An AI technology-based automatic drawing review system and method

Abstract

The application relates to an AI technology-based automatic drawing review system and method, and relates to the technical field of automatic drawing review.The system comprises a drawing uploading module, which is used for judging whether a CAD drawing to be reviewed needs parallel transmission, and if yes, performing block processing to generate multiple drawing blocks and uploading the multiple drawing blocks; a drawing preprocessing module, which is used for merging the multiple drawing blocks to generate a CAD drawing to be reviewed, performing partition processing to generate multiple area drawings to be reviewed; a drawing review module, which is used for determining a review sequence based on the component features of each area drawing to be reviewed, performing review based on the review sequence, determining review items of the area drawing to be reviewed based on the component features of the area drawing to be reviewed, performing review on the area drawing to be reviewed, and generating a review result; and a review allocation module, which is used for determining a target manual review end, and transmitting the CAD drawing to be reviewed and the review result to the target manual review end, so that the drawing review efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of automated drawing review technology, and in particular to an automated drawing review system and method based on AI technology. Background Technology

[0002] In the field of architectural engineering, CAD drawing design plays a crucial role, impacting every stage of the construction project lifecycle. For example, in the architectural design phase, architects creatively layout and conceive of the building space, generating feasible architectural drawings using CAD software. This process considers technical specifications such as building structure, building materials, water and electricity requirements, and fire safety requirements, ensuring precise design and control of the entire construction project. In the construction drawing design phase, precise CAD drawings are needed for various architectural disciplines, such as reinforcement design schemes, concrete design schemes, mechanical and electrical design schemes, and water and electricity design schemes, all requiring complete representation using CAD drawings. During the construction phase, CAD drawings play a supporting and guiding role. Construction steps, construction sequence, and protective measures marked on the drawings are essential information based on CAD design, helping project managers coordinate communication and anticipate and resolve potential problems.

[0003] Existing technologies are inefficient in the review and analysis of engineering design drawings, lack precise analysis and interoperability of options, and do not provide automated methods for reviewing and analyzing engineering design drawings and collecting and comparing data.

[0004] Therefore, there is a need to provide an automated drawing review system and method based on AI technology to improve drawing review efficiency. Summary of the Invention

[0005] This invention provides an automated drawing review system based on AI technology, comprising: a drawing upload module, used to determine whether the CAD drawing to be reviewed needs to be transmitted in parallel; if it is determined that the CAD drawing to be reviewed needs to be transmitted in parallel, the CAD drawing to be reviewed is divided into blocks to generate multiple drawing blocks, and the multiple drawing blocks are uploaded; a drawing preprocessing module, used to merge the multiple drawing blocks transmitted in parallel to generate the CAD drawing to be reviewed, and also used to partition the CAD drawing to be reviewed to generate multiple areas of drawing to be reviewed; a drawing review module, used to determine the review order of multiple areas of drawing to be reviewed based on the component characteristics of each area of ​​drawing to be reviewed, and to review the area of ​​drawing to be reviewed based on the review order of the multiple areas of drawing to be reviewed; to determine the review items of the area of ​​drawing to be reviewed based on the component characteristics of the area of ​​drawing to be reviewed, to review the area of ​​drawing to be reviewed based on the review items of the area of ​​drawing to be reviewed, and to generate the review result of the area of ​​drawing to be reviewed; and to generate the review result of the CAD drawing to be reviewed based on the review result of each area of ​​drawing to be reviewed; and a review assignment module, used to determine the target human review end, and to transmit the CAD drawing to be reviewed and the review result to the target human review end.

[0006] Furthermore, the drawing upload module determines whether the CAD drawing to be reviewed needs to be transmitted in parallel, including: predicting the future CAD drawing transmission volume based on the CAD drawing transmission volume at multiple historical time points; determining the real-time drawing size threshold based on the predicted future CAD drawing transmission volume; and determining whether the CAD drawing to be reviewed needs to be transmitted in parallel based on the drawing size of the CAD drawing to be reviewed and the real-time drawing size threshold.

[0007] Furthermore, the drawing upload module performs block processing on the CAD drawings to be reviewed, generating multiple drawing blocks, including: determining the optimal block size based on the predicted future CAD drawing transmission volume; dividing the CAD drawings to be reviewed into multiple layer blocks according to the layers of the CAD drawings to be reviewed; for each layer block, performing block processing according to the component distribution characteristics of the layer block and the optimal block size, generating the drawing blocks corresponding to the layer block and auxiliary merging tags for each drawing block.

[0008] Furthermore, the drawing preprocessing module merges multiple drawing blocks transmitted in parallel to generate CAD drawings to be reviewed, including: merging multiple drawing blocks transmitted in parallel based on auxiliary merging tags for each drawing block to generate CAD drawings to be reviewed.

[0009] Furthermore, the drawing preprocessing module partitions the CAD drawings to be reviewed, generating multiple review area drawings, including: dividing the CAD drawings to be reviewed into layer drawings according to their layers; for each layer drawing, identifying the components in the layer drawing, determining the component distribution characteristics of the layer drawing, clustering the components in the layer drawing according to the component distribution characteristics of the layer drawing, and generating the review area drawings corresponding to the layer drawing.

[0010] Furthermore, the drawing review module determines the review order of multiple drawing areas based on the component characteristics of each drawing area to be reviewed, including: for each drawing area to be reviewed, determining a first weight of the drawing area based on the component distribution characteristics of the drawing area; determining a second weight of the drawing area based on the component weights corresponding to the components in the drawing area; determining a third weight of the drawing area based on the association weights of any two components included in the drawing area; determining a comprehensive weight of the drawing area based on the first, second, and third weights; and determining the review order of multiple drawing areas based on the comprehensive weight of each drawing area.

[0011] Furthermore, the drawing review module determines the review items of the drawings in the area to be reviewed based on the component characteristics of the drawings in the area to be reviewed, including: determining the review items of multiple template area drawings; determining similar template area drawings based on the component characteristics of the drawings in the area to be reviewed and the component characteristics of each template area drawing; and determining the review items of the drawings in the area to be reviewed based on the review items of the similar template area drawings.

[0012] Furthermore, the drawing review module reviews the drawings in the area to be reviewed according to the review items, and generates review results for the drawings in the area to be reviewed. This includes: identifying components in the drawings in the area to be reviewed, reviewing the components, and generating component review results for the drawings in the area to be reviewed; identifying associated component pairs in the drawings in the area to be reviewed, reviewing the component pairs, and generating component pair review results for the drawings in the area to be reviewed; and reviewing the drawings in the area to be reviewed according to the review items, generating area review results for the drawings in the area to be reviewed. The review results for the drawings in the area to be reviewed include component review results, component pair review results, and area review results for the drawings in the area to be reviewed.

[0013] Furthermore, the review allocation module determines the target manual review end by: determining the target manual review end based on the component characteristics of the CAD drawing to be reviewed and the review results.

[0014] This invention provides an automated drawing review method based on AI technology, applied to the aforementioned automated drawing review system based on AI technology. The method includes: determining whether the CAD drawing to be reviewed needs parallel transmission; if so, dividing the CAD drawing into blocks to generate multiple drawing blocks and uploading these blocks; merging the multiple drawing blocks transmitted in parallel to generate the CAD drawing to be reviewed, and further partitioning the CAD drawing to be reviewed to generate multiple areas of drawing to be reviewed; determining the review order of the multiple areas of drawing to be reviewed based on the component characteristics of each area of ​​drawing to be reviewed; reviewing the areas of drawing to be reviewed based on the review order, determining the review items based on the component characteristics of each area of ​​drawing to be reviewed, reviewing the areas of drawing to be reviewed according to the review items, and generating the review results of the areas of drawing to be reviewed; generating the review results of the CAD drawing to be reviewed based on the review results of each area of ​​drawing to be reviewed; determining the target human review end based on the component characteristics and review results of the CAD drawing to be reviewed, and transmitting the CAD drawing to be reviewed and the review results to the target human review end.

[0015] Compared to existing technologies, the AI-based automated drawing review system and method provided in this specification have at least the following advantages:

[0016] 1. The drawing upload module can determine whether the CAD drawings to be reviewed need to be transmitted in parallel. If so, it processes them in chunks and uploads them simultaneously. This fully utilizes multi-threading or multi-processing technology, breaking the time limitations of traditional single-threaded transmission and processing. For example, for large architectural CAD drawings, traditional serial transmission may take a long time, while parallel transmission can divide the drawings into multiple small chunks and upload them simultaneously, greatly shortening the upload time.

[0017] The drawing review module determines the review order and items for the drawings in the areas to be reviewed based on component characteristics, and performs the review automatically, eliminating the need for manual determination of review priorities and order. Taking mechanical design drawing review as an example, the system can quickly identify key components in the drawings (such as high-precision transmission parts), prioritize the review of these areas, and automatically match the corresponding review rules (such as dimensional tolerances, material strength, etc.), avoiding omissions or disordered order that may occur during manual review, and significantly speeding up the review process.

[0018] 2. The drawing review module determines the review items based on the component characteristics of the drawings in the area to be reviewed, enabling targeted reviews based on the characteristics and requirements of different components. In electrical drawing review, for different types of electrical components (such as switches and transformers), the system can set different review items according to their characteristics, such as the rated current of switches and the capacity matching of transformers, ensuring a comprehensive and accurate review and effectively avoiding quality problems caused by incomplete review items;

[0019] 3. By analyzing CAD drawing transmission volumes at multiple historical points in time, future transmission volumes can be predicted, allowing for advance understanding of network load trends. For example, in different project cycles within a design institute, drawing transmission volumes fluctuate with project progress; historical data prediction can accurately capture this pattern. Based on the predicted future transmission volumes, a real-time drawing size threshold is determined and compared with the size of CAD drawings awaiting review to determine whether parallel transmission is necessary. During peak transmission periods, the threshold is appropriately lowered to allow more drawings to be transmitted in parallel, fully utilizing network bandwidth; during off-peak periods, the threshold is raised to reduce unnecessary parallel transmission overhead and improve overall transmission efficiency. If all drawings, regardless of size, are transmitted in parallel, it will consume significant network and server processing resources, increasing operating costs. By rationally assessing parallel transmission needs and only performing parallel transmission when necessary, excessive resource consumption and waste are avoided. Dynamically adjusting the parallel transmission strategy based on transmission volume allows network and server resources to be allocated according to actual demand, improving resource utilization and reducing enterprise operating costs.

[0020] 4. Determining the optimal block size based on predicted future CAD drawing transmission volume can make the transmission of block-based drawings more efficient. For example, when transmission bandwidth is limited, appropriately reducing the block size reduces the transmission time of each block and lowers the risk of transmission failure; when transmission bandwidth is sufficient, appropriately increasing the block size reduces the number of blocks and lowers the complexity of merging processing. First, divide the drawing into multiple layer blocks according to layers, and then further subdivide them based on the component distribution characteristics of the layer blocks and the optimal block size. This block-division method is more scientific and reasonable. Different layers usually contain different types of components; dividing by layer can maintain the relative integrity of components, facilitating subsequent review and processing. At the same time, further subdivision based on component distribution characteristics can avoid dividing important components into different blocks, ensuring the integrity and consistency of drawing information. Generating auxiliary merging tags for each drawing block allows for accurate merging of each block into a complete drawing during the subsequent merging process, avoiding merging errors and improving the accuracy and efficiency of merging.

[0021] 5. Within each layer of drawings, components are clustered based on their distribution characteristics to generate drawings for the areas to be reviewed. This clustering method considers the spatial distribution and interrelationships of components, grouping related components into the same area for centralized review. For example, in mechanical drawings, the areas containing various parts of the same component are clustered together. During review, all related components of that component can be checked at once, avoiding frequent switching between different areas and improving review efficiency. Multiple drawings for review after partitioning can be reviewed in parallel, making full use of system resources and shortening the overall review time. Different reviewers or review threads can process different drawings for review simultaneously, improving the system's concurrent processing capabilities. By determining the first weight based on component distribution characteristics, the second weight based on component weights, and the third weight based on component association weights, and combining these three weights to obtain the comprehensive weight of the drawings for the areas to be reviewed, multiple factors affecting the review order are comprehensively considered. Component distribution characteristics reflect the spatial density and importance of components, component weights reflect the criticality of a single component in the drawing, and component association weights consider the mutual influence between components. This multi-dimensional weight determination method can more accurately assess the importance and review priority of each drawing for review. Using a template-based approach to define audit items ensures consistency in the audit items for similar drawing areas across different auditors or at different times. This avoids significant differences in audit items due to variations in auditors' individual experience and understanding, thus guaranteeing the standardization and normalization of the audit process.

[0022] 6. The drawings for the area under review are reviewed at three levels: component review, component-to-component review, and area review. Component review focuses on the compliance and accuracy of individual components, ensuring that each component conforms to relevant standards and specifications; component-to-component review considers the relationships and mutual influences between components, checking whether the coordination between components is reasonable; area review evaluates the drawings for the area under review as a whole, determining whether the area meets design requirements and functional requirements. This multi-level review approach can comprehensively and meticulously identify problems in the drawings, improving the accuracy and reliability of the review. Attached Figure Description

[0023] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:

[0024] Figure 1 This is a block diagram of an AI-based automated drawing review system shown in one embodiment of this application;

[0025] Figure 2This is a flowchart illustrating an automated drawing review method based on AI technology in one embodiment of this application. Detailed Implementation

[0026] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0027] Figure 1 This is a block diagram of an automated drawing review system based on AI technology, as shown in one embodiment of this application. Figure 1 As shown, an AI-based automated drawing review system may include a drawing upload module, a drawing preprocessing module, a drawing review module, and a review assignment module.

[0028] The drawing upload module is used to determine whether the CAD drawings to be reviewed need to be transmitted in parallel. If it is determined that the CAD drawings to be reviewed need to be transmitted in parallel, the CAD drawings to be reviewed are divided into blocks, multiple drawing blocks are generated, and multiple drawing blocks are uploaded.

[0029] In some embodiments, the drawing upload module determines whether the CAD drawing to be reviewed needs to be transmitted in parallel, including:

[0030] Predict future CAD drawing transfer volume based on CAD drawing transfer volume at multiple historical time points;

[0031] Determine the real-time drawing size threshold based on the predicted future CAD drawing transfer volume;

[0032] Based on the size of the CAD drawing to be reviewed and the real-time drawing size threshold, determine whether the CAD drawing to be reviewed needs to be transmitted in parallel.

[0033] Specifically, the drawing upload module can collect and analyze CAD drawing transmission volume data from multiple historical time points over a period of time (such as a week, a month, or a longer period). It uses time series analysis, machine learning prediction algorithms (such as ARIMA (AutoRegressive Integrated Moving Average), LSTM (Long Short-Term Memory), etc.) or simple statistical methods (such as moving average, exponential smoothing) to predict the CAD drawing transmission volume in the future.

[0034] As an example, historical CAD drawing transfer data can be extracted from server logs or monitoring systems, including transfer time, drawing size, and transfer duration. The data is cleaned, outliers are removed, and missing values ​​are filled or interpolated. A transfer volume prediction model is built and trained using historical data. The trained model is then used to predict the CAD drawing transfer volume for a future period (such as the next time period, a day, or a week).

[0035] Assuming the CAD drawing transfer volume at 10 AM each day over the past week was 100MB, 120MB, 110MB, 130MB, 125MB, 135MB, and 140MB respectively, time series analysis can be used to predict the CAD drawing transfer volume at 10 AM each day for the next week.

[0036] Based on the predicted future volume of CAD drawings to be transmitted, and considering factors such as current network bandwidth and server processing capacity, a real-time drawing size threshold is dynamically determined. When the size of the CAD drawings to be reviewed exceeds this threshold, parallel transmission is employed to improve transmission efficiency.

[0037] For example, suppose the predicted CAD drawing transfer volume at 10 AM the next day is 200MB, the current network bandwidth is 100Mbps (i.e., 12.5MB / s), and the server's processing capacity is 5 medium-sized drawings per second (assuming medium size is 20MB). To maintain smooth network transmission and balanced server processing, a real-time drawing size threshold can be set. For instance, if the predicted transfer volume exceeds 150MB, the threshold can be set to 50MB, meaning drawings exceeding 50MB will be considered for parallel transmission.

[0038] In some embodiments, the drawing upload module divides the CAD drawings to be reviewed into blocks, generating multiple drawing blocks, including:

[0039] Determine the optimal block size based on the predicted future CAD drawing transfer volume;

[0040] Based on the layers of the CAD drawings to be reviewed, the drawings are divided into multiple layer blocks. The drawing data for one layer constitutes one layer block. CAD drawings typically consist of multiple layers, each containing different graphic elements and information. Layer-based block processing preserves the logical structure of the drawings, facilitating subsequent block transfer, processing, and merging. Furthermore, layer-based block processing aligns with the design and usage habits of CAD drawings, better preserving their semantic information. For example, given an architectural CAD drawing containing three layers: "Walls," "Doors and Windows," and "Furniture," by parsing the drawing, separate "Wall Layer Blocks," "Doors and Windows Layer Blocks," and "Furniture Layer Blocks" are created, and the graphic elements from each layer are copied to their corresponding layer blocks.

[0041] For each layer block, based on the component distribution characteristics and optimal block size, the layer block is divided into blocks to generate the corresponding drawing blocks and auxiliary merging labels for each drawing block.

[0042] Specifically, the optimal chunk size is determined to ensure that the partitioned drawings can be transmitted efficiently without placing excessive processing pressure on the server. If the predicted future transmission volume is large and the network may be congested, a smaller chunk size helps improve the parallelism and fault tolerance of the transmission. If the predicted transmission volume is small and the network is relatively idle, the chunk size can be appropriately increased, the number of chunks can be reduced, and the server processing overhead can be lowered.

[0043] Metrics for evaluating the merits of chunk size can be identified, such as transmission time, server resource utilization, and data integrity assurance capabilities. Based on the predicted transmission volume, the performance of each evaluation metric during drawing transmission and processing is simulated under different chunk sizes (e.g., 1MB, 5MB, 10MB). This can be achieved by constructing a simulation model or conducting comparative analysis using historical data. By comprehensively considering all evaluation metrics, the chunk size that best performs overall while meeting the requirements of transmission efficiency, server processing capacity, and data integrity is selected as the optimal chunk size.

[0044] For example, suppose we predict that the amount of CAD drawings to be transferred at 10:00 AM the next day will reach 500MB, and the network bandwidth is 100Mbps (approximately 12.5MB / s). Simulations show that when the block size is 5MB, the transmission time is approximately 40 seconds, the server CPU utilization is within a reasonable range (e.g., not exceeding 70%), and data integrity is well guaranteed. However, when the block size is 10MB, although the transmission time is shortened to about 20 seconds, the server CPU utilization will rise to over 80%, potentially leading to a decrease in server performance and affecting other tasks. After comprehensive consideration, the optimal block size is determined to be 5MB.

[0045] The component distribution characteristics of a layer block can include the number of components per unit area corresponding to the optimal block size in the layer block. The variance of the number of components per unit area corresponding to the optimal block size in each layer block included in the layer block is calculated as the difference value of the number of components per unit area in the layer block. If the difference value is less than the difference value threshold, the component distribution in the layer block is relatively uniform. Then, the layer block is evenly divided into multiple drawing blocks according to the unit area corresponding to the optimal block size.

[0046] If the difference value is greater than or equal to the difference value threshold, then the unit area corresponding to the adjacent optimal block size that has a number of components greater than the component number threshold is merged, and the layer block is evenly divided into several irregular drawing blocks.

[0047] Preferably, auxiliary labels may include:

[0048] Location Labels: Record the location information of each drawing block in the original layer block, such as the coordinates of the top left corner and bottom right corner of the block.

[0049] Sequence Label: Assign a sequence number to each drawing block to determine the splicing order of the blocks during merging.

[0050] The drawing preprocessing module is used to merge multiple drawing blocks transmitted in parallel to generate CAD drawings to be reviewed. It is also used to partition the CAD drawings to be reviewed to generate multiple areas of drawings to be reviewed.

[0051] In some embodiments, the drawing preprocessing module merges multiple drawing blocks transmitted in parallel to generate CAD drawings for review, including:

[0052] Based on the auxiliary merging tag of each drawing block, multiple drawing blocks transmitted in parallel are merged to generate CAD drawings to be reviewed.

[0053] Specifically, the location labels record the position information of each drawing block within the original layer block, typically including the coordinates of the block's top-left and bottom-right corners. For example, a block's location labels might be top-left (100, 100) and bottom-right (300, 300), indicating that the block occupies a specific rectangular area within the original layer block. During merging, the drawing preprocessing module uses these location labels to accurately place each block at its corresponding coordinate position on the original drawing. It's like piecing together... Figure 1 Similarly, each block has its specific location, and guided by the location labels, the module can piece together the blocks into a complete graphic. For example, for the block in the example above, the module will place it in the area with the upper left corner coordinates (100, 100) and the lower right corner coordinates (300, 300) in the original drawing.

[0054] Each drawing block is assigned a unique sequence number by a sequence label to determine the splicing order during merging. For example, blocks are designated as Block 1, Block 2, Block 3, etc., with smaller numbers indicating earlier splicing. During the merging process, the drawing preprocessing module splices the blocks sequentially according to the sequence labels, from smallest to largest. This ensures that the merged drawing is visually and logically coherent, preventing content chaos caused by disordered block order. For instance, the block with sequence label 1 is spliced ​​first, followed by the block with sequence label 2, and so on.

[0055] After the merge is complete, the drawing preprocessing module needs to verify the data integrity of the merged CAD drawings. This involves checking whether all graphic elements and text information in the drawings are complete and whether there is any data loss or corruption that occurred during transmission or the merging process.

[0056] In some embodiments, the drawing preprocessing module partitions the CAD drawings to be reviewed, generating multiple review areas, including:

[0057] The CAD drawings to be reviewed are divided into layer drawings according to their layers, with one layer corresponding to one layer drawing.

[0058] For each layer of drawings, identify the components in the layer drawing, determine the component distribution characteristics of the layer drawing, cluster the components in the layer drawing according to the component distribution characteristics, and generate the corresponding area drawing to be reviewed.

[0059] Specifically, a component is a combination of graphic elements with specific functions and meanings in a CAD drawing. Identifying components requires following certain rules and algorithms to combine graphic elements in layered drawings into meaningful components. For example, in architectural CAD drawings, a wall can be viewed as a component composed of multiple straight lines and polylines, while doors and windows can be viewed as components composed of rectangles and specific symbols.

[0060] Based on different types of CAD drawings and review requirements, define component identification rules. For example, for architectural drawings, the identification rule for walls can be defined as continuous straight lines or combinations of polylines exceeding a certain threshold in length; the identification rule for doors and windows can be rectangles of specific sizes and shapes plus door / window symbols. Using image processing, pattern recognition, and other technologies, the defined identification rules are applied to layered drawings to identify the components. By traversing the graphic elements in the layered drawings and checking whether they conform to the component identification rules, graphic elements that conform to the rules are combined into components.

[0061] The component distribution characteristics of the layer drawing can include the location of the components. The K-means clustering algorithm is used to cluster the components according to their location to determine multiple component clusters. The drawing corresponding to each component cluster is extracted from the layer drawing as the area drawing to be reviewed.

[0062] The drawing review module is used to determine the review order of multiple drawing areas based on the component characteristics of each drawing area to be reviewed, and to conduct the review based on the review order of multiple drawing areas to be reviewed. Based on the component characteristics of the drawing areas to be reviewed, the review items of the drawing areas to be reviewed are determined, the drawing areas to be reviewed are reviewed according to the review items, and the review results of the drawing areas to be reviewed are generated. Based on the review results of each drawing area to be reviewed, the review results of the CAD drawings to be reviewed are generated.

[0063] In some embodiments, the drawing review module determines the review order of multiple drawing areas based on the component characteristics of each drawing area to be reviewed, including:

[0064] For each drawing area to be reviewed, the first weight of the drawing area is determined based on the component distribution characteristics of the drawing area. The second weight of the drawing area is determined based on the component weights corresponding to the components in the drawing area. The third weight of the drawing area is determined based on the association weights of any two components included in the drawing area. The comprehensive weight of the drawing area is determined based on the first weight, second weight and third weight of the drawing area.

[0065] The review order of multiple areas of drawings is determined based on the overall weight of each area of ​​drawings to be reviewed.

[0066] Specifically, the component distribution characteristics of the drawings in the area to be reviewed can include the number of components included in the drawings in the area to be reviewed. The first weight of the drawings in the area to be reviewed is calculated according to the following formula:

[0067]

[0068] in, Let be the first weight of the i-th drawing area to be reviewed. Let be the number of components included in the i-th drawing area to be reviewed. Let m be the number of components included in the m-th drawing area to be reviewed. This represents the total number of drawings for the area awaiting review.

[0069] Understandably, the number of components in the drawings for the area to be reviewed... Divide by the sum of the number of components in all the drawings for the areas to be reviewed. This normalizes the first weight to the [0,1] interval. The first weight represents the proportion of the number of components in the i-th drawing to be reviewed among all drawings to be reviewed, reflecting its relative complexity.

[0070] Different types of components have varying degrees of importance in drawings. For example, in architectural structural drawings, load-bearing walls, columns, and other structural components have higher weights because they are directly related to the structural safety of the building; while some decorative components have relatively lower weights. The secondary weight for each component can be manually set (e.g., by an expert). The average of the secondary weights for each component in the area of ​​drawings to be reviewed is used as the secondary weight for the area of ​​drawings to be reviewed.

[0071] Some components are functionally closely related. For example, in mechanical drawings, transmission components and power components are highly interconnected; a design problem in one component may affect the normal operation of the other. Spatial adjacency or interdependence of components also affects their association weight. For instance, in architectural drawings, adjacent walls and doors / windows are spatially related, and the size and location of doors and windows need to be adapted to the walls. In some complex drawings, components may have logical relationships, such as the logical relationship between control circuits and controlled equipment in an electrical system. The association weight of any two components can be manually set (e.g., by an expert). The average of the association weights of any two components included in the drawing area to be reviewed is used as the third weight of the drawing area to be reviewed.

[0072] For example, the overall weight of the drawings for the area to be reviewed can be calculated using the following formula:

[0073]

[0074] in, Let be the overall weight of the i-th drawing area awaiting review. , and For preset coefficients, , and Greater than 0, and , Let be the second weight of the i-th drawing area awaiting review. Let be the third weight of the i-th drawing area to be reviewed. The second weight is the weight of the m-th drawing in the area to be reviewed. This is the third weight of the m-th drawing area awaiting review.

[0075] Understandably, this formula uses coefficients , and Weights of different dimensions (i.e.) , and A linear combination reflects the relative contribution of each dimension to the overall weight. The overall score of all drawings is scaled proportionally to ensure that the sum of the weights is 1, which facilitates subsequent prioritization or resource allocation.

[0076] The greater the overall weight of the drawings in the area to be reviewed, the earlier the drawings in that area will be reviewed.

[0077] In some embodiments, the drawing review module determines the review items for the drawings in the area to be reviewed based on the component characteristics of the drawings in the area to be reviewed, including:

[0078] Determine the review items for drawings in multiple template areas;

[0079] Based on the component characteristics of the drawings in the area to be reviewed and the component characteristics of each template area drawing, similar template area drawings are identified.

[0080] Based on the review items for similar template area drawings, determine the review items for the area drawings to be reviewed. For example, use the review items for similar template area drawings as the review items for the area drawings to be reviewed.

[0081] For example, the similarity between the drawing of the area to be reviewed and the drawing of the template area can be calculated using the following formula:

[0082]

[0083] in, The similarity between the drawing of the area to be reviewed and the drawing of the j-th template area. Let K be the matching value of the Kth component included in the drawing of the area to be reviewed. If the Kth component exists in both the drawing of the area to be reviewed and the jth template area drawing, then... Otherwise, it is 0. This represents the total number of components included in the drawings for the area to be reviewed.

[0084] Understandably, this formula calculates a similarity score between 0 and 1 by counting the number of components shared by both parties and normalizing the results.

[0085] In some embodiments, the drawing review module reviews the drawings in the area to be reviewed based on the review items, and generates the review results for the drawings in the area to be reviewed, including:

[0086] Identify components in the drawings of the area to be reviewed, review the components, and generate the component review results for the drawings of the area to be reviewed;

[0087] Identify the associated component pairs in the drawings of the area to be reviewed, review the component pairs, and generate the component pair review results for the drawings of the area to be reviewed.

[0088] The drawings in the area to be reviewed are reviewed according to the review items, and the area review results of the drawings in the area to be reviewed are generated. The review results of the drawings in the area to be reviewed include the component review results, component pair review results and area review results of the drawings in the area to be reviewed.

[0089] Specifically, select appropriate component identification rules based on the type of drawing. For example, in architectural drawings, wall components can be identified by recognizing continuous straight lines or combinations of multiple lines, combined with certain length and width thresholds; door and window components can be identified based on specific graphic shapes (such as rectangles) and door / window symbols. Check whether the dimensions of the components meet the design specifications and standards. For example, in architectural drawings, the thickness of walls, the width and height of doors and windows, etc., must meet the relevant building codes. Verify that the component's attribute information is complete, such as the component's material, model, and serial number. Missing necessary attribute information may affect subsequent construction and management.

[0090] By analyzing the spatial, functional, and logical relationships between components, interrelated component pairs are identified. Their relative positions and distances on the drawings are checked to determine if spatial connections exist. The mating dimensions and precision between component pairs are verified to meet requirements. For example, in mechanical assembly drawings, the clearance between shafts and holes must meet design requirements; otherwise, assembly quality and mechanical performance will be affected. The functional compatibility of component pairs is assessed. For instance, in building water supply and drainage systems, pipes and valves, faucets, and other components must provide normal water supply and drainage. The design and installation of component pairs are confirmed to comply with relevant specifications and standards. For example, in electrical installations, the connection method between wires and terminals must comply with electrical safety regulations.

[0091] Based on design specifications and standards, detailed review rules were developed, and the drawings for the area to be reviewed were checked item by item. The overall layout of the drawings for the area to be reviewed was checked to ensure it met design requirements and functional requirements. For example, in the architectural floor plan, the layout of each room was examined to ensure it was reasonable and met requirements for lighting, ventilation, and traffic flow. The entire area's design was confirmed to comply with relevant specifications and standards. For example, in fire protection design, the width of evacuation routes and the layout of fire protection facilities were checked to ensure they met fire safety regulations.

[0092] The regional review results of each area of ​​drawings to be reviewed are merged to generate the review results of the CAD drawings to be reviewed.

[0093] The review assignment module is used to determine the target human review end and transmit the CAD drawings to be reviewed and the review results to the target human review end.

[0094] In some embodiments, the review assignment module determines the target human review end, including:

[0095] Based on the component characteristics of the CAD drawings to be reviewed and the review results, the target manual review end is determined.

[0096] Specifically, in the drawing review process, the review assignment module plays a crucial role in assigning CAD drawings awaiting review to appropriate human reviewers. The purpose of this step is to ensure that each drawing is reviewed by personnel with the relevant professional knowledge and skills, thereby improving the accuracy and efficiency of the review and guaranteeing that the drawing quality meets relevant standards and requirements.

[0097] Human reviewers from different professional fields are often more familiar with and skilled at reviewing specific types of components. For example, in architectural drawings, the review of structural components (such as beams, columns, and slabs) requires personnel with expertise in structural engineering; while the review of electrical components (such as distribution boxes and cable trays) requires electrical engineering experts. Based on the identified component type, the drawings are assigned to human reviewers with the corresponding professional background. For example, if the drawings mainly contain structural components, they are assigned to structural engineers for review; if they contain a large number of electrical components, they are assigned to electrical engineers. The complexity of the components directly affects the difficulty and workload of the review. Complex components may require more in-depth professional knowledge and more meticulous review. For example, some irregularly shaped structural components or electrical components with special functions require review by experienced and highly skilled personnel.

[0098] Figure 2 This is a flowchart illustrating an automated drawing review method based on AI technology in one embodiment of this application, as shown below. Figure 2 As shown, an automated drawing review method based on AI technology may include the following steps.

[0099] Determine whether the CAD drawings to be reviewed need to be transmitted in parallel.

[0100] If it is determined whether the CAD drawing to be reviewed needs to be transmitted in parallel, the CAD drawing to be reviewed is divided into blocks, multiple drawing blocks are generated, and multiple drawing blocks are uploaded.

[0101] It merges multiple drawing blocks transmitted in parallel to generate CAD drawings to be reviewed, and also partitions the CAD drawings to be reviewed to generate multiple areas of drawings to be reviewed.

[0102] Based on the component characteristics of each area of ​​drawings to be reviewed, determine the review order of multiple areas of drawings to be reviewed;

[0103] The review is conducted based on the review order of multiple areas of drawings to be reviewed. Based on the component characteristics of the areas of drawings to be reviewed, the review items of the areas of drawings to be reviewed are determined. The areas of drawings to be reviewed are then reviewed according to the review items, and the review results of the areas of drawings to be reviewed are generated.

[0104] Based on the review results of the drawings in each area to be reviewed, the review results of the CAD drawings to be reviewed are generated.

[0105] Based on the component characteristics and review results of the CAD drawings to be reviewed, the target manual review end is determined, and the CAD drawings to be reviewed and the review results are transmitted to the target manual review end.

[0106] An AI-based automated drawing review method can be applied to the aforementioned AI-based automated drawing review system, which will not be elaborated upon here.

[0107] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. An AI technology-based automatic drawing review system, characterized by, The method comprises the following steps: a drawing uploading module is configured to determine whether the CAD drawing to be audited needs to be transmitted in parallel, and if so, to perform block processing on the CAD drawing to be audited, to generate a plurality of drawing blocks, and to upload the plurality of drawing blocks; a drawing preprocessing module is configured to merge the plurality of drawing blocks transmitted in parallel to generate the CAD drawing to be audited, and to perform partition processing on the CAD drawing to be audited to generate a plurality of regional drawings to be audited; a drawing auditing module is configured to determine an auditing order of the plurality of regional drawings to be audited based on the component features of each regional drawing to be audited, to perform auditing based on the auditing order of the plurality of regional drawings to be audited, to determine auditing items of the regional drawing to be audited based on the component features of the regional drawing to be audited, to perform auditing on the regional drawing to be audited according to the auditing items of the regional drawing to be audited, and to generate an auditing result of the regional drawing to be audited, and to generate an auditing result of the CAD drawing to be audited based on the auditing result of each regional drawing to be audited; a review allocation module is configured to determine a target artificial review end and to transmit the CAD drawing to be audited and the auditing result to the target artificial review end; the drawing preprocessing module performs partition processing on the CAD drawing to be audited to generate a plurality of regional drawings to be audited, which comprises: dividing the CAD drawing to be audited into layer drawings according to the layers of the CAD drawing to be audited; for each layer drawing, identifying components in the layer drawing, determining component distribution features of the layer drawing, clustering the components in the layer drawing according to the component distribution features of the layer drawing, and generating a regional drawing to be audited corresponding to the layer drawing; the drawing auditing module determines an auditing order of the plurality of regional drawings to be audited based on the component features of each regional drawing to be audited, which comprises: for each regional drawing to be audited, determining a first weight of the regional drawing to be audited based on the component distribution features of the regional drawing to be audited, determining a second weight of the regional drawing to be audited according to the component weights of the components in the regional drawing to be audited, determining a third weight of the regional drawing to be audited according to the correlation weights of any two components included in the regional drawing to be audited, and determining a comprehensive weight of the regional drawing to be audited according to the first weight, the second weight and the third weight of the regional drawing to be audited; determining the auditing order of the plurality of regional drawings to be audited according to the comprehensive weight of each regional drawing to be audited.

2. The automated plan review system based on AI technology according to claim 1, wherein The drawing uploading module determines whether the CAD drawing to be audited needs to be transmitted in parallel, which comprises: predicting future CAD drawing transmission volume based on CAD drawing transmission volume at a plurality of historical time points; determining a real-time drawing size threshold based on the predicted future CAD drawing transmission volume; determining whether the CAD drawing to be audited needs to be transmitted in parallel according to the drawing size of the CAD drawing to be audited and the real-time drawing size threshold. 3.The AI technology-based automated plan review system of claim 2, wherein, The drawing uploading module performs block processing on the CAD drawing to be audited to generate a plurality of drawing blocks, which comprises: determining an optimal block size based on the predicted future CAD drawing transmission volume; According to the layers of the CAD drawing to be audited, the CAD drawing to be audited is divided into multiple layer blocks; For each layer block, according to the component distribution characteristics and the optimal block size of the layer block, the layer block is processed by block, and the drawing block corresponding to the layer block and the auxiliary merging label of each drawing block are generated.

4. The automated plan review system based on AI technology according to claim 3, wherein, The drawing preprocessing module merges the multiple drawing blocks transmitted in parallel to generate the CAD drawing to be audited, comprising: Based on the auxiliary merging label of each drawing block, the multiple drawing blocks transmitted in parallel are merged to generate the CAD drawing to be audited. 5.The AI technology-based automated drawing review system according to claim 1, wherein, The drawing auditing module determines the auditing items of the region drawing to be audited based on the component characteristics of the region drawing to be audited, comprising: Determine the auditing items of multiple template region drawings; Based on the component characteristics of the region drawing to be audited and the component characteristics of each template region drawing, determine the similar template region drawing; According to the auditing items of the similar template region drawing, determine the auditing items of the region drawing to be audited. 6.The AI technology-based automatic drawing review system according to any one of claims 1-4, characterized in that, The drawing auditing module audits the region drawing to be audited according to the auditing items of the region drawing to be audited, and generates the auditing result of the region drawing to be audited, comprising: Identify the components in the region drawing to be audited, perform component auditing, and generate the component auditing result of the region drawing to be audited; Identify the associated component pairs in the region drawing to be audited, perform component pair auditing, and generate the component pair auditing result of the region drawing to be audited; According to the auditing items of the region drawing to be audited, the region drawing to be audited is audited to generate the region auditing result of the region drawing to be audited, wherein the auditing result of the region drawing to be audited includes the component auditing result, the component pair auditing result and the region auditing result of the region drawing to be audited.

7. The automatic drawing review system based on AI technology according to any one of claims 1-4, characterized in that, The review allocation module determines the target artificial review end, comprising: According to the component characteristics and the auditing result of the CAD drawing to be audited, the target artificial review end is determined.

8. An AI technology-based automatic drawing review method, characterized by, An automatic drawing auditing system based on AI technology, comprising: Determine whether the CAD drawing to be audited needs to be transmitted in parallel; If it is determined whether the CAD drawing to be audited needs to be transmitted in parallel, the CAD drawing to be audited is processed by block to generate multiple drawing blocks, and the multiple drawing blocks are uploaded; Merge the multiple drawing blocks transmitted in parallel to generate the CAD drawing to be audited, and also used for partitioning the CAD drawing to be audited to generate multiple region drawings to be audited; Determine the auditing order of the multiple region drawings to be audited based on the component characteristics of each region drawing to be audited; Based on the auditing order of the multiple region drawings to be audited, auditing is performed, based on the component characteristics of the region drawing to be audited, the auditing items of the region drawing to be audited are determined, the region drawing to be audited is audited according to the auditing items of the region drawing to be audited, and the auditing result of the region drawing to be audited is generated; Based on the auditing result of each region drawing to be audited, the auditing result of the CAD drawing to be audited is generated. According to the component features of the CAD drawing to be audited and the audit result, a target manual review end is determined, and the CAD drawing to be audited and the audit result are transmitted to the target manual review end.

Images