Automatic bag removing device and method

Abstract

The invention relates to an automatic package removing device, and belongs to the technical field of package detection equipment. The device comprises a horizontal conveyor belt, a metal detection unit, a weight detection unit, a correlation grating, a servo push plate mechanism and a control system, the metal detection unit detects metal impurities through a metal detector above a conveying belt, the weight detection unit obtains weight data through a weighing platform below the conveying belt, a built-in encoder of a correlation grating collects the speed of the conveying belt in real time, and the servo push plate mechanism drives a push plate through a servo motor and monitors thrust through a pressure sensor. And the control system receives the detection signal, calculates the starting delay time of the push plate based on the speed of the conveyor belt and the conveying distance, and drives the servo motor to act according to preset parameters, so that accurate removal of metal, weight and composite defective packaging materials is realized. The device is mainly used for automatically detecting and removing unqualified packages on a packaging production line, and the packaging quality control efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of packaging detection equipment. More specifically, the present invention relates to an automatic bag rejection device and method. Background Art

[0002] During the packaging production process, packages may become defective due to problems such as metal impurities being mixed in or the weight not meeting the standard, and need to be removed from the production line through a rejection device. Traditional packaging rejection devices are usually designed only for a single type of defect (such as metal detection or weight detection), and it is difficult to handle packages with multiple types of defects simultaneously, resulting in the need to configure multiple sets of independent detection equipment on the production line, increasing equipment costs and installation space. In addition, in terms of the control of the rejection position of existing devices, most rely on fixed mechanical limits and are sorted manually after being pushed out by a push plate. There are many drawbacks to this type of method: on the one hand, the fixed mechanical limit device needs to be complexly installed and debugged in advance according to parameters such as the conveyor belt speed and the size of the package. Once packages of different specifications are replaced, if a statically set thrust is used, it is easy to cause the position of the package to shift when it is pushed out and it cannot accurately fall into the collection area. On the other hand, manual sorting after the push plate is pushed out not only increases labor costs, but also the sorting efficiency is greatly affected by factors such as the operator's proficiency and fatigue, and it is prone to misjudgment or missed inspection, resulting in defective products being mixed into qualified products and affecting product quality. In addition, the manual sorting method cannot be efficiently connected to the automated production line, hindering the intelligent upgrade of the entire production process and making it difficult to meet the requirements of modern manufacturing for high-efficiency and precise production.

[0003] In summary, there is an urgent need for an automatic bag rejection device and method that can accurately control the rejection of packages with multiple types of defects and has an adaptive adjustment function. Summary of the Invention

[0004] The present invention provides an automatic bag rejection device and method, which can achieve the accurate rejection of packages with metal, weight, and composite defects, and is used for the automated detection and rejection of unqualified packages on the packaging production line, improving the efficiency of packaging quality control.

[0005] To achieve these and other advantages in accordance with the present invention, an automatic bag rejection device is provided, including a horizontally arranged conveyor belt, and a metal detection unit, a weight detection unit, an opposed light grating, and a servo push plate mechanism are sequentially arranged in the conveying direction of the conveyor belt; The metal detection unit includes a metal detector fixedly installed above the conveyor belt. The weight detection unit includes a weighing platform arranged below the conveyor belt. The weighing platform is fixedly connected to the supporting surface of the conveyor belt. The opposed light grating is fixed above the conveyor belt, and an in-built encoder collects the conveyor belt speed in real time. The servo pusher mechanism includes a pusher driven by a servo motor and a pressure sensor. The plate surface of the pusher is perpendicular to the conveying direction of the conveyor belt. The detection threshold of the pressure sensor is 50 - 100 N. The device further includes a control system. The control system is electrically connected to the metal detection unit, the weight detection unit, the encoder, the servo motor, and the pressure sensor respectively. The control system receives the metal detection signal output by the metal detection unit and the weight difference signal output by the weight detection unit, calculates the pusher start delay time of metal packaging, weight packaging, or composite packaging based on the conveyor belt linear speed and the conveying distance, and drives the servo motor to act according to the preset thrust parameters.

[0006] Preferably, the control system has a built-in time-position conversion model, and calculates the packaging position coordinates in real time through encoder pulse counting. The calculation formula is x = x 0 + v ×( t - t 0 ) where x 0 is the position of the opposed light grating from the starting point of the conveyor belt, v is the conveyor belt linear speed, t 0 is the grating trigger timestamp; For metal packaging, weight packaging, or composite packaging, the control system presets three types of rejection station coordinates: the composite station is the first distance downstream of the opposed light grating; the metal station is the second distance downstream of the opposed light grating, and the second distance is greater than the first distance. The weight station is the third distance downstream of the opposed light grating, and the third distance is greater than the second distance. The metal station is the midpoint of the weight station and the composite station; Through the encoder pulse difference Δ N the station spacing Δ L = Δ N ×0.1 mm is calculated to ensure that the front end of the packaging reaches within ±5 mm of the corresponding station when the pusher starts.

[0007] Preferably, the control system has a built-in thrust adaptive algorithm: When detecting metal packaging, the pressure sensor monitors and maintains the thrust F m at 70 - 80 N in real time, and according to the formula Fm修正 = F m × K 1 Dynamic correction, where K 1 is a dynamic compensation coefficient of 0.95 - 1.05, calculated from the friction coefficient μ corresponding to the chute inclination angle of 15°. When exceeding this range, the servo motor torque is automatically adjusted through the PID algorithm to make S 实际 = S 目标 ± 15 mm; When detecting a weight-type packaging, the pressure sensor monitors and maintains the thrust in real time F w at 50 - 60 N and dynamically corrects according to the weight data fed back by the weighing platform. The correction formula is F w修正 = F w ×( m 实际 / m 标准 )× K 1 Dynamic correction, where m 实际 is the measured weight of the packaging; When detecting a composite-type packaging, the pressure sensor monitors and maintains the thrust F m&w at 60 - 70 N, combines the weight data fed back by the weighing platform, and dynamically corrects the thrust using the weighted average algorithm, F m&w修正 =( F m × α + F w ×(1 − α ))×( m 实际 / m 标准 )× K 1 , where the weight coefficient dominated by metal defects α = 0.6, applicable to the scenario where metal material rejection is prioritized. If the weight deviation amplitude > 1.5 times, the weight coefficient dominated by weight defects is automatically switched α = 0.4 to prioritize protecting the packaging structure.

[0008] Preferably, three sets of inclined chutes are sequentially arranged in the pushing direction of the push plate, with a chute inclination angle of 15°. A diffuse reflection sensor is also installed at the entrance of each set of chutes, and the diffuse reflection sensor is electrically connected to the control system; The control system pre-calculates the sliding trajectory of the packaged goods according to the pusher plate action parameters, so that the weight-type packaged goods fall into the first chute correspondingly, the composite-type packaged goods fall into the second chute correspondingly, and the metal-type packaged goods fall into the third chute correspondingly; When the diffuse reflection sensor of any chute fails to detect a signal within the preset time after the pusher plate action, the control system triggers parameter self-learning and calculates ΔF = (S 目标 -S 实际 )×m 实际 ×(g×sin15°+μ×g×cos15°) / S 推板行程 , where S 推板行程 is the acting distance of the pusher plate, and the basic thrust of this specification is corrected again F 修正 ' = F 修正 +Δ F ×β, where β is the empirical correction coefficient.

[0009] Automatic package rejection method, applying the described device, the method includes: The conveyor belt conveys the packaged goods in sequence along the conveying direction. The metal detector detects in real time whether the packaged goods contain metal and outputs a metal detection signal. The weighing platform detects the weight of the packaged goods in real time and outputs a weight difference signal. The encoder collects the conveyor belt speed in real time and transmits it to the control system; The control system receives the metal detection signal and the weight difference signal. Based on the conveyor belt linear speed collected by the encoder, through the trigger timestamp of the opposed photoelectric grating and the preset position of the opposed photoelectric grating from the starting point of the conveyor belt, using the formula x = x 0 + v ×( t - t 0 ) calculates the position coordinates of the packaged goods in real time, where x 0 is the position of the opposed photoelectric grating from the starting point of the conveyor belt, v is the conveyor belt linear speed, t 0 is the grating trigger timestamp; The control system determines the type of the packaged goods as metal type, weight type or composite type according to the metal detection signal and the weight difference signal. Three types of rejection station coordinates are preset for the three types of packaged goods. The composite type station is the first distance downstream of the opposed photoelectric grating. The metal type station is the second distance downstream of the opposed photoelectric grating and the second distance is greater than the first distance. The weight type station is the third distance downstream of the opposed photoelectric grating and the third distance is greater than the second distance. The metal type station is the midpoint of the weight type station and the composite type station. Through the encoder pulse difference Δ N the station spacing Δ L is converted as Δ N×0.1 mm to ensure that the front end of the package reaches the corresponding work station within ±5 mm when the pusher plate starts; The control system calculates the start delay time of the pusher plate based on the conveyor belt linear velocity, the conveying distance between the current position coordinates of the package and the coordinates of the corresponding rejection work station, and drives the servo motor to act according to the preset thrust parameters. When detecting a metal package, the pressure sensor monitors and maintains the thrust in real time F m at 70 - 80 N, and dynamically corrects according to the formula F m修正 = F m × K 1 where K 1 is a dynamic compensation coefficient of 0.95 - 1.05, calculated from the friction coefficient μ corresponding to the chute inclination angle of 15°. When exceeding this range, the servo motor torque is automatically adjusted through the PID algorithm to make S 实际 = S 目标 ±15 mm; when detecting a weight package, the pressure sensor monitors and maintains the thrust in real time F w at 50 - 60 N, and dynamically corrects according to the weight data fed back by the weighing platform through the formula F w修正 = F w ×( m 实际 / m 标准 )× K 1 where m 实际 is the measured weight of the package; when detecting a composite package, the pressure sensor monitors and maintains the thrust in real time F m&w at 60 - 70 N, combines the weight data fed back by the weighing platform, and dynamically corrects the thrust using the weighted average algorithm F m&w修正 =( F m × α + F w ×(1− α ))×( m 实际 / m 标准 )× K 1 , where the weight coefficient dominated by metal defects α= 0.6, applicable to the scenario where metal material rejection takes precedence. If the weight deviation amplitude > 1.5 times, automatically switch to the weight defect-dominated weight coefficient α = 0.4, giving priority to protecting the packaging structure; The servo motor drives the push plate to act. The plate surface of the push plate is perpendicular to the conveying direction of the conveyor belt. The thrust of the push plate is adjusted in real time through a pressure sensor, and the corresponding type of packaging is pushed to the preset rejection station.

[0010] Preferably, the control system pre-stores N groups of packaging specification parameters, including the weight of the packaging, the basic thrust, and the sliding distance-thrust mapping relationship, S 实际 = (F×S 推板行程 ) / (m 实际 ×(g×sin15° + μ×g×cos15°)), where t 推板 is the acting time of the push plate, g = 9.8 m / s². When the packaging passes through the opposed beam grating, it triggers the barcode scanners installed on both sides of the conveyor belt to read the specification code, and the control system matches the standard weight of the corresponding specification according to the code m 标准 Set the basic thrust F , if the barcode scanning fails or there is no code, by default, use the parameters of the previous same-type packaging. It only takes effect when the actual weight measured by the weighing platform m 实际 has a weight error < 10% with the previous same type, otherwise calculate the thrust with m 实际 When the deviation between the code read by the barcode scanner F and the actual measurement of the weighing platform m 标准 > 5%, the control system performs the following operations: trigger an alarm signal, calculate the thrust based on the actual measured weight m 实际 as the benchmark, and update the pre-stored parameter library. m 实际

[0011] Preferably, after each group of specifications is successfully rejected 3 times in a row, the system automatically records the current data m 实际 ,, F 修正 , S 实际, standardized by the Z-score method, the dataset is divided into a training set and a validation set in a ratio of 8:2. A neural network is constructed with a two-layer feedforward structure. The input layer receives the standardized package weight and target chute distance data. The hidden layer has 8 neurons and uses the ReLU activation function. After the output, it is connected to a Dropout layer with a dropout rate of 0.2. The output layer linearly outputs the standardized predicted thrust. This neural network uses the mean squared error as the loss function and iterates with a learning rate of 0.001 through the Adam optimizer.

[0012] The present invention has at least the following beneficial effects: First, by sequentially arranging a metal detection unit, a weight detection unit, a photoelectric grating, and a servo pusher mechanism on the conveyor belt and integrating multi-source detection data through the control system, the present invention realizes the classification detection and precise rejection of metal, weight, and composite defect packages.

[0013] Second, the present invention sets different thrust ranges for packages with different defect types. For metal packages, a higher thrust range is adopted and the torque is adjusted in real time. For weight packages, the thrust is corrected by combining the measured weight ratio. For composite packages, the weighted algorithm is used to balance different defect dominant factors, avoiding incomplete rejection or damage to the packaging structure caused by improper thrust, and improving the stability and applicability of the rejection process.

[0014] Third, by setting three groups of inclined chutes and diffuse reflection sensors and combining the pre-designed calculation of the sliding trajectory, the present invention realizes the classified collection of packages with different defect types.

[0015] Fourth, through the learning of historical data by the neural network model, the device of the present invention has the ability of self-learning, can automatically optimize the thrust parameters to adapt to different combinations of weights and chute distances, and is applicable to the complex production scenarios of multi-specification packages.

[0016] Other advantages, objectives, and features of the present invention will be partially reflected by the following description and partially understood by those skilled in the art through the research and practice of the present invention. Detailed implementation manners

[0017] The following further detailed description of the present invention is made in combination with details so that those skilled in the art can implement it according to the description in the specification.

[0018] It should be understood that the terms such as "having", "comprising", and "including" used herein do not exclude the presence or addition of one or more other elements or their combinations.

[0019] It should be noted that the experimental methods described in the following implementation schemes are all conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial channels unless otherwise specified.

[0020] The present invention provides an automatic bag rejection device, which includes a horizontally arranged conveyor belt. A metal detection unit, a weight detection unit, a through-beam grating and a servo push plate mechanism are sequentially arranged in the conveying direction of the conveyor belt; The metal detection unit includes a metal detector fixedly installed above the conveyor belt, which is used to detect whether the packages on the conveyor belt contain metal impurities. It is installed directly above the conveyor belt and monitors in real time through the principle of electromagnetic induction. The transmitting coil covers the entire width of the conveyor belt. When the package contains metal impurities, the receiving coil senses a signal mutation and outputs a 0-5 V level signal to the control system. The weight detection unit includes a weighing platform arranged below the conveyor belt. The weighing platform is supported by 4 groups of S-type weighing sensors on the supporting surface of the conveyor belt. The implementation of being fixedly connected to the supporting surface of the conveyor belt is that the sensor is connected to the supporting surface through a spherical bearing to eliminate the interference of the conveyor belt tension and output a weight analog signal in real time. The through-beam grating is fixed above the conveyor belt and includes a transmitting end and a receiving end. The built-in encoder is directly connected to the driven shaft of the conveyor belt through a gear to collect the conveyor belt speed in real time. When a package passes through, it blocks the light beam and triggers the encoder to record the timestamp. The servo push plate mechanism includes a push plate driven by a servo motor and a pressure sensor. The plate surface of the push plate is perpendicular to the conveying direction of the conveyor belt, and a 3 mm thick silicone pad is pasted on the plate surface to buffer the impact force. The servo motor drives the push plate to act according to the instructions of the control system, and the pressure sensor monitors the thrust of the push plate in real time. The detection threshold of the pressure sensor is 50-100 N. The metal detector continuously emits electromagnetic signals. When the package contains metal impurities, the signal changes and outputs a metal detection signal to the control system. The weighing platform uses a strain gauge sensor to measure the weight of the package in real time, calculates the difference between the measured weight and the standard weight, and outputs a weight difference signal. The encoder of the through-beam grating is connected to the driving shaft of the conveyor belt through a gear or a belt. Each rotation generates a fixed number of pulses, and the linear speed of the conveyor belt is calculated by pulse counting. v .

[0021] The device further includes a control system, which is electrically connected to the metal detection unit, the weight detection unit, the encoder, the servo motor, and the pressure sensor respectively. The control system is an industrial-grade PLC or an embedded controller. The control system receives the metal detection signal output by the metal detection unit and the weight difference signal output by the weight detection unit, and pre-stores the determination logics for metal, weight, and composite defects. For example, if the metal detection signal triggers the metal type, it is marked as a metal defect and the red warning light is triggered; if the weight difference exceeds the threshold, it triggers the weight type, such as the actual weight < 90% or > 110% of the standard value, which is marked as a weight defect and the yellow warning light is triggered; if both are triggered simultaneously, it is the composite type, marked as a composite defect and the dual-color warning light is triggered. Based on the conveyor belt linear speed and the conveying distance, the push plate start delay time for metal-packaged goods, weight-packaged goods, or composite-packaged goods is calculated. The control system presets the thrust parameters according to the defect type, drives the servo motor to act according to the preset thrust parameters, reads the pressure sensor data in real time when driving the motor, and adjusts the motor torque through the PID algorithm when it exceeds the range to ensure stable thrust.

[0022] In the above technical solution, by sequentially arranging a metal detection unit, a weight detection unit, an opposed-beam grating, and a servo push plate mechanism on the conveyor belt, and integrating multi-source detection data through the control system, the classification detection and precise rejection of defective packages of metal, weight, and composite types are realized. The metal detector and the weighing platform respectively monitor metal impurities and weight deviations in real time, the encoder provides conveyor belt speed data, and the control system dynamically calculates the push plate start timing and thrust parameters based on these data to ensure that packages of different defect types are accurately rejected at the correct positions, improving the rejection efficiency and reliability.

[0023] In another technical solution, the control system internally sets up a time-position conversion model, and calculates the package position coordinates in real time through encoder pulse counting. The calculation formula is x = x 0 + v ×( t - t 0 ) where x 0 is the position of the opposed-beam grating from the starting point of the conveyor belt, v is the conveyor belt linear speed, t 0 is the grating trigger timestamp; For metal packaging, weight packaging or composite packaging, the control system presets three types of rejection station coordinates: the composite station is at the first distance downstream of the opposed grating; the metal station is at the second distance downstream of the opposed grating, and the second distance is greater than the first distance; the weight station is at the third distance downstream of the opposed grating, and the third distance is greater than the second distance. The metal station is the midpoint between the weight station and the composite station. Convert the station spacing Δ N through the encoder pulse difference Δ L = Δ N ×0.1 mm to ensure that the front end of the packaging reaches within ±5 mm of the corresponding station when the pusher plate starts.

[0024] Install a mechanical limit block (E) at the starting point of the conveyor belt (at the drive shaft) and set an optoelectronic switch (F) at the corresponding position. When the conveyor belt runs empty, the drive shaft rotates to make the limit block trigger the optoelectronic switch, and the control system records the encoder pulse count at this time as the initial zero point, and automatically calibrates the position of the opposed grating from the starting point of the conveyor belt. x 0 .

[0025] When the leading edge of the packaging enters the detection area of the opposed grating, the signal at the receiving end changes from "on" to "off", and the control system records the trigger timestamp. t 0 , and synchronously reads the current encoder pulse count. N 0 .

[0026] Conveyor belt linear velocity v is calculated through the encoder pulse change rate: the pulse increment ΔN is statistically counted every 10 ms and converted into the actual displacement Δ L = Δ N ×0.1 mm. The displacement per single pulse is 0.1 mm, so as to obtain the real-time speed v = Δ L / 0.01 s.

[0027] Real-time position coordinates x are calculated as follows: starting from the trigger timestamp t 0 , the position is updated every 1 ms, and the formula can be simplified to x = x 0 + v ×( t - t 0 ), that is, "initial position + speed × time difference", to ensure that the position calculation is synchronized with the movement of the conveyor belt.

[0028] Taking metal defects as an example, when the metal detector (H) detects a metal signal, the control system marks it as a metal defect and retrieves the preset target pulse count for the metal work station. N 2 , continuously monitors the encoder pulse count N in real time, calculates the pulse difference Δ between the current position and the target work station N 剩余 = N 2 - N 实时 .

[0029] When Δ N 剩余 = the pulse count corresponding to the pusher width + the pulse count corresponding to the package length (estimated by the occlusion time of the through-beam grating, for example, an occlusion of 20 ms corresponds to a length of 200 mm, converted to 2000 pulses), triggers the pusher start command to ensure that when the pusher contacts the package, its leading edge just reaches the target work station.

[0030] In the above technical solution, through the time-position conversion model and encoder pulse counting, the position coordinates of the package are accurately calculated in real time. Combining the preset coordinates of the three types of rejection work stations and the conversion method of the work station spacing, the front end of the package can accurately reach within the allowable error range (±5 mm) of the corresponding work station when the pusher starts, solving the problem of large rejection position deviation of the traditional device, improving the accuracy and consistency of the rejection action, and avoiding rejection failure or equipment collision caused by position deviation.

[0031] In terms of thrust control, the pusher thrust of the traditional device is mostly a fixed parameter or only simply adjusted proportionally based on a single detection parameter (such as weight), without considering the different thrust requirements for different defect types (metal, weight, composite), nor dynamically compensating for the sliding resistance of the package on the inclined chute (such as the friction coefficient and chute inclination). In another technical solution, the control system has a built-in thrust adaptive algorithm: When detecting a metal package, the pressure sensor monitors and maintains the thrust F m at 70 - 80 N in real time and dynamically corrects it according to the formula F m修正 = F m × K 1 . Among them K 1 is a dynamic compensation coefficient of 0.95 - 1.05, calculated from the friction coefficient μ corresponding to the chute inclination of 15°. When exceeding this range, the servo motor torque is automatically adjusted through the PID algorithm, for example, ±3% each time, so that S 实际 = S 目标±15 mm to compensate for the friction change of the slideway (such as the difference in the surface material of the packaging), ensure that the thrust is stable in the range of 70 - 80 N, and avoid the retention of metal impurities on the production line due to insufficient thrust or damage to the packaging due to excessive thrust; When detecting weight-type packaging, the pressure sensor monitors and maintains the thrust in real time F w at 50 - 60 N, and dynamically corrects it according to the weight data feedback from the weighing platform. The correction formula is F w修正 = F w ×( m 实际 / m 标准 )× K 1 for dynamic correction, where m 实际 is the measured weight of the packaging. Through the linear mapping of weight-thrust, different weight packages can obtain a matching pushing force to ensure a stable sliding distance; When detecting composite-type packaging, the pressure sensor monitors and maintains the thrust in real time F m&w at 60 - 70 N, combines the weight data feedback from the weighing platform, and dynamically corrects the thrust using the weighted average algorithm F m&w修正 =( F m × α + F w ×(1− α ))×( m 实际 / m 标准 )× K 1 , where the weight coefficient dominated by metal defects α = 0.6, applicable to the scenario where metal material rejection is prioritized. If the weight deviation amplitude > 1.5 times, automatically switch to the weight coefficient dominated by weight defects α = 0.4, which gives priority to protecting the packaging structure. While removing metal impurities, the main thrust component is reduced to protect the packaging structure.

[0032] The control system collects the pressure sensor and weighing data every 20 ms, and uses moving average filtering (5 samples) to eliminate signal fluctuations. When the detected values exceed the preset range for 3 consecutive times, the motor torque is adjusted to ensure that the thrust correction response time < 100 ms, adapting to the real-time control requirements of the conveyor belt speed ≤ 2 m / s, and realizing the "detection - calculation - execution" closed-loop control to improve the rejection reliability of different defect type packages.

[0033] In the above technical solution, different thrust ranges are set for packages with different defect types, and through a pressure sensor for real-time monitoring and a dynamic correction algorithm, it is ensured that the thrust adapts to the characteristics of the packages. For metal packages, a higher thrust range is adopted and the torque is adjusted in real time. For weight packages, the thrust is corrected by combining the measured weight ratio. For composite packages, the weighted algorithm is used to balance different dominant defect factors, avoiding incomplete rejection or damage to the packaging structure caused by improper thrust, and improving the stability and applicability of the rejection process.

[0034] In another technical solution, three sets of inclined chutes are sequentially arranged in the pushing direction (transverse) of the push plate, with different sliding distances. The inclination angle of the chutes is 15°. A diffuse reflection sensor is also installed at the entrance of each set of chutes. The diffuse reflection sensor is electrically connected to the control system. The transmitting end and the receiving end of the sensor are aligned with the center line of the chute. When the package slides in, the reflected light triggers a signal (high level), which is fed back to the control system in real time. A 0.3 mm thick PE anti-slip film is pasted on the inner side of the chute, and the entrance width is 10 mm wider than both sides of the conveyor belt to avoid jamming. The control system pre-calculates the sliding trajectory of the package according to the action parameters of the push plate, so that the weight packages fall into the first chute correspondingly, achieving short-distance sliding with a lighter weight and a smaller thrust. The composite packages fall into the second chute correspondingly, with a medium sliding distance. The metal packages fall into the third chute correspondingly, and a large thrust ensures long-distance sliding to the farthest chute. When the diffuse reflection sensor of any chute does not detect a signal within the preset time after the push plate action, the system determines that the rejection fails, and the control system triggers parameter self-learning to calculate Δ F = (S 目标 -S 实际 )×m 实际 ×(g×sin15°+μ×g×cos15°) / S 推板行程 , where S 推板行程 is the acting distance of the push plate, converting the physical law (the relationship between resistance and mass, angle) into calculable engineering parameters, avoiding the blindness of pure empirical adjustment, and correcting the basic thrust of this specification again F 修正 ’= F 修正 +Δ F ×β, where β is an empirical correction coefficient, and its value range is 0.2 - 0.8, calibrated through historical data. Based on the thrust correction logic of measured experience, the adjustment amplitude is controlled by the β coefficient, seamlessly connecting with the servo motor torque control, ensuring both the response speed and preventing package damage, and improving the measured classification accuracy.

[0035] In the above technical solution, by setting three groups of inclined chutes and diffuse reflection sensors and combining with the pre-designed calculation of the sliding trajectory, the classified collection of packaging products with different defect types is realized. The diffuse reflection sensor monitors in real time whether the packaging product falls into the target chute. If the detection fails, the secondary rejection program is triggered, and the problem of inaccurate classification and dependence on manual intervention after rejection by the traditional device is solved by adjusting the stroke and speed of the push plate, improving the automation degree and classification reliability.

[0036] An automatic packaging product rejection method, applying the device described above, the method includes: (1) The conveyor belt conveys the packaging products in sequence along the conveying direction. The metal detector detects in real time whether the packaging product contains metal and outputs a metal detection signal. The weighing platform detects the weight of the packaging product in real time and outputs a weight difference signal. The encoder collects the conveyor belt speed in real time and transmits it to the control system.

[0037] A metal detector is installed above the conveyor belt to continuously scan the passing packaging products and immediately send a signal to the control system when metal impurities are detected. A weighing platform is arranged below the conveyor belt to measure the weight of the packaging product in real time and output a weight anomaly signal (too light or too heavy) after comparing with the standard weight. An opposed beam grating above the conveyor belt triggers a signal when the packaging product passes through, and the built-in encoder synchronously records the real-time speed of the conveyor belt to provide data for subsequent position calculation. The metal detector, weighing platform, and encoder transmit the detection signals to the control system in real time to form multi-source data fusion (metal presence, weight deviation, conveyor belt speed).

[0038] (2) The control system receives the metal detection signal and the weight difference signal. Based on the conveyor belt linear speed collected by the encoder, through the trigger timestamp of the opposed beam grating and the preset position of the opposed beam grating from the starting point of the conveyor belt, using the formula x = x 0 + v × ( t - t 0 ) calculates the position coordinates of the packaging product in real time, where x 0 is the position of the opposed beam grating from the starting point of the conveyor belt, v is the conveyor belt linear speed, t 0 is the grating trigger timestamp.

[0039] When the opposed beam grating is triggered, the time and the initial position of the conveyor belt are recorded, and the position of the packaging product on the conveyor belt is continuously updated in combination with the real-time speed to ensure that when the push plate starts, the front end of the packaging product accurately reaches the target station.

[0040] (3)The control system determines the type of the packaged object as metal, weight, or composite based on the metal detection signal and the weight difference signal, and presets the coordinates of three rejection stations for the three types of packaged objects. The coordinate of the composite and weight station is the first distance downstream of the opposed grating, the coordinate of the metal station is the second distance downstream of the opposed grating and the second distance is greater than the first distance, the coordinate of the weight station is the third distance downstream of the opposed grating and the third distance is greater than the second distance, and the metal station is the midpoint between the weight station and the composite station. The encoder pulse difference Δ N is used to convert the station spacing Δ L = Δ N ×0.1 mm to ensure that the front end of the packaged object reaches within ±5 mm of the corresponding station when the push plate starts.

[0041] Along the conveyor belt direction, three rejection stations are preset for the three types of defects: Composite station: Close to the opposed grating, used for quickly rejecting the packaged objects with both defects at the same time; Metal station: At the middle distance, preferentially rejecting the metal-containing packaged objects that may damage the equipment; Weight station: Far from the opposed grating, leaving enough time to confirm the abnormal weight and avoiding misjudgment.

[0042] The position of the station is tracked in real time by the conveyor belt encoder to ensure that the packaged objects of different defect types are accurately rejected at the corresponding stations.

[0043] (4)The control system calculates the start delay time of the push plate based on the conveyor belt linear velocity and the conveying distance between the current position coordinate of the packaged object and the coordinate of the corresponding rejection station, and drives the servo motor to act according to the preset thrust parameter. When the detected packaged object is of the metal type, the pressure sensor monitors and maintains the thrust F m at 70 - 80 N, and dynamically corrects it according to the formula F m修正 = F m × K 1 where K 1 is a dynamic compensation coefficient of 0.95 - 1.05, calculated from the friction coefficient μ corresponding to the chute inclination angle of 15°. When it exceeds this range, the servo motor torque is automatically adjusted through the PID algorithm to make S 实际 = S 目标 ±15 mm; when the detected packaged object is of the weight type, the pressure sensor monitors and maintains the thrust F w at 50 - 60 N, and according to the weight data fed back by the weighing platform, through the formula F w修正 = Fw ×( m 实际 / m 标准 )× K 1 Dynamic correction, where m 实际 is the measured weight of the packaging; when the detected packaging is a composite type, the pressure sensor monitors in real time and maintains the thrust F m&w at 60 - 70 N. Combining with the weight data feedback from the weighing platform, the thrust is dynamically corrected using the weighted average algorithm. F m&w修正 =( F m × α + F w ×(1− α ))×( m 实际 / m 标准 )× K 1 , where the weight coefficient dominated by metal defects α = 0.6, applicable to the scenario where metal material rejection is prioritized. If the weight deviation amplitude > 1.5 times, the weight defect - dominated weight coefficient α = 0.4 is automatically switched to prioritize protecting the packaging structure.

[0044] For metal types, a higher thrust is used to ensure that the packaging containing metal overcomes the resistance and slides into the farthest chute. For weight types, the thrust is dynamically adjusted according to the measured weight. For lightweight packaging, a smaller thrust is used to avoid sliding out of the chute. For composite types, the thrust is between the two, and the weighted strategy is used to balance the metal rejection priority and the packaging protection requirements.

[0045] (5) The servo - motor drives the push - plate to act. The surface of the push - plate is perpendicular to the conveying direction of the conveyor belt. The thrust of the push - plate is adjusted in real time through the pressure sensor, and the corresponding type of packaging is pushed to the preset rejection station.

[0046] Detection stage: The metal detector, weighing platform, and opposed - beam grating work synchronously to collect the packaging defect information, weight data, and position and speed in real time. Decision - making stage: The control system integrates multi - source data, judges the defect type, matches the corresponding rejection station, and calculates the starting time and required thrust of the push - plate. Execution stage: The servo - push - plate mechanism acts according to the instruction, and the pressure sensor calibrates the thrust in real time to ensure that the packaging slides into the corresponding chute along the preset trajectory. Optimization stage: The chute sensor feeds back the rejection result, and the system self - learns to adjust the thrust parameters to continuously improve the rejection accuracy and reliability.

[0047] In the above technical solution, by integrating the real-time data of each unit of the device, a complete process from detection, position calculation, defect classification to thrust control is formed, ensuring that each rejection action is based on accurate detection and calculation results, realizing the automatic and precise rejection of metal, weight and composite defect packaging materials, and improving the quality control efficiency and intelligent level of the packaging production line.

[0048] In another technical solution, the control system pre-stores N a set of packaging material specification parameters, including the weight of the packaging material, the basic thrust and the sliding distance-thrust mapping relationship, S 实际 = (F×S 推板行程 ) / (m 实际 ×(g×sin15° + μ×g×cos15°)), where t 推板 is the acting time of the push plate, g = 9.8 m / s². When the packaging material passes through the opposed beam grating, it triggers the barcode scanners installed on both sides of the conveyor belt to read the specification code (such as QR code, bar code), and transmits it to the control system. The control system matches the standard weight of the corresponding specification according to the code m 标准 to set the basic thrust F . If the barcode scanning fails or there is no code, the parameters of the previous same-type packaging material are used by default. It only takes effect when the actual weight measured by the weighing platform m 实际 has a weight error of less than 10% from the previous same type. Otherwise, the thrust is calculated m 实际 . When the deviation between the F read by the barcode scanner m 标准 and the actual measurement of the weighing platform m 实际 is greater than 5%, the control system performs the following operations: triggers an alarm signal, calculates the thrust based on the actual measured weight m 实际 , and updates the pre-stored parameter library.

[0049] In the above technical solution, through pre-storing specification parameters and barcode recognition technology, rapid parameter matching of packaging materials with different weight specifications is realized. Combining with the thrust correction algorithm based on the physical model, it ensures that no matter how the weight of the packaging material changes, the thrust of the push plate can be automatically adjusted, so that the pushing distance is accurately aligned with the target slideway, solving the problem that the traditional device needs to manually adjust parameters frequently, and improving the versatility and production adaptability of the device.

[0050] In another technical solution, after 3 consecutive successful rejections for each group of specifications, that is, when all the slideway sensors are triggered, the system automatically captures the weight, defect type, target slideway, actual thrust, and environmental parameters, and records the current data m 实际 ,F 修正 , S 实际 , the data is normalized and stored in the data set for targeted training to avoid interference from abnormal data. The data set is divided into a training set and a validation set in an 8:2 ratio to start training, and a neural network is constructed to learn the mapping relationship between weight, target chute distance, and required thrust. The neural network adopts a two-layer feedforward structure (input layer → hidden layer → output layer). The input layer receives the normalized package weight and target chute distance data. The hidden layer corresponds to 8 basic thrust strategies through 8 neurons, equipped with a ReLU activation function and a dropout rate of 0.2 to prevent overfitting. The output layer linearly outputs the normalized predicted thrust. The neural network uses the mean squared error as the loss function and iterates with a learning rate of 0.001 through the Adam optimizer. When the data is insufficient (such as less than 3 successful cases for a new specification), it automatically switches to traditional formula control (guaranteed mechanism); when the data is sufficient (>50 groups), it enables the neural network output (intelligent mode). The physical formula relied on by the traditional scheme requires accurate measurement of parameters such as the friction coefficient and inclination angle; the neural network directly learns the implicit association of "weight → chute → thrust" from the data, can handle complex variables that are difficult to describe by formulas, and realizes "data-driven" instead of "model-driven". Traditional PID control is an immediate feedback of "deviation → correction"; the self-learning of the neural network is a long-term evolution of "experience → law → prediction", forming a large closed-loop of "data acquisition → model training → strategy output → result verification" to achieve self-optimization.

[0051] In the above technical solution, through the learning of historical data by the neural network model, the device is equipped with self-learning ability and can automatically optimize the thrust parameters to adapt to different combinations of weight and chute distance. Compared with the traditional thrust control based on fixed formulas, the prediction accuracy is higher and no manual intervention is required, especially suitable for complex production scenarios of multi-specification packages, further improving the intelligence and precision level of the rejection process.

[0052] The number of devices and the processing scale described here are used to simplify the description of the present invention. The application, modification, and variation of the present invention are obvious to those skilled in the art.

[0053] Although the embodiments of the present invention have been disclosed as above, it is not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those skilled in the art, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to the specific details and the details shown and described here.

Claims

1. Automatic bag picking device, characterized in that: It comprises a horizontally arranged conveyor belt, wherein a metal detection unit, a weight detection unit, a corresponding grating and a servo push plate mechanism are sequentially arranged in the conveying direction of the conveyor belt; The metal detection unit includes a metal detector fixedly installed above the conveyor belt, the weight detection unit includes a weighing platform arranged below the conveyor belt, the weighing platform is fixedly connected to the supporting surface of the conveyor belt, the reflected grating is fixed above the conveyor belt, and the built-in encoder collects the conveyor belt speed in real time; the servo push plate mechanism includes a push plate driven by a servo motor and a pressure sensor, the plate surface of the push plate is perpendicular to the conveying direction of the conveyor belt, and the detection threshold of the pressure sensor is 50-100 N; The device also includes a control system, which is electrically connected to the metal detection unit, the weight detection unit, the encoder, the servo motor and the pressure sensor respectively. The control system receives the metal detection signal output by the metal detection unit and the weight difference signal output by the weight detection unit, calculates the push plate start-up delay time of the metal packaging, the weight packaging or the composite packaging based on the conveyor belt linear speed and the conveying distance, and drives the servo motor to operate according to the preset thrust parameters.

2. The automatic bag picking device according to claim 1, characterized in that: The control system has a built-in time-position conversion model, which calculates the position coordinates of the package in real time through encoder pulse counting. The calculation formula is: x = x 0 + v ×( t - t 0 ),in x 0 is the distance between the reflected grating and the starting point of the conveyor belt. v is the conveyor belt linear speed, t 0 Timestamp for grating trigger; For metal packaging, weight packaging or composite packaging, the control system presets three types of rejection station coordinates: the composite station is the first distance downstream of the reflected grating; the metal station is the second distance downstream of the reflected grating, the second distance is greater than the first distance, the weight station is the third distance downstream of the reflected grating, the third distance is greater than the second distance, and the metal station is the midpoint of the weight station and the composite station; By encoder pulse difference Δ N Conversion station spacing Δ L = Δ N ×0.1 mm, ensuring that the front end of the package reaches the corresponding station within ±5 mm when the push plate is started.

3. The automatic bag picking device according to claim 1, characterized in that: The control system has a built-in thrust adaptive algorithm: When metal packaging is detected, the pressure sensor monitors and maintains thrust in real time F m At 70-80 N, and according to the formula F m修正 = F m × K 1 Dynamic correction, where K 1 The dynamic compensation coefficient is 0.95-1.05, which is calculated from the friction coefficient μ corresponding to the slideway inclination of 15°. When it exceeds this range, the PID algorithm automatically adjusts the servo motor torque to make S 实际 =S 目标 ±15 mm; When a heavy package is detected, the pressure sensor monitors and maintains the thrust in real time F w At 50-60 N, it is dynamically corrected according to the weight data fed back by the weighing platform. The correction formula is: F w修正 = F w ×( m 实际 / m 标准 )× K 1 ,in m 实际 The actual measured weight of the package; When a composite package is detected, the pressure sensor monitors and maintains the thrust in real time. F m&w The thrust is 60-70 N. Combined with the weight data fed back by the weighing platform, the weighted average algorithm is used to dynamically correct the thrust. F m&w修正 =( F m × α + F w ×(1− α ))×( m 实际 / m 标准 )× K 1 , where the weight coefficient dominated by metal defects is α = 0.6, suitable for metal material rejection priority scenarios. If the weight deviation is greater than 1.5 times, the weight factor dominated by weight defects will be automatically switched. α = 0.4, giving priority to protecting the packaging structure.

4. The automatic bag picking device according to claim 3, characterized in that: Three groups of inclined slideways are arranged in sequence in the pushing direction of the push plate, and the slideway inclination angle is 15°. A diffuse reflection sensor is also installed at the entrance of each group of slideways, and the diffuse reflection sensor is electrically connected to the control system; The control system calculates the sliding trajectory of the packages according to the preset parameters of the push plate action, so that the heavy packages fall into the first slideway, the composite packages fall into the second slideway, and the metal packages fall into the third slideway; When the diffuse reflection sensor of any slideway does not detect a signal within the preset time after the push plate action, the control system triggers parameter self-learning and calculates ΔF = (S 目标 -S 实际 )×m 实际 ×(g×sin15°+μ×g×cos15°) / S 推板行程 , where S 推板行程 For the distance of the push plate, the basic thrust of this specification is revised again F 修正 ' = F 修正 +Δ F ×β, where β is the empirical correction coefficient.

5. Automatic package removal method, characterized in that: Using the device according to any one of claims 1 to 4, the method comprises: The conveyor belt conveys the packages in sequence along the conveying direction. The metal detector detects whether the packages contain metal in real time and outputs a metal detection signal. The weighing platform detects the weight of the packages in real time and outputs a weight difference signal. The encoder collects the conveyor belt speed in real time and transmits it to the control system. The control system receives the metal detection signal and the weight difference signal, and based on the conveyor belt linear speed collected by the encoder, the time stamp of the photoelectric grating trigger and the preset position of the photoelectric grating from the conveyor belt starting point, uses the formula x = x 0 + v ×( t - t 0 ) calculates the package position coordinates in real time, where x 0 is the distance between the reflected grating and the starting point of the conveyor belt. v is the conveyor belt linear speed, t 0 Timestamp for grating trigger; The control system determines the type of packaging as metal, weight or composite according to the metal detection signal and the weight difference signal. Three types of rejection station coordinates are preset for the three types of packaging. The composite station is the first distance downstream of the reflected grating, the metal station is the second distance downstream of the reflected grating and the second distance is greater than the first distance, the weight station is the third distance downstream of the reflected grating and the third distance is greater than the second distance. The metal station is the midpoint of the weight station and the composite station. The encoder pulse difference Δ N Conversion station spacing Δ L = Δ N ×0.1 mm, ensuring that the front end of the package reaches the corresponding station within ±5 mm when the push plate is started; The control system calculates the push plate start delay time based on the conveyor belt linear speed and the conveying distance between the current position coordinates of the package and the corresponding rejection station coordinates, and drives the servo motor to operate according to the preset thrust parameters. When a metal package is detected, the pressure sensor monitors and maintains the thrust in real time. F m At 70-80 N, and according to the formula F m修正 = F m × K 1 Dynamic correction, where K 1 The dynamic compensation coefficient is 0.95-1.05, which is calculated from the friction coefficient μ corresponding to the slideway inclination of 15°. When it exceeds this range, the PID algorithm automatically adjusts the servo motor torque to make S 实际 =S 目标 ±15 mm; when the object is detected as a heavy package, the pressure sensor monitors and maintains the thrust in real time F w At 50-60 N, and according to the weight data fed back by the weighing platform, the formula F w修正 = F w ×( m 实际 / m 标准 )× K 1 Dynamic correction, where m 实际 Measure the actual weight of the package; when it is detected as a composite package, the pressure sensor monitors and maintains the thrust in real time F m&w The thrust is 60-70 N. Combined with the weight data fed back by the weighing platform, the weighted average algorithm is used to dynamically correct the thrust. F m&w修正 =( F m × α + F w ×(1− α ))×( m 实际 / m 标准 )× K 1 , where the weight coefficient dominated by metal defects is α = 0.6, suitable for metal material rejection priority scenarios. If the weight deviation is greater than 1.5 times, the weight factor dominated by weight defects will be automatically switched. α = 0.4, giving priority to protecting the packaging structure; The servo motor drives the push plate to move. The surface of the push plate is perpendicular to the conveying direction of the conveyor belt. The push force of the push plate is adjusted in real time through the pressure sensor to push the corresponding type of packaging to the preset rejection station.

6. The automatic package removal method according to claim 5, characterized in that: The control system pre-stores N Assemble the packaging specifications, including packaging weight, basic thrust, and glide distance-thrust mapping, S 实际 = (F×S 推板行程 ) / (m 实际 ×(g×sin15°+μ×g×cos15°)), where t 推板 is the push plate action time, g=9.8 m / s². When the package passes through the beam grating, the scanner installed on both sides of the conveyor belt is triggered to read the specification code. The control system matches the standard weight of the corresponding specification according to the code. m 标准 Setting base thrust F If the code scanning fails or there is no code, the parameters of the last package of the same type will be used by default. m 实际 It will take effect when the weight error of the same type as the last time is less than 10%, otherwise m 实际 Calculating thrust F , when the scanner reads m 标准 Measured with weighing platform m 实际 When the deviation is greater than 5%, the control system performs the following operations: trigger an alarm signal to measure the weight m 实际 Calculate thrust for the benchmark and update the pre-stored parameter library.

7. The automatic package removal method according to claim 5, characterized in that: After each group of specifications is successfully rejected for 3 consecutive times, the system automatically records the current data m 实际 , F 修正 , S 实际 , the Z-score method is used for standardization, and the data set is divided into a training set and a validation set in a ratio of 8:

2. A neural network is constructed with a 2-layer feedforward structure. The input layer receives the standardized packaging weight and target slide distance data. The hidden layer passes through 8 neurons and uses the ReLU activation function. The output is connected to the Dropout layer with a discard rate of 0.

2. The output layer linearly outputs the standardized predicted thrust. The neural network uses the mean square error as the loss function and iterates with a learning rate of 0.001 through the Adam optimizer.