Metal product standard box packing scheduling optimization system based on data mining

Abstract

This invention relates to the field of intelligent manufacturing and industrial automation control technology, specifically to a data mining-based scheduling optimization system for standard box packaging of metal products. The system includes: a data acquisition module for receiving multi-source physical data from packaging stations; a signal verification and box entry identification module for performing signal denoising based on the multi-source physical data and updating the current box status; a historical data analysis module for performing offline statistics based on historical scheduling data, generating a packing rule table containing judgment parameters and a threshold correspondence table, and sending it to the field control module; a field control module for converting the scheduling parameter set into underlying hardware control instructions for direct hardware drive execution when a real part is detected entering the box; and a feedback adjustment module for adjusting the judgment parameters of the packing rule table using a deviation calculation method to update the rules. This invention reduces the risk of localized off-center loading of the box and sudden load changes in the actuator.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing and industrial automation control technology, specifically to a data mining-based scheduling optimization system for standard box packaging of metal products. Background Technology

[0002] The standard box packaging scheduling system for metal products is a key link in the automation and efficiency improvement of the production line. The accuracy of scheduling execution directly affects the packaging quality and hardware operation status. The system mainly includes photoelectric sensors, weighing sensors, robotic arms and conveyor belts, etc., and realizes the task of packing metal parts through the physical movement coordination between multiple devices. Existing technologies mainly rely on single sensor input or fixed packing rules, which cannot fully utilize multi-source data for joint verification. They are also difficult to adapt to complex working conditions such as dust obstruction and transient high-frequency mechanical oscillations. Abnormal vibrations and mechanical burrs are easily amplified into multiple packing events, resulting in scheduling input distortion and duplicate recognition. At the same time, when faced with complex load conditions such as the concentrated arrival of heavy items, existing solutions lack pre-calculation mechanisms. Solving each item one by one causes the computational burden on edge devices to exceed the rated computational load and cause instruction lag. Furthermore, the scheduling branch switching is not timely, increasing the risk of local off-center loading of the box and sudden load changes in the actuator. Furthermore, existing scheduling strategies often lack direct hardware-driven interfaces and quantitative closed-loop evaluation mechanisms. Long-term operation of fixed rules can easily deviate from actual working conditions, resulting in insufficient production line response and unclear direction for rule optimization. Therefore, a solution is urgently needed to address the problems existing in the current technology. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a data mining-based scheduling optimization system for standard box packaging of metal products. Specifically, the technical solution of this invention includes: The data acquisition module is used to receive multi-source physical data from the packaging station. The multi-source physical data includes the switch sequence of the diffuse reflection photoelectric sensor that characterizes the spatial occlusion state, the original strain analog voltage output by the bottom weighing sensor, and the real-time current feedback value of the robotic arm servo motor. The signal verification and box entry identification module is used to perform signal denoising based on multi-source physical data. When dust interference conditions are detected, the triggering effect of the switch sequence is blocked, and the module switches to multi-sensor verification mode. It combines the original strain simulation voltage and real-time current feedback value to identify the actual part entering the box event and updates the current box status including the current box's cumulative weight. The historical data analysis module is used to acquire historical scheduling data, perform offline statistics based on the historical scheduling data, generate a packing rule table containing judgment parameters and a threshold correspondence table, and send it to the field control module. The field control module is used to match the scheduling parameter group in the packing rule table based on the current box status including the current box's cumulative weight when a real part is detected entering the box, and convert the scheduling parameter group into low-level hardware control instructions to execute the hardware directly. The feedback adjustment module is used to collect scheduling quality evaluation parameters after a standard container is sealed, and adjust the judgment parameters of the packing rule table according to the deviation calculation formula constructed based on the scheduling quality evaluation parameters to update the rules.

[0004] Optionally, the data acquisition module is specifically used for: A real-time data set is constructed by collecting multi-source physical data in real time with a fixed sampling period. The real-time data set includes: the switching sequence of diffuse reflection photoelectric sensors that characterize the spatial occlusion state, the original strain analog voltage output by the bottom weighing sensor, the real-time current feedback value of the robotic arm servo motor that characterizes the transient load torque of the current grasping action, and the real-time linear speed of the conveyor belt that characterizes the current material supply cycle.

[0005] Optionally, the signal verification and box entry identification module is used to identify actual part box entry events, including: Calculate the switching frequency of the switching sequence within a preset time window representing the time length; When the switching frequency exceeds the preset dust characteristic threshold representing the abnormal switching density, it is determined that the dust interference period has begun, the triggering effect of the switching sequence is blocked, and the system switches to multi-sensor verification mode; when the switching frequency is less than or equal to the preset dust characteristic threshold, it is determined that the dust interference period has not begun, and the triggering effect of the switching sequence is maintained. In multi-sensor verification mode, the original strain simulation voltage is smoothed and filtered to output the dynamic filtered weight, and the weight step increment is calculated. When the real-time current feedback value is greater than the preset servo grab current peak threshold, and the weight step increment is greater than the minimum metal part weight recognition threshold, and both occur simultaneously within the preset synchronization time window, the event of a real part entering the box is determined to be triggered, and the current box status, which includes the current box's cumulative weight, is updated.

[0006] Optionally, the historical data analysis module is used to generate a packing rule table, including: Order parameters are extracted from historical scheduling data, and K-means clustering algorithm is used to generate packing pattern categories; order parameters include metal part density, volume, and arrival frequency. For each packing mode category, pre-calculate the packing posture and delivery order that satisfy the anti-eccentric load constraint, and encapsulate them into a packing rule table; The packing rule table includes physical characteristics such as the current cumulative weight range of the box and the number of consecutive heavy items arriving, and the execution result includes the set of robotic arm trajectory numbers and the set of engineering instructions for the frequency derating factor of the inverter.

[0007] Optionally, the field control module is used to match the scheduling parameter set, including: When a real part is identified as being placed in the box, the current cumulative weight of the box is extracted, and the average servo current estimate of multiple consecutive parts placed in the box is calculated based on the real-time current feedback value to construct an index key value. Input the index key value into the packing rule table for condition matching; If the number of consecutive heavy components arriving that continuously meet the average servo current estimation value greater than the heavy component current threshold reaches a preset number, the judgment logic jumps to the heavy load anti-eccentricity branch, matches and outputs the scheduling parameter group under the current working condition; if the preset number is not reached, the regular scheduling branch is executed and the corresponding scheduling parameter group is output. The scheduling parameter set includes the corresponding robot arm trajectory number extracted from the set of robot arm trajectory numbers, the frequency derating factor of the inverter, and the torque limiting amplification factor.

[0008] Optionally, the field control module is used to convert the scheduling parameter set into low-level hardware control instructions for direct hardware drive execution, including: The target frequency is calculated by combining the conveyor belt reference operating frequency with the inverter frequency derating factor through the field industrial bus protocol, and the inverter frequency control word register is overwritten. The robot arm trajectory number is parsed into a servo controller program segment call instruction, and the kinematic interpolation path is switched. This instruction maps the torque limit amplification factor to the overwrite instruction of the servo driver's torque limit register.

[0009] Optionally, the feedback adjustment module is used to update the rules, including: Calculate scheduling quality evaluation parameters, which include physical cycle fluctuation rate, empty box rate calculated based on final static weight, and number of servo overload alarms. Based on preset weighting constants, the physical cycle fluctuation rate, empty box rate, and servo overload alarm count are weighted and calculated to construct a deviation calculation formula; With the goal of minimizing the deviation calculation formula, the judgment thresholds and corresponding output instructions of each condition in the packing rule table are adjusted by the gradient descent optimization method, and the updated packing rule table is generated and issued.

[0010] Optionally, the system adopts a hierarchical architecture of master station and slave station, wherein: The historical data analysis module and the feedback adjustment module are deployed in the remote dispatch center; The data acquisition module, signal verification and box entry identification module, and field control module are deployed on the field control terminal that is close to the physical production line. The remote dispatch center and the field control terminal establish a connection through industrial Ethernet. The remote dispatch center sends the packing rule table and threshold correspondence table to the field control terminal, and the field control terminal uploads the scheduling quality evaluation parameters to the remote dispatch center. The field control terminal establishes a direct addressing or binary search structure in memory to complete the conversion of complex multivariate scheduling logic into a fixed search memory table structure.

[0011] Compared with the prior art, the present invention has the following beneficial effects: 1. When dust interference is detected, the photoelectric trigger is blocked and the system switches to a multi-sensor verification mode. The system combines the smoothed and filtered weight step increment with the real-time current feedback of the servo motor for synchronous verification. This mechanism overcomes the false triggering caused by dust obstruction and high-frequency mechanical oscillation, avoids amplifying jitter and burrs into multiple box entry events, and improves the authenticity and recognition accuracy of the scheduling input. 2. This invention generates a packing rule table containing anti-off-center load constraints through offline clustering, and performs fast table lookup matching on-site based on the cumulative weight and servo current estimation value; when heavy items arrive in a concentrated manner, it automatically jumps to the heavy load anti-off-center load branch, avoiding the excessive edge computing burden and instruction lag caused by real-time item-by-item solving, and reducing the risk of local off-center load on the box and sudden load changes in the actuator; 3. This invention directly converts scheduling parameters into low-level control commands for frequency converters and servos, improving the response speed of production line actions; at the same time, it introduces a closed-loop evaluation mechanism, which integrates cycle time fluctuation rate, empty box rate and overload alarm count to construct a deviation calculation formula and dynamically optimize and adjust rule thresholds, overcoming the defect of fixed rules deviating from actual working conditions in long-term operation, and making the direction of rule optimization more quantitative and clear. Attached Figure Description

[0012] The above and other objects, features and advantages of the present invention will become clearer from the following description of embodiments of the invention with reference to the accompanying drawings, in which: Figure 1 This is a schematic diagram of the module of the data mining-based standard box packaging scheduling optimization system for metal products provided in the embodiments of this application. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0014] A data mining-based scheduling optimization system for standard box packaging of metal products includes: The data acquisition module is used to receive multi-source physical data from the packaging station. The multi-source physical data includes the switch sequence of the diffuse reflection photoelectric sensor that characterizes the spatial occlusion state, the original strain analog voltage output by the bottom weighing sensor, and the real-time current feedback value of the robotic arm servo motor. The signal verification and box entry identification module is used to perform signal denoising based on multi-source physical data. When dust interference conditions are detected, the triggering effect of the switch sequence is blocked, and the module switches to multi-sensor verification mode. It combines the original strain simulation voltage and real-time current feedback value to identify the actual part entering the box event and updates the current box status including the current box's cumulative weight. The historical data analysis module is used to acquire historical scheduling data, perform offline statistics based on the historical scheduling data, generate a packing rule table containing judgment parameters and a threshold correspondence table, and send it to the field control module. The field control module is used to match the scheduling parameter group in the packing rule table based on the current box status including the current box's cumulative weight when a real part is detected entering the box, and convert the scheduling parameter group into low-level hardware control instructions to execute the hardware directly. The feedback adjustment module is used to collect scheduling quality evaluation parameters after a standard container is sealed, and adjust the judgment parameters of the packing rule table according to the deviation calculation formula constructed based on the scheduling quality evaluation parameters to update the rules. The data acquisition module is specifically used to: acquire multi-source physical data in real time at a fixed sampling period to construct a real-time data set, which includes: The switching sequence of the diffuse reflection photoelectric sensor characterizing the spatial occlusion state, the original strain analog voltage output by the bottom weighing sensor, the real-time current feedback value of the robotic arm servo motor characterizing the transient load torque of the current grasping action, and the real-time linear speed of the conveyor belt characterizing the current material supply cycle. The signal verification and box entry identification module is used to identify actual part entry events, including: Calculate the switching frequency of the switching sequence within a preset time window representing the time length; When the switching frequency exceeds the preset dust characteristic threshold representing the abnormal switching density, it is determined that the dust interference period has begun, the triggering effect of the switching sequence is blocked, and the system switches to multi-sensor verification mode; when the switching frequency is less than or equal to the preset dust characteristic threshold, it is determined that the dust interference period has not begun, and the triggering effect of the switching sequence is maintained. In multi-sensor verification mode, the original strain simulation voltage is smoothed and filtered to output the dynamic filtered weight, and the weight step increment is calculated. When the real-time current feedback value is greater than the preset servo grab current peak threshold, and the weight step increment is greater than the minimum metal part weight recognition threshold, and both occur simultaneously within the preset synchronization time window, the event of a real part entering the box is determined to be triggered, and the current box status, which includes the current box's cumulative weight, is updated.

[0015] This embodiment provides a scheduling optimization system for the standard box packaging station of metal products, such as... Figure 1 As shown; the system takes multi-source physical data from the packaging station as input, performs joint judgment on photoelectric switching quantity, weighing analog quantity and servo current feedback, and after identifying the real part entering the box event, matches the corresponding scheduling parameter group according to the issued packing rule table, and directly converts the scheduling parameter group into the low-level instructions that can be executed by the frequency converter and servo controller. After a standard box is sealed, the evaluation results such as cycle time, empty box rate, and overload alarm are used to correct the judgment parameters in the rule table; thus, even in the case of dust accidental triggering, mechanical vibration burrs, and concentrated arrival of heavy items, the box entry event can still be stably judged, and the scheduling actions of the packaging station can be kept consistent with the hardware actions. Acquire multi-source physical data from the packaging station, and synchronously collect the multi-source physical data according to a fixed sampling period to generate real-time data sets; where the fixed sampling period refers to the sampling interval at which the field control terminal polls each sensor channel according to a unified system cycle; Real-time data sets refer to the current observation set formed within the same sampling period, including at least the switching sequence of diffuse reflection photoelectric sensors, the raw strain analog voltage output by the bottom weighing sensor, the real-time current feedback value of the robotic arm servo motor, and the real-time linear speed of the conveyor belt. In practice, the switching sequence can be read through the digital input port, the original strain analog voltage can be written into the sampling buffer of the field control terminal through the analog-to-digital conversion channel, and the real-time current feedback value and the real-time linear speed of the conveyor belt can be read from the servo driver and encoder feedback register through the fieldbus and written into the local circular buffer according to the sampling period, so that the identification logic in the current period can directly index it. After reading, it will be transferred to the next period for overwriting. By adopting synchronous sampling and buffering under a unified period, all signals are compared and contrasted in coordination on the same judgment basis, thus avoiding timing misalignment caused by asynchronous acquisition. This embodiment can achieve multi-source data alignment within the same period based on synchronous acquisition under a fixed sampling period, avoiding misjudgment caused by sampling lag in different channels, thereby improving the stability of bin entry identification; The switch signal sequence is acquired, and the switching signal sequence is statistically analyzed based on a preset time window that represents the duration of time to determine the switching signal reversal frequency. The preset time window that represents the duration of time refers to the observation length used to statistically analyze the intensity of photoelectric signal jitter. The switching signal reversal frequency refers to the ratio of the number of complete pulse cycles within the time window to the length of the time window, which is used to characterize the degree of abnormal reversal of the photoelectric signal under dust obstruction. To identify whether a photoelectric sensor is in a dust-triggered state, the number of complete pulse cycles from low to high and back to low within a preset time window, representing the duration of the pulse, is specifically used. Divide by the length of the time window used for statistics Its unit is seconds (s), which gives the flip frequency per unit time. The unit is Hz, which can be represented as: This embodiment first quantifies the flip density of the photoelectric signal in the current cycle, captures the high-frequency false triggering pattern, and transforms the characteristic of photoelectric sensors being prone to shaking in dusty environments into identifiable interference features; it can realize the pre-identification of dust interference based on the flip frequency statistics, avoids directly treating abnormal shaking as a box entry event, and prevents premature box sealing. The switching frequency is obtained, and the switching frequency is compared with the dust characteristic threshold to determine whether the dust interference period has been entered. The dust characteristic threshold refers to the judgment boundary that distinguishes normal shading from abnormal high-frequency jitter. Entering the dust interference period means that the field control terminal sets the trigger flag position of the photoelectric signal to a shielded state, only retains its monitoring record, and no longer allows it to directly trigger the box entry judgment. In practice, the shielding state can be written into the status bit in the memory of the field control terminal. Subsequent identification logic only reads this status bit to determine whether to enable the photoelectric branch. This embodiment uses a threshold shielding strategy to accurately suppress abnormal high-frequency flips, so that subsequent judgments are switched to a multi-sensor verification mode instead of continuing to be dominated by a single photoelectric channel. Trigger source switching can be achieved based on dust characteristic thresholds to avoid dust false triggers from continuously affecting scheduling judgments, thereby ensuring the accuracy of identification results. The original strain simulation voltage is obtained, and according to the smoothing filtering rule, the original strain simulation voltage is de-glitched to generate dynamic filtered weight, and the weight step increment is further calculated. Among them, dynamic filtered weight refers to the stable weight value obtained after smoothing the original voltage of the weighing sensor; weight step increment refers to the difference between the current stable weight and the previous observation reference weight, which is used to indicate whether a new effective weight gain has occurred in the box. By calculating and differentiating high-frequency burrs caused by mechanical vibration from the net weight increase caused by actual parts entering the box, specifically by calculating the dynamic filtered weight of the current cycle, the system can effectively differentiate these burrs from the net weight increase caused by actual parts entering the box. With observation delay stable weight before The difference between them yields the weight step increment. , can be represented as: This embodiment captures the continuous pattern of weight increase by combining smoothing filtering and front-to-back differential processing, weakens the interference of high-frequency impact spikes into short-term fluctuations, and extracts the actual steady-state weight gain of the box. In the context of metal product packaging, a common high-frequency interference condition of weighing signals is that transient mechanical high-frequency oscillations occur after the parts fall into the box, and the original voltage may show continuous spike peaks. If the method of comparing only the instantaneous peaks is used, it is easy to split a single drop vibration into multiple false events. This embodiment uses the combination of dynamic filtering weight and weight step increment to participate in subsequent judgment only when the stable weight forms an effective step. Therefore, multiple false triggers caused by mechanical vibration can be compressed into a single candidate event. This embodiment can achieve real weight gain recognition based on filtered weight and weight step increment, avoiding the amplification of mechanical vibration burrs into multiple box entry events. Ultimately, it solves the core technical defect of existing technology that leads to repeated recognition due to the lack of a deburring and net weight gain linkage mechanism. The system acquires real-time current feedback values ​​and weight step increments. Based on the servo gripping current peak threshold, the minimum metal part weight identification threshold, and the synchronization time window, it jointly judges the candidate events for entering the box, determines the actual part entering the box event, and updates the current box status. If the above conditions are not met simultaneously, it is determined that the actual part entering the box event has not been triggered. The servo gripping current peak threshold refers to the current boundary that distinguishes between normal idle travel and gripping load action. The minimum metal part weight recognition threshold refers to the minimum effective weight condition that excludes the influence of slight vibration or stray attachments; the synchronization time window refers to the allowable timing range in which the current peak and weight increase should occur together; the current box status includes at least the current cumulative weight of the box and can be synchronously updated in memory along with the box entry count of this cycle; joint discrimination is used to anchor the robotic arm action information and the box weight increase information to the same event; Specifically, when the real-time current feedback value of the current cycle Greater than the servo capture current peak threshold And the weight step increment The weight of the metal part is greater than the minimum metal part recognition threshold. And both are in the same time window When both events occur simultaneously, determine the logical state result of the actual part entering the box event. , indicates that the event has been triggered; otherwise This means it has not been triggered. This embodiment uses collaborative change discrimination to confirm the entry into the box only when both the peak capture and the net weight increase of the box are met within the synchronous time window. Therefore, it can downgrade the occasional anomaly of a single channel to a multi-condition auxiliary verification item. In this type of workstation, the evaluation criterion is not simply whether a single sensor is sensitive, but whether the consistency of event recognition can be maintained when dust obstruction, drop impact, and rhythm fluctuations coexist. With this embodiment, even if the photoelectric channel is shielded, the judgment basis can still be restored by using the synchronization relationship between current and weight, so that the recognition of the box entry event is closer to the actual delivery behavior. This embodiment can realize the identification of real parts entering the box based on the joint constraints of current peak, net weight gain and synchronization time window, avoid the box status update error caused by single path false triggering, and finally solve the core technical defect of the existing technology that causes scheduling input distortion due to the lack of multi-sensor verification mechanism. Historical scheduling data is acquired, and based on offline statistical results, the historical scheduling data is summarized and processed to generate a packing rule table and a threshold correspondence table, which are then written into the local lookup structure of the field control module. The packing rule table is a deterministic lookup table consisting of condition judgment items and execution result items. The threshold correspondence table is a set of thresholds that are associated with the rule table and are used to uniformly interpret weight ranges, heavy part judgment boundaries, and other working conditions. In practice, the remote dispatch center can send rule table data packets to the field control terminal via industrial Ethernet. After receiving the data, the field control terminal deserializes it and writes it into its local memory, and establishes a direct addressing structure or a binary search structure for quick matching after subsequent events are triggered. This embodiment completes the high-computation rule generation offline, compressing on-site judgment into table lookup and instruction mapping, thus avoiding the computational power squeeze of edge devices under high-cycle conditions. On-site fast matching can be achieved based on the rule table generated offline, avoiding the direct deployment of complex algorithms on edge computing nodes for real-time computation, and ultimately solving the core technical defect of existing technologies that cause control instruction lag due to on-site computation delay. The system acquires real parts loading events and current box status, performs condition matching on the current working conditions according to the packing rule table, generates a scheduling parameter set, and converts the scheduling parameter set into underlying hardware control instructions for direct hardware drive. The scheduling parameter set refers to a set of control parameters used to drive field equipment under the current working conditions, which may include at least the robotic arm trajectory number, inverter frequency derating factor, and load-related control coefficients. The underlying hardware control instructions refer to the execution commands that can be directly written into the target registers of the frequency converter, servo controller, or servo drive. In specific implementation, after the field control terminal confirms that the box entry event is established, it reads the corresponding scheduling parameter group from the local lookup structure and then writes the parameter group into the instruction buffer. Among them, the relevant parameters of the frequency converter can be encapsulated as frequency control word overwrite commands, and the robotic arm trajectory number can be encapsulated as servo program segment call commands. The contents of the buffer after reading are overwritten and updated in the next control cycle. This embodiment directly maps the rule matching results to register overwriting and program segment calls, enabling the scheduling results to have definite physical execution actions; it can realize the complete action from identification to execution based on the packing rule table and field instruction mapping, avoiding the scheduling results from being decoupled from the underlying hardware; After the standard container is sealed, the scheduling quality evaluation parameters are obtained. Based on the deviation calculation formula, the judgment parameters of the packing rule table are adjusted to generate an updated packing rule table. The scheduling quality evaluation parameters include at least the physical cycle time fluctuation rate, empty container rate, and the number of servo overload alarms. The deviation calculation formula refers to the comprehensive evaluation function after weighting the above evaluation parameters according to preset weights; after the box is sealed, the field control terminal uploads the evaluation parameters within the current standard box sealing cycle to the remote dispatch center via industrial Ethernet; after the remote dispatch center completes the parameter adjustment, it reissues the updated box packing rule table; This embodiment uses evaluation results to drive rule correction, which allows the judgment parameters to be updated as historical performance changes, avoiding long-term deviation of fixed rules from actual working conditions; the rule table can be updated based on evaluation parameters and deviation calculation formula, avoiding continuous accumulation of errors by fixed parameters; In actual deployment, the remote dispatch center can be located on the dispatch server side, and the field control terminal can be located on the packaging production line side. Multi-source physical data continuously flows into the sampling buffer of the field control terminal according to the system cycle. After the real part entering the box event is identified, the local table lookup and control command generation are triggered to complete the field movement adjustment of the conveyor belt and robotic arm. After a standard box is sealed, the evaluation parameters are uploaded to the remote scheduling center for rule correction. This forms a complete application process from data collection, event recognition, rule matching, hardware execution to rule update, which can be directly used in the standard box packaging scenario of metal products with multiple production lines.

[0016] Furthermore, the historical data analysis module is used to generate a packing rule table, including: Extract order parameters from historical scheduling data and use K-means clustering algorithm to generate packing pattern categories; Order parameters include the density, volume, and arrival frequency of the metal parts; For each packing mode category, pre-calculate the packing posture and delivery order that satisfy the anti-eccentric load constraint, and encapsulate them into a packing rule table; Among them, the condition judgment item of the packing rule table includes the physical characteristic conditions of the current cumulative weight range of the box and the number of consecutive heavy items arriving, and the execution result item includes the set of robotic arm trajectory numbers and the engineering instruction set of the frequency derating coefficient of the inverter. The field control module is used to match the scheduling parameter set, including: When a real part is identified as being placed in the box, the current cumulative weight of the box is extracted, and the average servo current estimate of multiple consecutive parts placed in the box is calculated based on the real-time current feedback value to construct an index key value. Input the index key value into the packing rule table for condition matching; if the number of consecutive heavy parts arriving that continuously meet the average servo current estimation value greater than the heavy part current threshold reaches the preset number, the judgment logic jumps to the heavy load anti-eccentricity branch, matches and outputs the scheduling parameter group under the current working condition. If the preset quantity is not reached, the regular scheduling branch is executed and the corresponding scheduling parameter group is output. The scheduling parameter group includes the corresponding robot arm trajectory number extracted from the robot arm trajectory number set, the inverter frequency derating factor, and the torque limiting amplification factor. The field control module is used to convert scheduling parameter sets into low-level hardware control instructions for direct hardware driver execution, including: The target frequency is calculated by combining the conveyor belt reference operating frequency with the inverter frequency derating factor through the field industrial bus protocol, and the inverter frequency control word register is overwritten. The robot arm trajectory number is parsed into a servo controller program segment call instruction, and the kinematic interpolation path is switched. This instruction maps the torque limit amplification factor to the overwrite instruction of the servo driver's torque limit register.

[0017] Based on the above implementation method, this implementation method further explains the generation method of the packing rule table, the on-site table lookup and matching method, and the mapping method of scheduling parameter group to hardware control instructions; The inputs to this section are historical scheduling data and real-time container status. The processing mechanism is to first generate a searchable packing rule table in the remote dispatch center, and then the field control terminal uses the current cumulative weight of the container and the characteristics of continuous heavy items to complete the condition matching. The output is the low-level control commands related to the robotic arm trajectory, the inverter target frequency, and the servo torque limit. In this way, the computationally complex packing decision can be transformed into rule matching that can be executed quickly on site. Acquire historical scheduling data, and perform clustering processing on the historical scheduling data according to order parameters to generate packing pattern categories; among them, order parameters refer to the basic inputs used to characterize the differences in packing of metal parts, including at least metal part density, volume and arrival frequency; metal part density is used to reflect the mass distribution corresponding to unit volume, volume is used to reflect the space occupied by a single piece in the box, and arrival frequency is used to reflect the tightness of the supply cycle of materials within the system cycle. Packing pattern category refers to a set of similar working conditions formed after clustering, which is used to merge a large number of historical orders into a limited number of categories. In specific implementation, the remote scheduling center extracts the above order parameters from the historical scheduling database, forms a sample set, and then uses the K-means clustering algorithm for classification. The clustered category labels and corresponding parameter boundaries are written into the rule generation process. This embodiment captures common patterns among orders by clustering and merging density, volume, and arrival frequency, transforming discrete and unstructured historical records into reusable packing pattern categories. Historical orders can be categorized based on order parameters, avoiding repeated calculations of packing schemes under similar working conditions on-site, and ultimately solving the core technical defect of existing technologies that result in excessive computational burden due to solving each order individually. Obtain the packing mode category, pre-calculate each packing mode category according to the anti-eccentric load constraint, determine the packing posture and delivery order, and encapsulate it into a packing rule table; among them, the anti-eccentric load constraint refers to the constraint conditions that need to control the force and center of gravity offset of the standard box during the loading process; the packing posture refers to the orientation of the metal parts when they are put into the box; the delivery order refers to the sequential arrangement of multiple parts within a box cycle. The packing rule table consists of condition judgment items and execution result items. The condition judgment items include at least the current cumulative weight range of the box and the number of consecutive heavy items arriving. The execution result items include at least the set of robotic arm trajectory numbers and the frequency derating factor of the inverter. In addition, for special working conditions of heavy load and anti-eccentric load, the execution result items also have a built-in corresponding torque limiting amplification factor. In practice, the remote dispatch center pre-calculates several executable schemes that meet the anti-eccentric load requirements for each packing mode category, and writes the scheme into the data packet in the form of a lookup table. Since the required overload protection level is different under different trajectories and cycle times, the system packages the corresponding torque limit amplification factor into the engineering instruction set to ensure the integrity of subsequent parameter matching. This embodiment completes the complex calculations related to anti-eccentric loading in advance and converts them into a rule table with clear condition items and result items, so that the field end is no longer under the pressure of three-dimensional binning solution; In this type of packaging system, a common extreme condition is the arrival of consecutive heavy items. If a single fixed robotic arm trajectory is still used, the local off-center load on the box and the sudden change in the load on the robotic arm will be amplified at the same time. This embodiment pre-calculates the posture and sequence under different weight ranges and continuous heavy item conditions, which can map the risk of heavy item concentration to the rule branch changes in advance, thereby reducing on-site emergency adjustments. This embodiment can form a rule table based on the pre-calculated packing posture and delivery order, avoiding the need for complex anti-eccentric load judgment to be carried out in real time on site, and ultimately solving the core technical defect of the existing technology that causes scheduling delay due to the lack of pre-calculated rules; The system acquires real part loading events and current cumulative weight of the box. Based on the real-time current feedback values ​​of multiple consecutive loaded parts, it constructs an index for the current operating condition and generates index key values. The average servo current estimate of multiple consecutive loaded parts refers to the estimated result after averaging the servo current feedback in several consecutive confirmed loading events. It is used to represent the overall mass distribution trend of the part load exceeding the limit within this time interval. The number of consecutive heavy parts arriving refers to the statistical quantity of parts that continuously meet the average servo current estimation value greater than the heavy part current threshold. The index construction process incorporates the current box volume occupancy rate and whether the quality of the recent sequence material exceeds the calibration tolerance into the table lookup basis. That is, by combining the current box cumulative weight, the average servo current estimation value of multiple consecutive box-entry parts, and the number of consecutive heavy parts arriving, a multi-dimensional index key value is formed. In practice, after each confirmation of box entry, the field control terminal extracts the current cumulative weight of the box and writes the current feedback into the local sliding cache. It then calculates the average servo current estimate according to a preset number window. Simultaneously, based on the logical judgment result of this estimate and the heavy item current threshold, it updates the continuous heavy item count, i.e. the number of continuous heavy items arriving. Finally, the above multi-dimensional features are packaged together to form an index key value, so that the condition dimension of the lookup index is strictly consistent with the condition judgment item of the packing rule table. This embodiment captures the implicit relationship between weight distribution stages and recent load trends by coordinating the judgment of cumulative weight and average servo current, transforming the limitation of insufficient sensitivity of single-dimensional judgment into the advantage of multi-condition indexing; it can build index key values ​​based on cumulative weight, average servo current and continuous heavy item count, avoiding the coarse branch selection caused by relying on a single weight condition, and ultimately solving the core technical defect of inaccurate matching caused by insufficient table lookup conditions in the existing technology. Obtain the index key value, perform condition matching on the index key value according to the packing rule table, and determine the scheduling parameter group; among them, the heavy component current threshold refers to the current judgment boundary used to distinguish between ordinary components and heavy components; the preset quantity refers to the number of consecutive heavy components required to determine whether to enter the heavy load anti-eccentric load branch; the scheduling parameter group includes at least the robot arm trajectory number, the inverter frequency derating factor, and the torque limiting amplification factor; In practice, the field control terminal reads the matching result from the local direct addressing structure or the binary search structure; if the number of consecutive heavy parts reaches the preset number, the logic jumps to the heavy load anti-eccentricity branch and outputs the corresponding scheduling parameter group; if it does not reach the preset number, it enters the regular scheduling branch and outputs the corresponding scheduling parameter group. This embodiment triggers branch jumps by counting heavy items continuously, making the rule selection more closely match the load changes and avoiding mistreating the sudden concentrated arrival of heavy items as ordinary cycle fluctuations. In the context of metal product packaging, an important benchmark for evaluating the effectiveness of on-site rules is whether the off-center loading and cycle time can still be controlled when encountering continuous heavy loads. If a single conventional branch is still used, continuous heavy loads will cause the load on the robotic arm and the weight distribution inside the box to exceed the upper limit of the control threshold at the same time. After introducing the heavy-load anti-eccentricity branch in this embodiment, the trajectory and cycle time parameters can change synchronously, thereby reducing the scheduling mismatch under this type of working condition; this embodiment can achieve scheduling parameter group matching based on index key value lookup table, avoiding the direct application of ordinary working condition rules to heavy-load working conditions, and ultimately solving the core technical defect of the existing technology that increases the risk of eccentricity and overload due to untimely branch switching; The scheduling parameter group is obtained, and the scheduling parameter group is mapped with instructions according to the field industrial bus protocol to generate and issue underlying hardware control instructions; among them, the frequency derating factor of the frequency converter is used to reduce the material arrival speed based on the reference operating frequency of the conveyor belt. The robotic arm trajectory number is used to select the preset kinematic interpolation path in the servo controller; the torque limit amplification factor is used to increase the upper limit of the output torque allowed by the servo drive when the load increases suddenly; the target frequency is calculated to convert the upper-level cycle adjustment into the frequency setting value that the frequency converter can execute, that is, to obtain the target frequency by multiplying the conveyor belt reference operating frequency by the frequency converter frequency derating factor. In practice, the field control terminal encapsulates the target frequency into a frequency control word overwrite command via the field industrial bus protocol and writes it into the frequency control word register of the frequency converter; it parses the robotic arm trajectory number into a program segment call instruction of the servo controller and writes it into the corresponding call register or instruction buffer; it maps the torque limit amplification factor to the torque limit register overwrite value of the servo drive for temporary torque amplification under the current working condition; each instruction takes effect immediately after being written within the control cycle and can be overwritten by the new scheduling parameter group in the next cycle; This embodiment anchors each scheduling result to a specific register or program segment call action, so that cycle slowdown, trajectory switching and torque adjustment all have clear execution paths; it can realize the conversion of scheduling parameter groups to hardware actions based on field industrial bus protocol and register overwrite mechanism, avoid scheduling decisions remaining at the logical level and unable to be implemented, and ultimately solve the core technical defect of insufficient production line action response caused by the lack of direct drive interface in the existing technology. In practice, the remote dispatch center first completes historical clustering and rule table generation, and then sends the rule table to the field control terminal; the field control terminal updates the current cumulative weight of the container after each confirmation of container entry. Average servo current estimation value and the number of consecutive heavy items arriving Based on this, the table lookup is completed, and the calculated target frequency, trajectory call, and torque limit overwrite instructions are sent to the frequency converter, servo controller, and servo drive. As a result, new box entry data is constantly being entered into the local lookup process, and new control commands are constantly being applied to the conveying and gripping actions, making it suitable for deployment in metal product packaging production lines with high requirements for cycle continuity and anti-eccentric load.

[0018] Furthermore, the feedback adjustment module is used to update the rules, including: calculating scheduling quality evaluation parameters, which include physical beat fluctuation rate, empty box rate calculated based on final static weight, and number of servo overload alarms; Based on preset weighting constants, the physical cycle fluctuation rate, empty box rate, and servo overload alarm count are weighted and calculated to construct a deviation calculation formula; With the goal of minimizing the deviation calculation formula, the judgment thresholds and corresponding output instructions of each condition in the packing rule table are adjusted by the gradient descent optimization method, and the updated packing rule table is generated and issued. The system adopts a hierarchical architecture of master station and slave station, in which: The historical data analysis module and the feedback adjustment module are deployed in the remote dispatch center; The data acquisition module, signal verification and box entry identification module, and field control module are deployed on the field control terminal that is close to the physical production line. The remote dispatch center and the field control terminal establish a connection through industrial Ethernet. The remote dispatch center sends the packing rule table and threshold correspondence table to the field control terminal, and the field control terminal uploads the scheduling quality evaluation parameters to the remote dispatch center. The field control terminal establishes a direct addressing or binary search structure in memory to complete the conversion of complex multivariate scheduling logic into a fixed search memory table structure.

[0019] Based on the above implementation method, this implementation method further describes the rule update and system deployment structure; the input of this part is the scheduling quality evaluation parameter after a standard box is sealed, the processing mechanism is that the remote scheduling center constructs a deviation calculation formula based on the evaluation parameter, and corrects the judgment threshold and output instructions in the rule table through the gradient descent optimization method; the field control terminal is responsible for receiving the rule table, storing it in memory and searching it quickly; thus, the rule table is not fixed, but is gradually adjusted according to the production line operation results; The system acquires the real-time linear speed, final static weight, and servo overload alarm records of the conveyor belt after the standard box is sealed. Based on the preset evaluation rules, the sealing cycle is statistically analyzed to generate scheduling quality evaluation parameters. Among them, the physical cycle fluctuation rate refers to the ratio of the standard deviation to the mean of the real-time linear speed of the conveyor belt within the box cycle, which is used to reflect the stability of the cycle. Empty container rate refers to the container utilization rate calculated based on the final static weight, used to reflect whether premature container sealing or insufficient utilization of the container has occurred; Servo overload alarm count refers to the overload alarm count generated by the servo drive within the container cycle, used to reflect the load pressure of the actuator under this strategy. In practice, after the box is sealed, the field control terminal reads the accumulated linear velocity samples, final static weight and alarm records of the box within the cycle from the local cache. After the calculation is completed, it generates an evaluation parameter data packet and uploads it to the remote dispatch center via industrial Ethernet. This embodiment quantifies cycle time, packing utilization, and drive load in a unified manner, so that rule adjustments have a clear basis, rather than relying on qualitative manual judgment without quantitative standards; it can generate scheduling quality evaluation parameters based on packing cycle statistics, avoiding the lack of objective reference for rule updates, and ultimately solving the core technical defect of existing technologies that lead to ambiguous optimization directions due to unclear evaluation mechanisms. Obtain scheduling quality evaluation parameters, and calculate the deviation by weighting each evaluation parameter according to the preset weight constant. The preset weight constant is used to balance the influence of different evaluation parameters in rule updates. The weight of physical cycle fluctuation rate is used to reflect the importance of production line cycle stability, the weight of empty box rate is used to reflect the importance of box utilization, and the weight of servo overload alarm times is used to reflect the importance of equipment load safety. The purpose of the deviation calculation formula is to compress the quality results of different dimensions into a comparable single evaluation quantity. That is, the physical clock fluctuation rate, empty box rate and servo overload alarm number are multiplied by their respective preset weight constants and then summed to obtain a comprehensive deviation calculation result. This embodiment incorporates multiple evaluation objectives into the same calculation framework through weighted calculation, so that rule updates are no longer overly sensitive to a certain indicator, but can achieve a quantitative balance between cycle time, empty boxes, and overload. In this type of system, if only a low empty box rate is pursued, it may lead to an increase in overload alarms; if only a low overload is pursued, the conveying cycle time may be overly conservative. This embodiment unifies the three types of results into a single target quantity through a deviation calculation formula, making it more suitable as the basis for subsequent threshold adjustments. The deviation calculation formula can be constructed based on a preset weight constant, avoiding the bias caused by a single indicator, and ultimately solving the core technical defect of the existing technology that the optimization results are one-sided due to the lack of a comprehensive evaluation function. Obtain the deviation calculation formula, and adjust the judgment thresholds and corresponding output instructions for each condition in the packing rule table according to the objective of minimizing the deviation calculation formula. Generate an updated packing rule table and issue it. The judgment thresholds may include at least the heavy component current threshold, the weight range boundary, and other thresholds related to branch selection. The corresponding output instructions may include at least the inverter frequency derating factor, the selection result of the robotic arm trajectory number set, and the output settings related to the execution intensity; the physical meaning of the gradient descent optimization method here is to gradually modify the rule table parameters along the direction that reduces the output result of the comprehensive deviation calculation formula, so that the updated rules are more in line with the recent production line performance. In practice, considering that the judgment threshold and instruction mapping in the rule table are often discrete or step-distributed, making it impossible to directly obtain the analytical gradient, the remote dispatch center uses the numerical difference approximation method to calculate the local gradient. That is, during each update, positive and negative perturbations with preset step sizes are applied to the current parameter to be adjusted, such as the current threshold of heavy components. The historical multi-source physical data collected recently is used for fast offline playback to obtain the deviation calculation results before and after the perturbation. The approximate gradient of the cost value for the discrete parameter is obtained by finite difference. Then, the parameter is updated along the negative gradient direction with a fixed step size. After receiving the evaluation parameters for the multi-box cycle and completing the above iterative adjustments, the remote dispatch center generates a new packing rule table and threshold correspondence table, and sends it to the field control terminal via industrial Ethernet. After receiving it, the field control terminal first marks the old rule table as to be replaced, and then writes the new rule table into the direct addressing or binary search structure in memory to complete the switch. This embodiment uses targeted numerical difference approximation optimization to directly apply the evaluation results to the rule conditions and output instructions, overcoming the problem of non-differentiability of discrete control parameters and enabling the rule table to be corrected according to the running results. The rule table can be updated based on the gradient descent optimization method, avoiding long-term mismatch of production line conditions with fixed thresholds, and ultimately solving the core technical defect of existing technology that causes the scheduling quality to gradually decline due to static rules. The deployment nodes of the remote dispatch center and field control terminals are obtained. According to the hierarchical architecture of master station and slave station, each module is allocated and an industrial Ethernet connection and local lookup structure are established. Among them, the remote dispatch center, as the master station, deploys the historical data analysis module and the feedback adjustment module, and undertakes the tasks of historical statistics, rule generation and rule update. The field control terminal, acting as a slave station, deploys a data acquisition module, a signal verification and box entry identification module, and a field control module, undertaking real-time sampling, event identification, and hardware driving tasks; the industrial Ethernet connection is used to carry the data exchange of rule tables and evaluation parameters; the local lookup structure refers to the direct addressing or binary lookup structure established in memory by the field control terminal, which is used to convert complex scheduling logic into a fixed lookup memory table structure; In practice, the data packets sent by the remote dispatch center may contain a packing rule table and a threshold correspondence table. After receiving the data packets, the field control terminal will deserialize them and write them into the memory lookup area. The data packets uploaded by the field control terminal may contain scheduling quality evaluation parameters formed according to the packing cycle. This embodiment adopts a hierarchical deployment of master and slave stations. The former is responsible for rule generation and updating, while the latter is responsible for real-time table lookup and execution. This avoids both excessive on-site computing power burden and direct remote intervention in high-frequency control. Based on the hierarchical architecture of master and slave stations, rule generation, rule distribution, on-site lookup, and evaluation feedback can be coordinated, avoiding the mixing of control and computing responsibilities. Ultimately, it solves the core technical defect of existing technologies that make it difficult to balance real-time performance and computational load due to unreasonable deployment structure. During application, the field control terminal continuously collects and identifies box entry events according to the system cycle, and performs scheduling control according to the local lookup structure; after each standard box is sealed, the evaluation parameters are packaged and uploaded to the remote dispatch center. The remote dispatch center adjusts the rules according to the deviation calculation formula and reissues the updated rule table to the field control terminal. Once the new rule table is entered into the memory lookup structure, it is used for table lookup control in subsequent box cycles. Through this deployment method, the metal product packaging schedule can be continuously updated without changing the basic hardware connection relationship on site.

[0020] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A scheduling optimization system for standard box packaging of metal products based on data mining, characterized in that, The system includes: The data acquisition module is used to receive multi-source physical data from the packaging station. The multi-source physical data includes the switch sequence of the diffuse reflection photoelectric sensor that characterizes the spatial occlusion state, the original strain analog voltage output by the bottom weighing sensor, and the real-time current feedback value of the robotic arm servo motor. The signal verification and box entry identification module is used to perform signal denoising based on the multi-source physical data, and when dust interference conditions are detected, it shields the triggering effect of the switch sequence and switches to multi-sensor verification mode. It combines the original strain simulation voltage and the real-time current feedback value to identify the actual part entering the box event and updates the current box status including the current box cumulative weight. The historical data analysis module is used to acquire historical scheduling data, perform offline statistics based on the historical scheduling data, generate a packing rule table containing judgment parameters and a threshold correspondence table, and send it to the field control module. The field control module is used to match the scheduling parameter group in the packing rule table according to the current box status including the current box's cumulative weight when the actual part is detected to be put into the box, and convert the scheduling parameter group into the underlying hardware control instructions to execute the hardware direct drive. The feedback adjustment module is used to collect scheduling quality evaluation parameters after a standard container is sealed, and adjust the judgment parameters of the packing rule table according to the deviation calculation formula constructed based on the scheduling quality evaluation parameters to update the rules.

2. The data mining-based scheduling optimization system for standard box packaging of metal products according to claim 1, characterized in that, The data acquisition module is specifically used for: A real-time data set is constructed by collecting the multi-source physical data in real time with a fixed sampling period. The real-time data set includes: the switching sequence of the diffuse reflection photoelectric sensor representing the spatial occlusion state, the original strain analog voltage output by the bottom weighing sensor, the real-time current feedback value of the robotic arm servo motor representing the transient load torque of the current grasping action, and the real-time linear speed of the conveyor belt representing the current material supply cycle.

3. The data mining-based scheduling optimization system for standard box packaging of metal products according to claim 2, characterized in that, The signal verification and box entry identification module is used to identify actual part box entry events, including: Calculate the switching frequency of the switching sequence within a preset time window representing the time length; When the switching frequency is greater than a preset dust characteristic threshold representing abnormal switching density, it is determined that a dust interference period has been entered, the triggering effect of the switching sequence is blocked, and the system switches to multi-sensor verification mode; when the switching frequency is less than or equal to the preset dust characteristic threshold, it is determined that a dust interference period has not been entered, and the triggering effect of the switching sequence is maintained. In the multi-sensor verification mode, the original strain simulation voltage is smoothed and filtered to output the dynamic filtered weight, and the weight step increment is calculated. When the real-time current feedback value is greater than the preset servo grab current peak threshold, and the weight step increment is greater than the minimum metal part weight recognition threshold, and both occur simultaneously within the preset synchronization time window, it is determined that the real part entering the box event is triggered, and the current box state containing the current box's cumulative weight is updated.

4. The data mining-based scheduling optimization system for standard box packaging of metal products according to claim 1, characterized in that, The historical data analysis module is used to generate a packing rule table, including: Order parameters are extracted from historical scheduling data, and K-means clustering algorithm is used to generate packing pattern categories; the order parameters include metal part density, volume, and arrival frequency. For each packing mode category, pre-calculate the packing posture and delivery order that satisfy the anti-eccentric load constraint, and encapsulate them into the packing rule table; The condition judgment item in the packing rule table includes physical characteristic conditions such as the current cumulative weight range of the box and the number of consecutive heavy items arriving, and the execution result item includes the set of robotic arm trajectory numbers and the engineering instruction set of the frequency derating coefficient of the inverter.

5. The data mining-based scheduling optimization system for standard box packaging of metal products according to claim 4, characterized in that, The field control module is used to match the scheduling parameter group, including: When the actual part is detected to be placed in the box, the current cumulative weight of the box is extracted, and the average servo current estimate of multiple consecutive parts placed in the box is calculated based on the real-time current feedback value to construct an index key value; The index key value is input into the packing rule table for condition matching; If the number of consecutive heavy components arriving that continuously meet the average servo current estimation value greater than the heavy component current threshold reaches a preset number, the judgment logic jumps to the heavy load anti-eccentricity branch, matches and outputs the scheduling parameter group under the current working condition; if the preset number is not reached, the regular scheduling branch is executed and the corresponding scheduling parameter group is output. The scheduling parameter set includes the corresponding robotic arm trajectory number extracted from the set of robotic arm trajectory numbers, the frequency derating factor of the inverter, and the torque limiting amplification factor.

6. The data mining-based scheduling optimization system for standard box packaging of metal products according to claim 5, characterized in that, The field control module is used to convert the scheduling parameter group into low-level hardware control instructions for direct hardware drive execution, including: The target frequency is calculated by combining the reference operating frequency of the conveyor belt with the frequency derating factor of the inverter through the field industrial bus protocol, and the frequency control word register of the inverter is overwritten. The robotic arm trajectory number is parsed into a servo controller program segment call instruction, and the kinematic interpolation path is switched. The torque limiting amplification factor is mapped to an overwrite instruction for the torque limiting register of the servo driver.

7. The data mining-based scheduling optimization system for standard box packaging of metal products according to claim 1, characterized in that, The feedback adjustment module is used for rule updates, including: Calculate the scheduling quality evaluation parameters, which include physical cycle fluctuation rate, empty box rate calculated based on final static weight, and number of servo overload alarms; The deviation calculation formula is constructed by weighting the physical beat fluctuation rate, the empty box rate, and the number of servo overload alarms according to a preset weight constant. With the goal of minimizing the deviation calculation formula, the judgment thresholds and corresponding output instructions of each condition in the packing rule table are adjusted by the gradient descent optimization method, and an updated packing rule table is generated and issued.

8. The data mining-based scheduling optimization system for standard box packaging of metal products according to claim 1, characterized in that, The system is deployed using a hierarchical architecture of master station and slave station, wherein: The historical data analysis module and the feedback adjustment module are deployed in a remote dispatch center; The data acquisition module, the signal verification and box entry identification module, and the field control module are deployed on the field control terminal that is close to the physical production line; The remote dispatch center and the field control terminal establish a connection through an industrial Ethernet. The remote dispatch center sends the packing rule table and threshold correspondence table to the field control terminal, and the field control terminal uploads the scheduling quality evaluation parameters to the remote dispatch center. The field control terminal establishes a direct addressing or binary search structure in memory to complete the conversion of complex multivariate scheduling logic into a fixed lookup memory table structure.

Images