A medical dressing production line equipment state intelligent monitoring method and system

Abstract

This invention discloses an intelligent monitoring method and system for the status of equipment in a medical dressing production line, specifically relating to the field of intelligent monitoring technology for industrial equipment status. The method simultaneously acquires multiple material tension sequences, infrared thermal distribution matrices, high-frequency acoustic emission signals, and spindle rotation angle sequences; calculates the tension mismatch gradient; combines the rotation angle with spatial partial correlation comparison in the thermal distribution matrix to construct a spatial mapping map of colloidal micro-accumulation characterizing hot melt adhesive overflow; generates an adaptive angle filter to filter the acoustic emission signal, extracts the intrinsic energy spectrum, and inputs it into a temporal convolutional network to generate the blade wear evolution trajectory; and calculates the coating degradation probability distribution matrix in parallel; finally, it matches and calculates this matrix with the production scheduling sequence, issues tension balance compensation coefficients, and generates targeted maintenance work orders within the order roll change time window. The system is used to implement the above method.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for industrial equipment status, specifically to a method and system for intelligent monitoring of the status of medical dressing production line equipment. Background Technology

[0002] Medical dressings are essential supplies in clinical medicine and daily care, and their manufacturing process highly relies on continuous, automated multi-layer composite roll processing production lines. In modern medical manufacturing scenarios, the operating status of key components such as rotary cutting rollers and heat-sealing rollers not only directly determines the cutting accuracy and packaging quality of dressing products, but also profoundly affects the continuous manufacturing capability of the entire production line. Ensuring the stable operation of medical dressing production line equipment under high loads and reducing the frequency of unexpected downtime has become a core aspect of maintaining the operational efficiency of medical device manufacturing companies.

[0003] However, in actual production, fluctuations in the physical properties of multilayer composite dressings can alter the operating status of equipment. Existing equipment condition monitoring solutions often employ a single vibration amplitude exceeding-limit alarm mode. When tension imbalance between dressing layers causes localized overflow of internal hot melt adhesive and adhesion to the surface of the actuators, the high-speed operation of the equipment generates strong mechanical friction noise and abnormal vibrations. This normal disturbance caused by abnormal material fluctuations is easily confused with signals generated by actual physical wear of the cutting tools, leading to frequent false alarms. Furthermore, existing maintenance decisions are often disconnected from current order scheduling. Forced shutdowns for inspection after receiving alarms directly disrupt the continuous roll processing rhythm, resulting in large-scale scrapping of work-in-process and low roll changeover efficiency. Therefore, it is still necessary to provide an intelligent monitoring method for the equipment condition of medical dressing production lines to improve the accuracy of equipment degradation characteristic identification and the coordination of maintenance and production scheduling. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for intelligent monitoring of the status of medical dressing production line equipment to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent monitoring method for the status of medical dressing production line equipment, comprising the following steps: S1, synchronously acquiring the multi-channel material tension sequence of the multi-layer composite unwinding station, the infrared thermal distribution matrix of the heat sealing roller station, the high-frequency acoustic emission signal of the rotary cutting station, and the real-time rotation angle sequence of the rotary cutting spindle; S2, calculating the tension mismatch gradient of the multi-channel material tension sequence, and performing spatial partial correlation comparison in the infrared thermal distribution matrix in conjunction with the real-time rotation angle sequence to locate the local thermal impedance abrupt change region, and constructing a spatial mapping map of the microscopic accumulation of colloids characterizing the overflow adhesion of hot melt adhesive; S3, based on the microscopic accumulation space of the colloids... The inter-mapping diagram generates a phase-constrained adaptive angle filter, which synchronously filters out pseudo-sound waves of peeling, adhesion and tearing on a loop-by-loop basis for high-frequency acoustic emission signals. The intrinsic acoustic emission energy spectrum is extracted and input into a time-series convolutional network to generate the physical wear evolution trajectory of the cutting blade. The contour expansion rate of the local thermal impedance abrupt change region is calculated in parallel to generate the coating decay probability distribution matrix. S4, The physical wear evolution trajectory of the cutting blade and the coating decay probability distribution matrix are matched and calculated with the cross-batch production scheduling sequence in the manufacturing execution system. The tension balance compensation coefficient of the current processing batch is output and sent to the unwinding servo end. A directional maintenance work order is generated within the order roll change time window of the cross-batch production scheduling sequence.

[0006] In a preferred embodiment, the specific process of synchronously acquiring the multi-channel material tension sequence of the multi-layer composite unwinding station, the infrared thermal distribution matrix of the heat-sealing roller station, the high-frequency acoustic emission signal of the rotary cutting station, and the real-time rotation angle sequence of the rotary cutting spindle is as follows: Configure time-sensitive network nodes to uniformly send global timing pulses to the multi-layer composite unwinding station, the heat-sealing roller station, and the rotary cutting station; acquire the multi-channel material tension sequence through a distributed tension sensor array, trigger the infrared thermal imager array to scan the surface of the heat-sealing roller to generate an infrared thermal distribution matrix, acquire the high-frequency acoustic emission signal through an acoustic emission probe, and simultaneously parse the absolute encoder output message to extract the real-time rotation angle sequence; perform timestamp binding and alignment on the multi-channel material tension sequence, the infrared thermal distribution matrix, the high-frequency acoustic emission signal, and the real-time rotation angle sequence based on the global timing pulse to construct a time synchronization benchmark for multi-source heterogeneous data.

[0007] In a preferred embodiment, the specific process of calculating the tension mismatch gradient of the multi-channel material tension sequence and performing spatial partial correlation comparison in the infrared thermal distribution matrix in conjunction with the real-time rotation angle sequence to locate the local thermal impedance abrupt change region is as follows: First-order difference feature vectors of the multi-channel material tension sequence within the sliding time window are extracted; cross-layer orthogonal projection is performed on these first-order difference feature vectors to extract the interlayer tension mismatch gradient; the real-time rotation angle sequence is substituted into the infrared thermal distribution matrix as a phase offset parameter to reconstruct a three-dimensional cylindrical thermal field sequence with circumferential unfolding characteristics; the base heating temperature is filtered out from the three-dimensional cylindrical thermal field sequence to extract residual thermal fluctuation characteristics; the dynamic Pearson partial correlation coefficient between the interlayer tension mismatch gradient and the residual thermal fluctuation characteristics is calculated; the spatial extreme value coordinate point cluster area corresponding to the dynamic Pearson partial correlation coefficient is extracted to locate the local thermal impedance abrupt change region.

[0008] In a preferred embodiment, the specific process of constructing a spatial mapping map of the micro-accumulation of colloids characterizing the overflow adhesion of hot melt adhesive is as follows: extracting the edge contour features of the local thermal impedance abrupt change region, performing morphological closing operations on the edge contour features to fill and generate accumulated connected domains; performing a spatial affine transformation on the pixel coordinate system parameters of the accumulated connected domains, mapping them to the three-dimensional physical space geometric coordinate system of the heat sealing roller, and obtaining the initial accumulation physical coordinate points of the colloids; fusing the real-time rotation angle sequence to assign dynamic phase labels to the initial accumulation physical coordinate points of the colloids, aggregating the initial accumulation physical coordinate points of the colloids carrying dynamic phase labels, and generating a spatial mapping map of the micro-accumulation of colloids.

[0009] In a preferred embodiment, the specific process of generating a phase-constrained adaptive angle filter based on the colloidal micro-accumulation spatial mapping map and synchronously filtering and peeling off adhesion and tearing pseudo-sound waves of the high-frequency acoustic emission signal is as follows: the physical accumulation coordinates in the colloidal micro-accumulation spatial mapping map are converted into the absolute mechanical angle boundary of the rotary cutting spindle, and the periodic interference phase interval is defined according to the absolute mechanical angle boundary; the phase blocking operation interval of the phase-constrained adaptive angle filter is configured based on the periodic interference phase interval; when the real-time rotation angle sequence enters the phase blocking operation interval, a time-domain windowing truncation operation is performed on the high-frequency acoustic emission signal; the high-frequency acoustic emission signal after time-domain windowing truncation is reconstructed to peel off adhesion and tearing pseudo-sound waves.

[0010] In a preferred embodiment, the specific process of extracting the intrinsic acoustic emission energy spectrum and inputting it into a temporal convolutional network to generate the physical wear evolution trajectory of the cutting edge, and parallel calculating the contour expansion rate of the local thermal impedance abrupt change region to generate the coating degradation probability distribution matrix is ​​as follows: A wavelet packet transform is performed on the high-frequency acoustic emission signal after peeling, adhesion, tearing, and pseudo-acoustic waves to extract the intrinsic acoustic emission energy spectrum. This intrinsic acoustic emission energy spectrum is then input into a temporal convolutional network to extract low-frequency cumulative fatigue feature components along the time dimension, and the physical wear evolution trajectory of the cutting edge is fitted and output. A two-dimensional topological boundary pixel set of the local thermal impedance abrupt change region within adjacent sampling periods is extracted. Deformation curvature evolution analysis is performed on the two-dimensional topological boundary pixel set to extract the contour expansion rate. The contour expansion rate is mapped to the grid coordinate system of the heat-sealing roller surface to generate the coating degradation probability distribution matrix.

[0011] In a preferred embodiment, the specific process of matching and calculating the physical wear evolution trajectory of the cutting blade, the coating decay probability distribution matrix, and the cross-batch production scheduling sequence within the manufacturing execution system, and outputting the tension balance compensation coefficient of the current processing batch to the unwinding servo is as follows: The physical wear evolution trajectory of the cutting blade and the coating decay probability distribution matrix are analyzed to extract the equipment's fault-free operating time margin and obtain the current batch completion time node of the cross-batch production scheduling sequence within the manufacturing execution system; a time axis cross-comparison is performed between the equipment's fault-free operating time margin and the current batch completion time node; when the equipment's fault-free operating time margin is less than the current batch completion time node, the tension optimization control algorithm is activated, and the tension balance compensation coefficient is output by reverse derivation of the interlayer tension mismatch gradient, and the tension balance compensation coefficient is encapsulated into an industrial communication control message and sent to the unwinding servo.

[0012] In a preferred embodiment, the specific process of generating a targeted maintenance work order within the order roll changeover time window of a cross-batch production scheduling sequence is as follows: Scan the cross-batch production scheduling sequence and extract the order roll changeover time window between adjacent batches; aggregate the coordinates of the wear extreme points in the physical wear evolution trajectory of the cutting blade with the high-risk peeling areas in the coating degradation probability distribution matrix to generate a set of maintenance action attributes; perform knowledge graph semantic matching mapping between the set of maintenance action attributes and the spare parts inventory status and personnel scheduling matrix in the factory asset management system; before the start node of the order roll changeover time window, package the set of maintenance action attributes and the knowledge graph semantic matching mapping results to generate a targeted maintenance work order.

[0013] A medical dressing production line equipment status intelligent monitoring system, used to execute the aforementioned medical dressing production line equipment status intelligent monitoring method, includes: a multi-source synchronous acquisition module, used to synchronously acquire the multi-channel material tension sequence of the multi-layer composite unwinding station, the infrared thermal distribution matrix of the heat sealing roller station, the high-frequency acoustic emission signal of the rotary cutting station, and the real-time rotation angle sequence of the rotary cutting spindle; a micro-accumulation mapping module, used to calculate the tension mismatch gradient of the multi-channel material tension sequence, perform spatial partial correlation comparison in the infrared thermal distribution matrix in conjunction with the real-time rotation angle sequence, locate local thermal impedance abrupt change regions, and construct a colloidal micro-accumulation spatial mapping map characterizing hot melt adhesive overflow adhesion; and a state evolution prediction module, used for... A phase-constrained adaptive angle filter is generated based on the colloidal micro-accumulation spatial mapping map. This filter synchronously filters out pseudo-sound waves of peeling, adhesion, and tearing from the high-frequency acoustic emission signal round by round. The intrinsic acoustic emission energy spectrum is extracted and input into a temporal convolutional network to generate the physical wear evolution trajectory of the cutting edge. The contour expansion rate of the local thermal impedance abrupt change region is calculated in parallel to generate the coating degradation probability distribution matrix. The collaborative operation and maintenance decision module is used to match and calculate the physical wear evolution trajectory of the cutting edge and the coating degradation probability distribution matrix with the cross-batch production scheduling sequence in the manufacturing execution system. It outputs the tension balance compensation coefficient of the current processing batch and sends it to the unwinding servo end. It also generates a targeted maintenance work order within the order roll change time window of the cross-batch production scheduling sequence.

[0014] The technical effects and advantages of this invention are as follows: (1) A method for intelligent monitoring of equipment status in a medical dressing production line. This invention simultaneously collects multiple material tension sequences, infrared thermal distribution matrices, high-frequency acoustic emission signals, and real-time rotation angle sequences during the monitoring process. It calculates the tension mismatch gradient of the material tension sequences and performs spatial partial correlation comparison in the infrared thermal distribution matrix in conjunction with the real-time rotation angle sequences. This allows for the location of local thermal impedance abrupt changes and the construction of a spatial mapping map characterizing the microscopic accumulation of colloids during hot melt adhesive overflow. This improves upon the prior art's misjudgment of mechanical disturbances caused by material property fluctuations as equipment failures, accurately identifies the location of hot melt adhesive overflow caused by interlayer tension imbalance, and provides a spatial position conversion benchmark for the accurate filtering of subsequent interference signals.

[0015] (2) An intelligent monitoring system for the equipment status of a medical dressing production line generates a phase-constrained adaptive angle filter based on the spatial mapping diagram of colloidal micro-accumulation. This filter synchronously filters high-frequency acoustic emission signals round by round to remove pseudo-acoustic waves of adhesion and tearing. The extracted intrinsic acoustic emission energy spectrum is input into a temporal convolutional network to generate the physical wear evolution trajectory of the cutting blade. Simultaneously, the system calculates the contour expansion rate of the thermal impedance abrupt change region to generate a coating degradation probability distribution matrix. Subsequently, the system performs cross-matching calculations between the above equipment degradation evolution results and the cross-batch production sequence of the manufacturing execution system, issues tension balance compensation coefficients, and generates targeted maintenance work orders within the order roll change time window. This effectively solves the problems of environmental noise masking the true degradation characteristics and forced shutdowns disrupting the production rhythm under the traditional single-threshold alarm mode, improves the prediction accuracy of the equipment degradation status of the medical dressing production line, and enhances the dynamic adaptive capability of equipment maintenance and actual production scheduling.

[0016] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0017] Figure 1 This is a flowchart of an intelligent monitoring method for the status of equipment in a medical dressing production line according to the present invention; Figure 2 This is a schematic diagram comparing adaptive filtering of acoustic emission signals in an embodiment of the present invention; Figure 3 This is a schematic diagram of the spatial mapping of colloidal micro-accumulation in an embodiment of the present invention; Figure 4 This is a flowchart of an intelligent monitoring system for the status of equipment in a medical dressing production line according to the present invention. Detailed Implementation

[0018] This application provides an intelligent monitoring method and system for the status of equipment in a medical dressing production line. This solves the problems of existing equipment monitoring technologies being susceptible to false alarms caused by material fluctuations during multi-layer composite processing of medical dressings, as well as the disconnect between maintenance actions and production scheduling.

[0019] The overall approach of the scheme in this application embodiment is as follows: Simultaneously acquire the multi-channel material tension sequence of the multi-layer composite unwinding station, the infrared thermal distribution matrix of the heat-sealing roller station, the high-frequency acoustic emission signal of the rotary cutting station, and the real-time rotation angle sequence of the rotary cutting spindle; calculate the tension mismatch gradient of the multi-channel material tension sequence, and perform spatial partial correlation comparison in the infrared thermal distribution matrix in conjunction with the real-time rotation angle sequence to locate local thermal impedance abrupt change regions, and construct a spatial mapping map of the microscopic accumulation of the colloid characterizing the overflow adhesion of hot melt adhesive; generate a phase-constrained adaptive angle filter based on the spatial mapping map of the microscopic accumulation of the colloid. The system synchronously filters and peels off pseudo-sound waves of adhesion and tearing from high-frequency acoustic emission signals round by round, extracts the intrinsic acoustic emission energy spectrum, inputs it into a temporal convolutional network to generate the physical wear evolution trajectory of the cutting edge, and calculates the contour expansion rate of the local thermal impedance abrupt region in parallel to generate the coating decay probability distribution matrix. The physical wear evolution trajectory of the cutting edge and the coating decay probability distribution matrix are matched and calculated with the cross-batch production scheduling sequence in the manufacturing execution system, and the tension balance compensation coefficient of the current processing batch is output and sent to the unwinding servo end. A targeted maintenance work order is generated within the order roll change time window of the cross-batch production scheduling sequence.

[0020] Example 1; please refer to Figure 1 This invention provides a technical solution: an intelligent monitoring method for the status of equipment in a medical dressing production line, comprising the following steps: S1, synchronously acquiring the multi-channel material tension sequence of the multi-layer composite unwinding station, the infrared thermal distribution matrix of the heat-sealing roller station, the high-frequency acoustic emission signal of the rotary cutting station, and the real-time rotation angle sequence of the rotary cutting spindle; S2, calculating the tension mismatch gradient of the multi-channel material tension sequence, and performing spatial partial correlation comparison in the infrared thermal distribution matrix in conjunction with the real-time rotation angle sequence to locate local thermal impedance abrupt change regions and construct a spatial mapping map of the microscopic accumulation of colloids characterizing hot melt adhesive overflow adhesion; S3, based on the spatial mapping map of the microscopic accumulation of colloids... The phase-constrained adaptive angle filter is generated from the radiograph to synchronously filter out pseudo-sound waves of peeling, adhesion and tearing on a loop-by-loop basis for high-frequency acoustic emission signals. The intrinsic acoustic emission energy spectrum is extracted and input into a time-series convolutional network to generate the physical wear evolution trajectory of the cutting blade. The contour expansion rate of the local thermal impedance abrupt change region is calculated in parallel to generate the coating decay probability distribution matrix. S4, the physical wear evolution trajectory of the cutting blade and the coating decay probability distribution matrix are matched and calculated with the cross-batch production scheduling sequence in the manufacturing execution system. The tension balance compensation coefficient of the current processing batch is output and sent to the unwinding servo end. A directional maintenance work order is generated within the order roll change time window of the cross-batch production scheduling sequence.

[0021] In this implementation scheme, step S1 performs low-level physical state perception and time reference alignment operations for multi-source heterogeneous data. In the continuous roll processing of medical dressings, the system acquires the tensile force change data of each layer of nonwoven fabric or absorbent pad during the unfolding process, forming a multi-path material tension sequence. Simultaneously, it records the microscopic elastic waveforms released when the cutting tool cuts the material lattice or breaks fibers, constituting a high-frequency acoustic emission signal. Combined with the two-dimensional temperature field matrix data of the heat-sealing roller surface and the instantaneous rotation position of the cutting spindle, the system establishes a strict synchronous correlation between the operating parameters of the electromechanical components and the processing parameters of the dressing material, providing fundamental data for subsequent signal decoupling and interference troubleshooting.

[0022] Step S2 performs spatial partial correlation analysis on the cross-modal data to locate abnormal physical accumulation points of materials on the production line. The system calculates the rate of change of tensile force difference between different coating layers, i.e., the tension mismatch gradient. This physical quantity reflects the degree to which uneven stress within the multilayer composite material causes abnormal extrusion and overflow of hot melt adhesive. The system substitutes the spindle rotation position as a phase variable into the temperature field matrix to find local thermal impedance abrupt change regions on the surface of the heat-sealing roller where the original normal thermal conductivity is suddenly blocked or altered due to the adhesion of insulating adhesive. Then, it calculates the specific adhesion location distribution map of these overflowing adhesives on the three-dimensional physical surface of the mechanical roller shaft.

[0023] Step S3 utilizes a spatial physical coordinate inverse constraint signal processing algorithm to extract the true mechanical degradation characteristics of the equipment. The system converts the adhesive adhesion coordinates found in the previous step into a specific angular range during roller rotation, and configures a digital algorithm logic, namely a phase-constrained adaptive angle filter, that activates shielding or attenuation functions only within the specific rotation angle range. This algorithm eliminates the accompanying friction noise when the spindle rotates to the specific angle where the adhesive adheres, and extracts the intrinsic acoustic emission energy spectrum that only represents the cutting action of the tool. The system inputs this effective feature into a deep learning network that performs convolution operations on time-series data to predict the trend of tool dulling, and simultaneously tracks the expansion rate of the insulating adhesive accumulation area to quantitatively assess the probability of damage to the anti-stick coating on the heat-sealing roller surface.

[0024] Step S4 integrates the equipment degradation prediction results with the enterprise production scheduling system to execute fault-tolerant maintenance decisions within the industrial cyber-physical system. The system reads the sequence of each processing order and the estimated roll changeover time, i.e., the cross-batch production scheduling sequence, recorded in the factory manufacturing system. Based on the predicted remaining tool life and coating damage risk, the system calculates a set of tension balance compensation coefficients and sends them to the front-end unwinding servo motor. By fine-tuning the driver torque, it actively corrects the difference in coating tension to reduce further overflow of hot melt adhesive. Simultaneously, the system automatically schedules work orders for maintenance personnel during the material roll changeover gaps between alternating shutdowns of two production orders, achieving physical coordination between equipment condition adjustment and maintenance and production cycle.

[0025] Specifically, the process of synchronously acquiring the multi-channel material tension sequence of the multi-layer composite unwinding station, the infrared thermal distribution matrix of the heat-sealing roller station, the high-frequency acoustic emission signal of the rotary cutting station, and the real-time rotation angle sequence of the rotary cutting spindle is as follows: Configure time-sensitive network nodes to uniformly send global timing pulses to the multi-layer composite unwinding station, the heat-sealing roller station, and the rotary cutting station; acquire the multi-channel material tension sequence through a distributed tension sensor array, trigger the infrared thermal imager array to scan the surface of the heat-sealing roller to generate an infrared thermal distribution matrix, acquire the high-frequency acoustic emission signal through an acoustic emission probe, and simultaneously parse the absolute encoder output message to extract the real-time rotation angle sequence; perform timestamp binding and alignment on the multi-channel material tension sequence, the infrared thermal distribution matrix, the high-frequency acoustic emission signal, and the real-time rotation angle sequence based on the global timing pulse to construct a time synchronization benchmark for multi-source heterogeneous data.

[0026] In this implementation scheme, the physical sensing and clock synchronization of the underlying equipment in the production line are performed based on a master-slave architecture at the hardware level. Time-sensitive network nodes, acting as boundary clock masters, periodically send global timing pulses conforming to a precision time protocol to the tension transmitters at the multi-layer composite unwinding station, the infrared thermal imager array at the heat-sealing roller station, the high-frequency acoustic emission acquisition card at the rotary cutting station, and the servo driver. Upon receiving the pulse signal, each distributed hardware node locks its local numerically controlled oscillator, completing the phase and frequency alignment of the underlying hardware clock. Subsequently, the distributed tension sensor array measures the physical tensile force changes of each layer of the composite material using a strain gauge bridge; the infrared thermal imager array acquires the two-dimensional temperature matrix of the heat-sealing roller surface through non-contact thermal radiation scanning; the acoustic emission probe is attached to the sidewall of the cutting blade holder, capturing high-frequency mechanical elastic waves using a piezoelectric ceramic array; and the absolute encoder directly reads the Gray code grating signal of the spindle to output the physical rotation angle. While acquiring physical quantities, each node uses its synchronized local clock to stamp each sampling point with an initial timestamp, forming a multi-source heterogeneous data stream carrying a time tag. Because the sampling frequencies of acoustic emission, infrared vision, and mechanical angles differ by orders of magnitude, the system needs to resample and align the out-of-order and heterogeneous discrete data to a unified time grid to construct a time synchronization reference. Specifically, the initial timestamp recorded by the hardware clock is first corrected for network fluctuations to obtain the corrected absolute physical timestamp. The calculation formula is as follows: ;in, This represents the corrected absolute physical timestamp; Indicates the initial timestamp recorded by the local hardware clock; This represents the static transmission delay between the master and slave nodes as resolved by the network time synchronization protocol; Indicates the frequency offset tracking compensation term; Indicates the serial number of the data source node; This represents a discrete index indicating the local sampling period. Subsequently, a unified global time grid node is established based on the rotation period of the cutting spindle. The corrected heterogeneous data from each path are interpolated and mapped to this grid node to obtain the heterogeneous data feature values ​​aligned to the unified benchmark. The computational logic is expressed as follows: ;in, Represents the feature values ​​of heterogeneous data aligned to a unified benchmark; This represents a unified global time grid node; x represents the discrete length index of the global time grid. This represents the set of valid sample indexes located within the neighborhood of a unified global time grid node; This represents the original monitoring sequence value acquired by the u-th data source node in the v-th local sampling period; This represents the dynamic interpolation weighting coefficient based on temporal proximity. The method for determining this weighting coefficient follows an exponential decay law, and the specific calculation relationship is as follows: ;in, This represents the normalization balancing factor; This represents the attenuation control parameter. The attenuation control parameter here is based on the formula... Confirmed; among them Indicates the rated operating angular velocity of the cutting roller; This indicates the maximum allowable phase jitter threshold of the mechanical spindle. This threshold is directly read from the tool spindle mechanical tolerance configuration table provided by the original equipment manufacturer, thereby solidifying the underlying mechanical constraints into the data alignment algorithm.

[0027] Specifically, the process of calculating the tension mismatch gradient of the multi-channel material tension sequence and performing spatial partial correlation comparison in the infrared thermal distribution matrix in conjunction with the real-time rotation angle sequence to locate the local thermal impedance abrupt change region is as follows: First-order difference feature vectors of the multi-channel material tension sequence within the sliding time window are extracted; cross-layer orthogonal projection is performed on these first-order difference feature vectors to extract the interlayer tension mismatch gradient; the real-time rotation angle sequence is substituted into the infrared thermal distribution matrix as a phase offset parameter to reconstruct a three-dimensional cylindrical thermal field sequence with circumferential unfolding characteristics; the base heating temperature is filtered out from the three-dimensional cylindrical thermal field sequence to extract residual thermal fluctuation characteristics; the dynamic Pearson partial correlation coefficient between the interlayer tension mismatch gradient and the residual thermal fluctuation characteristics is calculated; the spatial extreme coordinate point clusters corresponding to the dynamic Pearson partial correlation coefficient are extracted to locate the local thermal impedance abrupt change region.

[0028] In this implementation scheme, the system opens a sliding time window in the industrial control computer's memory and extracts the first-order time difference of multiple material tension sequences within this time window to obtain the dynamic rate of change of tension. Then, the tension difference vectors corresponding to different material layers are orthogonally projected to each other, thereby separating the pure tensile force difference caused entirely by interlayer entanglement, i.e., the interlayer tension mismatch gradient. The calculation formula is: ;in, This represents the interlayer tension mismatch gradient between the p-th layer and the q-th layer within the k-th time window; The first-order difference eigenvector representing the tension of the p-th layer of material; The first-order differential eigenvector represents the tension of the q-th layer of material, which serves as a reference physical benchmark. This represents the vector norm 2. Simultaneously, the system reads data from the infrared pixel array and, based on the real-time rotation angle provided by the absolute encoder, maps the planar two-dimensional thermal image onto a cylindrical coordinate system, generating a three-dimensional cylindrical thermal field sequence that coincides with the physical appearance of the heat-sealing roller. To eliminate background interference from the normal operation of the electric heating element and ambient temperature, the system subtracts the average temperature of historical fault-free cycles (i.e., the base heating temperature) pixel by pixel to extract residual thermal fluctuation characteristics. Subsequently, within the same time-domain sliding window, the dynamic Pearson partial correlation coefficient between the tension mismatch gradient sequence and the residual thermal fluctuation characteristic sequence is calculated pixel by pixel. The calculation logic is as follows: ;in, Represents spatial pixels in cylindrical coordinate system The dynamic Pearson partial correlation coefficient; This represents the total number of sampling steps within the sliding time window; This represents the residual thermal fluctuation characteristics of the pixel at step k; and These represent the mean of the corresponding gradient and the mean of the thermal fluctuation, respectively. After completing the surface calculations for the entire roll, the system uses a threshold to filter out clusters of highly correlated spatial extremum coordinate points. This threshold is... The method for determining it is as follows: ;in, This represents the arithmetic mean of the global partial correlation coefficients; The standard deviation of the global partial correlation coefficient; This represents the relaxation tolerance coefficient. This coefficient is set by exhaustively searching for the extreme point with the lowest false alarm rate in the historical production cycle of desensitized machines that have no tool breakage but have normal glue overflow. This allows for the precise delineation of the local thermal resistance change region.

[0029] Specifically, the process of constructing a spatial mapping map of the micro-accumulation of colloids characterizing the overflow adhesion of hot melt adhesive is as follows: extract the edge contour features of the local thermal impedance abrupt change region, perform morphological closing operation on the edge contour features to fill and generate the accumulation connected domain; perform spatial affine transformation on the pixel coordinate system parameters of the accumulation connected domain, map it to the three-dimensional physical space geometric coordinate system of the heat sealing roller, and obtain the initial accumulation physical coordinate points of the colloid; fuse the real-time rotation angle sequence to give the initial accumulation physical coordinate points of the colloids a dynamic phase label, aggregate the initial accumulation physical coordinate points of the colloids carrying the dynamic phase label, and generate the spatial mapping map of the micro-accumulation of colloids.

[0030] In this implementation scheme, the system calls the edge detection operator to perform boundary gradient differentiation on the aforementioned local thermal impedance abrupt change region, extracting closed edge contour features. Considering that the hot melt adhesive overflowing on the roller surface in actual production often presents an irregular distribution and may contain tiny discontinuities that the infrared thermal imager cannot fully capture, the system calls the two-dimensional morphological closing operator, using a fixed-size rectangular structuring element to first expand the contour and then perform erosion contraction, thereby connecting and closing the discrete adhesive distribution points to generate an internally coherent aggregated connected domain. Based on this, the system reverse-projects all pixel points within the aggregated connected domain from the digital image plane onto the actual mechanical three-dimensional space of the heat-sealing roller. Due to the field-of-view distortion caused by the oblique installation of the camera, this mapping performs a strict spatial affine transformation and cylindrical surface projection compensation, the formula of which is: ;in, This represents the three-dimensional spatial coordinates of the m-th colloid's initial accumulation physical coordinate point in the device's geodetic coordinate system after transformation; Represents the row and column coordinates of pixels within the aggregated connected component; This represents the equivalent axial and circumferential physical dimensions of a single pixel obtained through vision system calibration. The mechanical drawing design radius of the heat-sealing roller; This represents the external rotation correction parameter matrix of the camera; This represents the three-dimensional translation compensation vector. Next, the system deeply binds the obtained static physical coordinates with the rotational phase of the electromechanical actuator, attaching a dynamic phase label to each initial colloid accumulation physical coordinate point. The specific calculation method for the label is as follows: ;in, Indicates the first Each coordinate point is assigned a dynamic phase label; This represents the absolute angle feedback value of the current rotating cutting spindle in real time; This indicates the fixed installation compensation angle relative to the mechanical zero point of the encoder, representing the center plane of the infrared camera's optical axis. This means that the calculation results are constrained to the range of a single mechanical circle. Combined with... Figure 3 As shown, after completing the above spatial affine transformation and assigning dynamic phase labels, the one-dimensional tension characteristics and the two-dimensional thermal impedance abrupt change characteristics are successfully mapped to a unified three-dimensional cylindrical space. Figure 3 The system visually displays high-density clusters of scattered points on the localized surface of the heat-sealing roller. These clustered points not only define the physical coordinates of the hot melt adhesive overflow and adhesion, but each coordinate point also uniquely corresponds to a rotational phase interval of the main shaft. This mapping structure from physical space to operational phase provides a rigid triggering benchmark for subsequent identification and filtering of periodic mechanical friction noise. Finally, the system structurally encapsulates and stores the three-dimensional coordinate set with such dynamic circumferential phase labels, generating a spatial mapping map of the adhesive's microscopic accumulation that characterizes the real-time changes in the adhesive's position as the roller rotates.

[0031] Specifically, the process of generating a phase-constrained adaptive angle filter based on the colloidal micro-accumulation spatial mapping map and synchronously filtering and peeling off adhesion and tearing pseudo-sound waves of the high-frequency acoustic emission signal round by round is as follows: the physical accumulation coordinates in the colloidal micro-accumulation spatial mapping map are converted into the absolute mechanical angle boundary of the rotary cutting spindle, and the periodic interference phase interval is defined according to the absolute mechanical angle boundary; the phase blocking operation interval of the phase-constrained adaptive angle filter is configured based on the periodic interference phase interval; when the real-time rotation angle sequence enters the phase blocking operation interval, a time-domain windowing truncation operation is performed on the high-frequency acoustic emission signal; the high-frequency acoustic emission signal after time-domain windowing truncation is reconstructed to peel off adhesion and tearing pseudo-sound waves.

[0032] In this implementation scheme, the system reads the three-dimensional physical coordinate set stored in the spatial mapping map of the colloidal micro-accumulation, projects it along the cylindrical surface normal of the heat-sealing roller onto the circumference of the rotating cutting spindle, and calculates the starting and ending absolute mechanical angles corresponding to each isolated colloidal adhesion zone. Based on this, the system constructs a synchronously operating phase mask group, i.e., a phase-constrained adaptive angle filter, within the digital signal processor of the industrial control computer according to these angular boundaries. When the real-time rotation angle sequence of the equipment advances and falls into the set phase-blocking operating range, the algorithm engine immediately triggers the time-domain windowing truncation operator, forcibly reducing the acoustic emission signal mixed with high-frequency noise from hot melt adhesive adhesion and tearing within the corresponding range to zero. To prevent energy leakage and spurious abrupt changes in the signal during frequency domain analysis due to truncation, the system further utilizes a cubic spline smoothing algorithm to continuously interpolate and reconstruct the data points at both ends of the truncation gap, thereby obtaining a pure waveform sequence that completely removes pseudo-sound waves. Please refer to [link to relevant documentation]. Figure 2 As shown in the figure, the dashed line represents the original high-frequency acoustic emission signal, which exhibits obvious spikes and high-frequency energy jumps within a specific rotation angle range. These peaks are essentially pseudo-noise caused by the tearing of hot melt adhesive, which is easily misjudged by existing technologies as blade chipping or severe wear. The solid line in the figure represents the intrinsic acoustic emission energy spectrum obtained after processing by the phase-constrained adaptive angle filter in this embodiment, and after implementing time-domain windowing truncation and cubic spline reconstruction. Figure 2 The comparison clearly shows that this method eliminates interference energy within the contaminated phase interval, allowing the final output waveform to smoothly and accurately reflect the physical cutting state of the cutting blade, fundamentally eliminating the interference of material property fluctuations on equipment status monitoring. The truncation and reconstruction logic of this process relies on the following mathematical model for calculation: ;in, This represents the reconstructed truncated intrinsic acoustic emission sequence value; This represents the cubic spline reconstruction smoothing operator; This represents the original high-frequency acoustic emission sequence values; C represents the discrete-time sampling point sequence number; C represents the total number of discrete colloidal saturation clusters in the colloidal micro-aggregation space mapping diagram; c represents the index number of a specific saturation cluster. Represents a rectangular time-domain windowing operator; Indicates the first The absolute rotation angle feedback parameter read at each moment; This represents the center phase point projected onto the principal axis from the c-th cluster; This represents the threshold width of the phase blocking operating range configured for this accumulation cluster. The method for determining this threshold width incorporates colloidal morphology characteristics, according to the formula... ;Calculated and obtained; where, This represents the projected surface area of ​​the cluster as calculated visually; This indicates the rated circumference parameter of the roller; This represents the redundant phase margin constant used to accommodate mechanical assembly clearance errors.

[0033] Specifically, the process of extracting the intrinsic acoustic emission energy spectrum and inputting it into a temporal convolutional network to generate the physical wear evolution trajectory of the cutting edge, and parallel calculating the contour expansion rate of the local thermal impedance abrupt change region to generate the coating degradation probability distribution matrix is ​​as follows: Wavelet packet transform is performed on the high-frequency acoustic emission signal after peeling, adhesion, tearing pseudo-sound waves to extract the intrinsic acoustic emission energy spectrum. This intrinsic acoustic emission energy spectrum is then input into a temporal convolutional network to extract low-frequency cumulative fatigue feature components along the time dimension, and the physical wear evolution trajectory of the cutting edge is fitted and output. Two-dimensional topological boundary pixel sets of the local thermal impedance abrupt change region within adjacent sampling periods are extracted. Deformation curvature evolution analysis is performed on the two-dimensional topological boundary pixel sets to extract the contour expansion rate. The contour expansion rate is mapped to the grid coordinate system of the heat-sealing roller surface to generate the coating degradation probability distribution matrix.

[0034] In this implementation scheme, the system performs wavelet packet decomposition and reconstruction on the high-frequency acoustic emission sequence obtained by the above reconstruction to extract the intrinsic acoustic emission energy spectrum containing cutting characteristics, and uses it as the bottom layer input of the neural network. To improve the generalization ability of the model training, historical operation and maintenance time series data after removing abnormal noise points is used as the training set, and a hard-wired interruption protection circuit based on the physical limit displacement of the tool holder is set at the bottom layer of the controller to prevent mechanical interference caused by prediction deviation. At the bottom layer of the algorithm logic, the system has made substantial improvements to the network structure and evaluation benchmark of the conventional temporal convolutional network. On the one hand, the receptive field step size of the dilated dilated convolutional layer is aligned and anchored with the physical rotation period of the cutting tool. On the other hand, when optimizing the backpropagation of network parameters, the mean square error loss that treats positive and negative errors equally is abandoned, and an asymmetric loss function that deeply fits the physical characteristics of the tool's unidirectional wear is designed for iterative updating, thereby fitting the physical wear evolution trajectory of the output cutting edge. The calculation logic of the asymmetric loss function is as follows: ;in This represents the value of the customized optimized asymmetric wear loss function; b represents the total number of samples in a single training batch; b represents the sample data index. This represents the predicted wear state quantity output by the forward propagation of the network model; Labels representing the actual wear data of the measured alignment; The basic fitting weight parameter that controls the overall prediction convergence; This represents the hysteresis penalty weight parameter used to severely punish prediction lag. The configuration of this hysteresis penalty weight parameter is based on a dynamically adaptive establishment rule, and the calculation method is as follows: ;in, This represents the maximum permissible physical wear clearance scale. Simultaneously, the system extracts the outermost pixel boundary of the local thermal impedance abrupt change region within adjacent time slices during the image processing flow, and performs deformation curvature calculation on the boundary nodes to obtain the contour expansion rate. The expansion rate calculation model is as follows: ;in, This represents the contour expansion rate of the j-th independent pixel node on the two-dimensional topological boundary; This represents the unit normal vector pointing outwards from the point; This represents the spatial coordinate vector of the point within the current monitoring period; This represents the spatial coordinate vector of the same tracking point within the previous monitoring period; Indicates the physical time span between adjacent sampling periods; This represents the local geometric curvature value at that point obtained using the difference operator; This represents the tear fatigue sensitivity coefficient of the anti-stick coating material on the heat-sealing roller surface. Finally, the system spreads the expansion rate parameters of all boundary nodes along the spatial grid matrix of the heat-sealing roller to form a distribution map that quantitatively presents the probability of surface anti-stick coating degradation and peeling.

[0035] Specifically, the process of matching and calculating the physical wear evolution trajectory of the cutting blade, the coating decay probability distribution matrix, and the cross-batch production scheduling sequence within the manufacturing execution system, and then outputting the tension balance compensation coefficient for the current processing batch to the unwinding servo is as follows: The physical wear evolution trajectory of the cutting blade and the coating decay probability distribution matrix are analyzed to extract the equipment's fault-free operating time margin and obtain the current batch completion time node of the cross-batch production scheduling sequence within the manufacturing execution system; a time axis cross-comparison is performed between the equipment's fault-free operating time margin and the current batch completion time node; when the equipment's fault-free operating time margin is less than the current batch completion time node, the tension optimization control algorithm is activated, and the tension balance compensation coefficient is derived in reverse by combining the interlayer tension mismatch gradient. The tension balance compensation coefficient is then encapsulated into an industrial communication control message and sent to the unwinding servo.

[0036] In this implementation scheme, the system reads the cross-batch production scheduling sequence in the Manufacturing Execution System database in real time through an industrial communication interface, extracting the planned completion time node of the currently processed medical dressing batch. Simultaneously, the system analyzes the physical wear evolution trajectory of the cutting blade output by the temporal convolutional network to determine the time point when the blade size reaches the allowable limit, and synchronously retrieves the time point when the breakage probability crosses the tolerance red line in the coating degradation probability distribution matrix. The smaller of the two time scales is selected as the overall fault-free operating time margin for the equipment. When it is found that this operating time margin is shorter than the completion time node of the current batch, it means that the equipment will not be able to support the current order until its completion. At this time, the system does not directly trigger a shutdown alarm, but instead activates the underlying tension optimization control algorithm for adaptive intervention. This control algorithm uses the pre-extracted interlayer tension mismatch gradient as input, combined with the urgency of the equipment's lifespan, to deduce the tension balance compensation coefficient sent to the unwinding servo. The specific solution model is designed as an adaptive proportional-integral controller with a lifespan decay constraint, and the calculation formula is as follows: ;in, Indicates output to the first Tension balance compensation coefficient of each unwinding shaft driver; The instantaneous scalar projection representing the interlayer tension mismatch gradient; This represents the constant integral gain. This represents the discrete cumulative number of steps in the current control cycle; This represents the time scalar required to complete the current batch, extracted from the Manufacturing Execution System. This represents the calculated overall trouble-free operating time margin for the equipment; This represents the lifespan tolerance smoothing control factor; This represents the dynamic proportional gain. To improve the dynamic response performance of the control system in the later stages of equipment operation, the method for calculating and determining this dynamic proportional gain is as follows: ,in, This is the factory-standard proportional gain. The device is subjected to aggressive adjustment of the coefficients to mitigate equipment degradation. After the controller completes the solution, the edge computing gateway encapsulates the multi-axis compensation coefficients into communication messages conforming to the industrial fieldbus protocol specification and writes them into the underlying register of the unwinding servo controller at regular intervals. By dynamically applying reverse torque compensation, further abnormal overflow of hot melt adhesive is suppressed, and the maintenance operating limit of the device is forcibly extended.

[0037] Specifically, the process of generating a targeted maintenance work order within the order roll changeover time window of a cross-batch production scheduling sequence is as follows: Scan the cross-batch production scheduling sequence and extract the order roll changeover time window between adjacent batches; aggregate the coordinates of the wear extreme points in the physical wear evolution trajectory of the cutting blade with the high-risk peeling areas in the coating degradation probability distribution matrix to generate a set of maintenance action attributes; perform knowledge graph semantic matching mapping between the set of maintenance action attributes and the spare parts inventory status and personnel scheduling matrix in the factory asset management system; before the start node of the order roll changeover time window, package the set of maintenance action attributes and the knowledge graph semantic matching mapping results to generate a targeted maintenance work order.

[0038] In this implementation plan, the system performs a sliding scan of the cross-batch production scheduling sequence in the background, actively capturing the machine cleaning or material preparation pauses during the alternation of orders for different specifications of medical dressings, and strictly defining these pauses as order roll change time windows. Subsequently, the system aggregates the three-dimensional coordinates of the physical wear extreme points in the cutting blade trajectory with the coordinates of high-risk peeling areas on the heat-sealing roller surface, structurally generating a set of maintenance action attributes reflecting the underlying defects of the equipment. At this time, the system retrieves the real-time updated spare parts inventory status table and maintenance personnel skill scheduling matrix from the factory asset management system across network segments, incorporating them as entity nodes into the pre-constructed equipment maintenance knowledge graph to perform deep semantic mapping. The matching calculation logic for maintenance needs and resources depends on the structural feature inner product model shown in the figure below: ;in, Indicates the first The maintenance action requirements and the first The overall matching degree of the knowledge graph between maintenance resource entities (materials, spare parts, or maintenance personnel); This represents a high-dimensional feature embedding vector generated by encoding the maintenance action attributes through a graph attention network; Represents the attribute embedding vector that maintains the association between resource entities; The graph-based weighted transition matrix represents the mapping strength between spatially represented maintenance actions and their dependent resources. This represents the scalar value for adjusting inventory balance based on weight. This indicates the current actual available inventory of the spare part retrieved synchronously from the asset database; This represents the safety stock threshold for this type of spare parts. The method for determining this safety stock threshold is based on the supply chain logistics model, and the formula is as follows: ,in, This indicates the average lead time for historically procured materials. This indicates the average daily consumption rate of the spare part on the production line. This represents the confidence interval control constant set based on the historical shortage probability distribution. This represents the standard deviation of the consumption rate fluctuation. The system retrieves and filters the scheduling personnel and available spare parts combinations with the highest matching score. Before the start of the order roll change time window, it packages the instruction data such as positioning coordinates, operating procedures, and material barcodes into a structured targeted maintenance work order and pushes it to the designated worker's handheld terminal via wireless network.

[0039] Example 2; please refer to Figure 4 A medical dressing production line equipment status intelligent monitoring system is used to execute the medical dressing production line equipment status intelligent monitoring method described in the embodiments. The system includes: a multi-source synchronous acquisition module for synchronously acquiring multi-channel material tension sequences at the multi-layer composite unwinding station, the infrared thermal distribution matrix at the heat-sealing roller station, the high-frequency acoustic emission signal at the rotary cutting station, and the real-time rotation angle sequence of the rotary cutting spindle; a micro-accumulation mapping module for calculating the tension mismatch gradient of the multi-channel material tension sequences, performing spatial partial correlation comparison in the infrared thermal distribution matrix based on the real-time rotation angle sequence, locating local thermal impedance abrupt change regions, and constructing a spatial mapping map of the micro-accumulation of the colloid characterizing hot melt adhesive overflow adhesion; and a state evolution prediction module. The system is used to generate a phase-constrained adaptive angle filter based on the colloidal micro-accumulation spatial mapping map, synchronously filter the pseudo-sound waves of peeling, adhesion and tearing on a loop-by-loop basis for high-frequency acoustic emission signals, extract the intrinsic acoustic emission energy spectrum and input it into a time-series convolutional network to generate the physical wear evolution trajectory of the cutting edge, and calculate the contour expansion rate of the local thermal impedance abrupt change region in parallel to generate the coating decay probability distribution matrix; the collaborative operation and maintenance decision module is used to match and calculate the physical wear evolution trajectory of the cutting edge and the coating decay probability distribution matrix with the cross-batch production scheduling sequence in the manufacturing execution system, output the tension balance compensation coefficient of the current processing batch and send it to the unwinding servo end, and generate a targeted maintenance work order within the order roll change time window of the cross-batch production scheduling sequence.

[0040] In this implementation scheme, the multi-source synchronous acquisition module serves as the system's data sensing input, directly connecting to various sensor hardware at the production line's bottom layer and performing data timestamp binding operations. Upon receiving a global timing pulse, this module uniformly drives the distributed tension sensor, infrared thermal imager array, acoustic emission probe, and absolute encoder to initiate synchronous sampling. Its internal communication interface is responsible for aggregating physical tensile force data from multiple materials, temperature field data from the heat-sealing roller surface, mechanical wave frequencies generated by cutting, and the current mechanical rotation angle of the spindle. It then aligns and stores this heterogeneous data according to a unified time grid, thereby transforming discrete underlying hardware signals into a structured data stream with strict temporal correspondence, serving as the input foundation for subsequent system analysis.

[0041] The microscopic accumulation mapping module is responsible for spatial localization of abnormal physical states and processing the structured data stream transmitted by the multi-source synchronous acquisition module. The module's internal computing unit performs differential and orthogonal projection processing on tension sequences at different levels to obtain the mismatch gradient reflecting uneven material stress. Subsequently, the module inputs the tension mismatch gradient and residual thermal characteristics (after removing background temperature) into the partial correlation analysis engine, and uses synchronously acquired rotation angle information as a phase coordinate reference to spatially map and compare one-dimensional tensile force changes with two-dimensional temperature abrupt changes within a three-dimensional cylindrical space. Through morphological processing and spatial affine transformation algorithms, the module ultimately calculates the specific physical coordinate set of the hot melt adhesive overflowing and adhering on the roller surface, generating a spatial mapping map.

[0042] The state evolution prediction module performs interference removal and quantitative calculations of equipment degradation trends, using the spatial mapping map output by the pre-module as filtering constraints. This module configures a digitized angle mask filter based on the physical coordinates of the aggregates. When receiving the underlying acoustic emission signal, it performs temporal windowing truncation and reconstruction upon encountering the rotation phase set by the mask, stripping away pseudo-acoustic waves generated by colloidal friction. Next, the module inputs the cleaned acoustic features into a built-in temporal convolutional network, calculating and fitting the physical wear trajectory of the cutting edge along the time axis. In parallel, the module performs deformation curvature tracking calculations on the boundaries of abrupt change regions in the thermal image, outputting a damage probability matrix of the anti-stick coating over time, completing the conversion from physical signals to equipment degradation indicators.

[0043] The collaborative operation and maintenance decision-making module is responsible for executing closed-loop control actions that integrate equipment health status with factory production scheduling. This module reads the production sequence and order completion nodes from the factory's manufacturing execution system via an industrial communication interface, and cross-compares them with the blade wear trajectory and coating decay matrix output by the state evolution prediction module over time. When it is determined that the remaining time to repair (MTBF) of the equipment is shorter than the current order completion time, the module's internal control algorithm engine calculates the compensation torque parameters in reverse based on the degree of tension mismatch, encapsulates it into a control message, and sends it to the unwinding servo driver for tension correction. Simultaneously, this module retrieves cross-batch roll change downtime windows from the production sequence, matches available personnel and spare parts with the asset management system, and packages them into a work order containing clearly defined maintenance coordinates.

[0044] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for intelligent monitoring of equipment status in a medical dressing production line, characterized in that, Includes the following steps: S1. Synchronously acquire the multi-channel material tension sequence of the multi-layer composite unwinding station, the infrared heat distribution matrix of the heat sealing roller station, the high-frequency acoustic emission signal of the rotary cutting station, and the real-time rotation angle sequence of the rotary cutting spindle. S2. Calculate the tension mismatch gradient of the multi-path material tension sequence, and perform spatial partial correlation comparison in the infrared thermal distribution matrix in combination with the real-time rotation angle sequence to locate the local thermal impedance abrupt region and construct a spatial mapping map of the colloidal micro-accumulation that characterizes the overflow adhesion of hot melt adhesive. S3. Generate a phase-constrained adaptive angle filter based on the colloidal micro-accumulation space mapping diagram, and synchronously filter the high-frequency acoustic emission signal in turn to peel off the adhesion and tear pseudo-sound waves. Extract the intrinsic acoustic emission energy spectrum and input it into the time-series convolutional network to generate the physical wear evolution trajectory of the cutting blade. Calculate the contour expansion rate of the local thermal impedance abrupt region in parallel to generate the coating decay probability distribution matrix. S4. Match and calculate the physical wear evolution trajectory of the cutting blade, the coating decay probability distribution matrix, and the cross-batch production scheduling sequence in the manufacturing execution system. Output the tension balance compensation coefficient of the current processing batch and send it to the unwinding servo end. Generate a targeted maintenance work order within the order roll change time window of the cross-batch production scheduling sequence.

2. The intelligent monitoring method for the status of medical dressing production line equipment according to claim 1, characterized in that: The specific process of synchronously acquiring the multi-channel material tension sequence of the multi-layer composite unwinding station, the infrared thermal distribution matrix of the heat sealing roller station, the high-frequency acoustic emission signal of the rotary cutting station, and the real-time rotation angle sequence of the rotary cutting spindle is as follows: Configure time-sensitive network nodes to uniformly send global timing pulses to the multi-layer composite unwinding station, heat sealing roll station, and rotary cutting station; acquire multi-channel material tension sequences through a distributed tension sensor array, trigger an infrared thermal imager array to scan the surface of the heat sealing roll to generate an infrared heat distribution matrix, acquire high-frequency acoustic emission signals through an acoustic emission probe, and simultaneously parse the output message of the absolute encoder to extract the real-time rotation angle sequence. Based on global timing pulses, timestamp binding and alignment are performed on multiple material tension sequences, infrared thermal distribution matrices, high-frequency acoustic emission signals, and real-time rotation angle sequences to construct a time synchronization benchmark for multi-source heterogeneous data.

3. The intelligent monitoring method for the status of medical dressing production line equipment according to claim 1, characterized in that: The specific process of calculating the tension mismatch gradient of multiple material tension sequences and performing spatial partial correlation comparison in the infrared thermal distribution matrix based on the real-time rotation angle sequence to locate local thermal impedance abrupt change regions is as follows: Extract the first-order difference feature vector of the multi-path material tension sequence within the sliding time window, perform cross-layer orthogonal projection on the first-order difference feature vector, and extract the interlayer tension mismatch gradient; The real-time rotation angle sequence is substituted into the infrared thermal distribution matrix as a phase shift parameter to reconstruct a three-dimensional cylindrical thermal field sequence with circumferential unfolding characteristics. The base heating temperature is filtered out from the three-dimensional cylindrical thermal field sequence to extract residual thermal fluctuation characteristics. The dynamic Pearson partial correlation coefficient between the interlayer tension mismatch gradient and the residual thermal fluctuation characteristics is calculated. The spatial extreme value coordinate points corresponding to the dynamic Pearson partial correlation coefficient are extracted to locate the local thermal impedance abrupt change region.

4. The intelligent monitoring method for the status of medical dressing production line equipment according to claim 1, characterized in that: The specific process for constructing a spatial mapping map of colloidal micro-accumulation characterizing hot melt adhesive overflow adhesion is as follows: Extract the edge contour features of the local thermal impedance abrupt region, and perform morphological closing operation on the edge contour features to fill and generate accumulated connected regions; Perform a spatial affine transformation on the pixel coordinate system parameters of the accumulated connected domain and map them to the three-dimensional physical space geometric coordinate system of the heat sealing roller to obtain the initial physical coordinate points of the colloid accumulation. By integrating real-time rotation angle sequences, dynamic phase labels are assigned to the initial physical coordinate points of colloid accumulation. These initial physical coordinate points carrying dynamic phase labels are then aggregated to generate a spatial mapping map of colloid micro-accumulation.

5. The intelligent monitoring method for the status of medical dressing production line equipment according to claim 1, characterized in that: The specific process of generating a phase-constrained adaptive angle filter based on the colloidal micro-accumulation space mapping map, and synchronously filtering the pseudo-acoustic waves of peeling, adhesion, and tearing of high-frequency acoustic emission signals round by round, is as follows: The physical accumulation coordinates in the colloidal micro-accumulation space mapping are converted into the absolute mechanical angle boundary of the rotary cutting spindle, and the periodic interference phase interval is defined based on the absolute mechanical angle boundary. The phase blocking operation range is configured based on the phase constraint adaptive angle filter in the periodic interference phase range; When the real-time rotation angle sequence enters the phase blocking operation range, a time-domain windowing truncation operation is performed on the high-frequency acoustic emission signal; Reconstruct the high-frequency acoustic emission signal after time-domain windowing truncation, and peel off the adhering and torn pseudo-sound waves.

6. The intelligent monitoring method for the status of medical dressing production line equipment according to claim 1, characterized in that: The specific process of extracting the intrinsic acoustic emission energy spectrum and inputting it into a temporal convolutional network to generate the physical wear evolution trajectory of the cutting edge, and parallel calculating the contour expansion rate of the local thermal impedance abrupt change region to generate the coating degradation probability distribution matrix is ​​as follows: Wavelet packet transform is performed on the high-frequency acoustic emission signal after peeling off the pseudo-sound waves of adhesion and tearing to extract the intrinsic acoustic emission energy spectrum. The intrinsic acoustic emission energy spectrum is input into a temporal convolutional network to extract low-frequency cumulative fatigue feature components along the time dimension and fit the output physical wear evolution trajectory of the cutting blade. Extract the two-dimensional topological boundary pixel set of the local thermal impedance abrupt change region within adjacent sampling periods, perform deformation curvature evolution analysis on the two-dimensional topological boundary pixel set, and extract the contour expansion rate; The contour expansion rate is mapped to the grid coordinate system on the surface of the heat-sealing roller to generate a coating degradation probability distribution matrix.

7. The intelligent monitoring method for the status of medical dressing production line equipment according to claim 1, characterized in that: The specific process of matching and calculating the physical wear evolution trajectory of the cutting blade, the coating degradation probability distribution matrix, and the cross-batch production scheduling sequence within the manufacturing execution system, and then outputting the tension balance compensation coefficient of the current processing batch to the unwinding servo is as follows: The physical wear evolution trajectory of the cutting blade and the probability distribution matrix of coating degradation are analyzed to extract the fault-free operation time margin of the equipment and obtain the current batch completion time node of the cross-batch production scheduling sequence in the manufacturing execution system. Cross-compare the equipment's fault-free operating time margin with the current batch's completion time node on the time axis; When the equipment's fault-free operating time margin is less than the current batch's completion time, the tension optimization control algorithm is activated. The tension balance compensation coefficient is then derived in reverse by combining the interlayer tension mismatch gradient. This tension balance compensation coefficient is then encapsulated into an industrial communication control message and sent to the unwinding servo.

8. The intelligent monitoring method for the status of medical dressing production line equipment according to claim 1, characterized in that: The specific process of generating a targeted maintenance work order within the order roll changeover time window across batch production scheduling sequences is as follows: Scan the production scheduling sequence across batches and extract the roll changeover time window between adjacent batches; By combining the coordinates of the wear extreme points in the physical wear evolution trajectory of the aggregated cutting blade with the high-risk peeling areas in the coating degradation probability distribution matrix, a set of maintenance action attributes is generated. The set of maintenance action attributes is mapped using knowledge graph semantic matching with the spare parts inventory status and personnel scheduling matrix in the factory asset management system. Before the start of the order roll change time window, package the maintenance action attribute set and the semantic matching mapping results of the knowledge graph, and generate a targeted maintenance work order.

9. A medical dressing production line equipment status intelligent monitoring system, used to execute the medical dressing production line equipment status intelligent monitoring method according to any one of claims 1-8, characterized in that, include: The multi-source synchronous acquisition module is used to synchronously acquire the multi-channel material tension sequence of the multi-layer composite unwinding station, the infrared heat distribution matrix of the heat sealing roller station, the high-frequency acoustic emission signal of the rotary cutting station, and the real-time rotation angle sequence of the rotary cutting spindle. The micro-aggregation mapping module is used to calculate the tension mismatch gradient of the tension sequence of multiple materials, and to perform spatial partial correlation comparison in the infrared thermal distribution matrix in combination with the real-time rotation angle sequence to locate the local thermal impedance abrupt region and construct a colloidal micro-aggregation spatial mapping map characterizing the overflow adhesion of hot melt adhesive. The state evolution prediction module is used to generate a phase-constrained adaptive angle filter based on the colloidal micro-accumulation space mapping map, synchronously filter the pseudo-sound waves of peeling, adhesion and tearing of high-frequency acoustic emission signals round by round, extract the intrinsic acoustic emission energy spectrum and input it into the temporal convolutional network to generate the physical wear evolution trajectory of the cutting edge, and calculate the contour expansion rate of the local thermal impedance abrupt region in parallel to generate the coating decay probability distribution matrix. The collaborative operation and maintenance decision module is used to match and calculate the physical wear evolution trajectory of the cutting blade, the coating decay probability distribution matrix, and the cross-batch production scheduling sequence in the manufacturing execution system. It outputs the tension balance compensation coefficient of the current processing batch and sends it to the unwinding servo end. It also generates a targeted maintenance work order within the order roll change time window of the cross-batch production scheduling sequence.

Images