Method and system for controlling a conveyor for the production of medical packaging bags

Abstract

The present application relates to the field of automation control technology, and more particularly to a conveyor control method and system for medical packaging bag production, comprising synchronously acquiring bag deformation state data and load torque fluctuation data, realizing multi-source heterogeneous sensing, providing comprehensive and complementary information basis for subsequent analysis; generating dynamic friction fingerprints by utilizing cross-correlation analysis, improving the detection sensitivity of small friction changes; dividing the conveying section and constructing a dynamic conveying adaptation map, realizing spatial and time sequence dual-dimension evaluation, quantitatively obtaining the conveying efficiency coefficient of each section, and providing clear basis for differentiated regulation; generating differentiated speed regulation and correction instructions according to the conveying efficiency coefficient, effectively suppressing the instability trend and reducing the failure rate; updating the dynamic conveying adaptation map according to the execution result, continuously optimizing the control strategy with the change of working conditions, and guaranteeing the long-term operation stability and adaptability.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, and in particular to a control method and system for a conveyor in the production of medical packaging bags. Background Technology

[0002] In the production of medical packaging bags, belt conveyors are the core conveying devices that connect the upstream bag-making station with the downstream sealing and cutting station. Their operational stability directly determines the forming accuracy of the final product.

[0003] However, medical packaging bags are ultra-thin, flexible film products, and in actual continuous production, there are many unavoidable operational disturbances: the surface roughness of packaging films varies slightly between different batches; fluctuations in workshop temperature and humidity alter the frictional characteristics between the conveyor belt and the bag; and localized wear of the conveyor belt after long-term operation can lead to uneven load distribution. Existing methods lack a correlation sensing mechanism between the bag's microscopic deformation and the driving load fluctuations, failing to capture the dynamic frictional state of the bag-belt coupling. When implicit slippage occurs between the bag and the conveyor belt, the instability trend cannot be identified in advance, easily leading to problems such as bag misalignment. This can result in batches of defective products due to positioning deviations at downstream workstations, or even bag tearing and jamming, triggering a complete line shutdown. In medical packaging production scenarios requiring high cleanliness and high yield rates, the reliability and adaptability of traditional conveyor control modes can no longer meet the demands of large-scale continuous production. Therefore, this invention proposes a conveyor control method and system for medical packaging bag production. Summary of the Invention

[0004] The purpose of this invention is to solve the problems in the background art, and to propose a conveyor control method and system for the production of medical packaging bags.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of this invention provides a conveyor control method for the production of medical packaging bags, comprising: S1. Obtain the deformation status data of the medical packaging bag to be conveyed output by the optical contour detection module in the packaging equipment, and simultaneously obtain the real-time load torque fluctuation data of the conveyor drive unit. S2. Analyze the cross-correlation response between bag deformation state data and real-time load torque fluctuation data to generate dynamic friction fingerprint; S3. Divide the conveying into multiple conveying sections and, based on dynamic friction fingerprints, construct a dynamic conveying adaptation map between the bag movement posture and driving energy consumption of the conveyor in different conveying sections. The dynamic conveying adaptation map is used to obtain the conveying efficiency coefficient. Among them, the conveyor is a belt conveyor device in packaging equipment responsible for conveying medical packaging bags from the upstream workstation to the downstream workstation. S4. Based on the conveying efficiency coefficient obtained from the dynamic conveying adaptation map, generate and execute differentiated speed regulation commands and correction commands for the conveyor drive unit. S5. Update the dynamic conveying adaptation map based on the execution results of the differentiated speed regulation command and the correction command.

[0006] Further, S1 includes: The optical contour detection module acquires the bag edge contour sequence under a preset strobe light source. After removing noise by median filtering of the bag edge contour sequence, the bag deformation state data is generated by analyzing the pixel-level displacement of the bag edge contour sequence. Meanwhile, a torque sensor embedded in the shaft of the conveyor drive motor collects real-time load torque fluctuation data at a rate higher than the sampling frequency of the optical profile detection module.

[0007] Further, S2 includes: Align the bag deformation data with the real-time load torque fluctuation data on a unified time axis, and extract the synchronous segments of the two within the current conveying cycle. In the synchronous segment, the cross-correlation function between the bag deformation state data and the real-time load torque fluctuation data is calculated using the cross-correlation analysis window, and the peak delay time and peak amplitude of the cross-correlation function are extracted. A dynamic triboelectric fingerprint is formed by jointly analyzing the peak delay time and peak amplitude.

[0008] Furthermore, the step of calculating the cross-correlation function between the bag deformation state data and the real-time load torque fluctuation data using a cross-correlation analysis window, and extracting the peak delay time and peak amplitude of the cross-correlation function, includes: Within the cross-correlation analysis window, the first time series of bag deformation state data and the second time series of real-time load torque fluctuation data are obtained, and mean-reduction processing is performed on them respectively to obtain the first zero-mean sequence and the second zero-mean sequence. Calculate the normalized cross-correlation function between the first zero-mean sequence and the second zero-mean sequence to obtain the correlation coefficient sequence; Peak search is performed on the correlation coefficient sequence to extract the correlation coefficient values ​​corresponding to all local peak points and the time delay corresponding to those correlation coefficient values; The correlation coefficient value corresponding to each local peak point is compared with the preset correlation threshold. False peak points with correlation coefficient values ​​lower than the correlation threshold are removed, and the effective peak points with the largest correlation coefficient values ​​are retained. The time delay of the effective peak points is extracted as the peak delay time, and the correlation coefficient of the effective peak points is extracted as the peak amplitude.

[0009] Further, S3 includes: The conveying path formed by the conveyor inside the packaging equipment is divided into multiple conveying sections, and a set of benchmark dynamic parameters is pre-set for each conveying section; For each transport section, the dynamic friction fingerprint is mapped to a phase lag factor characterizing the movement posture of the bag in that transport section and an amplitude attenuation factor characterizing the driving energy consumption through a preset nonlinear mapping table. By using phase lag factor and amplitude attenuation factor to perform multidimensional state reconstruction and spatiotemporal characteristic analysis, a multidimensional deviation measure reflecting the degree of deviation of the transport state is obtained. Based on the evolution of phase lag factor, amplitude attenuation factor and multidimensional deviation metric over time, a dynamic transport adaptation map is constructed.

[0010] Furthermore, the multidimensional state reconstruction and spatiotemporal feature analysis using phase lag factor and amplitude attenuation factor yields a multidimensional deviation measure reflecting the degree of deviation from the transport state, including: For each transport section, the phase lag factor and amplitude attenuation factor are reconstructed into a two-dimensional state vector, and the transition entropy of the two-dimensional state vector between multiple consecutive state transition analysis windows is calculated as a state complexity index. Meanwhile, dynamic time bending analysis was performed on the phase lag factor and the reference dynamic parameters of the transport section to obtain a phase deviation index that reflects the degree of phase distortion. The state complexity index and the phase deviation index are used together as a multidimensional deviation measure to characterize the degree of deviation of the current state.

[0011] Furthermore, the construction of a dynamic transport adaptation map based on the evolution of phase lag factor, amplitude attenuation factor, and multidimensional deviation metric over time series includes: The phase lag factor, amplitude attenuation factor and multidimensional deviation metric generated for each transport segment in the current transport cycle and multiple consecutive historical transport cycles are arranged in chronological order to form a time series feature matrix corresponding to each transport segment. An evolutionary truncation window of preset length slides across the time series feature matrix to extract evolutionary feature tensors within multiple local time windows; wherein, a local time window is a continuous time segment extracted by the evolutionary truncation window during the sliding process; A two-dimensional matrix framework is established with the transport segment as the first dimension and the timestamp of the local time window as the second dimension. Multidimensional temporal trend analysis is performed on each evolutionary feature tensor to extract the long-term memory decay slope of the phase lag factor and the fluctuation divergence rate of the amplitude decay factor within the local time window. Based on the coupling relationship between the long-term memory decay slope and the fluctuation divergence rate, the transmission efficiency coefficient of this local time window is determined. The transport efficiency coefficient is filled into the corresponding transport segment and time window position in the two-dimensional matrix frame to form a dynamic transport adaptation map with the transport efficiency coefficient as the matrix element.

[0012] Further, S4 includes: Based on the dynamic transport adaptation map, the multidimensional deviation metric corresponding to each transport section is read, and the transport instability risk index of each transport section is calculated to identify the target transport section that leads to the trend of transport instability. The conveying instability risk index of the target conveying section is used as an input signal and sent to the proportional-integral controller, which calculates the speed regulation compensation amount based on the sign and magnitude of the conveying instability risk index. Generate differentiated speed control commands based on the speed compensation amount; Simultaneously, the phase deviation index of the transport instability risk index in the target transport section is read, and the degree of deviation exceeding the limit is analyzed. Based on the degree of deviation exceeding the limit, the strength value of the correction command is calculated, and a correction command containing the strength value is generated.

[0013] Further, S5 includes: After executing the differentiated speed control command and the correction command, the bag deformation state data and real-time load torque fluctuation data of the new cycle are reacquired. Based on the bag deformation state data of the new cycle and the real-time load torque fluctuation data, a new dynamic friction fingerprint is regenerated. The new dynamic triboelectric fingerprint is compared with the historical dynamic triboelectric fingerprint, and the Euclidean distance between the two is calculated as a measure of difference to obtain the correction amount of the dynamic transport adaptation map. The correction amount is fused with the phase lag factor and amplitude attenuation factor in the dynamic transport adaptation map of the previous transport cycle. The phase lag factor and amplitude attenuation factor are then updated with weights. Based on the updated phase lag factor and amplitude attenuation factor, the multidimensional deviation metric is recalculated, and the dynamic transport adaptation map is finally updated.

[0014] A second aspect of the present invention provides a conveyor control system for the production of medical packaging bags, comprising: Bag Deformation and Torque Data Acquisition Module: Acquires the bag deformation status data of the medical packaging bag to be conveyed, output by the optical contour detection module in the packaging equipment, and simultaneously acquires the real-time load torque fluctuation data of the conveyor drive unit. Dynamic friction fingerprint generation module: Analyzes the cross-correlation response relationship between bag deformation state data and real-time load torque fluctuation data to generate dynamic friction fingerprint; Dynamic transport adaptation map construction module: Divide into multiple transport sections and construct a dynamic transport adaptation map based on dynamic triboelectric fingerprint. The dynamic transport adaptation map is used to obtain the transport efficiency coefficient. Speed ​​regulation and correction command generation module: Based on the conveying efficiency coefficient obtained from the dynamic conveying adaptation map, it generates and executes differentiated speed regulation and correction commands for the conveyor drive unit. Dynamic transport adaptation map update module: Updates the dynamic transport adaptation map based on the execution results of differentiated speed control commands and correction commands.

[0015] Compared with existing technologies, the beneficial effects of the conveyor control method and system for medical packaging bag production provided by this invention are as follows: 1) The deformation sequence of the bag edge is collected by the stroboscopic optical contour detection module, and the load fluctuation data is collected synchronously at a higher frequency by the shaft end torque sensor. After median filtering and noise reduction, the displacement and torque are aligned in time and space, laying a solid foundation for high signal-to-noise ratio perception for subsequent control. The synchronous conveying segment is extracted by time axis alignment, and the two sets of data are subjected to mean-free normalized cross-correlation analysis to extract the effective peak delay time and amplitude, and to characterize the nonlinear frictional coupling characteristics between the medical packaging bag and the conveyor belt. 2) By dividing the conveying path into multiple independent sections, the friction fingerprint is transformed into a phase lag factor and an amplitude attenuation factor through nonlinear mapping. Combined with the transfer entropy and dynamic time curvature calculation, a multidimensional deviation metric is obtained, and finally a two-dimensional map with the conveying efficiency coefficient as the core is generated, realizing a global visual representation of the bag's motion posture and driving energy consumption under different working conditions. By relying on the dynamic conveying adaptation map to identify high instability risk sections, the proportional-integral controller generates speed adjustment compensation based on the risk index, and at the same time, the phase deviation index is combined to match the corresponding intensity of the correction command, thereby improving the stability of the entire line's operating cycle. By calculating the map correction amount through the Euclidean distance between the new and old dynamic friction fingerprints, the phase lag factor and amplitude attenuation factor are weighted and fused to update, and the long-term adaptability and yield of the packaging equipment are continuously improved. Attached Figure Description

[0016] Figure 1 This is a flowchart of the conveyor control method for the production of medical packaging bags proposed in this invention.

[0017] Figure 2 This is a block diagram of the conveyor control system for the production of medical packaging bags proposed in this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 This invention provides a conveyor control method for the production of medical packaging bags, comprising: S1. Obtain the deformation status data of the medical packaging bag to be conveyed output by the optical contour detection module in the packaging equipment, and simultaneously obtain the real-time load torque fluctuation data of the conveyor drive unit. S2. Analyze the cross-correlation response between bag deformation state data and real-time load torque fluctuation data to generate dynamic friction fingerprint; S3. Divide the conveying into multiple conveying sections and, based on dynamic friction fingerprints, construct a dynamic conveying adaptation map between the bag movement posture and driving energy consumption of the conveyor in different conveying sections. The dynamic conveying adaptation map is used to obtain the conveying efficiency coefficient. Among them, the conveyor is a belt conveyor device in packaging equipment responsible for conveying medical packaging bags from the upstream workstation to the downstream workstation. S4. Based on the conveying efficiency coefficient obtained from the dynamic conveying adaptation map, generate and execute differentiated speed regulation commands and correction commands for the conveyor drive unit. S5. Update the dynamic conveying adaptation map based on the execution results of the differentiated speed regulation command and the correction command.

[0020] In this embodiment of the invention, the specific implementation method of S1 acquiring the deformation state data of the medical packaging bag to be conveyed output by the optical contour detection module in the packaging equipment, and simultaneously acquiring the real-time load torque fluctuation data of the conveyor drive unit is as follows: S11. Obtain the bag edge contour sequence collected by the optical contour detection module under the preset strobe light source. After removing noise by median filtering of the bag edge contour sequence, generate bag deformation state data by analyzing the pixel-level displacement of the bag edge contour sequence. Specifically, after the conveyor enters a stable operating phase, the image sensors in the optical contour detection module work synchronously with the preset strobe light source. It should be noted that the typical configuration of the optical contour detection module is as follows: at least two high-frame-rate CMOS image sensors are installed in parallel on a gantry-type detection frame erected along the width of the conveyor belt. The field of view of each image sensor covers the left and right edges of the packaging bag, and the lens focal length is selected to fully capture the maximum lateral displacement of the bag within the field of view. Each image sensor is equipped with an independent optical bandpass filter, and the center wavelength of the filter matches the output spectrum of the strobe light source. The strobe light source uses a high-power narrow-pulse LED array, arranged on both sides of the image sensor and tilted to illuminate the conveyor belt plane. The light source controller receives hardware triggering from the frame synchronization signal from the image sensor to achieve synchronous output of strobe pulses. The pixel resolution of the image sensor is converted into physical coordinates through offline calibration by a calibration board, so that pixel-level displacement can be converted into the actual physical displacement of the bag edge. The method for obtaining the strobe light source is as follows: based on the current running speed of the conveyor and the exposure sequence of the image sensor in the optical contour detection module, the optimal strobe frequency for freezing the motion blur of the bag is determined, and the optimal strobe frequency parameter is pre-written into the register of the light source controller, so that the duration of the strobe pulse is much smaller than the pixel displacement that the bag may produce in a single exposure, thereby obtaining a clear outline of the bag edge in each frame image. During the acquisition process, a strobe light source emits high-brightness pulsed light at the optimal strobe frequency, illuminating the surface of the moving medical packaging bag. The image sensor continuously captures multiple frames of images, forming a sequence of bag edge contours arranged in chronological order. Median filtering is performed on each frame of the bag edge contour sequence. Specifically, a 5×5 pixel sliding window is used, moving pixel by pixel on the image. The gray values ​​of all pixels within the window are sorted by size, and the median value is taken as the new gray value of the center pixel of the window to replace the original pixel value. This filters out isolated noise points caused by random thermal noise from the image sensor and dust scattering, maintaining edge sharpness while eliminating pulse interference. After median filtering is completed, the position of the bag edge is extracted for two adjacent filtered images. The pixel coordinates of the same edge segment between adjacent frames are differentially calculated to obtain the pixel-level displacement of each edge pixel in the image plane along the conveying direction and perpendicular to the conveying direction. All pixel-level displacements are collected to form a displacement vector field. The displacement vector field is compressed into bag deformation state data that characterizes the degree and distribution of bag deformation.

[0021] S12. Real-time load torque fluctuation data is collected at a rate higher than the sampling frequency of the optical profile detection module by a torque sensor embedded in the end of the conveyor drive motor shaft, specifically including: A strain gauge torque sensor is embedded between the output shaft of the conveyor drive motor and the reducer. The strain gauge torque sensor uses strain gauges attached to the surface of the shaft to form a Wheatstone bridge. When the motor bearing is subjected to torque and undergoes a small torsional deformation, the bridge outputs a voltage signal proportional to the torque. The signal conditioning circuit of the torque sensor amplifies the voltage signal and converts it into a digital quantity. The analog-to-digital conversion is performed at a preset high sampling frequency, which is at least 4 times the sampling frequency of the optical contour detection module, to ensure that the high-frequency fluctuation details in the torque signal can be captured. The acquired raw torque signal contains instantaneous spikes caused by motor cogging effect, transmission chain vibration, and high-frequency noise from the measurement circuit. These spikes are not caused by changes in the bag's friction characteristics and need to be suppressed. To this end, an adaptive moving average processing method is implemented: a sliding window is defined, and the window length is dynamically determined based on the number of sampling points corresponding to one revolution of the motor, ensuring that the window always covers the sampling range corresponding to one complete mechanical cycle of the motor. At each sampling moment, the raw torque values ​​of T consecutive sampling points before and including the current moment are collected, and the arithmetic mean of all raw torque values ​​is calculated as the moving average torque value at the current moment, i.e., the moving average value; where T is the number of sampling points. The torque fluctuation component after removing the quasi-static load component is obtained by subtracting the moving average value from the raw torque value at the current moment. As time progresses, the sliding window discards the oldest sampling point and incorporates the newest sampling point, realizing the recursive sliding of the window. At the same time, a pre-acquired data segment is pre-filled so that a valid torque fluctuation value can be output from the beginning of the window's operation. After the above moving average processing, real-time load torque fluctuation data is generated.

[0022] In this embodiment of the invention, the detailed implementation steps of S2, which analyzes the cross-correlation response relationship between the bag deformation state data and the real-time load torque fluctuation data to generate a dynamic friction fingerprint, include: S21. Align the bag deformation status data and the real-time load torque fluctuation data on a unified time axis, and extract the synchronous segment of the two within the current conveying cycle. Specifically, a master clock is set, and the exposure center moment of each frame of the optical contour detection module is timestamped by a hardware trigger signal, and each sampling moment of the torque sensor is also marked by the same master clock source. Subsequently, a resampling process based on timestamp linear interpolation is performed: an equally spaced target time grid is constructed from the start time to the end time of the current delivery cycle. The grid interval is selected as the least common multiple of the two original sampling intervals, so that the target time grid can retain the bag deformation state data without introducing the aliasing error of torque data. At each time node of the target time grid, the two nearest original deformation sampling points are found to determine the bag deformation state data at that time node. Similarly, the two nearest original torque fluctuation sampling points are found and interpolated to obtain the real-time load torque fluctuation data at that time node. Ultimately, a one-to-one correspondence is formed between deformation state data and load torque fluctuation data on a unified, equally spaced time grid, resulting in a complete time-aligned data set. The start and end times of the current conveying cycle are provided by the conveyor, and time segments belonging to the current conveying cycle are extracted from the time-aligned data set. This synchronous segment contains deformation state data and load torque fluctuation data at each identical absolute time point during the entire process from the bag entering the conveyor to leaving the conveyor.

[0023] S22. In the synchronous segment, the cross-correlation function between the bag deformation state data and the real-time load torque fluctuation data is calculated using the cross-correlation analysis window, and the peak delay time and peak amplitude of the cross-correlation function are extracted, including: S221. Within the cross-correlation analysis window, obtain the first time series of bag deformation state data and the second time series of real-time load torque fluctuation data, and perform mean-removal processing on them respectively to obtain the first zero-mean sequence and the second zero-mean sequence. The specific steps are as follows: In the synchronous segments aligned from a unified time axis, a cross-correlation analysis window with variable width is selected. The width of the cross-correlation analysis window is adaptively adjusted according to the length of the conveying cycle to meet the analysis requirements under different bag lengths and conveying speeds. The starting position of the window is selected after the bag has completely entered the conveying area and the data on bag deformation and load torque fluctuations have entered a steady-state fluctuation period. The ending position of the window is set at the moment just before the bag is about to leave the conveying surface to ensure that the fluctuation characteristics captured in the cross-correlation analysis window fully reflect the continuous interactive dynamics between the bag and the conveyor belt. Within a defined cross-correlation analysis window, the numerical sequence corresponding to the time interval in the bag deformation state data is extracted and denoted as the first time series. Each value in this sequence represents a deformation state amplitude. Simultaneously, the numerical sequence in the same time interval in the real-time load torque fluctuation data is extracted and denoted as the second time series. Subsequently, the two time series are subjected to mean-reduction processing. For the first time series, all data points are traversed, summed, and divided by the total number of points in the sequence to obtain the arithmetic mean of the first time series. Then, the arithmetic mean is subtracted from each data point in the first time series to generate the first zero-mean sequence. The values ​​in this sequence fluctuate around zero, eliminating the constant component caused by the static deformation bias of the bag. For the second time series, the arithmetic mean is calculated using the same method and subtraction is performed to generate the second zero-mean sequence, thus transforming the torque fluctuation sequence into a symmetrical fluctuation signal centered at zero.

[0024] S222. Calculate the normalized cross-correlation function of the first zero-mean sequence and the second zero-mean sequence to obtain the correlation coefficient sequence under different time delays. First, a time delay search range is defined, consisting of a positive upper limit and a negative lower limit, typically encompassing several positive and negative sampling intervals to cover physical scenarios where torque fluctuations may precede or lag the deformation response. For each discrete time delay value within this search range, a cyclic shift operation is performed on the second zero-mean sequence: if the time delay value is positive, the entire second zero-mean sequence is shifted backward by the number of sampling points corresponding to the time delay value, with the gaps at the beginning of the sequence being cyclically filled by the values ​​shifted out from the end of the sequence; if the time delay value is negative, the sequence is shifted forward and the gaps at the end are cyclically filled. After completing the sequence shift under this time delay, the first zero-mean sequence and the shifted second zero-mean sequence are calculated. The sum of products at the same index position is calculated. Simultaneously, the square root of the self-multiplied cumulative value of the first zero-mean sequence and the square root of the self-multiplied cumulative value of the second zero-mean sequence are calculated respectively. The sum of the aforementioned products is divided by the product of these two square roots, and the result is the normalized cross-correlation coefficient under the specific time delay. For each time delay value in the search range, the above shift, multiplication and normalization operations are repeated to obtain the correlation coefficient corresponding to each time delay. All correlation coefficients are arranged in ascending order of time delay value to form a correlation coefficient sequence. Each value in this sequence represents the linear similarity between bag deformation fluctuation and load torque fluctuation under a specific time misalignment.

[0025] S223. Perform peak search on the correlation coefficient sequence, and extract the correlation coefficient value corresponding to all local peak points and the time delay corresponding to the correlation coefficient value; Specifically, a point-by-point peak search is performed on the correlation coefficient sequence to locate the time delay position corresponding to the local correlation coefficient maximum. The search process starts from the second element of the correlation coefficient sequence and ends at the second-to-last element, visiting each correlation coefficient value point in turn. For the current visited point, its correlation coefficient value is compared with the correlation coefficient values ​​of the immediately preceding point and the next point. If the correlation coefficient value of the current point is greater than the correlation coefficient values ​​of the preceding point and the next point, then the current point is determined to be a local peak point. After identifying all local peak points, for each local peak point, record the correlation coefficient value corresponding to that point and its index position in the correlation coefficient sequence. Subtract the index position corresponding to the center point of the correlation coefficient sequence from the index position, and then multiply by the sampling time interval to calculate the time delay corresponding to that peak point. If multiple adjacent points in the correlation coefficient sequence have equal correlation coefficient values ​​that are all greater than their neighbors due to incomplete filtering, then the midpoint of the plateau region formed by all equal points is taken as the position of the local peak point. Finally, a set of candidate peaks is obtained, and each element contains the correlation coefficient value of that peak and the corresponding time delay.

[0026] S224. Compare the correlation coefficient value corresponding to each local peak point with the preset correlation threshold, remove false peak points with correlation coefficient values ​​lower than the correlation threshold, and retain the effective peak point with the largest correlation coefficient value. The correlation threshold is determined by collecting multiple segments of operational data from the conveyor under normal transport of standard bag samples without abnormal slippage or blockage. For each segment of operational data, the cross-correlation function sequence of the bag deformation state data and real-time load torque fluctuation data within the same cross-correlation analysis window is calculated. The standard deviation of the correlation coefficients of all sequences is then calculated, and a preset multiple of this standard deviation is used as the correlation threshold. Given that the distribution of torque fluctuations and deformation signals in actual conveyor operation may have slight tailing, the multiple can be set in a floating range of 2.5 to 3.5 times, with a preferred value of 3 times. The correlation threshold characterizes the lower limit of the observable correlation strength under stable interaction conditions. Peak values ​​below the correlation threshold are considered spurious correlations induced by random noise or non-frictional coupling factors.

[0027] S225. Extract the time delay of the effective peak points as the peak delay time, and extract the correlation coefficient value of the effective peak points as the peak amplitude. In the set of all local peak points, the correlation coefficient value of each local peak point is compared with the determined correlation threshold one by one. Local peak points with correlation coefficient values ​​less than the correlation threshold are discarded, and only candidate peak points with correlation coefficient values ​​not lower than the correlation threshold are retained. If the number of remaining effective peak points after screening is zero, the peak delay time and peak amplitude of the previous transmission cycle are directly used as the result of the current transmission cycle, and the current transmission cycle is marked as a correlation loss state. If the number of remaining effective peak points after screening is a single one, the time delay of the effective peak point is extracted as the peak delay time, and its correlation coefficient value is extracted as the peak amplitude. If there are multiple effective peak points after screening, the one with the largest correlation coefficient value is selected as the main peak point, the time delay corresponding to the main peak point is extracted as the peak delay time, and the correlation coefficient value corresponding to the main peak point is extracted as the peak amplitude.

[0028] S23. The specific process of forming a dynamic triboelectric fingerprint by jointly analyzing the peak delay time and peak amplitude includes: Dividing the peak delay time by the total duration of the current conveying cycle yields a dimensionless relative delay value, which is limited to the range of 0 to 1. This relative delay value characterizes the proportion of the lag in the bag deformation response relative to the torque fluctuation response across all time scales. Then, dividing the peak amplitude by a preset torque fluctuation amplitude benchmark yields a relative amplitude coefficient, which characterizes the strength of the torque fluctuation coupling generated by the interaction between the bag and the conveying surface relative to the standard steady state. The torque fluctuation amplitude benchmark is obtained by averaging the torque fluctuation amplitude when the bag and the conveyor belt are in stable contact under long-term statistical conditions. The normalized relative delay value and relative amplitude coefficient are used to construct a two-dimensional feature vector. The first component of this feature vector is the relative delay value, and the second component is the relative amplitude coefficient. This two-dimensional feature vector is the dynamic triboelectric fingerprint.

[0029] In this embodiment of the invention, step S3 divides the transport into multiple transport sections and constructs a dynamic transport adaptation map based on dynamic triboelectric fingerprints. Specific implementation methods for using the dynamic transport adaptation map to obtain the transport efficiency coefficient include: S31. Divide the conveying path formed by the conveyor inside the packaging equipment into multiple conveying sections, and pre-set a set of reference dynamic parameters for each conveying section; Specifically, based on the layout of the conveyor's physical structure, the spatial positions of the drive and driven rollers, and the process flow rhythm of the medical packaging bags in the production line, the entire conveying path is divided into three conveying sections along the direction of motion: an acceleration zone, a constant speed zone, and a deceleration zone. The acceleration zone begins at the point where the bag falls or is released from the upstream equipment onto the conveyor belt surface and begins to be dragged by friction, and ends at the critical point where the bag's speed reaches synchronization with the conveyor belt's linear speed. The constant speed zone covers the middle section where the bag and conveyor belt do not slide relative to each other and maintain a constant speed. The deceleration zone begins at the point where the bag approaches the downstream junction and begins to decelerate, and ends at the point where the bag completely detaches from the conveyor surface. For each defined conveying section, a set of baseline dynamic parameters is pre-set. The baseline dynamic parameters are obtained as follows: under no-load conditions and stable conveying of standard packaging bags, the linear velocity of the conveyor belt surface, the ideal real-time centroid position and attitude angle of the bag are measured using a laser velocimeter and a high-speed camera system, combined with the output torque of the motor driver and the encoder phase information. After data fitting, the baseline dynamic parameters of each section are formed and pre-stored in the controller. The baseline dynamic parameters specifically include the standard conveying linear velocity of the section, the ideal motion phase trajectory of the bag, the baseline phase sequence, and the average value of the baseline load torque. These parameters constitute a reference baseline for measuring deviations that occur during actual conveying.

[0030] S32. For each conveying section, the dynamic friction fingerprint is mapped to a phase lag factor characterizing the movement posture of the bag in the conveying section and an amplitude attenuation factor characterizing the driving energy consumption through a preset nonlinear mapping table. Understandably, in the historical operation database, for each specific transport section, dynamic friction fingerprints under different working conditions are collected, along with the actual phase lag and torque amplitude attenuation observed synchronously with the fingerprints, to construct input-output sample pairs; the samples are then interpolated to generate a nonlinear mapping table, where the input dimension is the dynamic friction fingerprint and the output dimensions are the phase lag factor and amplitude attenuation factor. In actual operation, for the current dynamic friction fingerprint, within each conveying section, the two components of the dynamic friction fingerprint are input into a preset nonlinear mapping table, and the phase lag factor and amplitude attenuation factor of the conveying section are output by looking up the table. The phase lag factor characterizes the degree of lag of the actual movement of the bag relative to the reference phase trajectory, and the amplitude attenuation factor characterizes the degree of attenuation of the motor drive torque oscillation amplitude relative to the reference torque oscillation amplitude. Together, they reflect the transmission result from friction characteristics to motion and energy consumption response.

[0031] S33. Using phase lag factor and amplitude attenuation factor, multidimensional state reconstruction and spatiotemporal characteristic analysis are performed to obtain a multidimensional deviation measure reflecting the degree of deviation from the transport state, including: S331. For each transport section, the phase lag factor and amplitude attenuation factor are reconstructed into a two-dimensional state vector. The specific process of calculating the transition entropy of the two-dimensional state vector between multiple consecutive state transition analysis windows, which serves as a state complexity index, is as follows: The phase lag factor and amplitude attenuation factor obtained for each transport section in each transport cycle are combined into a two-dimensional state vector. The first dimension of the vector is the phase lag factor, and the second dimension is the amplitude attenuation factor, thus forming a two-dimensional continuous state space. The continuous state space is discretized: based on the distribution range of the phase lag factor and amplitude attenuation factor in historical data, the range of values ​​for both is uniformly divided into several equally wide sub-intervals, forming a two-dimensional grid. Each grid cell corresponds to a discrete state symbol. The observation interval that slides continuously on the time axis and contains a fixed number of continuous transmission cycles is defined as the state transition analysis window. In each state transition analysis window, the two-dimensional state vector is mapped to a discrete state symbol sequence. The number of transitions between adjacent state pairs in the discrete state symbol sequence is counted, and the frequency distribution of all possible state transitions is obtained by counting. The frequency of all transitions starting from each state is proportionally normalized so that the sum of the probabilities of transitioning from the same state to each possible state is 1, thus representing the state transition law in the form of a probability table. The process of obtaining the state transition entropy is as follows: traverse all transition probability values ​​greater than 0 in the probability table, calculate the product of each probability value and its natural logarithm, sum all products and take the opposite number to obtain the transition entropy of the state transition analysis window.

[0032] S332. The method for obtaining the phase deviation index reflecting the degree of phase distortion by performing dynamic time bending analysis on the phase lag factor and the reference dynamic parameters of the transport section is as follows: Dynamic time warping analysis measures the degree of nonlinear distortion of the reference phase sequence on the time axis between the actual phase lag factor time series and the reference dynamic parameters. The actual phase lag factor time series is obtained by arranging the phase lag factors generated in the current transport cycle and the next number of historical cycles in the transport segment in chronological order. The reference phase sequence is an ideal phase sequence that is pre-generated and advances uniformly with time in the segment. The distance between each element of the actual phase lag factor time series and each element of the reference phase series in the reference dynamic parameters is defined as the absolute value of its difference, i.e., the absolute deviation between the two phase values. This forms a distance matrix, where the row index of the distance matrix corresponds to the time of the actual phase lag factor time series, and the column index corresponds to the time of the reference phase series. Next, starting from the lower left element of the distance matrix, the cumulative distance matrix is ​​calculated. Specifically, for each matrix element, its cumulative distance is equal to the sum of its own distance value and the minimum cumulative distance among the three adjacent elements to the left, below, and lower left. After filling the entire cumulative distance matrix, the cumulative distance value located at the upper right corner of the matrix is ​​the dynamic time curvature distance between the two sequences. Finally, the calculated dynamic time curvature distance is divided by the length of the reference phase series to obtain the normalized dynamic time curvature distance value, which is used as a phase deviation index. The larger the phase deviation index, the more severe the distortion of the actual phase lag factor time series relative to the reference phase series, and the worse the motion synchronization of the bag. Conversely, the smaller the index, the more closely the motion phase follows the ideal trajectory.

[0033] S333. Use both the state complexity index and the phase deviation index as a multidimensional deviation measure to characterize the degree of deviation of the current state.

[0034] S34. Based on the evolution of phase lag factor, amplitude attenuation factor, and multidimensional deviation metric over time, a dynamic transport adaptation map is constructed, including: S341. Arrange the phase lag factor, amplitude attenuation factor and multidimensional deviation measure generated for each transport segment in the current transport cycle and multiple historical consecutive transport cycles in chronological order to form a time series feature matrix corresponding to each transport segment. Specifically, for each transport segment, the phase lag factor, amplitude attenuation factor, state complexity index, and phase deviation index generated in each transport cycle are extracted and collectively constitute the feature vector of the transport segment corresponding to that transport cycle. These feature vectors are stacked vertically in chronological order from earliest to latest to form a two-dimensional matrix. The rows of the matrix represent different transport cycles, i.e., the time dimension, and the columns of the matrix represent the phase lag factor, amplitude attenuation factor, state complexity index, and phase deviation index, i.e., the feature dimension. If a feature is missing in a certain transport cycle due to sensor interruption or unqualified data, the valid feature value of the previous transport cycle is used to fill the gap, ensuring the integrity of the matrix and the uniformity of time between rows. Thus, each transport segment independently maintains a time series feature matrix, which retains the instantaneous state information of a single transport cycle and contains the evolution law of the transport state over time through continuous observation of multiple historical cycles. When a new transport cycle ends, the latest generated feature vector is appended to the last row of the transport, and the data of the earliest cycle in the first row of the matrix is ​​removed, realizing the rolling update of the matrix over time.

[0035] S342. An evolutionary truncation window of preset length is slid across the time series feature matrix to extract evolutionary feature tensors within multiple local time windows; wherein, a local time window is a continuous time segment extracted by the evolutionary truncation window during the sliding process. Specifically, for the time series feature matrix of each transport segment, an evolution truncation window of a preset length is defined; The starting row of the evolutionary truncation window is aligned with the earliest transport cycle row in the time series feature matrix. The window slides down the row direction of the time series feature matrix from top to bottom, moving one transport cycle step at a time. At each window stop position, a data block of L rows and all feature columns covered by the window is extracted. This data block is an L-row multiplied by four-column matrix. By adding dimension identifiers, it is transformed into a three-dimensional evolutionary feature tensor. The three dimensions of this tensor are the time step within the window, the feature type, and the identifier of the transport segment, respectively. Here, the number of rows L is the length of the evolutionary truncation window, indicating that feature vectors of L consecutive transport cycles are selected. The feature types include phase lag factor, amplitude attenuation factor, state complexity index, and phase deviation index. As the window continues to slide, a series of evolutionary feature tensors that overlap in time but are slightly shifted are obtained in each transport segment; each evolutionary feature tensor contains the temporal change details of multidimensional features within that local time window, in which the evolution trajectories of the phase lag factor and amplitude decay factor are fully preserved, while the fluctuations of the state complexity index and phase deviation index over time are also embedded in it.

[0036] S343. Establish a two-dimensional matrix framework with the transport segment as the first dimension and the timestamp of the local time window as the second dimension.

[0037] S344. Perform multidimensional temporal trend analysis on each evolutionary feature tensor, extracting the long-term memory decay slope of the phase lag factor and the fluctuation divergence rate of the amplitude decay factor within the local time window, including: For each extracted evolutionary feature tensor, it is expanded according to the feature dimension to obtain the time series of the phase lag factor over L periods within the window and the time series of the amplitude decay factor over the same L periods. Long-term memory decay analysis is performed on the time series of the phase lag factor: first, the value of the time series at different lag steps is calculated to identify the decay trend where the autocorrelation value gradually weakens as the lag step number increases; then, linear fitting is performed in a semi-logarithmic coordinate system, specifically using the natural logarithm of the autocorrelation value as the vertical axis and the lag step number as the horizontal axis to fit an optimal straight line. The slope of this line is the long-term memory decay slope. The larger the absolute value of the long-term memory decay slope, the faster the memory decay of the phase lag factor, and the more rapidly the current perturbation dissipates its influence on the future phase. If the decay slope is close to 0, it indicates that the phase lag has a long-term persistent memory effect. Simultaneously, a fluctuation divergence rate analysis was performed on the time series of the amplitude decay factor: the time series was centered, its cumulative deviation sequence was calculated, and then the cumulative deviation sequence was divided into multiple segments of equal length. The range within each segment was calculated, and the changing trend of the range of each segment with the increase of the segment length was statistically analyzed. After logarithmic transformation, linear fitting was performed to obtain the fluctuation divergence rate. The divergence rate reflects the degree of divergence of the amplitude decay factor fluctuation as the observation time scale increases. The higher the divergence rate, the more diffuse and unstable the fluctuation of the energy consumption-related torque decay amplitude tends to be.

[0038] S345. Based on the coupling relationship between the long-term memory decay slope and the fluctuation divergence rate, determine the transmission efficiency coefficient of this local time window. Understandably, when the long-term memory of the phase lag factor decays rapidly, while the amplitude decay factor's fluctuation divergence remains at a low level, it indicates that minor disturbances will not persist in the system for a long time, and fluctuations in energy dissipation are not easily amplified. The conveyor has robust and efficient energy transfer and phase-locking capabilities. Conversely, if the decay slope is gentle and the divergence rate is large, minor bag slippage or frictional abrupt changes will be memorized by the system for a long time and amplified into severe vibrations and additional energy losses, resulting in deteriorated conveying efficiency. The absolute value of the long-term memory decay slope is normalized, and the normalization reference value is the average value of the decay slope observed during the historical normal operation of the transmission section. At the same time, the fluctuation divergence rate is normalized in the same way with reference to its historical average value. Then, the normalized absolute value of the decay slope is multiplied by the proportional coefficient determined by the regression analysis of historical data, and divided by the normalized fluctuation divergence rate to obtain the coupling ratio. The coupling ratio is mapped and its value range is compressed to between 0 and 1. The resulting value is the transmission efficiency coefficient of this local time window. The closer the transmission efficiency coefficient is to 1, the better the transmission efficiency. The closer the transmission efficiency coefficient is to 0, the worse the transmission efficiency.

[0039] S346. Fill the corresponding transport segment and time window positions in the two-dimensional matrix frame with the transport efficiency coefficient as the matrix element to form a dynamic transport adaptation map, specifically: A two-dimensional matrix framework is pre-constructed. The first dimension of the matrix corresponds to each transport section, arranged from top to bottom in the spatial order of acceleration zone, constant speed zone, and deceleration zone. The second dimension corresponds to the timestamp of the local time window, which is the center time or start time of the window, arranged from left to right according to the time progress. When the corresponding transport efficiency coefficient of a transport section is calculated at a certain time window position, the row of the transport section and the column that matches or is closest to the timestamp are located in the two-dimensional matrix framework according to the timestamp of the window, and the transport efficiency coefficient is filled into the matrix cell. As the transport process continues, the evolution capture window slides forward, and new transport efficiency coefficients are successively filled into the new columns of the two-dimensional matrix framework, while old columns that have not been updated for a long time can be covered or shifted to the right; finally, a dynamic transport adaptation map is formed, in which each element is the transport efficiency coefficient of a specific segment within a specific time window.

[0040] In this embodiment of the invention, the detailed implementation steps of S4, which generates and executes differentiated speed control commands and correction commands for the conveyor drive unit based on the conveyor efficiency coefficient obtained from the dynamic conveyor adaptation map, include: S41. Based on the dynamic transport adaptation map, read the multidimensional deviation measure corresponding to each transport section, calculate the transport instability risk index of each transport section, and identify the target transport section that leads to the trend of transport stability instability. Specifically, while providing the transportation efficiency coefficient, the dynamic transportation adaptation map also stores the multi-dimensional deviation metrics, namely the state complexity index and the phase deviation index, in association with each matrix element during its generation process. At the end of the current transportation cycle, the state complexity index and phase deviation index corresponding to the latest time window of each transportation segment are read from the dynamic transportation adaptation map. The state complexity index is standardized and divided by the statistical median of the historical state complexity of the transport section to obtain the relative complexity ratio. Simultaneously, the phase deviation index is standardized and divided by the allowable tolerance range of the reference phase sequence of the transport section to obtain the relative deviation ratio. Then, the relative complexity ratio and the relative deviation ratio are multiplied by the first and second weighting coefficients determined through training with historical data, respectively, and summed to obtain the transport instability risk index for the transport section. If the value of the transport instability risk index exceeds a preset instability warning threshold, it indicates that the transport state of the transport section is trending towards instability. The instability warning threshold is determined by analyzing the upper limit of the transport instability risk index distribution during multiple stable transport batches. Among all transport sections, the transport section with the highest transport instability risk index that exceeds the warning line is selected, and priority is determined by combining the transport section with the fastest rate of increase in transport instability risk index. Finally, one or more target transport sections that cause the transport stability trend to become unstable are identified.

[0041] S42. The conveying instability risk index of the target conveying section is used as an input signal and sent to the proportional-integral controller, wherein the proportional-integral controller calculates the speed regulation compensation amount according to the sign and magnitude of the conveying instability risk index. Understandably, once the target conveying section is determined, the conveying instability risk index of that section is used as an input signal and input to the incremental proportional-integral controller. The current conveying instability risk index is multiplied by a preset proportional gain coefficient to generate an instantaneous compensation component proportional to the conveying instability risk index. The conveying instability risk index is accumulated over time. Specifically, in each control cycle, the current risk index is multiplied by the integral gain coefficient and accumulated to the integral accumulator. The accumulated compensation component output by the integral accumulator can reflect the persistent deviation caused by the continued existence of the risk. Among them, the proportional gain coefficient determines the strength of the controller's instantaneous response to the deviation; the integral gain coefficient determines the rate at which the controller continuously accumulates the deviation. The value range of the proportional gain coefficient is 0.3 to 1.5, and the value range of the integral gain coefficient is 0.05 to 0.3. The integral gain coefficient is set to 1 / 6 to 1 / 5 of the proportional gain coefficient to adapt to the inertia and hysteresis response characteristics of the conveying process. The controller adds the instantaneous compensation component to the cumulative compensation component and outputs a signed speed compensation amount. When the conveyor instability risk index is positive, it indicates that the current state is in the direction of instability aggravation, and the speed compensation amount is given a negative value to reduce the conveyor speed in order to reduce the impact of frictional mutation. When the conveyor instability risk index is negative, it indicates that the state has a tendency to recover to stability, and the speed compensation amount is given a positive value to restore the appropriate speed. Through continuous adjustment by the proportional-integral controller, the rotational speed of the conveyor drive unit is dynamically corrected according to the real-time conveyor instability risk index to suppress the further development of the instability state.

[0042] S43. Generate differentiated speed control commands based on speed compensation amount; The process of generating differentiated speed control commands is as follows: The absolute value of the speed compensation is algebraically summed with the reference set speed of the current conveying section to obtain the target speed value. The reference set speed is obtained through analysis of the standard conveying linear velocity given in the reference dynamic parameters of the conveying section. The difference between the target speed and the current speed is determined as the speed adjustment range, which limits the specific amount of revolutions per minute that the motor speed needs to increase or decrease. This parameter determines the severity of the speed control action. Simultaneously, based on the position of the target conveying section in the conveying path and the real-time running speed of the bag, the estimated time for the bag to reach the entrance of that section is calculated. The adjustment phase of the speed control command is then set according to this estimated time. Phase refers to the time offset of the speed control command from the start of the current conveying cycle. This parameter ensures that the new speed setting is completed just as the front end of the bag enters the target conveying section, thus applying intervention in spatial position. Based on the magnitude of the speed compensation and the smoothness requirements of the current operating state, the speed control frequency parameter is determined. The speed control frequency parameter specifies the ramp rate or speed command update frequency used for the motor speed to transition from the current value to the target value. The finally generated differentiated speed control command is encapsulated into a communication data frame containing the speed adjustment amplitude, adjustment phase, and speed control frequency parameters, and sent to the frequency converter of the corresponding motor in the target conveying section for execution via the fieldbus.

[0043] S44. Simultaneously, the process of reading the phase deviation index of the transport instability risk index in the target transport section and analyzing the degree of deviation exceeding the limit is as follows: A phase deviation index is extracted separately from the transport instability risk index of the target transport section. An allowable limit for the phase deviation of this section is maintained. This allowable limit is calculated based on the geometric gap between the bag width and the guide baffle of the transport channel, representing the maximum tolerance for lateral phase distortion of the bag that does not exceed the guide allowable boundary. When analyzing the degree of deviation exceeding the limit, the current phase deviation index is compared with the allowable limit: if the phase deviation index is less than or equal to the allowable limit, the degree of deviation exceeding the limit is 0, indicating that no additional airflow deviation intervention is required; if the phase deviation index is greater than the allowable limit, the absolute difference exceeding the allowable limit is calculated, and the absolute difference is divided by the allowable limit to obtain the exceedance ratio. This exceedance ratio is the degree of deviation exceeding the limit, and its value characterizes the relative severity of the bag's phase distortion exceeding the safety boundary.

[0044] S45. Based on the degree of deviation exceeding the limit, calculate the intensity value of the deviation correction command, generate a deviation correction command containing the intensity value, and send the deviation correction command to the control valve of the airflow auxiliary guide device located on the side of the target conveying section to adjust the airflow output. Understandably, based on the degree of deviation exceeding the limit, the opening percentage of the airflow-assisted guiding device control valve is directly calculated according to a preset proportional conversion relationship, which serves as the intensity value of the deviation correction command. The higher the degree of deviation exceeding the limit, the larger the calculated intensity value, indicating that a stronger airflow thrust needs to be applied from the side. The intensity value is encapsulated as a deviation correction command, which includes the channel number and direction of action of the airflow-assisted guiding device control valve on the side of the target conveying section. The final generated deviation correction command is immediately sent to the corresponding electronically controlled proportional valve, which adjusts the airflow output to correct the lateral attitude of the bag and causes the phase deviation to return to the allowable range.

[0045] In this embodiment of the invention, the specific implementation of step S5, which updates the dynamic transport adaptation map based on the execution results of the differentiated speed control command and the correction command, is as follows: S51. After executing the differentiated speed control command and the correction command, reacquire the bag deformation status data and real-time load torque fluctuation data for the new cycle.

[0046] S52. Based on the bag deformation state data of the new cycle and the real-time load torque fluctuation data, a new dynamic friction fingerprint is regenerated.

[0047] S53. Compare the new dynamic friction fingerprint with the historical dynamic friction fingerprint, calculate the Euclidean distance between the two as a difference measure, and obtain the correction amount of the dynamic transport adaptation map. The new dynamic friction fingerprint generated in the new transport cycle is compared with the baseline friction fingerprint obtained by statistically analyzing the dynamic friction fingerprints of several historical stable cycles. During the comparison, the square root of the sum of the squares of the differences between the corresponding components of the new dynamic friction fingerprint and the baseline friction fingerprint in two-dimensional space is calculated to obtain the Euclidean distance. This Euclidean distance quantifies the difference between the current friction characteristics and the recent stable state, and this value is directly used as the correction amount for the dynamic transport adaptation map. The larger the correction amount, the more drastic the change in the interaction characteristics between the bag and the transport surface, and the more necessary it is to make corresponding amplitude corrections to the dynamic transport adaptation map. When the correction amount approaches zero, it indicates that the friction characteristics remain stable.

[0048] S54. The correction amount is fused with the phase lag factor and amplitude attenuation factor in the dynamic transport adaptation map of the previous transport cycle. The phase lag factor and amplitude attenuation factor are updated with weights. Based on the updated phase lag factor and amplitude attenuation factor, the multidimensional deviation metric is recalculated, and the update of the dynamic transport adaptation map is finally completed. Specifically, the correction is decomposed into a phase correction term for the phase lag factor and an amplitude correction term for the amplitude attenuation factor. The phase lag factor and amplitude attenuation factor recorded in the dynamic transport adaptation map of the previous transport cycle are weighted and fused with the corresponding correction terms to obtain the updated phase lag factor and amplitude attenuation factor. Based on the updated phase lag factor and amplitude attenuation factor, the state complexity index and phase deviation index of each transport section are recalculated to form the updated multidimensional deviation metric, thereby refreshing the transport efficiency coefficient of each corresponding matrix element in the dynamic transport adaptation map. The update process is completed online at the end of the current transport cycle, and the updated dynamic transport adaptation map is immediately put into the entire decision-making process of calculating the transport instability risk index, identifying the target section, and generating speed regulation and correction commands for the next transport cycle.

[0049] Please see Figure 2 This invention provides a conveyor control system for the production of medical packaging bags, comprising: Bag Deformation and Torque Data Acquisition Module: Acquires the bag deformation status data of the medical packaging bag to be conveyed, output by the optical contour detection module in the packaging equipment, and simultaneously acquires the real-time load torque fluctuation data of the conveyor drive unit. Dynamic friction fingerprint generation module: Analyzes the cross-correlation response relationship between bag deformation state data and real-time load torque fluctuation data to generate dynamic friction fingerprint; Dynamic transport adaptation map construction module: Divide into multiple transport sections and construct a dynamic transport adaptation map based on dynamic triboelectric fingerprint. The dynamic transport adaptation map is used to obtain the transport efficiency coefficient. Speed ​​regulation and correction command generation module: Based on the conveying efficiency coefficient obtained from the dynamic conveying adaptation map, it generates and executes differentiated speed regulation and correction commands for the conveyor drive unit. Dynamic transport adaptation map update module: Updates the dynamic transport adaptation map based on the execution results of differentiated speed control commands and correction commands.

[0050] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0051] It should be noted that all formulas in this manual are calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0052] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A conveyor control method for medical packaging bag production, characterized in that, include: S1. Obtain the deformation status data of the medical packaging bag to be conveyed output by the optical contour detection module in the packaging equipment, and simultaneously obtain the real-time load torque fluctuation data of the conveyor drive unit. S2. Analyze the cross-correlation response between bag deformation state data and real-time load torque fluctuation data to generate dynamic friction fingerprint; S3. Divide the transport into multiple transport sections and construct a dynamic transport adaptation map based on dynamic friction fingerprints. The dynamic transport adaptation map is used to obtain the transport efficiency coefficient. S4. Based on the conveying efficiency coefficient obtained from the dynamic conveying adaptation map, generate and execute differentiated speed regulation commands and correction commands for the conveyor drive unit. S5. Update the dynamic conveying adaptation map based on the execution results of the differentiated speed regulation command and the correction command.

2. The method according to claim 1, characterized in that, S1 includes: The optical contour detection module acquires the bag edge contour sequence under a preset strobe light source. After removing noise by median filtering of the bag edge contour sequence, the bag deformation state data is generated by analyzing the pixel-level displacement of the bag edge contour sequence. Meanwhile, a torque sensor embedded in the shaft of the conveyor drive motor collects real-time load torque fluctuation data at a rate higher than the sampling frequency of the optical profile detection module.

3. The method according to claim 1, characterized in that, S2 includes: Align the bag deformation data with the real-time load torque fluctuation data on a unified time axis, and extract the synchronous segments of the two within the current conveying cycle. In the synchronous segment, the cross-correlation function between the bag deformation state data and the real-time load torque fluctuation data is calculated, and the peak delay time and peak amplitude of the cross-correlation function are extracted. A dynamic triboelectric fingerprint is formed by jointly analyzing the peak delay time and peak amplitude.

4. The method according to claim 3, characterized in that, The calculation of the cross-correlation function between the bag deformation state data and the real-time load torque fluctuation data, and the extraction of the peak delay time and peak amplitude of the cross-correlation function, includes: Within the cross-correlation analysis window, the first time series of bag deformation state data and the second time series of real-time load torque fluctuation data are obtained, and mean-reduction processing is performed on them respectively to obtain the first zero-mean sequence and the second zero-mean sequence. Calculate the normalized cross-correlation function between the first zero-mean sequence and the second zero-mean sequence to obtain the correlation coefficient sequence; Peak search is performed on the correlation coefficient sequence to extract the correlation coefficient values ​​corresponding to all local peak points and the time delay corresponding to those correlation coefficient values; The correlation coefficient value is compared with the preset correlation threshold. False peaks with correlation coefficient values ​​lower than the correlation threshold are removed, and the effective peaks with the largest correlation coefficient values ​​are retained. The time delay of the effective peak points is extracted as the peak delay time, and the correlation coefficient of the effective peak points is extracted as the peak amplitude.

5. The method according to claim 1, characterized in that, S3 includes: The conveying path formed by the conveyor inside the packaging equipment is divided into multiple conveying sections, and a set of benchmark dynamic parameters is pre-set for each conveying section; For each transport section, the dynamic friction fingerprint is mapped to a phase lag factor characterizing the movement posture of the bag in that transport section and an amplitude attenuation factor characterizing the driving energy consumption through a preset nonlinear mapping table. By using phase lag factor and amplitude attenuation factor to perform multidimensional state reconstruction and spatiotemporal characteristic analysis, a multidimensional deviation measure reflecting the degree of deviation of the transport state is obtained. Based on the evolution of phase lag factor, amplitude attenuation factor and multidimensional deviation metric over time, a dynamic transport adaptation map is constructed.

6. The method according to claim 5, characterized in that, The method utilizes phase lag factor and amplitude attenuation factor for multidimensional state reconstruction and spatiotemporal feature analysis to obtain a multidimensional deviation metric reflecting the degree of deviation from the transport state, including: For each transport section, the phase lag factor and amplitude attenuation factor are reconstructed into a two-dimensional state vector, and the transition entropy of the two-dimensional state vector between multiple consecutive state transition analysis windows is calculated as a state complexity index. Meanwhile, dynamic time bending analysis was performed on the phase lag factor and the reference dynamic parameters of the transport section to obtain the phase deviation index. The state complexity index and the phase deviation index are used together as a multidimensional deviation measure to characterize the degree of deviation of the current state.

7. The method according to claim 5, characterized in that, The dynamic transport adaptation map is constructed based on the evolution of phase lag factor, amplitude attenuation factor, and multidimensional deviation metric over time series, including: The phase lag factor, amplitude attenuation factor and multidimensional deviation metric generated for each transport segment are arranged in chronological order to form a time series feature matrix corresponding to each transport segment. An evolutionary truncation window of preset length is slid across the time series feature matrix to extract evolutionary feature tensors within multiple local time windows; A two-dimensional matrix framework is established with the transport segment as the first dimension and the timestamp of the local time window as the second dimension. Multidimensional temporal trend analysis is performed on each evolutionary feature tensor to extract the long-term memory decay slope of the phase lag factor and the fluctuation divergence rate of the amplitude decay factor within the local time window. Based on the coupling relationship between the long-term memory decay slope and the fluctuation divergence rate, the transmission efficiency coefficient of this local time window is determined. The transport efficiency coefficient is filled into the corresponding transport segment and time window position in the two-dimensional matrix frame to form a dynamic transport adaptation map with the transport efficiency coefficient as the matrix element.

8. The method according to claim 1, characterized in that, S4 includes: Based on the dynamic transport adaptation map, the multidimensional deviation metric corresponding to each transport section is read, and the transport instability risk index of each transport section is calculated to identify the target transport section that leads to the trend of transport instability. The conveying instability risk index of the target conveying section is used as an input signal and sent to the proportional-integral controller, which calculates the speed regulation compensation amount based on the sign and magnitude of the conveying instability risk index. Generate differentiated speed control commands based on the speed compensation amount; Simultaneously, the phase deviation index of the transport instability risk index in the target transport section is read, and the degree of deviation exceeding the limit is analyzed. Based on the degree of deviation exceeding the limit, the strength value of the correction command is calculated, and a correction command containing the strength value is generated.

9. The method according to claim 1, characterized in that, S5 includes: After executing the differentiated speed control command and the correction command, the bag deformation state data and real-time load torque fluctuation data of the new cycle are reacquired. Based on the bag deformation state data of the new cycle and the real-time load torque fluctuation data, a new dynamic friction fingerprint is regenerated. The new dynamic triboelectric fingerprint is compared with the historical dynamic triboelectric fingerprint, and the Euclidean distance between the two is calculated as a measure of difference to obtain the correction amount of the dynamic transport adaptation map. The correction amount is fused with the phase lag factor and amplitude attenuation factor in the dynamic transport adaptation map of the previous transport cycle. The phase lag factor and amplitude attenuation factor are then updated with weights. Based on the updated phase lag factor and amplitude attenuation factor, the multidimensional deviation metric is recalculated, and the dynamic transport adaptation map is finally updated.

10. A conveyor control system for the production of medical packaging bags, characterized in that, The system, applied to the method of any one of claims 1-9, comprises: Bag Deformation and Torque Data Acquisition Module: Acquires the bag deformation status data of the medical packaging bag to be conveyed, output by the optical contour detection module in the packaging equipment, and simultaneously acquires the real-time load torque fluctuation data of the conveyor drive unit. Dynamic friction fingerprint generation module: Analyzes the cross-correlation response relationship between bag deformation state data and real-time load torque fluctuation data to generate dynamic friction fingerprint; Dynamic transport adaptation map construction module: Divide into multiple transport sections and construct a dynamic transport adaptation map based on dynamic triboelectric fingerprint. The dynamic transport adaptation map is used to obtain the transport efficiency coefficient. Speed ​​regulation and correction command generation module: Based on the conveying efficiency coefficient obtained from the dynamic conveying adaptation map, it generates and executes differentiated speed regulation and correction commands for the conveyor drive unit. Dynamic transport adaptation map update module: Updates the dynamic transport adaptation map based on the execution results of differentiated speed control commands and correction commands.

Images