A structural health monitoring scheme intelligent generation method based on workflow orchestration

Abstract

This invention discloses an intelligent generation method for structural health monitoring schemes based on workflow orchestration. It automates the generation of structural health monitoring schemes using FastGPT workflow orchestration, innovatively introducing FastGPT as the core of the workflow driver to guide semantic parsing, task allocation, content retrieval, and scheme output. The standardized construction and high-hit-rate design of the structural health monitoring knowledge base involves creating a standardized scheme knowledge base where all scheme documents are divided into logical segments, using custom delimiters to control content boundaries; improving semantic retrieval accuracy and ensuring the contextual integrity of content called by FastGPT, avoiding fragmented references, and improving response accuracy. An independent management and accurate referencing mechanism for the monitoring equipment list is implemented, constructing an independent monitoring equipment knowledge base to avoid incorrect associations between equipment technical indicators and images; FastGPT searches this knowledge base when referencing equipment information, ensuring the accuracy and consistency of the referenced content in the output.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of structural health monitoring, in particular to a structural health monitoring scheme intelligent generation method based on workflow arrangement. BACKGROUND

[0002] In the field of structural health monitoring, traditional monitoring scheme design mainly relies on the following several ways: 1. Expert-led manual design. Currently, the development of most structural health monitoring systems relies on structural engineering experts and sensor system experts to manually configure according to engineering actual conditions, which mainly includes: monitoring object identification (such as bridges, slopes, tunnels, etc.); sensor type selection (such as strain gauges, accelerometers, displacement meters, etc.); monitoring point design; data acquisition and processing scheme setting; alarm threshold and early warning model setting. This process usually takes a long time, relies on intensive domain knowledge, and has poor universality and adaptability.

[0003] 2. Semi-automatic tools based on templates. Some engineering consulting or equipment supply companies have developed template-driven scheme assistance tools, which generate standardized documents through parameterized input, such as: Excel table-driven design logic; visual drag-and-drop configuration interface (such as limited monitoring types and object systems). However, this type of tool has limited scope of application and is difficult to support flexible strategy combinations in complex scenarios, and mainly serves as a post-document output, lacking intelligent generation capabilities.

[0004] 3. Automatic configuration systems based on rule engines. Some systems introduce rule engines to recommend scheme components based on a set of rule libraries. For example: if it is "highway bridge + wind load monitoring", recommend a certain type of anemometer; if "steel structure + fatigue monitoring", recommend a vibration analysis algorithm. However, this method still relies on expert pre-set rules and lacks the ability to generalize to new scenarios and new structural forms.

[0005] 4. Exploratory attempts of AI-assisted analysis. In recent years, some research institutions have attempted to use machine learning methods to identify abnormalities and locate damage in structural monitoring data. However, these methods focus on the data analysis stage and rarely apply to the monitoring scheme design stage. There is no system that can automatically generate a complete monitoring scheme from task objectives, combined with structural characteristics, monitoring requirements, and scenario types.

[0006] Although structural health monitoring has been widely applied to major engineering structures such as bridges, slopes, and tunnels, existing monitoring scheme design methods have the following shortcomings: 1. High dependence on manual experience and low intelligence. The current development of structural health monitoring schemes is usually completed manually by engineering technicians, relying on their experience in judging the structural characteristics, use environment, and functional requirements of the monitoring object. This approach has the following problems: (1) Lack of a unified design logic, making standardization difficult; (2) Strong reliance on experts and subjective experience differences lead to unstable solution quality; (3) Difficult to respond quickly to large-scale projects or new scenarios.

[0007] 2. Templated and static rules are difficult to adapt to complex and ever-changing engineering scenarios. Some semi-automatic tools or rule engines use preset templates or rule bases for generation, lacking an understanding of contextual semantics, and have the following drawbacks: (1) Poor scene generalization ability and poor scalability; (2) It is not possible to flexibly combine sensors and algorithms according to real-time task objectives or structural types.

[0008] 3. The lack of end-to-end workflow support makes it difficult to achieve full-process automation. The existing solution generation process is usually discrete, including: manual structural modeling → manual selection → document editing → report output. These steps are often completed by multiple systems in a decentralized manner, making it impossible to achieve integrated process generation driven by monitoring objectives, resulting in problems of information fragmentation and low collaboration efficiency.

[0009] 4. The design process and results lack knowledge traceability and reusability. Traditional solution design processes often lack structured records and knowledge accumulation, making it impossible to quickly reuse historical experience or trace the basis behind the solution design.

[0010] This invention aims to solve the following core technical problems existing in the design of current structural health monitoring schemes: 1. The lack of an intelligent solution generation mechanism means that the current monitoring solution design relies heavily on human experience, making it difficult to standardize and deploy in batches, resulting in low efficiency and high requirements for engineers' technical skills.

[0011] 2. The solution generation process is fragmented and lacks integrated support driven by the process. Traditional methods cannot start from the "monitoring target" and automatically complete steps such as structural type identification, sensor selection, deployment suggestions, and communication configuration. The design process is scattered and difficult to coordinate.

[0012] 3. Lacking semantic understanding and reasoning capabilities, it cannot adapt to complex and diverse monitoring scenarios. Existing systems cannot understand task requirements expressed in natural language (such as "fatigue monitoring of bridges"), let alone perform scheme reasoning and dynamic configuration.

[0013] 4. It is difficult to continuously learn and optimize, build a structured knowledge graph or model library, and lack a continuous improvement mechanism based on historical project data. Existing methods are difficult to form a reusable and self-learning "solution generation engine". Summary of the Invention

[0014] The purpose of this invention is to provide an intelligent generation method for structural health monitoring schemes based on workflow orchestration. By introducing a workflow orchestration mechanism based on a large language model, the invention achieves intelligent processing of the entire process of structural health monitoring schemes, from natural language requirement input to standardized graphic and text scheme output. This solves key problems in existing technologies, such as reliance on human experience, chaotic information organization, difficulty in knowledge reuse, and incorrect equipment reference.

[0015] To achieve the above objectives, the present invention adopts the following technical solution: A method for intelligently generating structural health monitoring schemes based on workflow orchestration is disclosed. The specific steps of this method are as follows: Step 1: User inputs solution requirements to trigger task startup. The user inputs monitoring solution requirements through the front-end interface, including the type of structure and monitoring factors. The system automatically starts the preset workflow accordingly. Step 2: Check if the user input is missing key information. The system uses AI nodes to perform semantic review on the user's input solution requirements to confirm whether the list of monitoring factors is clear and outputs a conclusion on whether monitoring factors are missing. If no monitoring factors are provided, an error message is given to guide the user to complete the input and ensure the effective start of the task. Step 3: Extract keywords. The system uses an AI big data model to perform semantic analysis and keyword extraction on the user's input solution requirements, and automatically generates structured parameters for the structure type and monitoring factor list. Step 4: Knowledge base search. Use extracted keywords to match the standardized solution knowledge base; each solution document is cut into complete paragraphs according to logical semantics using delimiters; FastGPT calls the matched paragraphs according to user input, and splices, supplements, and expands the content according to the context to ensure that each generated paragraph has engineering logic and semantic integrity. Step 5: Generating and Outputting the Solution Content. Combining the knowledge base references, the system guides FastGPT to generate complete monitoring solution content through configurable AI dialogue nodes and prompt word templates. The content covers: project overview, necessity and significance of monitoring, basis for preparation, monitoring content, system design, monitoring points and equipment, system data collection and transmission, data processing and hierarchical early warning, and software system. Finally, the generated content is integrated and output as an HTML page.

[0016] Preferably, in step 1, the structure type includes bridges and slopes, and the monitoring factors include cracks, settlement, and strain.

[0017] Preferably, in step 2, if the input information is complete, the system will continue to extract keywords, and the extracted keywords will serve as the basis for subsequent knowledge base retrieval and content generation.

[0018] Preferably, in step 4, for the monitoring site and equipment section, all equipment information, namely images and parameters, is retrieved only from a separate "monitoring equipment list" knowledge base.

[0019] Preferably, in step 4, the knowledge base search includes two paths. The first path retrieves sensor equipment information related to the monitoring factors from the monitoring equipment list knowledge base, including equipment name, technical parameters, and images. The second path retrieves solution content paragraphs that match the structure type and monitoring factors from the standardized solution knowledge base, ensuring that the generated results are professional and logically complete.

[0020] Preferably, the monitoring equipment information is decoupled from the text scheme and maintained independently to avoid mis-reference of equipment due to contextual confusion. An independent monitoring equipment library is maintained, which includes monitoring items, equipment names, technical specifications, and equipment images. When FastGPT references equipment, it is limited to the data of this module to avoid referencing incorrect images or technical information.

[0021] Preferably, in step 1, users can input the type of structure and monitoring factors, which can be in the form of keywords, phrases or natural language task descriptions; and a knowledge base module for input parameter verification and standardization processing of monitoring equipment lists is provided.

[0022] Preferably, in step 5, an HTML page is output, which supports tables, images, flowcharts, and numbered paragraph elements, making it easy to display and integrate with the system.

[0023] Compared with existing technologies, the beneficial effects of this invention are as follows: By introducing a workflow orchestration mechanism based on a large language model, the entire process of structural health monitoring solutions, from natural language requirement input to standardized graphic and text solution output, is intelligently processed. This solves key problems in existing technologies such as reliance on human experience, chaotic information organization, difficulty in knowledge reuse, and equipment misuse, achieving the following beneficial technical effects: 1. This application enables the automatic generation of structural health monitoring plans, significantly improving compilation efficiency. Compared with existing methods of manually writing or combining plans, this application drives the language model to automatically execute the "keyword extraction → knowledge retrieval → content generation → HTML output" process through a preset workflow template, significantly improving efficiency.

[0024] 2. By employing a standardized knowledge base segmentation mechanism, the accuracy and completeness of content generation are improved. In traditional technologies, large language models often encounter problems such as incomplete reference fragments or semantic breaks when calling unstructured documents, resulting in generated content lacking engineering logic. This invention adopts a semantic-based logical paragraph segmentation mechanism (custom delimiters) to structurally organize standardized solution documents into semantic units. When referencing knowledge, the language model uses "segments" as the smallest unit, significantly improving content accuracy and coherence, and avoiding content being taken out of context or generating ambiguous expressions.

[0025] 3. The equipment list knowledge base is designed independently, effectively avoiding equipment referencing errors. In existing technologies, AI models easily confuse sensor names, technical parameters, or images with information from other contexts, leading to incorrect equipment descriptions in the output solution and posing professional risks. This invention separates equipment information from the main knowledge base, establishing an independent monitoring equipment knowledge base, and restricts the language model's access to equipment information to this base, ensuring the accuracy of equipment names, technical parameters, and images, and meeting engineering usability requirements.

[0026] 4. The overall technical solution reduces reliance on human experience and promotes automatic knowledge reuse and sharing. Traditional structural health monitoring solutions heavily rely on the experience of senior technicians, leading to difficulties in knowledge reuse and rapid knowledge loss. This invention, by organizing and storing existing high-quality solutions in blocks and combining them with language model reasoning capabilities, achieves structured knowledge accumulation and intelligent retrieval, promoting the automatic dissemination and standardized implementation of engineering knowledge. Attached Figure Description

[0027] Figure 1 This is a flowchart of the system processing of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Example

[0029] This invention provides a method for generating structure monitoring data analysis reports based on workflow and industry knowledge base. Through configurable analysis templates and knowledge enhancement mechanisms, it enables one-click generation from raw data to professional reports.

[0030] 1. System Architecture This technical solution adopts a modular architecture design, leveraging AI large-scale model capabilities and a workflow engine to build an intelligent generation platform for structural health monitoring solutions. The system mainly includes the following functional modules: (1) User input module It supports user input of structure type and monitoring factors, which can be in the form of keywords, phrases or natural language task descriptions; and provides necessary input parameter validation and standardization processing.

[0031] (2) FastGPT workflow orchestration module By introducing a large language model as the core of workflow scheduling, subtasks such as "structure type identification → monitoring factor analysis → knowledge matching → content generation → format output" are broken down into ordered workflow nodes.

[0032] (3) Structural health monitoring knowledge base management module The knowledge base is based on standardized schemes for various structure types, and uses custom delimiters to divide it into blocks according to logical paragraphs, thereby enhancing the accuracy of paragraph-level semantic calls.

[0033] (4) Monitoring Equipment List Knowledge Base Module An independent monitoring equipment library is maintained, including monitoring items, equipment names, technical specifications, and equipment images. When FastGPT references equipment, it is limited to data from this module to avoid referencing incorrect images or technical information.

[0034] (5) Image and text output module The generated monitoring plan is organized into a structured HTML page with graphics and text, supporting elements such as tables, images, flowcharts, and numbered paragraphs, which facilitates display and system integration.

[0035] Figure 1 This is a system processing flowchart for the patent. Starting with user input, the flowchart uses artificial intelligence (AI) semantic parsing and knowledge-based task orchestration to automatically generate and output a structural health monitoring solution. Specifically, it includes the following steps: First, users input their monitoring requirements through the system interface, typically including the type of structure (such as bridges, slopes, etc.) and monitoring factors (such as cracks, settlement, strain, etc.). After receiving the input, the system calls upon its built-in AI dialogue node to perform semantic parsing on the input content and identify key information.

[0036] After parsing, the system includes a checkpoint to determine if the user input is missing the crucial monitoring factor. If a missing factor is detected, the system immediately returns an error message, guiding the user to supplement the necessary parameters to ensure the smooth progress of subsequent processes. If the input information is complete, the system continues with keyword extraction, which will serve as the basis for subsequent knowledge base retrieval and content generation.

[0037] Subsequently, the system initiates two knowledge retrieval paths in parallel. The first path retrieves sensor equipment information related to the monitoring factors from the monitoring equipment list knowledge base, including equipment names, technical parameters, and images. The second path retrieves solution content paragraphs matching the structure type and monitoring factors from the standardized solution knowledge base, ensuring that the generated results are professional and logically complete.

[0038] After knowledge retrieval is complete, the system enters the solution generation phase. Based on the reference results from the standardized solution knowledge base and the monitoring equipment list knowledge base, a complete monitoring site layout and equipment deployment plan is automatically generated. Simultaneously, AI, combined with references to the standardized solution knowledge base, generates a complete solution text including project overview, monitoring necessity and significance, basis for compilation, monitoring content, system design, system data collection and transmission, data processing and tiered early warning, and software system.

[0039] Finally, the system organizes all generated content into a structured HTML document with both text and images, achieving a visually appealing and formatted output.

[0040] This flowchart comprehensively illustrates the entire automated process from user requirement input to complete monitoring solution output, demonstrating the technical advantages of this invention in task-driven, semantic recognition, knowledge retrieval, and structured output.

[0041] To better understand the technical solution of this invention, the following is combined with... Figure 1 The flowchart shown illustrates a detailed description of specific embodiments of the present invention. This embodiment uses a bridge structure as a typical scenario and "displacement monitoring, stress monitoring, crack monitoring, and cable tension monitoring" as monitoring objects, describing how a monitoring scheme can be automatically generated based on user input and a preset process.

[0042] 1. Task initiation and semantic recognition Users can input the following requirements through the system front end: "Generate a bridge monitoring plan, including displacement monitoring, stress monitoring, crack monitoring, and cable tension monitoring." The system first calls the large language model module based on qwen-plus to perform semantic parsing on the above natural language instructions and determines that there are monitoring factors in the user's requirement description. Then, it uses the qwen-plus large language model to automatically extract the following key information: (1) Structure type: Bridge; (2) List of monitoring factors: displacement monitoring, stress monitoring, crack monitoring, cable force monitoring.

[0043] 2. Knowledge base matching and intelligent retrieval The system uses the extracted structure type and monitoring factors as search keywords to retrieve information from two types of knowledge bases: (1) Standardized scheme knowledge base: It calls standardized document paragraphs containing multi-dimensional combinations of content such as "bridge structure + displacement monitoring" and "bridge structure + stress monitoring". All content is organized into blocks according to a logical structure using a custom delimiter "=====". FastGPT will combine and derive these blocks to generate structurally complete paragraph content.

[0044] (2) Monitoring Equipment Knowledge Base: For each of the four monitoring factors, suitable sensor devices are searched, such as: ① Displacement monitoring: wire displacement sensor, displacement meter, and GNSS integrated unit; ② Stress monitoring: strain gauges, rebar gauges; ③ Crack monitoring: crack gauge, data acquisition system; ④ Cable force monitoring: magnetic flux sensor, magnetic flux acquisition system, accelerometer, cloud vibration acquisition instrument; The name, technical parameters, and images of each type of equipment are extracted by the system from the equipment knowledge base, avoiding duplicate references and incorrect calls.

[0045] 3. Solution Content Generation By integrating knowledge segments with equipment information, the system automatically generates structured solution content. The core modules include: (1) Project Overview and Basis for Compilation: The standard paragraphs are used to generate content such as the background of the bridge structure, the necessity of monitoring, and the design basis.

[0046] (2) Monitoring objectives and contents: The technical objectives and coverage of monitoring displacement, stress, cracks, and cable force in the plan should be clearly defined.

[0047] (3) Sensor placement and installation methods: Based on the bridge structure layout diagram, the system generates placement suggestions for each type of sensor (such as placing GNSS reference stations on the top of the piers and strain gauges on the main beams), and automatically provides installation instructions.

[0048] (4) Data acquisition and transmission: The system recommends using a wireless gateway + 4G / 5G communication module for data transmission, with the sampling frequency set according to the type of equipment (e.g., 1Hz for crack monitoring, 0.1Hz for cable tension monitoring, etc.).

[0049] (5) Data processing and early warning strategies: The system uses reference thresholds based on national standards (such as the "Technical Specification for Bridge Structure Monitoring") to set multiple early warning levels and generate a data processing logic diagram.

[0050] 4. Structured Output The generated text content, after being structured and organized, is uniformly output as an HTML-formatted graphic document, with specific features including: (1) The paragraph hierarchy is clear, and a directory structure is used; (2) The sensor devices cited are listed in tabular form, with accompanying pictures and technical parameters; (3) The system flowchart and layout diagram are automatically embedded.

[0051] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligently generating structural health monitoring schemes based on workflow orchestration, characterized in that: The specific steps of the intelligent generation method for the structural health monitoring solution based on workflow orchestration are as follows: Step 1: User inputs solution requirements to trigger task startup. The user inputs monitoring solution requirements through the front-end interface, including the type of structure and monitoring factors. The system automatically starts the preset workflow accordingly. Step 2: Check if the user input is missing key information. The system uses AI nodes to perform semantic review on the user's input solution requirements to confirm whether the list of monitoring factors is clear and outputs a conclusion on whether monitoring factors are missing. If no monitoring factors are provided, an error message is given to guide the user to complete the input and ensure the effective start of the task. Step 3: Extract keywords. The system uses an AI big data model to perform semantic analysis and keyword extraction on the user's input solution requirements, and automatically generates structured parameters for the structure type and monitoring factor list. Step 4: Knowledge base search. Use extracted keywords to match the standardized solution knowledge base; each solution document is cut into complete paragraphs according to logical semantics using delimiters; FastGPT calls the matched paragraphs according to user input, and splices, supplements, and expands the content according to the context to ensure that each generated paragraph has engineering logic and semantic integrity. Step 5: Generating and Outputting the Solution Content. Combining the knowledge base references, the system guides FastGPT to generate complete monitoring solution content through configurable AI dialogue nodes and prompt word templates. The content covers: project overview, necessity and significance of monitoring, basis for preparation, monitoring content, system design, monitoring points and equipment, system data collection and transmission, data processing and hierarchical early warning, and software system. Finally, the generated content is integrated and output as an HTML page.

2. The intelligent generation method for structural health monitoring schemes based on workflow orchestration according to claim 1, characterized in that: In step 1, the structure types include bridges and slopes, and the monitoring factors include cracks, settlement, and strain.

3. The intelligent generation method for structural health monitoring schemes based on workflow orchestration according to claim 1, characterized in that: In step 2, if the input information is complete, the system will continue to extract keywords, and the extracted keywords will serve as the basis for subsequent knowledge base retrieval and content generation.

4. The intelligent generation method for structural health monitoring schemes based on workflow orchestration according to claim 1, characterized in that: In step 4, for the monitoring site deployment and equipment section, all equipment information, namely images and parameters, is retrieved only from the independent "Monitoring Equipment List" knowledge base.

5. The intelligent generation method for structural health monitoring schemes based on workflow orchestration according to claim 1, characterized in that: In step 4, the knowledge base search includes two paths. The first path retrieves sensor equipment information related to the monitoring factors from the monitoring equipment list knowledge base, including equipment names, technical parameters, and images. The second path retrieves solution content paragraphs that match the structure type and monitoring factors from the standardized solution knowledge base, ensuring that the generated results are professional and logically complete.

6. The intelligent generation method for structural health monitoring schemes based on workflow orchestration according to claim 4, characterized in that: The monitoring equipment information is decoupled from the text scheme and maintained independently to avoid mis-reference of equipment due to contextual confusion. An independent monitoring equipment library is maintained, which includes monitoring items, equipment names, technical specifications, and equipment images. When FastGPT references equipment, it is limited to the data of this module to avoid referencing incorrect images or technical information.

7. The intelligent generation method for structural health monitoring schemes based on workflow orchestration according to claim 1, characterized in that: In step 1, users can input the type of structure and monitoring factors, which can be in the form of keywords, phrases or natural language task descriptions; a knowledge base module for input parameter verification and standardization processing and monitoring equipment list is provided.

8. The intelligent generation method for structural health monitoring schemes based on workflow orchestration according to claim 1, characterized in that: In step 5, an HTML page is output, which supports tables, images, flowcharts, and numbered paragraph elements for easy display and system integration.

Images