on the first evaluation data and the second evaluation data according to risk degrees to obtain a risk value, comprising:

classifying subway waterlogging attacks into three stages of early warning preparation, initial waterlogging, and severe waterlogging respectively, constructing an attack state matrix according to risk factors of subway waterlogging, and calculating a risk value with risk factors as nodes:

wherein α_wis a hazard degree of a node w, custom character

_wis an attack result of the node w, c_wis an attack stage of the node w, d_wis an attack state of the node w, custom character

is a sum of hazard degrees, attack results, attack stages and attack states, and weights thereof are ψ₁, ψ₂, ψ₃and ψ₄respectively;

determining a direction of directed edge according to causal relationships between risk factors, calculating a weight of a directed edge of a node:

wherein N* is a total number of attack states, α_wi,maxis a maximum hazard degree of the node w in an attack state i,

is a risk value of the node w in the attack state i,

is a probability that a node y is in the attack state i after attack by the node w, α_wiis a hazard degree of the node w in the attack state i, z is an attack constant, and custom character

_wyis a risk expectation that the node w attacks the node y;

constructing a directed weighted network according to directional risk factors and calculating an average probability of path-compromised nodes:

wherein

is the number of other nodes compromised from the node w, custom character

is a path for compromising other nodes from the node w, custom character

_wis an average probability of compromising other nodes from the node w, and custom character

is a probability of compromising other nodes from the mode w through a custom character

path;

calculating an average probability of being compromised:

wherein

is a probability of compromising the node w through the custom character

path,

₁is the number of possible paths where the node w is compromised, and custom character

is an average probability of compromising the node i;

constructing a weighted directed graph based on the directed weighted network and calculating an importance index of node:

wherein ε_wis an importance index of the node w, ϑ₁is a first balance factor, ϑ₂is a second balance factor, e is a natural constant, S_ot(w) is a weight of the node w in sending connections, S_in(w) is a weight of the node w in receiving connections, and υ is a compromise coefficient;

calculating a weighted risk expectation of node:

wherein

_wis a weighted risk expectation of the node w

is a probability that the node w is in the attack state i, λ is a weighting factor, and a risk value of the node is calculated as follows:

wherein φ₁is a first balance coefficient, φ₂is a second balance coefficient, and custom character

_wis a risk value of the node w;

in the memory, adjusting the risk value according to the monitoring data to determine a risk degree, comprising:

calculating risk significance of monitoring data:

wherein

_jis risk significance of j^thmonitoring data, k is a total number of the j^thmonitoring data, custom character

_jis a contribution degree of the j^thmonitoring data, custom character

_jis a risk tolerance of the j^thmonitoring data, custom character

_i,bfis a risk damage degree caused by a previous rainfall scenario indicated in the j^thmonitoring data, and χ_jis an intensity factor of the j^thmonitoring data;

integrating monitoring data with risk significance of greater than 0.362 into risk data, and calculating a risk degree according to the risk data:

wherein U_cis a risk degree of a c^thrainfall scenario, f_cjis a value of j^thrisk data of the c^thrainfall scenario, f_j* is a standard value of the j^thrisk data, custom character

_cjis risk significance of the j^thrisk data of the c^thrainfall scenario, ζ_cis an optimization factor of the c^thrainfall scenario, k is a total number of the j^thrisk data of the c^thrainfall scenario, and custom character

_cis a risk value of the c^thrainfall scenario;

in the memory, constructing an auxiliary decision-making model for waterlogging risk disposal according to the risk degree to optimize the auxiliary decision-making for waterlogging risk disposal, comprising:

taking the risk degree as a criterion of decision-making through the auxiliary decision-making model for waterlogging risk disposal:

when the risk degree is less than 0.29, the risk level is classified as Level 1; when the risk degree is greater than 0.29 but less than 0.53, the risk level is classified as Level 2; when the risk degree is greater than 0.53 but less than 0.71, the risk level is classified as Level 3; when the risk degree is greater than 0.71 but less than 0.89, the risk level is classified as Level 4; and when the risk degree is greater than 0.89 but less than 1, the risk level is classified as a special level;

the auxiliary decision-making model for waterlogging risk disposal includes an association rule mining algorithm, a Bayesian network algorithm, a long short-term memory network algorithm, and a fuzzy evaluation algorithm;

the association rule mining algorithm selects critical data based on correlation of waterlogging data;

the Bayesian network algorithm constructs a basic Bayesian network structure according to the critical data; risk factors are defined as nodes of the Bayesian network; a probability value is assigned to the state of each node; and general risk factor nodes, decision-making nodes and utility nodes are configured to obtain a first risk degree;

the long short-term memory network algorithm predicts a second risk degree according to temporal characteristics of the critical data; and

the fuzzy evaluation algorithm fuses the first risk degree and the second risk degree to obtain a risk degree.