CN-121981613-A - EVA (ethylene-vinyl acetate copolymer) -based laminated non-woven fabric product quality tracing method and system

CN121981613ACN 121981613 ACN121981613 ACN 121981613ACN-121981613-A

Abstract

The invention relates to the technical field of non-woven fabric production quality control, in particular to a quality tracing method and system for non-woven fabric products based on EVA (ethylene-vinyl acetate copolymer) laminating, which comprises the steps of synchronously collecting time sequence process parameters of a full production chain through a sensor network, generating an original parameter flow bound with production times, obtaining a base line, a rhythm and disturbance components through multi-scale decomposition, mapping the base line, the rhythm and the disturbance components to a raw material layer, an interface bonding layer and a surface layer of the product to generate a layered quality parameter set. And constructing a defect mode sample library, identifying a potential defect evolution path through unsupervised clustering, calculating the dynamic matching degree of the current parameter set and the early stage of the defect evolution path in real time, and triggering parameter backtracking and root positioning of a corresponding process link if the current parameter set and the defect evolution path exceed the threshold. The invention realizes layered association of technological parameters and product structures, can identify defect evolution characteristics in advance, and improves the accuracy and prepositivity of EVA coated non-woven fabric quality tracing.

Inventors

CHEN ZHAOLIE
YUAN ZHIBIN
WU CHENGWANG
WANG BO
WU WENLI

Assignees

福建惠亿美环保材料科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260403

Claims (10)

1. The EVA-based laminated non-woven fabric product quality tracing method is characterized by comprising the following steps of: Synchronously collecting time sequence technological parameters of a raw material feeding link, a melt blending link, a tape casting film coating link and a rolling link which form an EVA film coating non-woven fabric production process chain through a sensor network deployed on a production site, and generating an original parameter stream which is uniquely bound with each independent production roll; performing multi-scale decomposition on the original parameter flow, and separating a baseline component representing long-term process drift, a rhythm component representing periodic fluctuation and a disturbance component representing instantaneous abnormality; mapping the baseline component, the rhythm component and the disturbance component according to the physical structure level of the laminated non-woven fabric, and respectively associating the mapping to the raw material layer, the interface bonding layer and the surface layer to generate a layered quality parameter set; Establishing a defect mode sample library based on a layered quality parameter set of historical production volume times and a final finished product inspection record, and identifying a potential defect evolution path with commonality from the defect mode sample library by an unsupervised clustering method; And calculating the dynamic matching degree of the layered quality parameter set and the early stage of each potential defect evolution path in real time aiming at the current production volume, and triggering a quality tracing flow aiming at the current production volume when the dynamic matching degree exceeds a preset sensitive threshold, wherein the quality tracing flow comprises parameter tracing and root positioning of process links related to the matched potential defect evolution paths.
2. The method for tracing the quality of the EVA-based laminated non-woven fabric product according to claim 1, wherein the method is characterized in that the sequential process parameters of a raw material feeding link, a melt blending link, a casting laminated link and a rolling link which form a production process chain of the EVA-based laminated non-woven fabric are synchronously collected through a sensor network arranged on a production site to generate an original parameter stream which is uniquely bound with each independent production roll, and the method specifically comprises the following steps: In the raw material feeding link, collecting batch codes, feeding time and initial temperature parameters of EVA particles and non-woven fabric substrates in different batches; in the melt blending link, collecting continuous change curves of temperature, melt pressure and melt viscosity of each area of a screw of blending equipment; in the casting film coating link, collecting the temperature distribution of a die head, the temperature of a cooling roller, the traction speed and on-line thickness spectrum data; In the rolling link, collecting rolling tension, rolling diameter change and visual detection results of surface flaws; And distributing a global unique identification code for each independent production volume, splicing and packaging time sequence process parameters of all links aligned in time by taking the time stamp as an index to form an original parameter stream of the production volume.
3. The method for tracing the quality of the EVA-based laminated non-woven fabric product according to claim 2, wherein the original parameter flow is subjected to multi-scale decomposition, and a baseline component representing long-term process drift, a rhythm component representing periodic fluctuation and a disturbance component representing instantaneous abnormality are separated, and specifically comprises the following steps: Processing an original parameter flow by adopting a self-adaptive sliding window mean value filtering method, filtering high-frequency noise, and taking the obtained smooth trend line as a baseline component for representing long-term process drift; subtracting the baseline component from the original parameter stream to obtain a residual sequence, performing spectrum analysis on the residual sequence, identifying frequency components related to the equipment operation period and the charging period, extracting and reconstructing periodic components to serve as rhythm components for representing periodic fluctuation; The rhythm component is further removed from the residual sequence, the remaining signal portion is the disturbance component characterizing the transient anomaly, and a dynamic threshold is set for the disturbance component to mark the anomaly pulse.
4. The method for tracing quality of EVA-based laminated non-woven fabric product according to claim 3, wherein the baseline component, the rhythm component and the disturbance component are mapped according to the physical structure level of the laminated non-woven fabric and are respectively associated to a raw material layer, an interface bonding layer and a surface layer to generate a layered quality parameter set, and the method specifically comprises the following steps: mapping a baseline component of a melt blending link, particularly a long-term change trend of melt viscosity, into a raw material layer quality parameter affecting overall uniformity; mapping the rhythm component of the casting and laminating link, particularly the periodic fluctuation of temperature and speed, into an interface bonding layer quality parameter affecting the bonding strength of the EVA film and the non-woven fabric base material; mapping disturbance components of each link, particularly abnormal pulses exceeding a dynamic threshold, into surface quality parameters causing surface flaws or local strength weak points; and organizing the mapped raw material layer quality parameters, interface bonding layer quality parameters and surface layer quality parameters according to the same unique production volume identification codes to form a structured layered quality parameter set.
5. The method for tracing quality of EVA-based laminated non-woven fabric product according to claim 4, wherein a defect mode sample library is established based on a layered quality parameter set of historical production rolls and a final product inspection record, and potential defect evolution paths with commonality are identified from the defect mode sample library through an unsupervised clustering method, and the method specifically comprises the following steps: screening out production volume times of which the final finished product inspection record is a defective product from the historical production records, and extracting layering quality parameter sets corresponding to the production volume times to form an initial defect sample set; Slicing the layering quality parameters of each sample in the initial defect sample set at equal time intervals from the production start time to the time when the defect is detected to form a series of parameter state snapshots arranged in time sequence; inputting all the parameter state snapshots into an unsupervised clustering model, and classifying the snapshots into different clusters by the unsupervised clustering model according to the proximity degree of the parameter state snapshots in the multidimensional feature space; analyzing the transfer sequences of the series snapshots belonging to the same historical defect sample among all clusters, extracting the frequently-occurring transfer sequences, defining a potential defect evolution path evolving from an early process state to a final defect state, and establishing a path characteristic template for each path.
6. The method for tracing quality of EVA-based laminated non-woven fabric product according to claim 5, wherein the method is characterized by calculating the dynamic matching degree of the layering quality parameter set and the early stage of each potential defect evolution path in real time according to the current production volume, and specifically comprises the following steps: acquiring a layered quality parameter set from the beginning to the current moment of the current production volume in real time, and dividing the layered quality parameter set into a parameter state snapshot sequence with the same interval as that in the definition of the potential defect evolution path according to the time sequence; Comparing the current parameter state snapshot sequence with a path characteristic template of each potential defect evolution path step by step; The comparison process calculates the multidimensional feature distance between the current sequence and the corresponding path template at the center of the stage cluster at each stage, and performs weighted summation according to the time weight to obtain the integral dynamic matching degree score of the current production state and the potential defect evolution path; A real-time dynamic matching score is maintained for each potential defect evolution path.
7. The method for tracing quality of EVA-based laminated non-woven fabric product according to claim 6, wherein when the dynamic matching degree exceeds a preset sensitive threshold, triggering a quality tracing flow for production volume, wherein the quality tracing flow comprises parameter tracing and root positioning of process links involved in a matched potential defect evolution path, and specifically comprises the following steps: presetting sensitive thresholds for potential defect evolution paths of different severity levels; the dynamic matching degree scores of all potential defect evolution paths are monitored in real time, and when any score exceeds a corresponding sensitive threshold value, the system automatically triggers a quality tracing flow; The tracing flow firstly locks a potential defect evolution path with overrun matching degree, analyzes a characteristic template of the potential defect evolution path, and determines the characterized defect type and key influence stage; According to the key influence stage, reversely tracing to the corresponding raw material layer, interface bonding layer or surface layer parameters in the layering quality parameter set; And further correlating back to the original parameter flow, positioning to a production link, equipment and even a sensor which specifically generate abnormal parameter data, and generating a preliminary root cause analysis report containing a time range, a position, an abnormal parameter value and a correlation defect type.
8. The method for tracing quality of EVA-based laminated nonwoven fabric product according to claim 7, wherein said processing of the original parameter stream by adaptive sliding window mean filtering method filters out high frequency noise, and the obtained smoothed trend line is used as a baseline component for representing long-term process drift, specifically comprising: Firstly, setting an initial sliding window length for each process parameter time sequence data, wherein the sliding window length is predefined according to the physical change inertia and sampling frequency of parameters; in the data processing process, calculating the standard deviation of the data in the sliding window in real time to measure the fluctuation degree of the data in the window; Comparing the standard deviation in the current window with the standard deviation sequence of the history window, dynamically adjusting the length of the next sliding window, properly reducing the window length to improve the response speed to trend change if fluctuation is aggravated, and maintaining or properly increasing the window length to enhance the smoothing effect if the fluctuation is stable; Calculating the arithmetic average value of all time sequence data points covered by the window in the determined sliding window, and assigning the arithmetic average value as a baseline component value at the center moment of the window; And (3) enabling the sliding window to move point by point along a time axis, repeatedly calculating the average value in the window and assigning the average value to a center point, and finally obtaining a smooth trend line which penetrates through the whole production roll time axis and filters high-frequency fluctuation, wherein the smooth trend line is a baseline component representing long-term process drift.
9. The method for tracing quality of EVA-based laminated nonwoven fabric product according to claim 8, wherein the inputting all parameter status snapshots into an unsupervised clustering model, the unsupervised clustering model classifies the snapshots into different clusters according to the proximity of the parameter status snapshots in the multidimensional feature space, specifically comprises: Constructing a multidimensional feature space, wherein the dimension of the multidimensional feature space is composed of all key feature indexes of the quality parameters of the raw material layer, the quality parameters of the interface bonding layer and the quality parameters of the surface layer contained in the layering quality parameter set; Performing standardization processing on all parameter state snapshots extracted from the historical defect sample set, so as to eliminate the influence of different parameter dimensions and numerical ranges, and mapping the parameter state snapshots into a coordinate point in the multidimensional feature space; the unsupervised clustering model adopts a clustering algorithm based on density to scan the whole feature space and automatically find out the region with dense data point distribution; The clustering algorithm sets a neighborhood radius and a minimum point threshold value, and merges core points with the density reaching the core points and the density reaching the core points in the neighborhood radius, wherein the core points contain no less than the minimum point threshold value, into the same cluster; Through this process, all input parameter state snapshots are automatically divided into a number of different clusters, each cluster representing a class of process states that are similar in quality parameter characteristics, according to their spatial aggregations in the multi-dimensional feature space.
10. The EVA-based laminated non-woven fabric product quality tracing system comprises a memory, a processor and a computer program stored in the memory and running on the processor, and is characterized in that the processor realizes the steps of the EVA-based laminated non-woven fabric product quality tracing method according to any one of claims 1 to 9 when executing the computer program.

Description

EVA (ethylene-vinyl acetate copolymer) -based laminated non-woven fabric product quality tracing method and system Technical Field The invention relates to the technical field of non-woven fabric production quality control, in particular to a quality tracing method and system for non-woven fabric products based on EVA (ethylene-vinyl acetate) laminated films. Background The existing EVA laminated non-woven fabric production quality tracing technology is characterized in that a production site sensor is used for collecting technological parameters of raw material feeding, melt blending, tape casting lamination and winding links, the parameters and independent production winding times are simply bound and stored, only original parameter flows are subjected to integral recording and post-looking up, multi-scale decomposition processing is not carried out on the time-series parameter flows, and different types of change characteristics in the parameters cannot be distinguished. In the prior art, the corresponding relation between the technological parameters and the EVA laminated non-woven fabric physical structure level is not constructed, the parameter change cannot be associated and matched with the quality forming process of the raw material layer, the interface bonding layer and the surface layer, the quality tracing is only dependent on the finished product inspection result to carry out passive analysis, and the accurate correspondence between the parameter characteristics and the product structural quality cannot be realized. The existing tracing scheme can only conduct parameter investigation of a single process link on formed defects, cannot form a common defect evolution rule through historical data mining, lacks a quantitative recognition and dynamic matching mechanism for early-stage characteristics of defects, is difficult to trigger a tracing flow in an initial stage of defect formation, and has the problems of fuzzy characteristic distinguishing, wide link positioning and inaccurate root cause judgment in a tracing process. The invention needs to complete multi-scale decomposition of the production original parameter flow and hierarchical mapping with the product structure, and simultaneously identifies potential defect evolution paths through unsupervised clustering, so as to realize dynamic matching of the current production parameters and early defect characteristics and complete accurate backtracking and positioning of defect sources. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a quality tracing method and system for EVA-based laminated non-woven fabric products. In order to achieve the aim, the invention adopts the following technical scheme that the EVA-based laminated non-woven fabric product quality tracing method comprises the following steps: Synchronously collecting time sequence technological parameters of a raw material feeding link, a melt blending link, a tape casting film coating link and a rolling link which form an EVA film coating non-woven fabric production process chain through a sensor network deployed on a production site, and generating an original parameter stream which is uniquely bound with each independent production roll; performing multi-scale decomposition on the original parameter flow, and separating a baseline component representing long-term process drift, a rhythm component representing periodic fluctuation and a disturbance component representing instantaneous abnormality; mapping the baseline component, the rhythm component and the disturbance component according to the physical structure level of the laminated non-woven fabric, and respectively associating the mapping to the raw material layer, the interface bonding layer and the surface layer to generate a layered quality parameter set; Establishing a defect mode sample library based on a layered quality parameter set of historical production volume times and a final finished product inspection record, and identifying a potential defect evolution path with commonality from the defect mode sample library by an unsupervised clustering method; And calculating the dynamic matching degree of the layered quality parameter set and the early stage of each potential defect evolution path in real time aiming at the current production volume, and triggering a quality tracing flow aiming at the current production volume when the dynamic matching degree exceeds a preset sensitive threshold, wherein the quality tracing flow comprises parameter tracing and root positioning of process links related to the matched potential defect evolution paths. As a further scheme of the invention, through a sensor network deployed on a production site, time sequence process parameters of a raw material feeding link, a melt blending link, a tape casting film coating link and a rolling link which form an EVA film coating non-woven fabric production process chain are synchronously collect