Safety detection system and method for sealing ring press vulcanizer

Abstract

The invention discloses a safety detection system and method for a sealing ring flat vulcanizing machine, and solves the problems that a traditional static protection system is single in detection and poor in adaptability, the system comprises a detection module, a processing module, an early warning execution module and a self-learning module, acquiring the position and posture of a person and the motion trail of an equipment part in real time; the processing module generates a double-track dynamic comparison model and calculates a real-time risk, and dynamically adjusts a safe distance threshold value in combination with the self-learning module; according to the method, closed-loop control is formed through track comparison, distance verification, data collection and threshold value optimization, the system supports operator identity recognition and personalized safety strategies, intelligent safety protection in a complex human-computer interaction scene is achieved, and the safety of the human-computer interaction scene is improved. The detection precision and the response flexibility are obviously improved, and the production safety and the equipment operation efficiency are balanced.

Description

Technical Field

[0001] The present invention relates to the field of safe production of sealing ring vulcanizing machines, and in particular to a safety detection system and method for sealing ring flat plate vulcanizing machines. Background Art

[0002] In the production of rubber sealing rings, flat-plate vulcanizers are key production equipment. They achieve high-temperature, high-pressure vulcanization molding by controlling the closing of upper and lower flat-plate molds through hydraulic drive rods. Traditional safety protection systems have significant technical flaws. First, the flat-plate molds weigh several tons, and the hydraulic drive rods have complex motion trajectories. Static protection devices such as fixed photoelectric fences or emergency stop buttons are difficult to adapt to the dynamic interaction requirements of human-machine interaction. Second, existing systems cannot accurately capture the real-time positional relationship between operator posture and heavy moving parts in three-dimensional space, resulting in safety protection blind spots. Third, static safety threshold settings lack flexibility. They cannot dynamically adjust based on individual operator characteristics such as height differences and operating habits, nor can they respond to changes in equipment operating status. False production shutdowns or delayed protection are common. More importantly, existing systems lack the ability to intelligently analyze operator behavior trajectories, unable to establish safety baselines based on historical compliance operation data. This makes it difficult to implement timely, graded warning measures in the event of sudden abnormal conditions. These problems have led to a high risk of mechanical extrusion accidents. The industry urgently needs an intelligent safety solution that integrates multi-dimensional perception, real-time trajectory analysis, and adaptive protection strategies. Summary of the Invention

[0003] The present invention addresses the shortcomings of the prior art in how to accurately prevent dangerous approaches between personnel and moving parts of equipment and balance safety and equipment operating efficiency in complex operation scenarios of flat-plate vulcanizers through multi-modal real-time detection and dynamic safety strategies, and provides a sealing ring flat-plate vulcanizer safety detection system and method.

[0004] In order to solve the above technical problems, the present invention is solved by the following technical solutions: a sealing ring flat vulcanizing press safety detection system, comprising:

[0005] A detection module for real-time monitoring of the three-dimensional position of personnel and a plate vulcanizer, wherein the plate vulcanizer includes a movable plate mold and a hydraulic drive rod, and the detection module includes a laser ranging unit, a pressure sensing unit, and a visual recognition unit;

[0006] a processing module electrically connected to the detection module, configured to receive in real time the three-dimensional position coordinates output by the detection module, generate real-time movement trajectories of personnel and moving parts of the flat vulcanizer, and generate graded warning signals by dynamically comparing the real-time movement trajectories with preset normal trajectories and comparing the real-time distance between the personnel and the equipment with a safety distance threshold;

[0007] The early warning execution module is electrically connected with the processing module, and is used for triggering the sound-light prompt unit, the automatic deceleration unit and the emergency braking unit in sequence according to the hierarchical early warning signal.

[0008] The self-learning module is electrically connected with the processing module, and is used for dynamically adjusting the safety distance threshold.

[0009] The laser ranging unit is configured with a laser radar, laser is emitted and reflected laser is received through the laser radar, and distance information of the personnel and the moving part of the flat vulcanizing machine is acquired;

[0010] The pressure sensing unit is configured with a pressure sensor array arranged on the ground around the equipment, two-dimensional plane position information is acquired by sensing ground pressure changes generated when the personnel stands or moves;

[0011] The visual recognition unit is configured with an industrial camera, image information of the personnel and the flat vulcanizing machine is acquired through the industrial camera, and the personnel posture, the position and the moving track of the moving part of the flat vulcanizing machine are recognized in combination with a deep learning algorithm.

[0012] By adopting the above technical scheme, through the multi-modal fusion detection scheme, global perception of the personnel position, posture and equipment moving part track is realized, the problem of a traditional single sensor detection blind area is effectively solved, the processing module generates a dynamic comparison model in real time and combines a safety distance dynamic calibration mechanism, so that the system has intelligent protection capability under complex working conditions, the false alarm rate is significantly reduced, and the dangerous response time is shortened.

[0013] The application is further provided that the trigger logic of the early warning execution module is:

[0014] When the processing module generates a hierarchical early warning signal, the sound-light prompt unit is triggered;

[0015] If the abnormal situation is not improved within a preset time, the automatic deceleration unit is triggered;

[0016] If the abnormal situation is still not eliminated or the distance between the personnel and the moving part of the flat vulcanizing machine reaches the safety distance threshold again after a preset time, the emergency braking unit is triggered to cut off the power supply of the flat vulcanizing machine and start the hydraulic system pressure maintaining device.

[0017] By adopting the above technical scheme, a three-level progressive early warning trigger logic is adopted to realize risk hierarchical response: the primary early warning gives an early warning through sound-light prompt, the secondary early warning reserves buffer time for personnel evacuation through automatic deceleration, and the senior early warning cuts off power and starts pressure maintaining through emergency braking, so that the response is more accurate than that of a traditional emergency stop scheme, and production interruption caused by unnecessary shutdown is avoided.

[0018] The application is further provided that the image recognition algorithm of the visual recognition unit includes:

[0019] A personnel target detection model based on deep learning technology is used to locate key body parts of an operator in real time.

[0020] A device component detection model is used to identify the opening angle of a flat mold and the extension length of a hydraulic driving rod.

[0021] The deep learning model is pre-trained through a vulcanizing machine operation scene data set, and the detection accuracy meets the industrial safety detection standard.

[0022] By adopting the above technical scheme, the dual detection model based on deep learning realizes high-precision real-time monitoring of key parts of an operator and the state of equipment, combines with a vulcanizing scene exclusive data set for training, the detection accuracy meets the industrial safety level requirement, effectively identifies complex human-machine interaction postures, and solves the recognition failure problem of a traditional visual system in a special environment.

[0023] The processing module generates a preset normal trajectory by continuously collecting compliance operation data within a preset time through a laser radar, a pressure sensor array and an industrial camera configured by the detection module, and generating a typical operation trajectory as a preset normal trajectory through a trajectory extraction algorithm.

[0024] By adopting the above technical scheme, a typical operation trajectory template is generated through multi-sensor data fusion, a dynamic reference model is established to make the abnormal detection have working condition adaptability, a variety of non-compliance operation trajectories can be efficiently identified, and the generalization ability is significantly improved compared with an artificial preset trajectory scheme.

[0025] The safety distance threshold is dynamically adjusted according to the operation habit of an operator and a working scene, and the optimal distance calculated by the model and the lower limit value specified by a safety standard are comprehensively considered in the adjustment process, and the threshold after dynamic adjustment is not less than a preset value of the lower limit value of the safety standard.

[0026] By adopting the above technical scheme, an adaptive safety threshold adjustment mechanism is constructed, the threshold is dynamically optimized through model calculation and safety standard double constraints, the safety distance can be flexibly adjusted under different risk scenes, the safety and the equipment running efficiency are balanced, and the problems of over-severe false stop or over-loose missed detection existing in a traditional fixed threshold are solved.

[0027] The flat vulcanizing machine is further provided with a two-hand operation starting device, the two-hand operation starting device includes two buttons, the personnel target detection model of the visual recognition unit is configured to: real-time identify the positions and action postures of the hands of an operator, and only when it is detected that the left hand and the right hand of the same operator press the two buttons respectively and the pressing time exceeds a preset safety trigger time length, a control signal allowing the equipment to start is generated.

[0028] The control signal also needs to be processed by the processing module to verify that the real-time distance between the current personnel and the moving parts of the equipment meets the safety distance threshold requirement, and then the start-up locking state of the flat vulcanizing press is released.

[0029] By adopting the above technical solutions, an innovative two-hand operation and visual recognition linkage control solution is designed. By real-time monitoring of the position and pressing status of both hands and combining it with safety distance verification, dual startup protection is formed to effectively prevent illegal startup behavior. Compared with the traditional two-hand button solution, it adds spatial position constraints and complies with relevant safety standards.

[0030] The present invention is further configured as follows: the self-learning module integrates the operator identity recognition function, identifies the operator's identity by collecting the current person's facial features through an industrial camera or the gait features of a pressure sensor array, establishes a personalized safety distance adjustment model based on the height, operating habits and historical safety data of different operators, and dynamically adjusts the dynamic configuration of the safety distance threshold.

[0031] By adopting the above technical solutions, the personalized identification function of operators is integrated, and a dedicated safety model is established based on individual characteristics to realize dynamic configuration of thresholds, solve the differentiated protection needs of people of different body shapes, reduce individual adaptability false alarms, and improve the safety of human-machine collaboration.

[0032] The present invention is further configured as follows: the processing module has a built-in risk assessment model, which calculates the dynamic risk coefficient and maps it to a graded warning signal based on the deviation between the real-time moving trajectory and the preset normal trajectory, the rate of change of the distance between personnel and key components of the equipment, and the acceleration parameters of the moving parts of the flat vulcanizer. The graded warning signal includes at least a first-level yellow warning, a second-level orange warning, and a third-level red warning.

[0033] By adopting the above technical solution, a built-in multi-parameter dynamic risk assessment model is established, which integrates parameters such as trajectory deviation, distance change rate and component acceleration to achieve quantitative mapping of risk levels. Compared with traditional alarm methods, it can more accurately identify the risk evolution process, predict dangerous trends in advance, and provide data support for safety decision-making.

[0034] The present invention is further configured as follows: the emergency braking unit of the early warning execution module is equipped with a dual redundant power-off device, including a hydraulic system power supply cut-off module and a control circuit safety relay, which cuts off the main power supply and control signal of the flat-plate vulcanizer when triggered at the same time. The flat-plate vulcanizer is equipped with a hydraulic pressure sensor, and the hydraulic system pressure is monitored in real time through the hydraulic pressure sensor. When the pressure holding device fails, the backup energy storage device is automatically started to maintain the clamping force.

[0035] By adopting the above technical solution, a dual redundant power-off protection design is adopted in conjunction with hydraulic system pressure monitoring and backup energy storage devices to ensure that the power supply and control signal are cut off simultaneously during emergency braking, and the clamping force is automatically maintained when the pressure holding fails, thus avoiding equipment failure and quality accidents and meeting high safety integrity level requirements.

[0036] The present invention is further provided as follows: a method for safety detection of a sealing ring flat vulcanizing machine, comprising the following steps:

[0037] S1: Dynamically compare the real-time movement trajectory with the preset normal trajectory established through self-learning based on historical compliance operation data;

[0038] S2: Calculate the real-time distance between the personnel and the key components of the equipment and compare it with the dynamic safety distance threshold output by the self-learning module;

[0039] S3: When an abnormal movement trajectory is detected or the real-time distance is less than the safety distance threshold, a graded warning signal is generated;

[0040] S4: Collect and store operator's operation behavior data and equipment operation status data;

[0041] S5: Build a model based on the operator's working habits to calculate the optimal safety distance under different operating scenarios;

[0042] S6: Set the safety distance threshold to no less than the lower limit specified by industry safety standards to ensure that the dynamically adjusted threshold meets basic safety requirements.

[0043] By adopting the above technical solutions, the entire process is intelligentized through a data closed-loop driven detection method, forming an adaptive system from trajectory comparison to threshold optimization. The collected operation data is used for model iteration, and the optimal safety distance is calculated in combination with the operating condition parameters and the safety lower limit constraint is enforced, so that the system has the ability to continuously evolve, and the safety performance is significantly improved in long-term operation.

[0044] Due to the adoption of the above technical solution, the present invention has significant technical effects: a sealing ring flat vulcanizer safety detection system and method provided by the present application solves the dynamic protection blind spots, inflexible static thresholds and early warning lag problems existing in traditional protection systems through three-dimensional spatial monitoring, dynamic trajectory comparison analysis and adaptive safety threshold adjustment of multi-sensor fusion, and has the advantages of real-time accurate monitoring, hierarchical active protection and personalized safety configuration. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a structural diagram of a safety detection system for a sealing ring flat vulcanizing machine; DETAILED DESCRIPTION

[0046] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0047] Example:

[0048] In the existing technology, the rubber sealing ring flat vulcanizing machine controls the closing of the upper and lower flat molds through a hydraulic drive rod to achieve high-temperature and high-pressure vulcanization molding. The traditional safety protection system uses a fixed photoelectric fence or emergency stop button, which cannot adapt to the complex human-machine dynamic interaction. Due to the complex motion trajectory of the flat mold and the hydraulic drive rod, the static safety threshold is difficult to capture the real-time position relationship between the posture of the person and the heavy moving parts in the three-dimensional space, and the protection strategy cannot be dynamically adjusted according to the individual characteristics of the operator and the operating status of the equipment, resulting in erroneous triggering of production stoppage or protection lag causing mechanical extrusion accidents.

[0049] In order to address the defect that traditional systems cannot perceive the human-computer interaction status in real time, it is necessary to establish a three-dimensional dynamic monitoring mechanism; to address the limitations of static safety thresholds, it is necessary to introduce adaptive adjustment algorithms; to address the shortcomings of a single protection method, it is necessary to build a hierarchical early warning system, and through multi-dimensional data fusion and dynamic analysis, form an intelligent safety protection plan that can adapt to different operating scenarios.

[0050] This application proposes a safety detection system including a detection module, a processing module, a warning execution module and a self-learning module. The detection module monitors the three-dimensional position of personnel and flat vulcanizers in real time, and includes a laser ranging unit, a pressure sensing unit and a visual recognition unit. The processing module generates a real-time moving trajectory and compares it with a preset normal trajectory, compares the real-time distance with the safety distance threshold to output a graded warning signal, and the warning execution module triggers sound and light prompts, automatic deceleration and emergency braking in sequence according to the signal. The self-learning module dynamically adjusts the safety distance threshold.

[0051] The laser ranging unit transmits and receives reflected lasers through the laser radar to obtain the distance information between personnel and moving parts of the equipment, solving the problem of insufficient three-dimensional spatial positioning accuracy. The pressure sensing unit senses the pressure changes when personnel move through the ground pressure sensor array, obtains two-dimensional plane position information, and makes up for the need for visual blind spot monitoring. The visual recognition unit uses industrial cameras combined with deep learning algorithms to identify personnel postures and equipment component positions to achieve dynamic trajectory capture. The processing module generates early warning signals through trajectory comparison and distance threshold judgment, forming a hierarchical response logic. The self-learning module dynamically optimizes safety parameters according to operation data to improve system adaptability.

[0052] The detection module collects personnel location and equipment status data in real time through multi-sensor fusion. The processing module comprehensively analyzes the three-dimensional distance data of the laser ranging unit, the planar positioning data of the pressure sensing unit and the trajectory recognition results of the visual recognition unit to generate the real-time movement trajectory of personnel and moving parts of the equipment. By comparing the deviation between the preset normal trajectory and the real-time trajectory, and combining the comparison results of the real-time distance and the dynamic safety threshold, a graded early warning signal is triggered. The early warning execution module activates sound and light prompts, equipment deceleration and emergency braking in sequence according to the signal level, forming a progressive protection mechanism. The self-learning module continuously optimizes the safety distance threshold based on historical operation data and real-time feedback to ensure that the protection strategy adapts to different operation scenarios.

[0053] Compared with existing technologies, traditional systems rely on fixed thresholds and single sensors and are unable to cope with dynamic human-computer interaction scenarios. This solution achieves precise three-dimensional space monitoring through multi-dimensional sensor fusion and dynamic trajectory analysis; adjusts safety thresholds through self-learning algorithms to adapt to different operators and production environments; and balances safety protection and production efficiency through a graded early warning mechanism to avoid excessive intervention.

[0054] This application solves the problems of false triggering and protection lag in traditional systems, realizes real-time three-dimensional spatial monitoring of personnel and moving parts of heavy equipment, dynamically adjusts safety protection strategies, effectively reduces the risk of mechanical extrusion accidents, and reduces the impact of unnecessary downtime on production efficiency.

[0055] The triggering logic of the early warning execution module is as follows: when the processing module generates a graded early warning signal, the sound and light prompt unit is triggered; if the abnormal situation does not improve within the preset time, the automatic deceleration unit is triggered; if the abnormal situation is still not resolved after the preset time or the distance between the personnel and the moving parts of the flat vulcanizer reaches the safety distance threshold, the emergency braking unit is triggered to cut off the power supply of the flat vulcanizer and start the hydraulic system pressure holding device.

[0056] A graded warning signal refers to a safety alarm signal with different danger levels generated based on real-time detection data. It can be implemented by a multi-level logic judgment algorithm to distinguish abnormal conditions of different degrees of urgency. The preset time refers to the pre-set time window that allows operators to respond to the alarm. It can be implemented by a programmable timer module, for example, set to a range of 2 seconds to 5 seconds. The emergency brake unit refers to a safety device used to cut off equipment power and maintain system pressure. It can be implemented by a dual relay control circuit working in conjunction with a hydraulic pressure-maintaining valve group to ensure that the mold remains stable after power is interrupted. The safety distance threshold refers to the minimum allowable distance for triggering the final protective action. It can be calculated through a dynamic risk assessment model combined with the equipment movement speed parameters.

[0057] When an abnormal distance between a person and a moving part of the equipment is detected, the system first activates the sound and light warning device to remind the operator to evacuate; if the safe distance is not restored within the set time range, the automatic deceleration function is activated to reduce the equipment's movement speed; if the risk of collision still exists after deceleration or the distance continues to shrink to a critical value, the main power supply is immediately cut off and hydraulic pressure holding is started to prevent the mold from shifting due to sudden power failure. This logic uses a staged response mechanism to minimize unnecessary downtime while ensuring safety.

[0058] Compared with existing technologies, traditional safety systems mostly use a single emergency stop triggering method, which cannot distinguish risk levels and lacks a buffer processing process, making it easy for production to be interrupted due to false triggering. This solution uses a graded response mechanism to give operators the opportunity to make corrections in the primary warning stage, reduce the potential injury intensity by deceleration in the intermediate stage, and implement precise braking in the final stage, which not only ensures personnel safety but also reduces unplanned downtime losses.

[0059] It effectively solves the problem of over-response or delayed response of traditional protection systems. By handling abnormal conditions in stages, the probability of false triggering is reduced while ensuring safety. The dynamically adjusted response threshold can adapt to different working conditions. The double power-off mechanism ensures the reliability of emergency braking. The hydraulic pressure holding device avoids the risk of mold displacement caused by traditional emergency stops, forming a complete safety protection closed loop.

[0060] The image recognition algorithm of the visual recognition unit includes a personnel target detection model and an equipment component detection model. The personnel target detection model locates the key body parts of the operator in real time based on deep learning technology. The equipment component detection model identifies the opening and closing angles of the flat mold and the extension length of the hydraulic drive rod. The deep learning model is pre-trained using a vulcanizer operation scene data set, and the detection accuracy meets industrial safety testing standards.

[0061] The personnel target detection model refers to a target recognition framework constructed through a convolutional neural network. Specifically, it can be implemented using the YOLOv5 architecture combined with the human key point detection algorithm. It is used to capture the coordinate positions of key parts such as the operator's head and hands in real time in complex industrial scenarios. The equipment component detection model refers to a feature extraction model for the moving parts of a flat vulcanizer. Specifically, it can be implemented using the Mask R-CNN instance segmentation algorithm. The opening and closing angles and extension lengths are calculated by identifying the mold edge contours and the displacement of the hydraulic rod. The vulcanizer operation scene dataset refers to a collection of images containing multi-angle equipment operation status and personnel operation behaviors. Specifically, it can be constructed by continuously collecting equipment working status and personnel position data under different lighting conditions through industrial cameras, which is used to improve the model's generalization ability in a real production environment.

[0062] The visual recognition unit uses industrial cameras to capture live video streams in real time. The human target detection model analyzes the operator's body posture frame by frame and locates the spatial relationship between his or her hands and the equipment's dangerous areas. The equipment component detection model simultaneously analyzes the metal reflective edge features of the flat mold and calculates motion parameters based on the displacement trajectory of the hydraulic drive rod. The detection data output by the two models is input into the processing module after coordinate conversion and compared with the preset safe operating parameters. Before deployment, the deep learning model must be trained on a data set containing typical illegal operation samples, such as scenarios where a person approaches the mold area too close or the hydraulic rod extends abnormally, to ensure that the detection results meet the detection accuracy requirements of the GB 27607-2011 machinery safety standard.

[0063] Compared with existing technologies, traditional safety systems use fixed photoelectric sensors or two-dimensional positioning technology, which cannot accurately identify the relationship between a person's body posture and three-dimensional spatial position. This solution, by integrating deep learning target detection and instance segmentation technology, can dynamically capture the spatial relationship between the operator's hand movements and heavy moving parts, while accurately quantifying key parameters such as the mold opening and closing angle, solving the problem of misjudgment caused by equipment reflection and personnel obstruction in traditional methods.

[0064] It achieves simultaneous and accurate recognition of the operator's three-dimensional posture and the equipment's motion status, effectively avoiding the protection lag problem caused by visual detection blind spots. By learning typical illegal operation scenarios through pre-training models, it significantly improves the detection stability under interference conditions such as complex lighting and equipment reflections, ensuring that the safety judgment results meet the requirements of industrial detection standards.

[0065] The processing module generates a preset normal trajectory by using the laser radar, pressure sensor array, and industrial camera configured in the detection module to continuously collect compliance operation data within a preset time, and generates a typical operation trajectory as the preset normal trajectory through a trajectory extraction algorithm.

[0066] Compliance operation data refers to historical operation records that comply with safety operation specifications. Specifically, it can be implemented using a combined data set of laser ranging data, pressure distribution data, and image sequences. It is used to characterize the spatial position relationship between equipment and personnel in a safe state. The trajectory extraction algorithm refers to a mathematical method for extracting typical motion patterns from multi-source sensor data. Specifically, it can be implemented by combining cluster analysis with time series modeling. A baseline trajectory model is established by identifying recurring operation path features.

[0067] In the device initialization phase, the laser radar continuously scans the three-dimensional motion path of the flat mold and the hydraulic rod, the pressure sensor array records the safety station area of the operator, and the industrial camera synchronously captures the device running state and personnel action. When the compliance operation data accumulation reaches the preset time (for example, continuously collecting 8 hours of production shift data), the processing module performs spatio-temporal alignment on the multi-source data, selects the highest repetition rate device motion trajectory and personnel activity path through the trajectory extraction algorithm, and quantizes the typical operation trajectory into a trajectory template containing timestamp, coordinate point and velocity vector, which is used as a reference frame for subsequent real-time trajectory comparison.

[0068] Compared with the prior art, the traditional safety system uses a fixed threshold or manually preset trajectory template, which cannot adapt to the behavior differences of different operators and the fluctuations of device running parameters. The present scheme establishes a dynamic reference by collecting compliance operation data in the actual production environment, which can accurately reflect the normal operation mode under the combination of a specific device and personnel, and solves the problem of false judgment caused by mismatch between the preset trajectory and the actual working condition.

[0069] The preset normal trajectory matching the current operation scene can be automatically generated according to the actual production data, so that the safety detection system has dynamic adaptability. After personnel rotation or device maintenance, the system updates the trajectory template by re-collecting compliance data, effectively avoiding safety false judgment caused by changes in personnel operation habits or device state drift, and significantly improving the accuracy of trajectory comparison detection.

[0070] The safety distance threshold is dynamically adjusted according to the operation personnel habits and working scene, and the optimal distance calculated by the model and the lower limit value specified by the safety standard are considered in the adjustment process. The dynamically adjusted threshold is not less than the preset value of the lower limit value of the safety standard.

[0071] The safety distance threshold refers to the dynamic parameter used to judge the danger level between personnel and device moving parts. It can be calculated by a machine learning model combined with real-time sensor data. An adaptive adjustment mechanism is established by analyzing the historical behavior data of the operator and the device running state. The optimal distance refers to the personalized safety distance calculated based on the height of the operator, the action habit and the motion characteristics of the device. It can be realized by a three-dimensional posture capture system and a device motion trajectory prediction algorithm to balance safety and work efficiency in different operation scenes. The lower limit value of the safety standard refers to the minimum safety distance required by the industry specification. It can be verified by a real-time call to the stored safety standard database through an embedded safety controller to ensure that the dynamically adjusted threshold always meets the regulatory requirements.

[0072] During the operation of the vulcanizer, the self-learning module continuously collects the operator's standing position distribution, movement speed and acceleration data of the equipment's moving parts, and generates a personalized safety distance recommendation value through the trajectory prediction model. The recommended value is compared with the lower limit of the safety standard in real time. When the recommended value is lower than the lower limit, the system automatically adopts the lower limit as the actual effective threshold; when the recommended value is higher than the lower limit, the system uses the recommended value as the dynamic threshold. For example, for operators with taller height, the system can appropriately increase the safety distance threshold to match their limb movement range, while ensuring that the adjusted threshold is not lower than the minimum value specified by the industry standard.

[0073] Compared with existing technologies, traditional safety systems use fixed thresholds that cannot adapt to the height differences of different operators and dynamic work scenarios. This solution, by integrating personalized behavioral data with safety standard constraints, not only solves the problem of false triggering caused by static thresholds, but also avoids the safety hazards caused by excessive relaxation of thresholds.

[0074] It effectively solves the problem of insufficient adaptability of traditional safety protection systems in complex human-computer interaction scenarios, realizes the dynamic optimization configuration of safety distance thresholds, and significantly improves the matching accuracy between protection strategies and the actual working modes of operators while ensuring compliance with industry safety regulations.

[0075] The flatbed vulcanizer is equipped with a two-handed operation starting device, which includes two buttons. The personnel target detection model of the visual recognition unit is configured to identify the operator's hand position and movement posture in real time. Only when it is detected that the left and right hands of the same operator are respectively pressing the two buttons and the pressing time exceeds the preset safety trigger time, a control signal allowing the equipment to start is generated; the control signal must also be verified by the processing module. When the real-time distance between the current person and the moving parts of the equipment meets the safety distance threshold requirement, the start lock state of the flatbed vulcanizer is released.

[0076] A two-handed operation starting device refers to a physical trigger device that requires the operator to operate both hands at the same time. It can be implemented by a two-button switch with an independent circuit. The risk of accidental touch with one hand is eliminated through synchronous operation of both hands. The preset safety trigger time refers to the shortest time that the button must be pressed continuously before the device is started. It can be implemented by a programmable timer module. The instantaneous false touch signal is filtered through the time threshold. The personnel target detection model of the visual recognition unit refers to a limb positioning algorithm based on image analysis. It can be implemented by a convolutional neural network combined with bone key point detection technology. The operation intention is judged by identifying the hand joint coordinates. The safety distance threshold verification refers to comparing the real-time monitoring data with the dynamically adjusted critical value. It can be implemented by three-dimensional coordinate difference calculation combined with a threshold comparator. Personal safety is ensured through spatial distance judgment.

[0077] During the equipment startup phase, the industrial camera continuously captures images of the operating area and locates the operator's hands through a skeletal key point detection algorithm. When it detects that the left and right hands cover the two button areas respectively, the timer is triggered to start accumulating the pressing time. If the pressing duration exceeds the preset value, a start request signal is generated and transmitted to the processing module. The processing module synchronously calls the laser ranging unit to obtain the real-time distance data between the operator and the flat mold, and compares the measured value with the current safety distance threshold. Only when the distance data meets the safety requirements and the two-hand pressing conditions are met at the same time, an unlocking command is sent to the equipment control system to allow the hydraulic drive rod to execute the mold closing action.

[0078] Compared with existing technologies, traditional safety systems rely on a single physical button or fixed-position sensor and are unable to verify the spatial relationship between the operator's body posture and the moving parts of the equipment. This solution uses the dual conditional constraints of visual recognition and distance verification to prevent false start-ups caused by non-standard operations and avoid operators triggering equipment movement in dangerous areas, effectively solving the three-dimensional spatial safety blind spot problem of traditional protective devices.

[0079] A collaborative verification mechanism for two-handed operation and safe distance has been implemented to ensure that the operator is in a safe area and maintains a standard operating posture when the equipment is started. The risk of mechanical extrusion caused by accidental touch or improper standing position is eliminated from the two dimensions of spatial position and operating behavior, significantly improving the safety of the operation process of heavy-duty vulcanizing equipment.

[0080] The self-learning module integrates the operator identification function. It identifies the operator's identity by collecting the current person's facial features through industrial cameras or the gait features of the pressure sensor array. Based on the height, operating habits and historical safety data of different operators, it establishes a personalized safety distance adjustment model and dynamically adjusts the dynamic configuration of the safety distance threshold.

[0081] The operator identity recognition function refers to a technical means of distinguishing different operators through biometric or behavioral characteristics. Specifically, it can use facial images captured by industrial cameras for feature extraction, or gait pressure distribution patterns collected by pressure sensor arrays for identity matching, thereby achieving rapid confirmation of the operator's identity. The personalized safety distance adjustment model refers to a dynamic threshold calculation model established based on individual differences of operators. Specifically, it can generate dynamic safety distance parameters adapted to individual characteristics by collecting height data, historical operation trajectory data and safety event records of different operators, combined with machine learning algorithm training, thereby solving the problem that traditional fixed thresholds cannot adapt to individual differences.

[0082] During the operation of the flatbed vulcanizer, the industrial camera continuously collects image information in the operating area, extracts the operator's facial features through a face recognition algorithm, or captures the operator's gait pressure distribution pattern when walking through a pressure sensor array to complete identity authentication. After identity confirmation, the system retrieves the operator's historical operation data, such as height parameters, typical standing areas, and past safety distance adjustment records. Combined with the current equipment operating status, the machine learning model is used to calculate a personalized safety distance threshold. For example, for operators with taller heights, the system can appropriately increase the safety distance threshold to match their limb movement range; for people who are accustomed to performing delicate operations close to the equipment, the system can dynamically optimize the threshold range based on their historical safety data.

[0083] Compared with existing technologies, traditional safety systems use a unified and fixed safety distance threshold, which cannot adapt to the height differences and operating habits of different operators. It is easy to cause false alarms or protection delays due to unreasonable threshold settings. This solution dynamically adjusts the safety distance through identity recognition and personalized models, so that the safety protection parameters can accurately match the individual characteristics of the operator, avoiding false triggering and shutdown caused by too small a threshold, and can timely adjust the protection strategy according to the operator's behavior pattern, significantly improving the effectiveness of safety protection.

[0084] It can provide customized safety protection based on the physical conditions and operating habits of different operators, effectively reducing the risk of mechanical extrusion caused by unreasonable safety distance threshold settings, while reducing unplanned shutdowns caused by false triggering, and improving the safety and production efficiency of vulcanizing press operations.

[0085] The processing module has a built-in risk assessment model. Based on the deviation between the real-time moving trajectory and the preset normal trajectory, the rate of change of the distance between personnel and key components of the equipment, and the acceleration parameters of the moving parts of the flat vulcanizer, it calculates the dynamic risk coefficient and maps it to a graded warning signal. The graded warning signal includes at least a level one yellow warning, a level two orange warning, and a level three red warning.

[0086] The dynamic risk coefficient refers to a quantitative risk indicator calculated by the fusion of multi-dimensional parameters. Specifically, it can be achieved by using a weighted algorithm to comprehensively calculate the trajectory deviation, distance change rate and acceleration parameters. It is used to characterize the probability level of a safety accident in the current operating scenario. The deviation refers to the spatial offset between the real-time moving trajectory and the preset normal trajectory. Specifically, it can be achieved by using a trajectory similarity algorithm to calculate the Euclidean distance difference of the trajectory point set. It is used to determine whether the equipment movement complies with safety operating specifications. The distance change rate refers to the instantaneous change speed of the real-time distance between personnel and key components of the equipment. Specifically, it can be achieved by using differential calculations to calculate the distance difference between adjacent time points. It is used to identify sudden behaviors of personnel accidentally approaching the equipment. The acceleration parameter refers to the motion state parameter of the moving parts of the flat vulcanizer during the hydraulic drive process. Specifically, it can be achieved by collecting data through an acceleration sensor installed on the hydraulic drive rod. It is used to determine whether the equipment has the risk of abnormal acceleration or loss of control.

[0087] The processing module uses a laser ranging unit, a pressure sensor array, and a visual recognition unit to collect real-time data on personnel positions, equipment motion trajectories, and acceleration. The trajectory deviation, distance change rate, and acceleration parameters are input into the risk assessment model for fusion calculation. When the trajectory deviation exceeds a preset threshold, the model generates a dynamic risk coefficient based on the deviation amplitude and the influencing factor of the distance change rate, combined with the abnormal fluctuation characteristics of the equipment acceleration. This coefficient is converted into a three-level warning signal through preset mapping rules: a yellow warning triggers an audible and visual prompt, an orange warning initiates automatic deceleration, and a red warning activates emergency braking.

[0088] Compared with existing technologies, traditional safety systems rely solely on a single distance threshold for static judgment and are unable to identify complex risk factors such as trajectory deviation and motion mutation. This solution uses dynamic risk assessment based on multi-dimensional parameter fusion to identify in advance the associated risks of abnormal personnel movement trajectories and equipment motion status, and implement graded responses when equipment acceleration is abnormal or personnel approach unexpectedly, effectively avoiding misjudgments or delayed responses caused by a single threshold.

[0089] It solves the technical problem that traditional safety systems are unable to dynamically evaluate complex risk scenarios, and realizes hierarchical early warning control based on multi-dimensional parameter fusion. By calculating the associated risk of trajectory deviation and equipment acceleration in real time, it can predict potential collision risks before personnel enter the dangerous area. At the same time, it combines the distance change rate to identify sudden dangerous actions, significantly improving the accuracy of early warning signals and response timeliness, and reducing the incidence of unplanned shutdowns or safety accidents due to misjudgment.

[0090] The emergency brake unit of the sealing ring flat plate vulcanizer safety detection system is equipped with a dual redundant power-off device, including a hydraulic system power supply cut-off module and a control circuit safety relay. When triggered, they cut off the main power supply and control signal of the flat plate vulcanizer. The flat plate vulcanizer is equipped with a hydraulic pressure sensor, and the hydraulic system pressure is monitored in real time through the hydraulic pressure sensor. When the pressure holding device fails, the backup energy storage device is automatically started to maintain the clamping force.

[0091] A dual redundant power-off device refers to a device that includes two sets of independently operating power-off devices. For example, the hydraulic system power supply cut-off module uses an electromagnetic circuit breaker, and the control circuit safety relay uses a solid-state relay. The two sets of devices are connected to the main circuit in parallel. When any set of devices is triggered, the power can be cut off. The hydraulic pressure sensor refers to a pressure transmitter installed in the hydraulic oil circuit, such as a piezoresistive sensor, which collects oil pressure data in the hydraulic cylinder in real time and transmits it to the processing module through an analog signal. The backup energy storage device refers to an emergency boosting unit with a compressed gas storage tank, such as a nitrogen cylinder connected to a boosting pump, which automatically releases gas to replenish the system pressure when the pressure of the pressure maintaining device drops.

[0092] When the processing module determines that emergency braking needs to be triggered, the system will synchronously activate the hydraulic system power supply cut-off module and the control circuit safety relay. For example, the electromagnetic circuit breaker cuts off the three-phase AC power input, and the safety relay disconnects the PLC control signal output. At this time, the hydraulic pressure sensor continuously monitors the clamping cylinder pressure. If it is detected that the pressure holding device has dropped to a critical value due to pipeline leakage, the backup energy storage device will be immediately activated. For example, the solenoid valve is used to control the nitrogen bottle to inject high-pressure gas into the oil circuit to maintain the clamping force required for mold closing.

[0093] In some specific embodiments, the hydraulic system power supply cut-off module may adopt a circuit breaker with a mechanical interlocking structure, the control circuit safety relay may be configured with a dual-contact redundant design, and the starting threshold of the backup energy storage device may be set to a value 10% lower than the normal operating pressure. For example, when the system pressure drops from 20MPa to 18MPa, the gas replenishment action is triggered.

[0094] Compared with existing technologies, traditional safety systems rely solely on a single power cut-off device, which poses the risk of brake failure due to contact adhesion or coil burning. The dual redundant design significantly reduces the probability of single-point failure through two sets of independent actuators. Existing pressure-maintaining devices cannot maintain pressure in the event of a sudden leak, which can easily cause the mold to loosen and result in product scrapping. The backup energy storage device effectively avoids such problems through real-time pressure monitoring and gas compensation mechanisms.

[0095] Solve the risk of power remaining that may occur in traditional safety system during emergency power-off, ensure the equipment completely stop moving; at the same time, prevent the mold displacement caused by hydraulic system pressure loss, avoid the workpiece deformation or personnel injury accident caused by insufficient clamping force.

[0096] A safety detection method for sealing ring flat vulcanizer, comprising the following steps: dynamically comparing real-time moving track with preset normal track established based on historical compliance operation data self-learning; calculating real-time distance between personnel and equipment key components, and comparing with dynamic safety distance threshold value output by self-learning module; generating graded early warning signal when detecting abnormal moving track or real-time distance less than safety distance threshold value; collecting and storing operation personnel's operation behavior data and equipment running state data; establishing model based on operation personnel's working habit, calculating optimal safety distance under different operation scenes; setting safety distance threshold value not less than lower limit value specified by industry safety standard, ensuring that the dynamically adjusted threshold value meets the basic safety requirement.

[0097] Real-time moving track refers to the three-dimensional position change path of personnel and equipment moving parts obtained in real time through laser ranging unit, pressure sensing unit and visual recognition unit, which can be realized by laser radar scanning combined with image recognition algorithm, and is used to capture the spatial dynamics of human-machine interaction in operation process; dynamic safety distance threshold value refers to the minimum safety distance adjusted in real time according to operation personnel's height, operation habit and equipment running state, which can be realized by self-learning module to analyze historical operation data and establish individualized model, and is used to adapt to the protection needs under different operation scenes; graded early warning signal refers to the warning level divided according to track deviation degree and distance threshold value overrun, which can be generated by risk assessment model to calculate dynamic risk coefficient, and is used to trigger phased safety response mechanism.

[0098] During equipment operation, personnel position and equipment movement state data are continuously collected through multi-sensor fusion technology, real-time track is matched with reference track formed by historical compliance operation, when track deviation or insufficient safety distance is detected, the system triggers sound and light warning, equipment speed reduction and emergency braking measures according to risk level, at the same time, operation personnel's behavior data is continuously recorded and input into machine learning model, through analyzing operation characteristics under different working scenes, safety distance parameters are automatically optimized, ensuring that the adjusted threshold value meets both individual operation habit and industry safety specification.

[0099] Compared with existing technologies, traditional methods use fixed safety thresholds that cannot adapt to dynamic human-computer interaction scenarios. This solution solves the problems of high false triggering rate and delayed response of static protection systems through real-time trajectory comparison and adaptive threshold adjustment. In existing technologies, the safety distance setting only relies on the lower limit of industry standards, while this solution combines personalized operation data with standard requirements for dynamic balance, improving protection accuracy while ensuring basic safety.

[0100] It effectively reduces the probability of incorrect shutdowns caused by differences in personnel operating habits. At the same time, it shortens the response time to abnormal conditions through a graded early warning mechanism. By dynamically adjusting the safety distance threshold, it avoids the safety hazards caused by traditional systems with too small thresholds and overcomes the problem of reduced equipment utilization caused by too large thresholds. The self-learning mechanism enables the system to adapt to the working modes of different operators and achieve precise protection in complex production scenarios.

Claims

1. A sealing ring flat vulcanizing press safety detection system, comprising: A detection module for real-time monitoring of the three-dimensional position of personnel and a plate vulcanizer, wherein the plate vulcanizer includes a movable plate mold and a hydraulic drive rod, and the detection module includes a laser ranging unit, a pressure sensing unit, and a visual recognition unit; a processing module electrically connected to the detection module, configured to receive in real time the three-dimensional position coordinates output by the detection module, generate real-time movement trajectories of personnel and moving parts of the flat vulcanizer, and generate graded warning signals by dynamically comparing the real-time movement trajectories with preset normal trajectories and comparing the real-time distance between the personnel and the equipment with a safety distance threshold; The warning execution module is electrically connected to the processing module and is used to trigger the sound and light prompt unit, the automatic deceleration unit and the emergency braking unit in sequence according to the graded warning signal. The self-learning module is electrically connected to the processing module and is used to dynamically adjust the safety distance threshold. The laser ranging unit is equipped with a laser radar, which emits laser and receives reflected laser to obtain the distance information between the personnel and the moving parts of the flat vulcanizer; The pressure sensing unit is configured with a pressure sensor array arranged on the ground around the device, which obtains two-dimensional plane position information by sensing the ground pressure changes caused by a person standing or moving; The visual recognition unit is equipped with an industrial camera, which collects image information of personnel and the flat-plate vulcanizer through the industrial camera, and combines the deep learning algorithm to identify the personnel's posture, the position and movement trajectory of the moving parts of the flat-plate vulcanizer.

2. A sealing ring flat vulcanizing press safety detection system according to claim 1, characterized in that: The triggering logic of the early warning execution module is: When the processing module generates a graded warning signal, the sound and light prompt unit is triggered; If the abnormal situation does not improve within the preset time, the automatic deceleration unit is triggered; If the abnormal situation is not resolved after the preset time or the distance between the personnel and the moving parts of the plate vulcanizer reaches the safety distance threshold, the emergency brake unit is triggered to cut off the power supply of the plate vulcanizer and start the hydraulic system pressure holding device.

3. A sealing ring flat vulcanizing press safety detection system according to claim 1, characterized in that: The image recognition algorithm of the visual recognition unit includes: Human target detection model, which uses deep learning technology to locate key body parts of operators in real time; Equipment component detection model to identify the opening and closing angle of the flat mold and the extension length of the hydraulic drive rod; The deep learning model is pre-trained using a vulcanizer operation scenario dataset, and its detection accuracy meets industrial safety testing standards.

4. A sealing ring flat vulcanizing press safety detection system according to claim 1, characterized in that: The method for the processing module to generate a preset normal trajectory is: the laser radar, pressure sensor array and industrial camera configured in the detection module continuously collect compliance operation data within a preset time, and generate a typical operation trajectory as the preset normal trajectory through a trajectory extraction algorithm.

5. The sealing ring flat vulcanizing press safety detection system according to claim 1, characterized in that: The safety distance threshold is dynamically adjusted according to the operator's habits and work scenarios. During the adjustment process, the optimal distance calculated by the model and the lower limit value specified by the safety standard are comprehensively considered. The dynamically adjusted threshold value is not lower than the preset value of the lower limit value of the safety standard.

6. A sealing ring flat vulcanizing press safety detection system according to claim 1, characterized in that: The flatbed vulcanizing press is equipped with a two-handed start-up device, which includes two buttons. The human target detection model of the visual recognition unit is configured to: identify the position and movement posture of the operator's hands in real time, and generate a control signal allowing the equipment to start only when it is detected that the operator's left and right hands are respectively pressing the two buttons and the pressing time exceeds the preset safety trigger time. The control signal also needs to be processed by the processing module to verify that the real-time distance between the current personnel and the moving parts of the equipment meets the safety distance threshold requirement, and then the start-up locking state of the flat vulcanizing press is released.

7. The sealing ring flat vulcanizing press safety detection system according to claim 1, characterized in that: The self-learning module integrates the operator identification function, and identifies the operator's identity by collecting the current person's facial features through industrial cameras or the gait features of the pressure sensor array. Based on the height, operating habits and historical safety data of different operators, a personalized safety distance adjustment model is established to dynamically adjust the dynamic configuration of the safety distance threshold.

8. The sealing ring flat vulcanizing press safety detection system according to claim 1, characterized in that: The processing module has a built-in risk assessment model that calculates the dynamic risk coefficient and maps it to a graded warning signal based on the deviation between the real-time movement trajectory and the preset normal trajectory, the rate of change of the distance between personnel and key components of the equipment, and the acceleration parameters of the moving parts of the flat vulcanizer. The graded warning signal includes at least a level one yellow warning, a level two orange warning, and a level three red warning.

9. The sealing ring flat vulcanizing press safety detection system according to claim 2, characterized in that: The emergency braking unit of the early warning execution module is equipped with a dual redundant power-off device, including a hydraulic system power supply cut-off module and a control circuit safety relay, which cuts off the main power supply and control signal of the flat-plate vulcanizer when triggered at the same time. The flat-plate vulcanizer is equipped with a hydraulic pressure sensor, and the hydraulic system pressure is monitored in real time through the hydraulic pressure sensor. When the pressure holding device fails, the backup energy storage device is automatically started to maintain the clamping force.

10. A sealing ring vulcanizing machine safety detection method, according to a sealing ring vulcanizing machine safety detection system according to any one of claims 1 to 9, characterized in that: The following steps are involved: S1: Dynamically compare the real-time movement trajectory with the preset normal trajectory established through self-learning based on historical compliance operation data; S2: Calculate the real-time distance between the personnel and the key components of the equipment and compare it with the dynamic safety distance threshold output by the self-learning module; S3: When an abnormal movement trajectory is detected or the real-time distance is less than the safety distance threshold, a graded warning signal is generated; S4: Collect and store operator's operation behavior data and equipment operation status data; S5: Build a model based on the operator's working habits to calculate the optimal safety distance under different operating scenarios; S6: Set the safety distance threshold to no less than the lower limit specified by industry safety standards to ensure that the dynamically adjusted threshold meets basic safety requirements.

Images