PCB multi-unit intelligent layout and collaborative inspection system for AVI detection

Abstract

The application relates to a PCB multi-unit intelligent layout and collaborative inspection system for AVI detection, and particularly relates to the field of intelligent manufacturing. Through dynamic fusion of production order attributes and PCB design data, an enhanced task object containing value evaluation and an accurate outer frame is automatically generated for each order. According to a multi-objective optimization model, a multi-unit layout scheme that comprehensively optimizes space utilization, order priority and equipment switching cost is intelligently generated under the condition of meeting process and geometric constraints. After virtualization verification ensures feasibility, the detection equipment is driven to perform synchronous visual analysis on multiple PCB units, and the model parameters are continuously optimized through detection efficiency feedback, so that the single detection output efficiency is significantly improved, high-value orders are ensured to be preferentially transferred, and adaptive collaborative optimization of the production scheduling and detection process is realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing, and more specifically, to a PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection. Background Technology

[0002] In the printed circuit board (PCB) manufacturing industry, with the rapid development of electronic products towards personalization and diversification, flexible and agile production has become the mainstream trend. Production lines need to frequently handle orders with multiple varieties, small batches, and high mixing. Among these, automatic visual inspection is a key link in ensuring the quality of the final product, and its inspection efficiency directly affects the overall production cycle and order delivery capability. The AVI visual inspection machine is the core equipment for performing this process. Its working method is usually to place the PCB to be inspected in its fixed inspection area and perform image acquisition and defect analysis on each PCB. However, facing a continuous flow of orders with diverse sizes, different process requirements, and tight delivery deadlines, the equipment utilization and output efficiency under the traditional inspection mode face severe challenges. Production planning often needs to make real-time decisions on how to combine different types of PCB substrates in a single inspection task based on the dynamically changing order queue, so as to make full use of the equipment workbench and reduce the non-productive time loss caused by frequent program switching. This places extremely high demands on the real-time performance and intelligence of the production scheduling and resource optimization of the inspection process.

[0003] In existing technologies, to improve the effective output of a single AVI device, some solutions have attempted to use a multi-PCS synchronous inspection method. This involves using software to arrange and combine multiple PCB units with the same process requirements into a composite graphic file that conforms to the size of the equipment's inspection area. The equipment then scans and inspects multiple units simultaneously. Such solutions typically rely on preset layout templates and simple rule checks, such as arranging a specific number of units in a fixed matrix. However, this layout strategy based on static rules and offline processing has significant limitations. First, its layout logic is isolated from the upper-level production management system, making it impossible to obtain dynamic order queues, precise physical attributes of workpieces, process priorities, and real-time output data from the Manufacturing Execution System (MES) or Enterprise Resource Planning (ERP) system. The inability to handle information leads to a disconnect between the generated layout scheme and the overall production plan, making it merely a passive response rather than an active optimization tool. Secondly, its optimization dimensions are extremely limited, typically considering only a few geometric constraints such as the number of units and the minimum safe distance between units. It is a simple spatial arrangement and fails to incorporate key production optimization factors such as order delivery urgency, switching costs caused by mixed layout of different PCB models, maximizing the utilization of the inspection area, and balancing multiple objectives into a unified decision model for solution. Therefore, existing methods are unable to adaptively generate a comprehensive and optimal panel inspection scheme that matches the dynamic production state in a flexible manufacturing environment, making the AVI inspection station a bottleneck that may still restrict the improvement of overall production efficiency. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection. The system solves the problems mentioned in the background art through a data interface module, a central optimization engine, a layout execution module, and an inspection control module.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is as follows: it specifically includes: a data interface module, a central optimization engine, a typesetting execution module, and a detection and control module connected in sequence, wherein; Data Interface Module: When the production order queue is updated, it retrieves order data containing order attribute information from the Manufacturing Execution System, processes the order data, generates and outputs an ordered list of enhanced task data objects to the central optimization engine; the order attribute information includes part number, delivery date, face number, quantity, and process type obtained from the bill of materials; Central Optimization Engine: Receives an ordered list of enhanced task data objects to obtain the current set of enhanced task data objects. When the AVI visual inspection machine is idle or receives new order data, the central optimization engine calculates the current set of enhanced task data objects based on a predefined multi-objective optimization model. It selects at least one enhanced task data object from this set that meets preset constraints and determines the two-dimensional position coordinates of each selected enhanced task data object within the effective inspection area of ​​the AVI visual inspection machine, thereby generating a panelization instruction. The panelization instruction at least includes the unique identifier information of the selected enhanced task data object and its corresponding two-dimensional position coordinates. The layout execution module receives the imposition instruction, retrieves the corresponding standardized outline graphics from the enhanced task data object based on the unique identifier information of the enhanced task data object contained in the imposition instruction, positions and arranges all the retrieved standardized outline graphics on a virtual canvas corresponding to the effective inspection area size of the AVI appearance inspection machine according to the two-dimensional position coordinates in the imposition instruction, generates a virtual imposition image, and performs boundary and spacing compliance verification on the virtual imposition image; after the verification is passed, the layout execution module converts the virtual imposition image into an imposition graphic file that can be recognized by the AVI appearance inspection machine; Inspection and control module: Loads the panel layout graphic file and controls the AVI visual inspection machine to perform synchronous visual inspection on multiple PCB units laid out according to the virtual panel layout graphic file; In a preferred embodiment, the specific process of generating and outputting an ordered list of enhanced task data objects to the central optimization engine in the data interface module is as follows: Each piece of order data is validated to check whether the delivery date is earlier than the current system time and whether the number of rounds belongs to the preset set of allowed rounds. For order data with a delivery date earlier than the current system time or a number of rounds that does not belong to the set of allowed rounds, it is marked as abnormal data and temporarily stored in an abnormal queue independent of the main processing flow. At the same time, an alarm notification is triggered, and the order data that passes the validation continues to be processed. Calculate a dynamic value assessment score for each verified order data; Generate a corresponding enhanced task data object for each valid order data; Based on the dynamic value assessment score, all generated enhanced task data objects are sorted in descending order from high to low to form an ordered list of enhanced task data objects, which is then sent to the central optimization engine.

[0006] In a preferred embodiment, the process of calculating a dynamic value assessment score for each verified order data is as follows: The four weighting coefficients, which are the sum of the delivery urgency weighting coefficient, the batch size weighting coefficient, the process complexity weighting coefficient, and the customer priority weighting coefficient, are multiplied by the delivery urgency factor, the order batch size factor, the process complexity factor, and the static customer priority factor, and then summed to obtain a dynamic value assessment score.

[0007] In a preferred embodiment, the process of generating a corresponding enhanced task data object for each verified order data is as follows: The enhanced task data object includes the part number, delivery date, face number, and quantity from the order data, as well as the process type obtained from the bill of materials, the calculated dynamic value assessment score, and a globally unique task identifier within this system cycle. During the generation of the enhanced task data object, the data interface module calls a preprocessing subsystem of a communication connection through the part number it contains; the preprocessing subsystem performs the following: locates and loads the original Gerber design file corresponding to the part number, performs parsing and graphic alignment processing on the file including multiple standard layers including circuit layer, solder mask layer, silkscreen layer, extracts the precise outline representing the physical boundary of the PCB, and generates a standardized outline graphic defined by the width and height of the circumscribed rectangle. The standardized outline graphic is associated with the corresponding enhanced task data object, becoming part of that enhanced task data object.

[0008] In a preferred embodiment, the calculation process for the current enhanced task data object set based on a predefined multi-objective optimization model in the central optimization engine specifically includes: The central optimization engine receives an ordered list of enhanced task data objects from the data interface module, which serves as the current set of enhanced task data objects. Retrieve the following attributes for each enhanced task data object in the current enhanced task data object set: width and height of the associated standardized outline graphic, process type, dynamic value assessment score, face number, and unique identifier information; A multi-objective optimization model is constructed with the goal of maximizing the comprehensive utility value of a single assembly. The comprehensive utility value is obtained by multiplying the area utilization rate by the first weight coefficient, adding the weighted value urgency by the second weight coefficient, and then subtracting the process switching cost by the third weight coefficient. The area utilization rate is equal to the sum of the areas of the standardized outline graphics corresponding to all selected enhanced task data objects, divided by the area of ​​the effective inspection area of ​​the AVI appearance inspection machine; the area of ​​the standardized outline graphics is obtained by multiplying its width by its height, and the area of ​​the effective inspection area is obtained by multiplying the preset area length by the area width; The weighted value urgency term is equal to multiplying the dynamic value assessment score of all selected enhanced task data objects by a delivery urgency factor, and then summing all the product results. The process switching cost item is a binary indicator. When the dominant process type associated with the enhanced task data object in the current layout scheme is different from the dominant process type associated with the enhanced task data object in the previous layout scheme recorded by the AVI appearance inspection machine, the indicator value is one; otherwise, the value is zero. The calculation process satisfies the following constraints: all enhanced task data objects selected and placed in the same panel scheme have the same face number; all selected standardized outline graphics, after being placed according to the determined two-dimensional position coordinates, are located within the boundary of the effective inspection area, no two graphics overlap, and the minimum distance between them is not less than a safety distance value preset according to the equipment positioning accuracy and inspection requirements.

[0009] In a preferred embodiment, the specific operation for generating the imposition instruction is as follows: The central optimization engine uses a heuristic search algorithm, taking the current enhanced task data object set, the length and width of the effective inspection area of ​​the AVI appearance inspection machine, the safety distance value, and the dominant process type of the previous panelization scheme as input; Based on a multi-objective optimization model, optimization calculations are performed to obtain a feasible solution that maximizes the overall utility value. The heuristic search algorithm is one of the following: genetic algorithm, simulated annealing algorithm, or tabu search algorithm; Once a feasible solution is found that satisfies all constraints and has a satisfactory overall utility value, the central optimization engine generates a panelization instruction based on that feasible solution. This panelization instruction is a list, and each record in the list contains a unique identifier for a selected enhanced task data object, as well as the x-coordinate and y-coordinate values ​​of the two-dimensional position coordinates of the standardized outline graphic corresponding to that enhanced task data object.

[0010] In a preferred embodiment, in the layout execution module, firstly, the layout instruction is parsed to extract the unique identifier information of all enhanced task data objects contained therein; based on this unique identifier information, the width and height of the corresponding standardized outline graphic are obtained; combined with the two-dimensional position coordinates specified for each enhanced task data object in the layout instruction, a graphic instance containing its unique identifier information, width, height, and position coordinates is constructed for each processed enhanced task data object; and a virtual layout image is generated on the virtual canvas based on all such graphic instances. The specific steps for verifying the compliance of boundaries and spacing in a virtual mosaic image are as follows: Secondly, perform basic geometric constraint checks on the virtual jigsaw puzzle. Specifically, for each graphic instance, use the horizontal coordinate value of its two-dimensional position coordinates as the left boundary position, the horizontal coordinate value plus the width as the right boundary position, the vertical coordinate value as the bottom boundary position, and the vertical coordinate value plus the height as the top boundary position. Check whether the left and bottom boundary positions of all graphic instances are not less than zero, whether the right boundary position of all graphic instances is not greater than the length of the virtual canvas, and whether the top boundary position of all graphic instances is not greater than the width of the virtual canvas. Next, perform the inter-graphic interference and safety distance verification, specifically: For any two different graphic instances, calculate the horizontal projection overlap gap between them, which is equal to zero or the maximum of the following two differences: the larger of the left boundary positions of the corresponding graphics of the two graphic instances, minus the smaller of the right boundary positions of the corresponding graphics of the two graphic instances; calculate the vertical projection overlap gap between them, which is equal to zero or the maximum of the following two differences: the larger of the lower boundary positions of the corresponding graphics of the two graphic instances, minus the smaller of the upper boundary positions of the corresponding graphics of the two graphic instances; based on the horizontal and vertical projection overlap gaps, calculate the effective distance between the corresponding graphics of the two graphic instances; verify whether the effective distance between any two corresponding graphics of the graphic instances is greater than or equal to a preset engineering safety distance value; Finally, the graphic fill completeness and layout density assessment and verification are performed. Specifically, a graphic fill completeness index is calculated. The calculation process of the graphic fill completeness index is as follows: the sum of the first part value and the second part value is calculated. The first part value is calculated by multiplying the width and height of the standardized outer frame of each graphic instance to obtain the individual area, and then adding the individual areas of all graphic instances to obtain the total area of ​​the placed graphic. Then, this total area is divided by the total area of ​​the virtual canvas obtained by multiplying the length and width of the virtual canvas. The second part value is calculated by dividing the density weight factor by the sum of the effective spacing between the number one and all different graphic instance pairs, and verifying whether the graphic fill completeness index is greater than a preset fill completeness threshold.

[0011] In a preferred embodiment, the process of converting the virtual mosaic image into a mosaic graphic file recognizable by an AVI visual inspection machine specifically involves: After passing all compliance checks, the layout execution module performs a manufacturability risk assessment on the virtual mosaic. This assessment includes: calculating the size and position of the overall bounding rectangle formed by all standardized outline graphics in the current mosaic scheme; the left boundary of the overall bounding rectangle is equal to the minimum value among the left boundary positions of all graphic instances; the right boundary is equal to the maximum value among the right boundary positions of all graphic instances; the lower boundary is equal to the minimum value among the lower boundary positions of all graphic instances; and the upper boundary is equal to the maximum value among the upper boundary positions of all graphic instances. The module also calculates the weight of all standardized outline graphics on the virtual canvas. The centroid distribution coordinates are calculated as follows: the x-coordinate is equal to the sum of the products of the individual areas of all graphic instances and their respective x-coordinates, divided by the total area of ​​the placed graphic; the y-coordinate is equal to the sum of the products of the individual areas of all graphic instances and their respective y-coordinates, divided by the total area of ​​the placed graphic. The proximity of the overall bounding rectangle to the virtual canvas boundary and the deviation of the centroid coordinates from the canvas center are compared with the preset allowable extreme values ​​for proximity and deviation, respectively. If the allowable extreme values ​​are not exceeded, the manufacturability risk is judged to be low, and a risk warning log is generated; if the allowable extreme values ​​are exceeded, a high-risk warning is generated. Subsequently, the layout execution module performs device-specific format conversion, encoding the verified and risk-predicted virtual layout results according to the dedicated graphic file format specified by the target AVI appearance inspection machine. The virtual layout results include the unique identifier information of each enhanced task data object in the imposition instruction, as well as the width, height, and two-dimensional position coordinates of its corresponding standardized outer frame graphic. According to the equipment requirements, the encoding process maps the position, size, and unique identifier of each graphic to an instruction sequence or graphic element description that the device controller can parse, ultimately generating a structured imposition graphic file containing complete layout information.

[0012] In a preferred embodiment, the specific process of loading the panel graphic file and controlling the AVI appearance inspection machine to perform synchronous visual inspection based on the panel graphic file in the detection control module is as follows: First, the mosaic graphic file is parsed to obtain the unique identifier information of all the enhanced task data objects contained therein, as well as the width, height, and two-dimensional position coordinates of the standardized outline graphic corresponding to each enhanced task data object. The center point coordinates of the standardized outline graphic are calculated based on its two-dimensional position coordinates, and the Euclidean distance between each center point coordinate and the preset scanning start point is calculated. Then, this distance is divided by a preset distance reference value and normalized to obtain a dimensionless relative distance value. Secondly, a detection priority weight is calculated for each enhanced task data object. The value of the detection priority weight is equal to the dynamic value assessment score attached to the enhanced task data object, multiplied by a distance decay factor. The distance decay factor is based on the natural constant, and its exponent is the negative value obtained by multiplying the relative distance value by a preset distance decay coefficient that is greater than zero. Next, based on all the calculated detection priority weights, a heuristic path planning algorithm is used to generate one or more camera scanning paths; Finally, the generated camera scanning path, along with the unique identifier of each enhanced task data object and the precise coordinates and dimensions of its corresponding standardized outline graphic, are sent to the AVI visual inspection machine via the device communication protocol. The AVI visual inspection machine is then controlled to perform the following sequential actions: based on the sent camera scanning path and coordinate information, the motion system is driven to precisely position the first standardized outline graphic to be scanned within the camera's field of view; the camera is controlled to acquire images according to the path plan, and based on the process type associated with the unique identifier of each graphic, the corresponding visual inspection program is called to synchronously analyze the corresponding areas of multiple different standardized outline graphics covered in a single acquired image, thereby achieving synchronous visual inspection of multiple PCB units in a single scanning process.

[0013] In a preferred embodiment, during and after the synchronous visual inspection is performed by the AVI appearance inspection machine, the inspection control module also performs the following operations: During the synchronous visual inspection process, the average image contrast and system computing load rate are monitored in real time, and the local illumination intensity or computing resource allocation is dynamically fine-tuned according to the preset thresholds for the average image contrast and system computing load rate. After the inspection is completed, two types of data are collected and integrated: the first type is the defect detection results of the PCB unit corresponding to each enhanced task data object, including the good status, defect type and location coordinates; the second type is the process efficiency data of this panel inspection, including the actual total time spent from the start to the end of this inspection task, the actual scanning order of each enhanced task data object in the scanning path, and the average deviation between the scanning position and the theoretical position monitored online. The inspection and control module sends the defect detection results to the manufacturing execution system and packages the process performance data, sending it back to the central optimization engine through the feedback channel. The central optimization engine uses the process performance data to compare the actual total time with the model's predicted time to correct the inspection speed estimation parameters, analyzes the conformity between the actual scanning sequence and the theoretical inspection priority weights to optimize the distance attenuation coefficient used when calculating the inspection priority weights, and uses the average position deviation to assess the adequacy of the preset engineering safety distance value for layout verification. This enables iterative learning and adaptive adjustment of the inspection speed estimation parameters, distance attenuation coefficients, and engineering safety distance values ​​in the multi-objective optimization model.

[0014] The beneficial effects of this invention are as follows: By dynamically integrating production order attributes and PCB design data, an enhanced task object containing value assessment and precise outline is automatically generated for each order. Based on a multi-objective optimization model, it calculates in real time and intelligently generates a multi-unit panelization scheme that optimizes space utilization, order priority, and equipment switching costs while meeting process and geometric constraints. After ensuring feasibility through virtualization verification, it drives the inspection equipment to perform synchronous visual analysis on multiple PCB units and continuously optimizes model parameters through inspection performance feedback. This significantly improves the output efficiency of a single inspection, ensures the priority flow of high-value orders, and achieves adaptive collaborative optimization of the scheduling and inspection processes. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a block diagram of the system structure of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0018] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0019] Example 1 This embodiment provides, for example Figure 1-2 The system illustrates a PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection, specifically comprising: a data interface module, a central optimization engine, a layout execution module, and an inspection control module connected in sequence; wherein; Data Interface Module: When the production order queue is updated, it retrieves order data containing order attribute information from the manufacturing execution system, processes the order data, generates and outputs an ordered list of enhanced task data objects to the central optimization engine; the order attribute information includes part number, delivery date, face number and quantity, as well as the process type (such as immersion gold, tin plating) obtained from the bill of materials. Central Optimization Engine: Receives an ordered list of enhanced task data objects to obtain the current set of enhanced task data objects. When the AVI visual inspection machine is idle or receives new order data, the central optimization engine calculates the current set of enhanced task data objects based on a predefined multi-objective optimization model. It selects at least one enhanced task data object from this set that meets preset constraints and determines the two-dimensional position coordinates of each selected enhanced task data object within the effective inspection area of ​​the AVI visual inspection machine, thereby generating a panelization instruction. The panelization instruction at least includes the unique identifier information of the selected enhanced task data object and its corresponding two-dimensional position coordinates. The layout execution module receives the imposition instruction, retrieves the corresponding standardized outline graphics from the enhanced task data object based on the unique identifier information of the enhanced task data object contained in the imposition instruction, positions and arranges all the retrieved standardized outline graphics on a virtual canvas corresponding to the effective inspection area size of the AVI appearance inspection machine according to the two-dimensional position coordinates in the imposition instruction, generates a virtual imposition image, and performs boundary and spacing compliance verification on the virtual imposition image; after the verification is passed, the layout execution module converts the virtual imposition image into an imposition graphic file that can be recognized by the AVI appearance inspection machine; Inspection and control module: Loads the panel layout graphic file and controls the AVI visual inspection machine to perform synchronous visual inspection on multiple PCB units laid out according to the virtual panel layout graphic file.

[0020] In this embodiment, the specific process of generating and outputting an ordered list of enhanced task data objects to the central optimization engine in the data interface module is as follows: Each acquired order data is validated to ensure that the delivery date is not earlier than the current system time and that the number of facets belongs to the preset set of allowed facets. Validity validation can also be extended to include checking other business rules such as whether the order quantity is greater than zero. The preset set of allowed facets typically includes, but is not limited to, "component facets" and "welding facets," and is pre-set by the system based on the actual inspection capabilities of the connected AVI visual inspection machine. The "not earlier than the current system time" check means that the order's delivery date and timestamp must be later than or equal to the system clock time when the data interface module performs the validation operation. This validation mechanism ensures that orders that are expired or impossible to deliver on time will not be processed. The system allocates resources for task detection. For order data with a delivery date earlier than the current system time or a face count not belonging to the allowed face count set, it is marked as abnormal data and temporarily stored in an abnormal queue independent of the main processing flow. At the same time, an alarm notification is triggered. The alarm notification is implemented through audio and visual prompts, message push to the manufacturing execution system operation interface, or email to preset maintenance personnel. The independent abnormal queue is used to archive all order data records that failed verification for subsequent manual review and processing, preventing them from entering the automatic scheduling process. Order data that passed verification continues to be processed. This design achieves the robustness of the system, decouples data abnormalities from the core processing flow, and prevents the entire automated process from being interrupted due to individual erroneous data. A dynamic value assessment score is calculated for each verified order data. The dynamic value assessment score is used to quantify the overall priority of the order in subsequent layout and production scheduling decisions. The higher the score, the greater the value of prioritizing the order. This dynamic value assessment score unifies multiple heterogeneous dimensions such as the urgency of the order's delivery date, production batch, process complexity, and customer importance into a comparable scalar value, providing a quantitative decision-making basis for the central optimization engine. For each verified order data, a corresponding enhanced task data object is generated; the generation operation is a key step in data form transformation, which encapsulates the original order attributes from the manufacturing execution system into a structured object containing a globally unique identifier and standardized value score, which can be directly processed by downstream optimization and execution modules; Based on the dynamic value assessment score, all generated enhanced task data objects are sorted in descending order from high to low to form an ordered list of enhanced task data objects, which is then sent to the central optimization engine. This pre-sorting operation applies an initial global priority before the data enters the optimization engine, so that when subsequent layout optimization searches for feasible solutions, it can naturally tend to prioritize the combination of high-value tasks. Thus, while pursuing the local optimum of a single layout, it guides the system as a whole to evolve towards processing high-value order flows, achieving an initial synergy between global efficiency and value gain. The process of calculating a dynamic value assessment score for each verified order is as follows: The four weighting coefficients, which are the sum of the delivery urgency weighting coefficient, the batch size weighting coefficient, the process complexity weighting coefficient, and the customer priority weighting coefficient, are multiplied by the delivery urgency factor, the order batch size factor, the process complexity factor, and the static customer priority factor, and then summed to obtain a dynamic value assessment score. As an example, a set of feasible weighting coefficient configurations is as follows: the weighting coefficient for delivery urgency is set to 0.5, the weighting coefficient for batch size is set to 0.2, the weighting coefficient for process complexity is set to 0.2, and the weighting coefficient for customer priority is set to 0.1. These coefficients can be adjusted in the system configuration interface according to the production strategy (such as focusing on on-time delivery or focusing on batch efficiency). The calculation process of the delivery urgency factor is as follows: take the natural constant as the base, subtract the negative value of the difference between the current system time and the order delivery time, divide it by a preset time sensitivity coefficient that is greater than zero, and use it as an index for calculation. The resulting function value is the delivery urgency factor. The "time sensitivity coefficient" is a preset, positive constant used to adjust the intensity of the impact of delivery urgency on the factor value. For example, it can be set to 24 (unit: hours). This means that for orders within 24 hours of the delivery date, the delivery urgency factor will rapidly approach 1 as time approaches; while for orders several days away from the delivery date, the factor value will approach 0. This exponential decay calculation method, compared to linear calculation, can more sensitively reflect the extreme urgency of tasks approaching the delivery date. The calculation process for the order batch size factor is as follows: take 10 as the base, add one to the quantity as the argument, and the resulting logarithmic value is the order batch size factor. The purpose of using a base-10 logarithmic function to handle the quantity is to smooth out the huge impact of the difference in order of magnitude. For example, the factor of an order with a quantity of 1 is approximately log10(2)≈0.3; the factor of an order with a quantity of 9 is approximately log10(10)=1; and the factor of an order with a quantity of 999 is approximately log10(1000)=3. This method avoids a single super-large batch order completely dominating the sorting, while reserving a reasonable priority space for small and medium batch orders, which is in line with the consideration of order diversity in flexible manufacturing. The process complexity factor is obtained by querying a preset process type and complexity coefficient mapping table to obtain the value corresponding to the process type obtained from the bill of materials. This value is the process complexity factor. The "preset process type and complexity coefficient mapping table" is pre-established based on statistical analysis of historical production data. For example, the complexity coefficient of the "tin spraying" process can be defined as 1.0 (as a baseline), the coefficient of the "immersion gold" process is set to 1.3 because the process is more complex and the testing requirements are higher, and the coefficient of the "immersion silver" process is set to 1.1. This mapping table allows administrators to maintain and update it based on the actual testing time experience value of the production line. The static customer priority coefficient is obtained directly from the manufacturing execution system, or assigned a value in the range of one to three according to the preset customer importance rules. For example, the order priority coefficient of key customers can be preset to 3.0, important customers to 2.0, and ordinary customers to 1.0. This coefficient, as a static business rule, is combined with dynamically calculated factors such as delivery time and batch size to form a complete value assessment model. The process of generating a corresponding enhanced task data object for each validated order data is as follows: The enhanced task data object includes the part number, delivery date, face number, and quantity from the order data, as well as the process type obtained from the bill of materials, the calculated dynamic value assessment score, and a globally unique task identifier within this system cycle. The globally unique task identifier can be generated in the format of "year-month-day-hour-minute-second-millisecond-sequence number", such as "20240520103015001", where "20240520103015" is the current timestamp and "001" is the sequence number generated within the same millisecond. This design ensures the uniqueness of the identifier even under extremely high concurrency. This enhanced task data object exists in memory in a specific data structure (such as a JSON object or class instance), and its fields correspond one-to-one with database table fields or data units of subsequent communication protocols, which facilitates serialization transmission and persistent storage. During the generation of the enhanced task data object, the data interface module calls a preprocessing subsystem of a communication connection through the part number it contains; the preprocessing subsystem performs the following: locates and loads the original Gerber design file corresponding to the part number, performs parsing and graphic alignment processing on the file including multiple standard layers including circuit layer, solder mask layer, silkscreen layer, extracts the precise outline representing the physical boundary of the PCB, and generates a standardized outline graphic defined by the width and height of the circumscribed rectangle. The extraction process of the standardized outline graphic includes: reading the coordinate data of the Gerber file and identifying the outermost boundary points of all graphic elements; using the convex hull algorithm or the minimum bounding rectangle algorithm to calculate the smallest rectangle that can enclose all graphic elements; and recording the coordinates of the lower left corner, the width value, and the height value of the rectangle as the definition parameters of the standardized outline graphic. This process unifies PCB design graphics of arbitrary shapes into regular rectangular representations, which greatly simplifies the complexity of subsequent layout geometric calculations. The standardized outline graphic is associated with the corresponding enhanced task data object, becoming part of that enhanced task data object. The association method is to add two fields, "BoundingBox_Width" and "BoundingBox_Height", to the data structure of the enhanced task data object to store the width and height values ​​of the standardized outline graphic. This association makes the task object itself carry all the key geometric information required for layout optimization.

[0021] In this embodiment, it is specifically necessary to explain that the calculation process of the current enhanced task data object set based on the predefined multi-objective optimization model in the central optimization engine is as follows: The central optimization engine receives an ordered list of enhanced task data objects from the data interface module, which serves as the current set of enhanced task data objects. Retrieve the following attributes for each enhanced task data object in the current enhanced task data object set: width and height of the associated standardized outline graphic, process type, dynamic value assessment score, face number, and unique identifier information; A multi-objective optimization model is constructed with the goal of maximizing the overall utility value of a single assembly. The overall utility value is obtained by multiplying the area utilization rate by the first weight coefficient, adding the weighted value urgency by the second weight coefficient, and then subtracting the process switching cost by the third weight coefficient. The first, second, and third weight coefficients are all positive numbers and their sum is one. As an exemplary, and not restrictive, configuration, a typical set of weighting coefficients is as follows: the first weighting coefficient (area utilization rate weight) is 0.4, the second weighting coefficient (weighted value urgency weight) is 0.5, and the third weighting coefficient (process changeover cost weight) is 0.1. This configuration reflects the principle of prioritizing the delivery of high-value, high-urgency orders while ensuring high equipment space utilization, and appropriately considering the efficiency of continuous equipment operation. The system provides a configuration interface that allows users to dynamically adjust these coefficients according to actual production strategies. The area utilization rate is equal to the sum of the areas of the standardized outline graphics corresponding to all selected enhanced task data objects, divided by the area of ​​the effective inspection area of ​​the AVI appearance inspection machine; the area of ​​the standardized outline graphics is obtained by multiplying its width by its height, and the area of ​​the effective inspection area is obtained by multiplying the preset area length by the area width; The length and width of the effective inspection area are fixed physical parameters of the AVI visual inspection machine. For example, the length can be set to 400 mm and the width to 250 mm. The calculation of the area utilization rate directly reflects the utilization efficiency of the workbench area of ​​the expensive inspection equipment and is a key indicator for reducing the unit inspection cost. The weighted value urgency item is equal to multiplying the dynamic value assessment score of all selected enhanced task data objects by a delivery urgency factor, and then summing all the product results. The delivery urgency factor is calculated by dividing the negative value of the difference between the current system time and the order delivery date by a preset time sensitivity coefficient that is greater than zero, with the natural constant as the base. The preset time sensitivity coefficient is used to control the decay rate of the impact of delivery urgency. For example, it can be set to 24 (unit: hours). This means that for an order with 24 hours left before the delivery date, its delivery urgency factor is about 0.368; for an order with only 1 hour left before the delivery date, its delivery urgency factor is about 0.96, significantly amplifying its urgency. This exponential decay model can highlight the priority of tasks that are close to the deadline more effectively than the linear model. The process switching cost item is a binary indicator. When the dominant process type associated with the enhanced task data object in the current layout scheme is different from the dominant process type associated with the enhanced task data object in the previous layout scheme recorded by the AVI appearance inspection machine, the indicator value is one; otherwise, the value is zero. The determination rule for the "dominant process type" can be as follows: if all enhanced task data objects in the panel layout have the same process type, then that type is the dominant process type; if there are multiple process types, then the process type with the most occurrences is selected as the dominant process type. The introduction of the process switching cost item quantifies the non-productive time loss caused by changing the detection program, adjusting optical parameters, etc., and guides the optimization engine to merge orders with the same process type as much as possible in continuous production scheduling, thereby improving the overall equipment efficiency (OEE). The calculation process satisfies the following constraints: all enhanced task data objects selected and placed in the same panel scheme have the same face number; all selected standardized outline graphics, after being placed according to the determined two-dimensional position coordinates, are located within the boundary of the effective inspection area, no two graphics overlap, and the minimum distance between them is not less than a safety distance value preset according to the equipment positioning accuracy and inspection requirements. The safety clearance value is a key engineering parameter used to prevent interference between adjacent PCBs during inspection due to positioning errors or mechanical vibrations, and to ensure that the optical inspection system has sufficient field of view isolation. This value can be set comprehensively based on the positioning accuracy of the equipment and the size of the PCB. For example, it can be set to 2 mm. When calculating the minimum distance, it is necessary to consider that the standardized outer frame is a rectangle. Therefore, the calculation of the minimum distance between rectangles can be simplified to calculating the projection interval in the horizontal and vertical directions. The specific steps for generating imposition instructions are as follows: The central optimization engine uses a heuristic search algorithm, taking the current enhanced task data object set, the length and width of the effective inspection area of ​​the AVI appearance inspection machine, the safety distance value, and the dominant process type of the previous panelization scheme as input; Based on a multi-objective optimization model, optimization calculations are performed to obtain a feasible solution that maximizes the overall utility value. The feasible solution is defined as: the unique identifier information of one or more enhanced task data objects selected from the current set of enhanced task data objects for this assembly, and the two-dimensional position coordinates within the valid verification area assigned to each standardized outline graphic corresponding to these enhanced task data objects. The "optimization calculation" is specifically completed by the optimization solver built into the central optimization engine. Taking the genetic algorithm as an example, its implementation steps include: 1) Encoding: Encoding possible puzzle schemes (i.e., which task objects are selected and their positions) into chromosomes (such as binary encoding or real number encoding); 2) Initializing the population: Randomly generating a set of initial puzzle schemes; 3) Fitness evaluation: Using the comprehensive utility value U as the fitness function, evaluating the merits of each chromosome; 4) Selection, crossover, and mutation: Performing selection operations based on fitness, and performing crossover and mutation on the selected chromosomes to generate a new population; 5) Iteration: Repeating steps 3-4 until the preset maximum number of iterations is reached or the fitness converges; 6) Decoding: Decoding the optimal chromosome into a feasible solution. The entire optimization process is completed within seconds to meet the requirements of real-time response in the production site. The heuristic search algorithm is one of the genetic algorithm, simulated annealing algorithm, or tabu search algorithm; these algorithms are all commonly used approximate algorithms for solving NP-hard two-dimensional rectangular layout optimization problems, and can find high-quality feasible solutions in a reasonable time, balancing the quality of the solution and computational efficiency. Once a feasible solution is found that satisfies all constraints and meets the overall utility value requirements, the central optimization engine generates a panelization instruction based on the feasible solution. The panelization instruction is a list, and each record in the list contains a unique identifier for a selected enhanced task data object, as well as the x and y coordinates of the two-dimensional position coordinates determined for the standardized outline graphic corresponding to the enhanced task data object. Imposition instructions can be organized using a structured data exchange format (such as JSON). For example, a complete imposition instruction can contain the following fields: batch_id (the batch number of this imposition), layout_width, layout_height (the effective area size), and an array of panels. Each element in the array corresponds to a selected task object, containing its task_id, panel_x, and panel_y coordinates. This instruction format is clear and machine-readable, providing a solid foundation for the accurate parsing and execution of the typesetting execution module.

[0022] In this embodiment, it is specifically necessary to explain that in the layout execution module, firstly, the layout instruction is parsed to extract the unique identification information of all enhanced task data objects contained therein; based on these unique identification information, the width and height of the corresponding standardized outer frame graphic are obtained; combined with the two-dimensional position coordinates specified for each enhanced task data object in the layout instruction, a graphic instance containing its unique identification information, width, height and position coordinates is constructed for each processed enhanced task data object; and a virtual layout image is generated on the virtual canvas based on all such graphic instances. A “graphic instance” is a data structure temporarily built in memory that binds logical layout instructions from the central optimization engine to specific geometric attributes from enhanced task data objects, forming a graphic object with an identity that can be directly manipulated by verification and drawing software. The process of building a graphical instance completes the mapping from “task identifier-coordinate” pairs to “specific graphic-coordinate” pairs, which is the basis for performing subsequent physical space simulation verification. The specific steps for verifying the compliance of boundaries and spacing in a virtual mosaic image are as follows: Secondly, perform basic geometric constraint checks on the virtual jigsaw puzzle. Specifically, for each graphic instance, use the horizontal coordinate value of its two-dimensional position coordinates as the left boundary position, the horizontal coordinate value plus the width as the right boundary position, the vertical coordinate value as the bottom boundary position, and the vertical coordinate value plus the height as the top boundary position. Check whether the left and bottom boundary positions of all graphic instances are not less than zero, whether the right boundary position of all graphic instances is not greater than the length of the virtual canvas, and whether the top boundary position of all graphic instances is not greater than the width of the virtual canvas. This verification is the first step in ensuring the physical feasibility of the panelization scheme, preventing the theoretical position of any PCB unit from exceeding the actual mechanical travel range of the AVI equipment's worktable, and is a prerequisite for subsequent more complex verifications. Next, perform the inter-graphic interference and safety distance verification, specifically: For any two different graphic instances, calculate the horizontal projection overlap gap between them. This projection overlap gap is equal to zero or the maximum of the following two differences: the larger of the left boundary positions of the corresponding graphics of the two graphic instances, minus the smaller of the right boundary positions of the corresponding graphics of the two graphic instances; calculate the vertical projection overlap gap between them. This gap is equal to zero or the maximum of the following two differences: the larger of the lower boundary positions of the corresponding graphics of the two graphic instances, minus the smaller of the upper boundary positions of the corresponding graphics of the two graphic instances; based on the horizontal and vertical projection overlap gaps, calculate the effective distance between the corresponding graphics of the two graphic instances. This effective distance is equal to the square of the horizontal projection overlap gap plus the square of the vertical projection overlap gap, and then taking the square root of the sum; verify whether the effective distance between any two corresponding graphics of the graphic instances is greater than or equal to a preset engineering safety distance value. The "projection overlap gap" calculation is the core algorithm for determining whether two rectangles overlap and calculating the distance between non-overlapping areas. It cleverly handles three cases: graphic separation, contact, and overlap by comparing the maximum value with zero. When the graphics are separated, a positive gap is obtained; when they are in contact, it is zero; and when they overlap, it is negative but is zeroed by the max(0,...) function, thus providing the correct input for the effective gap calculation. The "engineering safety gap value" needs to take into account the mechanical positioning accuracy of the AVI equipment, the dimensional tolerance of the PCB board itself, and the repeatability of the loading mechanism. For example, it can be set to 2.0 mm. This verification aims to eliminate the risk of physical collisions between the PCB and the PCB during the inspection process due to accumulated errors and to ensure that the optical camera has a clear field of view to isolate adjacent boards. Finally, the graphic fill completeness and layout density assessment and verification are performed. Specifically, a graphic fill completeness index is calculated. The calculation process of the graphic fill completeness index is as follows: the sum of the first part value and the second part value is calculated. The first part value is calculated by multiplying the width and height of the standardized outer frame of each graphic instance to obtain the individual area, then adding the individual areas of all graphic instances to obtain the total area of ​​the placed graphic, and then dividing this total area by the total area of ​​the virtual canvas obtained by multiplying the length and width of the virtual canvas. The second part value is calculated by dividing a density weight factor greater than zero by the sum of the effective spacing between the number one and all different graphic instance pairs, and verifying whether the graphic fill completeness index is greater than a preset fill completeness threshold. The "Graphic Fill Completeness Index" is a composite evaluation index. Its "first part value" is the traditional area utilization rate, reflecting the degree of space occupation. Its "second part value" innovatively introduces the evaluation of layout compactness. When all graphics are arranged closely (the sum of effective spacing is small), this value increases, thereby improving the overall index and encouraging compact layout. The "compactness weight factor" is used to adjust the weight of "compactness" in the overall index, for example, it can be set to 0.1. The "fill completeness threshold" is used to set the passing line, for example, it can be set to 0.85. This verification comprehensively evaluates the layout quality from two dimensions: "space occupation" and "graphic aggregation degree", and selects the scheme that is not only usable but also makes efficient use of space, avoiding low-quality layouts that meet the safety spacing but have graphics that are too loose and waste effective area. The specific process of converting a virtual mosaic image into an AVI graphic file that can be recognized by an appearance inspection machine is as follows: After passing all compliance checks, the layout execution module performs a manufacturability risk assessment on the virtual mosaic. This assessment includes: calculating the size and position of the overall bounding rectangle formed by all standardized outline graphics in the current mosaic scheme; the left boundary of the overall bounding rectangle is equal to the minimum value among the left boundary positions of all graphic instances; the right boundary is equal to the maximum value among the right boundary positions of all graphic instances; the lower boundary is equal to the minimum value among the lower boundary positions of all graphic instances; and the upper boundary is equal to the maximum value among the upper boundary positions of all graphic instances. The module also calculates the weight of all standardized outline graphics on the virtual canvas. The centroid distribution coordinates are calculated as follows: the x-coordinate is equal to the sum of the products of the individual areas of all graphic instances and their respective x-coordinates, divided by the total area of ​​the placed graphic; the y-coordinate is equal to the sum of the products of the individual areas of all graphic instances and their respective y-coordinates, divided by the total area of ​​the placed graphic. The proximity of the overall bounding rectangle to the virtual canvas boundary and the deviation of the centroid coordinates from the canvas center are compared with the preset allowable extreme values ​​for proximity and deviation, respectively. If the allowable extreme values ​​are not exceeded, the manufacturability risk is judged to be low, and a risk warning log is generated; if the allowable extreme values ​​are exceeded, a high-risk warning is generated. The "overall circumscribed rectangle" reflects the maximum envelope space occupied by the panel layout on the equipment table. The "closeness" of its boundary to the canvas boundary can be quantified by calculating the distance from each side of the circumscribed rectangle to the corresponding edge of the canvas. For example, an allowable extreme value of 5 mm is set, meaning that the distance between any side of the circumscribed rectangle and the corresponding edge of the canvas should not be less than 5 mm to prevent part of the PCB from exceeding the camera's field of view due to table positioning deviation. The "center of gravity distribution coordinates" reflect the position of the center of mass of the panel layout. The "deviation" of its center from the canvas center can be quantified by calculating the Euclidean distance between the two. For example, an allowable extreme value of 10% of the canvas diagonal length is set. Excessive deviation of the center of gravity may cause additional vibration or uneven force on the equipment during high-speed movement. This predictive mechanism adds a forward-looking assessment of physical manufacturability on the basis of geometric verification, improving the stability and reliability of the solution in actual production. Subsequently, the layout execution module performs a device-specific format conversion, encoding the verified and risk-predicted virtual layout results according to the dedicated graphic file format specified by the target AVI appearance inspection machine. The virtual layout results include the unique identifier information of each enhanced task data object in the imposition instruction, as well as the width, height, and two-dimensional position coordinates of its corresponding standardized outline graphic. According to the equipment requirements, the encoding process maps the position, size, and unique identifier of each graphic to an instruction sequence or graphic element description that the equipment controller can parse, ultimately generating a structured imposition graphic file containing complete layout information. This file can be directly loaded by the inspection control module to drive the AVI appearance inspection machine to perform synchronous visual inspection. The "dedicated graphic file format" varies depending on the AVI device manufacturer. Common formats include text scripts containing specific instruction sets, DXF files containing vector graphics information, or binary formats customized by the device manufacturer. The encoding process requires accurately converting the x-coordinate, y-coordinate, width, height, and associated unique identifier information of each graphic instance into the syntax and data structure required by the target format. For example, for AVI devices that support scripts, a series of instruction lines such as PANEL_ID=TaskID_001,X=10.5,Y=20.0,WIDTH=50.0,HEIGHT=80.0 may be generated. This step is the "language translation" link in the interaction between the system and the physical device, and its accuracy and compatibility directly determine whether the intelligent typesetting scheme can be correctly recognized and executed by the device.

[0023] In this embodiment, it is specifically necessary to explain the process by which the detection control module loads the panel graphic file and controls the AVI appearance inspection machine to perform synchronous visual inspection based on the panel graphic file, as follows: First, the mosaic graphic file is parsed to obtain the unique identifier information of all the enhanced task data objects contained therein, as well as the width, height, and two-dimensional position coordinates of the standardized outline graphic corresponding to each enhanced task data object. The center point coordinates of the standardized outline graphic are calculated based on its two-dimensional position coordinates, and the Euclidean distance between each center point coordinate and the preset scanning start point is calculated. Then, this distance is divided by a preset distance reference value and normalized to obtain a dimensionless relative distance value. The "preset distance reference value" can usually be set to the diagonal length of the effective inspection area of ​​the AVI visual inspection machine, thereby normalizing the Euclidean distance and ensuring that the relative distance values ​​calculated on devices of different sizes are comparable. This reference value is adjustable as a system configuration parameter. Secondly, a detection priority weight is calculated for each enhanced task data object. The value of the detection priority weight is equal to the dynamic value assessment score attached to the enhanced task data object, multiplied by a distance decay factor. The distance decay factor is based on the natural constant, and its exponent is the negative value obtained by multiplying the relative distance value by a preset distance decay coefficient that is greater than zero. The "distance attenuation coefficient" is used to control the attenuation intensity of the impact of physical distance on detection priority. For example, it can be set to 2.0. The larger the coefficient, the faster the distance attenuation factor of PCB units that are farther away from the scanning start point decreases, and the less priority they are given in path planning. This calculation model integrates the "order value" from upstream scheduling decisions with the "physical distance" that affects the efficiency of equipment movement at the control layer. It aims to guide the equipment to prioritize the completion of high-value and conveniently located detection tasks, thereby shortening the overall production time of high-value orders. Furthermore, based on all the calculated detection priority weights, a heuristic path planning algorithm is used to generate one or more camera scanning paths. The camera scanning path is used to plan the camera's movement and image acquisition sequence within the effective inspection area. The path planning aims to maximize the cumulative sum of detection priority weights in the area covered by the scanning path in the early stage, while satisfying the device motion constraints. The “heuristic path planning algorithm” can be an improved bidirectional fast search random tree algorithm combined with priority guidance or a genetic algorithm with an elite strategy. The path planning not only considers the shortest total journey, but also strives to cover as many areas with high detection priority weights as possible in the early stages of scanning. This optimization goal enables the device motion control itself to have intelligent decision-making capabilities oriented towards production value, rather than simply geometric path optimization. Finally, the generated camera scanning path, along with the unique identifier of each enhanced task data object and the precise coordinates and dimensions of its corresponding standardized outline graphic, are sent to the AVI visual inspection machine via the device communication protocol. The AVI visual inspection machine is then controlled to perform the following sequential actions: based on the sent camera scanning path and coordinate information, the motion system is driven to precisely position the first standardized outline graphic to be scanned within the camera's field of view; the camera is controlled to acquire images according to the path plan, and based on the process type associated with the unique identifier of each graphic, the corresponding visual inspection program is called to synchronously analyze the corresponding areas of multiple different standardized outline graphics covered in a single acquired image, thereby achieving synchronous visual inspection of multiple PCB units in a single scanning process. "Synchronous analysis" is achieved through multi-threading or parallel computing technology. For a large-format image containing multiple PCB units acquired in a single frame, the system automatically divides multiple independent regions of interest based on the coordinate information of each standardized outline graphic, and simultaneously launches multiple detection instances, each loading the corresponding process detection algorithm to perform defect analysis on its respective region. This allows the concurrent detection of multiple PCB units to be completed within one hardware scan cycle. This concurrent processing mechanism is a key technology for improving the overall detection throughput. During and after the synchronous visual inspection is performed by the AVI appearance inspection machine, the inspection control module also performs the following operations: During the synchronous visual inspection process, the average image contrast and system computational load rate are monitored in real time, and the local illumination intensity or computational resource allocation is dynamically fine-tuned according to the preset thresholds for the average image contrast and system computational load rate, in order to maintain stable detection quality and efficiency. The "preset thresholds" include a lower limit threshold for average image contrast (e.g., set to 100) and an upper limit threshold for system computational load (e.g., set to 80%). When the average image contrast detected in real time is lower than the lower limit threshold, the system automatically increases the illumination brightness of the corresponding area. When the system computational load rate is consistently higher than the upper limit threshold, the frame rate or resolution of the image processing algorithm can be dynamically reduced to prevent system overload from causing detection interruption. This adaptive adjustment mechanism enhances the robustness of the system under different operating conditions. After the inspection is completed, two types of data are collected and integrated: the first type is the defect detection results of the PCB unit corresponding to each enhanced task data object, including the good status, defect type and location coordinates; the second type is the process efficiency data of this panel inspection, including the actual total time spent from the start to the end of this inspection task, the actual scanning order of each enhanced task data object in the scanning path, and the average deviation between the scanning position and the theoretical position monitored online. "Process performance data" and "defect detection results" together constitute a complete data profile of a detection task, providing a comprehensive data foundation for production quality traceability and process optimization; among them, the average deviation value reflects the long-term repeatability accuracy of the equipment positioning system and is a key indicator for assessing the health status of the equipment. The inspection and control module sends the defect inspection results to the manufacturing execution system and packages the process performance data, sending it back to the central optimization engine through the feedback channel. The central optimization engine uses the process performance data to compare the actual total time with the model's predicted time to correct the inspection speed estimation parameters, analyzes the conformity between the actual scanning sequence and the theoretical inspection priority weights to optimize the distance attenuation coefficient used when calculating the inspection priority weights, and uses the average position deviation to assess the adequacy of the preset engineering safety distance value for layout verification. This enables iterative learning and adaptive adjustment of the inspection speed estimation parameters, distance attenuation coefficients, and engineering safety distance values ​​in the multi-objective optimization model. The "iterative learning and adaptive adjustment" process constitutes the core self-optimization closed loop of the system. The central optimization engine can use gradient descent-based optimization methods or Bayesian optimization methods to periodically update the internal model parameters using new feedback process performance data. For example, if the actual detection time is shorter than the model prediction for several consecutive times, the detection speed estimation parameter is adjusted upwards, and vice versa. Through continuous learning, the system's layout optimization and path planning models can increasingly match the dynamic performance of actual equipment and the production environment, realizing an intelligent evolution from static rule configuration to dynamic adaptation, and continuously improving overall production efficiency and quality.

[0024] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0025] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0026] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0027] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0028] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0029] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0030] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection, characterized in that, Specifically, it includes: The data interface module, central optimization engine, typesetting execution module, and detection control module are connected sequentially, among which; Data Interface Module: When the production order queue is updated, it retrieves order data containing order attribute information from the manufacturing execution system, processes the order data, generates and outputs an ordered list of enhanced task data objects to the central optimization engine; Order attribute information includes part number, delivery date, face number, quantity, and process type obtained from the bill of materials; Central optimization engine: Receives the ordered list of enhanced task data objects and obtains the current set of enhanced task data objects; When the AVI visual inspection machine is idle or receives new order data, the central optimization engine calculates the current set of enhanced task data objects based on a predefined multi-objective optimization model. It then selects at least one enhanced task data object from this set that meets preset constraints and determines the two-dimensional position coordinates of each selected enhanced task data object within the effective inspection area of ​​the AVI visual inspection machine, thereby generating a panelization instruction. The panelization instruction at least includes the unique identifier information of the selected enhanced task data object and its corresponding two-dimensional position coordinates. The layout execution module receives the imposition instruction, retrieves the corresponding standardized outline graphics from the enhanced task data object based on the unique identifier information of the enhanced task data object contained in the imposition instruction, positions and arranges all the retrieved standardized outline graphics on a virtual canvas corresponding to the effective inspection area size of the AVI appearance inspection machine according to the two-dimensional position coordinates in the imposition instruction, generates a virtual imposition image, and performs boundary and spacing compliance verification on the virtual imposition image; after the verification is passed, the layout execution module converts the virtual imposition image into an imposition graphic file that can be recognized by the AVI appearance inspection machine; Inspection and control module: Loads the panel layout graphic file and controls the AVI visual inspection machine to perform synchronous visual inspection on multiple PCB units laid out according to the virtual panel layout graphic file.

2. The PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection according to claim 1, characterized in that: The specific process of generating and outputting an ordered list of enhanced task data objects to the central optimization engine in the data interface module is as follows: Each piece of order data is validated to check whether the delivery date is earlier than the current system time and whether the number of rounds belongs to the preset set of allowed rounds. For order data with a delivery date earlier than the current system time or a number of rounds that does not belong to the set of allowed rounds, it is marked as abnormal data and temporarily stored in an abnormal queue independent of the main processing flow. At the same time, an alarm notification is triggered, and the order data that passes the validation continues to be processed. Calculate a dynamic value assessment score for each verified order data; Generate a corresponding enhanced task data object for each valid order data; Based on the dynamic value assessment score, all generated enhanced task data objects are sorted in descending order from high to low to form an ordered list of enhanced task data objects, which is then sent to the central optimization engine.

3. The PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection as described in claim 2, characterized in that: The process of calculating a dynamic value assessment score for each verified order data is as follows: The four weighting coefficients, which are the sum of the delivery urgency weighting coefficient, the batch size weighting coefficient, the process complexity weighting coefficient, and the customer priority weighting coefficient, are multiplied by the delivery urgency factor, the order batch size factor, the process complexity factor, and the static customer priority factor, and then summed to obtain a dynamic value assessment score.

4. The PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection as described in claim 3, characterized in that: The process of generating a corresponding enhanced task data object for each verified order data is as follows: The enhanced task data object includes the part number, delivery date, face number, and quantity from the order data, as well as the process type obtained from the bill of materials, the calculated dynamic value assessment score, and a globally unique task identifier within this system cycle. During the generation of the enhanced task data object, the data interface module calls a preprocessing subsystem of a communication connection through the part number it contains; The preprocessing subsystem performs the following steps: locates and loads the original Gerber design file corresponding to the part number, performs parsing and graphic alignment of the file including multiple standard layers such as circuit layer, solder mask layer, and silkscreen layer, extracts the precise outline representing the physical boundary of the PCB, and generates a standardized outline graphic defined by the width and height of the circumscribed rectangle. The standardized outline graphic is associated with the corresponding enhanced task data object, becoming part of that enhanced task data object.

5. A PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection as described in claim 4, characterized in that: In the central optimization engine, the specific calculation process for the current enhanced task data object set based on the predefined multi-objective optimization model is as follows: The central optimization engine receives an ordered list of enhanced task data objects from the data interface module, which serves as the current set of enhanced task data objects. Retrieve the following attributes for each enhanced task data object in the current enhanced task data object set: width and height of the associated standardized outline graphic, process type, dynamic value assessment score, face number, and unique identifier information; A multi-objective optimization model is constructed with the goal of maximizing the comprehensive utility value of a single assembly. The comprehensive utility value is obtained by multiplying the area utilization rate by the first weight coefficient, adding the weighted value urgency by the second weight coefficient, and then subtracting the process switching cost by the third weight coefficient. The area utilization rate is equal to the sum of the areas of the standardized outline graphics corresponding to all selected enhanced task data objects, divided by the area of ​​the effective inspection area of ​​the AVI appearance inspection machine; the area of ​​the standardized outline graphics is obtained by multiplying its width by its height, and the area of ​​the effective inspection area is obtained by multiplying the preset area length by the area width; The weighted value urgency term is equal to multiplying the dynamic value assessment score of all selected enhanced task data objects by a delivery urgency factor, and then summing all the product results. The process switching cost item is a binary indicator. When the dominant process type associated with the enhanced task data object in the current layout scheme is different from the dominant process type associated with the enhanced task data object in the previous layout scheme recorded by the AVI appearance inspection machine, the indicator value is one; otherwise, the value is zero. The calculation process satisfies the following constraints: all enhanced task data objects selected and placed in the same panel scheme have the same face number; all selected standardized outline graphics, after being placed according to the determined two-dimensional position coordinates, are located within the boundary of the effective inspection area, no two graphics overlap, and the minimum distance between them is not less than a safety distance value preset according to the equipment positioning accuracy and inspection requirements.

6. The PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection as described in claim 5, characterized in that: The specific steps for generating the layout instruction are as follows: The central optimization engine uses a heuristic search algorithm, taking the current enhanced task data object set, the length and width of the effective inspection area of ​​the AVI appearance inspection machine, the safety distance value, and the dominant process type of the previous panelization scheme as input; Based on a multi-objective optimization model, optimization calculations are performed to obtain a feasible solution that maximizes the overall utility value. The heuristic search algorithm is one of the following: genetic algorithm, simulated annealing algorithm, or tabu search algorithm; Once a feasible solution is found that satisfies all constraints and has a satisfactory overall utility value, the central optimization engine generates a panelization instruction based on that feasible solution. This panelization instruction is a list, and each record in the list contains a unique identifier for a selected enhanced task data object, as well as the x-coordinate and y-coordinate values ​​of the two-dimensional position coordinates of the standardized outline graphic corresponding to that enhanced task data object.

7. A PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection as described in claim 6, characterized in that: In the layout execution module, firstly, the layout instruction is parsed to extract the unique identifier information of all enhanced task data objects contained therein; based on this unique identifier information, the width and height of the corresponding standardized outer frame graphic are obtained; combined with the two-dimensional position coordinates specified for each enhanced task data object in the layout instruction, a graphic instance containing its unique identifier information, width, height and position coordinates is constructed for each processed enhanced task data object; and a virtual layout image is generated on the virtual canvas based on all such graphic instances. The specific steps for verifying the compliance of boundaries and spacing in a virtual mosaic image are as follows: Secondly, perform basic geometric constraint checks on the virtual jigsaw puzzle. Specifically, for each graphic instance, use the horizontal coordinate value of its two-dimensional position coordinates as the left boundary position, the horizontal coordinate value plus the width as the right boundary position, the vertical coordinate value as the bottom boundary position, and the vertical coordinate value plus the height as the top boundary position. Check whether the left and bottom boundary positions of all graphic instances are not less than zero, whether the right boundary position of all graphic instances is not greater than the length of the virtual canvas, and whether the top boundary position of all graphic instances is not greater than the width of the virtual canvas. Next, perform the inter-graphic interference and safety distance verification, specifically: For any two different graphic instances, calculate the horizontal projection overlap gap between them, which is equal to zero or the maximum of the following two differences: the larger of the left boundary positions of the corresponding graphics of the two graphic instances, minus the smaller of the right boundary positions of the corresponding graphics of the two graphic instances; calculate the vertical projection overlap gap between them, which is equal to zero or the maximum of the following two differences: the larger of the lower boundary positions of the corresponding graphics of the two graphic instances, minus the smaller of the upper boundary positions of the corresponding graphics of the two graphic instances; based on the horizontal and vertical projection overlap gaps, calculate the effective distance between the corresponding graphics of the two graphic instances; verify whether the effective distance between any two corresponding graphics of the graphic instances is greater than or equal to a preset engineering safety distance value; Finally, the graphic fill completeness and layout density assessment and verification are performed. Specifically, a graphic fill completeness index is calculated. The calculation process of the graphic fill completeness index is as follows: the sum of the first part value and the second part value is calculated. The first part value is calculated by multiplying the width and height of the standardized outer frame of each graphic instance to obtain the individual area, and then adding the individual areas of all graphic instances to obtain the total area of ​​the placed graphic. Then, this total area is divided by the total area of ​​the virtual canvas obtained by multiplying the length and width of the virtual canvas. The second part value is calculated by dividing the density weight factor by the sum of the effective spacing between the number one and all different graphic instance pairs, and verifying whether the graphic fill completeness index is greater than a preset fill completeness threshold.

8. A PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection according to claim 7, characterized in that: The process of converting the virtual mosaic image into a mosaic graphic file recognizable by an AVI visual inspection machine is as follows: After passing all compliance checks, the layout execution module performs a manufacturability risk assessment on the virtual mosaic. This assessment includes: calculating the size and position of the overall bounding rectangle formed by all standardized outline graphics in the current mosaic scheme; the left boundary of the overall bounding rectangle is equal to the minimum value among the left boundary positions of all graphic instances; the right boundary is equal to the maximum value among the right boundary positions of all graphic instances; the lower boundary is equal to the minimum value among the lower boundary positions of all graphic instances; and the upper boundary is equal to the maximum value among the upper boundary positions of all graphic instances. The module also calculates the weight of all standardized outline graphics on the virtual canvas. The centroid distribution coordinates are calculated as follows: the x-coordinate is equal to the sum of the products of the individual areas of all graphic instances and their respective x-coordinates, divided by the total area of ​​the placed graphic; the y-coordinate is equal to the sum of the products of the individual areas of all graphic instances and their respective y-coordinates, divided by the total area of ​​the placed graphic. The proximity of the overall bounding rectangle to the virtual canvas boundary and the deviation of the centroid coordinates from the canvas center are compared with the preset allowable extreme values ​​for proximity and deviation, respectively. If the allowable extreme values ​​are not exceeded, the manufacturability risk is judged to be low, and a risk warning log is generated; if the allowable extreme values ​​are exceeded, a high-risk warning is generated. Subsequently, the layout execution module performs device-specific format conversion, encoding the verified and risk-predicted virtual layout results according to the dedicated graphic file format specified by the target AVI appearance inspection machine. The virtual layout results include the unique identifier information of each enhanced task data object in the imposition instruction, as well as the width, height, and two-dimensional position coordinates of its corresponding standardized outer frame graphic. According to the equipment requirements, the encoding process maps the position, size, and unique identifier of each graphic to an instruction sequence or graphic element description that the device controller can parse, ultimately generating a structured imposition graphic file containing complete layout information.

9. A PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection as described in claim 8, characterized in that: The specific process of loading the panel graphic file and controlling the AVI appearance inspection machine to perform synchronous visual inspection based on the panel graphic file in the detection control module is as follows: First, the mosaic graphic file is parsed to obtain the unique identifier information of all the enhanced task data objects contained therein, as well as the width, height, and two-dimensional position coordinates of the standardized outline graphic corresponding to each enhanced task data object. The center point coordinates of the standardized outline graphic are calculated based on its two-dimensional position coordinates, and the Euclidean distance between each center point coordinate and the preset scanning start point is calculated. Then, this distance is divided by a preset distance reference value and normalized to obtain a dimensionless relative distance value. Secondly, a detection priority weight is calculated for each enhanced task data object. The value of the detection priority weight is equal to the dynamic value assessment score attached to the enhanced task data object, multiplied by a distance decay factor. The distance attenuation factor is based on the natural constant, and its exponent is the negative value obtained by multiplying the relative distance value by a preset distance attenuation coefficient that is greater than zero. Next, based on all the calculated detection priority weights, a heuristic path planning algorithm is used to generate one or more camera scanning paths; Finally, the generated camera scanning path, along with the unique identifier of each enhanced task data object and the precise coordinates and dimensions of its corresponding standardized outline graphic, are sent to the AVI visual inspection machine via the device communication protocol. The AVI visual inspection machine is then controlled to perform the following sequential actions: based on the sent camera scanning path and coordinate information, the motion system is driven to precisely position the first standardized outline graphic to be scanned within the camera's field of view; the camera is controlled to acquire images according to the path plan, and based on the process type associated with the unique identifier of each graphic, the corresponding visual inspection program is called to synchronously analyze the corresponding areas of multiple different standardized outline graphics covered in a single acquired image, thereby achieving synchronous visual inspection of multiple PCB units in a single scanning process.

10. A PCB multi-unit intelligent panelization and collaborative inspection system for AVI inspection according to claim 9, characterized in that: During and after the synchronous visual inspection is performed by the AVI appearance inspection machine, the inspection control module also performs the following operations: During the synchronous visual inspection process, the average image contrast and system computing load rate are monitored in real time, and the local illumination intensity or computing resource allocation is dynamically fine-tuned according to the preset thresholds for the average image contrast and system computing load rate. After the inspection is completed, two types of data are collected and integrated: the first type is the defect detection results of the PCB unit corresponding to each enhanced task data object, including the good status, defect type and location coordinates; the second type is the process efficiency data of this panel inspection, including the actual total time spent from the start to the end of this inspection task, the actual scanning order of each enhanced task data object in the scanning path, and the average deviation between the scanning position and the theoretical position monitored online. The detection and control module sends the defect detection results to the manufacturing execution system, and at the same time packages the process performance data and sends it back to the central optimization engine through the feedback channel. The central optimization engine uses process performance data to compare the actual total time consumption with the model's predicted time consumption to correct the detection speed estimation parameters. It analyzes the conformity between the actual scanning sequence and the theoretical detection priority weight to optimize the distance attenuation coefficient used when calculating the detection priority weight. It also uses the average position deviation to assess the adequacy of the preset engineering safety distance value for layout verification. This enables iterative learning and adaptive adjustment of the detection speed estimation parameters, distance attenuation coefficient, and engineering safety distance value in the multi-objective optimization model.

Images