Artificial intelligence (ai) intelligent detection method and system for optical module manufacturing, terminal device, and medium

Abstract

The application is suitable for the technical field of optical module manufacturing, and provides an AI intelligent detection method and system for optical module manufacturing, a terminal device and a medium. The application not only realizes real-time management and control and defect interception of quality of each process, but also can automatically generate accurate process optimization instructions and apply them to subsequent production by collecting and analyzing cross-process deep detection data, so that the manufacturing system is freed from high dependence on manual offline analysis and experience debugging, adaptive adjustment of process parameters is realized, the optimization response cycle is significantly shortened, real-time adaptation to production fluctuations and prediction avoidance ability of potential quality risks are improved, and finally product yield and production efficiency are improved.

Description

Technical Field

[0001] This application belongs to the field of optical module manufacturing technology, and in particular relates to an AI intelligent detection method, system, terminal equipment and medium for optical module manufacturing. Background Technology

[0002] In automated manufacturing lines for optical modules, visual inspection devices are typically installed at each assembly stage to monitor quality during production. However, these devices often have limited functionality, primarily serving to determine the immediate conformity of products at their respective stages. This results in fragmented and isolated quality information, making adjustments and optimizations to process parameters heavily reliant on engineers' offline statistical analysis of production data, personal experience, and targeted equipment adjustments. This optimization approach has a long response time, struggles to adapt to minor fluctuations in materials, environment, or equipment in real time, and is unable to proactively predict and mitigate potential quality risks, thus hindering continuous improvements in product yield and production efficiency. Summary of the Invention

[0003] In view of this, the embodiments of this application provide an AI intelligent inspection method, system, terminal equipment and medium for optical module manufacturing, which can solve the problems of long response cycles, difficulty in adapting to small fluctuations in materials, environment or equipment in real time, inability to actively predict and avoid potential quality risks, thereby restricting the continuous improvement of product yield and production efficiency.

[0004] In the first aspect, the embodiments of this application provide an AI intelligent detection method for optical module manufacturing. The AI ​​intelligent detection method for optical module manufacturing is applied to the control module of the optical module manufacturing system. The optical module manufacturing system also includes a top cover feeding and gluing equipment, a handle and spring pre-assembly equipment, a protective cotton and PCBA assembly equipment, a bottom shell feeding and gluing equipment, a marking and cover closing equipment, and a locking and finished product unloading equipment arranged sequentially along the production line. Each piece of equipment is correspondingly provided with a detection module for performing visual detection and a sorting mechanism for performing sorting actions. The methods include: The equipment is controlled sequentially to perform the corresponding assembly processes on the optical modules that flow through it; In response to the optical module completing its corresponding process on any device, the detection result generated by the process-specific visual inspection performed on the optical module by the detection module set on that device is obtained, and the sorting mechanism of the device is controlled according to the detection result to sort the optical modules with unqualified detection results to the NG channel. When acquiring the detection results of the detection module on each device, the detection data of that detection module is collected and recorded synchronously; Data analysis is performed on the collected detection data to obtain the data analysis results; Based on the data analysis results, optimization instructions are generated and sent to the corresponding equipment or detection modules for execution in subsequent production cycles.

[0005] Secondly, this application provides an optical module manufacturing system, which includes a control module, a top cover feeding and gluing equipment, a handle and spring pre-assembly equipment, a protective cotton and PCBA assembly equipment, a bottom shell feeding and gluing equipment, an identification and cover closing equipment, and a locking and finished product unloading equipment. Each piece of equipment is correspondingly equipped with a detection module for performing visual inspection and a sorting mechanism for performing sorting actions. The control module is used for: The equipment is controlled sequentially to perform the corresponding assembly processes on the optical modules that flow through it; In response to the optical module completing its corresponding process on any device, the detection result generated by the process-specific visual inspection performed on the optical module by the detection module set on that device is obtained, and the sorting mechanism of the device is controlled according to the detection result to sort the optical modules with unqualified detection results to the NG channel. When acquiring the detection results of the detection module on each device, the detection data of that detection module is collected and recorded synchronously; Data analysis is performed on the collected detection data to obtain the data analysis results; Based on the data analysis results, optimization instructions are generated and sent to the corresponding equipment or detection modules for execution in subsequent production cycles.

[0006] Thirdly, embodiments of this application provide a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the AI ​​intelligent detection method for manufacturing the optical module described above.

[0007] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the AI ​​intelligent detection method for manufacturing the aforementioned optical module.

[0008] Fifthly, embodiments of this application provide a computer program product that, when run on a terminal device, causes the terminal device to execute the aforementioned AI intelligent detection method for optical module manufacturing.

[0009] The beneficial effects of this application embodiment compared with the prior art are as follows: This application embodiment not only realizes real-time quality control and defect interception of each process, but also automatically generates accurate process optimization instructions by collecting and analyzing in-depth detection data across processes and applying them to subsequent production. This enables the manufacturing system to get rid of its high dependence on manual offline analysis and experience-based debugging, realizes adaptive adjustment of process parameters, significantly shortens the optimization response cycle, improves the real-time adaptability to production fluctuations and the ability to predict and avoid potential quality risks, and ultimately improves product yield and production efficiency. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram illustrating the implementation process of the AI-powered intelligent detection method for optical module manufacturing provided in this application embodiment.

[0012] Figure 2 This is a schematic diagram of the optical module manufacturing system provided in the embodiments of this application.

[0013] Figure 3 This is a schematic diagram of the structure of the terminal device provided in the embodiments of this application.

[0014] Attached image labels: 1. Optical module manufacturing system; 11. Top cover feeding and gluing equipment; 12. Pull handle and spring pre-assembly equipment; 13. Protective cotton and PCBA assembly equipment; 14. Bottom shell feeding and gluing equipment; 15. Marking and cover closing equipment; 16. Locking and finished product unloading equipment. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are protected by this application.

[0016] It should be noted that the terms "comprising," "including," and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application, are intended to cover non-exclusive inclusion. For example, a process, method, terminal, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices. Terms such as "first" and "second" in the claims, specification, and accompanying drawings of this application, as well as relational terms, are used merely to distinguish one entity / operation / object from another entity / operation / object, and do not necessarily require or imply any such immediate relationship or order between these entities / operations / objects.

[0017] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0018] In automated manufacturing lines for optical modules, visual inspection devices are typically installed at each assembly stage to monitor quality during production. However, these devices often have limited functionality, primarily serving to determine the immediate conformity of products at their respective stages. This results in fragmented and isolated quality information, making adjustments and optimizations to process parameters heavily reliant on engineers' offline statistical analysis of production data, personal experience, and targeted equipment adjustments. This optimization approach has a long response time, struggles to adapt to minor fluctuations in materials, environment, or equipment in real time, and is unable to proactively predict and mitigate potential quality risks, thus hindering continuous improvements in product yield and production efficiency.

[0019] In view of this, the embodiments of this application provide a control method for an AI intelligent inspection system in optical module manufacturing. This method not only realizes real-time quality control and defect interception of each process, but also automatically generates accurate process optimization instructions by collecting and analyzing in-depth inspection data across processes and applying them to subsequent production. This frees the manufacturing system from its heavy reliance on manual offline analysis and experience-based debugging, enables adaptive adjustment of process parameters, significantly shortens the optimization response cycle, improves the real-time adaptability to production fluctuations and the ability to predict and avoid potential quality risks, and ultimately improves product yield and production efficiency.

[0020] To illustrate the technical solution of this application, specific embodiments are described below.

[0021] Figure 1 This illustration shows a schematic diagram of the control method implementation process of an AI intelligent inspection method system for optical module manufacturing provided in an embodiment of this application. This method can be applied to the control module in the optical module manufacturing system 1. The optical module manufacturing system 1 also includes a top cover feeding and gluing device 11, a handle and spring pre-assembly device 12, a protective cotton and PCBA assembly device 13, a bottom shell feeding and gluing device 14, an identification and cover closing device 15, and a locking and finished product unloading device 16. Each device is correspondingly equipped with an inspection module for performing visual inspection and a sorting mechanism for performing sorting actions.

[0022] The control module is the core processor (such as an industrial PC or PLC) of the optical module manufacturing system 1. It is responsible for receiving sensor signals from various devices, running preset control programs, sending action commands, and executing the training and inference of AI algorithms. It is the central hub of intelligent control.

[0023] An NG channel refers to a physical path or collection device, such as a chute, bin, or conveyor belt, specifically designed on a production line for receiving and transferring products or semi-finished products that have been determined to be non-conforming, and is separate from the main transfer channel for conforming products.

[0024] like Figure 2 As shown, the optical module manufacturing system 1 consists of six functional devices connected in a straight line along the production line. Each device realizes material flow through conveyor belts, jig return tracks, or robotic arms. It is equipped with a three-color warning light on the top and an operation panel and visualization window on the front. The upper shell tray loading + gluing machine (far left) transports the qualified upper shells to the pull handle loading assembly + spring loading assembly machine (right next to it) via conveyor belt. The latter uses a jig conveyor belt to send the pre-assembled body to the protective cotton loading assembly + PCB loading assembly machine, and then transfers the assembly to the lower shell loading + gluing machine via conveyor belt. The lower shell loading and dispensing machine delivers qualified lower shells via conveyor belt to the lower shell flipping and labeling machine and the upper and lower shell closing machine (including the scanning and input of barcodes into the MES). Finally, the upper and lower shell screw fastening and finished product unloading machine (far right end) completes the fastening, inspection and finished product / NG sorting. The fixture is then returned to the front-end equipment, forming a closed loop of loading-assembly-dispensing-closing-fastening-unloading. All equipment achieves full-process automated collaboration through physical connection and data linkage (such as visual inspection and MES barcode scanning).

[0025] Each piece of equipment integrates at least one inspection module, the core of which can be an industrial CCD camera. The camera is fixedly mounted above, to the side, or diagonally above a specific workstation, depending on the inspection requirements, to ensure clear imaging of the area being inspected (such as dispensing paths or assembly interfaces). The camera is equipped with a specialized light source (such as ring light, coaxial light, or strip light) to optimize imaging results. The inspection module's function is to non-contactly capture the visual state of the product after the process is completed and convert it into quantifiable quality information through image processing technology; it is the physical sensor for performing specialized visual inspection of the process.

[0026] Each piece of equipment is also equipped with a sorting mechanism, which can be located to the side or below the main flow path (the path for qualified products to exit). Common forms include pneumatic push rods, swing cylinder baffles, small robots, or reversing belt mechanisms. The sorting mechanism is directly driven by the control module. Its function is to quickly and accurately remove the currently non-conforming product from the main flow path after receiving a non-conforming judgment signal, and guide it to fall into a dedicated NG channel (such as a material box or chute), thereby achieving immediate physical isolation of defective products. It is the actuator that performs the final action of quality control.

[0027] Specifically, the control method of the AI ​​intelligent detection method system for manufacturing the above-mentioned optical module may include the following steps S101 to S105.

[0028] Step S101: The device is controlled to perform corresponding assembly processes on the optical modules that flow through it.

[0029] In the embodiments of this application, the control module can sequentially issue commands to start each piece of equipment according to a preset process sequence (upper shell processing, pull handle spring assembly, protective cotton and PCBA assembly, lower shell processing, marking and closing, and locking and unloading). Specifically, the control module can control the upper shell feeding and gluing equipment 11 to take out the upper shell from the hopper and complete the gluing in the designated area; after completion, the control module controls the conveying mechanism (such as a synchronous belt) to transfer the fixture carrying the upper shell to the pull handle and spring pre-assembly equipment 12, which then grabs the pull handle and spring and completes the pressing; this process is repeated until the product flows through all six pieces of equipment and completes all assembly steps, resulting in the assembled optical module.

[0030] Step S102: In response to the optical module completing its corresponding process in any device, obtain the detection result generated by the process-specific visual inspection performed by the detection module of the device on the optical module, and control the sorting mechanism of the device according to the detection result to sort the optical module with the unqualified detection result to the NG channel.

[0031] In the embodiments of this application, after a piece of equipment (e.g., a pull handle and spring pre-assembly equipment 12) completes its assembly action (spring pressing in), the sensors (e.g., photoelectric sensors) installed on it send a process completion signal back to the control module. The control module can control the detection module integrated on the equipment (e.g., a CCD camera mounted above the upper shell gluing station with a ring light source) to photograph the current product. The vision processing unit (which can be built into the control module or integrated into the camera) runs a pre-stored detection program for this process. For example, it identifies the spring contour through image processing algorithms and calculates its relative position to the pull handle groove edge, ultimately outputting a pass or fail judgment result (i.e., the detection result). The control module can receive this result in real time. If the detection result is fail, it immediately sends an action command to the sorting mechanism of the equipment (e.g., a small cylinder push rod directly controlled by the control module) to push the fail product laterally into the NG box on the side of the equipment, thereby ensuring that it does not flow to the next station. This allows for the immediate interception of defective products at each manufacturing stage, achieving real-time quality control and preventing the accumulation of defects and value loss.

[0032] Step S103: When acquiring the detection results of the detection module on each device, the detection data of that detection module is collected and recorded synchronously.

[0033] In the embodiments of this application, while the detection module performs visual analysis and generates detection results, the control module can synchronously acquire and record the raw or intermediate data generated during the detection process through a data interface (such as GigE, USB, or IO). This detection data includes, but is not limited to: raw high-definition images captured by the detection module, and key quantitative feature values ​​extracted during image processing (such as the calculated actual width pixel value of the colloid, the coordinate offset of the spring center, and the measured height difference between the screw head and the housing). Simultaneously, the production line MES system or a barcode scanner can assign a unique identifier (such as a QR code) to each moving fixture or product. The control module can timestamp all the data obtained in this detection and bind it to the unique identifier carried by the currently inspected optical module, forming a structured data record, which is then stored in a real-time database or data file. This captures in-depth process status information across the entire process and builds a traceable digital quality profile for each product for subsequent correlation analysis.

[0034] Step S104: Perform data analysis based on the collected detection data to obtain data analysis results.

[0035] In the embodiments of this application, the control module can invoke a built-in or connected AI data analysis engine to process massive amounts of inspection data (already associated with product IDs) retrieved from a database and accumulated within a preset time period (such as a production shift). Data analysis can include various specific implementation modes. One basic mode is that the control module, for a single piece of equipment, counts the number of recently produced defective products, and generates an analysis result of "abnormal process status of equipment X" when the count exceeds a threshold. Another more advanced mode is that the control module associates inspection data of the same product across different processes, uses data mining techniques (such as statistical analysis and machine learning models) to explore hidden correlations or patterns between cross-process parameters, and ultimately outputs rules or insights such as "feature 1 of process A is strongly correlated with the defect rate of process C" as the data analysis result. This embodiment can transform raw, discrete inspection data into process knowledge and insights with decision support value.

[0036] It should be noted that the "data mining analysis" performed in this step is a typical application of unsupervised learning and association rule mining in the field of artificial intelligence. Specifically, the control module can automatically identify statistically significant cross-process feature association pairs by performing statistical correlation calculations and cluster analysis on the multi-dimensional feature data generated by massive optical module products at each process stage throughout the entire process, and generate interpretable association rules. This process does not require manual pre-setting of association assumptions; it is entirely driven by the AI ​​algorithm to autonomously learn process patterns from the data.

[0037] Step S105: Based on the data analysis results, generate optimization instructions and send the optimization instructions to the corresponding equipment or detection module for execution in subsequent production cycles.

[0038] In the embodiments of this application, the control module can parse the data analysis results. If the analysis results indicate that "the variance of the top cover adhesive width is too large" or indicate that "the adhesive width of the lower shell is related to the final gap defect," the control module can automatically generate a clearly executable optimization instruction based on the results according to preset rules or optimization algorithms. The instruction is specific; for example, it could be "increase the adhesive dispensing volume setting of the No. 1 dispensing valve in the top cover feeding and adhesive coating equipment 11 by 5%" or "increase the lower limit threshold of the torque judgment of the No. 2 electric screwdriver in the fastening equipment by 0.1 N·m." After generating the optimization instruction, the control module can accurately send the instruction to the local controller of the target device or detection module via a network (such as EtherCAT, PROFINET) or IO signal. The recipient applies this new parameter at the start of the next production cycle (when the next product flows in). This allows the manufacturing system to proactively adjust itself based on data insights to achieve process optimization.

[0039] In one specific implementation, the control module can parse a selected association rule from the data mining analysis results, such as "the 'adhesive width variance' feature of the top cover feeding and gluing equipment 11 is strongly positively correlated with the 'screw floating defect rate' feature of the locking and finished product unloading equipment 16, with a correlation coefficient of 0.85," thereby determining the causal process feature, the result process feature, and the direction and strength of the association. Then, the control module can obtain the latest measurement sequence of adhesive width variance in the current production flow in real time, assuming that its average value has increased from the historical steady-state value of 0.15 mm² to 0.25 mm², exceeding the upper limit of the process specification. Simultaneously, by querying the equipment parameter database, it determines that the adjustable process parameter corresponding to this causal feature is the dispensing pressure. Then, the control module can call its built-in AI decision optimization engine, which belongs to the category of knowledge-based decision optimization in artificial intelligence. Its workflow includes: first, performing a state assessment, comparing the current real-time mean of "colloid width variance" with the target tolerance limit, and calculating the deviation; then, applying knowledge, inferring the optimization direction as reducing "colloid width variance" based on the parsed association rules; finally, combining the deviation, association strength (0.85), and historical feedback data (such as the empirical sensitivity coefficient in historical adjustment records showing that increasing "dispensing pressure" by 0.02 MPa can reduce "colloid width variance" by 0.01 mm²), and through a decision algorithm that integrates proportional control and learning factors, calculating the suggested adjustment amount as increasing the dispensing pressure by 0.08 MPa based on the current set value. Finally, the control module generates a structured optimization instruction based on this adjustment amount. Its core fields include instruction type, target device, target parameters, adjusted value, rule ID, and expected impact target, and sends it to the local controller of the top cover feeding and coating equipment 11 via the network, automatically applying the new parameters at the start of the next production cycle. This approach enables intelligent closed-loop control across the entire process, from data mining to discover cross-process correlations, to automatic decision-making based on AI knowledge models, and finally to precise execution parameter optimization.

[0040] The beneficial effects of this application embodiment compared with the prior art are as follows: This application embodiment not only realizes real-time quality control and defect interception of each process, but also automatically generates accurate process optimization instructions by collecting and analyzing in-depth detection data across processes and applying them to subsequent production. This enables the manufacturing system to get rid of its high dependence on manual offline analysis and experience-based debugging, realizes adaptive adjustment of process parameters, significantly shortens the optimization response cycle, improves the real-time adaptability to production fluctuations and the ability to predict and avoid potential quality risks, and ultimately improves product yield and production efficiency.

[0041] In some specific embodiments of this application, the step of performing data analysis based on the collected detection data to obtain data analysis results may specifically include steps S401 and S402.

[0042] Step S401: For any target device, count the number of non-compliant results detected by the detection module of the target device within a preset time period.

[0043] The target is any specific assembly equipment selected for monitoring and analysis in the optical module manufacturing system 1, such as the cover feeding and gluing equipment 11, or the locking and finished product unloading equipment 16.

[0044] A preset time period refers to a continuous time window set in advance for counting the number of non-conforming results, such as the most recent hour, the current production shift (e.g., 8 hours), or the 1000 most recent product cycles. This time period can be configured according to process stability and monitoring granularity requirements.

[0045] In the embodiments of this application, the control module can read all detection result records reported by the detection module of the selected target equipment within a set time window from its real-time database or production log. For example, if the target equipment is the locking and finishing unloading equipment 16, and the preset time period is the most recent two hours, the control module will filter out all detection results generated by this equipment in the past two hours. The system iterates through these records and accumulates the count of records marked as unqualified in the result field. This operation is performed independently and periodically for each piece of equipment. This transforms discrete unqualified events into a quantitative indicator reflecting the recent quality output stability of the equipment, providing a data basis for status judgment.

[0046] Step S402: When the quantity exceeds the first preset threshold, a data analysis result indicating an abnormal process state of the target equipment is generated.

[0047] The first preset threshold is a pre-set quantitative critical value for determining whether the process status of the target equipment is abnormal. This threshold can be determined based on historical yield data, process capability index, etc., such as the upper limit of the number of defective products allowed per hour.

[0048] In the embodiments of this application, the control module can compare the actual number of defective items with a first preset threshold set for the target equipment. For example, suppose the threshold set for the dispensing equipment is no more than 5 defective items per hour, and the number of defective items in the most recent hour is 8. Since 8 is greater than 5, exceeding the threshold, the judgment logic within the control module is triggered. The control module can generate a structured data analysis result, which at least includes the identifier of the target equipment determined to be abnormal, such as generating a message or status flag indicating an abnormal process status of the top cover feeding and dispensing equipment 11. This automatically identifies equipment that may experience performance drift or malfunction, transforming potential quality risks into a clear abnormal signal.

[0049] This application's implementation method, by setting independent monitoring windows and non-conforming product quantity thresholds for each device and automatically performing statistics and judgments, can promptly detect yield decline trends in individual devices caused by parameter drift, component wear, or minor faults, issuing accurate early warnings before the problem escalates into a batch quality incident. This significantly reduces reliance on manual periodic inspections and post-event data analysis, enabling the production line to possess basic self-sensing and self-diagnostic capabilities. This provides a clear basis for subsequent targeted parameter adjustments or preventative maintenance, improving the stability and controllability of the production process.

[0050] In some specific embodiments of this application, the step of generating optimization instructions based on the data analysis results may specifically include steps S501 to S503.

[0051] Step S501: Determine at least one specific defect type associated with the abnormal process state and the process parameters in the target equipment that cause the specific defect type.

[0052] Among them, specific defect types are quality problems specifically manifested by abnormal process conditions. For example, when the target equipment is the fastening and finished product unloading equipment 16, its specific defect types may include screw floating, screw stripping, or screw missing. Each type corresponds to different physical causes and process parameters.

[0053] The process parameters to be adjusted are adjustable variables on the target equipment whose values ​​can directly and significantly affect the occurrence rate of the specific defect type. For example, for a screw floating defect, the corresponding process parameter to be adjusted could be the target torque value of the fastening electric screwdriver; for a dispensing bubble defect, it could be the pressure or vacuum back suction value of the dispensing valve.

[0054] In the embodiments of this application, when the control module receives data analysis results indicating an abnormal process state of a target device, it can query the defect knowledge base in its internal or associated database. This knowledge base pre-stores mapping relationships between device types, common defect types, and key adjustable process parameters leading to each type of defect. The control module can use the target device identifier as an index to retrieve the specific defect type codes corresponding to all detected non-conforming results on that device within the current time period. Then, it performs frequency statistics on these defect types, identifying the one or more defect types with the highest frequency as the specific defect types associated with the current abnormal process state. It then queries the knowledge base again to obtain a list of adjustable process parameters with a strong causal relationship to these specific defect types, and uses these as the process parameters to be adjusted. For example, if the control module determines that the fastening equipment is abnormal and statistics show that screw floating defects account for 70%, it can determine the torque setting value of the fastening electric screwdriver as the process parameter to be adjusted based on the knowledge base mapping.

[0055] Step S502: Calculate the adjustment amount based on the historical set values ​​of the process parameters to be adjusted, the degree of anomaly in the data analysis results, and the feedback data on the effects of adjusting the same parameters in the past.

[0056] The effect feedback data refers to the changes in the number of nonconforming results related to the same specific defect type that were previously calculated after an optimization command (i.e., adjustment) was executed for the same process parameter to be adjusted in the past, within a subsequent preset time period. This data is used to evaluate the effectiveness of historical adjustment actions, determining whether it is positive feedback (defect reduction), negative feedback (defect increase), or ineffective.

[0057] The adjustment amount is the specific numerical change made to modify the set value of the process parameter to be adjusted. It can include the direction of adjustment (increase or decrease) and the magnitude of adjustment. For example, increasing the locking torque set value by 0.2 Newton-meters.

[0058] In the embodiments of this application, the control module can obtain the current setpoint of the process parameter to be adjusted, i.e., the historical setpoint, from the equipment parameter record. Simultaneously, it parses the quantified anomaly level value from the data analysis results. Then, it can retrieve all recent records of adjustments made to the same target equipment and the same process parameter from the historical optimization log. Each record includes the past adjustment amount, the anomaly level before adjustment, and the new defect rate change obtained after adjustment. The control module can use this feedback data to evaluate the effectiveness of historical adjustments, for example, by calculating an empirical approximate sensitivity coefficient between parameter adjustment and defect rate change. Subsequently, the control module can call a preset adaptive optimization algorithm function, such as an algorithm based on PID control principles and incorporating a learning factor. This algorithm aims to reduce the anomaly level, taking the current parameter value, the current anomaly level, and the historical adjustment effect (as a learning factor to correct the PID coefficient) as input. Under the constraint of ensuring process safety and stability, it calculates a theoretically helpful parameter recommendation value that helps the defect rate return to the normal range. The difference between this recommendation value and the historical setpoint is the calculated adjustment amount.

[0059] Step S503: Generate the optimization instruction based on the adjustment amount.

[0060] In the embodiments of this application, the control module can encapsulate the adjustment amount, along with the necessary execution information, into a complete optimization instruction that can be parsed and executed by the downstream equipment controller, according to a predefined instruction format. This optimization instruction may include the following core data fields: instruction type identifier (such as process parameter adjustment), target equipment number, name of the specific process parameter to be adjusted, the new set value after adjustment (derived from the historical set value plus the adjustment amount), and an optional planned effective time (starting from the next production cycle).

[0061] This application's implementation establishes a precise mapping from equipment anomalies to specific defect types and then to key process parameters. By comprehensively utilizing real-time anomaly severity and historical adjustment effect feedback data, and calculating the adjustment amount through an adaptive algorithm, it surpasses simple parameter fine-tuning that relies on fixed rules or human experience. This allows the system to dynamically optimize its decision-making strategy based on the actual effect of each adjustment, thereby more quickly and accurately bringing drifting process parameters back to a stable range. This effectively suppresses quality fluctuations caused by the slow degradation of equipment performance and significantly improves the adaptability, stability, and long-term yield of the production system.

[0062] In some other specific embodiments of this application, data analysis is performed based on the collected detection data to obtain data analysis results, which may specifically include steps S601 and S602.

[0063] Step S601: The detection data is associated with the unique identifier of the corresponding optical module and stored to form a cross-process detection data set corresponding to the unique identifier.

[0064] A unique identifier is a unique identification code assigned to each optical module product or the fixture that carries it during the production process. Common forms include serial numbers stored in one-dimensional barcodes, two-dimensional barcodes, or RFID tags. This identifier is generated and attached to the product at the beginning of the production line (such as the initial material loading or printing and labeling station) and is tracked and identified by reading and writing devices (such as barcode scanners) throughout the entire process, serving as a key index for data association.

[0065] A cross-process inspection data set refers to a data structure organized and stored in a database using the aforementioned unique identifier as the primary key. This set aggregates all inspection data synchronously collected and recorded by the inspection modules of each device as the same optical module product flows through different devices. Each sub-data record within the set includes a timestamp, process device identifier, and specific data content for that inspection (such as image feature vectors and measurement values).

[0066] In the embodiments of this application, when an optical module product flows through any equipment on the production line and completes inspection, the control module can not only acquire the inspection results and data, but also obtain the unique identifier on the current product in real time through the identification reading device of that workstation (such as a fixed barcode scanner). Subsequently, the control module can construct a data record containing a timestamp, equipment workstation code, unique identifier, and complete inspection data (such as the storage path of the original image file, extracted feature vectors, and intermediate algorithm results). This record is sent via industrial Ethernet and written to the central database server. The control module can logically aggregate all data records with the same unique identifier but from different process equipment into a single data object, namely, the cross-process inspection data set of the product. This ensures that all quality data of the same product is accurately collected from the first to the last process.

[0067] Step S602: Based on the cross-process detection data sets of each of the multiple optical modules, perform data mining analysis to obtain data mining analysis results.

[0068] In the embodiments of this application, the control module can retrieve a complete cross-process inspection data set of a sufficient number of optical module products produced within a specified production cycle (such as the past week) from the central database. The control module can first preprocess the data, including cleaning outliers and performing feature alignment and standardization on data from different devices. Then, it uses selected data mining algorithms to analyze this structured dataset. For example, it can calculate the Pearson correlation coefficient matrix between different process characteristic variables to discover potential linear associations; or use clustering algorithms to group products to distinguish different quality performance patterns; or train models such as decision trees to identify key process characteristic combinations that lead to final non-conformity. After the analysis is completed, the control module can organize and package the algorithm outputs, such as identified strongly correlated pairs, representative cluster center features, or decision rule sets, into structured data mining analysis results. Through the correlation and holistic analysis of massive amounts of product process data, it proactively discovers complex process interaction patterns, early signs of quality problems, or systemic influencing factors that are difficult to observe with the human eye, thereby providing in-depth insights for process optimization.

[0069] This application's implementation method systematically associates and integrates full-process inspection data using unique identifiers as clues, and performs data mining analysis on this basis, successfully constructing a quality data map and intelligent analysis capabilities covering the entire manufacturing process. This enables the manufacturing system to automatically mine implicit process relationships, defect transmission chains, and key influencing factors across processes from massive, multi-dimensional production data, thereby elevating the perspective of quality control and process optimization from isolated, post-event, single-point processing to a system-wide, pre-event or in-event, global insight.

[0070] In some specific embodiments of this application, the step of performing data mining analysis based on the cross-process detection data sets of each of the multiple optical modules to obtain data mining analysis results may specifically include steps S701 to S704.

[0071] Step S701: Perform time-series alignment and feature extraction on the cross-process detection data sets of the multiple optical modules to obtain the multi-dimensional feature sequence corresponding to each optical module.

[0072] Among them, the multidimensional feature sequence is a structured data representation. The sequence is arranged in the order of the process steps, where each step position corresponds to a feature vector, which consists of multiple quantifiable feature values ​​extracted from the detection data of that step. Therefore, the entire sequence can be mathematically represented as a matrix, with rows corresponding to the process steps and columns corresponding to the feature dimensions, fully depicting the quantitative process state trajectory of the product throughout the entire process.

[0073] In the embodiments of this application, the control module can obtain a set of cross-process inspection data for a batch of optical module samples from the database. First, time alignment is performed, ensuring that data from different processes within the same optical module are correctly arranged according to their actual production sequence based on the timestamp of each data record. Next, feature extraction is performed, running a feature extraction algorithm specific to the inspection content of each process for the inspection data. For example, for image data from the dispensing process, the average width, standard deviation, and area of ​​the adhesive are extracted using an image processing algorithm; for assembly alignment images, the center coordinate offset of key components is extracted. Multiple feature values ​​extracted from each process constitute a feature vector. The feature vectors obtained from all processes for the same optical module are then assembled strictly according to the process sequence, forming a multi-dimensional feature sequence for that optical module. This transforms the original heterogeneous inspection data into a unified numerical feature sequence suitable for mathematical and statistical analysis.

[0074] Step S702: Calculate the statistical correlation coefficient between any two feature dimensions belonging to different processes in the multidimensional feature sequence to obtain the correlation coefficient matrix. The correlation coefficient matrix is ​​used to characterize the linear correlation strength between the features of each process.

[0075] In this context, a feature dimension is a single quantifiable variable extracted from the inspection data to characterize a specific process quality attribute. For example, multiple feature dimensions can be extracted from the adhesive dispensing inspection of a top cover, such as the average adhesive width, adhesive width variance, and adhesive path continuity score. Each feature dimension is a basic data column that constitutes a multidimensional feature sequence.

[0076] The statistical correlation coefficient is a statistical indicator used to quantify the degree and direction of the linear association between two feature dimensions, such as the Pearson correlation coefficient. Its value ranges from -1 to 1, with positive values ​​indicating positive correlation and negative values ​​indicating negative correlation. The closer the absolute value is to 1, the stronger the linear association, and 0 indicates no linear association.

[0077] The correlation coefficient matrix is ​​a symmetric square matrix where each row and column represents all feature dimensions (covering all processes). The element in the i-th row and j-th column is the statistical correlation coefficient between the i-th and j-th feature dimensions. This matrix encapsulates the linear correlation information of all feature pairs between processes from a global perspective.

[0078] In the embodiments of this application, the control module can collect the multidimensional feature sequences of all sample optical modules and organize them into a large sample feature matrix. Each row of this matrix corresponds to an optical module sample, and each column corresponds to a feature dimension (covering all processes). The control module can then iterate through all possible combinations of feature dimension columns belonging to different processes. For each pair of feature dimension columns, the Pearson correlation coefficient is calculated over the entire sample set. This coefficient quantifies the degree of linear correlation between the changing trends of the two features among the samples. The control module can then fill all the calculated correlation coefficients into the corresponding positions of a symmetric matrix according to the order of the feature dimensions, thereby generating a complete correlation coefficient matrix, which is a global, quantified correlation map. This allows for a quantitative assessment of the overall linear correlation strength between any two process quality features from different processes during production.

[0079] Step S703: Based on the correlation coefficient matrix, identify feature dimension pairs whose correlation strength exceeds a second preset threshold.

[0080] The second preset threshold is a pre-set threshold value for the absolute value of the correlation coefficient, for example, 0.7, to screen out strongly correlated feature pairs with significant engineering implications. This threshold is used to distinguish between weak correlations caused by random fluctuations and strong correlations that may reflect potential process coupling.

[0081] A feature dimension pair is a combination of any two feature dimensions identified from the correlation coefficient matrix whose absolute correlation coefficient exceeds a second preset threshold. A feature dimension pair reveals a possible strong correlation between two specific process parameters or quality indicators.

[0082] In the embodiments of this application, the control module can read a pre-configured second preset threshold, which is a decimal between 0 and 1. The control module traverses the upper or lower triangular portion (excluding the diagonal) of the correlation coefficient matrix, checking the absolute value of the correlation coefficient corresponding to each off-diagonal element, i.e., each pair of different feature dimensions. The absolute value of each correlation coefficient is compared with the second preset threshold. Any two feature dimensions corresponding to correlation coefficients whose absolute values ​​are greater than the threshold are marked and recorded by the system as a strongly correlated feature dimension pair. The control module can store all identified feature dimension pairs, along with their corresponding correlation coefficient values ​​and the process information to which these two feature dimensions belong, as an intermediate result list. From a massive number of possible associations, statistically significant feature pairs with correlation strength reaching the level of engineering concern are filtered out, thereby focusing on the potential relationships most likely to reflect substantial process interactions or defect propagation paths, providing targets for subsequent rule-based interpretation.

[0083] Step S704: Based on the identified feature dimension pairs and their corresponding process information, association direction and strength, generate a set of association rules as the data mining analysis result.

[0084] The direction of the association is indicated by the sign of the correlation coefficient, indicating whether the trends of change in the two feature dimensions are in the same direction (positive correlation) or in opposite directions (negative correlation). The strength of the association is indicated by the magnitude of the absolute value of the correlation coefficient.

[0085] A set of association rules is a collection of readable and interpretable statements of process knowledge. Each association rule typically includes causal feature dimensions, result feature dimensions, association direction, association strength, and information about the process to which it belongs.

[0086] In the embodiments of this application, the control module can read a list of strongly correlated feature dimension pairs. For each pair in the list, the correlation direction is determined based on the sign of its correlation coefficient, and the correlation strength is determined based on its absolute value. Subsequently, the control module can organize the information such as the two features involved in the feature dimension pair, their respective processes, correlation direction, and strength into an easy-to-understand statement according to a predefined rule template. For example, if feature A comes from process 2, feature B comes from process 5, and the correlation coefficient is -0.85, a rule can be generated stating that feature A from process 2 and feature B from process 5 are strongly negatively correlated. By converting all identified feature dimension pairs into rule statements in this way, a structured set of association rules is formed. This set is the final data mining analysis result, which clearly and directly reveals the strong correlation patterns of process feature relationships across processes in historical data.

[0087] This application's implementation serializes and standardizes the process characteristics throughout the entire process, and systematically calculates and filters strong statistical correlations between cross-process characteristics, ultimately generating a clear set of association rules. This can reveal implicit quality correlation networks existing between different manufacturing stages that are difficult to detect through human observation and single-process analysis, thereby elevating the tracing of quality problems from empirical guesswork to the level of data verification.

[0088] In some specific embodiments of this application, the step of generating optimization instructions based on the data analysis results may specifically include steps S801 to S803.

[0089] Step S801: parse any association rule in the association rule set, and determine the causal process feature, the result process feature, and the direction and strength of the association between the two in the association rule.

[0090] The causal process feature is a process feature dimension that is identified as having an influence in association rules, and whose change may lead to a change in another feature. It is usually located at the causal end of the association.

[0091] Result process features: These refer to the process feature dimensions that are identified in the association rules as being affected by changes in the cause process features, and are usually located at the result end of the association.

[0092] In the embodiments of this application, the control module can select a rule to be applied from the generated set of association rules. This rule typically contains multiple fields in its data structure, such as rule ID, causal feature (including process code and feature name), result feature (including process code and feature name), and correlation coefficient (including direction and strength). The specific parsing process involves reading and extracting the values ​​of these fields. Based on the causal feature field, the control module determines which specific feature of which process needs to be monitored and intervened on. Based on the result feature field, it clarifies which indicator of which process the optimization aims to ultimately affect. Finally, based on the correlation coefficient field, it obtains the quantitative relationship (positive / negative correlation and correlation strength) between the two. For example, after parsing, the causal process feature is the glue width variance of the top cover gluing process, the result process feature is the screw floating defect rate of the fastening process, the correlation direction is negative, and the strength is 0.82.

[0093] Step S802: Obtain the real-time measurement value of the cause process characteristic in the current production flow, and calculate the adjustment suggestion amount for the equipment process parameters corresponding to the cause process in combination with the correlation direction and intensity.

[0094] The suggested adjustment amount refers to the recommended adjustment value for the equipment process parameters corresponding to the characteristics of the cause process. It indicates the magnitude and direction of the parameter adjustment, such as increasing the dispensing pressure setting value by 0.05 MPa.

[0095] In the embodiments of this application, the control module can monitor or retrieve the latest measurement sequence of the cause process characteristic in the current production flow in real time, based on the cause process characteristic. For example, if the cause is the variance of the glue width of the top cover, the actual measured values ​​of the glue width variance of the most recent products are obtained. Simultaneously, the control module can obtain the process parameters that can be directly adjusted on the equipment corresponding to the cause characteristic, such as the dispensing pressure. Then, the embedded optimization decision function is invoked. This function can take the current measured value of the cause characteristic, its historical statistical center value or target value, and the strength and direction of the association rule as the main inputs. The specific algorithm logic can be: determine whether the current measured value deviates from the desired state; if it does, calculate a parameter adjustment amount based on the association strength (influence factor size) and association direction (determining the adjustment direction), aiming to bring the cause characteristic back to the desired state, thereby indirectly driving the result characteristic to develop in the desired direction. For example, if the current glue width variance is large and negatively correlated with the screw float height, the algorithm will calculate an adjustment amount for the dispensing pressure to reduce the glue width variance. This combines static, correlated knowledge with dynamic, real-time production status, and uses a quantitative model to calculate the amount of equipment parameter adjustments required to achieve specific optimization goals, thus enabling intelligent decision-making that combines feedforward and feedback.

[0096] Step S803: Generate the optimization instruction based on the suggested adjustment amount, the associated rule identifier, and the target result process characteristics.

[0097] In the embodiments of this application, the control module can encapsulate the adjustment suggestion, along with the unique identifier of the association rule on which the current decision is based, and the target result process characteristic information, into a complete optimization instruction that can be reliably identified and executed by the downstream equipment controller, according to a predefined instruction data structure and communication protocol. This instruction may include fields such as an instruction header, target equipment address code, specific process parameter code to be adjusted, new parameter setting value, association rule ID, and timestamp. For example, the generated instruction in binary or text format may explicitly include: "Apply rule R001, to optimize the floating rate of the locking screws, update the pressure setting value of valve number one of the top cover dispensing machine to a certain value." This transforms the calculated adjustment suggestion containing intelligent decision-making logic into a standardized machine instruction with complete contextual information that is traceable and executable, ensuring that the optimization intent can be accurately and reliably transmitted and applied to the physical production equipment.

[0098] The implementation method of this application combines the data mining results characterizing the process coupling relationship with the real-time status perception of the production site, and generates specific equipment parameter adjustment instructions through a quantitative decision model. Based on the discovered cross-process quality correlation patterns, it can automatically and timely adjust the parameters of upstream key processes when potential adverse trends appear in the current production, thereby intervening before quality problems become explicit. This significantly enhances the production system's ability to resist complex quality risks and its adaptive adjustment capability to process fluctuations.

[0099] In some specific embodiments of this application, the process is specialized in visual inspection, including inspecting at least one of the following: dispensing quality, assembly alignment, insertion depth, marking position, cap alignment, and screw fastening status. Specifically, it may include steps S901 to S907.

[0100] Step S901: Inspect the dispensing quality.

[0101] In the embodiments of this application, after the dispensing equipment (such as the top cover feeding and coating equipment 11) completes the coating, the control module controls the dedicated CCD camera at that station to take pictures of the dispensing area with a light source of a specific wavelength (such as red ring light to enhance the contrast of the adhesive) to obtain a high-definition image. Then, the control module can run a preset algorithm to accurately extract the contour of the adhesive area through image segmentation technology, calculate the geometric features of the contour, including the continuity of the adhesive path centerline, the average and variance of the adhesive width, and the area of ​​the adhesive area, and compare these measured values ​​with the preset tolerance range; if all feature values ​​are within the qualified range and no adhesive overflow into non-designated areas is detected, the dispensing quality is determined to be qualified; otherwise, the specific unqualified type is marked (such as adhesive breakage, insufficient adhesive width, adhesive overflow), thereby ensuring that the coating process of the thermally conductive or sealant meets the reliability design requirements and avoiding the decrease in heat dissipation efficiency or sealing failure due to poor dispensing.

[0102] Step S902: Perform alignment detection during assembly.

[0103] In the embodiments of this application, after the pull handle and spring pre-assembly equipment 12 has completed press-fitting, the control module controls the high-speed CCD camera at this station to take pictures of the assembly from a preset angle (such as directly above or to the side); and performs feature matching and edge detection on the acquired images to identify the reference positioning features of the upper shell (such as positioning posts or edges) and the key feature points of the pull handle and spring assembly (such as the center of the pull handle shaft and the position of the spring end face); calculates the relative coordinate offset between these feature points and compares it with the position of the standard template, thereby quantifying the alignment error between the pull handle and the upper shell and the posture of the spring in the groove; if the calculated offset is within the allowable tolerance range, the assembly alignment is deemed qualified. This ensures the assembly accuracy of the moving parts, ensures smooth opening and closing of the pull handle and reliable spring action, and prevents functional defects or premature wear caused by misalignment.

[0104] Step S903: Detect the insertion depth.

[0105] In the embodiments of this application, after the protective cotton and PCBA assembly equipment 13 complete the PCBA board insertion operation, the control module controls the high-resolution CCD camera (possibly equipped with coaxial light) at this station to take pictures from a direction perpendicular to the PCBA board, acquiring clear images including the edge of the PCBA board and the interface area of ​​the upper shell slot; by analyzing the image grayscale gradient or using template matching, the gap between the top surface of the PCBA board and the upper shell bearing surface is accurately located, or it is detected whether the gold finger area of ​​the PCBA board is visually invisible (fully inserted); by measuring the gap width or calculating the matching degree score and comparing it with the preset qualified standard (such as a gap less than 0.1mm), the insertion depth of the PCBA board is judged to be in place. This verifies the assembly quality of the precision electrical connection components and prevents poor contact, signal transmission failure, or PCBA board damage caused by improper insertion.

[0106] Step S904: Detect the location of the marker.

[0107] In the embodiments of this application, after the label printing and pasting (or laser marking) of the labeling and capping device 15 is completed, the control module controls the CCD camera at that station to capture an image of the labeling area on the lower shell; it locates the outline of the lower shell or a specific reference mark in the image, and then identifies and extracts the outline or character area of ​​the label (such as a QR code or barcode); by calculating the deviation between the actual center coordinates and rotation angle of the labeling area and its theoretical position relative to the lower shell reference, the positional error and angular error of the label are obtained; the label content is read to verify its recognizability. If the positional error and angular error are within the allowable range and the label can be read correctly, the label position is deemed qualified. This ensures the accuracy and readability of product traceability information, meets the requirements of quality management and product traceability, and ensures that the product appearance conforms to specifications.

[0108] Step S905, performing detection for cover closing alignment.

[0109] In the embodiment of the present application, after the marking and cover closing apparatus 15 completes the press-fitting of the upper and lower shells, the control module controls one or more CCD cameras (such as a top camera and a side camera) at this station to shoot the product after cover closing from multiple angles, so as to obtain a complete shell seam image; an edge detection algorithm is adopted to extract continuous edge lines along the seam of the upper and lower shells, and measure the width value of the seam at a series of predefined detection points; the maximum value, minimum value, range (uniformity) and average value of these width measurement values are calculated, and compared with the gap width range and uniformity requirements specified in the process specification. If the gap width of all measurement points is within the specified range and the uniformity is good (for example, the range is less than the threshold), it is determined that the cover closing alignment is qualified. Thereby, the structural integrity, sealing performance and appearance quality of the final assembled product are ensured, and air tightness failure and stress concentration caused by poor cover closing are prevented.

[0110] Step S906, performing detection for the screw locking state.

[0111] In the embodiment of the present application, after the locking and finished product unloading apparatus 16 completes screw locking, the control module controls the CCD camera at this station to cooperate with a low-angle illumination light source (to highlight the height difference between the screw head and the shell surface) to shoot the locking surface of the product; through a pattern recognition or feature matching algorithm, the theoretical installation position of each screw is located in the image, and it is detected whether a screw head exists at this position; for the existing screws, the image of the screw head area is further analyzed, and it is judged whether there is a floating height (not locked) by measuring the gray difference between the top surface of the screw head and the surrounding shell surface or calculating the height feature of its contour; the screw groove shape is checked for completeness to judge whether there is thread slipping; if all screw installation positions are correctly identified and there are no defects such as floating height and thread slipping, it is determined that the screw locking state is qualified. Thereby, the reliability of the mechanical connection of the product is ensured, and loose product structure, poor electrical contact or potential safety hazards caused by screw assembly problems are prevented.

[0112] Figure 2 shows a schematic structural diagram of an optical module manufacturing system 1 provided by an embodiment of the present application. The above-mentioned optical module manufacturing system 1 includes a control module, an upper cover feeding and glue dispensing apparatus 11, a pull tab and spring pre-assembly apparatus 12, a protective cotton and PCBA assembly apparatus 13, a lower shell feeding and glue dispensing apparatus 14, a marking and cover closing apparatus 15, and a locking and finished product unloading apparatus 16. Each apparatus is correspondingly provided with a detection module for performing visual detection and a sorting mechanism for performing sorting actions. The control module is configured to; sequentially control the apparatuses to perform corresponding assembly processes on the flowing optical modules; In response to the optical module completing its corresponding process in any device, the detection result generated by the process-specific visual inspection performed on the optical module by the detection module set in that device is obtained, and the sorting mechanism of the device is controlled according to the detection result to sort the optical module with the unqualified detection result to the NG channel. When acquiring the detection results of the detection module on each device, the detection data of that detection module is collected and recorded synchronously; Data analysis is performed based on the collected detection data to obtain data analysis results; Based on the data analysis results, optimization instructions are generated and sent to the corresponding devices or detection modules for execution in subsequent production cycles.

[0113] The beneficial effects of this application embodiment compared with the prior art are as follows: This application embodiment not only realizes real-time quality control and defect interception of each process, but also automatically generates accurate process optimization instructions by collecting and analyzing in-depth detection data across processes and applying them to subsequent production. This enables the manufacturing system to get rid of its high dependence on manual offline analysis and experience-based debugging, realizes adaptive adjustment of process parameters, significantly shortens the optimization response cycle, improves the real-time adaptability to production fluctuations and the ability to predict and avoid potential quality risks, and ultimately improves product yield and production efficiency.

[0114] In some embodiments of this application, the control module is further configured to: count the number of non-conforming results detected by the detection module of the target device within a preset time period for any target device; When the quantity exceeds the first preset threshold, a data analysis result indicating an abnormal process state of the target equipment is generated.

[0115] In some embodiments of this application, the control module is further configured to: determine at least one specific defect type associated with the abnormal process state and the process parameters in the target equipment that cause the specific defect type; The adjustment amount is calculated based on the historical set values ​​of the process parameters to be adjusted, the degree of anomaly in the data analysis results, and the feedback data on the effects of adjusting the same parameters in the past. The optimization instruction is generated based on the adjustment amount.

[0116] In some embodiments of this application, the control module is further configured to: associate and store the detection data with the unique identifier of the corresponding optical module to form a cross-process detection data set corresponding to the unique identifier; Data mining analysis is performed on the cross-process detection data sets of each of the multiple optical modules to obtain the data mining analysis results.

[0117] In some embodiments of this application, the control module is further configured to: perform time-series alignment and feature extraction on the cross-process detection data sets of the plurality of optical modules respectively, to obtain a multi-dimensional feature sequence corresponding to each optical module; Calculate the statistical correlation coefficient between any two feature dimensions belonging to different processes in the multidimensional feature sequence to obtain a correlation coefficient matrix. The correlation coefficient matrix is ​​used to characterize the linear correlation strength between the features of each process. Based on the correlation coefficient matrix, feature dimension pairs with a correlation strength exceeding a second preset threshold are identified; Based on the identified feature dimensions and their corresponding process information, association direction and strength, a set of association rules is generated as the result of the data mining analysis.

[0118] In some embodiments of this application, the control module is further configured to: parse any association rule in the association rule set, and determine the causal process feature, the result process feature, and the association direction and strength between the two in the association rule; Obtain real-time measurements of the characteristics of the cause process in the current production flow, and calculate the recommended adjustment amount of the equipment process parameters corresponding to the cause process based on the correlation direction and intensity. The optimization instruction is generated based on the suggested adjustment amount, the associated rule identifier, and the characteristics of the target result process.

[0119] In some embodiments of this application, the process is specialized for visual inspection, including inspecting at least one of the following: dispensing quality, assembly alignment, insertion depth, marking position, cap alignment, and screw fastening status.

[0120] like Figure 3 The diagram shown is a schematic of a terminal device provided in an embodiment of this application. The terminal device 3 may include: a processor 301, a memory 302, and a computer program 303 stored in the memory 302 and executable on the processor 301, such as a control program for an AI intelligent detection method system for optical module manufacturing. When the processor 301 executes the computer program 303, it implements the steps in the control embodiments of the AI ​​intelligent detection method system for optical module manufacturing described above, for example... Figure 1 Steps S101 to S105 are shown.

[0121] A computer program can be divided into one or more modules / units. One or more modules / units are stored in memory 302 and executed by processor 301 to complete this application. One or more modules / units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in a terminal device.

[0122] The terminal device may include, but is not limited to, processor 301 and memory 302. Those skilled in the art will understand that... Figure 3 This is merely an example of a terminal device and does not constitute a limitation on the terminal device. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, a terminal device may also include input / output devices, network access devices, buses, etc.

[0123] The processor 301 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0124] The memory 302 can be an internal storage unit of the terminal device, such as the hard drive or RAM of the terminal device. The memory 302 can also be an external storage device of the terminal device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory 302 can include both internal and external storage units of the terminal device. The memory 302 is used to store computer programs and other programs and data required by the terminal device. The memory 302 can also be used to temporarily store data that has been output or will be output.

[0125] It should be noted that, for the sake of convenience and brevity, the structure of the terminal device described above can also be referred to the specific description of the structure in the method embodiment, which will not be repeated here.

[0126] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0127] This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement the steps in the control method of the AI ​​intelligent detection method system for manufacturing optical modules described above.

[0128] This application provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to implement the steps in the control method of the AI ​​intelligent detection method system for optical module manufacturing described above.

[0129] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0130] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for various specific applications, but such implementations should not be considered beyond the scope of this application.

[0131] In the embodiments provided in this application, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0132] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0133] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0134] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0135] The embodiments described above are merely illustrative of the technical solutions of this application and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An AI-powered intelligent inspection method for manufacturing optical modules, characterized in that, The AI ​​intelligent detection method for optical module manufacturing is applied to the control module of the optical module manufacturing system. The optical module manufacturing system also includes, in sequence along the production line, a top cover feeding and gluing device, a handle and spring pre-assembly device, a protective cotton and PCBA assembly device, a bottom shell feeding and gluing device, a marking and cover closing device, and a locking and finished product unloading device. Each device is equipped with a detection module for performing visual inspection and a sorting mechanism for performing sorting actions. The method includes: The equipment is controlled sequentially to perform corresponding assembly processes on the optical modules that flow through it; In response to the optical module completing its corresponding process in any device, the detection result generated by the process-specific visual inspection performed on the optical module by the detection module set in that device is obtained, and the sorting mechanism of the device is controlled according to the detection result to sort the optical module with the unqualified detection result to the NG channel. When acquiring the detection results of the detection module on each device, the detection data of that detection module is collected and recorded synchronously; Data analysis is performed based on the collected detection data to obtain data analysis results; Based on the data analysis results, optimization instructions are generated and sent to the corresponding devices or detection modules for execution in subsequent production cycles.

2. The AI ​​intelligent detection method for optical module manufacturing as described in claim 1, characterized in that, The step of performing data analysis based on the collected detection data to obtain data analysis results includes: For any target device, count the number of non-conforming results detected by the detection module of the target device within a preset time period; When the quantity exceeds the first preset threshold, a data analysis result indicating an abnormal process state of the target equipment is generated.

3. The AI ​​intelligent detection method for optical module manufacturing as described in claim 2, characterized in that, The step of generating optimization instructions based on the data analysis results includes: Identify at least one specific defect type associated with the abnormal process state and the process parameters in the target equipment that cause the specific defect type; The adjustment amount is calculated based on the historical set values ​​of the process parameters to be adjusted, the degree of anomaly in the data analysis results, and the feedback data on the effects of adjusting the same parameters in the past. The optimization instruction is generated based on the adjustment amount.

4. The AI ​​intelligent detection method for optical module manufacturing as described in claim 1, characterized in that, The step of performing data analysis based on the collected detection data to obtain data analysis results includes: The detection data is associated with the unique identifier of the corresponding optical module and stored to form a cross-process detection data set corresponding to the unique identifier; Data mining analysis is performed on the cross-process detection data sets of each of the multiple optical modules to obtain the data mining analysis results.

5. The AI ​​intelligent detection method for optical module manufacturing as described in claim 4, characterized in that, The step involves performing data mining analysis based on the cross-process detection data sets of each of the multiple optical modules to obtain data mining analysis results, including: Time-series alignment and feature extraction are performed on the cross-process detection data sets of the multiple optical modules to obtain a multi-dimensional feature sequence corresponding to each optical module; Calculate the statistical correlation coefficient between any two feature dimensions belonging to different processes in the multidimensional feature sequence to obtain a correlation coefficient matrix. The correlation coefficient matrix is ​​used to characterize the linear correlation strength between the features of each process. Based on the correlation coefficient matrix, feature dimension pairs with a correlation strength exceeding a second preset threshold are identified; Based on the identified feature dimensions and their corresponding process information, association direction and strength, a set of association rules is generated as the result of the data mining analysis.

6. The AI ​​intelligent detection method for optical module manufacturing as described in claim 5, characterized in that, The step of generating optimization instructions based on the data analysis results includes: Analyze any association rule in the set of association rules to determine the causal process characteristics, the result process characteristics, and the direction and strength of the association between them in the association rule; Obtain real-time measurements of the characteristics of the cause process in the current production flow, and calculate the recommended adjustment amount of the equipment process parameters corresponding to the cause process based on the correlation direction and intensity. The optimization instruction is generated based on the suggested adjustment amount, the associated rule identifier, and the characteristics of the target result process.

7. The AI ​​intelligent detection method for optical module manufacturing as described in claim 1, characterized in that, The process-specific visual inspection includes inspecting at least one of the following: dispensing quality, assembly alignment, insertion depth, marking position, cap alignment, and screw fastening status.

8. An optical module manufacturing system, characterized in that, The optical module manufacturing system includes a control module, a top cover feeding and gluing equipment, a handle and spring pre-assembly equipment, a protective cotton and PCBA assembly equipment, a bottom shell feeding and gluing equipment, a marking and cover closing equipment, and a locking and finished product unloading equipment. Each piece of equipment is equipped with a detection module for performing visual inspection and a sorting mechanism for performing sorting actions. The control module is used for: The equipment is controlled sequentially to perform corresponding assembly processes on the optical modules that flow through it; In response to the optical module completing its corresponding process in any device, the detection result generated by the process-specific visual inspection performed on the optical module by the detection module set in that device is obtained, and the sorting mechanism of the device is controlled according to the detection result to sort the optical module with the unqualified detection result to the NG channel. When acquiring the detection results of the detection module on each device, the detection data of that detection module is collected and recorded synchronously; Data analysis is performed based on the collected detection data to obtain data analysis results; Based on the data analysis results, optimization instructions are generated and sent to the corresponding devices or detection modules for execution in subsequent production cycles.

9. A terminal device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program and the processor runs the computer program to enable the terminal device to perform the AI ​​intelligent detection method for manufacturing an optical module according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the AI ​​intelligent detection method for manufacturing optical modules as described in any one of claims 1-7.

Images