Road and bridge construction multi-sensor data fusion safety early warning system

Abstract

The invention relates to the technical field of bridge construction safety monitoring, and discloses a road and bridge construction multi-sensor data fusion safety early warning system. The system comprises a data preprocessing and quantum state characterization module, a heterogeneous tensor field construction and quantum correlation analysis module, a quantum topological structure analysis and risk area identification module, a non-equilibrium tensor field evolution and risk early warning module and an early warning output and visualization module. The method comprises the following steps: uniformly mapping multi-source heterogeneous sensor data into a density matrix in a high-dimensional Hilbert space, calculating quantum correlation degree quantity among the data to construct a multi-dimensional heterogeneous tensor field and a quantum correlation topological graph, and identifying a risk area and a propagation path by calculating topological invariants of the topological graph. And the risk critical point is predicted by monitoring entropy generation rate evolution of the tensor field. According to the method, the heterogeneous data can be deeply fused, the weak links of the structure can be identified from the system level, the risk evolution trend can be prospectively pre-judged, and the early-stage and interpretable early warning of the construction risk can be realized.

Description

Technical Field

[0001] This invention relates to the field of bridge construction safety monitoring technology, specifically a multi-sensor data fusion safety early warning system for road and bridge construction. Background Technology

[0002] During the construction of large infrastructure projects such as roads and bridges, various types of sensors are typically deployed at critical structural locations to ensure structural stability and the safety of construction workers, enabling real-time structural health monitoring. These sensors, such as strain gauges, accelerometers, displacement gauges, and image acquisition devices, generate massive amounts of monitoring data reflecting the structural condition.

[0003] Existing safety early warning methods typically rely on analyzing data from single or a few sensors. However, due to the complex environment of construction sites and the varying types, data formats, and sampling frequencies of sensors used, monitoring data exhibits typical multi-source heterogeneous characteristics. Current data processing technologies have limitations in deeply fusing these heterogeneous data, often processing different types of data in isolation or performing simple weighted fusion, making it difficult to effectively reveal the potential and complex nonlinear correlations between different physical measurements.

[0004] This lack of data correlation mining makes it difficult for early warning systems to identify systemic risks from a global perspective. Early signs of risk often do not manifest as a single sensor reading exceeding a threshold, but rather as subtle changes in the correlation patterns between multiple monitoring parameters. Because existing technologies cannot capture these subtle evolutions in system states, they struggle to identify areas where risks converge within a structure and their potential propagation paths.

[0005] Furthermore, most early warning systems rely on fixed thresholds for judgment, triggering an alarm when monitored values ​​exceed preset safety ranges. This is essentially a delayed response mechanism. Such methods lack the ability to predict the evolutionary trends of structures as complex dynamic systems and cannot provide proactive warnings before risks accumulate and approach a critical state, thus limiting their effectiveness in preventing sudden structural safety incidents. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a multi-sensor data fusion safety early warning system for road and bridge construction, which solves the problems of packaging box decay and shortened product shelf life caused by moisture accumulation during storage.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of this invention provides a multi-sensor data fusion safety early warning system for road and bridge construction, the system comprising: The data acquisition module is used to collect various sensor data at the road and bridge construction site. The various sensor data are heterogeneous and contain timestamps and spatial location information. The data preprocessing and quantum state characterization module is used to preprocess and characterize the various types of sensor data, and to map the various types of sensor data into a density matrix in a high-dimensional Hilbert space; The heterogeneous tensor field construction and quantum correlation analysis module is used to calculate the quantum correlation measure between various types of sensor data based on the density matrix in the high-dimensional Hilbert space, and to construct a multidimensional heterogeneous tensor field. Each element of the multidimensional heterogeneous tensor field stores the quantum correlation measure and contains the time, spatial location and sensor type information of various types of sensor data. The quantum topology analysis and risk region identification module is used to construct a quantum correlation topology graph based on the multidimensional heterogeneous tensor field and calculate the topological invariants of the quantum correlation topology graph to identify the risk regions and risk propagation paths of the road and bridge structure. The non-equilibrium tensor field evolution and risk warning module is used to monitor the dynamic evolution of the multidimensional heterogeneous tensor field. The dynamic evolution is characterized by the entropy generation rate and is used to predict the risk evolution trend and identify the risk critical point. The early warning output and visualization module is used to generate and output early warning information based on the risk area, the risk propagation path, and the risk critical point.

[0008] In a specific embodiment, the data preprocessing and quantum state characterization module is specifically used for: performing spatiotemporal alignment and noise suppression on the various types of sensor data; mapping the various types of sensor data to a density matrix in the high-dimensional Hilbert space, wherein the mapping includes: extracting time-frequency features from the numerical or time-series sensor data and encoding them into the density matrix in the high-dimensional Hilbert space for mapping; extracting semantic features from the image or video feature data in the various types of sensor data and mapping the semantic features to a density matrix in the high-dimensional Hilbert space; and applying quantum random walks to the quantum states of the mapped numerical or time-series sensor data for feature enhancement.

[0009] Preferably, the spatiotemporal alignment and noise suppression are achieved by dynamically adjusting the filtering intensity based on the inherent uncertainty of the various types of sensor data, and performing adaptive filtering on the various types of sensor data.

[0010] Preferably, the quantum random walk feature enhancement is specifically used to: abstract the time series of the numerical or time-series sensor data into a graph structure, where the nodes of the graph structure represent the data states at different time steps; and extract the nonlocal correlations and evolution patterns of the numerical or time-series sensor data by simulating the evolution of quantum particles on the graph structure.

[0011] In a specific embodiment, the heterogeneous tensor field construction and quantum correlation analysis module is implemented through the following steps: First, a joint density matrix describing the overall state of the construction site is constructed by performing tensor product operations on the density matrices representing different sensor data; then, based on the joint density matrix, the entanglement entropy or quantum mutual information between any two or more sets of density matrices is calculated, and the calculation result is determined as the quantum correlation metric between the various types of sensor data; finally, the quantum correlation metric and its corresponding time, spatial location, and sensor type information are stored as elements in the corresponding locations, thereby constructing the multidimensional heterogeneous tensor field.

[0012] In a specific embodiment, the quantum topology analysis and risk region identification module is specifically used to: divide the road and bridge structure into multiple spatial regions, and use the spatial regions as nodes to construct the quantum correlation topology graph; extract the quantum correlation metric corresponding to the spatial region from the multidimensional heterogeneous tensor field, and use the extracted quantum correlation metric as the edge weight of the quantum correlation topology graph; calculate the Betti number or persistent cohomology feature of the quantum correlation topology graph, and use the calculated Betti number or persistent cohomology feature as the topological invariant; and based on the analysis of the topological invariant, identify the risk regions of the road and bridge structure and the risk propagation path.

[0013] Preferably, the quantum topology analysis and risk region identification module achieves risk identification by analyzing the topological invariants. The risk identification includes: determining the weak links or risk convergence areas of the road and bridge structure based on the abnormal changes in the Betti number or the short-lifetime characteristics in the continuous cohomology features; and determining the critical paths for the propagation of the risk in the road and bridge structure based on the long-lifetime characteristics in the continuous cohomology features.

[0014] In a specific embodiment, the non-equilibrium tensor field evolution and risk warning module is specifically used to: treat the multidimensional heterogeneous tensor field as a non-equilibrium system; obtain the entropy generation rate by calculating the product of the flux of the quantum correlation metric change and the thermodynamic force within the multidimensional heterogeneous tensor field, and use the entropy generation rate to monitor the dynamic evolution of the multidimensional heterogeneous tensor field; and identify the risk critical point by monitoring the nonlinear accelerated growth of the entropy generation rate, the sudden reconstruction of the correlation mode of the multidimensional heterogeneous tensor field, or the rapid change of the topological invariants.

[0015] Preferably, the non-equilibrium tensor field evolution and risk warning module is further used to: raise the warning level when the risk accumulation occurs in the risk region or the risk propagation path identified by the quantum topology analysis and risk region identification module.

[0016] In a specific embodiment, the early warning output and visualization module is specifically used to: receive the risk region and the risk propagation path identified by the quantum topology analysis and risk region identification module, and the risk critical point identified by the non-equilibrium tensor field evolution and risk early warning module; and generate and output the early warning information based on the received risk region, the risk propagation path, and the risk critical point. The generation and output of the early warning information includes: highlighting the risk region and the risk propagation path on the three-dimensional model or map of the road and bridge structure; and explaining the risk causes of the early warning information by reverse analysis of the changes in the quantum correlation metric, the entropy generation rate, or the topological invariant that lead to the early warning.

[0017] This invention provides a multi-sensor data fusion safety early warning system for road and bridge construction. It has the following beneficial effects: 1. This invention achieves unified and deep fusion of multi-source heterogeneous sensor data from road and bridge construction sites by constructing a unique data processing and representation framework. The system's data preprocessing and quantum state representation modules uniformly map various types and dimensions of sensor data into density matrices in a high-dimensional Hilbert space, and further utilize quantum random walks for feature enhancement. This quantum state-based representation method enables the system to transcend the limitations of traditional data formats, calculating quantum correlation metrics (such as entanglement entropy or quantum mutual information) between various types of sensor data within a unified mathematical framework, thereby capturing deeper nonlinear and nonlocal correlations that are difficult to reveal using traditional statistical methods.

[0018] 2. This invention, by introducing topological data analysis and non-equilibrium system theory, achieves a shift from data monitoring to risk prediction, providing early and forward-looking warning capabilities. The quantum topological structure analysis and risk region identification module in the system, by calculating topological invariants (such as Betti numbers) of the quantum correlation topological graph, can identify weak links or risk convergence areas that already exist in the entire structure, even when sensor readings are still within normal ranges. Simultaneously, the non-equilibrium tensor field evolution and risk warning module, by monitoring the entropy generation rate of the multidimensional heterogeneous tensor field, can capture the dynamic trend of the system's evolution from a stable state to an unstable state, thereby identifying the risk critical point before a qualitative change occurs, achieving early prediction of potential risks.

[0019] 3. This invention provides a global, system-level risk assessment and interpretation mechanism, enhancing the interpretability and decision-making value of early warning information. The system integrates dispersed sensor data into a dynamic network reflecting the overall structural state by constructing a multidimensional heterogeneous tensor field and a quantum correlation topology graph. This enables the system not only to identify risk areas but also to reveal the key paths of risk propagation within the structure by analyzing topological invariants such as persistent cohomology characteristics. Furthermore, the early warning output and visualization module can provide a clear explanation of the risk causes for the generated early warning information through reverse analysis of changes in quantum correlation metrics, entropy generation rates, or topological invariants leading to the early warning, thus providing managers with an intuitive and traceable basis for decision-making. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall system structure of the present invention; Figure 2 This is a schematic diagram of the deployment of the data acquisition module of the present invention; Figure 3 This is a schematic diagram of the functional flow of the data preprocessing and quantum state characterization module of the present invention; Figure 4 This is a schematic diagram of the functional flow of the heterogeneous tensor field construction and quantum correlation analysis module of the present invention; Figure 5 This is a schematic diagram of the functional flow of the quantum topology analysis and risk region identification module of the present invention; Figure 6 This is a schematic diagram of the functional flow of the non-equilibrium tensor field evolution and risk warning module of the present invention. Detailed Implementation

[0021] 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.

[0022] See attached document Figure 1 This invention provides a multi-sensor data fusion safety early warning system for road and bridge construction. The system may include: a data acquisition module, a data preprocessing and quantum state characterization module, a heterogeneous tensor field construction and quantum correlation analysis module, a quantum topological structure analysis and risk area identification module, a non-equilibrium tensor field evolution and risk early warning module, and an early warning output and visualization module.

[0023] In one specific embodiment, the data acquisition module is used to collect data from various sensors deployed at the road and bridge construction site. This sensor data is heterogeneous and includes timestamps and spatial location information. The types of data collected include, but are not limited to, numerical, time-series, image, or video data.

[0024] The data preprocessing and quantum state characterization module is connected to the data acquisition module to receive data from various sensors. This module performs preprocessing operations such as spatiotemporal alignment and noise suppression on the received data, and then maps the preprocessed sensor data to a density matrix ρ in a high-dimensional Hilbert space. This mapping process provides the basic data structure for subsequent unified correlation analysis.

[0025] The heterogeneous tensor field construction and quantum correlation analysis module receives the density matrix output from the previous module. Based on the density matrix, this module calculates the quantum correlation metric between any two or more sets of sensor data, where the quantum correlation metric is entanglement entropy or quantum mutual information. Subsequently, this module constructs a multidimensional heterogeneous tensor field and stores the calculated quantum correlation metric, along with its corresponding time, spatial location, and sensor type information, into the elements of this tensor field.

[0026] The quantum topology analysis and risk region identification module constructs a quantum correlation topology graph based on data stored in a multidimensional heterogeneous tensor field. This module identifies risk regions and risk propagation paths in road and bridge structures by calculating topological invariants of the quantum correlation topology graph, such as Betti numbers or persistent cohomology features.

[0027] The non-equilibrium tensor field evolution and risk early warning module is used to monitor the dynamic process of multidimensional heterogeneous tensor fields changing over time. This module characterizes the degree of dynamic evolution by calculating the entropy generation rate σ within the multidimensional heterogeneous tensor field, and identifies risk critical points based on criteria such as the nonlinear growth of the entropy generation rate, thereby enabling the prediction of risk evolution trends.

[0028] The early warning output and visualization module receives the risk areas, risk propagation paths, and risk thresholds identified by the aforementioned modules. Based on the received information, this module generates and outputs early warning information. The output of the early warning information includes highlighting the risk areas and paths on a 3D model or map of the road and bridge structure, and explaining the causes of the risks through reverse analysis of the data changes that led to the early warning.

[0029] See attached document Figure 2 The data acquisition module's function is to acquire raw monitoring data from the physical world of the road and bridge construction site, providing input for subsequent data fusion and safety early warning. In one specific embodiment, the data acquisition module includes various types of sensors deployed at key locations on the road and bridge structure, which are connected to the data acquisition unit via wired or wireless means.

[0030] Sensor data is heterogeneous, including numerical or time-series sensor data, as well as image or video feature data. Numerical or time-series sensor data is collected by sensors deployed at locations such as the main girder, main tower, piers, or foundation of the bridge. These sensors include, but are not limited to: strain gauges for measuring structural strain, accelerometers for measuring structural vibration characteristics, displacement gauges for measuring structural displacement, temperature sensors for monitoring the heat of hydration of concrete or ambient temperature, and global navigation satellite system receivers for acquiring the overall deformation and spatial position of the structure.

[0031] Image or video feature data is collected by fixed high-definition cameras deployed at the construction site or by inspection cameras mounted on drones. These cameras are used to acquire visual images of the bridge structure, videos of construction activities, and visual information about the surrounding environment.

[0032] All sensor data collected by the data acquisition module is accompanied by timestamps and spatial location information. Timestamps are synchronized via a network time protocol to ensure consistency across all heterogeneous data in the time dimension. Spatial location information consists of pre-calibrated, precise coordinates of each sensor within the three-dimensional coordinate system of the road and bridge structure, corresponding to the coordinate system in the design drawings or building information model.

[0033] In one specific embodiment, the data acquisition module further includes a data transmission network. The analog or digital signals acquired by each sensor are first processed and digitized by the data acquisition unit to which they are connected. Then, the raw data stream with timestamps and spatial location information is transmitted to the central processing system or edge computing node through an optical fiber network or wireless communication network, forming a unified raw dataset to be processed.

[0034] See attached document Figure 3 This module receives the raw heterogeneous dataset with timestamps and spatial location information transmitted by the data acquisition module, and converts it into a unified quantum state data structure required for subsequent analysis.

[0035] This module first performs spatiotemporal alignment and noise suppression on the received sensor data. Spatiotemporal alignment is based on the timestamps attached to each data stream. Through interpolation or resampling, all data are unified to a common time base to ensure that data at the same point in time can be analyzed synchronously. Noise suppression is achieved through adaptive filtering. Specifically, the system calculates the statistical variance or signal-to-noise ratio of each sensor data stream within a preset time window, using this as a measure of the inherent uncertainty of the data. The system dynamically adjusts the filter parameters based on this measure. For example, when the data variance is small, indicating signal stability, a lower cutoff frequency is used; when the data variance is large, indicating signal fluctuations or potential anomalies, a higher cutoff frequency is used, thereby filtering out high-frequency noise while retaining effective signal change information.

[0036] After preprocessing, the module maps various sensor data into a density matrix ρ in a high-dimensional Hilbert space. This mapping process is performed differently depending on the data type.

[0037] For numerical or time-series sensor data, the module first employs time-frequency analysis methods, such as short-time Fourier transform or wavelet transform, to extract features within each time window. These extracted features form a feature vector. After normalization, this feature vector is encoded into a pure quantum state. Among them, c i These are the components of the normalized eigenvector, satisfying |ψ represents a quantum state, |i> represents a ground state of the system, and N represents the total number of ground states.

[0038] Subsequently, the corresponding density matrix ρ is obtained through outer product operation. num =|ψ><ψ|.

[0039] For image or video feature data, the module utilizes a pre-trained deep convolutional neural network model to extract semantic features from image or video frames and outputs a high-dimensional feature vector. Similar to processing numerical data, this semantic feature vector is also normalized and mapped to a corresponding density matrix ρ. img Through this process, raw data from different sources and with different structures are uniformly represented as a density matrix, thus solving the compatibility problem of heterogeneous data.

[0040] To further extract the implicit nonlocal correlations and evolution patterns in time-series sensor data, the module applies quantum random walks to enhance the features of the mapped numerical or time-series sensor data. This process first involves mapping the density matrix sequence {ρ} over a time period. t Abstract this into a graph structure G = (V, E), where the nodes v of the graph are... t ∈V represents the data state ρ at time step t. t In a graph, edges e∈E connect nodes that are adjacent in time.

[0041] Subsequently, the evolution of a quantum particle over discrete time steps is simulated on this graph structure. The evolution is driven by a unitary operator U, and after a specified number of steps, a final probability distribution is obtained. This probability distribution reflects the strength of the nonlocal correlation between the initial state and all other nodes in the graph. Finally, this probability distribution information is used to update or reconstruct the original density matrix ρ. num This generates a density matrix ρ′ that enhances the features. num Then output it to the next module.

[0042] See attached document Figure 4 This module receives a set of density matrices representing the data states of each sensor, output by the data preprocessing and quantum state characterization module. Its core function is to quantify the correlation strength between different sensor data and construct a structured multidimensional data field from these correlation information.

[0043] In one specific embodiment, the module first processes the density matrix {ρ1,ρ2,…,ρ...} representing different sensor data. N Perform tensor product operations to construct a joint density matrix ρ that describes the overall state of the construction site. joint This operation is represented as: Where, ρ joint The joint density matrix of the system describes the joint state of multiple quantum systems, ρ. i Let the density matrix of the i-th system be denoted as . Represents the tensor product, used to combine the states of two or more quantum systems. This represents the set of tensor products of all system density matrices from i=1 to N, generating the overall joint density matrix, where N is the total number of sensors.

[0044] Subsequently, the module calculates a quantum correlation metric between any two or more sets of sensor data based on the joint density matrix. In one specific embodiment, quantum mutual information is used as this metric. For a composite system ρ consisting of any two subsystems (e.g., sensor A and sensor B), AB The calculation steps for the quantum mutual information I(A:B) are as follows: First, calculate the von Neumann entropy S(ρ), which is given by the formula: S(ρ)=-Tr(ρlog2ρ); Where S(ρ) represents the entropy of the quantum state ρ, Tr represents the trace of the matrix, and ρlog2ρ is the core part of the von Neumann entropy, representing the uncertainty of the quantum state.

[0045] Then, by analyzing the density matrix ρ of the composite system AB Perform partial trace calculations to obtain the reduced density matrix ρ of each subsystem. A =Tr B (ρ AB ) and ρ B =Tr A (ρ AB Finally, the quantum mutual information I(A:B) is calculated: I(A:B)=S(ρ A )+S(ρ B )-S(ρ AB ); Where I(A:B) represents the quantum mutual information between system A and system B, ρ A and ρ B These are the density matrices of system A and system B, respectively, ρ AB This represents the joint density matrix when systems A and B are combined. By calculating the entropy of the joint density matrix, the amount of information shared by A and B can be obtained.

[0046] The calculation result of the quantum mutual information I(A:B) quantifies the total correlation between sensor A and sensor B, including classical correlation and quantum correlation.

[0047] Finally, the module stores the calculated quantum correlation metric and its corresponding time, spatial location, and sensor type information as elements in the corresponding locations, thereby constructing a multidimensional heterogeneous tensor field. This tensor field is a high-dimensional array indexed by time, spatial coordinates, and sensor type. Specifically, an element in the tensor field... What is stored is the spatial location at time t. The type is s i The sensor and its location in space The type is s j The quantum mutual information values ​​between sensors are used to transform discrete, multi-source sensor data into a unified, spatiotemporally continuous field that reflects the internal interconnected structure of the system. This constructed multidimensional heterogeneous tensor field is then transmitted to the quantum topology analysis and risk region identification module.

[0048] See attached document Figure 5 This module receives a multidimensional heterogeneous tensor field generated by the heterogeneous tensor field construction and quantum correlation analysis module. Its function is to identify potential risk areas in road and bridge structures and the propagation path of risks within the structure through topological data analysis methods.

[0049] In one specific embodiment, the module first divides the physical structure of the road and bridge into multiple discrete spatial regions in a digital model. This division can be achieved using spatial mesh generation or finite element mesh generation methods. These discrete spatial regions are used as nodes to construct a quantum interconnected topology graph.

[0050] Subsequently, the module extracts the quantum correlation metric between any two spatial region nodes from the multidimensional heterogeneous tensor field. This extracted quantum correlation metric is assigned as the weight of the edge connecting the corresponding two nodes. Thus, a weighted undirected graph with spatial regions as nodes and the quantum correlation metric between regions as edge weights is constructed.

[0051] Next, the module computes topological invariants based on the weighted graph. In one embodiment, a continuous homology method is used for computation. This process first filters the graph according to the edge weights to construct a serialized simplicial complex. For example, a threshold ∈ is set, including all edges with weights higher than ∈ and their connected nodes in the simplicial complex. By continuously changing the threshold ∈ from high to low, a nested sequence of simplicial complexes is obtained.

[0052] During this filtering process, the system continuously tracks the evolution of topological features, specifically including the birth and death of zero-dimensional features (connected components), one-dimensional features (cycles or loops), and two-dimensional features (holes or cavities). The lifecycle of each topological feature, i.e., the threshold range from its birth to its death, is recorded. This lifecycle information can be represented as a persistent graph or barcode. Simultaneously, at specific thresholds, the Betti numbers b0, b1, and b2 of the simplex complex are also calculated, representing the number of connected components, independent cycles, and holes at that threshold, respectively.

[0053] Finally, based on the analysis of topological invariants, the module identifies risk areas and risk propagation paths in the road and bridge structures. The specific criteria for risk identification are as follows: based on the short-lifecycle characteristic in the continuous homology features, weak links or risk convergence areas in the road and bridge structures are determined. A short-lifecycle topological feature indicates a correlation structure that exists only within a very small correlation metric range, corresponding to unstable or localized correlation patterns in the structure. Simultaneously, anomalous changes in the Betti number, such as a sharp increase in b0, indicate a split in the overall correlation network and are also used to identify risk areas.

[0054] Based on the long lifecycle characteristic of persistent cohomology features, the critical paths for risk propagation in road and bridge structures are identified. A long lifecycle topological feature represents a macroscopically correlated structure that is stable across a wide range of correlation metrics. This stable, strongly correlated path is identified as the main channel for risk or stress propagation in the structure.

[0055] The module ultimately outputs a set of spatial nodes identifying risk areas, as well as a set of nodes and edges constituting risk propagation paths. This data is then transmitted to the non-equilibrium tensor field evolution and risk early warning module and the early warning output and visualization module.

[0056] See attached document Figure 6 This module receives a sequence of multidimensional heterogeneous tensor fields that evolve over time, output by the heterogeneous tensor field construction and quantum correlation analysis module. Its function is to monitor the dynamic evolution of the field theory system, thereby predicting the risk evolution trend and identifying the impending risk critical point.

[0057] In one specific embodiment, the module treats the time-evolving multidimensional heterogeneous tensor field as a non-equilibrium thermodynamic system. In this system, the continuous change of quantum correlation metrics in time and space is considered a macroscopic manifestation of irreversible processes within the system. To quantify the drastic nature of the system's state evolution, the module introduces and calculates the entropy generation rate σ.

[0058] The entropy production rate σ is calculated based on non-equilibrium thermodynamics, and is expressed as the sum of the products of thermodynamic flows and thermodynamic forces of all irreversible processes within the system. Specifically, its calculation formula is: σ=∑ k J k X k ; Where, ∑ k J k J represents the summation over index k. k It is a thermodynamic flux, defined as the rate of change of the quantum correlation metric X at a specific location or on a specific correlation pair k in a tensor field over time. kIt is the corresponding thermodynamic force, defined as the gradient of the quantum correlation metric in space, where σ represents the symbol of a physical quantity or operator.

[0059] This module identifies risk thresholds by continuously monitoring the entropy generation rate σ and other relevant indicators. The criteria for identifying risk thresholds include one or more of the following: monitoring a non-linear accelerating growth trend in the total entropy generation rate of the system over time; monitoring a sudden, global reconstruction of the distribution pattern of highly correlated metric regions in a multidimensional heterogeneous tensor field; or monitoring a rapid change over time in topological invariants (such as the Betti number) calculated by the quantum topological structure analysis and risk region identification module, with the rate of change exceeding a preset threshold.

[0060] Furthermore, this module is used to adjust the warning level. The module receives spatial location information of risk areas or risk propagation paths determined by the quantum topology analysis and risk area identification module. Simultaneously, the module itself calculates spatial regions with high values ​​in the entropy generation rate σ field. By performing overlap analysis on these two sets of spatial region data, when the calculated area or proportion of the overlapping region exceeds a preset judgment threshold, it indicates that the dynamic accumulation process of risk is occurring at a critical weak point in the structure, and the system accordingly raises the output warning level.

[0061] The module's final output includes: the time and spatial location information of the identified risk thresholds, and the dynamically adjusted warning level. This information is then transmitted to the warning output and visualization module to generate the final warning command.

[0062] The early warning output and visualization module is the final output of the entire early warning system. Its function is to integrate the analysis results of upstream modules and present the early warning information to users in an intuitive and interpretable way.

[0063] In one specific embodiment, the module is equipped with a specific data interface for receiving risk region and risk propagation path data output by the quantum topology analysis and risk region identification module, as well as risk critical points and warning level information output by the non-equilibrium tensor field evolution and risk warning module. The received risk region and path data includes a precise set of spatial coordinates or node numbers.

[0064] Based on the received risk area, risk propagation path, and risk threshold, this module generates and outputs early warning information. The generation and output of early warning information includes the following two aspects: First, on the 3D model or map of the road and bridge structure, risk areas and risk propagation paths are highlighted. This module internally loads or externally links to a Building Information Model (BIM) or 3D Geographic Information Model corresponding to the road and bridge structure. The module maps the received spatial coordinates of the risk areas onto this 3D model and, through graphics rendering technology, changes the color of the corresponding parts in the model to a preset warning color (e.g., red or yellow), or applies visual effects such as flashing or highlighting. For risk propagation paths, the module renders vector arrows or dynamic light strips with directional indicators connecting the corresponding area nodes on the 3D model.

[0065] Secondly, by reverse-engineering changes in quantum correlation metrics, entropy generation rates, or topological invariants that lead to warnings, the module explains the risk causes of the warning information. When a warning event is triggered, the module automatically initiates a query request to the historical database or upstream analysis module, containing the time and spatial information that triggered the warning. Based on this request, the system traces back and locates the specific original data changes that led to the warning, such as: which two or more sensors exceeded the quantum mutual information value threshold, which spatial region experienced a non-linear increase in entropy generation rate, or which abnormal topological feature appeared in the lifecycle. The module integrates these identified root causes into a structured text report, which is presented on the user interface along with the visualization results, thus providing traceable data evidence for warnings.

Claims

1. A multi-sensor data fusion safety early warning system for road and bridge construction, characterized in that, include: The data acquisition module is used to collect various sensor data at the road and bridge construction site. The various sensor data are heterogeneous and contain timestamps and spatial location information. The data preprocessing and quantum state characterization module is used to preprocess and characterize the various types of sensor data, and to map the various types of sensor data into a density matrix in a high-dimensional Hilbert space; The heterogeneous tensor field construction and quantum correlation analysis module is used to calculate the quantum correlation measure between various types of sensor data based on the density matrix in the high-dimensional Hilbert space, and to construct a multidimensional heterogeneous tensor field. Each element of the multidimensional heterogeneous tensor field stores the quantum correlation measure and contains the time, spatial location and sensor type information of various types of sensor data. The quantum topology analysis and risk region identification module is used to construct a quantum correlation topology graph based on the multidimensional heterogeneous tensor field and calculate the topological invariants of the quantum correlation topology graph to identify the risk regions and risk propagation paths of the road and bridge structure. The non-equilibrium tensor field evolution and risk warning module is used to monitor the dynamic evolution of the multidimensional heterogeneous tensor field. The dynamic evolution is characterized by the entropy generation rate and is used to predict the risk evolution trend and identify the risk critical point. The early warning output and visualization module is used to generate and output early warning information based on the risk area, the risk propagation path, and the risk critical point.

2. The multi-sensor data fusion safety early warning system for road and bridge construction according to claim 1, characterized in that, The data preprocessing and quantum state characterization module is specifically used for: Spatiotemporal alignment and noise suppression are performed on the data from the various sensors. The various sensor data are mapped to a density matrix in the high-dimensional Hilbert space, and the mapping includes: Numerical or time-series sensor data from the various types of sensor data are mapped by extracting time-frequency features and encoding them into a density matrix in the high-dimensional Hilbert space. Extract semantic features from the image or video feature data of the various types of sensor data, and map the semantic features into a density matrix in the high-dimensional Hilbert space; Quantum random walks are applied to the quantum states of the mapped numerical or temporal sensor data to enhance their features.

3. The multi-sensor data fusion safety early warning system for road and bridge construction according to claim 2, characterized in that, The spatiotemporal alignment and noise suppression are achieved in the following ways: Based on the inherent uncertainty of the various types of sensor data, the filtering intensity is dynamically adjusted, and adaptive filtering is performed on the various types of sensor data.

4. The multi-sensor data fusion safety early warning system for road and bridge construction according to claim 2, characterized in that, The quantum random walk feature enhancement is specifically used for: The time series of the numerical or time-series sensor data is abstracted into a graph structure, where the nodes of the graph structure represent the data state at different time steps. By simulating the evolution of quantum particles on the graph structure, the nonlocal correlations and evolution patterns of the numerical or temporal sensor data are extracted.

5. A multi-sensor data fusion safety early warning system for road and bridge construction according to claim 1, characterized in that, The heterogeneous tensor field construction and quantum correlation analysis module is implemented through the following steps: First, by performing tensor product operations on the density matrices representing different sensor data, a joint density matrix describing the overall state of the construction site is constructed. Then, based on the joint density matrix, the entanglement entropy or quantum mutual information between any two or more sets of the density matrices is calculated, and the calculation result is determined as the quantum correlation measure between the various types of sensor data; Finally, the quantum correlation metric and its corresponding time, spatial location and sensor type information are stored as elements in the corresponding locations to construct the multidimensional heterogeneous tensor field.

6. The multi-sensor data fusion safety early warning system for road and bridge construction according to claim 1, characterized in that, The quantum topology analysis and risk region identification module is specifically used for: The road and bridge structure is divided into multiple spatial regions, and the spatial regions are used as nodes to construct the quantum interconnection topology graph; From the multidimensional heterogeneous tensor field, the quantum correlation metric corresponding to the spatial region is extracted, and the extracted quantum correlation metric is used as the edge weight of the quantum correlation topology graph; Calculate the Betti number or persistent cohomology feature of the quantum correlation topology graph, and use the calculated Betti number or persistent cohomology feature as the topological invariant; Based on the analysis of the topological invariants, the risk areas of the road and bridge structure and the risk propagation paths are identified.

7. A multi-sensor data fusion safety early warning system for road and bridge construction according to claim 6, characterized in that, The quantum topology analysis and risk region identification module identifies risks by analyzing the topological invariants. The risk identification includes: Based on the abnormal changes in the Betty number or the short life cycle characteristic in the continuous homology feature, the weak links or risk convergence areas of the road and bridge structure are identified. Based on the long lifecycle characteristic in the continuous coherence feature, the critical path for the propagation of the risk in the road and bridge structure is determined.

8. A multi-sensor data fusion safety early warning system for road and bridge construction according to claim 1, characterized in that, The non-equilibrium tensor field evolution and risk warning module is specifically used for: The multidimensional heterogeneous tensor field is considered as a non-equilibrium system. The entropy generation rate is obtained by calculating the product of the flux of the quantum correlation metric change and the thermodynamic force within the multidimensional heterogeneous tensor field, and the entropy generation rate is used to monitor the dynamic evolution of the multidimensional heterogeneous tensor field. The risk threshold is identified by monitoring the nonlinear accelerated growth of the entropy generation rate, the sudden reconstruction of the correlation mode of the multidimensional heterogeneous tensor field, or the rapid change of the topological invariants.

9. A multi-sensor data fusion safety early warning system for road and bridge construction according to claim 1, characterized in that, The warning output and visualization module is specifically used for: Receive the risk region and the risk propagation path identified by the quantum topology analysis and risk region identification module, and the risk critical point identified by the non-equilibrium tensor field evolution and risk warning module; Based on the received risk area, risk propagation path, and risk threshold, the warning information is generated and output, and the generation and output of the warning information includes: On the three-dimensional model or map of the road and bridge structure, the risk area and the risk propagation path are highlighted; By reverse analysis of the changes in the quantum correlation metric, the entropy generation rate, or the topological invariant that cause the warning, the risk causes of the warning information can be explained.

10. A multi-sensor data fusion safety early warning system for road and bridge construction according to claim 1, characterized in that, The non-equilibrium tensor field evolution and risk warning module is further used for: When the risk accumulation occurs in the risk area or the risk propagation path identified by the quantum topology analysis and risk area identification module, the warning level is raised.

Images