A smart management platform system integrating data from the entire construction process of the seawall project.

Abstract

This application relates to the fields of smart construction and digital twin technology, and discloses a smart management platform system for the integration of data throughout the entire construction process of a seawall project. The system includes: a multi-source heterogeneous data acquisition module for acquiring and standardizing construction data; a core data fusion and reasoning module for constructing a causal graph representing the relationships between elements in the construction process, and calculating the credibility of each state node through a multi-source evidence theory fusion engine; an intelligent application and evolution module for performing forward risk propagation early warning and root cause tracing diagnosis based on the causal graph; and a comprehensive visualization and decision support module for integrating and displaying the analysis results with a BIM / GIS model. This invention overcomes the problems of data silos and information uncertainty by constructing a causal knowledge graph and an evidence fusion reasoning mechanism, achieving proactive early warning, accurate source tracing, and dynamic self-adaptation of construction risks, significantly improving the smart management level of seawall projects.

Description

Technical Field

[0001] This invention relates to the fields of smart construction and digital twin technology, specifically a smart management platform system for the integration of data throughout the entire construction process of seawall projects. Background Technology

[0002] Seawall projects are critical infrastructure projects to ensure the safety of life and property in coastal areas. Their construction involves multiple disciplines, complex procedures, and dynamic changes in the geological and hydrological environment, which places extremely high demands on the management of the entire construction process.

[0003] Currently, the application of information technology in seawall construction has become increasingly widespread. For example, Building Information Modeling (BIM) technology is used for 3D visualization design and collaboration, Geographic Information System (GIS) technology is used for macro-environmental and spatial data management, and Internet of Things (IoT) technology uses various sensors to monitor key indicators in real time. However, in practice, these technologies are often applied independently. Different systems (such as quality, safety, schedule, and environmental monitoring) generate data with inconsistent standards, leading to inefficient information flow and deep integration between departments and systems. When faced with conflicting or uncertain information from different sources, managers struggle to form a comprehensive, accurate, and reliable understanding of a specific state of the project.

[0004] Furthermore, existing construction risk management models still largely rely on the personal experience of project managers and post-event remediation. Due to the lack of quantitative models for the complex causal relationships between construction procedures, material conditions, and environmental factors, the system cannot proactively deduce the propagation paths of potential risks and issue early warnings based on the current state. When quality or safety issues occur, tracing the root cause often relies on manually reviewing a large number of construction records, a cumbersome and inefficient process that makes it difficult to quickly and accurately pinpoint the root cause, thus missing the optimal time for intervention.

[0005] More importantly, existing management systems and analytical models are typically static. Once their analytical logic and parameters are set, they are difficult to change and cannot dynamically adapt to sudden disturbances at the construction site, such as severe weather or material batch changes. These events temporarily alter the causal relationship between existing processes, and static models cannot capture such changes, leading to analytical results that deviate from reality. Furthermore, the valuable experience and lessons learned from completed projects lack a systematic mechanism for mining, verifying, and storing this knowledge, making it difficult to form a reusable knowledge system to guide subsequent new projects. This results in inefficient cross-project knowledge transfer and iteration. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a smart management platform system for the entire construction process of seawall projects, which integrates data. The system aims to solve the problems of inaccurate status assessment due to difficulties in data fusion, passive and delayed risk warnings, low efficiency in tracing the root causes of abnormal events, and static and rigid analysis models that lack dynamic adaptation and cross-project self-learning capabilities.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solution: a smart management platform system for the whole process of seawall construction data integration. This system integrates multi-source data acquisition, causal knowledge graph construction, evidence theory fusion reasoning, risk propagation and tracing, dynamic evolution of the graph, and three-dimensional visualization technology.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] The intelligent management platform system for the entire construction process of the seawall project integrates data, including a multi-source heterogeneous data acquisition module, a core data fusion and reasoning module, an intelligent application and evolution module, and a comprehensive visualization and decision support module.

[0010] The multi-source heterogeneous data acquisition module, as the data input terminal of the system, is responsible for acquiring raw heterogeneous data from various data sources such as IoT sensors, BIM / GIS models, manual data entry records, and external environmental data.

[0011] This module has an internal data acquisition unit that performs data acquisition through a configured dedicated interface; it also has a data preprocessing and standardization unit that cleans the acquired raw data to remove outliers, performs spatiotemporal alignment to unify the spatiotemporal benchmark, and performs formatting to finally generate a standardized information package with a unified data structure for use by downstream modules.

[0012] The core data fusion and inference module is connected to the multi-source heterogeneous data acquisition module and is used to receive and parse standardized information packets. The core function of this module is to construct and maintain a construction process causal graph (P-CG). The data structure of this graph is defined by the following elements:

[0013] Event nodes that represent construction procedures or actions ( );

[0014] Status nodes representing the state of engineering quality, safety, or environment ( );

[0015] Process dependency edges connecting event nodes with a construction sequence ( );

[0016] And causal influence edges connecting events and state nodes or states and state nodes, with influence strength weights. ).

[0017] To calculate the credibility of state nodes, this module incorporates a multi-source evidence theory fusion engine.

[0018] This engine is for any state node. First, configure it with an identification framework consisting of mutually exclusive propositions. .

[0019] Subsequently, evidence from different standardized information packages (such as sensor readings and detection reports) is transformed into multiple basic probability allocation functions (BPA) through a preset function.

[0020] Finally, these BPAs are fused using Dempster's combination rule to address potential conflicts between pieces of evidence. The fused BPA function is then obtained. The calculation method is as follows:

[0021] ;

[0022] in, , where is the conflict coefficient between pieces of evidence. and For focal elements under different evidence sources, the system calculates that this state node belongs to the proposition based on the fused BPA function. State Trust Zone This data is then output as the credibility data of the state nodes.

[0023] The intelligent application and evolution module is connected to the core data fusion and reasoning module, receives state node credibility data, and performs analysis and evolution based on the construction process causal graph (P-CG). This module internally includes a risk propagation and causal tracing unit. During forward risk propagation, when any upstream node ( abnormal trust level If the risk threshold is exceeded, the system performs forward probability propagation along the causal influence edges in the graph, updates the risk status of downstream nodes, and generates an early warning. During root cause tracing, when a terminal state node ( If a node is identified as an anomaly, the system starts from that node and searches backwards on the graph for all causal paths, calculating the causal path for each path. confidence score This is used to identify the most likely root cause. The scoring formula is as follows:

[0024] ;

[0025] in, Abnormal trust level of the terminal node, The abnormal trust level of intermediate nodes on the path. For connecting nodes and The weight of the causal influence edge.

[0026] In addition, this module includes a construction disturbance event dynamic graph reconstruction unit to improve the system's adaptability. When external construction disturbance events such as rainstorms or material replacements are detected, this unit matches the event type from the reconstruction rule base and temporarily adjusts the weights of relevant causal influence edges in the P-CG to make the system's reasoning more consistent with the current working conditions.

[0027] After the disturbance event ends, the weights automatically revert to their original values. This module also includes a cross-project causal graph self-learning unit, which extracts historical data from completed projects, uses causal discovery algorithms to uncover potential causal relationships, and updates the domain knowledge base with new knowledge after human-machine collaborative verification by domain experts, thereby achieving knowledge accumulation and iteration.

[0028] The integrated visualization and decision support module, connected to the intelligent application and evolution module, is responsible for presenting the analysis results to project managers in an intuitive way. This module includes a BIM / GIS fusion visualization unit that associates status node credibility data with the BIM / GIS 3D model, displaying risk distribution on specific physical components of the model through heatmaps or highlight rendering.

[0029] Simultaneously, an interactive causal chain browser unit is set up to highlight the risk propagation path or root cause tracing path on the P-CG topology, and support users to interactively query the detailed evidence source of any node.

[0030] This invention provides a smart management platform system for the entire construction process of seawall projects, integrating data. It offers the following advantages:

[0031] 1. This invention standardizes the data from the entire seawall project process by setting up a multi-source heterogeneous data acquisition module, and uses a core data fusion and inference module to construct a construction process causal graph (P-CG) to uniformly describe the complex relationships between construction elements. In particular, through the application of a multi-source evidence theory fusion engine, the system can quantify conflicting or uncertain information from different channels, calculate accurate state trust intervals, overcome the assessment inaccuracies caused by data silos and information inconsistencies in traditional methods, and significantly improve the accuracy and reliability of construction status assessment.

[0032] 2. This invention utilizes the risk propagation and causal tracing unit within the intelligent application and evolution module, based on the topological structure of the construction process causal graph (P-CG), to achieve proactive risk warning and precise root cause diagnosis. Through the risk forward propagation algorithm, the system can automatically deduce potential risks and issue early warnings; when an anomaly occurs, through the root cause backward tracing algorithm and the quantitative ranking of path confidence scores, the most likely root cause path can be quickly located. This changes the previous passive management model that relied on manual experience judgment, providing decision-makers with timely and interpretable decision-making basis.

[0033] 3. This invention, by setting up a dynamic graph reconstruction unit for construction disturbance events, enables the system to temporarily adjust the weights of the causal graph when encountering sudden construction disturbance events, allowing the inference model to dynamically adapt to on-site conditions. Simultaneously, a cross-project causal graph self-learning unit mines and verifies new causal knowledge from historical data using causal discovery algorithms. This design frees the system from the limitations of static models, giving it adaptive capabilities to dynamic conditions and self-learning capabilities for cross-project knowledge iterative evolution. Attached Figure Description

[0034] Figure 1 A schematic diagram of the functional modules of a smart management platform system for data integration of the entire construction process of a seawall project, according to an embodiment of the present invention;

[0035] Figure 2 This is a functional unit diagram of a multi-source heterogeneous data acquisition module according to an embodiment of the present invention;

[0036] Figure 3 This is a schematic diagram of the construction process cause-effect graph according to an embodiment of the present invention;

[0037] Figure 4 This is a schematic diagram of the workflow of a multi-source evidence theory fusion engine according to an embodiment of the present invention;

[0038] Figure 5 This is a schematic diagram of the risk propagation and causal tracing process according to an embodiment of the present invention;

[0039] Figure 6 This is a schematic diagram of the dynamic graph reconstruction process for construction disturbance events according to an embodiment of the present invention;

[0040] Figure 7 This is a schematic diagram of a cross-project causal graph self-learning process according to an embodiment of the present invention;

[0041] Figure 8 This is a functional unit diagram of the integrated visualization and decision support module 40 according to an embodiment of the present invention.

[0042] The module comprises: 10. Multi-source heterogeneous data acquisition module; 11. Data acquisition unit; 12. Data preprocessing and standardization unit; 20. Core data fusion and reasoning module; 21. Construction process causal graph construction and storage unit; 22. Multi-source evidence theory fusion engine; 30. Intelligent application and evolution module; 31. Risk propagation and causal tracing unit; 32. Construction disturbance event dynamic reconstruction unit; 33. Cross-project causal relationship mining and graph self-learning unit; 40. Comprehensive visualization and decision support module; 41. BIM / GIS integrated visualization unit; and 42. Interactive causal chain browser unit. Detailed Implementation

[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] See attached document Figure 1 The present invention provides a smart management platform system for data integration of the entire construction process of seawall projects. The system may include: a multi-source heterogeneous data acquisition module 10, a core data fusion and reasoning module 20, an intelligent application and evolution module 30, and a comprehensive visualization and decision support module 40.

[0045] The multi-source heterogeneous data acquisition module 10 is designed to acquire data from multiple sources at the construction site, including IoT sensors, building information models, and manually entered construction process records. After preliminary processing, the multi-source heterogeneous data acquisition module 10 transmits the acquired data to the core data fusion and inference module 20.

[0046] The core data fusion and reasoning module 20 is connected to the multi-source heterogeneous data acquisition module 10. The core function of this module 20 is to transform the received multi-source heterogeneous data into probabilistic evidence based on the causal graph of the construction process, and to apply a multi-source evidence fusion algorithm to perform uncertainty reasoning on key state nodes during the construction process, ultimately outputting quantitative data characterizing the credibility of each state node. The core data fusion and reasoning module 20 then transmits the calculated quantitative data to the intelligent application and evolution module 30.

[0047] The intelligent application and evolution module 30 is connected to the core data fusion and reasoning module 20. Based on the received state node credibility data, this module 30 performs forward propagation of risks and backward tracing of fault root causes. Furthermore, it is configured to dynamically adjust the parameters of the causal graph of the construction process according to real-time construction disturbance events, and can perform cross-project causal relationship mining based on historical project data to achieve self-learning and evolution of the knowledge base. The intelligent application and evolution module 30 transmits the analysis and prediction results to the integrated visualization and decision support module 40.

[0048] The integrated visualization and decision support module 40 is connected to the intelligent application and evolution module 30. The function of this integrated visualization and decision support module 40 is to perform three-dimensional fusion visualization of risk warning, traceability results, and the credibility of status nodes with building information model or geographic information system model, providing project managers with an intuitive and interactive decision support interface.

[0049] To implement this invention, the system can be physically deployed in a hardware environment consisting of a cloud server, an edge computing gateway, and on-site IoT terminals. At the software level, the system can be constructed as a layered architecture, including a perception layer for data acquisition, a data layer for data storage and management, a model and service layer for implementing core algorithms, and an application layer for human-computer interaction.

[0050] See attached document Figure 2 The function of the multi-source heterogeneous data acquisition module 10 is to collect various types of data generated throughout the entire construction process of the seawall project and process them into a standardized data format that can be used by the core data fusion and inference module 20. This module may include a data acquisition unit 11 and a data preprocessing and standardization unit 12.

[0051] Data acquisition unit 11 is responsible for communicating and acquiring data from multiple different data sources both inside and outside the construction site. To ensure compatibility with different types of data sources, data acquisition unit 11 integrates various data interfaces. In one specific implementation, the data interfaces may include, but are not limited to:

[0052] The sensor data interface is used to receive data from various IoT sensor terminals deployed in the field. This interface supports multiple IoT communication protocols, such as, but not limited to, Message Queuing Telemetry Transport Protocol (MQTT) or a RESTful API interface based on Hypertext Transfer Protocol (HTTP). The data packets reported by the sensor terminals must include at least a unique device identifier, a data acquisition timestamp, and one or more monitored values, such as settlement, seepage pressure, and ambient temperature and humidity.

[0053] The spatial information data interface is used to access Building Information Modeling (BIM) data and Geographic Information System (GIS) data for engineering projects. This interface has the ability to parse industry-standard data formats; for example, it can parse files conforming to the Industrial Foundation Class (IFC) standard to extract geometric information, material properties, and engineering parameters of components, or parse geospatial data formats (such as GeoJSON and Shapefile) to obtain the geographic coordinates and topographic information of the engineering area.

[0054] The construction process data interface is used to receive process information entered by site management personnel through specific terminals (such as mobile applications or web browsers). To ensure the structured nature of the entered data, the interface provides standardized data entry forms. The fields of the forms correspond to event nodes or status nodes in the cause-and-effect graph of the construction process. The types of data entered include construction logs, material inspection reports, quality assessment records, and image data of concealed works acceptance.

[0055] The external system data interface is used to obtain external data related to the engineering environment from third-party service systems. This interface periodically obtains information such as regional weather forecasts, real-time hydrological and tidal conditions, and geological disaster early warnings by calling external application programming interfaces (APIs), serving as an important supplementary data source for the engineering environment status.

[0056] The data preprocessing and standardization unit 12 is connected to the data acquisition unit 11. Its function is to clean, align and format the received raw data to generate a unified, standardized information package that can be processed later.

[0057] The data preprocessing and standardization unit 12 performs data cleaning operations to identify and process noise and anomalies in the raw data stream. For continuous sensor time-series data, this unit can use statistical methods, such as those based on the 3-sigma principle or local anomaly factor algorithms, to detect outliers that exceed the normal fluctuation range. For detected anomalous data, they can be marked, removed, or repaired using interpolation algorithms (such as linear interpolation or spline interpolation) according to preset strategies.

[0058] The data preprocessing and standardization unit 12 performs spatiotemporal alignment operations, assigning a unified time and spatial reference to data from all sources. In the time dimension, this unit converts the timestamps of all data to Coordinated Universal Time (UTC) to eliminate time zone differences. In the spatial dimension, based on the data content or its metadata, this unit uses the BIM or GIS model parsed by the spatial information data interface to associate each data point with a unique three-dimensional spatial coordinate or engineering station number within the project's unified coordinate system, achieving precise physical location for all data.

[0059] The data preprocessing and standardization unit 12 ultimately performs a formatting operation, encapsulating the cleaned and aligned data into a standardized information package. This information package is a structured data object, such as a JSON object, which contains at least the following fields:

[0060] A unique data identifier; a source identifier indicating the data's origin; a standard timestamp indicating the data acquisition time; standard spatial coordinates indicating the data's spatial location; specific numerical values ​​or states representing the data's content; and a confidence score characterizing the initial credibility of the data. This confidence score can be preset based on prior knowledge such as the reliability of the data source and the accuracy of the sensors, serving as the initial input for subsequent evidence reasoning.

[0061] Finally, the multi-source heterogeneous data acquisition module 10 will transmit the generated standardized information package to the core data fusion and reasoning module 20 in real time through the internal data bus or message queue, so that it can perform subsequent evidence modeling and fusion reasoning.

[0062] See attached document Figure 3 The core data fusion and reasoning module 20 includes a construction process causal graph construction and storage unit 21. The function of this unit 21 is to provide the entire intelligent management platform system with a knowledge model framework that describes the inherent logic and physical laws of the seawall project. This knowledge model is formally defined as a construction process causal graph (P-CG), mathematically expressed as a weighted directed acyclic graph. .

[0063] In the figure In the definition, This represents the set of all nodes in the graph. (Node set) It consists of two subsets, namely the set of event nodes. With the set of state nodes An event node It represents a discrete action or operation in the construction process, such as "excavation of the foundation trench in section K1+200" or "concrete pouring in section A03".

[0064] A state node It represents a continuous engineering entity or environmental property that can be measured or assessed at a certain point in time, such as "3-day strength of concrete in AO3 compartment" or "settlement rate at monitoring point S5".

[0065] This represents the set of all directed edges in the graph, which also consists of two subsets: the set of process-dependent edges. With causal influence edge set A process depends on an edge. This indicates the necessary sequence of construction procedures between two event nodes. A causal influence edge. Represents a predecessor node (It can be an event node or a state node) for a successor state node There is a physical or logical causal relationship.

[0066] Representative and causal influence edge set The associated set of weights. For each causal influence edge. Each of them has a corresponding causal influence weight. ,and This weight Used for quantizing predecessor nodes State changes affect successor nodes The contribution of state changes can be pre-calibrated by a domain expert knowledge base or obtained through statistical analysis of historical data.

[0067] The construction process causal graph construction and storage unit 21 internally contains a configurable domain knowledge base, which forms the basis for generating P-CG. This knowledge base encapsulates general engineering knowledge, including national and industry construction standards, general principles for water conservancy engineering design, and verified expert experience.

[0068] When a specific seawall project is initiated, the construction process cause-effect graph construction and storage unit 21 performs a graph instantiation operation. In this operation, the construction process cause-effect graph construction and storage unit 21 first parses the input files such as the BIM model, construction organization design, and schedule of the specific project, and extracts the key engineering components, construction procedures, and quality control nodes from them.

[0069] Subsequently, the construction process causal graph construction and storage unit 21 matches and binds these project-specific entities with the general node templates in the domain knowledge base, thereby generating a unique P-CG instance specific to the project. For example, a general "concrete pouring" event node in the knowledge base is instantiated into a "K1+200 section A03 concrete pouring" node, and is uniquely associated with the corresponding slab component in the BIM model.

[0070] To achieve efficient management and querying of the P-CG, the construction process causal graph construction and storage unit 21 uses graph database technology for physical storage. In one optional implementation, Neo4j, JanusGraph, or a similar graph database can be used. Each event node or state node in the P-CG is stored as a node object in the database, and the node's attributes include at least a unique identifier, node type (event or state), and node name. Each process dependency edge or causal influence edge in the P-CG is stored as a directed relation (edge) object in the database, and the relation's attributes include at least the relation type (process dependency or causal influence). For causal influence edges, their attributes must also include quantified causal influence weights. This storage method can fully leverage the performance advantages of graph databases in handling complex relationships, providing efficient graph traversal support for subsequent risk propagation and causal tracing algorithms.

[0071] See attached document Figure 4 The core data fusion and reasoning module 20 includes a multi-source evidence theory fusion engine 22. This engine works in conjunction with the construction process causal graph construction and storage unit 21, and its core function is to process each state node in the P-CG. Uncertainty reasoning and quantitative evaluation are conducted on the true state.

[0072] Before executing inference, the multi-source evidence theory fusion engine 22 needs to perform a process for each state node. Configure its recognition framework Identification Frame It is a complete and mutually exclusive set containing all possible states of the state node. For example, for a state node representing "base compaction", its identification framework can be configured as follows: .

[0073] The multi-source evidence theory fusion engine 22 receives standardized information packets directed to specific state nodes from the multi-source heterogeneous data acquisition module 10, and transforms the data carried in each information packet into a Basic Probability Assignment (BPA) function. This function is denoted as... , indicating from the Evidence from a data source for the state node Identification Frame The degree of support for each subset. The conversion process uses different mapping functions depending on the data type.

[0074] In one specific implementation, when the received data is continuous numerical (such as sensor readings), the multi-source evidence theory fusion engine 22 uses a preset membership function for transformation. For example, for compaction test values... Membership functions for three state intervals—"qualified," "critical," and "unqualified"—can be preset. When When a value falls within a certain range, the function will calculate one or more BPA values. For example, a value close to the lower limit of the "acceptable" range might be mapped to a value that falls within the range of acceptable values. The trust allocation in Proposition 0.7, for Trust allocation for proposition 0.1, and its impact on the entire identification framework. Trust allocation of 0.2 (representing uncertainty).

[0075] In another specific implementation, when the received data is of discrete categorical type (such as a quality inspection report labeled "qualified" or "unqualified"), the engine employs discrete mapping rules. For example, a "qualified" report can be mapped to... The trust assignment for proposition 0.9 is used, while the trust for proposition 0.1 is assigned to the entire recognition framework. This indicates that the report itself is not entirely reliable.

[0076] This function The following mathematical constraints must be satisfied:

[0077] ;

[0078] in, It is an identification framework any subset, It is evidence "Proposition" node The true state belongs to the set Support level.

[0079] When there are multiple independent BPA functions (e.g.) and Pointing to the same state node At that time, the multi-source evidence theory fusion engine 22 uses the Dempster combination rule for fusion calculation to generate a new BPA function that integrates all the evidence. For the identification framework any non-empty subset The formula for calculating the BPA value after fusion is as follows:

[0080] ;

[0081] The conflict coefficient in the formula The formula used to quantify the degree of contradiction between different pieces of evidence is as follows:

[0082] ;

[0083] in, and Evidence respectively and Supported A subset of the evidence. When there are more than two pieces of evidence, this fusion process can be performed iteratively.

[0084] After completing evidence fusion, the engine further calculates arbitrary state hypotheses. Trust function Similarity function This is used to form a trust interval for that state.

[0085] Trust function This indicates that all evidence fully supports the proposition. Total trust:

[0086] ;

[0087] plausible function Indicates all that are not related to the proposition The greatest support that contradictory evidence can provide:

[0088] ;

[0089] Finally, the multi-source evidence theory fusion engine 22 will calculate the representations of each state node. BPA function after state fusion and the trust intervals for each critical state. The data is output to the intelligent application and evolution module 30 as a quantitative input for risk propagation and causal tracing.

[0090] See attached document Figure 5 The intelligent application and evolution module 30 includes a risk propagation and causal tracing unit 31. This risk propagation and causal tracing unit 31 is connected to the core data fusion and reasoning module 20, receives the state node trust data output by the module, and performs proactive risk warning and diagnostic fault tracing based on the construction process causal graph (P-CG).

[0091] The risk propagation and causal tracing unit 31 includes a risk forward propagation algorithm, which aims to provide proactive early warning of potential construction risks. The triggering condition for this algorithm is:

[0092] When the core data fusion and inference module 20 calculates any upstream node in P-CG A subset of abnormal states Trust level Exceeding a preset risk threshold The algorithm is activated at that time.

[0093] After the algorithm is activated, from the triggered node Begin by tracing all causal influence edges in P-CG that originate from it. Perform a forward graph traversal. For each affected downstream state node... The algorithm assigns the trust level of the abnormal state of the upstream node based on the causal influence weight. This is transformed into new evidence representing "transmission risk." This evidence is constructed as a basic probability assignment function (BPA), denoted as... The specific assignment method is as follows:

[0094] ;

[0095] ;

[0096] in, It is a downstream node The corresponding subset of abnormal states, yes The identification framework. This BPA function. Subsequently, it was input as new evidence into the multi-source evidence theory fusion engine 22 of the core data fusion and reasoning module 20, and linked with the node... The existing evidence is merged. This merging operation updates the node. The state trust level. If the updated It also exceeded the risk threshold. Then the propagation process will be in the form of Starting from a new point, continue iterating downstream until the graph reaches a leaf node or the risk confidence level of the propagation decays below the threshold.

[0097] The risk propagation and causal tracing unit 31 also includes a root cause backward tracing algorithm, which automatically diagnoses and locates the most likely root cause when an anomaly occurs in a critical quality or safety state. The algorithm is triggered when a terminal or critical state node in the PCG experiences an anomaly. When identified as an abnormal state, such as a subset of its abnormal states Trust level Reaching a confirmation threshold that characterizes a fault.

[0098] After the algorithm is activated, from this abnormal node Starting from the point (P-CG), perform a depth-first or breadth-first search of the graph along the opposite direction of the causal influence edges to find all causal paths that can reach the node. For each identified causal path... The algorithm calculates a path confidence score. This is used to quantify the likelihood that the path is the root cause. A specific scoring method is the product of the anomaly trust level and the causal influence weight of each node on the path:

[0099] ;

[0100] in, It is a path The directed edge on, It is the causal influence weight of that edge. It is a node Trust level for abnormal states.

[0101] The backward tracing algorithm traverses all possible upstream causal paths and calculates the path confidence score accordingly. The risk propagation and causal tracing unit 31 then sorts the results. Finally, it outputs the one or more causal paths with the highest scores as the most credible root cause chain of the failure.

[0102] Finally, the risk propagation and causal tracing unit 31 transmits the early warning information generated by forward risk propagation and the diagnostic path generated by backward root cause tracing to the integrated visualization and decision support module 40 for presentation.

[0103] See attached document Figure 6 The intelligent application and evolution module 30 further includes a construction disturbance event dynamic reconstruction unit 32. The core function of this construction disturbance event dynamic reconstruction unit 32 is to endow the system of the present invention with the ability to adapt to dynamic emergencies at the construction site, ensuring the timeliness and accuracy of the inference model.

[0104] The construction disturbance event dynamic reconstruction unit 32 is configured as a continuously running monitoring service. This unit has a pre-configured "construction disturbance event library." This library programmatically associates common disturbance events in seawall projects (e.g., but not limited to "typhoon warning," "heavy rainfall," "major equipment failure," and "emergency design changes") with a specific set of data triggering conditions.

[0105] The construction disturbance event dynamic reconstruction unit 32 continuously monitors the standardized real-time data stream from the multi-source heterogeneous data acquisition module 10. When the monitored data meets the predefined triggering conditions of a disturbance event in the event library (for example, the data value received by the 'environmental wind force' status node continuously exceeds a certain threshold, or the 'construction machinery status' node receives a 'fault' code), the construction disturbance event dynamic reconstruction unit 32 is activated.

[0106] Once activated, the unit retrieves the currently triggered disturbance event from the event database. The corresponding "dynamic reconstruction rule set" explicitly defines which causal influence edges exist in the construction process causal graph (P-CG). Will be affected by this disturbance event The impact of this, and how to quantify and adjust for this impact.

[0107] The construction disturbance event dynamic reconstruction unit 32 then performs a dynamic graph reconstruction operation. This operation does not change the topology of the PG (i.e., it does not add or delete nodes or edges), but rather adjusts the weights of the affected causal edges. Make temporary dynamic adjustments to update it to a scenario weight. The weight adjustment process can be described by the following formula:

[0108] ;

[0109] in: It is a protest event After it happened, the side Temporary, adjusted weights of causal effects;

[0110] It is the original baseline weight of the edge stored in P-CG, representing the normal operating condition;

[0111] Represents the specific perturbation event that has been identified;

[0112] It is related to a disturbance event and specific edges The relevant perturbation impact factor function; the value of this factor is predefined by the rule set, for example, This indicates an enhancement of the causal effect. This indicates a reduction in the causal effect.

[0113] In one specific implementation, when a "heavy rainfall" event is identified... At that time, the dynamic reconstruction unit 32 of the construction disturbance event will query the rule set to locate the "environmental rainfall" status node. Pointing to the "Slope Stability of Foundation Pit" status node The causal influence edge, and its weight Multiply by a factor greater than 1 .

[0114] This dynamic adjustment of weights enables the core data fusion and reasoning module 20 to automatically and quantitatively increase the consideration of the disturbance factor when performing M-SEF evidence fusion and the risk propagation and causal tracing unit 31 performs risk propagation, thereby improving the timeliness and scenario accuracy of risk warning.

[0115] Dynamic reconfiguration is temporary. The construction disturbance event dynamic reconfiguration unit 32 also includes a state monitoring and rollback mechanism. This mechanism continuously monitors the triggering event. The data conditions. When it is detected that the condition has been resolved (e.g., heavy rainfall stops and ambient humidity returns to below the threshold), the unit will automatically perform a rollback operation, removing all modified temporary weights. Restore to its original baseline weights This allows the P-CG to return to its normal operating state.

[0116] See attached document Figure 7 The intelligent application and evolution module 30 further includes a cross-project causal relationship mining and graph self-learning unit 33. The function of this cross-project causal relationship mining and graph self-learning unit 33 is to enable the system of this invention to automatically learn and discover new causal relationships not defined by initial expert knowledge from completed project data, thereby achieving self-evolution and iterative enhancement of the core knowledge base.

[0117] The self-learning process of this cross-project causal relationship mining and graph self-learning unit 33 is triggered upon completion of one or more seawall engineering projects. The first step of this process is to construct a historical data matrix for causal mining. The cross-project causal relationship mining and graph self-learning unit 33 extracts full-process data of completed projects or key segments of projects from the system's historical database. Specifically, for each state node in the construction process causal graph (P-CG). The system extracts quantitative characterization values ​​of the project after it reaches a steady state during the construction period (e.g., the expected value of its final confidence distribution). Simultaneously, it extracts outcome indicators characterizing the project's final effectiveness, such as final settlement and durability ratings of key components. The extracted characterization values ​​of each state node and the project outcome indicators together constitute a historical data matrix. The column vectors of the matrix. Each row represents an independent data sample, that is, a completed project or a segment of a project.

[0118] In historical data matrix Once constructed, the cross-project causal relationship mining and graph self-learning unit 33 calls the internally integrated causal discovery algorithm module to process the matrix. The analysis is performed. This algorithm module aims to infer the causal structure between variables from observed data.

[0119] In one specific implementation, the algorithm module can employ a constraint-based causal discovery algorithm, such as the PC algorithm. This algorithm first assumes that all variables (matrices) The edges are fully joined between all columns, and then edges that do not satisfy the causal condition are gradually removed by performing a series of conditional independence tests. For any two variables... and The algorithm will check whether they are in a given subset of other variables. Under the condition that the conditions are independent. If the condition of independence holds, then it is considered that... and There is no direct causal relationship between them, so remove the edges between them.

[0120] Another alternative implementation approach is to employ an algorithm based on a functional causal model, such as the LiNGAM algorithm. This algorithm, assuming the data does not follow a Gaussian distribution, can identify unique, directed linear causal relationships between variables using techniques such as independent component analysis. The output of this algorithm module is a newly discovered set of potential causal influence edges with high-confidence statistical support, denoted as .

[0121] Unit 33, which focuses on cross-project causal relationship mining and graph self-learning, includes a human-computer collaborative knowledge verification and updating mechanism.

[0122] The algorithm outputs a set of potential causal influence edges. The newly discovered causal edges are not directly written to the knowledge base, but are first sent to an expert review interface. This interface visually displays each newly discovered causal edge to domain experts. The key statistical indicators supporting this relationship should be included, such as the p-value of the conditional independence test or the path coefficient.

[0123] Domain experts review these potential causal relationships discovered through data-driven analysis. For causal relationships confirmed by experts to align with the actual logic of the engineering process, the system formally embeds them into the initial domain knowledge base of the construction process causal graph. This embedding process includes adding a new causal influence edge type to the graph structure definition of the knowledge base and using the causal strength value obtained from data analysis as the initial causal influence weight for that edge.

[0124] Through the closed-loop process of constructing a historical data matrix, analyzing causal discovery algorithms, and updating the human-machine collaborative knowledge base, the system of this invention can transform its experience in historical projects into reusable, structured causal knowledge. This new knowledge will be invoked during the P-CG instantiation process of future new projects, thereby enabling the system's knowledge model to be continuously enriched and refined with the accumulation of project experience, achieving long-term self-learning and self-evolution.

[0125] See attached document Figure 8 The integrated visualization and decision support module 40 is connected to the data output end of the intelligent application and evolution module 30. Its function is to transform the abstract data and reasoning results generated by the core data fusion and reasoning module 20 and the intelligent application and evolution module 30 into visualized information that project managers can intuitively understand and interact with, thereby providing decision support. This integrated visualization and decision support module 40 may include a BIM / GIS integrated visualization unit 41 and an interactive causal chain browser unit 42.

[0126] The BIM / GIS fusion visualization unit 41 functions to accurately map the inference results of each state node in the construction process cause-effect graph (P-CG) to the three-dimensional physical space of the project. This BIM / GIS fusion visualization unit 41 first loads the building information model (BM) or geographic information system (GIS) model of the project as the basic scene for three-dimensional visualization. For any state node output by the core data fusion and inference module 20... State trust data, such as its subset of abnormal states. Trust function value The BIM / GIS fusion visualization unit 41 locates the specific engineering component corresponding to the state node in the 3D model by querying the association between the node and the unique identifier of the BIM component established during P-CG instantiation.

[0127] The BIM / GIS fusion visualization unit 41 has a built-in color mapping function that maps the input trust function value. The color is mapped to a specific color value. In one specific implementation, the color mapping function can be set as a continuous gradient color spectrum from green through yellow to red. When the trust function value is low, the corresponding component is rendered in green, representing a safe state; as the trust function value increases and exceeds a preset attention threshold, the color gradually changes to yellow; when the trust function value further increases and exceeds a risk threshold, the color gradually changes to red. In this way, the BIM / GIS fusion visualization unit 41 presents the risk status of different areas of the entire engineering model in real time and intuitively in the form of a heat map, enabling managers to immediately perceive the distribution and severity of risks. Users can interact in the 3D view, for example, by clicking on a component rendered in an abnormal color, to query the specific trust function and plausibility function values ​​of the status nodes associated with that component.

[0128] The interactive causal chain browser unit 42 provides an interpretable and traceable risk warning and fault diagnosis interface. When the risk propagation and causal tracing unit 31 of the intelligent application and evolution module 30 outputs one or more high-confidence causal paths, the interactive causal chain browser unit 42 visualizes and renders the construction process causal graph (P-CG) in the form of a node-link graph in a dedicated graphical user interface.

[0129] In the rendered graph, all event nodes, state nodes, and causal influence edges belonging to the output causal path will be specially highlighted, for example, by changing their color, increasing their line width, or adding glowing effects, so that they stand out from the entire complex graph background.

[0130] The interface is interactive. Users can click on any node in the highlighted causal path using an input device, such as a mouse. This click triggers an information query and display event. An information panel displays detailed information related to the clicked node, including at least: the node's name and unique identifier; a historical graph showing the node's state confidence level over time; and a detailed list of all original evidence contributing to the node's current state. The list of original evidence clearly shows the data source, original observations, and their specific values ​​after being transformed into the Basic Probability Assignment Function (BPA) for each piece of evidence. This provides managers with a complete and traceable chain of evidence, from the final inference result to the most original input data, greatly enhancing the credibility and reliability of decision-making.

[0131] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart management platform system for integrating data throughout the whole process of seawall engineering construction, characterized in that, The application relates to a construction process causal graph construction method and system. The application comprises: a multi-source heterogeneous data acquisition module for acquiring original heterogeneous data from Internet of Things sensors, BIM / GIS models, manual input records and external environment data and processing the original heterogeneous data into standardized information packages; a core data fusion and reasoning module for receiving the standardized information packages, constructing a construction process causal graph containing event nodes, state nodes, process dependency edges and causal influence edges and calculating state node credibility data of the state nodes; the data structure of the construction process causal graph is defined as: event nodes for representing specific processes or actions in the construction process; state nodes for representing engineering quality, safety or environmental states triggered or influenced by the event nodes; process dependency edges for connecting two event nodes with a construction sequence; causal influence edges for connecting event nodes and state nodes or connecting state nodes and state nodes and being attached with a weight representing influence intensity; the core data fusion and reasoning module is internally provided with a multi-source evidence theory fusion engine for converting evidence from different information packages into a plurality of basic probability assignment functions for any state node and fusing the basic probability assignment functions through a Dempster combination rule; an intelligent application and evolution module for receiving the state node credibility data, performing risk analysis on the construction process causal graph and outputting risk early warning and traceability results; the intelligent application and evolution module is internally provided with a construction disturbance event dynamic graph reconstruction unit, and the working process of the unit comprises: after listening to an external construction disturbance event, the event type is matched from a reconstruction rule library, corresponding causal influence edge weight adjustment rules are acquired, the weight in the construction process causal graph is temporarily modified, and after listening to the end of the disturbance event, the weight is restored to the original value; 2. The intelligent management platform system for whole-process data integration of seawall engineering construction according to claim 1, characterized in that, a comprehensive visualization and decision support module for receiving the risk early warning and traceability results and generating a visual decision support interface for project managers. The multi-source heterogeneous data acquisition module comprises: a data acquisition unit provided with interfaces for different data sources and used for collecting original heterogeneous data; 3. The intelligent management platform system for whole-process data integration of seawall engineering construction according to claim 1, characterized in that, a data preprocessing and standardization unit used for performing data cleaning, time-space alignment and format processing on the original heterogeneous data to generate the standardized information packages. The intelligent application and evolution module is internally provided with a risk propagation and causal tracing unit, and the implementation mode of the risk forward propagation is:

4. The intelligent management platform system for whole-process data integration of seawall engineering construction according to claim 3, characterized in that, when the abnormal credibility of any upstream node is monitored to exceed a preset risk threshold, the abnormal credibility of the node is taken as a conduction risk evidence to perform forward probability propagation along the causal influence edges in the construction process causal graph, so that the risk state of a downstream node is updated and early warning is generated. The implementation mode of the root cause backward tracing of the risk propagation and causal tracing unit is: when a terminal state node is confirmed to be abnormal, all causal paths pointing to the node are searched in reverse on the construction process causal graph, a confidence score of each path is calculated, and all potential root cause paths are sorted according to the confidence scores.

5. The intelligent management platform system for whole-process data integration of seawall engineering construction according to claim 1, characterized in that, The intelligent application and evolution module is internally provided with a cross-project causal graph self-learning unit, and the workflow thereof includes: Data is extracted from the completed historical projects to construct an event-state-result matrix, potential causal relationships are calculated and generated by using a causal discovery algorithm, and after being verified by a man-machine collaborative verification of a domain expert, the verified new causal relationships are updated to a domain knowledge base, which is used to guide the initialization construction of a causal graph of a new project construction process.

6. The intelligent management platform system for whole-process data integration of seawall engineering construction according to claim 1, characterized in that, The comprehensive visualization and decision support module is internally provided with a BIM / GIS fusion visualization unit, which is used to associate the core data fusion and the state node reliability data output by the reasoning module with a BIM / GIS three-dimensional space model, and to visually display the model corresponding physical components or regions in the form of a heat map or high-light rendering.

7. The intelligent management platform system for whole-process data integration of seawall engineering construction according to claim 1, characterized in that, The comprehensive visualization and decision support module is internally further provided with an interactive causal chain browser unit, which is used to highlight the path display of the risk early warning and traceability results output by the intelligent application and evolution module on the topological structure of the construction process causal graph, and allows a project manager to click any node to query the detailed evidence source and reliability data thereof.

Images