Workshop equipment remote point inspection method based on digital twinning

Abstract

The invention is suitable for the technical field of workshop digital management, and provides a workshop equipment remote point inspection method based on digital twinning, which comprises the following steps: constructing a three-dimensional digital twinning model associated with physical equipment parts and measuring points, and driving the model to dynamically update through real-time data; receiving and analyzing a multi-modal inspection instruction, and automatically generating a virtual inspection focusing scene taking a target object as a core; on this basis, according to the importance weight of the real-time parameter, a virtual camera inspection path with the goal of maximizing continuous observation of a high-importance area is intelligently planned; and finally, controlling the virtual camera to perform immersive tour along the path, synchronously displaying data and states, and completing abnormality diagnosis and report generation in tour. According to the invention, an active fusion intelligent inspection mode is realized, the immersion, target positioning efficiency and observation pertinence of remote inspection are greatly improved, and the timeliness of abnormality discovery and the accuracy of root positioning are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of digital factory management technology, and in particular to a remote point inspection method for workshop equipment based on digital twins. Background Technology

[0002] In current industrial workshop equipment management, daily inspection and maintenance of production equipment are crucial for ensuring production safety and continuity. Existing technologies generally employ a combination of manual on-site inspections and fixed monitoring. Operators need to periodically visit the equipment site, recording operating parameters through visual inspection and instrument measurements, and relying on experience to judge the equipment's condition. Simultaneously, the workshop typically deploys independent video surveillance systems for real-time monitoring, as well as distributed data acquisition and monitoring control systems to collect process parameters such as current, temperature, and vibration from some key equipment. These systems operate independently, storing the generated data and video streams on different platforms. Managers use multiple independent software interfaces to view equipment status, retrieve monitoring footage, or review historical data reports.

[0003] However, manual inspection relies heavily on personnel experience, resulting in inconsistent inspection quality and an inability to achieve continuous, 24 / 7 monitoring. It also makes it difficult to detect sudden anomalies promptly. When an abnormal temperature is detected in equipment, operators struggle to quickly pinpoint the exact location within the corresponding video feed and observe its operational details. Furthermore, they cannot intuitively understand the interrelationships between parameters on the equipment's three-dimensional structure. This leads to a fragmented inspection process, requiring personnel to perform tedious information comparisons and comprehensive judgments across multiple systems, resulting in low efficiency and a risk of missing crucial clues. Consequently, it fails to meet the demands of modern intelligent manufacturing for transparent equipment status, predictive maintenance, and efficient operation and maintenance.

[0004] In view of this, a remote point inspection method for workshop equipment based on digital twins is proposed. Summary of the Invention

[0005] This invention provides a remote point inspection method for workshop equipment based on digital twins, which is used to solve the problem of difficulty in meeting the requirements of modern intelligent manufacturing for transparent equipment status, predictive maintenance and efficient operation and maintenance.

[0006] This invention provides a remote point inspection method for workshop equipment based on digital twins, comprising: Construct a three-dimensional digital twin model of the workshop equipment, and establish an association between each component in the three-dimensional digital twin model and the corresponding physical measurement point in the data acquisition system; The operating parameters of the workshop equipment are collected in real time, and the operating parameters are matched with the components and physical measurement points associated in the three-dimensional digital twin model, and the three-dimensional digital twin model is driven to update its status. The system receives multimodal inspection commands input by inspection personnel, parses the multimodal commands to determine the target inspection object and inspection intention, and automatically generates a virtual inspection focus scene based on the three-dimensional digital twin model and the target inspection object. Based on the current real-time operating parameters of the workshop equipment, the importance weight of each monitoring parameter is evaluated; with the goal of maximizing the continuous observation effect of the area where the high importance parameters are located, the inspection path of the virtual camera is dynamically planned in combination with the spatial structure of the virtual inspection focusing scene; The virtual camera is controlled to patrol along the inspection path in the virtual inspection focus scene, and the real-time data and status of the corresponding components are displayed simultaneously. During the inspection, anomalies are diagnosed, and a comprehensive inspection report is generated based on the diagnosis results.

[0007] Furthermore, the construction of a three-dimensional digital twin model of the workshop equipment, and the association between each component in the three-dimensional digital twin model and the corresponding physical measurement point in the data acquisition system, includes: Identify the outlines of functional components and measurement point symbols in the acquired technical drawings; For each identified functional component outline, label the corresponding component type; and for each identified measuring point symbol, label the corresponding measuring point identifier. Based on the component type indicated in the annotation, the corresponding 3D component model is retrieved from the preset 3D standard component library; Based on the spatial relationship between the functional component outlines and measuring point symbols in the technical drawings, the called three-dimensional component models are combined to generate a three-dimensional digital twin model of the workshop equipment. Based on the marked measurement point identifiers, establish the data association relationship between each component and measurement point in the three-dimensional digital twin model and the corresponding physical measurement points in the data acquisition system.

[0008] Furthermore, the matching of the operating parameters with the associated components and physical measurement points in the three-dimensional digital twin model includes: The real-time data values ​​and status identifiers corresponding to each physical measurement point in the data acquisition system are parsed from the collected operating parameters. Based on the established relationships in the three-dimensional digital twin model, the real-time data values ​​and status identifiers are mapped to the corresponding three-dimensional component models according to the physical measurement point identifiers.

[0009] Furthermore, the step of parsing the multimodal instructions to determine the target inspection object and inspection intention includes: The multimodal inspection instructions are parsed to extract one or more instruction features that include at least text keywords, image region identifiers, and entity names in voice instructions; Based on a preset equipment knowledge graph, the extracted instruction features are matched with the corresponding workshop equipment, components or measuring points in the three-dimensional digital twin model to determine at least one target inspection object; Based on the parsed instruction modal features and contextual semantics, the inspection intent type expressed by the multimodal inspection instruction is determined, and the inspection intent type is at least one of status query, anomaly location, historical backtracking, and comprehensive inspection.

[0010] Furthermore, the automatic generation of a virtual inspection focus scene based on the three-dimensional digital twin model and the target inspection object includes: The scene center is the three-dimensional component model corresponding to the target inspection object in the three-dimensional digital twin model; The spatial range and associated object set of the virtual inspection focus scene are determined according to the inspection intention type; wherein, the associated object set includes equipment adjacent to the target inspection object in the process flow, pipelines or lines directly connected to the target inspection object, and one or more measuring points characterizing the key operating status of the target inspection object; The three-dimensional models and data measurement points involved in the scene center, spatial range and related object set are cut and recombined from the three-dimensional digital twin model to generate an independent virtual inspection focus scene.

[0011] Furthermore, the step of evaluating the importance weight of each monitoring parameter based on the current real-time operating parameters of the workshop equipment includes: Obtain the current real-time data value, preset safety threshold, and historical trend data of the monitoring parameters; The first evaluation score is calculated based on how close the current real-time data value is to the preset security threshold. Based on the historical trend data, determine whether the monitoring parameter is in an abnormal fluctuation state, and calculate the second evaluation score; By combining the first evaluation score and the second evaluation score, the importance weight of each monitoring parameter is calculated; among them, the monitoring parameter that is closer to the safety threshold and has more abnormal fluctuations has a higher importance weight.

[0012] Furthermore, the step of dynamically planning the inspection path of the virtual camera, with the goal of maximizing continuous observation of the area containing high-importance parameters and in conjunction with the spatial structure of the virtual inspection focusing scene, includes: Based on the importance weight, high-importance monitoring parameters with importance higher than a preset threshold are selected, and the corresponding three-dimensional component regions in the virtual inspection focus scene are determined. In the virtual inspection focusing scene, a set of candidate observation positions and attitudes of the virtual camera are determined; Analyze the visibility characteristics of the virtual camera at each candidate observation position to the three-dimensional component area where the high-importance monitoring parameter is located; Based on the visibility characteristics, with the goal of maximizing the continuous coverage of the area where the high-importance monitoring parameters are located by the virtual camera during its movement, an optimal inspection path is planned from the candidate observation location sequence.

[0013] Furthermore, the analysis of the visibility characteristics of the virtual camera at each candidate observation position to the three-dimensional component region where the high-importance monitoring parameter is located includes: For each of the candidate observation positions, calculate the spatial overlap metric between the virtual camera's preset field of view cone and a target three-dimensional component region, where the target three-dimensional component region is the spatial range of the three-dimensional component model associated with the high importance monitoring parameter; For the calculated candidate observation positions, the proportion of the display area of ​​the target 3D component region in the virtual camera image and the degree of visual center offset at that position are evaluated according to the overlapping spatial metric. Based on the assessed display area ratio and visual center offset for the candidate observation location, a visibility feature value characterizing the observation quality of the candidate observation location is generated.

[0014] Furthermore, the process of performing anomaly diagnosis during the inspection and generating a comprehensive inspection report based on the diagnosis results includes: As the virtual camera patrols along the inspection path, it acquires real-time data values ​​of the high-importance monitoring parameters. The real-time data values ​​are compared with preset process parameter safety thresholds to obtain abnormal parameters that exceed the thresholds. By combining the context of the three-dimensional component model associated with the abnormal parameters in the process flow, the data coupling relationship with other monitoring parameters is analyzed to locate the root cause component of the abnormality; Based on the diagnostic results of the aforementioned abnormal root cause components, a preliminary abnormal diagnosis conclusion is generated in conjunction with the equipment maintenance knowledge base.

[0015] Furthermore, the comprehensive inspection report includes: a panoramic view snapshot of the virtual inspection focus scene at the end of the inspection, the anomaly diagnosis conclusion, maintenance and handling suggestions based on the anomaly diagnosis conclusion, the inspection path information on which this inspection was based, key data records during the inspection process, and the identification information of the virtual inspection focus scene.

[0016] As can be seen from the above technical solutions, the present invention has the following advantages: This invention first constructs a 3D digital twin model associated with physical equipment components and measurement points, and dynamically updates the model through real-time data. It receives and parses multimodal inspection commands, automatically generating a virtual inspection focus scene centered on the target object. Based on this, according to the importance weight of real-time parameters, it intelligently plans a virtual camera inspection path aimed at maximizing continuous observation of highly important areas. Finally, it controls the virtual camera to conduct an immersive inspection along this path, simultaneously displaying data and status, and completing anomaly diagnosis and report generation during the inspection. This invention, by creating a unified digital mirror platform, deeply integrates real-time data, video streams, and 3D models, achieving a proactive and intelligent inspection mode. It significantly improves the immersive experience of remote inspection, target location efficiency, and observation targeting, effectively enhancing the timeliness of anomaly detection and the accuracy of root cause localization. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a remote point inspection method for workshop equipment based on digital twins provided by this invention; Figure 2 This is a schematic diagram of the process for constructing a three-dimensional digital twin model in this invention; Figure 3 This is a flowchart illustrating the process of matching runtime parameter data and driving the model in this invention. Figure 4 This is a flowchart illustrating the process of parsing instructions to determine the target inspection object and inspection intention in this invention. Figure 5 This is a schematic diagram of the process for generating a virtual inspection focus scene in this invention; Figure 6 This is a schematic diagram of the process for evaluating the importance weight of each monitoring parameter in this invention; Figure 7 This is a flowchart illustrating the process of dynamically planning the inspection path of a virtual camera in this invention. Figure 8 This is a schematic diagram of the process for abnormal diagnosis during the inspection in this invention. Detailed Implementation

[0018] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] Example 1 The implementation of the method provided by this invention can rely on a digital twin workshop equipment monitoring system that integrates data acquisition, model processing, and interactive applications. This system, deployed on a server or in the cloud, is responsible for constructing and driving three-dimensional digital twin models that correspond one-to-one with the physical workshop equipment. Simultaneously, it aggregates multi-source data from field sensors, programmable logic controllers, and cameras in real time via an industrial network. Inspection personnel interact with this unified digital twin environment through client software on their terminals. This environment is a virtual mirror that is synchronously mapped to and deeply integrated with the real-time operating status of the physical workshop. Please refer to [link to relevant documentation]. Figure 1 The method provided in this application includes the following steps: S1. Construct a three-dimensional digital twin model of the workshop equipment and establish a connection between each component in the three-dimensional digital twin model and the corresponding physical measurement point in the data acquisition system; A 3D digital twin model is an interactive 3D geometric and semantic model of physical workshop equipment in virtual space, serving as a digital mirror of the physical equipment. The data acquisition system, a hardware and software suite consisting of an industrial gateway, sensor network, programmable logic controller, and data interfaces, is responsible for real-time acquisition of operational status data from the physical equipment. A physical measuring point refers to a sensor installed on the workshop equipment to monitor specific parameters, or its corresponding unique data acquisition channel. Please refer to [link to relevant documentation]. Figure 2 The model construction includes the following sub-steps: S11. Identify the functional component outlines and measurement point symbols in the acquired technical drawings; S12. Label the component type for each identified functional component outline, and label the corresponding measurement point identifier for each identified measurement point symbol; S13. Based on the labeled component type, call the corresponding 3D component model from the preset 3D standard component library; S14. Based on the spatial relationship between the functional component outlines and measuring point symbols in the technical drawings, combine the called three-dimensional component models to generate a three-dimensional digital twin model of the workshop equipment. S15. Based on the marked measurement point identifiers, establish the data association relationship between each component and measurement point in the three-dimensional digital twin model and the corresponding physical measurement points in the data acquisition system.

[0020] Specifically, image recognition technology is used to automatically identify the geometric contours of different functional components and specific graphic symbols representing sensor installation locations in the drawings, i.e., measurement point identifiers. Next, each identified functional component contour is assigned a component type label, such as a motor or valve, and a unique measurement point identifier string is generated for each identified measurement point identifier. Then, the component type labeled in the layout is used to call the corresponding 3D component model from a pre-built and stored 3D standard component library containing various standard 3D model files. Based on the coordinates, dimensions, and relative positional relationships between the identified functional component contours and measurement point identifiers in the design drawings, all called 3D component models are positioned, aligned, and assembled in virtual 3D space, thereby generating a complete 3D digital twin model consistent with the physical equipment. Finally, based on the previously generated measurement point identifiers, the system internally matches and binds the corresponding components or location points in the 3D digital twin model with specific physical measurement points with the same identifiers in the data acquisition system, thereby establishing a precise data association between the two.

[0021] S2. Real-time acquisition of workshop equipment operating parameters, matching of operating parameters with associated components and physical measurement points in the 3D digital twin model, and driving the 3D digital twin model to update its status; Operating parameters are collected through an industrial data acquisition gateway deployed on-site in the workshop. This gateway connects to the equipment's programmable logic controller (PLC), various sensors, and monitoring and data acquisition systems, continuously acquiring process parameters such as current, voltage, temperature, vibration, and pressure, as well as equipment on / off status signals. Driving the 3D digital twin model for status updates enables real-time dynamic mapping and visualization of the virtual model's actual operating status of the physical equipment, allowing remote inspection personnel to intuitively perceive the equipment's condition. Specifically, model updates involve changing the visualization attributes of the 3D component model based on the incoming real-time data. For example, changing the color of a model area based on temperature values, or triggering corresponding operational animations based on the equipment's start / stop status, and displaying the values ​​of key parameters in real-time at the corresponding measurement points on the model. Please refer to [link / reference]. Figure 3 The data matching and driving process includes the following sub-steps: S21. Extract the real-time data values ​​and status indicators corresponding to each physical measurement point in the data acquisition system from the collected operating parameters; S22. Based on the relationships established in the three-dimensional digital twin model, map the real-time data values ​​and status indicators to the corresponding three-dimensional component models according to the physical measurement point identifiers.

[0022] Specifically, from the acquired raw operating parameter data stream, the specific measured values ​​corresponding to each physical measuring point within the data acquisition system are parsed out, i.e., real-time data values. Simultaneously, labels representing the data quality or alarm status of that measuring point are parsed out, i.e., status identifiers, such as normal, out of limit, or disconnected. Then, based on the pre-established relationships during the construction of the 3D digital twin model, the parsed real-time data values ​​and status identifiers are accurately mapped and bound to the corresponding specific 3D component models in the 3D digital twin model through the key index of the physical measuring point identifiers. For example, the temperature data value of 85℃ and the status identifier "normal" for the physical measuring point T-203 are mapped to a specific component representing the No. 2 reactor in the 3D model, thereby driving the component model to update its appearance color according to preset rules (temperature range color mapping) and displaying the 85℃ numerical label next to it.

[0023] S3. Receive multimodal inspection instructions input by inspection personnel, parse the multimodal instructions to determine the target inspection object and inspection intention; automatically generate a virtual inspection focus scene based on the three-dimensional digital twin model and the target inspection object; Multimodal inspection commands refer to instructions input by inspection personnel through the system interface, using one or more methods such as text input, voice input, or directly selecting areas on a two-dimensional plant site map. These commands express the specific equipment, areas, or points of interest that the personnel wish to view or inspect. After receiving these commands, the system needs to perform in-depth analysis to understand the user's intent. Please refer to [link to relevant documentation]. Figure 4 This can be achieved through the following steps: S311. Parse the multimodal inspection instructions and extract one or more instruction features that contain at least text keywords, image region identifiers, and entity names in the voice instructions; S312. Based on the preset equipment knowledge graph, the extracted instruction features are matched with the corresponding workshop equipment, components or measuring points in the three-dimensional digital twin model to determine at least one target inspection object; S313. Based on the parsed instruction modal features and context semantics, determine the inspection intent type expressed by the multimodal inspection instruction. The inspection intent type is at least one of status query, anomaly location, historical backtracking, and comprehensive inspection.

[0024] Specifically, the system uses natural language processing (NLP) technology to process text or speech-to-text content, extracting text keywords. For image selection commands, it identifies the included pixel regions and converts them into corresponding image region identifiers. These extracted text keywords, image region identifiers, or entity names in speech are collectively referred to as command features. The system calls a preset equipment knowledge graph for matching. This knowledge graph stores the names, aliases, functional descriptions, spatial locations, and technological connections of all equipment and components in the workshop in a graph structure. The system performs similarity calculations and semantic matching between the extracted command features and nodes in the knowledge graph, thereby locating the corresponding specific workshop equipment, components, or measurement points in the 3D digital twin model and identifying these located entities as target inspection objects. At the same time, the system combines the input method of the command (modal features) and the possible contextual semantics in the command to analyze and determine the core purpose of the user's inspection, i.e., the type of inspection intent. For example, it may be a quick check of the current status of a certain device, an anomaly location to pinpoint the root cause of an alarm, a historical review of the operation of a certain device over a period of time, or a comprehensive inspection of a certain area without omissions.

[0025] After clarifying the target inspection object and the inspection intent, the system automatically begins to construct a virtual environment conducive to immersive inspection. Please refer to [link / reference needed]. Figure 5 Specifically, the virtual inspection focus scene is generated through the following steps: S321. The scene center is the three-dimensional component model corresponding to the target inspection object in the three-dimensional digital twin model; S322. Determine the spatial range and associated object set of the virtual inspection focus scene according to the inspection intention type; wherein, the associated object set includes equipment adjacent to the target inspection object in the process flow, pipelines or lines directly connected to the target inspection object, and one or more measuring points characterizing the key operating status of the target inspection object; S323. Cut and reassemble the three-dimensional models and data measurement points involved in the scene center, spatial range and related object set from the three-dimensional digital twin model to generate an independent virtual inspection focus scene.

[0026] Specifically, the system uses the spatial location of the specific 3D component model corresponding to the target inspection object in the 3D digital twin model as the core, setting it as the visual and logical center of the scene to be generated. Based on the previously determined inspection intent type, the system dynamically determines the size of the physical spatial range to be covered by the focused scene and intelligently selects the set of related objects to be displayed together. For example, if the intent is anomaly localization, the set of related objects will mainly include upstream or downstream equipment directly adjacent to the target inspection object in the process flow, auxiliary facilities directly connected to it via pipes or cables, and several key status measurement points used to characterize its operational health. Finally, based on the determined spatial range, the system cuts out the scene center object within that spatial range, the 3D models of all related objects, and their key data measurement points from the complete 3D digital twin model. These cut-out 3D models and data measurement points are then reassembled and rendered in an independent 3D view environment, generating a virtual inspection focused scene that eliminates irrelevant interference and focuses on the core equipment and its related context of the current inspection task, providing a dedicated operating space for subsequent path planning and inspection.

[0027] S4. Based on the current real-time operating parameters of the workshop equipment, assess the importance weight of each monitoring parameter; with the goal of maximizing the continuous observation effect of the area where the high importance parameters are located, and combined with the spatial structure of the virtual inspection focusing scene, dynamically plan the inspection path of the virtual camera; Current real-time operating parameters are continuously provided by the data acquisition system and serve as a direct basis for assessing equipment status. Importance weight is a quantifiable value used to characterize the priority of different monitoring parameters in terms of the time required for attention and inspection. The spatial structure of the virtual inspection focus scene refers to the geometric positions, distances, and orientational relationships between various equipment, component models, and measuring points within this 3D scene. The inspection path refers to the sequence of continuous positions and orientations followed by the virtual camera during its movement and observation in the 3D space of the focus scene. This step first dynamically assesses the importance of the parameters; please refer to [link / reference]. Figure 6 Specifically, it includes the following sub-steps: S411. Obtain the current real-time data values, preset safety thresholds, and historical trend data of the monitoring parameters; S412. Calculate the first evaluation score based on how close the current real-time data value is to the preset safety threshold; S413. Based on historical trend data, determine whether the monitoring parameters are in an abnormal fluctuation state, and calculate the second evaluation score; S414. The importance weight of each monitoring parameter is calculated by combining the first evaluation score and the second evaluation score; among them, the monitoring parameter that is closer to the safety threshold and has more abnormal fluctuations has a higher importance weight.

[0028] Specifically, the system first acquires two types of core data for each monitoring parameter: the latest measurement value from the data acquisition system, i.e., the current real-time data value; the alarm or warning limit set in advance for the parameter according to the equipment's safe operation procedures, i.e., the preset safety threshold; and the historical reading sequence of the parameter over a period of time, i.e., historical trend data. The principle for calculating the first evaluation score is to measure the degree to which the current real-time data value deviates from the preset safety threshold, using normalized distance calculation; the closer to or exceeding the threshold, the higher the score. The second evaluation score is calculated by analyzing the historical trend data, using a mutation point detection algorithm to determine whether the parameter has recently experienced abnormal fluctuations; the more severe or unstable the fluctuations, the higher the score. Finally, the system calculates the final importance weight of the monitoring parameter by weighted fusion of the first and second evaluation scores. The fusion rule is set so that a parameter that simultaneously possesses both the characteristics of being closer to the safety threshold and exhibiting more abnormal fluctuations receives a higher importance weight.

[0029] After obtaining the importance weights of each monitoring parameter, the system uses these weights to plan an efficient observation route for the virtual camera within the generated virtual inspection focusing scene. Please refer to [link / reference]. Figure 7 Specifically, this is achieved through the following steps: S421. Select high-importance monitoring parameters with importance higher than the preset threshold according to the importance weight, and determine the corresponding three-dimensional component area in the virtual inspection focus scene; S422. Determine a set of candidate observation positions and orientations for the virtual camera in the virtual inspection focusing scene; S423. Analyze the visibility characteristics of the virtual camera at each candidate observation position to the three-dimensional component area where the high-importance monitoring parameters are located; 1. For each candidate observation position, calculate the spatial overlap metric between the virtual camera's preset field of view cone and a target 3D component region, where the target 3D component region is the spatial range of the 3D component model associated with the high-importance monitoring parameter; 2. For the calculated candidate observation positions, evaluate the proportion of the display area of ​​the target 3D component region in the virtual camera image and the degree of visual center offset at that position based on the overlap spatial metric; 3. Based on the assessed display area ratio and visual center offset for the candidate observation location, generate a visibility feature value that characterizes the observation quality of the candidate observation location.

[0030] S424. Based on visibility features, with the goal of maximizing the continuous coverage of the area where high-importance monitoring parameters are located by the virtual camera during its movement, an optimal inspection path is planned from the candidate observation location sequence.

[0031] Specifically, the system sets an importance threshold as a preset threshold, filtering out monitoring parameters with importance weights higher than this value and defining them as high-importance monitoring parameters. Each high-importance monitoring parameter is associated with a specific component in the 3D digital twin model. The 3D spatial volume occupied by this component in the virtual inspection focusing scene is determined as its corresponding 3D component region, which is also the target to be focused on for subsequent observation. Based on the spatial structure of the scene, a set of discrete spatial points and a recommended camera orientation for each point are automatically generated within the scene using a spatial sampling algorithm, forming a set of candidate observation positions and poses for the virtual camera. Then, the system performs a detailed analysis of the observation view quality of the virtual camera at each candidate observation position for each 3D component region where a high-importance monitoring parameter is located, i.e., it calculates its visibility characteristics. The system's core optimization objective is to maximize the overall effect of the virtual camera continuously and stably covering multiple high-importance 3D component regions as it moves along a path. Based on this objective, and by integrating the visibility features of all candidate observation locations, the system uses path search algorithms, such as graph-based search or optimization algorithms, to calculate and plan a moving route that optimizes continuous coverage from the candidate observation location sequence, which serves as the final virtual camera inspection path.

[0032] The specific implementation of visibility feature analysis involves three levels of calculation: First, for each candidate observation position and each target 3D component region being evaluated, the system constructs a field-of-view cone with a certain opening angle emanating from the camera position in 3D space, and calculates the volume or area of ​​the intersection between this cone and the target component region's 3D bounding box or precise model. This metric is the overlap space metric, reflecting the theoretically visible area size. Second, based on the calculated overlap space metric, the system further estimates the percentage of the screen area occupied by the visible portion of the target component region in the virtual camera's 2D rendered image, i.e., the display area ratio; simultaneously, it calculates the pixel distance or angular difference between the target region's position in the image and the image center, i.e., the visual center offset degree. The smaller the offset, the closer the target is to the center of the field of view. Finally, the system combines the display area ratio and visual center offset degree calculated for the candidate observation position into a scalar value through a comprehensive evaluation function, i.e., generating a visibility feature value representing the observation quality of the region from that position. The higher the value, the better the viewing angle of the region from that position.

[0033] S5. Control the virtual camera to perform a patrol along the inspection path in the virtual inspection focus scene, and simultaneously display the real-time data and status of the corresponding components; Based on the optimal inspection path planned in step S4, the system automatically controls the virtual camera to move within the virtual inspection focus scene in a programmed manner. The camera's position and orientation smoothly transition along the path points, simulating a continuous flight or walking observation perspective of the scene. At each moment of the inspection, the system synchronously acquires the latest operating parameters of the physical measurement points associated with the visible components within the current field of view from the data acquisition system, and dynamically binds these real-time data values ​​and status indicators to the corresponding 3D component models based on the data association established in step S2. The data and status are displayed in the following ways: the values ​​of key parameters are refreshed in real time as floating text labels next to the 3D model or at preset annotation points; simultaneously, based on whether the parameter values ​​exceed limits or whether the status indicator is an alarm, the 3D model is driven to change its local or overall color or play preset status animations. This process ensures that inspection personnel can observe the latest operating data and intuitive status feedback of the target equipment and its associated components without delay or interruption while conducting an immersive inspection following the virtual camera's perspective. In addition, the system usually provides interactive controls that allow inspectors to pause, adjust the speed of the inspection, or manually fine-tune the viewing angle for a more detailed inspection during the automatic inspection process.

[0034] S6. Conduct anomaly diagnosis during the inspection and generate a comprehensive inspection report based on the diagnosis results.

[0035] Anomaly diagnosis refers to the process by which the system intelligently analyzes the operating status of equipment during remote point inspections, identifies abnormal situations, and traces their root causes. Please refer to [link to relevant documentation]. Figure 8 Specifically, this is achieved through the following sub-steps: S61. Acquire real-time data values ​​of high-importance monitoring parameters while the virtual camera patrols along the inspection path; S62. Compare the real-time data values ​​with the preset process parameter safety thresholds to obtain abnormal parameters that exceed the thresholds; S63. Combine the contextual relationship of the three-dimensional component model associated with abnormal parameters in the process flow, analyze the data coupling relationship with other monitoring parameters, and locate the component that is the root cause of the abnormality; S64. Based on the diagnostic results of the abnormal root cause component, and combined with the equipment maintenance knowledge base, generate a preliminary abnormal diagnosis conclusion.

[0036] The comprehensive inspection report includes: a panoramic view snapshot of the virtual inspection focus scene at the end of the inspection, anomaly diagnosis conclusions, maintenance and handling suggestions based on the anomaly diagnosis conclusions, inspection path information used in this inspection, key data records during the inspection process, and identification information of the virtual inspection focus scene.

[0037] Specifically, while the virtual camera automatically patrols along the inspection path, the system simultaneously performs anomaly diagnosis and analysis. First, it continuously acquires the latest measurement values ​​of monitoring parameters identified as high-importance in previous steps from the data acquisition system—the real-time data values ​​of these high-importance monitoring parameters. The system compares the real-time data value of each high-importance monitoring parameter with its respective preset process parameter safety threshold. This safety threshold is a pre-set range based on the equipment manufacturer's specifications, historical operating data, and safe operating procedures. Once the real-time data value of a parameter exceeds or falls below its safety threshold, the parameter is marked as an abnormal parameter exceeding the threshold. Then, for each abnormal parameter, the system analyzes the data interaction relationships between the abnormal parameter and other relevant monitoring parameters, i.e., the data coupling relationships, in conjunction with the upstream and downstream relationships (contextual relationships) of its associated 3D component model within the entire production process. For example, by analyzing the changing trends and correlations between parameters, the system infers the initial component most likely to cause the anomaly—the root cause component of the anomaly. Finally, the system accesses the equipment maintenance knowledge base, which stores common equipment failure modes, mapping relationships between root cause components and symptoms, and expert experience rules. Based on the located abnormal root cause components and associated symptom data, the system matches the most likely cause of the failure and generates a preliminary abnormal diagnosis conclusion that includes the abnormal phenomenon, possible cause, and confidence level.

[0038] This invention constructs a novel remote point inspection mode for workshop equipment by building a three-dimensional digital twin model associated with physical equipment and deeply integrating real-time multi-source data. It achieves intelligent guidance and immersive interaction during the inspection process, improves the accuracy of status perception and anomaly diagnosis, enhances the standardization, efficiency, and reliability of inspection operations, and provides strong data support for predictive maintenance of equipment.

[0039] It is understood that those skilled in the art can combine various implementation methods in the above embodiments under the guidance of the above examples to obtain technical solutions with multiple implementation methods.

[0040] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for remote point inspection of workshop equipment based on digital twins, characterized in that, include: Construct a three-dimensional digital twin model of the workshop equipment, and establish an association between each component in the three-dimensional digital twin model and the corresponding physical measurement point in the data acquisition system; The operating parameters of the workshop equipment are collected in real time, and the operating parameters are matched with the components and physical measurement points associated in the three-dimensional digital twin model, and the three-dimensional digital twin model is driven to update its status. Receive multimodal inspection instructions input by inspection personnel, and parse the multimodal instructions to determine the target inspection object and inspection intention; Based on the aforementioned three-dimensional digital twin model and the target inspection object, a virtual inspection focus scene is automatically generated; Based on the current real-time operating parameters of the workshop equipment, the importance weight of each monitoring parameter is evaluated; with the goal of maximizing the continuous observation effect of the area where the high importance parameters are located, the inspection path of the virtual camera is dynamically planned in combination with the spatial structure of the virtual inspection focusing scene; The virtual camera is controlled to patrol along the inspection path in the virtual inspection focus scene, and the real-time data and status of the corresponding components are displayed simultaneously. During the inspection, anomalies are diagnosed, and a comprehensive inspection report is generated based on the diagnosis results.

2. The method for remote point inspection of workshop equipment based on digital twins according to claim 1, characterized in that, The process involves constructing a three-dimensional digital twin model of the workshop equipment and associating each component in the three-dimensional digital twin model with the corresponding physical measurement points in the data acquisition system, including: Identify the outlines of functional components and measurement point symbols in the acquired technical drawings; For each identified functional component outline, label the corresponding component type; and for each identified measuring point symbol, label the corresponding measuring point identifier. Based on the component type indicated in the annotation, the corresponding 3D component model is retrieved from the preset 3D standard component library; Based on the spatial relationship between the functional component outlines and measuring point symbols in the technical drawings, the called three-dimensional component models are combined to generate a three-dimensional digital twin model of the workshop equipment. Based on the marked measurement point identifiers, establish the data association relationship between each component and measurement point in the three-dimensional digital twin model and the corresponding physical measurement points in the data acquisition system.

3. The method for remote point inspection of workshop equipment based on digital twins according to claim 1, characterized in that, The process of matching the operating parameters with the associated components and physical measurement points in the three-dimensional digital twin model includes: The real-time data values ​​and status identifiers corresponding to each physical measurement point in the data acquisition system are parsed from the collected operating parameters. Based on the established relationships in the three-dimensional digital twin model, the real-time data values ​​and status identifiers are mapped to the corresponding three-dimensional component models according to the physical measurement point identifiers.

4. The method for remote point inspection of workshop equipment based on digital twins according to claim 1, characterized in that, The process of parsing the multimodal commands to determine the target inspection object and inspection intent includes: The multimodal inspection instructions are parsed to extract one or more instruction features that include at least text keywords, image region identifiers, and entity names in voice instructions; Based on a preset equipment knowledge graph, the extracted instruction features are matched with the corresponding workshop equipment, components or measuring points in the three-dimensional digital twin model to determine at least one target inspection object; Based on the parsed instruction modal features and contextual semantics, the inspection intent type expressed by the multimodal inspection instruction is determined, and the inspection intent type is at least one of status query, anomaly location, historical backtracking, and comprehensive inspection.

5. The remote inspection method for workshop equipment based on digital twins according to claim 4, characterized in that, The automatic generation of a virtual inspection focus scene based on the three-dimensional digital twin model and the target inspection object includes: The scene center is the three-dimensional component model corresponding to the target inspection object in the three-dimensional digital twin model; The spatial range and associated object set of the virtual inspection focus scene are determined according to the inspection intention type; wherein, the associated object set includes equipment adjacent to the target inspection object in the process flow, pipelines or lines directly connected to the target inspection object, and one or more measuring points characterizing the key operating status of the target inspection object; The three-dimensional models and data measurement points involved in the scene center, spatial range and related object set are cut and recombined from the three-dimensional digital twin model to generate an independent virtual inspection focus scene.

6. The method for remote point inspection of workshop equipment based on digital twins according to claim 1, characterized in that, The step of evaluating the importance weight of each monitoring parameter based on the current real-time operating parameters of the workshop equipment includes: Obtain the current real-time data value, preset safety threshold, and historical trend data of the monitoring parameters; Calculate a first evaluation score based on how close the current real-time data value is to the preset security threshold; Based on the historical trend data, determine whether the monitoring parameter is in an abnormal fluctuation state, and calculate the second evaluation score; By combining the first evaluation score and the second evaluation score, the importance weight of each monitoring parameter is calculated; among them, the monitoring parameter that is closer to the safety threshold and has more abnormal fluctuations has a higher importance weight.

7. The method for remote point inspection of workshop equipment based on digital twins according to claim 6, characterized in that, The goal is to maximize continuous observation of areas containing high-importance parameters. This is achieved by dynamically planning the inspection path of the virtual camera, taking into account the spatial structure of the virtual inspection focusing scene. This includes: Based on the importance weight, high-importance monitoring parameters with importance higher than the preset threshold are selected, and the corresponding three-dimensional component regions in the virtual inspection focus scene are determined. In a virtual inspection focusing scenario, a set of candidate observation positions and attitudes of the virtual camera are determined; Analyze the visibility characteristics of the virtual camera at each candidate observation position to the three-dimensional component area where the high-importance monitoring parameter is located; Based on the visibility characteristics, with the goal of maximizing the continuous coverage of the area where the high-importance monitoring parameters are located by the virtual camera during its movement, an optimal inspection path is planned from the candidate observation location sequence.

8. The method for remote point inspection of workshop equipment based on digital twins according to claim 7, characterized in that, The analysis of the visibility characteristics of the virtual camera at each candidate observation position to the three-dimensional component region where the high-importance monitoring parameter is located includes: For each of the candidate observation positions, calculate the spatial overlap metric between the virtual camera's preset field of view cone and a target three-dimensional component region, where the target three-dimensional component region is the spatial range of the three-dimensional component model associated with the high importance monitoring parameter; For the calculated candidate observation positions, the proportion of the display area of ​​the target 3D component region in the virtual camera image and the degree of visual center offset at that position are evaluated according to the overlapping spatial metric. Based on the assessed display area ratio and visual center offset for the candidate observation location, a visibility feature value characterizing the observation quality of the candidate observation location is generated.

9. The method for remote point inspection of workshop equipment based on digital twins according to claim 1, characterized in that, The process of performing anomaly diagnosis during the inspection and generating a comprehensive inspection report based on the diagnosis results includes: As the virtual camera patrols along the inspection path, it acquires real-time data values ​​of the high-importance monitoring parameters. The real-time data values ​​are compared with preset process parameter safety thresholds to obtain abnormal parameters that exceed the thresholds. By combining the context of the three-dimensional component model associated with the abnormal parameters in the process flow, the data coupling relationship with other monitoring parameters is analyzed to locate the root cause component of the abnormality; Based on the diagnostic results of the aforementioned abnormal root cause components, a preliminary abnormal diagnosis conclusion is generated in conjunction with the equipment maintenance knowledge base.

10. The method for remote point inspection of workshop equipment based on digital twins according to claim 9, characterized in that, The comprehensive inspection report includes: a panoramic view snapshot of the virtual inspection focus scene at the end of the inspection, the anomaly diagnosis conclusion, maintenance and handling suggestions based on the anomaly diagnosis conclusion, the inspection path information on which this inspection was based, key data records during the inspection process, and the identification information of the virtual inspection focus scene.

Images