Question and answer system control method, device and storage medium based on engineering management

Abstract

The application discloses an engineering management-based question and answer system control method and device and a storage medium, and belongs to the technical field of knowledge question and answer systems. The method comprises the following steps: receiving an inquiry question, decomposing the inquiry question into at least one sub-question, positioning and acquiring an entity state parameter subset corresponding to the sub-question in a digital twin, fusing field knowledge information with the entity state parameter subset and the sub-question, establishing context information of the sub-question, generating at least one sub-question hypothesis information according to the context information through an inference module, simulating and verifying the hypothesis information through the digital twin, forming an inference path according to a simulation verification result, synthesizing a thinking chain based on the inference path, and generating an inquiry result of the inquiry question according to the thinking chain. The application synthesizes the thinking chain based on the inference path and generates the inquiry result, ensures that the output conclusion is logically coherent and consistent with the actual state of the engineering entity and the field specification, and improves the reliability of the output result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of knowledge question and answer systems, and in particular to a question and answer system control method based on engineering management, equipment and a storage medium. BACKGROUND

[0002] In order to improve the intelligence level of the question and answer system, the related technology usually accesses a large language model (LLM) in the engineering management system. When the system receives a user's consultation question, the question and the relevant static data text retrieved from the system database are submitted to the LLM, and a general knowledge built-in in the LLM is relied on to generate and directly return a text as an answer.

[0003] In actual application, the answer generation of the LLM seriously depends on the static general knowledge of the model pre-training, and cannot be deeply combined with the real-time data and field rules specific to the project and dynamically changing, and cannot be logically closed loop verified. This leads to the system that may generate general or wrong answers that are out of the actual state of the current project, lack of data and rule support, for the consultation questions that need to be closely combined with the specific project context for logical deduction and causal analysis, and the credibility of the answers is difficult to guarantee.

[0004] The above content is only used to assist in understanding the technical solutions of the present application, and does not represent the acknowledgement of the above content as prior art. SUMMARY

[0005] The main purpose of the present application is to provide a question and answer system control method based on engineering management, equipment and a storage medium, aiming at solving the technical problem of low answer credibility of the large language model accessed in the engineering management system.

[0006] To achieve the above purpose, the present application provides a question and answer system control method based on engineering management, which comprises the following steps:

[0007] receiving a consultation question, decomposing the consultation question into at least one sub-question, and positioning and obtaining an entity state parameter subset corresponding to the sub-question in a digital twin;

[0008] associating and fusing the field knowledge information with the entity state parameter subset and the sub-question, and establishing the context information of the sub-question;

[0009] generating at least one hypothesis information of the sub-question according to the context information through an inference module;

[0010] verifying the hypothesis information through the digital twin, and forming an inference path according to the simulation verification result;

[0011] Based on the reasoning path, a thought chain is synthesized, and a consultation result for the consultation question is generated based on the thought chain.

[0012] In one embodiment, prior to the steps of receiving a consultation question, decomposing the consultation question into at least one sub-question, and locating and obtaining a subset of entity state parameters corresponding to the sub-question in the digital twin, the method further includes:

[0013] Acquire real-time image data;

[0014] Object recognition and state analysis are performed on the real-time image data to extract entity state information of the construction site;

[0015] The entity state information is matched with the corresponding entity in the digital twin, and the entity state parameters in the digital twin are updated according to the matching result.

[0016] In one embodiment, the steps of receiving a consultation question, decomposing the consultation question into at least one sub-question, and locating and obtaining a subset of entity state parameters corresponding to the sub-question in the digital twin include:

[0017] Natural language processing is performed on the consultation question to identify entity information in the consultation question;

[0018] Based on the domain knowledge base, the entity information is mapped to a predefined conceptual model;

[0019] Based on the entity relationship information in the conceptual model, the consultation problem is decomposed into at least one of the sub-problems;

[0020] Based on the correspondence between the sub-problems and the entity information, the subset of entity state parameters is queried in the entity model of the digital twin.

[0021] In one embodiment, the step of associating and fusing domain knowledge information with the subset of entity state parameters and the sub-problem to establish context information for the sub-problem includes:

[0022] Based on the sub-problem, a target knowledge unit is retrieved from the domain knowledge base, wherein the target knowledge unit includes business rules, constraints and / or cases;

[0023] By using a semantic matching algorithm, the target knowledge unit is associated with the subset of entity state parameters to construct the context graph of the sub-problem;

[0024] The context information is formed by fusing the context graph and the sub-problems.

[0025] In one embodiment, the step of generating hypothesis information for at least one of the sub-problems based on the context information via the reasoning module includes:

[0026] In the reasoning module, a set of reasoning rules corresponding to the problem type of the sub-problem is loaded;

[0027] By applying the inference rule set, logical reasoning is performed on the context graph to generate at least one candidate hypothesis;

[0028] Calculate the confidence level of the candidate hypothesis, and generate the hypothesis information based on the confidence level and the candidate hypothesis.

[0029] In one embodiment, the step of simulating and verifying the hypothetical information using the digital twin, and forming a reasoning path based on the simulation and verification results, includes:

[0030] Based on the assumed information, the simulation parameters of the digital twin are calculated;

[0031] Based on the simulation parameters and the physical model of the digital twin, a simulation model corresponding to the hypothetical information is constructed.

[0032] According to the preset verification rules, the target confidence level of the hypothesis information is calculated by comparing the simulation model with the hypothesis information.

[0033] Based on the simulation verification results corresponding to the target confidence level, the target hypothesis information is determined, and the inference path is formed based on the target hypothesis information.

[0034] In one embodiment, the step of synthesizing a thought chain based on the reasoning path and generating a consultation result for the consultation question based on the thought chain includes:

[0035] Extract key reasoning nodes and logical connections from the reasoning path to construct a logical chain;

[0036] Using a templated natural language generation algorithm, the logical chain is converted into a structured text description, forming the thought chain;

[0037] The consultation results are generated based on the thought chain using a large model.

[0038] In one embodiment, prior to the steps of receiving a consultation question, decomposing the consultation question into at least one sub-question, and locating and obtaining a subset of entity state parameters corresponding to the sub-question in the digital twin, the method further includes:

[0039] Based on the entity state parameters of the digital twin, identify the risk information corresponding to the model nodes in the digital twin;

[0040] Obtain the association information of the digital twin and its mapping relationship with the model nodes;

[0041] Based on the mapping relationship, and according to the risk information and the association information, the annotation information of the model nodes is generated;

[0042] The annotation information is added to the digital twin to form the entity state parameters of the digital twin.

[0043] In addition, to achieve the above objectives, this application also provides a question-and-answer system control device based on engineering management, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the question-and-answer system control method based on engineering management as described above.

[0044] In addition, to achieve the above objectives, this application also provides a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the question-and-answer system control method based on engineering management as described above.

[0045] One or more technical solutions proposed in this application have at least the following technical effects:

[0046] This application decomposes received consultation questions into sub-questions, locates and obtains the entity state parameter subsets corresponding to the sub-questions in a digital twin, and integrates domain knowledge information with the entity state parameter subsets and the sub-questions to establish context information for the sub-questions. This enables the system to reason under the dual constraints of specific domain rules and specific project states. Then, the reasoning module generates hypothesis information, and the hypothesis information is verified through the digital twin to form a reasoning path. Based on the reasoning path, a thought chain is synthesized and consultation results are generated, thereby ensuring that the output conclusions not only have logical coherence but also highly match the actual state of the engineering entity and domain specifications, significantly improving the reliability and professionalism of the system output results. Attached Figure Description

[0047] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0048] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart illustrating the first embodiment of the question-and-answer system control method based on engineering management of this application.

[0050] Figure 2 This is a flowchart illustrating the second embodiment of the question-and-answer system control method based on engineering management in this application.

[0051] Figure 3 This is a schematic diagram of the structure of a question-and-answer system control device based on engineering management, which is part of the hardware operating environment involved in the embodiments of this application.

[0052] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0053] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0054] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0055] The main solution of this application embodiment is as follows: receiving a consultation question, decomposing the consultation question into at least one sub-question, locating and obtaining the entity state parameter subset corresponding to the sub-question in a digital twin, associating and fusing domain knowledge information with the entity state parameter subset and the sub-question, establishing context information for the sub-question, generating hypothesis information for at least one sub-question based on the context information through a reasoning module, performing simulation verification on the hypothesis information through a digital twin, forming a reasoning path based on the simulation verification results, synthesizing a thought chain based on the reasoning path, and generating a consultation result for the consultation question based on the thought chain.

[0056] Current technologies typically integrate large language models (LLMs) into engineering management systems. When the system receives a user's inquiry, it submits the inquiry along with relevant static text data retrieved from the system database to the LLM. The LLM then generates and returns a text as the answer, relying on its built-in general knowledge. In practice, the LLM's answer generation heavily depends on the static general knowledge pre-trained in the model, failing to deeply integrate with project-specific, dynamically changing real-time data and domain rules for logical closure and verification. This leads to situations where, for inquiries requiring close integration with the specific project context for logical deduction and causal analysis, the system may generate vague or incorrect answers that are detached from the current project status, lack data and rule support, and whose credibility is difficult to guarantee.

[0057] This application decomposes received consultation questions into sub-questions, locates and obtains the entity state parameter subsets corresponding to the sub-questions in a digital twin, and integrates domain knowledge information with the entity state parameter subsets and the sub-questions to establish context information for the sub-questions. This enables the system to reason under the dual constraints of specific domain rules and specific project states. Then, the reasoning module generates hypothesis information, and the hypothesis information is verified through the digital twin to form a reasoning path. Based on the reasoning path, a thought chain is synthesized and consultation results are generated, thereby ensuring that the output conclusions not only have logical coherence but also highly match the actual state of the engineering entity and domain specifications, significantly improving the reliability, professionalism, and decision support value of the system output results.

[0058] To better understand the above technical solutions, exemplary embodiments of this application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.

[0059] It should be noted that the executing entity in this embodiment can be an engineering management system, or a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, or a question-and-answer system control device based on engineering management, etc. This embodiment does not specifically limit it. The following uses an engineering management system as an example to describe this embodiment and the following embodiments.

[0060] Based on this, embodiments of this application provide a question-and-answer system control method based on engineering management, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the question-and-answer system control method based on engineering management in this application.

[0061] In this embodiment, the question-and-answer system control method based on engineering management includes steps S10 to S40:

[0062] Step S10: Receive the consultation question, decompose the consultation question into at least one sub-question, and locate and obtain the subset of entity state parameters corresponding to the sub-question in the digital twin;

[0063] In this embodiment, the engineering management system receives the original consultation questions raised by users through a multimodal interaction interface, then initiates a question understanding and decomposition process to transform them into structured queries, while simultaneously activating the data service layer of the digital twin to obtain real-time status data. Here, the consultation question refers to the engineering management-related questions expressed by the user in natural language, voice, or image form; the digital twin refers to a virtual simulation model constructed using IoT sensors, Building Information Modeling (BIM), and business system data, which maintains a dynamic mapping with the physical engineering entity; and the entity status parameter subset refers to the set of attribute data extracted from the digital twin that is related to the current sub-question, including geometric attributes, physical status, and / or business indicators.

[0064] It should be noted that the project management system has a built-in Large Language Model (LLM) and a digital twin. The LLM can receive user questions and, during the answer generation process, query the digital twin and / or domain knowledge base to obtain relevant information and knowledge to assist in logical decomposition, generation of logical chains, and generation of answer text.

[0065] Specifically, in the engineering management system, the system performs natural language processing on the consultation question, identifies the entity information in the consultation question, maps the entity information to a predefined conceptual model based on the domain knowledge base, decomposes the consultation question into at least one sub-question based on the entity relationship information in the conceptual model, and queries the entity state parameter subset in the entity model of the digital twin based on the correspondence between the sub-question and the entity information.

[0066] Optionally, the multimodal input processing unit employs an end-to-end speech recognition model and an attention-based image description generation model to uniformly convert non-text input into standard text format. The question parsing engine utilizes a dependency parser and semantic role labeling tools to identify the core predicate structure and argument relations in the consultation question. A sequence-to-sequence architecture-based question decomposition model, based on a pre-defined engineering domain question pattern library, breaks down complex questions into a logically related sequence of atomic sub-questions. Simultaneously, the digital twin query agent matches entity names mentioned in the question with entity identifiers in the twin using entity linking technology, and performs multi-hop traversal in the entity relationship network using graph database query language, aggregating entity state parameter subsets through a data acquisition interface.

[0067] As an optional implementation, the problem decomposition process can employ a deep parsing method based on domain ontology reasoning. The system first constructs a domain knowledge base containing engineering management concepts, attributes, and constraints, explicitly defining core concepts such as "construction activities," "resource allocation," and / or "risk factors" and their relationships. Upon receiving a consultation question, the system uses a semantic parser to convert the natural language question into an ontology-based logical expression, and then uses a descriptive logic inference engine to detect implicit query requirements within the expression. The query planner automatically generates the optimal sub-question query plan based on the concept hierarchy and attribute constraints defined in the ontology, ensuring that the decomposed sub-questions not only cover all aspects of the original question but also conform to the structured characteristics of domain knowledge.

[0068] Optionally, the project management system can optimize the decomposition process of consultation questions using an intent recognition model. First, a trained text classification model is used to categorize the consultation questions by intent, determining if they belong to categories such as schedule management, quality management, or safety management. Different categories correspond to different decomposition frameworks. For example, schedule management questions use a "current situation analysis - reasons for gaps - adjustment plan" framework, while quality management questions use a "problem manifestation - influencing factors - solutions" framework. Then, based on the classification results, the decomposition rules of the corresponding framework are invoked, and sub-questions are generated by combining entity recognition results. Simultaneously, the accuracy of the sub-questions is confirmed through real-time interaction with the user. If the user reports that a sub-question deviates from the requirements, the decomposition logic is readjusted.

[0069] As an alternative implementation, the engineering management system can incorporate knowledge graphs for sub-problem decomposition. First, a knowledge graph for the engineering management domain is constructed, containing entities (such as processes, resources, and standards), relationships (such as "includes," "impacts," and "basis"), and attributes (such as process duration and resource quantity). The consultation question is converted into a query statement for the knowledge graph. Entities and relationships related to the question are retrieved through the graph database, and the knowledge modules involved are determined based on the search results. Then, according to the hierarchical structure and association strength of the knowledge modules, the consultation question is broken down into sub-problems corresponding to each knowledge module, ensuring that each sub-problem can be mapped to a specific entity cluster in the knowledge graph, providing precise guidance for subsequent information retrieval.

[0070] Optionally, digital twin data acquisition can employ an enhanced retrieval strategy based on spatiotemporal indexing. The system establishes a four-dimensional spatiotemporal index structure within the digital twin, mapping entity state change records to spatiotemporal coordinates. When processing inquiries involving historical state retrospection or future trend prediction, the query engine not only retrieves the entity's current state but also obtains state sequence data within a specific time period through spatiotemporal range queries. Simultaneously, the system integrates a streaming computing framework to perform real-time anomaly detection and feature extraction on the entity state data, ensuring that the returned subset of entity state parameters contains preprocessed time-series feature indicators and anomaly marker information.

[0071] For example, when a user submits a query asking "Analyze the stability of the foundation pit support structure under current rainfall conditions," the system first breaks it down into sub-questions such as "Obtain real-time displacement data of the support structure monitoring points," "Query historical rainfall records and geological survey data," and "Evaluate the design safety factor of the support structure." Then, it locates entities such as "support pile L-12" and "tilt monitoring point S-08" in the digital twin and obtains their current displacement value, tilt angle, concrete strain value, and other real-time monitoring data. Simultaneously, it retrieves the rainfall intensity sequence and groundwater level change data from the past 24 hours through a spatiotemporal index, collectively forming a subset of entity state parameters.

[0072] Step S20: Associate and integrate domain knowledge information with the subset of entity state parameters and sub-problems to establish contextual information for sub-problems;

[0073] In this embodiment, the engineering management system constructs a contextual environment through knowledge retrieval and data fusion technologies, providing comprehensive information support for subsequent reasoning. Domain knowledge information refers to structured professional knowledge stored in a knowledge graph, including technical specifications, process standards, and case libraries. Contextual information refers to a comprehensive representation structure that integrates real-time data, domain knowledge, and problem semantics, organized, for example, in the form of text information or a contextual graph.

[0074] Specifically, the engineering management system vectorizes sub-problems through a knowledge retrieval engine and searches for semantically similar knowledge nodes in the domain knowledge graph using approximate nearest neighbor search. The entity association module then employs an attention-based neural network to calculate the association weights between entity state parameters and knowledge nodes, establishing cross-modal semantic alignment. The context assembler then organizes the associated elements into a graph structure, where nodes represent entities, parameters, or knowledge concepts, and edges represent semantic relationships, spatial relationships, or logical constraints. A quality validator performs a completeness check on the generated context graph to ensure that key entities and constraints have been correctly included.

[0075] Optionally, the engineering management system retrieves target knowledge units from the domain knowledge base based on the sub-problem. These target knowledge units include business rules, constraints, and / or cases. A semantic matching algorithm is then used to associate the target knowledge units with subsets of entity state parameters, constructing a context graph for the sub-problem. The context graph and the sub-problem are then fused to form contextual information.

[0076] As an optional implementation, knowledge retrieval can employ a multi-granularity semantic matching strategy. The system constructs a multi-layered knowledge graph structure that includes macro-level normative clauses, meso-level process methods, and micro-level parameter constraints. During the retrieval process, the system simultaneously uses three different similarity calculation methods—bag-of-words model, semantic embedding, and graph neural network—to evaluate the relevance of sub-questions to knowledge nodes at the lexical, semantic, and structural levels, respectively. A weighted fusion is used to obtain a comprehensive relevance score, ensuring that the retrieval results include both directly relevant specific clauses and indirectly relevant background knowledge.

[0077] As an alternative implementation, context fusion can employ a knowledge-enhanced representation learning approach. The system uses a pre-trained language model to uniformly encode sub-question text, entity state data, and domain knowledge, learning deep interaction features between them through a cross-modal attention mechanism. Specifically, the system introduces a contrastive learning objective during training, ensuring that representations of different modalities within the same context are as close as possible in the vector space, while representations of heterogeneous contexts are kept far apart. This representation learning method can effectively capture the complex correlation patterns between numerical data and textual knowledge.

[0078] For example, for the sub-problem of "assessing the operational safety of tower cranes under strong wind conditions," the system retrieves relevant knowledge nodes such as "tower crane wind resistance design level," "wind speed warning threshold," and "special weather operation specifications" from the domain knowledge graph. Then, it compares and correlates the "real-time wind speed monitoring value" and "wind speed warning threshold" in the entity state parameters, and correlates the "tower crane model information" and "wind resistance design level" as entity attributes, generating a context graph that includes numerical comparison relationships, specification constraints, and real-time monitoring data.

[0079] Step S30: Using the reasoning module, generate hypothesis information for at least one of the sub-problems based on the context information;

[0080] In this embodiment, the reasoning module derives verifiable hypothetical propositions from contextual information based on a hybrid reasoning mechanism that combines symbolic reasoning and numerical computation. Here, the reasoning module refers to an intelligent reasoning system that integrates rule-based reasoning, case-based reasoning, and quantitative analysis; the hypothetical information refers to the system's speculative conclusions, based on existing information, regarding the causes, trends, or solutions to problems, which are yet to be verified.

[0081] Specifically, the reasoning module searches the rule base for production rules that match the current context pattern, activates the reasoning rules that meet the conditions, collects the premise evidence that supports each rule, calculates the confidence weight of the evidence, and then performs logical reasoning operations to generate candidate hypotheses and calculate the initial confidence. Based on the confidence score and the quality of the evidence, the generated hypotheses are prioritized and sorted, and the most reliable set of hypotheses is output.

[0082] As an optional implementation, rule-based reasoning can employ a combination of fuzzy logic and confidence propagation. The system uses fuzzy set theory to handle uncertain concepts in the context, such as fuzzy predicates like "high wind speed" and "high risk." During the reasoning process, the system establishes a confidence propagation network, using a message-passing algorithm to propagate the support of evidence within the rule network. Optionally, the system introduces a damping factor to control the scope of confidence propagation, avoiding confidence decay caused by excessively long reasoning chains, and employs conflict resolution strategies to handle contradictory reasoning conclusions.

[0083] As an alternative implementation, the system can integrate a numerical inference engine based on a physical model. For problems involving physical processes such as engineering mechanics and fluid mechanics, the inference module calls the built-in physical equation solver. For example, for structural safety assessment problems, the system, based on a finite element analysis model, directly calculates the stress distribution and deformation of the structure according to the load parameters, material properties, and boundary conditions in the context, using these quantitative analysis results as the basis for hypothesis generation. This numerical inference can provide reliable assumptions based on first principles.

[0084] For example, in the deep foundation pit safety monitoring scenario, the inference module generates the hypothesis that "there may be piping risk around the foundation pit" based on the context information of "accelerated displacement rate of support piles" and "abnormal drop in water level of dewatering wells" and the domain rule that "rapid dewatering under quicksand geological conditions may cause soil loss". The initial confidence level of this hypothesis is calculated to be 0.76 based on the displacement change rate and the drop in water level.

[0085] Optionally, based on the context graph in the context information, the engineering management system loads a set of inference rules corresponding to the problem type of the sub-problem in the inference module, applies the inference rule set to perform logical inference on the context graph, and generates at least one candidate hypothesis. The confidence level of the candidate hypothesis is calculated, and hypothesis information is selected from the candidate hypotheses based on the confidence level.

[0086] Furthermore, after decomposing the consultation problem into sub-problems, the engineering management system needs to analyze the dependencies between sub-problems based on entity information and other data within each sub-problem to determine their logical order. The system can use these dependencies as constraints to regulate the generation of assumptions, ensuring their reliability. Alternatively, the system can add the dependencies between sub-problems to the context of preceding and following problems, and then generate assumptions sequentially according to the order of the problems corresponding to these dependencies. When generating assumptions for the current sub-problem, the system can combine previously generated assumptions with the corresponding sub-problems to make the assumptions.

[0087] Step S40: Using the digital twin, the hypothetical information is simulated and verified, and a reasoning path is formed based on the simulation verification results;

[0088] In this embodiment, the engineering management system utilizes the simulation capabilities of a digital twin to empirically test hypotheses and construct a traceable reasoning record. Simulation verification refers to the process of testing the rationality of hypotheses by simulating the assumed conditions in the digital twin and observing the system's behavior. The reasoning path refers to structured data that records the complete reasoning process from initial evidence to the final conclusion.

[0089] Specifically, the initial parameters and boundary conditions of the digital twin are configured according to the assumptions, and then a physical process simulation or discrete event simulation is run in a virtual environment through a simulation engine. The simulation output data is extracted and compared with the expected results of the assumptions. Key decision points, intermediate results, and final conclusions in the verification process are organized into a structured reasoning path document. Specifically, the engineering management system calculates the simulation parameters of the digital twin based on the assumption information. Based on these simulation parameters and the entity model of the digital twin, a simulation model corresponding to the assumption information is constructed. According to preset verification rules, the target confidence level of the assumption information is calculated by comparing the simulation model with the assumption information. Based on the simulation verification results corresponding to the target confidence level, the target assumption information is determined, and a reasoning path is formed based on the target assumption information.

[0090] As an alternative implementation, simulation verification can employ a robustness testing method based on sensitivity analysis. The system not only runs simulations under the assumed parameter conditions but also performs Latin hypercube sampling in the parameter space, constructing a surrogate model of the input parameters and output response through multiple simulation experiments. Based on this surrogate model, the system calculates the conditional probability of the hypothesis's validity and identifies the most critical influencing factors. This statistical verification method can effectively evaluate the robustness of the hypothesis under different operating conditions.

[0091] As an alternative implementation, the system can achieve parallel verification of multiple hypotheses and dynamic path optimization. When multiple competing hypotheses exist, the system creates an independent simulation instance for each hypothesis and executes the verification task in parallel through a high-performance computing cluster. During the verification process, the system monitors the verification progress and intermediate results of each hypothesis in real time and dynamically adjusts the allocation of computing resources. Simultaneously, the path optimizer dynamically simplifies the recorded content of the inference path based on the importance of the verification results, ensuring that the final generated inference path is both complete and concise.

[0092] For example, to verify the hypothesis that "increasing the pumping rate of the dewatering well can control the deformation of the foundation pit," the system gradually increases the pumping rate of the dewatering well from its current value to the design maximum value in the digital twin, runs a seepage-stress coupling simulation model, and records the entire process of the foundation pit deformation changing with the pumping rate. The simulation results show that when the pumping rate increases to 80% of the design value, the foundation pit deformation tends to stabilize, thus confirming the hypothesis. The verification process and key data points are recorded in the inference path.

[0093] Step S50: Synthesize a thought chain based on the reasoning path, and generate a consultation result for the consultation question based on the thought chain.

[0094] It's important to note that Chain of Thought (CoT) is a technique used to encourage large language models to answer difficult questions through "thinking." It involves breaking down logically complex problems and forming a complete thought process through a series of logically related steps. The aim is to show large language models a small number of examples and explain the reasoning process through these examples, allowing the models to learn to incorporate the reasoning process when generating answers, leading to more accurate results. Chain of Thought is divided into explicit and implicit chains. Implicit chains refer to the machine-readable structured process data that the system relies on for internal reasoning. This data is invisible to the user and represents the system's internal thought process, serving as the computational foundation for correct reasoning and reliable verification. This ensures the logical rigor, traceability, and reproducibility of the reasoning process, enabling the system to manage and execute complex reasoning steps. Explicit chains, on the other hand, are human-readable natural language descriptions that the system translates from implicit chains to communicate with the user. It provides interpretability and transparency, clearly showing users how the conclusions are derived step by step from known data, domain rules, and simulation verification, thereby building trust, assisting decision-making, and facilitating experts to review and question the reasoning process.

[0095] In this embodiment, the engineering management system transforms the structured reasoning path into an easily understandable reasoning narrative, i.e., a thought chain, and generates final decision recommendations. This thought chain can be either an explicit thought chain or an implicit thought chain.

[0096] Specifically, the engineering management system analyzes the logical dependencies in the reasoning path, identifying key reasoning nodes and turning points. The semantic enhancer then adds natural language descriptions to the numerical data and logical relationships, enhancing the readability of the narrative. The results assembler then selects an appropriate report template based on the question type, integrating the thought process with specific conclusions into a structured report. Finally, the quality checker verifies the consistency and completeness of the generated consultation results, ensuring the professionalism of the final output.

[0097] As an alternative implementation, thought chain synthesis can employ a deep narrative generation method based on logical templates. The system constructs a logical template library containing various reasoning modes, such as causal reasoning templates, comparative analysis templates, and decision-making trade-off templates. During generation, the system first identifies the type of reasoning mode used in the reasoning path, then instantiates the corresponding logical template, filling in the appropriate positions of the template with specific parameter values, evidence, and conclusions. Optionally, the system can also introduce rhetorical structure theory to optimize narrative coherence, using appropriate conjunctions and transitional sentences to naturally connect each reasoning step.

[0098] As an alternative implementation, the consultation results generation can integrate multimodal interactive display technology. The system not only generates text-based consultation reports but also automatically creates a visual interface that supports interactive exploration. Users can view key time points in the reasoning process using a timeline control, delve into detailed information about specific reasoning steps through focused links, and compare the verification results of different hypotheses using a comparison view. Simultaneously, the system provides a voice-guided navigation function, using speech synthesis technology to convert core conclusions and key reasoning steps into audio narration, meeting information acquisition needs in different scenarios.

[0099] Optionally, LLM can construct corresponding answer texts based on different logical nodes in the thought chain and / or the dependencies between logical nodes, and combine the answer texts into a consultation answer.

[0100] For example, for a complete consultation question on foundation pit safety assessment, the system generates the following thought chain: "Monitoring data shows that the cumulative displacement of the support piles has reached 45mm, exceeding the alarm value of 30mm; geological data indicates the presence of a silt layer in the area; simulation verification shows that continuous pumping will accelerate silt loss; based on comprehensive judgment, the current deformation is caused by piping triggered by dewatering; it is recommended to immediately adjust the dewatering plan and strengthen deformation monitoring." Based on this thought chain, the system generates a complete consultation report including risk level, cause analysis, treatment recommendations, and prognosis assessment, and uses visual charts to show the correlation between displacement development and pumping volume.

[0101] This application embodiment decomposes the received consultation question into sub-questions, locates and obtains the entity state parameter subset corresponding to the sub-question in the digital twin, and associates and integrates domain knowledge information with the entity state parameter subset and the sub-question to establish context information for the sub-question. This enables the system to perform reasoning under the dual constraints of specific domain rules and specific project status. Then, the reasoning module generates hypothesis information, and the hypothesis information is verified through the digital twin to form a reasoning path. Based on the reasoning path, a thought chain is synthesized and a consultation result is generated, thereby ensuring that the output conclusion not only has logical coherence, but also highly matches the actual state of the engineering entity and domain specifications, significantly improving the reliability and professionalism of the system output results.

[0102] Based on the same inventive concept, this application also provides a second embodiment, referring to... Figure 2 , Figure 2 This is a flowchart illustrating the second embodiment of the question-and-answer system control method based on engineering management in this application.

[0103] In this embodiment, the question-and-answer system control method based on engineering management further includes steps S11 to S13:

[0104] Step S11: Acquire real-time image data;

[0105] Step S12: Perform object recognition and state analysis on real-time image data to extract entity state information of the construction site;

[0106] Step S13: Match the entity state information with the corresponding entity in the digital twin, and update the entity state parameters in the digital twin according to the matching result.

[0107] In this embodiment, to achieve real-time dynamic management of construction projects by the engineering management system, the system can also acquire real-time image data of the construction site. Through depth analysis, object recognition, and other methods, it performs object recognition and state analysis on the real-time image data, extracting entity state information such as object location, site layout information, and depth point cloud data. Furthermore, by matching the entity state information with the corresponding entities in the currently stored digital twin, the engineering management system can update the entity state parameters in the digital twin based on the matching results.

[0108] Furthermore, the digital twin deployed in the engineering management system also includes a digital twin model, such as a BIM model, and entity status parameters associated with each model node of the digital twin model. These entity status parameters may include the model parameters corresponding to the model node in the digital twin model, and / or relevant information associated with the model node, such as construction precautions or electronic archives such as national construction standards.

[0109] Optionally, based on the entity state parameters of the digital twin, risk information corresponding to model nodes in the digital twin is identified, association information of the digital twin and mapping relationship with model nodes are obtained, and based on the mapping relationship, annotation information of model nodes is generated according to risk information and association information, and the annotation information is added to the digital twin to form entity state parameters of the digital twin.

[0110] Optionally, the project management system also includes an electronic records management system, which, after receiving electronic records input by the user, can construct the association between the electronic records and the model nodes in the digital twin model based on the record entity information determined by semantic analysis of the electronic records.

[0111] Since the system described in Embodiment 2 of this application is a system used to implement the method of Embodiment 1 of this application, those skilled in the art can understand the specific structure and variations of the system based on the method described in Embodiment 1 of this application, and therefore will not be described again here. All systems used in the method of Embodiment 1 of this application fall within the scope of protection of this application.

[0112] This application provides a question-and-answer system control device based on engineering management. The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the question-and-answer system control method based on engineering management in the first embodiment described above.

[0113] The following is for reference. Figure 3 This document illustrates a structural schematic diagram of a question-and-answer system control device suitable for implementing embodiments of this application. The question-and-answer system control device based on engineering management in these embodiments may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 3 The illustrated question-and-answer system control device based on engineering management is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0114] like Figure 3As shown, the question-and-answer system control device based on engineering management may include a processing unit 1001 (e.g., a core processor, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 1002 or a program loaded from storage device 1003 into random access memory (RAM) 1004. The random access memory 1004 also stores various programs and data required for the operation of the question-and-answer system control device based on engineering management. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the engineering-managed question-and-answer system control device to communicate wirelessly or wiredly with other devices to exchange data. Although an engineering-managed question-and-answer system control device with various systems is shown in the figure, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented or possessed alternatively.

[0115] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0116] The question-and-answer system control device based on engineering management provided in this application, employing the question-and-answer system control method based on engineering management in the above embodiments, can solve the technical problem of low answer credibility of large language models accessed in engineering management systems. Compared with the prior art, the beneficial effects of the question-and-answer system control device based on engineering management provided in this application are the same as the beneficial effects of the question-and-answer system control method based on engineering management provided in the above embodiments, and other technical features in this question-and-answer system control device are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0117] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0118] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0119] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the question-and-answer system control method based on engineering management in the above embodiments.

[0120] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, radio frequency (RF), etc., or any suitable combination thereof.

[0121] The aforementioned computer-readable storage medium may be included in an engineering management-based question-and-answer system control device; or it may exist independently and not be assembled into an engineering management-based question-and-answer system control device.

[0122] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by a question-and-answer system control device based on engineering management, the question-and-answer system control device based on engineering management performs the following actions: receiving a consultation question, decomposing the consultation question into at least one sub-question, locating and obtaining a subset of entity state parameters corresponding to the sub-question in a digital twin, associating and fusing domain knowledge information with the subset of entity state parameters and the sub-question, establishing context information for the sub-question, generating hypothesis information for at least one sub-question based on the context information through a reasoning module, performing simulation verification on the hypothesis information through a digital twin, forming a reasoning path based on the simulation verification results, synthesizing a thought chain based on the reasoning path, and generating a consultation result for the consultation question based on the thought chain.

[0123] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0124] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion 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 indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0125] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0126] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., computer programs) for executing the above-described question-answering system control method based on engineering management. This addresses the technical problem of low answer reliability from large language models accessed in engineering management systems. Compared to existing technologies, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the question-answering system control method based on engineering management provided in the above embodiments, and will not be elaborated upon here.

[0127] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A question-and-answer system control method based on engineering management, characterized in that, The method includes the following steps: Receive a consultation question, decompose the consultation question into at least one sub-question, and locate and obtain the subset of entity state parameters corresponding to the sub-question in the digital twin; The domain knowledge information is associated and fused with the subset of entity state parameters and the sub-problem to establish the context information of the sub-problem; The reasoning module generates at least one hypothesis information for the sub-problem based on the context information. The dependency relationship between the sub-problems is analyzed based on the entity information in the sub-problems, and the dependency relationship is used as constraint information to constrain the generation of hypothesis information. When generating hypothesis information for the current sub-problem, the hypothesis is made by combining the previously generated hypothesis information and the corresponding sub-problems. Based on the assumed information, the simulation parameters of the digital twin are calculated; Based on the simulation parameters and the physical model of the digital twin, a simulation model corresponding to the hypothetical information is constructed. According to the preset verification rules, the target confidence level of the hypothesis information is calculated by comparing the simulation model with the hypothesis information. Based on the simulation verification results corresponding to the target confidence level, the target hypothesis information is determined, and an inference path is formed based on the target hypothesis information; Based on the reasoning path, a thought chain is synthesized, and a consultation result for the consultation question is generated based on the thought chain.

2. The question-and-answer system control method based on engineering management as described in claim 1, characterized in that, Before the steps of receiving the consultation question, decomposing the consultation question into at least one sub-question, and locating and obtaining the subset of entity state parameters corresponding to the sub-question in the digital twin, the method further includes: Acquire real-time image data; Object recognition and state analysis are performed on the real-time image data to extract entity state information of the construction site; The entity state information is matched with the corresponding entity in the digital twin, and the entity state parameters in the digital twin are updated according to the matching result.

3. The question-and-answer system control method based on engineering management as described in claim 1, characterized in that, The steps of receiving a consultation question, decomposing the consultation question into at least one sub-question, and locating and obtaining a subset of entity state parameters corresponding to the sub-question in the digital twin include: Natural language processing is performed on the consultation question to identify entity information in the consultation question; Based on the domain knowledge base, the entity information is mapped to a predefined conceptual model; Based on the entity relationship information in the conceptual model, the consultation problem is decomposed into at least one of the sub-problems; Based on the correspondence between the sub-problems and the entity information, the subset of entity state parameters is queried in the entity model of the digital twin.

4. The question-and-answer system control method based on engineering management as described in claim 1, characterized in that, The step of associating and fusing domain knowledge information with the subset of entity state parameters and the sub-problem to establish context information for the sub-problem includes: Based on the sub-problem, a target knowledge unit is retrieved from the domain knowledge base, wherein the target knowledge unit includes business rules, constraints and / or cases; By using a semantic matching algorithm, the target knowledge unit is associated with the subset of entity state parameters to construct the context graph of the sub-problem; The context information is formed by fusing the context graph and the sub-problems.

5. The question-and-answer system control method based on engineering management as described in claim 4, characterized in that, The step of generating hypothesis information for at least one of the sub-problems based on the context information through the reasoning module includes: In the reasoning module, a set of reasoning rules corresponding to the problem type of the sub-problem is loaded; By applying the inference rule set, logical reasoning is performed on the context graph to generate at least one candidate hypothesis; Calculate the confidence level of the candidate hypothesis, and generate the hypothesis information based on the confidence level and the candidate hypothesis.

6. The question-and-answer system control method based on engineering management as described in claim 1, characterized in that, The steps of synthesizing a thought chain based on the reasoning path and generating a consultation result for the consultation question based on the thought chain include: Extract key reasoning nodes and logical connections from the reasoning path to construct a logical chain; Using a templated natural language generation algorithm, the logical chain is converted into a structured text description, forming the thought chain; The consultation results are generated based on the thought chain using a large model.

7. The question-and-answer system control method based on engineering management as described in claim 1, characterized in that, Before the steps of receiving the consultation question, decomposing the consultation question into at least one sub-question, and locating and obtaining the subset of entity state parameters corresponding to the sub-question in the digital twin, the method further includes: Based on the entity state parameters of the digital twin, identify the risk information corresponding to the model nodes in the digital twin; Obtain the association information of the digital twin and its mapping relationship with the model nodes; Based on the mapping relationship, and according to the risk information and the association information, the annotation information of the model nodes is generated; The annotation information is added to the digital twin to form the entity state parameters of the digital twin.

8. A question-and-answer system control device based on engineering management, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the question-and-answer system control method based on engineering management as described in any one of claims 1 to 7.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the question-and-answer system control method based on engineering management as described in any one of claims 1 to 7.

Images