Escalator intelligent manufacturing production line collaborative control method and system

Abstract

This invention relates to the technical field of escalator production line control, specifically to a collaborative control method and system for an intelligent escalator manufacturing production line. The method includes the following steps: acquiring workpiece model information input by the operator; determining the cause of the current physical state deviation of the workpiece based on the workpiece model information and the production line operating status; obtaining the workpiece's physical characteristic parameters from pre-stored product parameter information based on the cause of the physical state deviation, and adjusting the subsequent workstation's operation control strategy for the workpiece based on the physical characteristic parameters. The adjustment of the operation control strategy includes adjusting the conveyor belt operating parameters and adjusting the timing of the positioning fixture's actions; providing feedback on the adjusted operation control strategy to the operator and receiving confirmation information from the operator regarding the workpiece's positioning status. This helps improve the adaptability and reliability of the production line.

Description

Technical Field

[0001] This invention relates to the technical field of escalator production line control, specifically to a collaborative control method and system for an intelligent escalator manufacturing production line. Background Technology

[0002] In modern industrial production, intelligent manufacturing production lines strive for efficient and precise production through automated equipment and centralized control systems. However, when production lines need to frequently switch to accommodate small-batch, multi-model customized orders, traditional, fixed-cycle production models face severe challenges. This is especially true in intelligent escalator production lines, where the introduction of new products leads to subtle, unmodeled differences in their physical properties (e.g., surface friction coefficient) compared to the internal physical model of the collaborative control system. Even with manual input of the model number by the operator, the system struggles to consistently and accurately perceive and correct the minute physical position deviations accumulated during multi-station transmission caused by these differences. Summary of the Invention

[0003] The purpose of this invention is to address the aforementioned shortcomings by proposing a collaborative control method and system for an intelligent manufacturing production line of escalators. The present invention adopts the following technical solution: A collaborative control method for an intelligent manufacturing production line of escalators, the method comprising the following steps: Obtain the workpiece model information input by the operator; Based on the workpiece model information and the production line operating status, determine the cause of the current physical state deviation of the workpiece; Based on the cause of the physical state deviation, the physical characteristic parameters of the workpiece are obtained from the pre-stored product parameter information. Based on the physical characteristic parameters, the operation control strategy of the subsequent workstations for the workpiece is adjusted. The adjustment of the operation control strategy includes the adjustment of the conveyor belt operation parameters and the adjustment of the timing of the positioning tooling actions. It provides operators with feedback on adjustments to the operation control strategy and receives confirmation from operators regarding the workpiece positioning status. This technical solution effectively solves the problem of workpiece physical position deviation caused by the mismatch between the physical characteristics of new products and the internal physical model of the system. By dynamically adjusting the operation control strategy, it ensures the accurate positioning and stable transmission of workpieces on the production line, thereby avoiding problems such as visual recognition failure, inaccurate positioning and reduced welding quality, and significantly improving the adaptability and reliability of the production line. This application also discloses a collaborative control system for an escalator intelligent manufacturing production line, applied to the aforementioned collaborative control method for the escalator intelligent manufacturing production line. The system includes: The acquisition module retrieves the workpiece model information input by the operator. The judgment module determines the cause of the current physical state deviation of the workpiece based on the workpiece model information and the production line operating status. The adjustment module obtains the physical characteristic parameters of the workpiece from the pre-stored product parameter information based on the cause of the physical state deviation, and adjusts the operation control strategy of the subsequent workstations on the workpiece based on the physical characteristic parameters. The adjustment of the operation control strategy includes the adjustment of the conveyor belt operation parameters and the adjustment of the timing of the positioning tooling actions. The confirmation module provides feedback to the operator regarding adjustments to the operation control strategy and receives confirmation from the operator regarding the workpiece positioning status. This technical solution provides a system for implementing the aforementioned collaborative control method. Through modular design, it enables the determination of the cause of workpiece physical state deviation, adjustment of operation control strategy, and receipt of operator confirmation information. This effectively solves problems such as inaccurate workpiece positioning and unstable transmission encountered by production lines when facing multi-model, small-batch customized production, thereby improving the intelligence level and production efficiency of the production line. This application significantly improves the adaptability and flexibility of the intelligent escalator manufacturing production line to multi-model, small-batch customized production by introducing a dynamic operation control strategy adjustment mechanism based on workpiece physical characteristic parameters. It effectively solves the safety and quality problems of the production line caused by the mismatch of physical models in the existing technology, realizes the intelligent, precise and efficient production process, and has significant and excellent technical effects. To further understand the features and technical content of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are for reference and illustration only and are not intended to limit the present invention. Attached Figure Description

[0004] Figure 1 This is a flowchart of the collaborative control method for an intelligent manufacturing production line of escalators according to the present invention; Figure 2 This is a schematic diagram of the collaborative control system for the intelligent manufacturing production line of escalators according to the present invention. Detailed Implementation

[0005] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of the present invention. Furthermore, the accompanying drawings of the present invention are for simple illustrative purposes only and are not depictions of actual dimensions; this is stated in advance. The following embodiments will further describe the relevant technical content of the present invention in detail, but the disclosed content is not intended to limit the scope of protection of the present invention. This embodiment provides a collaborative control method and system for an intelligent manufacturing production line of escalators, combined with... Figure 1 and Figure 2 As shown. refer to Figure 1 A collaborative control method for an intelligent manufacturing production line of escalators, the method comprising the following steps: Obtain the workpiece model information input by the operator; Based on the workpiece model information and the production line operating status, determine the cause of the current physical state deviation of the workpiece; Based on the cause of the physical state deviation, the physical characteristic parameters of the workpiece are obtained from the pre-stored product parameter information. Based on the physical characteristic parameters, the operation control strategy of the subsequent workstations for the workpiece is adjusted. The adjustment of the operation control strategy includes the adjustment of the conveyor belt operation parameters and the adjustment of the timing of the positioning tooling actions. It provides operators with feedback on adjustments to the operation control strategy and receives confirmation from operators regarding the workpiece positioning status. The "Workpiece Model Information" refers to the product identifier entered by the operator through the human-machine interface, such as "Escalator Truss Model A" or "Escalator Truss Model B." This information is used by the system to identify the type of workpiece currently being produced. "Production Line Operation Status" covers the real-time operating status of each piece of equipment on the production line, including conveyor belt speed, positioning fixture position, sensor feedback data, and the occupancy status of each workstation. "Physical State Deviation Reasons" refers to the fundamental factors that cause the workpiece to fail to reach the designated position on the production line according to the expected trajectory or time, such as changes in the coefficient of friction or accumulated dimensional tolerances. "Product Parameter Information" is detailed physical attribute data about different workpiece models, such as weight, dimensions, surface friction coefficient, and center of gravity position, pre-stored in the system database. "Operation Control Strategy" refers to the set of operating instructions for the equipment on the production line, including the start / stop, speed, and acceleration of the conveyor belt, as well as the sequence, time interval, and force of the positioning fixture's actions. The collaborative control method for intelligent manufacturing production lines of escalators disclosed in this application may include the following steps in its specific implementation: First, the workpiece model information input by the operator is obtained. This step can be achieved in several ways. For example, the operator can manually input the model code of the escalator workpiece to be processed through the human-machine interface of the production line control system. Another way is for the operator to scan the barcode or QR code on the workpiece, and the scanning device will automatically read and input the workpiece model information. In addition, a voice recognition system can be used, where the operator verbally states the workpiece model, and the system converts it into digital information. Secondly, based on the workpiece model information and the production line's operating status, the system determines the cause of the workpiece's current physical state deviation. After obtaining the workpiece model information, the system analyzes it in conjunction with real-time production line operating status data. For example, the system can retrieve the theoretical physical characteristic parameters of the workpiece model from the database based on the workpiece model information. Simultaneously, the system continuously monitors sensor data on the production line, such as the workpiece arrival time detected by photoelectric sensors and the workpiece position captured by vision sensors. By comparing the differences between the actual monitored data and the theoretically expected data, and considering the characteristics of the workpiece model, the system can deduce the specific cause of the workpiece's physical state deviation. For example, if the system finds that the workpiece arrives at the positioning station later than expected, and the surface friction coefficient of this workpiece model shows significant fluctuations in historical data, the system may determine that the deviation is caused by slippage due to changes in the friction coefficient. Secondly, based on the cause of the physical state deviation, the system retrieves the workpiece's physical characteristic parameters from the pre-stored product parameter information. Based on these parameters, it adjusts the operation control strategy for subsequent workstations, including adjustments to the conveyor belt's operating parameters and the timing of the positioning fixture's actions. Once the cause of the physical state deviation is determined, the system further queries the pre-stored product parameter information to obtain detailed physical characteristic parameters related to the current workpiece model and the cause of the deviation. For example, if the cause of the deviation is a change in the coefficient of friction, the system obtains the range of the coefficient of friction for that workpiece model under different surface treatments. Based on these physical characteristic parameters, the system intelligently adjusts the operation control strategy for subsequent workstations. Specifically, regarding the adjustment of conveyor belt operating parameters, the system can dynamically adjust the conveyor belt's speed, acceleration, or deceleration based on the workpiece's actual coefficient of friction to compensate for slippage or premature arrival trends. For example, if the workpiece is prone to slippage, the system can appropriately reduce the conveyor belt's operating speed or use a smoother acceleration / deceleration curve in critical positioning areas. For adjusting the timing of positioning fixture actions, the system can fine-tune the start time or duration of the positioning fixture (such as a pneumatic locking pin) based on the actual arrival time or positional deviation of the workpiece, to ensure that it locks when the workpiece reaches the precise position. For example, if the workpiece tends to arrive earlier, the timing of the positioning fixture actions can be appropriately advanced. Finally, the system provides feedback to the operator on the adjustment of the operation control strategy and receives confirmation from the operator regarding the workpiece positioning status. After the system completes the adjustment of the operation control strategy, it will provide feedback to the operator on the specific details of the adjustment (e.g., "conveyor belt speed reduced by 5%", "positioning fixture movement advanced by 0.2 seconds") through the human-machine interface, indicator lights, or audible and visual alarms. After observing the actual positioning status of the workpiece after the adjustment, the operator can input confirmation information through the human-machine interface, such as clicking the "Confirm Positioning Accurate" button, or after visual inspection confirms that there are no errors, the system will automatically receive the confirmation signal. The collaborative control method for intelligent manufacturing production line of escalators in this application achieves precise transmission and positioning of escalator workpieces by introducing an intelligent judgment mechanism for the cause of workpiece physical state deviation and dynamically adjusting the operation control strategy of each workstation on the production line based on this. Specifically, after the operator inputs the workpiece model information, the system intelligently analyzes and determines the cause of the workpiece's current physical state deviation, based on the real-time operating status of the production line. For example, if the system detects abnormal slippage of the workpiece on the conveyor belt, or if the time it takes to reach the positioning station is not as expected, the system will infer the specific physical cause of the deviation based on this information and the characteristics of the workpiece model, such as changes in the coefficient of friction or accumulated dimensional tolerances. Once the cause of the deviation is determined, the system retrieves the detailed physical characteristic parameters of the workpiece from the pre-stored product parameter information. Based on these parameters, the system dynamically adjusts the operation control strategy of subsequent stations for the workpiece. For example, if the deviation is caused by slippage due to an excessively low coefficient of friction, the system will adjust the conveyor belt's operating parameters, such as reducing the speed or optimizing the acceleration / deceleration curve, to reduce slippage. Simultaneously, the system will adjust the timing of the positioning fixture's actions to ensure it locks when the workpiece reaches the precise position, thereby compensating for accumulated physical position deviations. After the strategy adjustment is completed, the system will feed back the adjustment information to the operator and receive confirmation of the workpiece's positioning status from the operator. This closed-loop control mechanism enables the production line to adapt to the differences in physical characteristics of different workpiece models in real time, effectively solving the problem of inaccurate workpiece positioning caused by differences in physical characteristics when traditional production lines face new products or multiple model switching. This application achieves intelligent perception and diagnosis of the workpiece's physical state by introducing a core step of "determining the cause of the workpiece's current physical state deviation based on workpiece model information and production line operating status." This allows the system to move beyond relying solely on preset models and instead identify and analyze the root causes of workpiece positioning deviations in real time. Furthermore, based on the determined cause of the physical state deviation, this application retrieves the workpiece's physical characteristic parameters from pre-stored product parameter information and dynamically adjusts the subsequent workstation's operation control strategy for the workpiece, including adjusting the conveyor belt's operating parameters and the timing of the positioning fixture's actions. This dynamic adjustment mechanism enables the production line to proactively adapt to the actual physical characteristics of the workpiece, effectively compensating for accumulated deviations during transmission and ensuring the workpiece's precise positioning at each workstation. For example, when the system determines that the workpiece is sliding due to changes in the coefficient of friction, it can immediately adjust the conveyor belt speed and the timing of the positioning fixture's actions to ensure that the gantry is precisely locked when it reaches the positioning fixture. Furthermore, this application also includes providing operators with feedback on adjustments to the operational control strategy and receiving confirmation from operators regarding the workpiece positioning status, forming a human-machine collaborative closed-loop control system. This not only improves the automation level of the production line but also fully utilizes the operator's experience and judgment, further enhancing the system's robustness and reliability. Through the above innovations, the method of this application can effectively avoid problems such as decreased welding quality, equipment damage, and production line downtime caused by inaccurate workpiece positioning, significantly improving the flexibility, efficiency, and product quality of the escalator intelligent manufacturing production line. This application further proposes that the above-mentioned collaborative control method for intelligent manufacturing production lines of escalators also includes the following steps: A vision sensor is integrated into the end of the welding torch of the welding robot to acquire image data of the molten pool; Analyze the image data of the molten pool, extract the geometric shape and temperature distribution characteristics of the molten pool, and compare them with the preset ideal reference characteristics to quantify the deviation of the molten pool; Based on the deviation of the molten pool, welding parameter adjustment instructions are generated and sent to the controller of the welding robot to adjust the welding parameters of the welding torch. Based on the deviation of the molten pool, fine-tuning instructions for the welding torch posture and trajectory are generated and sent to the controller of the welding robot to adjust the posture or trajectory of the welding torch. The system continuously acquires image data of the molten pool and iteratively adjusts the welding parameters, posture, and trajectory of the welding torch based on the deviation of the molten pool. Specifically, a vision sensor is integrated into the end of the welding torch of the welding robot. This vision sensor can be understood as a sensor capable of capturing the state of the weld pool in real time. Examples include high-speed industrial cameras, infrared thermal imagers, or multispectral sensors. Its purpose is to provide high-resolution images of the weld pool or temperature distribution data. This vision sensor is precisely mounted at the end of the welding torch to ensure that it can acquire image data of the weld pool at close range and without obstruction, thereby enabling real-time monitoring of the welding process. Analyzing the image data of the molten pool involves processing and parsing the acquired image data to extract feature information closely related to welding quality. Specifically, this includes using image processing algorithms (such as edge detection, region segmentation, and morphological processing) to identify the boundaries of the molten pool, thereby calculating its geometric features such as width, length, and area. Simultaneously, through grayscale analysis or temperature calibration, the temperature distribution characteristics within the molten pool can be extracted, such as the highest temperature point and temperature gradient. These extracted features are then compared with preset ideal reference features. These preset ideal reference features are established based on extensive experimental data, process specifications, or expert experience, representing the characteristics of the molten pool under high-quality welding conditions. By comparing, the deviation between the molten pool and the ideal state can be quantified, indicating, for example, an excessively wide or narrow, asymmetrical, or temperature-abnormal molten pool, or the presence of potential defects such as porosity or slag inclusions. In practical applications, welding parameter adjustment instructions are generated based on the deviation of the molten pool. The purpose is to correct abnormal states of the molten pool by adjusting the welding parameters. For example, when the molten pool width is too large, it may be necessary to reduce the welding current or increase the welding speed; when the molten pool temperature is too low, it may be necessary to increase the welding voltage or reduce the wire feed speed. These adjustment instructions are sent to the welding robot's controller in real time, and the controller modifies the corresponding welding parameters according to the instructions to make the molten pool state approach the ideal state. Furthermore, fine-tuning instructions for the welding torch's attitude and trajectory are generated based on the deviation of the molten pool. The purpose is to optimize the welding process by adjusting the spatial position and movement path of the welding torch. For example, if the molten pool is asymmetrical or deviates from the weld centerline, it may be necessary to fine-tune the welding torch's tilt angle, oscillation amplitude, or its trajectory on the workpiece. These fine-tuning instructions are also sent to the welding robot's controller, which precisely adjusts the welding torch's attitude or trajectory accordingly to ensure that the weld formation meets design requirements. Furthermore, by continuously acquiring image data of the molten pool and iteratively adjusting the welding parameters, attitude, and trajectory of the welding torch based on the deviation of the molten pool, the aim is to achieve a closed-loop adaptive control process. This means that the system does not make adjustments all at once, but continuously monitors the state of the molten pool and makes continuous, small-scale adjustments based on the latest deviation. This iterative adjustment mechanism enables the welding process to dynamically adapt to various uncertainties such as batch differences in materials, fluctuations in ambient temperature, and equipment wear, thereby ensuring the stability and consistency of welding quality. This application's solution integrates a vision sensor at the end of the welding torch of a welding robot, enabling real-time, non-contact monitoring of the weld pool. Image data of the weld pool is continuously acquired and analyzed, allowing for precise extraction of its geometric shape and temperature distribution characteristics. This real-time, quantified deviation information enables the system to promptly generate targeted welding parameter adjustment commands and torch posture and trajectory fine-tuning commands, which are then sent to the welding robot's controller. Consequently, the welding robot can dynamically adjust its welding parameters, torch posture, and trajectory based on the actual weld pool condition, forming a closed-loop adaptive control system. This continuous iterative adjustment mechanism allows the welding process to proactively address various uncertainties, effectively avoiding quality problems caused by localized deviations during processing. Through the above technical solution, this application overcomes the limitations of traditional methods that lack real-time feedback and adaptive adjustment capabilities during the welding process. Real-time visual perception and quantification of molten pool deviation enable welding quality control to shift from passive detection to proactive prevention. This significantly improves the stability and consistency of welding quality, effectively reduces welding defects, and lowers scrap rates and rework costs. Furthermore, this solution achieves intelligent and automated welding processes, reducing reliance on human experience and improving the overall efficiency and reliability of the production line. In some preferred embodiments, this application is implemented as follows: Assume that on an escalator intelligent manufacturing production line, a welding robot is automatically welding the key load-bearing structure of the escalator truss. Before the welding operation begins, the workpiece's model information is acquired, and the conveyor belt's operating parameters and the timing of the positioning fixture's actions are adjusted based on the cause of its physical state deviation to ensure the truss is accurately positioned at the welding station. After the welding operation starts, the vision sensor integrated into the welding robot's welding torch continuously acquires image data of the molten pool. For example, when the system detects that the width of the molten pool slightly exceeds the preset range and there are slight irregularities at the edge of the molten pool, this indicates a possible trend of excessively high welding current or excessively slow welding speed. At this time, the system immediately analyzes these image data, quantifies the geometric deviation of the molten pool, and generates corresponding welding parameter adjustment instructions, such as slightly reducing the welding current by 2A and slightly increasing the welding speed by 0.5mm / s. These instructions are sent to the welding robot's controller. Simultaneously, if the vision sensor further detects localized overheating or asymmetry in the temperature distribution of the molten pool, the system will generate fine-tuning instructions for the welding torch's posture and trajectory. For example, adjusting the torch's tilt angle by 0.5 degrees or performing micrometer-level corrections to the torch's trajectory. Upon receiving these instructions, the welding robot controller will adjust the welding parameters, posture, and trajectory of the welding torch in real time. Subsequently, the vision sensor continues to acquire molten pool image data and iteratively adjusts based on new molten pool deviations until the entire weld is completed. Through this real-time, closed-loop feedback and adjustment mechanism, even when encountering interference such as material batch differences or ambient temperature fluctuations during the welding process, the quality of the weld can be ensured to remain at its optimal state, thereby significantly improving the structural strength and reliability of the escalator truss. This application further proposes a collaborative control method for an intelligent manufacturing production line of escalators, which includes the following steps: The image data of the molten pool is continuously monitored for clarity, contrast and noise level to obtain image quality index. When the image quality index is lower than the preset quality threshold for multiple consecutive sampling periods, it is determined that there is contamination on the surface of the protective cover. When contamination is detected on the surface of the protective cover, the welding robot is controlled to pause the welding operation and move the welding torch to the preset cleaning station; At the cleaning station, start the cleaner to physically clean the surface of the protective cover; After cleaning, updated image quality metrics are calculated based on the sharpness, contrast, and noise level of the image data of the cleaned molten pool. When the updated image quality index reaches the preset quality threshold, control the welding robot to return to the original welding position and resume the welding operation. An alarm is triggered if the updated image quality index is still lower than the preset quality threshold. Specifically, continuously monitoring the sharpness, contrast, and noise level of molten pool image data refers to real-time analysis of the molten pool image data acquired by the vision sensor using image processing algorithms to calculate key indicators reflecting image quality. For example, sharpness can be measured by the image's gradient information or high-frequency components, contrast can be evaluated by the dynamic range of pixel grayscale values, and noise level can be detected by the image's statistical characteristics or specific filters. A comprehensive evaluation of these indicators forms an image quality index, used to quantify the usability of the current image data. The preset quality threshold can be understood as the minimum image quality standard to ensure that the molten pool image data meets the accuracy requirements of subsequent analysis and welding control. This threshold can be calibrated and set according to the actual application scenario, welding process requirements, and the performance of the vision sensor. When the image quality index consistently falls below this threshold for multiple consecutive sampling periods, it indicates that the decline in image quality is not an accidental fluctuation but a persistent problem, such as contamination on the protective cover surface. In practical applications, when contamination is detected on the protective cover surface, the system immediately suspends the welding robot's current welding operation to avoid welding defects caused by image quality issues. Simultaneously, the welding torch is precisely moved to a pre-set cleaning station, typically equipped with specialized cleaning equipment. At the cleaning station, the cleaner is activated to physically clean the protective cover surface of the vision sensor, removing contaminants through methods such as air jetting, brushing, or wiping. After cleaning, the system re-acquires image data of the molten pool and recalculates updated image quality metrics based on the clarity, contrast, and noise level of the cleaned image data. By comparing the updated image quality metrics with a preset quality threshold, the effectiveness of the cleaning operation can be evaluated. If the updated image quality metrics reach the preset quality threshold, the cleaning is considered successful, and the welding robot can safely return to its original welding position and resume welding. Conversely, if the updated image quality metrics remain below the preset quality threshold, it may indicate incomplete cleaning, damage to the protective cover, or other problems. In this case, the system will trigger an alarm, prompting the operator to conduct manual inspection or intervention. This application's solution effectively addresses the aforementioned issue of degraded image quality in visual sensors by introducing an intelligent detection and automatic cleaning mechanism for surface contamination on the protective cover of the visual sensor. Specifically, by continuously monitoring the clarity, contrast, and noise level of the molten pool image data, the system can assess the working status of the visual sensor in real time. When these image quality indicators continuously fall below preset thresholds, the system can promptly and accurately determine the presence of contamination on the protective cover surface, thereby preventing reduced welding control precision due to image quality degradation. It is precisely this proactive contamination detection that allows the system to take measures at the initial stage of the problem. Furthermore, when contamination is detected, the system can intelligently control the welding robot to pause welding and move to a cleaning station, activating the cleaner for physical cleaning. This automated cleaning process avoids the tediousness and delays of manual intervention, ensuring the continuity and efficiency of the production line. After cleaning, the system reassesses the image quality, forming a closed-loop feedback mechanism. If the image quality recovers to an acceptable level, welding operations resume, ensuring the accuracy of subsequent welding processes; if the image quality still fails to meet standards, an alarm is triggered, promptly reminding the operator to take further action, thereby avoiding potential welding quality problems. This solution, through preventative maintenance and adaptive adjustment, ensures that the vision sensor is always in optimal working condition, providing reliable image data support for high-precision welding control. In some preferred embodiments, it is assumed that a welding robot is welding an escalator truss in an escalator intelligent manufacturing production line. The welding robot's welding torch tip integrates a vision sensor for real-time monitoring of the molten pool. After continuous operation for a period of time, slight contamination begins to appear on the surface of the vision sensor's protective cover due to the accumulation of welding spatter and fumes. At this point, the system continuously monitors the image data of the molten pool and calculates its sharpness, contrast, and noise level. For example, the system detects that the image sharpness index is below a preset threshold of 0.7 for 10 consecutive sampling periods, the contrast index has decreased by 15%, and the noise level has increased by 20%. Based on this data, the system determines that the surface of the vision sensor's protective cover is contaminated. Once contamination is detected, the collaborative control system immediately sends a command to the welding robot controller to pause the current welding operation. Subsequently, the welding robot precisely moves the welding torch to a preset cleaning station. At the cleaning station, an integrated pneumatic cleaner is activated, using high-pressure airflow and a soft brush to physically clean the surface of the vision sensor's protective cover to remove the attached contaminants. After the cleaning process is complete, the system re-acquires image data of the molten pool and recalculates the image quality indicators. Assuming that after cleaning, the image sharpness index recovers to above 0.9, the contrast improves by 10%, and the noise level decreases by 15%, all reaching or exceeding the preset quality thresholds, the system determines that the cleaning was successful and controls the welding robot to return to its original welding position to resume the previous welding operation. The entire process requires no manual intervention, ensuring the continuity of the production line and welding quality. If, after cleaning, the image quality indicators still do not reach the preset quality threshold (e.g., sharpness only recovers to 0.6), the system will trigger an alarm and issue a prompt to the operator via the human-machine interface, indicating that manual inspection or replacement of the protective cover may be necessary. This prevents continued welding under poor image quality conditions and ensures product quality. This application further proposes steps for continuously monitoring the sharpness, contrast, and noise level of image data of the molten pool, including: Within the image acquisition area of ​​the visual sensor, multiple image sub-regions are divided; Image quality metrics for each image sub-region are continuously monitored. These metrics include sharpness, contrast, and noise level. When the image quality index of a single image sub-region is lower than the preset index threshold for multiple consecutive sampling periods, it is determined that the single image sub-region has local contamination. When the image quality index of multiple image sub-regions is lower than the preset index threshold for multiple consecutive sampling periods, it is determined that there is widespread contamination on the surface of the protective cover. Based on the type of localized or widespread contamination, corresponding cleaning strategy instructions are generated. Specifically, when the visual sensor acquires image data of the molten pool, its image acquisition area can be logically or physically divided into multiple image sub-regions. For example, the entire image frame can be divided into an M-row, N-column grid, with each grid cell being an image sub-region. This division aims to achieve refined perception of the contamination status of the protective shield surface. Continuous monitoring of image quality indicators for each image sub-region refers to calculating image quality indicators such as sharpness, contrast, and noise level for each independent image sub-region. These indicators can be calculated using industry-standard algorithms. For example, sharpness can be measured by the gradient magnitude or high-frequency component energy of the image; contrast can be calculated by the standard deviation or maximum / minimum range of pixel grayscale values; and noise level can be assessed by pixel fluctuations in smooth areas of the image. In practical applications, when the image quality indicator of a certain image sub-region remains below a preset threshold for multiple consecutive sampling periods, the system determines that there is localized contamination in that specific image sub-region. This indicates that the protective shield surface may only be affected by dust, splashes, or other substances adhering to a localized area. Furthermore, when the image quality indicators of multiple image sub-regions are all below a preset threshold for several consecutive sampling periods, it is determined that there is widespread contamination on the surface of the protective cover. This usually means that the entire protective cover is contaminated, such as a large area covered by smoke or dust or uniform oil stains. Therefore, based on the determined type of localized or widespread contamination, the system can generate corresponding cleaning strategy instructions. For example, for localized contamination, the cleaner can be instructed to perform localized cleaning only on the contaminated image sub-regions; for widespread contamination, the cleaner can be instructed to perform a complete cleaning. This application's solution achieves refined perception of surface contamination on the protective cover by dividing the image acquisition area of ​​the visual sensor into multiple image sub-regions and continuously monitoring image quality indicators for each sub-region. This regionalized monitoring allows the system to distinguish between localized and widespread contamination. When localized contamination is detected in a single image sub-region, it indicates limited contamination, eliminating the need to clean the entire protective cover and avoiding unnecessary downtime. Conversely, when contamination is present in multiple image sub-regions, it is determined to be widespread contamination, requiring comprehensive cleaning. This mechanism of generating corresponding cleaning strategy instructions based on contamination type makes cleaning operations more targeted and efficient, effectively solving the problem of insufficient precision in cleaning strategies under traditional overall monitoring methods. In some preferred embodiments, assuming a welding robot is welding an escalator truss, a vision sensor at the end of its welding torch continuously acquires image data of the molten pool. If the image acquisition area of ​​the vision sensor is divided into nine image sub-regions (3x3 grid), and the system detects that only the sharpness and contrast of the top-left image sub-region are below a preset threshold for five consecutive sampling periods, while the image quality indicators of the other eight image sub-regions are normal, the system determines that there is localized contamination on the protective cover surface. In this case, the system generates a localized cleaning strategy instruction, directing the cleaner to perform targeted cleaning only on the top-left area of ​​the protective cover. This avoids prolonged interruptions to the entire welding operation, requiring only a brief pause and localized processing before resumption. For example, if the system detects that the sharpness, contrast, and noise levels of all nine image sub-regions are below a preset threshold for three consecutive sampling periods, the system determines that there is widespread contamination on the protective cover surface. In this case, the system generates a comprehensive cleaning strategy instruction, directing the cleaner to thoroughly clean the entire protective cover surface. This differentiated approach ensures that the most appropriate measures can be taken in different pollution situations, thereby optimizing the operating efficiency and maintenance costs of the production line. This application further proposes steps for generating corresponding cleaning strategy instructions, including: Obtain information on production task priority, remaining cleaning agent quantity, and wear and tear of the cleaning equipment; Obtain historical cleaning performance data; Based on the priority information of production tasks, the remaining amount of cleaning agent, the wear and tear of the cleaner, and historical cleaning effect data, a cleaning strategy instruction is generated. Adjust the timing of cleaning strategy instructions based on the priority information of production tasks. Specifically, production task priority information refers to the importance or urgency of various tasks on the current production line. This information can be obtained from the production management system, for example, by querying the planned delivery date and urgency level of production orders. Its purpose is to ensure the smooth execution of critical production tasks and avoid unnecessary delays caused by cleaning operations. Cleaning agent remaining quantity information refers to the current inventory or available quantity of cleaning agent used to clean the protective cover surface. This information can be obtained in real time through devices such as level sensors and flow meters. Its purpose is to ensure that there is sufficient cleaning agent available when cleaning is needed, avoiding interruptions to cleaning operations due to insufficient cleaning agent. Cleaner wear information refers to the current wear condition of cleaning equipment (such as brushes and scrapers). This information can be obtained through equipment operating time, cleaning frequency statistics, or wear sensors. Its purpose is to assess the cleaning capacity of the cleaner and arrange maintenance or replacement as necessary to ensure cleaning effectiveness. Historical cleaning effect data refers to records of past cleaning operations, including contamination type, cleaning strategy, cleaning agent usage, cleaning time, and post-cleaning image quality indicators. This data can be stored in a database to provide experience and optimization basis for the generation of current cleaning strategies. The generation of cleaning strategy instructions can be understood as the system comprehensively analyzing the aforementioned information and using preset rules, algorithms, or machine learning models to calculate and determine the most suitable cleaning solution for the current situation. For example, when localized contamination is detected, the system will decide whether to perform immediate localized cleaning, postpone cleaning, or adopt a specific cleaning mode based on the priority of the production task, the status of the cleaning agent and cleaner, and historical data. In practical applications, adjusting the execution timing of cleaning strategy instructions specifically involves dynamically scheduling the start time or duration of cleaning operations based on the priority information of the production tasks. For example, for high-priority production tasks, cleaning operations may be postponed to production breaks or low-load periods; while for low-priority tasks, cleaning operations may be executed immediately. This application's solution comprehensively considers production task priority information, remaining cleaning agent information, cleaner wear information, and historical cleaning effect data. This allows the generation of cleaning strategies to move beyond a single contamination type and dynamically adapt to the actual operating conditions and resource constraints of the production line. Specifically, production task priority information ensures the continuity of high-priority tasks, avoiding unnecessary downtime; remaining cleaning agent information and cleaner wear information guarantee the feasibility and effectiveness of cleaning operations, preventing cleaning failures due to insufficient resources or equipment malfunctions; historical cleaning effect data provides valuable experience, enabling the system to learn and optimize cleaning parameters and processes, thereby generating more accurate and efficient cleaning strategy instructions. Furthermore, adjusting the execution timing of cleaning strategy instructions based on production task priority information effectively balances cleaning needs and production efficiency, preventing unnecessary interruptions or delays to critical production tasks caused by cleaning operations. In some preferred embodiments, it is assumed that the vision sensor continuously monitors a sub-region of an image containing localized contamination on the end cap of the welding robot's torch. In this case, the system first acquires the priority information of the current production task. For example, if the workpiece currently being welded belongs to a high-priority order requiring urgent delivery, the system assesses the impact of immediate cleaning on that order. Simultaneously, the system acquires information on the remaining amount of cleaning agent (e.g., the cleaning agent tank shows 80% remaining) and the wear level of the cleaner (e.g., the cleaner has been running for 500 hours and has moderate wear). Next, the system queries historical cleaning performance data and finds that for similar contamination types and cleaner configurations, a typical cleaning time of 5 minutes is required, and historical records show that cleaning operations were successfully scheduled during workpiece changeovers when production tasks were of high priority. Based on this information, the system generates a cleaning strategy instruction, for example, suggesting cleaning 3 minutes after the current high-priority workpiece is welded and before the next workpiece enters the workstation. In this way, the timing of the cleaning operation is intelligently adjusted, ensuring the cleanliness of the end cap while minimizing the impact on high-priority production tasks. This application further proposes steps for obtaining priority information for production tasks, including: Query the production management system to obtain information on all currently executing production orders. The production order information includes the order number, product model, planned delivery date, and urgency level indicator. Based on production order information, and combined with the real-time occupancy status of each workstation on the production line and the production schedule, calculate the process flow path and estimated completion time for each production order; Based on the difference between the planned delivery date and the current date, the urgency level indicator, and the estimated completion time of the process flow, an initial priority value is assigned to each production order to generate priority information for production tasks; When a cleaning operation request is detected, assess the impact of the cleaning operation on the process flow path and estimated completion time of the currently executing high-priority production order; Based on the assessed impact and the urgency of high-priority production orders, the priority of cleaning operations is adjusted, and a resource coordination request is sent to the production line scheduler. Specifically, querying the production management system aims to obtain detailed information on all currently active production orders on the production line. This production order information is fundamental data for assessing the priority of production tasks. Among these, the order number is used to uniquely identify each order; the product model indicates the type of workpiece to be produced, which may affect its production process and required time; the planned delivery date is the deadline by which the order must be completed and is a key factor in determining its urgency; the urgency level indicator directly reflects the customer's urgent requirement for the order, such as "rush" or "normal". Calculating the workflow path and estimated completion time based on production order information can be understood as determining, through simulation or actual calculation, the various workstations and their sequence required for each production order to proceed from its current state to completion, and estimating the total time under the current production line load. This requires considering the real-time occupancy status of each workstation on the production line, such as which workstations are occupied and which are idle, as well as the preset production schedule, to ensure the accuracy of the calculation results. Its purpose is to provide a quantifiable time basis for subsequent priority allocation. In practical applications, assigning an initial priority value to each production order involves considering multiple factors. The smaller the difference between the planned delivery date and the current date, the more urgent the order typically is; the urgency level indicator directly assigns an initial urgency to the order; and the estimated completion time of the workflow reflects the actual production cycle required to complete the order. By combining these factors through weighted averages or other algorithms, an objective initial priority value can be generated for each order. Furthermore, intelligent management is required when a cleaning operation request is detected. Assessing the impact of a cleaning operation on high-priority production orders involves analyzing which high-priority order workflows would be disrupted if cleaning were performed at this time, and by how much their estimated completion time would be extended. This can be done by simulating the occupancy of production line resources (such as specific workstations and conveyor belts) during the cleaning operation, and by combining this with the current progress and subsequent paths of high-priority orders for prediction. Therefore, based on the assessed impact and the urgency of high-priority production orders, the priority of cleaning operations is adjusted, and a resource coordination request is sent to the production line scheduler. This means that if a cleaning operation has a significant impact on a very high-priority order, its priority may be lowered or postponed; conversely, if the impact is minor, the cleaning operation can be given higher priority for expedited execution. Sending a resource coordination request to the production line scheduler aims to notify the scheduling system so that it can re-plan resource allocation and production scheduling to ensure that cleaning operations are carried out without severely impacting critical production tasks. This application's solution effectively addresses the conflict between cleaning strategy generation and high-priority production tasks in the aforementioned basic solution by introducing a dynamic priority assessment and adjustment mechanism. Specifically, firstly, comprehensive production order information is obtained by querying the production management system, providing a data foundation for priority assessment. Secondly, based on this order information and the real-time status of the production line, the process flow path and estimated completion time for each order are calculated, making the priority allocation of production tasks more accurate and quantifiable. More importantly, when a cleaning operation request is triggered, the system no longer simply executes it but intelligently assesses the impact of the cleaning operation on currently executing high-priority production orders. This assessment mechanism can identify potential production bottlenecks or delay risks. Finally, based on the assessment results and the urgency of high-priority orders, the priority of the cleaning operation is dynamically adjusted, and a resource coordination request is sent to the production line scheduler, thereby ensuring that cleaning and maintenance activities can efficiently coordinate with production tasks and avoid unnecessary interference with critical production processes. Through the above technical solution, this application enables refined management of production task priorities and optimizes the timing of cleaning operations. Compared to basic solutions that only obtain static priority information, this solution can dynamically assess the impact of cleaning operations on high-priority production orders, thereby avoiding production interruptions or delays caused by cleaning and maintenance. This significantly improves the overall collaborative efficiency and flexibility of the production line, ensuring the smooth progress and timely delivery of critical production tasks, while also guaranteeing the timeliness of equipment maintenance, effectively balancing production efficiency and equipment maintenance needs. In some preferred embodiments, suppose an escalator intelligent manufacturing production line is executing three production orders: Order A (product model X, planned delivery date T+1 day, urgency level "urgent"), Order B (product model Y, planned delivery date T+5 days, urgency level "normal"), and Order C (product model Z, planned delivery date T+3 days, urgency level "high"). At this time, a vision sensor detects localized contamination on the surface of the welding robot's protective cover and triggers a cleaning operation request. The system first queries the production management system to obtain detailed information for orders A, B, and C. Then, combining this information with the real-time occupancy status of each workstation on the production line and the production schedule, it calculates that order A is expected to be completed in T+0.5 days, order B in T+4 days, and order C in T+2 days. Based on this information, the system assigns order A the highest initial priority (e.g., 90 points), order C the next highest (e.g., 70 points), and order B the lowest (e.g., 50 points). When a cleaning operation request is detected, the system begins to assess the impact of the cleaning operation on currently executing high-priority production orders (Order A and Order C). Assume the cleaning operation will occupy the welding station for approximately 30 minutes. System simulations show that if cleaning is performed immediately, the estimated completion time of Order A will be delayed by 30 minutes, potentially resulting in non-delivery within T+1 days; the estimated completion time of Order C will also be delayed by 30 minutes. Based on the assessed impact and the urgency of order A, the system determines that immediate cleaning would severely impact order A. Therefore, the system adjusts the priority of the cleaning operation, lowering its priority, and sends a resource coordination request to the production line scheduler, suggesting that the cleaning operation be postponed until after order A is completed, or during a break in the production line when load is low. For example, the scheduler might decide to perform cleaning at the next shift handover after order A is completed, or to clean the welding robot after the welding process of order A is completed, while other workstations are handling subsequent processes for order A, thereby minimizing the impact on high-priority production tasks. The steps to obtain historical cleaning performance data include: Identify the workpiece model and configuration information of the cleaning equipment currently in use on the production line; Based on the workpiece model and the configuration information of the cleaning equipment currently in use on the production line, query the historical cleaning operation records that match the specific workpiece model and cleaning equipment configuration; Filter the retrieved historical cleaning operation records to exclude data that does not match the current type of contamination; The filtered historical cleaning operation records will be used as historical cleaning effect data. The identification of the workpiece model and the configuration information of the cleaning equipment currently in use on the production line refers to the system automatically or through sensors acquiring the specific specifications and types of the escalator components currently being produced, such as treads and trusses, as well as the model of the cleaning equipment, the type of cleaning medium, and nozzle parameters. Its purpose is to provide accurate matching conditions for subsequent queries. In practical applications, based on the workpiece model and the configuration information of the cleaning equipment currently in use on the production line, historical cleaning operation records matching specific workpiece models and cleaning equipment configurations can be retrieved from a preset database or historical data storage system to obtain past cleaning data similar to the current production environment and equipment conditions. Furthermore, the retrieved historical cleaning operation records are filtered to exclude data that does not match the current type of contamination. For example, if the current protective cover surface contains oil stains, historical records of oil stain treatment should be filtered out, while records of dust or welding slag treatment should be excluded. The purpose is to ensure that the acquired historical data is highly relevant to the current actual problem. Therefore, the filtered historical cleaning operation records are used as historical cleaning effect data, which will be used to generate subsequent cleaning strategy instructions to improve the accuracy and effectiveness of the strategy. This application's solution first identifies the workpiece model and cleaner configuration information of the current production line, providing precise context for historical data retrieval. Subsequently, based on this contextual information, the system can accurately locate cleaning operation records matching the current production conditions from a massive amount of historical data. More importantly, by filtering the retrieved historical records and actively excluding data inconsistent with the current contamination type, this application ensures that the acquired historical cleaning effect data is highly targeted and effective. It is precisely this refined data acquisition and filtering mechanism that allows subsequent generated cleaning strategy instructions to fully utilize past successful experiences, avoiding strategy deviations caused by data mismatches, thereby effectively solving the problem of low historical data utilization efficiency that may exist in basic solutions. In some preferred embodiments, it is assumed that the current intelligent escalator manufacturing production line is welding a specific model of escalator tread, and a vision sensor detects localized contamination on the surface of the welding torch shield caused by welding spatter. First, the system will identify the workpiece model of the current production line as "escalator tread - model A" and the cleaning device being used as "ultrasonic cleaner - model X". Next, the system searches the historical database for all relevant cleaning operation records based on the two criteria: "escalator tread - Model A" and "ultrasonic cleaner - Model X". These records may include past instances where "ultrasonic cleaner - Model X" was used to treat various contaminants during welding of "escalator tread - Model A". The system then filters the retrieved historical records. Since the current contamination type is "localized contamination caused by welding spatter," the system excludes records of treatments for oil, dust, or other non-spatter contamination, retaining only those related to welding spatter treatment. For example, it might filter out a cleaning operation record detailing the use of an "ultrasonic cleaner - model X" at a specific frequency and duration to clean welding spatter on an "escalator tread - model A," ultimately achieving the desired image quality. Ultimately, these selected historical cleaning operation records, which closely match the current type of contamination, will be used as historical cleaning effectiveness data to guide the generation of cleaning strategy instructions for the current welding spatter contamination. For example, they may recommend adjusting the power of the ultrasonic cleaner or the cleaning duration to achieve the best cleaning results. The steps for filtering historical cleaning operation records to exclude data that does not match the current contamination type include: Based on the pollution characteristic information recorded in the historical cleaning operation records, extract historical pollution characteristics; Extract real-time features of current surface contamination on the protective shield; Compare historical pollution characteristics with real-time characteristics to calculate the similarity between historical pollution types and current pollution types; Based on a preset similarity threshold, historical cleaning operation records that reach the preset similarity threshold are selected. "Historical pollution characteristics" refer to descriptive data about past pollution events obtained from stored historical cleaning operation records. These characteristics include the type of pollution (such as oil, dust, and particulate matter), distribution patterns, severity, and physicochemical properties of the pollutants. These characteristics can be stored in structured data formats, such as numerical values, classification labels, or feature vectors. The process of extracting historical pollution characteristics may involve reading, parsing, and formatting this information from a database. "Real-time features of current protective shield surface contamination" refers to a quantitative description of the current surface contamination of the protective shield obtained through image data acquired in real time via a visual sensor, after image processing and analysis. For example, this may include the area of ​​the contaminated region, the color, texture, brightness, contrast, particle size distribution, and spectral information of the contaminants. The extraction of these real-time features typically involves computer vision techniques such as image segmentation, feature point detection, texture analysis, and color histogram analysis. "Comparing historical pollution features with real-time features to calculate the similarity between historical and current pollution types" refers to using specific algorithms or models to quantitatively compare extracted historical and real-time pollution features, thereby deriving a numerical value representing the degree of similarity between the two. For example, distance or similarity measures such as Euclidean distance, cosine similarity, and Pearson correlation coefficient can be used. In more complex scenarios, machine learning models (such as support vector machines and neural networks) can also be used to classify or regress features to output a similarity score. The "preset similarity threshold" refers to a numerical limit set by the system before performing similarity comparisons. When the calculated similarity reaches or exceeds this threshold, the historical pollution type is considered sufficiently similar to the current pollution type, and only then will the historical cleaning operation record be included in the valid historical cleaning effect data. This threshold can be optimized based on actual application needs, empirical data, or through machine learning training. This application's solution introduces a quantitative comparison of contamination characteristics, making the screening process of historical cleaning operation records more objective and accurate. Specifically, firstly, structured historical contamination characteristics are extracted from historical records, laying the data foundation for subsequent comparisons. Secondly, the contamination characteristics of the current protective cover surface are acquired and extracted in real time, ensuring the timeliness and relevance of the comparison. Subsequently, by calculating the similarity between historical and real-time contamination characteristics, the abstract concept of "contamination type mismatch" is transformed into a quantifiable value, thus avoiding subjective judgment. Finally, screening is performed based on a preset similarity threshold, ensuring that only historical data highly correlated with the current contamination status is adopted, thereby providing a reliable basis for generating more precise cleaning strategy instructions. In some preferred embodiments, it is assumed that the current visual sensor detects a general contamination on the protective cover surface, primarily composed of fine particulate matter. Its real-time characteristics are manifested as a specific texture pattern in the image, low contrast, and uniformly distributed noise levels. The system first extracts various historical contamination features from historical cleaning operation records. For example, record A describes oil contamination (characterized by high brightness and irregular patches), record B describes coarse particulate contamination (characterized by large-sized, high-contrast dots), and record C describes fine dust contamination (characterized by uniformly distributed low-contrast texture). The system compares the current real-time contamination features with the historical contamination features of records A, B, and C, calculating their similarity. For example, the similarity with record A is 0.2, with record B is 0.3, and with record C is 0.9. If the preset similarity threshold is 0.8, only record C reaches this threshold. Therefore, the historical cleaning operation record corresponding to record C will be selected as valid historical cleaning effect data to guide the generation of cleaning strategies for current fine particulate matter contamination. This application further proposes a step for filtering historical cleaning operation records that meet a preset similarity threshold, including: Based on the workpiece model of the current production line, the configuration information of the cleaner in use, and the current type of contamination, select a threshold adjustment rule that matches the current production conditions from the preset threshold adjustment rule set. According to the threshold adjustment rules, the similarity threshold is adjusted, including raising or lowering the value of the similarity threshold. Based on the adjusted similarity threshold, historical cleaning operation records that reach the adjusted similarity threshold are selected. Specifically, when acquiring historical cleaning performance data, the first step is to select the most suitable rule from a pre-stored set of threshold adjustment rules, based on the current workpiece model on the production line, the configuration information of the cleaning equipment being used, and the type of contamination detected in real time. This set of threshold adjustment rules can include similarity threshold adjustment strategies for different combinations of workpieces, cleaning equipment, and contamination. For example, for workpieces made of a specific material, surface contamination may be more sensitive to the impact on cleaning performance, requiring a higher similarity threshold to ensure that the selected historical data is highly relevant; while for another type of workpiece, the threshold may be appropriately lowered to obtain a wider range of reference data. The threshold adjustment rules can be understood as a series of predefined logic or algorithms used to guide the dynamic changes of the similarity threshold. The adjustment process involves raising or lowering the similarity threshold value, with the aim of ensuring a higher degree of match between the selected historical cleaning operation records and the current situation. For example, when the current type of contamination is rare or the cleaner is worn out, it may be necessary to lower the similarity threshold to broaden the search scope and find more potentially useful historical data; conversely, when the production task has extremely high priority and stringent requirements for cleaning effectiveness, it may be necessary to raise the threshold to ensure that only the most precisely matching historical data is selected. In practical applications, after the similarity threshold is adjusted, the system reassesses the similarity between historical and real-time contamination characteristics based on the new threshold, and selects historical cleaning operation records whose similarity reaches the adjusted threshold. This process ensures that the acquired historical cleaning effect data more accurately reflects the cleaning needs and potential effects under current production conditions, providing a more reliable basis for generating subsequent cleaning strategy instructions. This application's solution effectively addresses the limitations of fixed thresholds in filtering historical data under varying production environments by introducing a dynamic adjustment mechanism for similarity thresholds. Specifically, when the workpiece model, cleaner configuration, or contamination type on the production line changes, the system can intelligently select and apply the most suitable adjustment rule from a preset threshold adjustment rule set based on these real-time conditions. For example, for fragile workpieces or scenarios with extremely high cleaning requirements, the threshold is raised to ensure that only historical data highly similar to the current situation is adopted, avoiding deviations in cleaning strategies due to inappropriate data. Conversely, in certain special cases, such as when the contamination type is uncommon or historical data is scarce, the threshold can be appropriately lowered to broaden data sources and ensure valuable references can be found even with limited data. It is precisely this adaptive threshold adjustment that makes the filtering process of historical cleaning effect data more accurate and flexible, thus laying the foundation for generating more targeted and effective cleaning strategy instructions. In some preferred embodiments, suppose the escalator intelligent manufacturing production line is processing a new type of escalator tread, the surface material of which is highly sensitive to the corrosiveness of cleaning agents, and the current visual sensor detects a mixed type of oil contamination on the protective cover surface. Without a dynamic threshold adjustment mechanism, the system might use a generic similarity threshold to filter historical cleaning data. However, due to the specific characteristics of this new tread model, a generic threshold may not be effective in filtering historical cleaning data specific to the sensitive material and the mixed oil contamination, resulting in an inaccurate cleaning strategy. According to the solution in this application, when the system identifies that the current workpiece model is a new escalator tread, the cleaner is configured as a specific type, and the contamination type is mixed oil stains, it will select a threshold adjustment rule specifically for the "sensitive material workpiece + mixed oil stains" scenario from a preset threshold adjustment rule set. This rule may indicate that the similarity threshold is increased from the default 0.7 to 0.85. Based on the adjusted similarity threshold of 0.85, the system will re-filter historical cleaning operation records. For example, if there is a cleaning record in the history of "similar sensitive material workpiece + similar mixed oil stains" with a similarity of 0.88, the record will be successfully filtered out; while if another record has a similarity of 0.75, it will be excluded. In this way, the system can ensure that only historical cleaning data that highly matches the current high-sensitivity production conditions is selected, thereby generating safer and more effective cleaning strategy instructions, such as recommending the use of specific mild detergents and adjusting the physical cleaning intensity of the cleaner to avoid damage to the new escalator tread and ensure cleaning effectiveness. refer to Figure 2 This application further proposes a collaborative control system for an intelligent manufacturing production line of escalators, the system comprising: The acquisition module retrieves the workpiece model information input by the operator. The judgment module determines the cause of the current physical state deviation of the workpiece based on the workpiece model information and the production line operating status. The adjustment module obtains the physical characteristic parameters of the workpiece from the pre-stored product parameter information based on the cause of the physical state deviation, and adjusts the operation control strategy of the subsequent workstations on the workpiece based on the physical characteristic parameters. The adjustment of the operation control strategy includes the adjustment of the conveyor belt operation parameters and the adjustment of the timing of the positioning tooling actions. The confirmation module provides feedback to the operator regarding adjustments to the operation control strategy and receives confirmation from the operator regarding the workpiece positioning status. The acquisition module can be a data input interface, such as a touch screen, keyboard, or interface for data communication with a higher-level control system. Its purpose is to receive instructions or data from the operator or automated scheduling system regarding the model of the workpiece to be processed. The judgment module can be understood as a data processing and analysis unit. It receives workpiece model information from the acquisition module and monitors in real time the workpiece's position, attitude, speed, and other operating status data fed back by various sensors on the production line (such as vision sensors, laser rangefinders, encoders, etc.). Through built-in logical judgment rules, machine learning models, or expert systems, this module can analyze this data and identify possible physical state deviations (such as misalignment, tilting, deformation, etc.) of the workpiece during the production process and their specific causes. In practical applications, the adjustment module receives the physical state deviation caused by the judgment module and accesses a pre-stored product database containing physical characteristic parameters of different workpiece models, such as dimensions, weight, center of gravity position, and material properties. Based on these parameters and the cause of the deviation, the adjustment module generates an optimized operation control strategy. This includes fine-tuning the operating parameters such as the speed, acceleration, and start / stop timing of the conveyor belt, as well as precisely controlling the timing of clamping, releasing, and moving actions of positioning fixtures (such as clamps and lifting mechanisms) to effectively correct the physical state deviation of the workpiece and ensure its accurate processing in subsequent workstations. The confirmation module can be a human-machine interface or a communication interface. Its purpose is to provide feedback to the operator on the operation control strategy adjustment plan generated by the adjustment module through a visual interface or alarm information. At the same time, this module is also responsible for receiving confirmation instructions or feedback information from the operator after observation, verification, or manual adjustment regarding whether the workpiece positioning status meets the requirements, thereby forming a closed-loop control to ensure the effectiveness and safety of the adjustment. This application's solution concretizes the abstract steps of the collaborative control method for escalator intelligent manufacturing production lines into system modules, achieving automated and intelligent execution of the method. The acquisition module, as the information input, ensures the production line can acquire key workpiece model information in a timely and accurate manner, laying the foundation for subsequent intelligent decision-making. The judgment module utilizes this workpiece model information and real-time production line operating status data, employing advanced analysis algorithms to accurately identify potential physical state deviations of the workpiece during production and their specific causes, thus providing a clear basis for corrective measures. The adjustment module, the core execution unit of the entire system, dynamically calculates and generates operating control strategies for subsequent production stations based on the deviation causes output by the judgment module and pre-stored workpiece physical characteristic parameters. This includes fine-tuning the conveyor belt operating parameters and the timing of positioning tooling actions, effectively correcting physical state deviations of the workpiece and ensuring precise positioning of the workpiece at each processing stage. The confirmation module provides a human-machine interface, allowing operators to participate in the control loop and ultimately confirm the adjustment effects, improving the system's reliability and safety. This systematic design enables the originally abstract methods and steps to operate efficiently, accurately, and reliably in the actual production environment, significantly improving the intelligence level of the production line. The content disclosed above is only a preferred and feasible embodiment of the present invention, and is not intended to limit the scope of protection of the present invention. Therefore, all equivalent technical changes made based on the content of the present invention specification and drawings are included within the scope of protection of the present invention. Furthermore, the elements therein can be updated as technology develops.

Claims

1. A collaborative control method for an intelligent manufacturing production line of escalators, characterized in that, The method includes the following steps: Obtain the workpiece model information input by the operator; Based on the workpiece model information and the production line operating status, determine the cause of the current physical state deviation of the workpiece; Based on the cause of the physical state deviation, the physical characteristic parameters of the workpiece are obtained from the pre-stored product parameter information. Based on the physical characteristic parameters, the operation control strategy of the subsequent workstations for the workpiece is adjusted. The adjustment of the operation control strategy includes the adjustment of the conveyor belt operation parameters and the adjustment of the timing of the positioning tooling actions. It provides operators with feedback on adjustments to the operation control strategy and receives confirmation from operators regarding the workpiece positioning status.

2. The collaborative control method for an intelligent manufacturing production line of escalators as described in claim 1, characterized in that, The method also includes the following steps: A vision sensor is integrated into the end of the welding torch of the welding robot to acquire image data of the molten pool; Analyze the image data of the molten pool, extract the geometric shape and temperature distribution characteristics of the molten pool, and compare them with the preset ideal reference characteristics to quantify the deviation of the molten pool; Based on the deviation of the molten pool, welding parameter adjustment instructions are generated and sent to the controller of the welding robot to adjust the welding parameters of the welding torch. Based on the deviation of the molten pool, fine-tuning instructions for the welding torch posture and trajectory are generated and sent to the controller of the welding robot to adjust the posture or trajectory of the welding torch. The system continuously acquires image data of the molten pool and iteratively adjusts the welding parameters, posture, and trajectory of the welding torch based on the deviation of the molten pool.

3. The collaborative control method for an intelligent manufacturing production line of escalators as described in claim 2, characterized in that, The method also includes the following steps: The image data of the molten pool is continuously monitored for clarity, contrast and noise level to obtain image quality index. When the image quality index is lower than the preset quality threshold for multiple consecutive sampling periods, it is determined that there is contamination on the surface of the protective cover. When contamination is detected on the surface of the protective cover, the welding robot is controlled to pause the welding operation and move the welding torch to the preset cleaning station; At the cleaning station, start the cleaner to physically clean the surface of the protective cover; After cleaning, updated image quality metrics are calculated based on the sharpness, contrast, and noise level of the image data of the cleaned molten pool. When the updated image quality index reaches the preset quality threshold, control the welding robot to return to the original welding position and resume the welding operation. An alarm is triggered if the updated image quality index is still lower than the preset quality threshold.

4. The collaborative control method for an intelligent manufacturing production line of escalators as described in claim 3, characterized in that, The steps for continuously monitoring the sharpness, contrast, and noise level of the molten pool image data include: Within the image acquisition area of ​​the visual sensor, multiple image sub-regions are divided; Image quality metrics for each image sub-region are continuously monitored. These metrics include sharpness, contrast, and noise level. When the image quality index of a single image sub-region is lower than the preset index threshold for multiple consecutive sampling periods, it is determined that the single image sub-region has local contamination. When the image quality index of multiple image sub-regions is lower than the preset index threshold for multiple consecutive sampling periods, it is determined that there is widespread contamination on the surface of the protective cover. Based on the type of localized or widespread contamination, corresponding cleaning strategy instructions are generated.

5. The collaborative control method for an intelligent manufacturing production line of escalators as described in claim 4, characterized in that, The steps to generate the corresponding cleaning strategy instructions include: Obtain information on production task priority, remaining cleaning agent quantity, and wear and tear of the cleaning equipment; Obtain historical cleaning performance data; Based on the priority information of production tasks, the remaining amount of cleaning agent, the wear and tear of the cleaner, and historical cleaning effect data, a cleaning strategy instruction is generated. Adjust the timing of cleaning strategy instructions based on the priority information of production tasks.

6. The collaborative control method for an intelligent manufacturing production line of escalators as described in claim 5, characterized in that, The steps to obtain priority information for production tasks include: Query the production management system to obtain information on all currently executing production orders. The production order information includes the order number, product model, planned delivery date, and urgency level indicator. Based on production order information, and combined with the real-time occupancy status of each workstation on the production line and the production schedule, calculate the process flow path and estimated completion time for each production order; Based on the difference between the planned delivery date and the current date, the urgency level indicator, and the estimated completion time of the process flow, an initial priority value is assigned to each production order to generate priority information for production tasks; When a cleaning operation request is detected, assess the impact of the cleaning operation on the process flow path and estimated completion time of the currently executing high-priority production order; Based on the assessed impact and the urgency of high-priority production orders, the priority of cleaning operations is adjusted, and a resource coordination request is sent to the production line scheduler.

7. The collaborative control method for an intelligent manufacturing production line of escalators as described in claim 5, characterized in that, The steps to obtain historical cleaning performance data include: Identify the workpiece model and configuration information of the cleaning equipment currently in use on the production line; Based on the workpiece model and the configuration information of the cleaning equipment currently in use on the production line, query the historical cleaning operation records that match the specific workpiece model and cleaning equipment configuration; Filter the retrieved historical cleaning operation records to exclude data that does not match the current type of contamination; The filtered historical cleaning operation records will be used as historical cleaning effect data.

8. The collaborative control method for an intelligent manufacturing production line of escalators as described in claim 7, characterized in that, The steps for filtering historical cleaning operation records to exclude data that does not match the current contamination type include: Based on the pollution characteristic information recorded in the historical cleaning operation records, extract historical pollution characteristics; Extract real-time features of current surface contamination on the protective shield; Compare historical pollution characteristics with real-time characteristics to calculate the similarity between historical pollution types and current pollution types; Based on a preset similarity threshold, historical cleaning operation records that reach the preset similarity threshold are selected.

9. The collaborative control method for an intelligent manufacturing production line of escalators as described in claim 8, characterized in that, The steps for filtering historical cleaning operation records that meet the preset similarity threshold include: Based on the workpiece model of the current production line, the configuration information of the cleaner in use, and the current type of contamination, select a threshold adjustment rule that matches the current production conditions from the preset threshold adjustment rule set. According to the threshold adjustment rules, the similarity threshold is adjusted, including raising or lowering the value of the similarity threshold. Based on the adjusted similarity threshold, historical cleaning operation records that reach the adjusted similarity threshold are selected.

10. A collaborative control system for an escalator intelligent manufacturing production line, applied to the collaborative control method for an escalator intelligent manufacturing production line as described in claim 1, characterized in that, The system includes: The acquisition module retrieves the workpiece model information input by the operator. The judgment module determines the cause of the current physical state deviation of the workpiece based on the workpiece model information and the production line operating status. The adjustment module obtains the physical characteristic parameters of the workpiece from the pre-stored product parameter information based on the cause of the physical state deviation, and adjusts the operation control strategy of the subsequent workstations on the workpiece based on the physical characteristic parameters. The adjustment of the operation control strategy includes the adjustment of the conveyor belt operation parameters and the adjustment of the timing of the positioning tooling actions. The confirmation module provides feedback to the operator regarding adjustments to the operation control strategy and receives confirmation from the operator regarding the workpiece positioning status.

Images