Intelligent supervision method and system of state-owned assets based on digital twins

Abstract

The present invention provides a method and system for intelligent supervision of state-owned assets based on digital twins, which relates to the technical field of asset management. The method includes establishing a digital twin model of state-owned assets, analyzing the potential causal relationship between various variables in the evolution of asset behavior based on the digital twin model, constructing a causal graph, and constructing a reinforcement learning scheduling strategy model through the causal graph and the digital twin model to output the optimal scheduling strategy. Supervisors remotely control and manage state-owned assets through smart terminals based on the optimal scheduling strategy. The operation instructions are processed by the data processing center and fed back to the corresponding Internet of Things sensors and execution devices to realize intelligent management and control of state-owned assets. The present invention realizes real-time visual supervision and intelligent management and control of state-owned assets, improves supervision efficiency and accuracy, reduces management costs and risks, and effectively improves the operational efficiency and security of state-owned assets.

Description

Technical Field

[0001] The present invention relates to the technical field of asset management, and in particular to a method and system for intelligent supervision of state-owned assets based on digital twins. Background Art

[0002] In recent years, with the development of information technology, various regions have gradually established state-owned asset management platforms, achieving basic records of the asset life cycle through asset registration, ledger management, and transfer approval. However, existing systems generally exhibit the characteristics of "static ledger management", that is, they mainly rely on manual input and regular updates to maintain asset information, and lack real-time perception, behavior analysis, and intelligent decision-making capabilities. This makes it impossible to accurately capture and proactively judge issues such as status changes, value loss, and abuse risks of assets during dynamic operation. In addition, traditional regulatory methods are mostly carried out in a decentralized manner by institutions, lacking a unified perspective to comprehensively evaluate the overall efficiency of state-owned assets, scheduling structure, and allocation rationality.

[0003] To address these issues, existing research has attempted to introduce the concept of "digital twins," which involves constructing virtual digital models to map and synchronously update physical assets, enabling real-time visualization of asset status and monitoring of operational processes. However, these applications are primarily limited to engineering equipment or industrial manufacturing scenarios, and their implementation in the field of state-owned asset management is still in its early stages. On the one hand, state-owned asset management involves multi-dimensional and multi-departmental collaboration, and the assets themselves are complex (such as fixed assets, office space, and rental assets). Their operational behavior is often influenced by multiple factors, including policy, market, geography, and budget. Existing digital twin solutions struggle to fully model these influencing factors. On the other hand, while existing systems can visualize asset status, they still lack in-depth data understanding capabilities. This is particularly true when identifying scenarios such as abnormal asset use, hidden vacancies, duplicate configurations, and inefficient usage, and they are unable to provide reliable risk warnings and resolution recommendations.

[0004] More critically, existing management systems are unable to analyze the causes of asset behavior, nor can they conduct simulations and optimize strategies based on historical experience. For example, when an asset remains unused for an extended period, the system cannot determine whether its inactivity is due to obsolete functionality, management oversight, or delayed allocation processes. Similarly, when multiple units frequently swap similar assets, the system lacks a mechanism to assess whether there are irrational scheduling or resource misallocations. Existing systems, in terms of data analysis and decision support, are characterized by "information display" rather than "intelligent supervision," making it difficult to achieve the leap from "visibility" to "effective management." Summary of the Invention

[0005] In response to the shortcomings of current state-owned asset management systems in terms of data perception, behavioral analysis capabilities, intelligent scheduling, and interpretability of supervisory decisions, this paper proposes a digital twin intelligent supervision method and system for state-owned asset management scenarios. This method achieves in-depth modeling of assets in their actual operating environments by constructing a multi-dimensional digital twin model covering asset status, circulation behavior, environmental information, and management strategies. Furthermore, by combining causal analysis of asset behavior changes with a policy learning mechanism, the system can identify the potential causes of abnormal asset behavior and, based on this, generate goal-oriented scheduling or configuration recommendations, thereby achieving closed-loop dynamic supervision and optimization.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is to provide a method for intelligent supervision of state-owned assets based on digital twins, comprising the following steps:

[0007] S1. Map data from different sources into a state vector of a unified structure through an alignment function, and build a digital twin model based on the time series status of all assets;

[0008] S2. Analyze the potential causal relationships between variables in the evolution of asset behavior based on the digital twin model and construct a causal graph. The causal graph is used to extract the causal path of abnormal behavior at a certain moment for each asset.

[0009] S3. Build a reinforcement learning scheduling strategy model based on the acquired asset status and abnormal causal path in the digital twin model, and output the optimal scheduling strategy;

[0010] S4. Apply the optimal scheduling strategy to the supervision system and collect actual execution feedback while executing asset scheduling decision actions.

[0011] Preferably, in S1, data from different sources include ledger data, allocation and usage logs, sensor data, geographic data and policy data, which form a time series expression based on the current state and historical behavior to construct an institutionally sensitive twin state.

[0012] Preferably, in S2, the causal graph constructs a local abnormal causal path for each asset at a specific time to obtain a path chain, and the path chain is used to represent the causal source chain of the current abnormal behavior.

[0013] Preferably, in S3, by introducing a strategy learning objective function, the strategy learning objective function is in the following form:

[0014] ;

[0015] in, Represents the optimal strategy function, which will be deployed to the asset supervision platform for scheduling execution; Represents the scheduling strategy model, from the state Mapping to Action ; Reward function representing the dual constraint construction; Indicates the total duration of the scheduling cycle; Indicates that in the strategy expectations under Indicates the current scheduling period.

[0016] Better yet, a monitoring interface is developed based on the optimal strategy function and visualization technology, the scheduling strategy model is combined with real-time data, and the key status parameters and operating indicators of the physical assets are displayed in an intuitive manner on the interface.

[0017] Preferably, in said S4, the state-owned assets are remotely controlled and managed through the operation terminal according to the scheduling strategy that outputs the best scheduling, and the operation instructions are processed by the data processing center and then fed back to the corresponding Internet of Things sensors and execution devices.

[0018] Preferably, in S4, the feedback writing function is introduced to write the reality feedback back to the asset state of the digital twin model, and the edge weight confidence in the causal graph is adjusted.

[0019] In a second aspect, the present invention further provides a state-owned asset intelligent supervision system based on digital twins, comprising:

[0020] A digital twin model construction module is used to map data from different sources into a state vector of a unified structure through an alignment function and to build a digital twin model;

[0021] The causal graph construction module is used to analyze the potential causal relationship between various variables in the evolution of asset behavior based on the digital twin model, construct a causal graph, and extract the abnormal causal path of each asset's behavior at a certain moment;

[0022] A reinforcement learning scheduling strategy model building module is used to build a reinforcement learning scheduling strategy model based on the acquired asset status and abnormal causal path in the digital twin model;

[0023] The execution and feedback module executes the optimal scheduling strategy output by the reinforcement learning scheduling strategy model in the supervision system, writes the real feedback back to the asset status of the digital twin model through the feedback writing function, and adjusts the edge weight confidence in the causal graph.

[0024] The beneficial effects of the present invention are as follows: compared with traditional technologies, the state-owned assets intelligent supervision method and system based on digital twins provided by the present invention no longer stop at static visualization of asset status, but go deep into the entire asset use cycle, integrating "explainable anomaly identification mechanism" and "adaptive policy optimization capabilities" on the basis of digital twins. The former enhances the logic and accountability of regulatory behavior by exploring the driving factors behind asset behavior; the latter enables the system to have the ability to respond to complex management needs in real time through simulation feedback and policy evolution, thereby effectively dealing with practical problems such as cross-regional and cross-institutional asset scheduling optimization, asset sedimentation risk prevention and control, etc. This method is suitable for state-owned asset scenarios of different types and sizes, has strong generalization capabilities and deployability, and can significantly improve the intelligence level and resource allocation efficiency of the current state-owned asset supervision system. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a flow chart of the state-owned assets intelligent supervision method based on digital twins of the present invention.

[0026] Figure 2 This is the block diagram of the state-owned assets intelligent supervision system based on digital twins in the present invention. DETAILED DESCRIPTION

[0027] See also Figure 1 As shown, the present invention is about a method for intelligent supervision of state-owned assets based on digital twins, comprising the following steps:

[0028] S1. Map data from different sources into a state vector of a unified structure through an alignment function, and build a digital twin model based on the time series status of all assets;

[0029] This step aims to establish a multidimensional, dynamic digital twin model of state-owned assets to support subsequent causal modeling (S2), scheduling strategy generation (S3), and state feedback updates (S4). Its core task is to transform real-world data, often from dispersed, heterogeneous, and complex sources, into a twin state sequence with a unified structure, temporal continuity, and institutional awareness.

[0030] This includes various IoT sensors deployed on state-owned assets (such as the aforementioned production equipment), including temperature, pressure, vibration, current, and voltage sensors. These sensors collect real-time operational and environmental data on the assets, including ledger data, allocation and usage logs, sensor data, geographic data, and policy data. Based on the current state and historical behavior, they form a time-series representation for the construction of an institutionally sensitive twin state. This data is then transmitted to the system's data processing center via wireless communication technology for use in building a digital twin model.

[0031] Define structure alignment function , mapping data from different sources into a state vector of unified structure :

[0032] ;

[0033] in, Indicates the current asset status, Represents ledger data, such as asset type, acquisition year, and affiliated institution; Indicates the allocation and usage log; Indicates sensor data, such as operating status and whether it is online; Represents geographic data, such as latitude and longitude, and regional classification; Represents policy data, such as transfer limits and budget priorities; The multimodal structure aligner defined in this invention is used to integrate the above five types of data and output a state vector in a unified format. ; represents a vector space with d real dimensions.

[0034] Then construct the "time-aware state fusion function" , used to synthesize the current state and historical behavior to form a temporal expression :

[0035] ;

[0036] in, Indicates the current asset status; :Forward The state at the moment; K is the total length of history; Represents the decay weight of the historical state, is the decay rate and ; , Indicates the weight balance between the current state and the historical state. ; Represents the twin state expression after time fusion.

[0037] Introducing institutional vectors With embedded mapping , building an institutionally sensitive twin state :

[0038] ;

[0039] in, represents the institutional factor vector, such as policy priority, constraint level, and unit weight; represents the policy embedding mapping function, which converts the institutional factors into embedding vectors; represents the institutional influence coefficient, which indicates the degree of intervention of the institution on the behavior state (e.g. indicating moderate intervention); Represents the final system-enhanced twin state, which is used as input for downstream tasks.

[0040] Finally, the final digital twin model set is generated based on the above data :

[0041] The time series status of all assets is organized into the following structure:

[0042] ;

[0043] in, Indicates the asset number; Indicates the Assets at time the institutionally enhanced state; Represents assets the total length of time tracked; Indicates the total amount of assets; Represents a collection of digital twin models, which are used for causal modeling (S2), strategy training (S3), and feedback update (S4).

[0044] This structure provides a standardized, dynamic, and system-aware state representation for the entire patent system, with temporal consistency, structural consistency, and semantic stability, and is the necessary foundation for realizing "digital twin + intelligent supervision".

[0045] S2. Analyze the potential causal relationships between variables in the evolution of asset behavior based on the digital twin model and construct a causal graph. The causal graph is used to extract the causal path of abnormal behavior at a certain moment for each asset.

[0046] The goal of this step S2 is to build a digital twin model based on the step 1 , analyze the potential causal relationship between variables in the process of asset behavior evolution and construct a causal diagram , and accordingly extract the abnormal behavior path of each asset at a certain moment This structural result will serve as input to the subsequent strategy generation phase.

[0047] First, the state sequence data output in step 1 needs to be formatted into the time window required for mapping, providing a data basis for subsequent causal relationship learning.

[0048] Obtained through S1 As input, choose a fixed time window length , for each asset Construct a time window sequence:

[0049] ;

[0050] in, Indicates the A unique identifier for an asset; Represents assets In time the institutionally enhanced twin state (from S1); Represents assets Total tracking time length; Represents the window length, which is used for local data sampling for structure learning; Represents assets At the moment The state time window sequence of .

[0051] By designing a structural cause-effect diagram Learning mechanism:

[0052] By extracting causal structure from asset state variables, we provide structured priors for downstream policy generation instead of relying on black-box predictions.

[0053] This invention proposes a cross-asset joint mapping method , based on covariance dependence and structural stability constraints, construct a causal structure diagram of learning variables :

[0054] ;

[0055] in, , both represent any two variables in the asset behavior variable set; Represents the full set of asset behavior variables, including the asset state characteristic dimensions used to construct all nodes in the causal graph; represents the covariance calculated in the window samples, which is used to measure the strength of causal dependence; Indicates the structural stability (standard deviation) of the causal edge in different units or regions; The adjustment coefficient of the institutional disturbance factor is used to control the importance of the stability constraint (recommended range ); Represents the objective function of structural graph learning, which is used to select the optimal set of causal edges.

[0056] The purpose of extracting abnormal causal paths is to use the causal structure to explain single assets and guide subsequent actions, ensuring that the generated strategy has a clear behavioral traceability path.

[0057] Based on cause-effect diagram , for each asset In time Construct its local abnormal causal path :

[0058] First test Is there a variable in Extreme behavior occurs (e.g., exceeding a quantile threshold);

[0059] from Backtracking the causal path in , up to the root variable;

[0060] Get the path chain , indicating the causal source chain of the abnormal behavior.

[0061] The final output structure will be directly used as the core input for the next step of policy training to ensure that the scheduling strategy operates on the basis of structural interpretability. The final output of this step S2 is as follows:

[0062] : Asset behavior causal structure diagram (globally shared), provided to the subsequent strategy training stage;

[0063] :assets In time The causal path of abnormal behavior provides structural priors for subsequent strategy generation.

[0064] S3. Build a reinforcement learning scheduling strategy model based on the acquired asset status and abnormal causal path in the digital twin model, and output the optimal scheduling strategy;

[0065] The goal of this step S3 is to: (From S1 ) and abnormal causal paths (From S2 ) based on which a reinforcement learning scheduling strategy model is constructed. , realizing dynamic optimization allocation, shared management, idle release and risk avoidance of assets.

[0066] Digital twin status following step 1 , directly using it as the state input of reinforcement learning ensures the consistency between the state space and the actual asset operation characteristics, which is the basic premise for the operability of strategy training.

[0067] By directly using the twin state vector in S1 As a state space for reinforcement learning , this state has integrated historical behavior and institutional characteristics to avoid duplication of processing.

[0068] Represents assets At the moment The institutional enhancement state of ; : Reinforcement learning state space.

[0069] By defining the action space , including scheduling operations that can be executed for state-owned asset management. This step defines the set of scheduling behaviors that can be executed in the present invention, which is also the action range output by the policy model and must be consistent with the business process and authority boundaries of state-owned asset management. Including:

[0070] :Transfer to unit ; : Mark as shared available; : Released as idle;

[0071] : Return to the warehouse; : Restrict usage permissions, etc.

[0072] Action screening mechanism guided by causal paths:

[0073] Abnormal path based on the output of step 2 ,design a structured action screening method to exclude non-compliant or risky actions during strategy search,,improve training efficiency and strengthen causal consistency.

[0074] The present invention designs a structure-guided action screening function , which is used to exclude high-risk actions that conflict with abnormal path variables. The action screening function calculation formula is as follows:

[0075] ;

[0076] in, Represents assets At the moment abnormal causal paths; Represents each variable in the path; Indicates action For variables The institutional risk conflict score, with higher values ​​indicating greater non-compliance; Indicates the action system compliance threshold (e.g., 1.0); Represents the final subset of legal actions as the action space of the policy model.

[0077] The present invention proposes a reward function mechanism that simultaneously considers causal path perturbations and institutional compliance. This is the core control point for the rationality of reinforcement learning strategies, ensuring that the strategies are not purely profit-oriented, but give equal weight to behavioral structures and institutional goals.

[0078] We also designed a reward function driven by structural and institutional constraints. , the reward function calculation formula is as follows:

[0079] ;

[0080] in, represents the basic benefit function, such as asset utilization improvement (e.g. ); Indicates the abnormal causal path at the current moment; represents the causal edge in the path; Representing variables In performing actions The change value after represents the path perturbation penalty coefficient (e.g. ); Indicates that in a given asset state Next, execute the action The degree of conflict of institutional constraints triggered. The calculation method is: the system will act The affected institutional fields (such as the transfer destination unit type, budget level, asset category) and the current asset status The asset is compared with the institutional constraints in the asset (such as non-transfer across regions, non-transfer to non-shared units), and a basic penalty point is accumulated for each violation condition met. For example, if the asset status It specifies that “transfers are only allowed to units in the same region”. If the actual transfer target is across regions, If the "asset category restriction" is also violated, then , and so on, to form a positive integer value, which is used to punish non-compliance in policy behavior. If the current action violates the policy system, the value is positive; represents the regular weight of institutional consistency (e.g. ); Represents the final reward value, which is used to train intelligent strategies.

[0081] By defining the strategy optimization goal of reinforcement learning, the ultimately trained strategy model can form a stable, reasonable, and deployable scheduling function under the joint guidance of structural constraints and institutional penalties.

[0082] By introducing the strategy learning objective function, the strategy learning objective function is as follows:

[0083] ;

[0084] in, Represents the optimal strategy function, which will be deployed to the asset supervision platform for scheduling execution; Represents the scheduling strategy model, from the state Mapping to Action ; Reward function representing the dual constraint construction; Indicates the total duration of the scheduling cycle; Indicates that in the strategy expectations under Indicates the current scheduling period.

[0085] The training process of this strategy model will use the constrained action space With customized reward functions, we ensure that the generated scheduling policies are not only effective, but also compliant, explainable, and consistent with causal logic and institutional goals.

[0086] A monitoring interface is developed based on the optimal strategy function and visualization technology, which combines the scheduling strategy model with real-time data to display the key status parameters and operating indicators of physical assets in an intuitive manner on the interface.

[0087] S4. Apply the optimal scheduling strategy to the supervision system and collect actual execution feedback while executing asset scheduling decision actions.

[0088] The responsibility of this step S4 is to output the reinforcement learning strategy of S3 Applied to the real-world supervision system, while executing asset scheduling decisions, actual execution feedback is collected and written back to the digital twin model. The asset status and the cause-effect structure diagram Fine-tune the edge weight confidence in . This process does not involve any modeling behavior and is limited to state updates and parameter writing. It is the last step in building the system's closed-loop control capabilities.

[0089] According to the dispatching strategy with the best output, state-owned assets are remotely controlled and managed through the operation terminal. The operation instructions are processed by the data processing center and then fed back to the corresponding Internet of Things sensors and execution devices.

[0090] For each asset status , the action is executed by the policy model output: ;

[0091] in, The optimal policy function output by the policy model in S3; The current asset status, from the twin model of S1 ; Indicates execution of actions, such as allocation, recovery, sharing, restriction, etc.; a new state is generated after execution , that is, the result state in reality.

[0092] By introducing a feedback writing function, the real feedback is written back to the asset status of the digital twin model, and the edge weight confidence in the causal graph is adjusted. The mathematical expression is as follows:

[0093] ;

[0094] in, is the state update function, which is used to describe how to derive the next state from the current state and observation value. Indicates the feedback fusion weight, controls the update amplitude, and is generally ; Indicates the final updated state write , without changing its structure, only the value;

[0095] The response of a variable after the policy is executed is used to adjust the weight of the relevant edge in the causal graph (instead of structural learning):

[0096] ;

[0097] in, Representing a cause-effect diagram Middle side confidence level; Indicates the response after the variable is executed; Represents the learning rate coefficient, which is generally taken as .

[0098] This step S4 does not involve model training, learning, or reconstruction. It only updates the results into the model to ensure that the model remains up-to-date.

[0099] This step S4 does not change the edge set of the causal graph, but only fine-tunes the weights of the existing edges to ensure Evolvability and dynamism.

[0100] like Figure 2 As shown, in a second aspect, the present invention further provides a state-owned assets intelligent supervision system based on digital twins, comprising:

[0101] A digital twin model construction module is used to map data from different sources into a state vector of a unified structure through an alignment function and to build a digital twin model;

[0102] The causal graph construction module is used to analyze the potential causal relationship between various variables in the evolution of asset behavior based on the digital twin model, construct a causal graph, and extract the abnormal causal path of each asset's behavior at a certain moment;

[0103] A reinforcement learning scheduling strategy model building module is used to build a reinforcement learning scheduling strategy model based on the acquired asset status and abnormal causal path in the digital twin model;

[0104] The execution and feedback module executes the optimal scheduling strategy output by the reinforcement learning scheduling strategy model in the supervision system, writes the real feedback back to the asset status of the digital twin model through the feedback writing function, and adjusts the edge weight confidence in the causal graph.

[0105] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. The state-owned assets intelligent supervision method based on digital twins is characterized by: The following steps are involved: S1. Map data from different sources into a state vector of a unified structure through an alignment function, and build a digital twin model based on the time series status of all assets; S2. Analyze the potential causal relationships between variables in the evolution of asset behavior based on the digital twin model and construct a causal graph. The causal graph is used to extract the causal path of abnormal behavior at a certain moment for each asset. S3. Build a reinforcement learning scheduling strategy model based on the acquired asset status and abnormal causal path in the digital twin model, and output the optimal scheduling strategy; S4. Apply the optimal scheduling strategy to the supervision system and collect actual execution feedback while executing asset scheduling decision actions; Wherein, the S3 includes: Using the digital twin model as the state space of the scheduling strategy model; Action screening based on abnormal causal paths is used to exclude non-compliant or risky actions during policy search. The calculation formula for the action screening function that meets the causal graph conditions is as follows: ； in, Represents assets At the moment abnormal causal paths; Represents each variable in the path; Indicates action For variables The institutional risk conflict score, with higher values ​​indicating greater non-compliance; Indicates the action system compliance threshold; Represents the final subset of legal actions as the action space of the scheduling strategy model of the reinforcement learning model; Designing the reward function , the reward function calculation formula is as follows: ； in, represents the basic benefit function; represents the causal edge in the path; Representing variables In performing actions The change value after represents the path disturbance penalty coefficient; Indicates that in a given asset state Next, execute the action the degree of conflict between institutional constraints triggered; By introducing the strategy learning objective function, the strategy learning objective function is as follows: ； in, Represents the optimal strategy function, which will be deployed to the asset supervision platform for scheduling execution; Represents the scheduling strategy model, from the state Mapping to Action ; Reward function representing the dual constraint construction; Indicates the total duration of the scheduling cycle; Indicates that in the strategy expectations under Indicates the current scheduling period.

2. The method for intelligent supervision of state-owned assets based on digital twins according to claim 1 is characterized in that: In S1, data from different sources include ledger data, allocation and usage logs, sensor data, geographic data, and policy data, which form a time series expression based on the current state and historical behavior, and are used to construct an institutionally sensitive twin state.

3. The method for intelligent supervision of state-owned assets based on digital twins according to claim 1 is characterized in that: In S2, the causal graph constructs a local abnormal causal path for each asset at a specific time to obtain a path chain, which is used to represent the causal source chain of the current abnormal behavior.

4. The method for intelligent supervision of state-owned assets based on digital twins according to claim 1 is characterized in that: A monitoring interface is developed based on the optimal strategy function and visualization technology, which combines the scheduling strategy model with real-time data to display the key status parameters and operating indicators of physical assets in an intuitive manner on the interface.

5. The method for intelligent supervision of state-owned assets based on digital twins according to claim 1 is characterized in that: In S4, according to the optimal scheduling strategy, the state-owned assets are remotely controlled and managed through the operation terminal, and the operation instructions are processed by the data processing center and then fed back to the corresponding Internet of Things sensors and execution devices.

6. The method for intelligent supervision of state-owned assets based on digital twins according to claim 1 is characterized in that: In S4, the feedback writing function is introduced to write the reality feedback back to the asset status of the digital twin model, and the edge weight confidence in the causal graph is adjusted.

7. The state-owned assets intelligent supervision system based on digital twins is characterized by: include: A digital twin model construction module is used to map data from different sources into a state vector of a unified structure through an alignment function and to build a digital twin model; The causal graph construction module is used to analyze the potential causal relationship between various variables in the evolution of asset behavior based on the digital twin model, construct a causal graph, and extract the abnormal causal path of each asset's behavior at a certain moment; A reinforcement learning scheduling strategy model building module is used to build a reinforcement learning scheduling strategy model based on the acquired asset status and abnormal causal path in the digital twin model; The execution and feedback module executes the optimal scheduling strategy output by the reinforcement learning scheduling strategy model in the supervision system, writes the real feedback back to the asset status of the digital twin model through the feedback writing function, and adjusts the edge weight confidence in the causal graph.

Images