Multimodal data-driven engineering progress risk intelligent assessment method and system

Abstract

The application relates to the technical field of engineering progress management, and provides a multi-modal data-driven intelligent engineering progress risk assessment method and system. Multi-modal engineering information is mapped into a three-order tensor, a tensor slice of the three-order tensor is generated as a space-time tensor based on a preset projection function; a time-varying graph composed of engineering tasks and execution relationships is constructed based on the engineering information, task features are extracted from the space-time tensor, edge weights are generated based on the task features, the node relationship in the time-varying graph is adjusted through the edge weights, and a dynamic graph structure is output; based on the space-time tensor and the dynamic graph structure, a risk equation of the engineering progress in the space-time dimension is generated, the risk equation is solved through a differential equation, and a risk term of the engineering progress is output; and based on the risk term and actual progress data in the engineering information, an engineering progress risk assessment file is generated. The risk assessment error is reduced, the efficiency of the engineering progress risk assessment is improved, and engineering decision-making is supported to change from passive response to active intervention.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engineering progress management, in particular to a multi-modal data driven engineering progress risk intelligent evaluation method and system. BACKGROUND

[0002] Modern engineering projects, such as large infrastructure and industrial manufacturing, involve multidisciplinary collaboration, multi-stage tasks and dynamic environmental changes. Traditional single data source analysis cannot capture multi-dimensional risk factors. With the popularization of industrial Internet of Things, sensor networks and unmanned aerial vehicle inspection technologies, the data collection capability of engineering sites has been significantly improved. Multi-modal data such as video, sensor data and text logs provide more abundant information sources for risk assessment, but intelligent methods are needed to achieve efficient integration and analysis.

[0003] The prior art needs to process large-scale operations by fusing multi-source data, which requires high computing resources. The model structure needs to be optimized or edge computing and distributed computing technologies need to be used to improve efficiency. At the same time, structured and unstructured data are processed separately, which makes it difficult to capture cross-modal correlations, so it is not possible to dynamically capture engineering progress changes and risks, resulting in low efficiency of engineering progress risk assessment. SUMMARY

[0004] The present application provides a multi-modal data driven engineering progress risk intelligent evaluation method and system, which can at least partially solve the problem of low efficiency of engineering progress risk assessment.

[0005] Other characteristics and advantages of the present application will become apparent from the following detailed description, or will be learned by practice of the present application.

[0006] According to one aspect of the present application, a multi-modal data driven engineering progress risk intelligent evaluation method is provided, comprising: acquiring multi-modal engineering information, the engineering information including structured data, unstructured data and time series data; mapping the engineering information into a three-order tensor, generating a tensor slice of the three-order tensor based on a preset projection function, as a spatio-temporal tensor; wherein the three-order tensor includes an engineering entity dimension, a time dimension and a feature dimension; constructing a time-varying graph composed of engineering tasks and execution relationships based on engineering information, extracting task features from the spatio-temporal tensor, generating edge weights based on the task features, adjusting node relationships in the time-varying graph through the edge weights, and outputting a dynamic graph structure; based on the spatio-temporal tensor and the dynamic graph structure, generating a risk equation of engineering progress in the spatio-temporal dimension, and solving the risk equation through a differential equation, and outputting a risk term of engineering progress; generating an engineering progress risk assessment file according to the risk term and actual progress data in the engineering information.

[0007] In the present application, based on the foregoing scheme, the mapping of the engineering information into a three-order tensor, generating a tensor slice of the three-order tensor as a space-time tensor based on a preset projection function, comprises: extracting first information belonging to entity dimension, time dimension and feature dimension in the engineering information, and mapping the first information into a three-order tensor; determining attention weights of each modality based on a learnable weight matrix; generating a tensor slice of the three-order tensor as a space-time tensor based on the attention weights of each modality and a preset projection function.

[0008] In the present application, based on the foregoing scheme, the determination of the attention weights of each modality based on the learnable weight matrix comprises: obtaining the original input feature of the mth modality Performing spatial dimension compression to generate compressed data; determining the attention weight corresponding to the mth modality based on the compressed data and the learnable weight matrix of the mth modality

[0009]

[0010] wherein, represents the original input feature of the mth modality, represents the compressed data, k and m represent the identification of the modality, and M represents the total number of the modality, represents the learnable weight matrix of the kth modality.

[0011] In the present application, based on the foregoing scheme, the generation of the tensor slice of the three-order tensor as a space-time tensor based on the attention weights of each modality and the preset projection function comprises: mapping the original input feature of the mth modality in the engineering entity dimension n and the time dimension t to a unified dimension to generate dimension data; generating a tensor slice of the three-order tensor as a space-time tensor based on the attention weights of each modality and the dimension data

[0012]

[0013] wherein, represents the original input feature of the mth modality in the engineering entity n and the time window t, represents the dimension data, represents an activation function.

[0014] ​​​In the present application, based on the foregoing scheme, the time-varying graph composed of engineering tasks and execution relationships is constructed based on engineering information, task features are extracted from the spatio-temporal tensor, edge weights are generated based on the task features, node relationships in the time-varying graph are adjusted through the edge weights, and a dynamic graph structure is output. The method comprises the following steps: obtaining the engineering tasks and execution relationships in the engineering information, simulating nodes, edges and edge attributes based on the engineering tasks and execution relationships, and constructing a time-varying graph; extracting task features from the spatio-temporal tensor, and generating edge weights based on the task features; adjusting the node relationships in the time-varying graph through the edge weights, and outputting a dynamic graph structure.

[0015] In the present application, based on the foregoing scheme, the risk equation of engineering progress in the spatio-temporal dimension is generated based on the spatio-temporal tensor and the dynamic graph structure, and the risk equation is solved by a differential equation to output the risk term of engineering progress. The method comprises the following steps: determining a diffusion term of risk through the dynamic graph structure based on the dynamic graph structure; determining a driving term of risk caused by multi-modal feature changes based on the spatio-temporal tensor; determining a risk equation of engineering progress risk in the spatio-temporal dimension based on the diffusion term and the driving term; and solving the risk equation by a differential equation to output the risk term of engineering progress.

[0016] In the present application, based on the foregoing scheme, the diffusion term of risk through the dynamic graph structure is determined based on the dynamic graph structure, which comprises the following steps:

[0017]

[0018] wherein, respectively represent the risk probability of task node i and task node j, represents the edge weight between task node i and task node j, represents the neighbor set of task node i.

[0019] According to one aspect of the present application, a multi-modal data driven intelligent engineering progress risk evaluation system is provided, which comprises:

[0020] An acquisition unit is configured to acquire multi-modal engineering information, wherein the engineering information comprises structured data, unstructured data and time series data;

[0021] A tensor unit is configured to map the engineering information to a third-order tensor, generate a tensor slice of the third-order tensor as a spatio-temporal tensor based on a preset projection function; wherein the third-order tensor comprises an engineering entity dimension, a time dimension and a feature dimension.

[0022] ​a structure unit configured to construct a time-varying graph composed of engineering tasks and execution relationships based on the engineering information, extract task features from the spatio-temporal tensor, generate edge weights based on the task features, adjust node relationships in the time-varying graph through the edge weights, and output a dynamic graph structure;

[0023] a risk unit configured to generate a risk equation of engineering progress in a spatio-temporal dimension based on the spatio-temporal tensor and the dynamic graph structure, solve the risk equation through a differential equation, and output a risk term of the engineering progress;

[0024] an evaluation unit configured to generate an engineering progress risk evaluation file according to the risk term and actual progress data in the engineering information.

[0025] According to an aspect of the present application, a computer readable medium having a computer program stored thereon is provided, the computer program being executed by a processor to implement the multi-modal data driven engineering progress risk intelligent evaluation method as described in the above embodiments.

[0026] According to an aspect of the present application, an electronic device is provided, comprising: one or more processors; a storage device configured to store one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the multi-modal data driven engineering progress risk intelligent evaluation method as described in the above embodiments.

[0027] According to an aspect of the present application, a computer program product or computer program is provided, the computer program product or computer program comprising computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to cause the computer device to perform the multi-modal data driven engineering progress risk intelligent evaluation method provided in the various optional implementation manners described above.

[0028] In the technical solution of the present application, multi-modal data is uniformly coded by a three-order tensor, structured, unstructured and time sequence features are dynamically weighted and fused by using an attention mechanism, and the problem of traditional data fragmentation is broken through. The dynamic graph structure adaptively updates edge weights based on spatio-temporal convolution, captures task dependency and resource flow changes in real time, and solves the lag of static models. The risk equation fuses a graph diffusion term (task network conduction), a tensor gradient term (multi-modal feature driven) and a nonlinear interaction term (burst event amplification effect), and realizes accurate modeling of spatio-temporal evolution of risk. Compared with traditional methods, the dynamic fusion and mechanism driven characteristics can reduce risk evaluation errors and improve the efficiency of engineering progress risk evaluation, supporting engineering decision-making to shift from passive response to active intervention.

[0029] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application, as claimed. BRIEF DESCRIPTION OF DRAWINGS

[0030] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and serve to explain the principles of the application. It is readily apparent to one skilled in the art that the following description is merely exemplary and explanatory of the application and that various embodiments of the application can be made without departing from the spirit and scope of the application.

[0031] Figure 1 A flow chart of a multi-modal data-driven engineering progress risk intelligent assessment method in an embodiment of the application is schematically shown.

[0032] Figure 2 A flow chart of generating a spatio-temporal tensor in an embodiment of the application is schematically shown.

[0033] Figure 3 A schematic diagram of a multi-modal data-driven engineering progress risk intelligent assessment system in an embodiment of the application is schematically shown.

[0034] Figure 4 A structural schematic diagram of a computer system of an electronic device suitable for implementing embodiments of the application is shown. DETAILED DESCRIPTION

[0035] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations may, however, be implemented in many different forms and should not be construed as limited to the implementations set forth herein; rather, these implementations are provided as non-limiting examples so that this disclosure will fully convey its scope to those skilled in the art. Like reference numerals refer to like elements throughout.

[0036] Additionally, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the application. One skilled in the relevant art will recognize, however, that the

[0037] The block diagrams shown in the drawings are merely functional entities, and do not necessarily correspond to physically independent entities. That is, the functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0038] The flowcharts shown in the drawings are merely exemplary illustrations, and do not necessarily include all contents and operations / steps, nor are they necessarily executed in the order described. For example, some operations / steps can be further decomposed, and some operations / steps can be combined or partially combined, so the actual execution order can be changed according to actual conditions.

[0039] The implementation details of the technical solutions of the present application are described in detail as follows:

[0040] Figure 1 A flowchart of a multi-modal data-driven engineering progress risk intelligent assessment method according to an embodiment of the present application is shown. Referring to Figure 1 The multi-modal data-driven engineering progress risk intelligent assessment method includes at least steps S110 to S150, which are described in detail as follows:

[0041] S110, obtaining multi-modal engineering information, the engineering information including structured data, unstructured data, and time series data.

[0042] In an embodiment of the present application, multi-modal engineering information is obtained, wherein the engineering information includes structured data, unstructured data, and time series data. Based on the obtained engineering industry big data, an industrial cloud platform based on industrial cloud computing is constructed, and an engineering progress risk identification system based on industrial networks is generated in combination with an industrial internet platform architecture.

[0043] Specifically, in the present embodiment, the structured data includes progress plan information and resource allocation information, wherein the progress plan information includes table data such as Gantt charts for time scheduling and dependency relationships of engineering tasks. The resource allocation information includes quantitative data such as matrices or database tables for resource allocation of manpower, equipment, and materials.

[0044] Specifically, in the present embodiment, the unstructured data includes field images and text reports, wherein the field images include visual data such as pictures or video frames taken by a camera for construction scene. The text reports include natural language descriptions such as construction logs and inspection reports.

[0045] Specifically, in the present embodiment, the time series data includes physical quantity data such as temperature, vibration, and displacement time series signals collected in real time by sensors.

[0046] S120, map the engineering information into a three-order tensor, generate a tensor slice of the three-order tensor as a space-time tensor based on a preset projection function; wherein the three-order tensor comprises an engineering entity dimension, a time dimension, and a feature dimension.

[0047] As shown in the embodiment of the present application, the engineering information is mapped into a three-order tensor, and a tensor slice of the three-order tensor is generated as a space-time tensor based on a preset projection function, comprising: Figure 2

[0048] S210, extract first information belonging to the entity dimension, the time dimension, and the feature dimension in the engineering information, and map the first information into a three-order tensor;

[0049] S220, determine attention weights of each modality based on a learnable weight matrix;

[0050] S230, generate a tensor slice of the three-order tensor as a space-time tensor based on the attention weights of each modality and a preset projection function.

[0051] In an embodiment of the present application, the multi-modal data is mapped into a unified mathematical representation to form a three-order tensor, which includes an engineering entity dimension, a time dimension, and a feature dimension. The first information belonging to the entity dimension, the time dimension, and the feature dimension in the engineering information is extracted and mapped into a three-order tensor.

[0052] Specifically, in the engineering entity dimension, such as a task or a process node, for example, a specific task such as pouring concrete or installing steel structure. The time dimension is divided according to fixed intervals (such as days / weeks) to capture dynamic changes. The multi-modal feature dimension integrates the features of each data source, such as progress features, image features, and text semantics. Through tensorization, heterogeneous data such as tables, images, texts, and time series are encoded into a unified three-dimensional structure as a three-order tensor to solve the problem of multi-modal data alignment difficulty in traditional methods.

[0053] In an embodiment of the present application, the attention weights of each modality are determined based on a learnable weight matrix, comprising:

[0054] The original input features of the mth modality are spatially compressed to generate compressed data;

[0055] Based on the compressed data and the learnable weight matrix of the mth modality , the attention weight corresponding to the mth modality is determined as:

[0056]

[0057] ​wherein, represents the original input feature of the mth modality, represents compressed data, k and m represent the identification of the modality, M represents the total number of modalities, and T represents the transpose operation of the learnable weight matrix, represents the learnable weight matrix of the kth modality.

[0058] The importance of each modality is automatically learned through the attention mechanism, for example: in the early stage of construction, the structured data of the progress plan modality has a higher weight; in the event of an emergency, the weight of the time series data of the sensor modality is significantly improved.

[0059] In an embodiment of the present application, based on the attention weight of each modality and the preset projection function, a tensor slice of the third-order tensor is generated as a spatiotemporal tensor, comprising:

[0060] The original input feature of the mth modality in the engineering entity dimension n and the time dimension t is mapped to a unified dimension to generate dimension data;

[0061] Based on the attention weight of each modality and the dimension data, a tensor slice of the third-order tensor is generated as a spatiotemporal tensor :

[0062]

[0063] wherein, represents the original input feature of the mth modality in the engineering entity n and the time window t, represents the dimension data, represents an activation function.

[0064] The above process maps structured data (such as progress plan), unstructured data (such as field image, text report) and time series data (such as sensor monitoring) to a third-order tensor, maps heterogeneous data to a unified space through a projection function, solves the problem of mismatch between multi-modal feature dimension and semantics, and realizes the unified representation of heterogeneous data. At the same time, through the activation function, the fitting ability of the model to complex features is enhanced, and the contribution degree of the modality is dynamically adjusted combined with the attention weight. For example, capture the synergistic effect of progress delay and resource shortage, in the early stage of construction, the weight of the progress plan modality is higher; in the event of an emergency, the weight of the sensor data modality is improved. Through the above process, the multi-modal data is uniformly encoded into a spatiotemporal tensor, solving the problem of multi-modal data alignment difficulty and information fragmentation in traditional methods, significantly improving the precision and flexibility of data fusion, and providing a basis for subsequent dynamic graph construction and risk propagation modeling.

[0065] S130, based on the engineering information, constructing a time-varying graph composed of engineering tasks and execution relationships, extracting task features from the space-time tensor, generating edge weights based on the task features, adjusting the node relationship in the time-varying graph through the edge weights, and outputting a dynamic graph structure.

[0066] In an embodiment of the present application, based on the engineering information, a time-varying graph composed of engineering tasks and execution relationships is constructed, task features are extracted from the space-time tensor, edge weights are generated based on the task features, the node relationship in the time-varying graph is adjusted through the edge weights, and a dynamic graph structure is outputted, comprising:

[0067] Obtaining the engineering tasks and execution relationships in the engineering information, simulating nodes, edges and edge attributes based on the engineering tasks and execution relationships, and constructing a time-varying graph;

[0068] Task features are extracted from the space-time tensor, and edge weights are generated based on the task features;

[0069] The node relationship in the time-varying graph is adjusted through the edge weights, and a dynamic graph structure is outputted.

[0070] In an embodiment of the present application, the engineering tasks and execution relationships in the engineering information are obtained. Among them, the execution relationship of the engineering task describes the logical order and dependence between tasks, such as task A must be started after task B is completed, which is usually represented as a directed edge, such as A→B. For example, in a construction project, the foundation pouring depends on the excavation of the foundation pit, forming a dependent edge.

[0071] In addition, the resource flow network can also be obtained from the engineering information to describe the transfer path of resources (manpower, equipment, materials) between different tasks. For example, the crane moves from task X to task Y; the scheduling relationship between the concrete pump truck and the tasks of pouring floor and pouring beam column.

[0072] Based on the above information obtained, a time-varying graph, i.e. an industrial knowledge graph, is constructed. It includes a node set, a time-dimension edge set and edge attributes. Specifically, each node in the node set represents an engineering task, such as installing pipes; the time-dimension edge set includes two types of edges: dependent edges based on task logical relationship, and resource flow edges based on resource transfer path; the edge attributes include dependent strength and resource flow amount, etc.

[0073] Task features are extracted from the space-time tensor, mapped to a unified space through a projection matrix, the cross-task similarity is calculated and normalized to a weight value, and the edge weight is generated. At the same time, according to the current graph state and the space-time feature, the incremental change of the edge weight is predicted, the adjustment range is controlled by combining the evolution rate parameter, the node relationship in the time-varying graph is adjusted through the edge weight, and a dynamic graph structure is outputted, realizing the progressive update of the graph structure.

[0074] The process outputs a dynamic graph structure, each time window corresponds to an updated graph structure, dynamically reflecting the changes of task dependency and resource flow. The technical effect is that the adaptive adjustment of edge weight can accurately capture the abnormality in engineering progress, such as resource conflict weakening the flow edge weight, while integrating task logic and physical constraints to reveal the conduction path of risk along the high weight edge, providing structured support for subsequent risk item modeling.

[0075] The above process constructs a time-varying graph based on engineering task dependency relationship and resource flow network, calculates task feature similarity through spatio-temporal convolution, and dynamically updates edge weight. For example, device failure leads to reduced resource flow edge weight, and the risk conduction path of dependent tasks is adjusted accordingly. This dynamic graph structure can capture the topological changes in engineering progress in real time (such as the addition of emergency tasks, resource conflicts), breaking through the limitations of traditional static graph models that cannot respond to real-time changes, making risk assessment more timely and scenario-adaptive.

[0076] S140, based on the spatio-temporal tensor and the dynamic graph structure, generating a risk equation of the engineering progress in the spatio-temporal dimension, and solving the risk equation through a differential equation to output the risk item of the engineering progress.

[0077] In an embodiment of the present application, based on the spatio-temporal tensor and the dynamic graph structure, a risk equation of the engineering progress in the spatio-temporal dimension is generated, and the risk equation is solved through a differential equation to output the risk item of the engineering progress, comprising:

[0078] Based on the dynamic graph structure, determine the diffusion term of risk through the dynamic graph structure;

[0079] Based on the spatio-temporal tensor, determine the driving term of risk of multi-modal feature change;

[0080] Based on the diffusion term and the driving term, determine the risk equation of the engineering progress risk in the spatio-temporal dimension;

[0081] Solve the risk equation through a differential equation to output the risk item of the engineering progress.

[0082] In an embodiment of the present application, based on the dynamic graph structure, the diffusion term of risk through the dynamic graph structure is determined, comprising:

[0083] In an embodiment of the present application, based on the risk probability of the task node in the dynamic graph structure, the diffusion term of risk through the dynamic graph structure is determined For:

[0084]

[0085] Wherein, respectively represent the risk probability of task node i and task node j, denotes the edge weight between task node i and task node j, denotes the neighbor set of task node i.

[0086] In this embodiment, the diffusion term describes the degree of risk diffusion along the task dependency and resource flow edges, for example, the risk of a critical path task is transmitted to adjacent tasks through high weight edges. The high risk difference between tasks amplifies the diffusion effect through edge weights, such as the delay of a critical path task triggering a chain reaction.

[0087] In an embodiment of the present application, based on the spatio-temporal tensor, the driving term of the multi-modal feature change on the risk is determined is:

[0088]

[0089] wherein, denotes the sensitivity of the risk to the feature , denotes the time derivative of the feature , i.e. the rate of change of the feature over time, D denotes the total number of feature dimensions, and d denotes the identification of the feature dimension.

[0090] In this embodiment, the driving term quantifies the driving effect of the multi-modal feature (such as resource shortage, progress deviation) on the risk over time. If a certain feature dimension worsens over time ( ), such as the resource shortage index, and the risk is sensitive to this feature ( ), then the risk accumulates rapidly.

[0091] In an embodiment of the present application, based on the diffusion term and the driving term, the risk equation of the engineering progress risk in the spatio-temporal dimension is:

[0092]

[0093] wherein, denotes a preset hyperparameter, respectively controlling the contribution degree of graph diffusion and tensor gradient.

[0094] wherein, , denotes a nonlinear interaction term. Used to capture the amplification effect of sudden events through a deep neural network, such as the superposition of adverse weather and equipment failure leading to a sharp increase in risk.

[0095] wherein, W denotes a learnable weight matrix, R denotes a risk term, and T denotes a feature term. When a high-risk task (high R value) is combined with a specific feature (such as a high-temperature environment ), it may trigger a nonlinear risk surge. For example, abnormal concrete solidification in a high-temperature environment leads to an exponential increase in progress risk.

[0096] In an embodiment of the present application, the risk equation is solved by a differential equation to output the risk term of the engineering progress. The risk equation is converted into a numerical integration problem by a neural differential equation solver, and the risk term is calculated by adaptive iteration. In a specific implementation, the current risk term of each time window t is Predicting the rate of change through a neural network and updating to .

[0097] The output risk term R is in the form of a matrix, and each element represents the risk value of task n at time t. Through the fusion of physical mechanisms and data-driven, it can not only explain the transmission path of risk along the task network, such as the chain reaction caused by the change of dependency edge weight, but also dynamically respond to the nonlinear influence of multi-modal features, such as the risk correction triggered by abnormal sensor signals. In practical applications, the risk term can be visualized through a heat map to help managers quickly locate high-risk tasks and optimize resource allocation, such as automatically triggering an early warning when the risk value exceeds a threshold, reducing the probability of engineering delays.

[0098] S150, generating an engineering progress risk assessment file according to the risk term and actual progress data in the engineering information.

[0099] In an embodiment of the present application, the system analyzes the risk term to identify the risk value of each task, which is divided into three levels of low, medium and high according to the threshold. For example, high-risk tasks are marked as progress delay exceeding 7 days or resource consumption exceeding budget by 20%, and combined with the edge weight in the dynamic graph, the risk transmission source is located, such as a certain key task delay causing a chain of risks in subsequent tasks.

[0100] Convert the abstract risk value output by the mathematical model into an engineering language description. For example: Task A (steel structure installation) has a high current risk level, the main cause is equipment failure causing a 15% delay in the construction period, which has affected Task B (roof construction) through the dependency edge, and it is recommended to prioritize the allocation of spare equipment and extend the daily working hours. The report synchronously associates actual progress data, such as Gantt chart deviations and resource inventory, to verify the credibility of risk calculation.

[0101] According to the analysis of risk transmission path and nonlinear effect, targeted measures are proposed. For example: for high-risk tasks affected by resource flow edges, it is recommended to increase buffer inventory; for risks driven by multi-modal features (such as sensor data in the rainy season), it is recommended to adjust the construction plan or introduce an emergency plan. The report presents recommendations in priority order to ensure that the engineering team responds quickly. By integrating theoretical models and field data, complex mathematical results are converted into executable decision support to help managers intervene proactively before risk escalation, improving the overall controllability of the project.

[0102] The technical scheme of the application obtains multi-modal engineering information, the engineering information includes structured data, unstructured data and time series data, maps the engineering information to a three-order tensor, generates a tensor slice of the three-order tensor based on a preset projection function as a space-time tensor, constructs a time-varying graph composed of engineering tasks and execution relationships based on the engineering information, extracts task features from the space-time tensor, generates edge weights based on the task features, adjusts the node relationship in the time-varying graph through the edge weights, and outputs a dynamic graph structure; based on the space-time tensor and the dynamic graph structure, generates a risk equation of engineering progress in the space-time dimension, and solves the risk equation through a differential equation to output a risk term of engineering progress; and generates an engineering progress risk assessment file according to the risk term and actual progress data in the engineering information. The above process unifies the coding of multi-modal data through a three-order tensor, dynamically weights and fuses structured, unstructured and time series features by using an attention mechanism, and breaks through the problem of traditional data fragmentation. The dynamic graph structure adaptively updates the edge weights based on space-time convolution, captures task dependency and resource flow changes in real time, and solves the lag of static models. The risk equation fuses a graph diffusion term (task network conduction), a tensor gradient term (multi-modal feature driving) and a nonlinear interaction term (burst event amplification effect), and solves the risk equation by combining a neural differential equation to realize accurate modeling of the risk space-time evolution. Compared with traditional methods, the dynamic fusion and mechanism driving characteristics can reduce the risk assessment error, improve the efficiency of engineering progress risk assessment, and support the transition of engineering decision-making from passive response to active intervention.

[0103] The following describes an apparatus embodiment of the application, which can be used to execute the multi-modal data driven engineering progress risk intelligent assessment method in the above embodiments of the application. It can be understood that the apparatus can be a computer program (including program code) running in a computer device, for example, the apparatus is an application software; the apparatus can be used to execute corresponding steps in the method provided by the embodiments of the application. For details not disclosed in the apparatus embodiment of the application, please refer to the above embodiments of the multi-modal data driven engineering progress risk intelligent assessment method.

[0104] Figure 3 A block diagram of a multi-modal data driven engineering progress risk intelligent assessment system according to an embodiment of the application is shown.

[0105] Referring to Figure 3 The multi-modal data driven engineering progress risk intelligent assessment system according to an embodiment of the application includes:

[0106] The acquisition unit 310 is configured to acquire multi-modal engineering information, and the engineering information includes structured data, unstructured data and time series data.

[0107] The tensor unit 320 is configured to map the engineering information into a third-order tensor, generate a tensor slice of the third-order tensor as a space-time tensor based on a preset projection function, and the third-order tensor comprises an entity dimension, a time dimension and a feature dimension.

[0108] The structure unit 330 is configured to construct a time-varying graph constituted by engineering tasks and execution relationships based on the engineering information, extract task features from the space-time tensor, generate edge weights based on the task features, adjust node relationships in the time-varying graph through the edge weights, and output a dynamic graph structure.

[0109] The risk unit 340 is configured to generate a risk equation of the engineering progress in a space-time dimension based on the space-time tensor and the dynamic graph structure, solve the risk equation through a differential equation, and output a risk term of the engineering progress.

[0110] The evaluation unit 350 is configured to generate an engineering progress risk evaluation file according to the risk term and actual progress data in the engineering information.

[0111] In the present application, based on the foregoing scheme, the engineering information is mapped into a third-order tensor, and a tensor slice of the third-order tensor is generated as a space-time tensor based on a preset projection function. The scheme comprises: extracting first information belonging to an entity dimension, a time dimension and a feature dimension in the engineering information, and mapping the first information into a third-order tensor; determining attention weights of each modality based on a learnable weight matrix; and generating a tensor slice of the third-order tensor as a space-time tensor based on the attention weights of each modality and a preset projection function.

[0112] In the present application, based on the foregoing scheme, the attention weights of each modality are determined based on a learnable weight matrix. The scheme comprises: performing spatial dimension compression on the original input features of the mth modality to generate compressed data; and determining the attention weight corresponding to the mth modality based on the compressed data and a learnable weight matrix of the mth modality .

[0113]

[0114] wherein, represents the original input features of the mth modality, represents the compressed data, k and m represent the identification of the modality, and M represents the total number of the modality.

[0115] ​In the present application, based on the foregoing scheme, the tensor slice of the third-order tensor is generated as a spatiotemporal tensor based on the attention weight of each modality and the preset projection function, comprising: mapping the original input features of the mth modality in the engineering entity dimension n and the time dimension t to a unified dimension to generate dimension data; based on the attention weight of each modality and the dimension data, generating a tensor slice of the third-order tensor as a spatiotemporal tensor :

[0116]

[0117] wherein, represents the original input features of the mth modality in the engineering entity n and the time window t, represents the dimension data, represents an activation function.

[0118] In the present application, based on the foregoing scheme, the task feature is extracted from the spatiotemporal tensor based on the engineering information to construct a time-varying graph composed of engineering tasks and execution relationships, the edge weight is generated based on the task feature, the node relationship in the time-varying graph is adjusted through the edge weight, and a dynamic graph structure is output, comprising: obtaining the engineering tasks and execution relationships in the engineering information, simulating nodes, edges and edge attributes based on the engineering tasks and execution relationships to form a time-varying graph; extracting a task feature from the spatiotemporal tensor, generating an edge weight based on the task feature; adjusting the node relationship in the time-varying graph through the edge weight, and outputting a dynamic graph structure.

[0119] In the present application, based on the foregoing scheme, the risk equation of the engineering progress in the spatiotemporal dimension is generated based on the spatiotemporal tensor and the dynamic graph structure, and the risk equation is solved by a differential equation to output the risk term of the engineering progress, comprising: determining the diffusion term of the risk through the dynamic graph structure based on the dynamic graph structure; determining the driving term of the risk of the multi-modal feature change based on the spatiotemporal tensor; determining the risk equation of the engineering progress risk in the spatiotemporal dimension based on the diffusion term and the driving term; solving the risk equation by a differential equation to output the risk term of the engineering progress.

[0120] In the present application, based on the foregoing scheme, the diffusion term of the risk through the dynamic graph structure is determined based on the risk probability of the task node in the dynamic graph structure :

[0121]

[0122] wherein, respectively represent the risk probability of task node i and task node j, denotes an edge weight between task node i and task node j, denotes a neighbor set of task node i.

[0123] The technical scheme of the present application obtains multi-modal engineering information, which includes structured data, unstructured data and time series data; maps the engineering information into a three-order tensor, generates a tensor slice of the three-order tensor based on a preset projection function as a space-time tensor; constructs a time-varying graph composed of engineering tasks and execution relationships based on the engineering information, extracts task features from the space-time tensor, generates edge weights based on the task features, adjusts the node relationship in the time-varying graph through the edge weights, and outputs a dynamic graph structure; generates a risk equation of engineering progress in the space-time dimension based on the space-time tensor and the dynamic graph structure, and solves the risk equation through a differential equation to output a risk term of engineering progress; and generates an engineering progress risk assessment file according to the risk term and actual progress data in the engineering information. The above process unifies the coding of multi-modal data through a three-order tensor, dynamically weights and fuses structured, unstructured and time series features by using an attention mechanism, and breaks through the problem of traditional data fragmentation. The dynamic graph structure adaptively updates the edge weights based on space-time convolution, captures the changes of task dependence and resource flow in real time, and solves the lag of the static model. The risk equation fuses a graph diffusion term (task network conduction), a tensor gradient term (multi-modal feature driving) and a nonlinear interaction term (burst event amplification effect), and solves the risk equation by combining a neural differential equation, so as to realize accurate modeling of the risk space-time evolution. Compared with traditional methods, the dynamic fusion and mechanism driving characteristics can reduce the risk assessment error, improve the efficiency of engineering progress risk assessment, and support the transition of engineering decision-making from passive response to active intervention.

[0124] Figure 4 A structural schematic diagram of a computer system of an electronic device suitable for implementing the embodiments of the present application is shown.

[0125] It should be noted that the computer system of the electronic device in the present embodiment is only an example, and should not impose any limitation on the functions and use range of the embodiments of the present application.

[0126] The computer system in the present embodiment includes a central processing unit 401, which can perform various appropriate actions and processes according to programs stored in a read-only memory 402 or programs loaded into a random access memory 403 from a storage part 408, such as the multi-modal data driven engineering progress risk intelligent assessment method described in the above embodiments. Various programs and data required for system operation are also stored in the random access memory 403. The central processing unit 401, the read-only memory 402 and the random access memory 403 are connected to each other through a bus 404. An input / output interface 405 is also connected to the bus 404.

[0127] The following components are connected to the input / output interface 405: an input portion 406 including input devices such as a keyboard and a mouse; an output portion 407 including output devices such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and a speaker; a storage portion 408 including a hard disk; and a communication portion 409 including a network interface card such as a LAN (Local Area Network) card, a modem, and the like. The communication portion 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output interface 405 as necessary. A removable media 411 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is attached to the drive 410 as necessary, so that a computer program read therefrom is installed in the storage portion 408 as necessary.

[0128] In particular, the processes described above with reference to the flowcharts can be implemented as a computer software program according to embodiments of the present application. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program containing computer programs for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network by the communication portion 409, and / or installed from the removable media 411. When the computer program is executed by the central processing unit 401, various functions defined in the system of the present application are executed.

[0129] It should be noted that the computer-readable medium in the embodiments of the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination thereof. The computer-readable storage medium may, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination thereof. More specific examples of the computer-readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (Compact Disc Read-Only Memory, CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or apparatus. In the present application, the computer-readable signal medium can include a data signal carrying computer-readable computer programs in a baseband or as a part of a carrier wave. Such a propagated data signal can take various forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than the computer-readable storage medium, which can send, propagate or transmit programs for use by or in conjunction with an instruction execution system, device or apparatus. The computer programs contained in the computer-readable medium can be transmitted by any suitable medium, including but not limited to wireless, wired, or the like, or any suitable combination thereof.

[0130] The flowcharts and block diagrams in the drawings illustrate the possible implementation architectures, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In the flowcharts or block diagrams, each block can represent a module, a program segment or a part of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different orders than that shown in the drawings. For example, two blocks represented in succession can actually be executed substantially in parallel, and sometimes in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams or flowcharts, and the combination of blocks in the block diagrams or flowcharts, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0131] The units described in the embodiments of the present application can be implemented by software, or can be implemented by hardware, and the units described can also be arranged in a processor. In some cases, the names of the units do not constitute a limitation on the units themselves.

[0132] According to an aspect of the present application, a computer program product or computer program is provided, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method provided in the various optional implementation manners described above.

[0133] As another aspect, the present application also provides a computer readable medium, which can be included in the electronic device described in the above embodiments, or can exist separately without being assembled into the electronic device. The computer readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to implement the multi-modal data driven engineering progress risk intelligent evaluation method described in the above embodiments.

[0134] It should be noted that although several modules or units of the device for action execution are mentioned in the above detailed description, such division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into several modules or units.

[0135] From the above description of the embodiments, those skilled in the art can easily understand that the example embodiments described herein can be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solutions according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, U disk, mobile hard disk, etc.) or network, and includes several instructions to make a computing device (which can be a personal computer, server, touch terminal, or network device, etc.) execute the methods according to the embodiments of the present application.

[0136] Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the embodiments disclosed herein. The present application is intended to cover any variations, uses, or adaptations of the application following the general principles thereof and including such departures from the present disclosure as come within known use or custom in the art.

[0137] It is to be understood that the application is not limited to the precise construction already described above and shown in the drawings, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application. The scope of the application should only be limited by the claims appended hereto.

Claims

1. A multi-modal data-driven intelligent assessment method for engineering progress risk, characterized in that, The method comprises the following steps: acquiring multi-modal engineering information, the engineering information comprising structured data, unstructured data and time series data; mapping the engineering information into a three-order tensor, generating tensor slices of the three-order tensor as a space-time tensor based on a preset projection function; wherein the three-order tensor comprises an engineering entity dimension, a time dimension and a feature dimension; the engineering entity dimension comprises a task node; constructing a time-varying graph comprising engineering tasks and execution relationships based on the engineering information, extracting task features from the space-time tensor, generating edge weights based on the task features, adjusting node relationships in the time-varying graph through the edge weights, and outputting a dynamic graph structure; based on the space-time tensor and the dynamic graph structure, generating a risk equation of engineering progress in the space-time dimension, and solving the risk equation through a differential equation to output a risk term of the engineering progress; generating an engineering progress risk assessment file according to the risk term and actual progress data in the engineering information; wherein, based on the space-time tensor and the dynamic graph structure, generating a risk equation of engineering progress in the space-time dimension, and solving the risk equation through a differential equation to output a risk term of the engineering progress, comprises: determining a diffusion term of risk through the dynamic graph structure based on the dynamic graph structure; determining a driving term of risk caused by multi-modal feature changes based on the space-time tensor; determining a risk equation of engineering progress risk in the space-time dimension based on the diffusion term and the driving term; solving the risk equation through a differential equation to output a risk term of the engineering progress; wherein, determining a diffusion term of risk through the dynamic graph structure based on the dynamic graph structure, comprises: determining a risk diffusion term of a risk passing through the dynamic graph structure based on the risk probability of the task node in the dynamic graph structure is: ； wherein, respectively denote the risk probability of task node i and task node j, denotes the edge weight between task node i and task node j, denotes the neighbor set of task node i; wherein, based on the spatiotemporal tensor, a driving term of the multimodal feature change to the risk is determined is: ； wherein, represents the sensitivity of the risk to the feature , represents the time derivative of the feature , i.e. the rate of change of the feature over time, D represents the total number of feature dimensions, and d represents the identity of a feature dimension; wherein, determining a risk equation of engineering progress risk in the space-time dimension based on the diffusion term and the driving term, is: ； wherein, denotes a preset hyper-parameter; denotes a nonlinear interaction term; W denotes a learnable weight matrix, R denotes a risk term, and T denotes a feature term.

2. The multi-modal data-driven engineering schedule risk intelligence assessment method of claim 1, wherein, mapping the engineering information into a three-order tensor, generating tensor slices of the three-order tensor as a space-time tensor based on a preset projection function, comprises: extracting first information belonging to the engineering entity dimension, the time dimension and the feature dimension in the engineering information, and mapping the first information into a three-order tensor; determining attention weights of each modality based on a learnable weight matrix; generating tensor slices of the three-order tensor as a space-time tensor based on the attention weights of each modality and a preset projection function.

3. The multi-modal data-driven engineering schedule risk intelligence assessment method of claim 2, wherein, determining attention weights of each modality based on a learnable weight matrix, comprises: raw input features of the mth modality perform spatial dimension compression to generate compressed data determining attention weights corresponding to the mth modality based on the compressed data and a learnable weight matrix of the mth modality , determining attention weights corresponding to the mth modality based on the compressed data and a learnable weight matrix of the mth modality 4. The multi-modal data-driven engineering schedule risk intelligence assessment method of claim 3, wherein, generating tensor slices of the three-order tensor as a space-time tensor based on the attention weights of each modality and a preset projection function, comprises: mapping original input features of the mth modality in the engineering entity dimension n and the time dimension t to a unified dimension to generate dimension data; generating tensor slices of the three-order tensor as a space-time tensor based on the attention weights of each modality and the dimension data.

5. The multi-modal data driven engineering schedule risk intelligence assessment method of claim 1, wherein, constructing a time-varying graph comprising engineering tasks and execution relationships based on the engineering information, extracting task features from the space-time tensor, generating edge weights based on the task features, adjusting node relationships in the time-varying graph through the edge weights, and outputting a dynamic graph structure, comprises: Obtaining the engineering tasks and execution relationships in the engineering information, simulating nodes, edges and edge attributes based on the engineering tasks and execution relationships, and constructing a time-varying graph; Extracting task features from the space-time tensor, and generating edge weights based on the task features; Adjusting the node relationships in the time-varying graph through the edge weights, and outputting a dynamic graph structure.

6. A multi-modal data-driven intelligent assessment system for engineering progress risk, the system comprising: Comprise: An acquisition unit is configured to acquire multi-modal engineering information, the engineering information comprising structured data, unstructured data and time series data; A tensor unit is configured to map the engineering information into a three-order tensor, generate tensor slices of the three-order tensor as a space-time tensor based on a preset projection function; wherein the three-order tensor comprises an engineering entity dimension, a time dimension and a feature dimension; the engineering entity dimension comprises task nodes; A structure unit is configured to construct a time-varying graph composed of engineering tasks and execution relationships based on engineering information, extract task features from the space-time tensor, generate edge weights based on the task features, adjust the node relationships in the time-varying graph through the edge weights, and output a dynamic graph structure; A risk unit is configured to generate a risk equation of engineering progress in the space-time dimension based on the space-time tensor and the dynamic graph structure, solve the risk equation through a differential equation, and output a risk term of engineering progress; An evaluation unit is configured to generate an engineering progress risk assessment file according to the risk term and actual progress data in the engineering information; Wherein, based on the space-time tensor and the dynamic graph structure, generating a risk equation of engineering progress in the space-time dimension, solving the risk equation through a differential equation, and outputting a risk term of engineering progress, comprise: Based on the dynamic graph structure, determining a diffusion term of risk through the dynamic graph structure; Based on the space-time tensor, determining a driving term of risk caused by multi-modal feature changes; Based on the diffusion term and the driving term, determining a risk equation of engineering progress risk in the space-time dimension; Solving the risk equation through a differential equation, and outputting a risk term of engineering progress; Wherein, based on the dynamic graph structure, determining a diffusion term of risk through the dynamic graph structure, comprises: determining a risk diffusion term of a risk passing through the dynamic graph structure based on the risk probability of the task node in the dynamic graph structure is: ； wherein, respectively denote the risk probability of task node i and task node j, denotes the edge weight between task node i and task node j, denotes the neighbor set of task node i; wherein, based on the spatiotemporal tensor, a driving term of the multimodal feature change to the risk is determined is: ； wherein, represents the sensitivity of the risk to the feature , represents the time derivative of the feature , i.e. the rate of change of the feature over time, D represents the total number of feature dimensions, and d represents the identity of a feature dimension; Wherein, based on the diffusion term and the driving term, the risk equation of engineering progress risk in the space-time dimension is: ； wherein, represents a preset hyperparameter; represents a nonlinear interaction term; W represents a learnable weight matrix, R represents a risk term, and T represents a feature term.

7. The multi-modal data-driven engineering schedule risk intelligence assessment system of claim 6, wherein, Mapping the engineering information into a three-order tensor, generating tensor slices of the three-order tensor as a space-time tensor based on a preset projection function, comprise: Extracting first information belonging to the engineering entity dimension, the time dimension and the feature dimension in the engineering information, and mapping the first information into a three-order tensor; Based on the learnable weight matrix, determining attention weights of each modality; Based on the attention weights of each modality and the preset projection function, generating tensor slices of the three-order tensor as a space-time tensor.

8. The multi-modal data-driven engineering schedule risk intelligence assessment system of claim 7, wherein, Based on the learnable weight matrix, determining attention weights of each modality, comprises: raw input features of the mth modality perform spatial dimension compression to generate compressed data based on the compressed data and a learnable weight matrix of the mth modality determine an attention weight corresponding to the mth modality is: ； wherein, represents the original input features of the mth modality, represents compressed data, k and m represent the identity of the modality, and M represents the total number of modalities, represents the learnable weight matrix of the kth modality.

Images