A tarpaulin coating thickness uniformity laser detection method

Abstract

The application discloses a tarpaulin coating thickness uniformity laser detection method, and belongs to the field of on-line laser detection of coiled material coating. In order to solve the problems that the total thickness is mixed by the base cloth thickness and the tension deformation, the single station measurement is difficult to distinguish the coating abnormality, and the abnormal position is unstable when the speed changes, the application samples according to the equal material length through the pre-coating and post-coating double laser profile stations, and carries out the same material coordinate mapping and base cloth strain thickness conversion. Then, the effective anchor points of the uncoated reference area are used to solve the base cloth thickness correction field and the uncertainty field, the net coating thickness map is obtained, the uniformity, the abnormal area and the measurement state are output, and the technical effects of reducing the base cloth change misjudgment and stably positioning the coating abnormality are realized.

Description

Technical Field

[0001] This invention relates to the field of online laser inspection of coil coatings, and more particularly to a laser inspection method for the uniformity of coating thickness on tarpaulins. Background Technology

[0002] Tarpaulins are typically coated with a waterproof, weather-resistant, or abrasion-resistant coating on the surface of a base fabric using a continuous conveyor belt coating process. During the coating process, variations in base fabric tension, coating viscosity, coating roller gap, and production speed can cause coating thickness fluctuations in both the width and roll length directions. To adjust the process promptly and pinpoint defects, it is necessary to continuously obtain coating thickness information covering the effective width of the fabric on the production line.

[0003] Existing spot thickness measurements can only reflect limited locations and are insufficient to cover the entire width and continuous roll length. Single-station laser measurements measure the total thickness of the base fabric and coating. Differences in the base fabric's partial thickness, longitudinal and transverse strain caused by tension, and the belt's movement all contribute to the measurement results, making it easy to mistake changes in the base fabric for coating anomalies. If the two measurement data points are only aligned according to the sampling time, mismatches can occur in the same material location when the production speed changes, making it difficult to stably output the roll length coordinates of abnormal areas.

[0004] Therefore, there is a need for a laser detection method for the uniformity of tarpaulin coating thickness that can overcome the shortcomings of the existing technology. Summary of the Invention

[0005] One objective of this invention is to propose a laser detection method for the uniformity of tarpaulin coating thickness. Addressing the problems in existing technologies such as the mixing of base fabric thickness and tension deformation into the total post-coating thickness, difficulty in aligning cross-station data to the same material unit, potential contamination of compensation results by low-quality echoes, and instability of abnormal positions under variable speed conditions, this invention proposes a method where two laser profilometry stations collect pulses based on the material length (e.g., encoder) before and after coating. After compensation by the measurement station, the pulses are mapped to a unified material coordinate system composed of the cumulative strain-free travel length and the transverse material position. The predicted base fabric thickness is generated using the longitudinal and transverse strain of the base fabric and its thickness calibration relationship. Furthermore, the base fabric thickness correction field and uncertainty field are solved using effective anchor points in the uncoated reference area. This invention achieves the technical effects of separating the net coating thickness from the total thickness, reducing misjudgments of base fabric variations, preventing invalid data from entering the over-thinness determination, and stably outputting abnormal material coordinates.

[0006] This invention provides a laser detection method for the uniformity of tarpaulin coating thickness, comprising: S1. At the upstream and downstream of the tarpaulin coating station, laser profile measurement stations with opposite directions are used to collect the thickness of the base fabric before coating, the total thickness of the tarpaulin after coating, tension, width edge and laser echo quality according to the material length pulse of the encoder, and after compensation, the full-width thickness matrix before coating and the full-width total thickness matrix after coating are formed. S2. Based on the equal material length pulse, tension and width edge, map the two thickness matrices to a unified material coordinate system represented by the cumulative strain-free travel length and the lateral material position to obtain the same material correspondence between the post-coating mesh and the pre-coating base fabric unit. S3. Based on the corresponding material relationship and the longitudinal and transverse strain and thickness calibration relationship of the base fabric, the thickness of the base fabric unit before coating is converted into the predicted base fabric thickness under tension after coating. S4. Map the measured thickness of the base fabric in the uncoated reference area after coating to the same material element, select the grids whose laser echo quality meets the preset effective conditions as reference anchor points, take the difference between the measured and predicted base fabric thickness at the anchor point as the boundary residual, and use the transverse tension or transverse width strain partition as the propagation constraint to solve the base fabric thickness correction field and uncertainty field. S5. Correct the predicted base fabric thickness using the base fabric thickness correction field, subtract the corrected predicted base fabric thickness from the total thickness matrix after coating to obtain the net coating thickness map, calculate the transverse uniformity index and longitudinal uniformity index, merge coating abnormal areas, and output the measurement status and material coordinate information according to the uncertainty field.

[0007] Optionally, S1 includes: at each laser profile measuring station, the upper laser profile probe and the lower laser profile probe measure the upper surface profile and the lower surface profile of the tarpaulin respectively, and obtain the thickness based on the upper surface profile and the lower surface profile at the same lateral position. In the absence of strip material, the reference distance and optical axis tilt angle between the upper and lower laser profile probes are calibrated. Temperature compensation is performed using the calibration relationship between the temperature of the measuring station and the drift of the reference distance. Roll runout compensation is performed using the periodic displacement sequence corresponding to the angular position of the roll body. The compensated thickness, tension distribution, width edge, conveyor posture, and laser echo quality are associated with and saved as pulse identifiers of the same material length.

[0008] Optionally, S2 includes: reading the tension-longitudinal strain calibration relationship and the strain-free reference width from the detection parameter file associated with the base fabric type; determining the pre-coating longitudinal strain and post-coating longitudinal strain based on the tension collected by the pre-coating laser profilometry station and the post-coating laser profilometry station respectively; and determining the pre-coating transverse width strain and post-coating transverse width strain based on the change of the measured width defined by the two sides of the width edge detected by each station relative to the strain-free reference width. Based on the increase in the length of the conveyor belt corresponding to adjacent pulses of equal material length and the longitudinal strain of the corresponding measuring station, the increase in the length without strain is calculated segment by segment and then accumulated to obtain the cumulative length without strain. The transverse sampling position is converted into the transverse material position according to its position ratio relative to the two sides of the width edge of the corresponding measurement station; The same material correspondence is established between the pre-coated base fabric unit and the post-coated mesh with the same cumulative strain-free travel length and the same transverse material position.

[0009] Optionally, S3 includes: applying multiple longitudinal and transverse strain states to a calibration sample of the same type as the base fabric to be tested, recording the change in thickness relative to the unstrained thickness under each strain state, and forming the relationship between longitudinal and transverse strain and thickness of the base fabric. Based on the pre-coating longitudinal strain and pre-coating transverse width strain of the pre-coating base fabric unit determined in step S2, and the post-coating longitudinal strain and post-coating transverse width strain of the same material post-coating mesh, the thickness conversion amount is determined from the relationship between the longitudinal and transverse strains and the thickness of the base fabric, and the compensated thickness of the pre-coating base fabric unit is converted into the predicted base fabric thickness according to the thickness conversion amount.

[0010] Optionally, S4 includes: obtaining the uncoated reference area from a continuous uncoated edge strip disposed on at least one side of the coating edge or from uncoated reference windows spaced apart along the conveying direction; For each candidate grid in the uncoated reference area, the laser echo quality is determined to meet the preset effective conditions based on the comparison results of echo intensity, light stripe width, saturated pixel ratio, thickness residual of upper and lower laser profile probes and thickness change of adjacent frames with the corresponding effective range or effective upper limit. Candidate meshes that meet the preset valid conditions and complete the same material mapping are determined as reference anchor points, and the boundary residuals and their material coordinates at the reference anchor points are input into the solution process of the base fabric thickness correction field and uncertainty field. Furthermore, each of the aforementioned effective ranges and effective upper limits is determined by a calibration sample of the base fabric with no coating and a qualified thickness; The distribution of echo intensity, light stripe width, saturated pixel ratio, thickness residual of upper and lower laser profile probes, and thickness variation of adjacent frames of the base fabric calibration samples are statistically analyzed. The range of values ​​covering the preset sample ratio is determined as the corresponding effective range, and the upper limit of the residual or the upper limit of the variation covering the preset sample ratio is determined as the corresponding effective upper limit. The preset sample ratio is recorded in a detection parameter file associated with the base fabric type; Furthermore, the solution process for the base fabric thickness correction field and uncertainty field includes: dividing the propagation zone according to the boundary position of the transverse tension change or the boundary position of the transverse width strain change; Within each propagation zone, the goal is to minimize the weighted sum of the boundary residual fitting error of the reference anchor point, the difference in correction amount between adjacent grids, and the difference in correction amount across zones. The weight of the boundary residual fitting error is determined by the laser echo quality of the reference anchor point, and the correction amount of each grid is determined under the constraint of maintaining the continuity of the correction amount across zones. The uncertainty of each grid is determined based on the material coordinate distance from each grid to the reference anchor point, the degree of dispersion of the boundary residual within the propagation zone, and the laser echo quality of the reference anchor point.

[0011] Optionally, S5 includes: determining the lateral uniformity index based on the dispersion of the net coating thickness of each lateral grid at the same cumulative strain-free travel length, and determining the longitudinal uniformity index based on the dispersion of the net coating thickness of each longitudinal grid within a preset roll length interval at the same lateral material position. The upper limit of coating thickness, lower limit of coating thickness, minimum physical length and minimum physical area are read from the coating process formula. The connected domains of grids that are edge-adjacent or corner-adjacent in the same material coordinate system and have the same direction of exceeding the limit are merged. The connected domains whose longitudinal physical length reaches the minimum physical length or whose area reaches the minimum physical area are identified as coating abnormal regions. When the laser echo quality of the mesh does not meet the preset valid conditions, an invalid state is output. When the laser echo quality of the mesh meets the preset valid conditions and there is no available reference anchor point, an invalid state is output. When the laser echo quality of the mesh meets the preset valid conditions, there is an available reference anchor point and the uncertainty does not exceed the preset uncertainty upper limit, an valid state is output. When the laser echo quality of the mesh meets the preset valid conditions, there is an available reference anchor point and the uncertainty exceeds the preset uncertainty upper limit, a retest state is output, and it is prohibited to judge the mesh in the retest state or invalid state as too thin anomaly. Furthermore, the material coordinate information of the abnormal coating area includes the starting position of the width, the ending position of the width, the starting position of the cumulative unstrained roll length, the ending position of the cumulative unstrained roll length, the area, the extreme value of the net coating thickness, and the measurement confidence value. The measurement confidence value is determined based on the laser echo quality and uncertainty of the effective state grid within the region, and the measurement confidence value does not decrease when the number of effective items corresponding to the laser echo quality increases, nor does it increase when the uncertainty increases. Furthermore, the material coordinate information, lateral uniformity index, longitudinal uniformity index, and valid, retested, or invalid measurement status of the abnormal coating area are associated with the corresponding equal material length pulse markers to form a detection record; The roll material verification positioning interval is generated based on the starting position and ending position of the cumulative unstrained roll length, and the material space interval between adjacent detection records is determined by pulses of equal material length after the production speed changes.

[0012] The beneficial effects of this invention are: 1. Using encoders and other material length pulses and unified material coordinates, the data before and after coating are mapped to the same material, so that the sampling points before and after the production speed change are still arranged according to the material space interval, which makes it easy to locate abnormal areas according to the width position and cumulative strain-free roll length.

[0013] 2. By using the relationship between longitudinal and transverse strain and thickness calibration of the base fabric, the thickness of the base fabric before coating is converted to the tension state after coating, and the thickness of the base fabric is corrected by the online measured anchor point of the uncoated reference area. This helps to reduce the probability of base fabric layout differences and cross-site model drift being misjudged as coating anomalies.

[0014] 3. Based on the laser echo quality, the availability of the reference anchor point, and the uncertainty, a mutually exclusive state of valid, retested, and invalid is formed, and retested or invalid meshes are prohibited from entering the overly thin judgment, which helps to avoid low-quality echo contamination compensation and anomaly judgment. Attached Figure Description

[0015] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a laser detection method for the uniformity of tarpaulin coating thickness according to the present invention.

[0016] Figure 2 This is a flowchart of the uncertainty correction process for the S4 anchor point screening and partitioning solution in this invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0018] refer to Figures 1-2 A laser detection method for the uniformity of tarpaulin coating thickness, comprising: S1. At the upstream and downstream of the tarpaulin coating station, laser profile measurement stations with opposite directions are used to collect the thickness of the base fabric before coating, the total thickness of the tarpaulin after coating, tension, width edge and laser echo quality according to the material length pulse of the encoder, and after compensation, the full-width thickness matrix before coating and the full-width total thickness matrix after coating are formed. S2. Based on the equal material length pulse, tension and width edge, map the two thickness matrices to a unified material coordinate system represented by the cumulative strain-free travel length and the lateral material position to obtain the same material correspondence between the post-coating mesh and the pre-coating base fabric unit. S3. Based on the corresponding material relationship and the longitudinal and transverse strain and thickness calibration relationship of the base fabric, the thickness of the base fabric unit before coating is converted into the predicted base fabric thickness under tension after coating. S4. Map the measured thickness of the base fabric in the uncoated reference area after coating to the same material element, select the grids whose laser echo quality meets the preset effective conditions as reference anchor points, take the difference between the measured and predicted base fabric thickness at the anchor point as the boundary residual, and use the transverse tension or transverse width strain partition as the propagation constraint to solve the base fabric thickness correction field and uncertainty field. S5. Correct the predicted base fabric thickness using the base fabric thickness correction field, subtract the corrected predicted base fabric thickness from the total thickness matrix after coating to obtain the net coating thickness map, calculate the transverse uniformity index and longitudinal uniformity index, merge coating abnormal areas, and output the measurement status and material coordinate information according to the uncertainty field.

[0019] In this specific embodiment, S1 includes: In this specific embodiment, the object of inspection is a polyester fiber tarpaulin base fabric with a width of 2.10 m and a nominal base fabric thickness of 520 μm. The coating controller establishes a roll inspection task record for the current roll, recording the roll identification, base fabric type, coating formula version, inspection parameter file version, upstream pre-coating laser profile measurement station identification, and downstream post-coating laser profile measurement station identification. Both measurement stations are set in the tension stabilizing roller section, with the upstream station located in front of the coating head and the downstream station located after the drying and shaping section. Each measurement station includes vertically opposite linear laser profile probes, a tension acquisition device, left and right edge detection channels, a measuring roller angle encoder, an in-station temperature sensor, and an equal material length pulse latch. The controller generates an equal material length pulse identifier with a material feed amount of 1 mm, and uses this identifier as the longitudinal primary key for the data records of the two measurement stations. Each line laser profile probe outputs 2048 profile sampling points along the width direction when the same pulse arrives. The upper probe records the upper surface distance, echo intensity, light stripe width, and saturation pixel ratio, while the lower probe records the corresponding fields on the lower surface. The in-station data acquisition unit converts the pixel coordinates to millimeter coordinates according to the probe's factory pixel calibration table and completes the lateral indexing and registration of the upper and lower probes using a rigid calibration plate at the same measurement station. When the lateral position difference between the upper and lower profiles does not exceed 0.15 mm, they are assigned to the same lateral sampling position. Sampling points exceeding this difference are first linearly interpolated according to adjacent valid points. When the width of continuous missing points exceeds 5 mm, the corresponding grid is marked as echo missing and does not form an effective thickness. In the calibration state with no strip material and the measuring roller stationary, traceable ceramic plates are placed sequentially at the center and left and right quarter-width positions of the width. The no-load reference distance and optical axis tilt angle between the upper and lower probes are collected to form a station-level geometric calibration record. The measuring station is heated from 20 ℃ to 45 ℃ and the no-load measurement is repeated at 5 ℃ intervals. The drift coefficient of the reference distance with temperature is fitted by least squares linear fitting. The geometric calibration record saves the reference distance, upper and lower optical axis tilt angle, reference temperature, temperature drift coefficient, calibration date and upper limit of calibration residual. Before the current roll material is started, the record that matches the measuring station identification and has not exceeded the 30-day validity period is read. If there is no match or the residual exceeds 3 μm, the data of that station is stopped from being entered into the matrix. For the measurement station identification The pulse sequence number of the material length is... The horizontal sampling sequence number is The grid, the controller according to Calculate the thickness after compensation; This represents the compensated thickness of the corresponding mesh, in μm. Station markings indicating whether the station is painted before or after the station. Indicates the pulse number for materials of equal length. Indicates the horizontal sampling sequence number. This represents the unloaded reference distance between the upper and lower probes at the reference temperature, in μm. This represents the temperature drift coefficient at the reference distance, expressed in μm / ℃. This indicates the temperature inside the station at the time of sampling, in °C. This represents the reference temperature in the geometric calibration record, in °C. and These represent the surface distances output by the upper and lower probes, respectively, in μm. and These represent the tilt angles of the upper and lower probe optical axes relative to the calibration normal, in rad. This represents the cosine function, whose input is the optical axis tilt angle in units of rad, and whose output is the dimensionless projection coefficient. and These represent the periodic runout displacements obtained by looking up the upper and lower measuring roller angle positions, in μm. and These represent the angular positions of the upper and lower measuring rollers during sampling, in rad. The roll runout displacement sequence is established by collecting three empty distances at 0.5° angular position intervals under no-strip conditions. The median of the three displacements at the same angular position is taken and the mean value of the whole circumference is subtracted to obtain a periodic displacement sequence of 720 angular position key values ​​for one circumference. The sequence fields include measurement station identifier, roll identifier, angular position, displacement value, displacement unit, and version number. During operation, the roll runout is compensated by using the periodic displacement sequence corresponding to the angular position of the roll. Specifically, the latch value of the measuring roll angle encoder is interpolated according to the adjacent angular positions. The interpolated displacements of the upper and lower rolls are respectively included in the compensation thickness calculation. When the roll angle signal is lost for more than two consecutive pulses, the thickness value is retained but the runout compensation effective bit in the echo quality record is cleared to avoid the uncompensated thickness from participating in the reference anchor point screening. The tension acquisition device collects data from the tension sensors on the left and right sides within a 2 ms window before and after each pulse of equal material length. After discarding readings that exceed the sensor's range, the median is taken to form the left tension, right tension, and transverse tension distribution obtained by linear interpolation of the two, all in N / m. The left and right edge detection channels determine the width edge based on the transition position from the continuous effective point to the no-echo zone of the echo intensity, and suppress flypoints using the median of the edge positions of the most recent five pulses. If the amplitude of the pulse changes more than 8 mm relative to the median, the edge state is marked as to be retested, and the edge position of the previous effective pulse is used only for belt conveyor attitude estimation and is not used for subsequent uniformity statistics. Laser echo quality recording stores echo intensity, light stripe width, saturation pixel ratio, upper and lower contour registration difference, independent reconstruction thickness residual of two subframes, thickness before and after compensation, thickness change between adjacent frames, effective bits of jitter compensation, and edge state on a per-grid basis. The echo intensity of the upper and lower probes is normalized to 0 to 1 according to their respective full-scale range; among them, the upper and lower relative probes are recorded at the beginning and end of two 0.5 mm pulses of the same material length. The ms subframe reads the distance between the upper and lower surfaces respectively, and independently reconstructs the thickness according to the compensated thickness formula in this step. The controller takes the absolute value of the difference between the reconstructed thicknesses of the two subframes to form the dual-subframe independent reconstructed thickness residual, with the unit being μm. If any subframe field is missing, the residual is written as a null value and a quality invalid reason code. The dual-subframe independent reconstructed thickness residual is used to check the repeatability of the thickness measurement. It is a different quality quantity from the upper and lower contour registration difference, which is in mm and used for lateral index alignment. The controller also forms the belt skew angle based on the angle between the left and right edge lines and the lateral reference axis of the measuring station, and forms the lateral displacement based on the offset of the width center relative to the station center. The skew angle, lateral displacement, left and right tension, and lateral tension distribution are all associated with the same material length pulse identifier for S2 to perform material coordinate conversion. The controller arranges the effective lateral sampling positions from left to right for the pre-coating station and the post-coating station respectively, and writes them into the thickness matrix using the pulse sequence number of equal material length as the row index and the lateral sampling sequence number as the column index. The pre-coating matrix unit stores the base fabric thickness after compensation and its echo quality reference, and the post-coating matrix unit stores the tarpaulin total thickness and its echo quality reference. The metadata of the two matrices stores the roll material identifier, station identifier, pulse start and end sequence number, lateral coordinate calibration version, temperature drift record version, and roller runout sequence version. Missing units are written with null values ​​and missing reason codes, and are not replaced with zero thickness. After completing a 256-pulse data block, the pre-coating full-width thickness matrix block, the post-coating full-width total thickness matrix block, tension record, width edge record, belt travel posture record, and echo quality record are written into the roll material detection cache, and a coordinate mapping task containing the data block identifier and pulse range is issued to S2.

[0020] In this specific embodiment, S2 includes: The coordinate mapping module of S2 reads the same roll material detection cache written by S1 and pairs the pre-coating station and post-coating station records according to the roll material identifier, data block identifier, and equal material length pulse range. The detection parameter file uses the base fabric type and batch grade as the joint key to save the strain-free reference width, tension value nodes, longitudinal strain corresponding to each tension node, transverse position interpolation tolerance, and file version. In this specific embodiment, the strain-free reference width corresponding to the 2.10 m detection object is 2.100 m, the tension nodes cover 0 to 2500 N / m, the longitudinal strain is applied step by step by the same batch of base fabric strips under constant temperature and humidity conditions and calibrated by an extensometer. During operation, piecewise linear interpolation is performed between adjacent tension nodes. When the tension exceeds the calibration range, the strain is taken as the nearest endpoint value and the coordinate state is marked as extrapolated. For any measuring station, the coordinate mapping module uses the average value of the left and right tension sensors as the longitudinal tension input for that pulse, obtains the longitudinal strain based on the detection parameter file, and then... Calculate the transverse width strain; The station identifier is And the pulse sequence number is The transverse width strain at that time is a dimensionless quantity. This indicates the measured width between the left and right edges of the swath at this station, in meters (m). This indicates the strain-free reference width associated with the base fabric type, in meters (m). Station markings indicating whether the station is painted before or after the station. The pulse number indicates the material length; before publishing the same material relationship, the coordinate mapping module verifies that the transverse width strain falls within the calibration node range or allowable extrapolation range of the S3 base fabric calibration model. If it exceeds the allowable extrapolation range, the predicted input state is marked as invalid and the unit is prevented from entering the subsequent correction and coating determination; the original field is still retained when the extrapolation state, the edge state to be retested, or any edge is missing, but the corresponding unit is not set as an available same material relationship; To form a local transverse width strain field with transverse position index, the detection parameter file also stores a piecewise linear calibration table of transverse tension and local transverse width strain using the base fabric type, batch grade, and transverse material segment as joint keys. This calibration table is obtained by sampling the same batch of base fabric every 100 mm in the width direction and applying transverse linear loads stepwise on a biaxial tensioning table, recording the transverse material segment, tension node, corresponding local transverse width strain, repeatability standard deviation, and version number. For each pulse of equal material length and each 2 mm transverse grid column, the coordinate mapping module reads the transverse position tension value obtained by linear interpolation of S1 from the left and right tension sensors, and performs piecewise linear interpolation in the calibration table of the corresponding transverse material segment to obtain the local transverse width strain. The recorded fields include station identifier, pulse number, transverse grid column, transverse material segment, local tension, local transverse width strain, calibration table version, and local strain state. When local tension is missing, the corresponding segment calibration table is missing, or the tension exceeds the calibration node range, the local strain state of that grid column is written as missing and no endpoint extrapolation is performed; linear interpolation is only performed when there are valid local strain values ​​within 4 mm to the left and right of a missing grid column, and no interpolation is performed when there are more than one consecutive missing grid column or when it is located at the edge of the width. The aforementioned full-width transverse strain continues to be used as the input for S3 thickness calibration conversion, and the local transverse strain field is only used as the input for S4 to form the propagation boundary along the transverse material position, and is published with the unified material coordinate index; For the station identifier is Two adjacent pulses of equal material length, according to Convert the unstrained length increment and follow the instructions. Cumulative strain-free conveyor length; Indicates the first The pulse to the first The unstrained length increment between pulses, in meters. This represents the increment of belt feed length under the current tension condition, calculated by the encoder based on roller diameter and pulse count, in meters. This represents the longitudinal strain obtained by interpolating the current station tension in the calibration relationship; it is a dimensionless quantity. Indicates as of the date The cumulative strain-free transport length of each pulse, in meters. This indicates the cumulative strain-free travel length up to the previous pulse, in meters (m). and These represent the measuring station identifier and pulse sequence number, respectively; when the denominator is less than 0.98 or greater than 1.08, this segment is marked as strain exceeding the limit and will not participate in the same material matching. Both the pre-coating station and the post-coating station establish a zero point for the cumulative strain-free travel length at the beginning of the roll material using the same temperature-resistant reflective mark manually attached. After detecting the echo characteristics of the mark, the coordinate mapping module latches the corresponding pulse and uses the zero-point pulses of the two stations to eliminate the fixed offset caused by the installation distance between the stations. If the reflective mark is not recognized at any station during production, the encoder pulse difference recorded at the start of coating and the inter-station length calculated by strain-free transformation are used as the zero-point offset. At the same time, the registration source is written as encoder calculation. Subsequently, the remaining longitudinal offset of no more than 10 mm is corrected at the first reliable uncoated reference window using the thickness texture cross-correlation peak value. If it exceeds this range, the release of the same material relationship is suspended. The station identifier is Pulse sequence number is Horizontal sampling position is The unit, according to Convert the horizontal material position; This indicates the lateral material position in a unified material coordinate system, in meters (m). Indicates the first under the transverse reference axis of the measuring station The location of each sampling point is shown in meters (m). and These represent the left and right edge positions detected by the same pulse, respectively, in meters (m). This indicates the strain-free reference width, in meters (m). , and These represent the measurement station identifier, pulse number, and transverse sampling number, respectively. This proportional conversion simultaneously eliminates transverse displacement and width scaling. The transverse material position is limited to the range of 0 to the strain-free reference width. Points exceeding the boundary are deleted and the reason for exceeding the boundary is recorded. The coordinate mapping module establishes a material coordinate record for each thickness matrix unit. The fields include roll material identifier, source station identifier, source pulse sequence number, source transverse sampling sequence number, cumulative strain-free tape length, transverse material position, longitudinal strain, transverse width strain, coordinate status, and original echo quality reference. The cumulative strain-free tape length is quantized with a 1 mm grid, and the transverse material position is quantized with a 2 mm grid. When multiple sampling points fall within the same grid, the thickness value is formed by weighting the values ​​based on the inverse of the distance from the grid center. At the same time, the number of participating points and the maximum coordinate deviation are written into the record. When the participating points are empty, the grid is left blank. For each post-coating mesh, the module searches the pre-coating material coordinate index for candidate base fabric elements with a cumulative strain-free band length difference not exceeding 0.6 mm and a lateral material position difference not exceeding 1.2 mm, and sorts them in ascending order of the sum of squares of the two normalized coordinate differences. If the minimum value is unique, the pre-coating element is identified as a homogeneous element. If there are identical minimum values, the element with more valid echo quality items is selected first, followed by the element with the lower source pulse number. If there are no candidates or the minimum longitudinal difference exceeds the tolerance, the homogeneous state is written as unmatched. For post-coating meshes located between four pre-coating meshes and all four meshes are valid, the four pre-coating element identifiers and bilinear interpolation weights are saved. The weights are normalized to 1 to maintain the continuity of the thickness texture in the unified material coordinates. The final material correspondence record saves the post-coating mesh identifier, one or four pre-coating base fabric unit identifiers, corresponding interpolation weights, longitudinal and transverse strains at both ends, longitudinal and transverse coordinate differences, matching status, and coordinate parameter versions. The coordinate mapping module sorts and writes the data into the unified material coordinate index according to the cumulative strain-free strip length. It does not delete unmatched meshes, but instead uses them to carry the reason code to subsequent status judgments. After completing the current data block, the unified material coordinate index, pre-coating and post-coating strain fields, and material correspondence are published to S3. The values ​​used to correct the longitudinal offset between stations are written back to the roll material inspection task record to ensure that subsequent data blocks use the same material coordinate reference.

[0021] In this specific embodiment, S3 includes: The base fabric thickness conversion module of S3 reads the same material correspondence and longitudinal and transverse strains at both ends of S2, and reads the base fabric longitudinal and transverse strain and thickness calibration model that matches the base fabric type, fabric structure and batch grade from the test parameter file; the calibration sample is cut from three strips each from the left, middle and right positions of the same batch of base fabric of the roll to be tested, and conditioned for 24 h at 23 ℃ and 50% relative humidity. The combined states of longitudinal strain 0%, 1%, 2% and 3% and transverse strain -1%, 0%, 1% and 2% are applied using a biaxial tensile fixture. After each combination is stabilized for 30 s, the thickness is collected with the same upper and lower laser probes as the production measurement station, and the thickness in the strain-free state is used as the normalization benchmark. The calibration module removes thickness values ​​that deviate from the median of the same group by more than three times the median absolute deviation for each strain combination, calculates the thickness ratio of the retained samples, and establishes a mapping rule between the longitudinal and transverse strains and the thickness of the base fabric at the longitudinal and transverse strain nodes, forming a bilinear interpolation surface. The calibration model records the base fabric type, longitudinal strain nodes, transverse strain nodes, thickness ratio of each node, number of samples at each node, standard deviation of repeated measurements, applicable temperature and humidity range, version identifier, and effective roll batch. An execution record shows a thickness ratio of 0.973, a sample size of 9, and a standard deviation of 2.1 μm when the longitudinal strain is 2% and the transverse strain is 1%. When the production strain falls within the node range, bilinear interpolation is performed. When it falls outside the range but is no more than 0.5 percentage points away from the boundary, boundary slope extrapolation is used and the subsequent uncertainty is increased. When it exceeds this range, the predicted state is written as invalid. For the post-coating mesh identifier is And the corresponding pre-coating base fabric unit is identified as Based on the material relationship, the thickness conversion module first reads the effective bit of the pre-coating source unit's jitter compensation using the pre-coating base fabric unit identifier and original echo quality reference saved in S2; only when this effective bit is one, according to... Calculate the predicted base fabric thickness under tension after coating. If the effective bit is zero or missing, write the predicted thickness as null, write the predicted state as invalid, and record the invalid reason code for the pre-coating source unit periodic jump compensation. This indicates the predicted base fabric thickness of the coated mesh, in μm. and These represent the grid number of the cumulative strain-free travel band after coating and the grid number of the lateral material location, respectively. This indicates the compensated thickness of the base fabric unit before coating, in μm. and These represent the grid number of the cumulative strain-free travel strip length before coating and the grid number of the lateral material location, respectively. Indicates the base fabric type The thickness ratio under the combined longitudinal strain and transverse width strain after coating is a dimensionless quantity. This represents the thickness ratio of the same base fabric type under the combination of longitudinal strain and transverse width strain before coating; it is a dimensionless quantity. and These represent the longitudinal strain and the transverse width strain after coating, respectively, and are dimensionless quantities. and These represent the longitudinal strain and the transverse width strain before coating, respectively, and are dimensionless quantities. Indicates the base fabric type identifier; When the same material relationship includes four pre-coating base fabric elements, the thickness conversion module reads the jump compensation valid bits one by one through the original echo quality references of each of the four elements. Only when all four valid bits are one, the weighted sum of the four compensated base fabric thicknesses is calculated according to the bilinear interpolation weights saved in S2, and the same weights are used to calculate the weighted values ​​of the four pre-coating longitudinal strains and transverse width strains, and then substituted into the thickness conversion. When the jump compensation valid bit of any of the four elements is zero or missing, or the echo quality state of any element is missing, the remaining weights are not re-normalized to avoid changing the corresponding material area. Instead, the predicted thickness of the post-coating mesh is written as null, the predicted state is set to invalid, and the corresponding pre-coating source element identifier and reason code are recorded. When the only same material element is valid but its thickness differs from the median of the five adjacent valid elements by more than 40 μm, it is marked as a texture mutation point and the original value is retained. Subsequently, the uncertainty field of S4 reflects the sparse influence of the neighborhood anchor points, and unsupported smoothing is not used to replace the real base fabric texture. The thickness conversion module synchronously calculates the sources of uncertainty in the calibration model interpolation, including the standard deviation of calibration node repeatability, the normalized distance from the current strain to the nearest node, and the repeatability of the pre-coating thickness measurement. The sum of the squares of these three items is used to form the additional uncertainty in the prediction, in μm. This additional uncertainty is only used as the baseline input for solving the uncertainty field in S4 and does not change the predicted base fabric thickness itself. The extrapolation grid adds an extrapolation amount of 5 μm to the baseline. Grids with inconsistent calibration model versions, denominator thickness ratios less than 0.90 or greater than 1.10, and mismatched material conditions are all output with an invalid prediction reason code. To verify the traceability of the conversion chain, the predicted base fabric thickness record saves the post-coating mesh identifier, the pre-coating unit identifier of the same material, the original pre-coating thickness, the longitudinal and transverse strains before and after coating, the two thickness ratios, the predicted base fabric thickness, the additional uncertainty of the prediction, the calibration model version, the interpolation or extrapolation flag, and the prediction status. In this specific embodiment, the thickness ratio within the normal production strain range is limited to 0.94 to 1.03. If the calculation result exceeds this range, the thickness value is not truncated, but the prediction status is changed to verification in order to distinguish between the actual anomaly of the base fabric and the calibration model going out of bounds. After completing the current unified material coordinate data block, the module forms a predicted base fabric thickness matrix and a predicted additional uncertainty matrix according to the post-coating mesh order, and writes the correspondence between the two and the original same material into the roll material detection cache. The matrix cell is not deleted due to invalidity, and the numerical field, status field and reason code are saved respectively. Then, the module publishes the predicted base fabric thickness data block identifier, material coordinate range, calibration model version and mesh set that can be used for comparison with the uncoated reference area to S4, so that the measured thickness of the reference anchor point and the predicted thickness are directly subtracted on the same material cell.

[0022] In this specific embodiment, S4 includes: The online correction module of S4 reads the predicted base fabric thickness matrix and the predicted additional uncertainty matrix of S3 and the total thickness matrix of the full width after coating of S1, and reads the definition of the uncoated reference area from the coating process formula. In this specific embodiment, a continuous uncoated edge strip with a width of 20 mm is reserved on the left coating edge, and an uncoated reference window with a longitudinal length of 100 mm and a transverse width of 30 mm is set every 50 m roll length. The formula record saves the reference area type, transverse material position range, longitudinal interval, window length, formula version and effective time. The online correction module selects candidate grids that fall into the reference area based on the unified material coordinates rather than the sampling time. The online calibration module matches the uncoated reference area mesh with the S2 material correspondence according to the material coordinate difference from low to high, and maps the measured thickness of the base fabric in the uncoated reference area after coating to the material element; when one measured mesh corresponds to four pre-coating elements, the predicted values ​​of the four elements are matched according to the bilinear weights saved in S2; when the material coordinate difference exceeds the S2 tolerance, the mesh is deleted from the anchor point candidate set and the material mapping failure reason code is written to ensure that the measured thickness, predicted thickness and boundary residual use the same material area; For each candidate grid, the module reads the normalized echo intensity of the upper and lower probes, the light stripe width, the saturation pixel ratio, the independent reconstruction thickness residual of the two subframes, the thickness variation between adjacent frames, and the effective bits of post-coating jitter compensation from the post-coating echo quality references stored in S1. It then reads all the effective bits of pre-coating source unit jitter compensation that form the predicted base fabric thickness for that candidate grid, one by one, through one or four pre-coating base fabric unit identifiers and their original echo quality references in the same material correspondence relationship in S2. The module reads the effective ranges of each of the five quality quantities from the detection parameter file associated with the base fabric type. Effective upper limits; These thresholds are established using calibration samples of the same type of base fabric without coating and whose thickness has been confirmed as qualified by a contact thickness gauge. The calibration samples are no less than 10 rolls and cover the left, middle and right width positions. The empirical distribution of the five quality quantities is statistically analyzed. The center value interval covering 99.5% of the samples is taken as the effective interval of echo intensity and light stripe width. The 99.5% quantile value is taken as the effective upper limit of saturation pixel ratio, independent reconstruction thickness residual of dual subframes and change in adjacent frames. The preset sample ratio, sample batch, statistical date and threshold version are all written into the detection parameter file. The module first checks the valid bits of post-coating jitter compensation and all valid bits of pre-coating source unit jitter compensation. Then, it sequentially filters the anchor point candidate set according to five conditions: echo intensity, light stripe width, saturation pixel ratio, thickness residual, and thickness change between adjacent frames. For each check, it saves the input value, corresponding station identifier, pre-coating source unit identifier, threshold version, and pass flag. When the valid bits of post-coating jitter compensation are zero or missing, or the valid bits of jitter compensation of one pre-coating source unit are zero or missing, or the valid bits of jitter compensation of any one of the four pre-coating source units are zero or missing, the candidate grid is immediately deleted from the reference anchor point set, and the post-coating periodic jitter compensation invalid reason code or the pre-coating source unit is written respectively. If the periodic fluctuation compensation is invalid, the anchor point qualification cannot be restored even if all five subsequent quality conditions are passed. When the fluctuation compensation valid bit of all source elements after coating and before coating is one, but any quality item does not meet the corresponding valid range or valid upper limit, the candidate mesh will be deleted from the reference anchor point set but the quality record will be retained. Only when the fluctuation compensation valid bit of all source elements after coating and before coating is one, all five quality conditions are passed, the same material mapping status is valid, and the S3 prediction status is valid, will the candidate mesh be updated as a reference anchor point and its material coordinates, measured base fabric thickness, and predicted base fabric thickness be published. Mesh with an S3 prediction status of verification or invalid shall not be included in the reference anchor point set. In the detection parameter recording of this specific embodiment, the effective range of normalized echo intensity is 0.18 to 0.92, the effective range of light stripe width is 2.0 to 8.0 pixels, the upper limit of saturated pixel ratio is 0.08, the upper limit of thickness residual for independent reconstruction of dual subframes is 12 μm, and the upper limit of thickness variation between adjacent frames is 35 μm. μm; the detection parameter file also saves continuous normalized scoring rules and scoring rule versions from 0 to 1 for the five quality quantities. A score of 1 indicates that it is at the center of the calibration sample or the zero defect end, and a score of 0 indicates that it has reached the invalid boundary. The value is linearly taken according to the normalized margin from the center or the upper limit of the effective range within the interval; during operation, the jump compensation effective bit hard threshold is first applied to the post-coating mesh and all pre-coating source units that form its predicted base fabric thickness, and then the five quality conditions are compared item by item. If any original quality field is missing, the corresponding item is invalid; the candidate mesh with the jump compensation effective bit of the post-coating and all pre-coating source units being one, all five quality quantities being effective, the same material mapping being completed, and the S3 prediction state being effective is determined as the reference anchor point. Any related jump compensation effective bit being zero or missing, the S3 prediction state being verification or invalid, and other unqualified candidate meshes are retained in the reference area record but are not included in the boundary residual fitting so that they can be recalculated after the detection parameters are updated; when the continuous uncoated side strip is in the current 200 If there is no reference anchor point within the mm vertical window, then search the most recent uncoated reference window. If neither of them has an anchor point, mark the current propagation partition as having no available reference anchor point. Online calibration module according to Form a reference anchor point quality coefficient, and according to Boundary residuals are formed; Indicates reference anchor point The continuous mass coefficient is a dimensionless quantity ranging from 0 to 1. Indicates reference anchor point No. The quality score obtained according to the test parameter file and scoring rule version is a normalized quality score, which is a dimensionless quantity ranging from 0 to 1. Indicates the sequence number of the five quality items. Indicates the reference anchor index. Indicates reference anchor point Boundary residuals at the location, in μm. This indicates the thickness of the base fabric measured by the post-coating station in the uncoated reference area, in μm. The predicted base fabric thickness of the same material unit under post-coating tension is expressed in μm; the five effective thresholds are still used as hard screening conditions for the reference anchor points. Different anchor points after hard screening obtain different results based on the continuous margin within the effective range. ,Should It can be directly used to calculate the fitting weights and the average quality of the nearest anchor point; The module reads the transverse tension distribution saved in S1 and the local transverse strain field with transverse grid column index published in S2 along the transverse material position. It calculates the tension difference only for adjacent 2 mm grids where the tension state on both sides is valid, and calculates the local transverse strain difference only for adjacent 2 mm grids where the local strain state on both sides is valid. Locations where the tension difference exceeds 120 N / m or the local transverse strain difference exceeds 0.20 percentage points are identified as propagation boundary locations. If a local strain side is missing, the strain boundary branch for that adjacent pair is not executed, but the tension boundary branch is retained. If both tension and local strain are missing, the adjacent pair is directly marked as the input missing propagation boundary and smoothing terms are prohibited from crossing it. The module connects adjacent boundary locations within the current 200 mm longitudinal solution window to form a propagation partition. The partition record saves the partition identifier, the lateral start and end material positions, the longitudinal solution window, the grid columns on both sides, the boundary source, the tension difference between the two sides, the local lateral width strain difference on both sides, and the input status. When an isolated boundary lasts only one pulse, it is replaced by the midpoint of three adjacent pulses to avoid pseudo-partitions with a width of only one grid caused by measurement noise. The partition boundary formed by tension difference or local strain difference does not block the correction amount, but reduces the cross-partition smoothing weight and applies a continuity upper limit during the solution. Within each propagation partition, the module updates the set of adjacent grid pairs within the partition according to the four-adjacency relationship, and filters the set of adjacent grid pairs across partitions according to the boundary crossing flag. When there are no reference anchors in a partition, the partition is removed from the independent solution list and enters the no-anchor state. When the number of anchors in a partition is one or two and the total number of reference anchors in the current solution window is not less than three, the solution for that partition is retained and the robust standard deviation of the boundary residuals of the reference anchors in the entire window is adopted. When the number of anchors in a partition is not less than three, the robust standard deviation of the boundary residuals of the reference anchors in this partition is adopted. When the number of anchors in a partition is one or two and the total number of reference anchors in the entire window is still less than three, the solution is not continued, and the partition is written into the sparse anchor retest status and cause code, so that the partition, the adjacency set and the source of uncertainty are consistent. For each grid within the current solver window Set correction amount ,by To obtain the base fabric thickness correction field for the target; This represents the weighted squared error objective to be minimized, in μm. 2 , This represents the set of reference anchor points within the current window. Indicates reference anchor point The fitting weights, determined by the quality coefficients, are dimensionless and take... , Indicates anchor point The correction amount for the grid in question, in μm. Indicates anchor point The boundary residuals, in μm. This represents the smoothing weight within the same propagation partition, which is a dimensionless quantity and is set to 0.35 in this specific implementation. This represents a set of horizontally or vertically adjacent grid pairs within the same propagation partition. and Indicates the adjacent grid index, and These represent the correction amounts for the corresponding grid cells, in μm. This represents the continuous weight across propagation partitions, is a dimensionless quantity, and is set to 0.08 in this specific implementation. Represents the set of adjacent grid pairs that cross partition boundaries; The solver employs a constrained quadratic optimization process that combines an outer loop of active sets with an inner loop of preconditional conjugate gradients, and adds cross-partition adjacent meshes. Continuity constraints; This represents the absolute value of the difference in correction between adjacent grid cells across different partitions, in μm. This indicates the upper limit of cross-partition contiguous space given in the detection parameter file, in μm, and is taken as 25 μm in this specific implementation. and This indicates the adjacent grid indices located on either side of the partition boundary. and These represent the correction amounts, in μm. Specifically, candidate correction fields are first formed using quadratic objectives without activated hard constraints. Then, each cross-partition adjacent grid pair is checked: when the difference in correction amount exceeds the positive upper limit or negative lower limit by more than 0.01 μm, the grid pair is added to the active constraint table according to the direction of the exceedance, and the corresponding difference is fixed at the positive upper limit or negative lower limit. Subsequently, the active constraints are used as equation conditions to form the current quadratic subproblem. A symmetric positive definite linear equation for the correction amount is obtained through Lagrange multiplier elimination. This linear equation is solved using the diagonal preconditioned conjugate gradient method, and the Lagrange multipliers of each active constraint are obtained by back substitution. If the direction of the Lagrange multiplier of an active constraint indicates that continuing to fix the constraint will increase the objective, the constraint is released from the active constraint table and the solution is repeated. In each outer loop, all cross-partition adjacent grid pairs are rechecked until the maximum constraint exceedance does not exceed 0.01 μm. If both μm and active constraints satisfy the multiplier direction condition and the relative change of the target value between two adjacent rounds is less than 0.0001, then the constraint solution is considered to have converged. The upper limit of the inner iteration of the preconditional conjugate gradient is 200 times, and the upper limit of the outer loop of the active set is 30 times. If any layer reaches the upper limit but still does not meet the above conditions, the correction field of the previous effective window is retained only for display and the current window is marked as a retest. This retained value is not allowed to enter the overly thin anomaly judgment. For each grid according to Calculate the uncertainty; Represents a grid The uncertainty, in μm, This represents the additional uncertainty in the prediction given by S3, in μm. This represents the anchor point distance contribution coefficient, in μm / m, and in this specific implementation, it is taken as 2.0 μm / m. Represents a grid The Euclidean distance in material coordinates to the nearest reference anchor point within the same propagation zone, in meters. This represents the contribution coefficient to the dispersion of the boundary residuals; it is a dimensionless quantity and is set to 1.2. Represents a grid Robust standard deviation of the boundary residuals of the reference anchor point within the propagation zone, in μm. This represents the echo quality gap contribution factor, in μm and taken as 8 μm. Represents a grid The number of available reference anchors within the propagation partition is a non-negative integer. Indicates 8 and The minimum of the two, Indicates the distance to the grid within the same propagation partition. Recent The arithmetic mean of the quality coefficients of available reference anchor points, with the actual number of participating anchor points as the denominator, is a dimensionless quantity ranging from 0 to 1; when there are fewer than three anchor points within a partition but no fewer than three reference anchor points across the entire window, Take the robust standard deviation of the boundary residuals of the full window reference anchor points; if there are still fewer than three full window reference anchor points, do not calculate the finite uncertainty and write it into the sparse anchor point retest state; if there are no anchor points at all, write it into the no anchor point state. The correction field data block stores material coordinates, correction amount, propagation zone, nearest anchor distance, number of anchors, solution convergence status, and parameter version grid by grid. The uncertainty field stores the prediction additional uncertainty, residual discrete contribution, distance contribution, mass contribution, and final uncertainty. The online correction module sets a 50 mm overlap area for adjacent solution windows. Within the overlap area, the correction amount and uncertainty are fused according to the linear weight from the center of the window to ensure longitudinal continuity. Then, the base fabric thickness correction field, uncertainty field, reference anchor index, no-anchor zone identifier, and solution status are written into the roll material detection cache and published to S5.

[0023] In this specific embodiment, S5 includes: The coating evaluation module of S5 reads the post-coating full-width total thickness matrix and its echo quality reference from S1, the predicted base fabric thickness matrix from S3, the base fabric thickness correction field, uncertainty field, and uncoated reference area mask from S4, connects them one by one according to the grid identifier in the unified material coordinate system, and reads the effective bits of the jump compensation for each post-coating grid from the post-coating echo quality reference; grids falling into continuous uncoated edges or uncoated reference windows are written to the reference area state separately; for non-reference area grids... Only when the effective bit of the post-application jitter compensation is one, the S3 prediction state is valid, and there is a usable anchor point in the correction field, should the following be followed first. The corrected predicted base fabric thickness is obtained, and then according to Obtain the net coating thickness; Represents a grid The corrected predicted base fabric thickness, in μm. This indicates the predicted base fabric thickness output by S3, in μm. This represents the base fabric thickness correction amount output by S4, in μm. Represents a grid Net coating thickness, in μm. This indicates the total thickness of the coating after application, in μm. This indicates a unified material coordinate grid index; when any input thickness is missing, the effective bit of post-coating jump compensation is zero or missing, S3 writes the prediction state as invalid due to invalid jump compensation of any pre-coating source unit, other prediction states are invalid, or the correction field has no anchor point, the net coating thickness retains a null value instead of being replaced by a zero value. The coating evaluation module writes the net coating thickness, cumulative strain-free walk length, lateral material position, laser echo quality reference, uncertainty, reference area mask, and input state into the net coating thickness map; for the same cumulative strain-free walk length row... ,according to Calculate the lateral uniformity index; Indicates the first The horizontal uniformity index of the row, in percentage. This indicates the row number representing the cumulative strain-free conveyor belt length. This represents the set of horizontal grid cells within the row that are in a valid state and do not belong to the reference area. Indicates the horizontal grid number. This indicates the net coating thickness for the corresponding grid, in μm. This represents the average net coating thickness within the set, in μm. Indicates the number of valid non-reference area grids. This represents the lower limit of the same unit to prevent the average thickness from approaching zero, with the unit being μm and taken as 10 μm. This expression represents the non-negative square root operation; the input unit is μm. 2 The output unit is μm. This indicates that the maximum value of two inputs with the same unit is taken, and the output unit of this formula is μm; the larger the index, the greater the lateral dispersion. When the effective non-reference area grid coverage width is less than 80%, no value is output and the state of insufficient coverage is output instead. transverse material position column Within each preset volume length range, according to Calculate the longitudinal uniformity index; Indicates the first The vertical uniformity index listed within the current volume length interval, in percentage. Indicates the column number of the horizontal material position. This indicates that the column is a set of vertical grids that is valid within the current volume length range and does not belong to the reference area. This indicates the row number representing the cumulative strain-free conveyor belt length. This indicates the net coating thickness for the corresponding grid, in μm. This represents the average net coating thickness within the set, in μm. Indicates the number of vertical grid cells in the effective non-reference area. This indicates the lower limit of the same unit for 10 μm. and The complete explanations of the non-negative square root operation and the maximum value operation of the same unit in the previous horizontal uniformity formula segment are used respectively; the preset volume length interval of this specific implementation is a continuous 10 m and the adjacent intervals do not overlap. The larger the index, the greater the vertical dispersion. When the effective non-reference area length coverage is less than 90%, the output coverage is insufficient. The coating process formulation uses product specifications and coating surface as the joint key, storing the lower and upper limits of net coating thickness, allowable values ​​for transverse and longitudinal uniformity, minimum physical length, minimum physical area, upper limit of uncertainty, and formulation version. In this specific embodiment, the allowable range of net coating thickness is 280 to 340 μm, the allowable values ​​for transverse and longitudinal uniformity are both 6%, the minimum physical length is 50 mm, and the minimum physical area is 200 mm². 2 The upper limit of uncertainty is 20 μm. The formula record is locked at the beginning of the roll material inspection task. If the formula is changed in the middle, it will only take effect from the new equal material length pulse mark and generate a new inspection record segment to avoid using different thresholds for the same abnormal area. Measurement status is determined before anomaly type, and after identification in the reference area, mutual exclusion judgment is performed in the order of invalid, retest, and valid: when the mesh falls into the uncoated edge or uncoated reference window defined by S4, the reference area status is written first; for non-reference area meshes, all invalid conditions are checked first. When the post-coating jitter compensation valid bit is zero or missing, S3 writes the prediction status as invalid because any jitter compensation valid bit in one or four pre-coating source elements is zero or missing, S3 outputs an invalid prediction for other reasons, S4 propagation partition is in an anchorless state, or any echo quality item does not meet the preset valid condition of S4, the invalid status is written. The invalid status takes precedence over any retest condition and subsequent judgments cannot overwrite it; retest conditions are checked only when invalid conditions are not triggered. When the S3 prediction status is verification, the current solution window of S4 has not converged, the S4 propagation partition is a sparse anchor retest status, or the uncertainty exceeds 20, the invalid status is written as invalid. Write the retest status when the value is μm, and the retest status must not be rewritten as valid subsequently; write the valid status only when the invalid condition and the retest condition are not triggered, the post-coating jump compensation valid bit is one, the S3 prediction status is valid, the S4 solution converges, there is a usable reference anchor point in the propagation partition, all echo quality items meet the valid conditions and the uncertainty does not exceed 20 μm; only valid and non-reference area state grids participate in the horizontal or vertical uniformity statistics, net coating thickness upper and lower limit comparison and connected component merging, and reference area, retest or invalid state grids do not participate and are neither judged as too thin or too thick, it is especially prohibited to write the near zero net coating thickness, missing net coating thickness or low confidence negative difference value of the reference area as too thin anomaly; the reference area data is only used for S4 anchor point, correction and quality traceability. For valid non-reference area state meshes, if the net coating thickness is below the lower limit of the formulation, it is marked as an overly thin candidate; if it is above the upper limit of the formulation, it is marked as an overly thick candidate; otherwise, it is marked as normal. The evaluation module performs an eight-neighbor connected domain search on the unified material coordinate mesh for both overly thin and overly thick candidates. Mesh that are edge-adjacent or corner-adjacent and have the same over-limit direction are grouped into the same candidate region, without establishing connections across the reference region, retest, or invalid meshes. When a real anomaly candidate is adjacent to the reference region, the candidate region is truncated by the reference region mask boundary, and the two are not merged. The candidate region record is stored separately with adjacent reference region flags. The longitudinal physical length of each candidate region is obtained by adding a longitudinal mesh size to the difference between the maximum and minimum cumulative strain-free walk length. The region area is obtained by summing the areas represented by the member meshes in the strain-free material coordinate system. The longitudinal physical length reaches 50 mm or the region area reaches 200 mm. 2 If the area is identified as an abnormal coating region, it is otherwise retained as a transient candidate and no alarm is output. The detection parameter file also saves normalized quality scoring rules and their versions ranging from 0 to 1 for echo intensity, light stripe width, saturation pixel ratio, independent reconstruction thickness residual of two subframes, and thickness variation between adjacent frames. A score of 1 indicates that it is at the center of the calibration sample or at the zero-defect end, while a score of 0 indicates that it has reached or exceeded the invalid boundary. Within the interval, the score is linearly taken according to the normalized margin from the center or the upper limit. For each valid and non-reference area grid ,according to Forming the grid laser echo quality coefficient, Represents a grid The laser echo quality coefficient is a dimensionless quantity ranging from 0 to 1. Represents a grid No. The echo quality item is a normalized quality score obtained from the test parameter file, which is a dimensionless quantity ranging from 0 to 1. This indicates the serial number of one of the five echo quality items. This represents a unified material coordinate grid index; the coefficient and scoring rule version are written into the net coating thickness map; for each coating anomaly area... ,according to Calculate the measurement confidence value; Indicates abnormal area The measurement confidence value is a dimensionless quantity ranging from 0 to 1. Indicates areas of coating abnormality. Represents the valid and non-reference area state member grid within the region. The arithmetic mean is a dimensionless quantity ranging from 0 to 1. This represents the average uncertainty of the valid, non-reference state grid within the region, in μm. This represents the upper limit of uncertainty in the coating process formulation, expressed in μm and taken as 20 μm. Represented by natural constant An exponential function with base 0.5, this expression is input. Both the output and the output are dimensionless quantities. This represents a cutoff function that restricts the result to the range of 0 to 1; this relationship ensures that the confidence value does not decrease when the echo quality improves, and does not increase when the uncertainty increases; the confidence value is not calculated when there is no valid and non-reference area state grid in the region. The aforementioned grid laser echo quality coefficient is calculated for each non-reference area grid before the hard threshold of the measurement state: The evaluation module generates effective indication values ​​for each of the five echo quality items. An item is assigned a value of one if it meets the corresponding effective interval or upper limit of S4, otherwise it is assigned zero. The integer sum of the five effective indication values ​​is recorded as the number of effective items for that grid. Items with an effective indication value of zero have their continuous normalized quality score forcibly set to zero. Items with an effective indication value of one use the 0 to 1 non-negative score given in the detection parameter file, and are then calculated using the aforementioned grid laser echo quality coefficient formula with the fixed five items as the denominator. Therefore, when the scores of other items remain unchanged, any item changing from invalid to valid only replaces the zero score with a non-negative score. The grid laser echo quality coefficient does not decrease when the number of effective items increases. The regional quality average is formed by the grid laser echo quality coefficient, and when the member set and uncertainty remain unchanged, increasing the number of effective items in the member grid does not decrease the measurement confidence value. After completing the recording of the quality coefficient and the number of valid items, the five hard thresholds of S5 are executed. Any grid with a valid indicator of zero is still written as invalid. Its quality coefficient and the number of valid items are only used for quality traceability and do not enter the area quality average or measurement confidence value of the coating anomaly area. Only grids with all five items being valid can enter the anomaly area confidence value calculation according to other state conditions. The abnormal area material coordinate record saves the starting and ending positions of the width, the starting and ending positions of the cumulative strain-free roll length, the area, the longitudinal physical length, the minimum and maximum net coating thickness, the abnormal direction, the measurement confidence value, the number of effective grids, the number of retest grids, and the formula version. The starting and ending positions of the width are determined by the transverse material position boundary of the member grid, and the starting and ending positions of the roll length are determined by the cumulative strain-free strip length boundary of the member grid. The extreme value of the net coating thickness is only obtained from the effective state members. When the abnormal area is in contact with the retest area, an adjacent retest mark is added to the record for the roll material review personnel to determine the expansion of the retest range. The coating evaluation module associates the material coordinate information of abnormal areas, the lateral and longitudinal uniformity indices of each 10 m interval, the grid-level valid, retested or invalid states, and the corresponding uncertainties with the original material length pulse identifiers to form a test record. The test record uses the roll material identifier, formula version and pulse identifier as the joint primary key, and saves the unified material coordinate version, the pre-coating and post-coating matrix block identifiers, the same material correspondence version, the base fabric calibration model version, the correction field solution status and generation time. When the production speed changes, the material spatial interval represented by the 1 mm material length pulse does not change. Therefore, adjacent test records are still sorted by pulse identifier and cumulative strain-free travel length, not by sampling time interval. For confirmed coating anomaly areas, the system generates a roll material verification positioning interval by extending 20 mm forward and backward based on the starting and ending positions of the cumulative strain-free roll length, respectively. This interval, along with the start and end positions of the roll width, the direction of the anomaly, the extreme value of the net coating thickness, and the measurement confidence value, is written into the verification queue. The rewinding equipment reads the roll material identification and the cumulative strain-free roll length position to drive the encoder for positioning. Before entering the interval, it decelerates and is verified by the offline thickness measurement station. The verification results are written back to the detection record according to the anomaly area identification. The final output includes the net coating thickness map, the transverse uniformity index, the longitudinal uniformity index, the material coordinate record of the anomaly area, the grid measurement status, and the verification positioning interval. The system compares online detection records with verification records according to the roll material identification, unified material coordinate version, and equal material length pulse identification. When the verification positioning error exceeds 5 mm, the encoder zero-point offset of the subsequent interval is updated and the original detection record is retained. When the extreme difference between the offline verification thickness and the net coating thickness exceeds 25 μm, the status of the corresponding abnormal area is updated to pending analysis without rewriting the locked base fabric correction field. The comparison results save the verification equipment identification, gauge calibration version, positioning error, thickness difference, and handling status. The coating evaluation module calculates the percentage of valid, retested, and invalid grids for each roll and each formulation version and updates the roll material inspection task summary. When the percentage of invalid grids in a continuous 1 m of material exceeds 20%, a measurement chain maintenance prompt is output. When the percentage of retested grids exceeds 10% but the percentage of invalid grids does not exceed 20%, the retest interval is output instead of the conclusion that the coating is too thin. The laser detection method for tarpaulin coating thickness uniformity thus forms a method where the detection results maintain the mutual traceability of grid status, uniformity index, abnormal areas, and verification positioning intervals, thereby completing a continuous detection chain from pre-coating same material unit, strain calculation, online anchor point correction to roll material spatial positioning.

[0024] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0025] This invention connects equal material length sampling, dual-station same material mapping, base fabric strain thickness conversion, and net coating thickness evaluation into a continuous data processing chain, so that the base fabric change in the total coating thickness can be deducted according to the corresponding pre-coating material unit and post-coating tension state, and abnormal areas can be recorded using material coordinates rather than sampling time.

[0026] This invention introduces the measured values ​​of the online base fabric in the uncoated reference area, and solves the correction field and uncertainty field after screening the echo quality of the reference anchor point. It also processes the low confidence area according to the rule that no anchor point is directly invalid and the uncertainty is diverted according to the rule that there is an anchor point. In this way, it corrects the cross-station model drift and reduces the error of underthin judgment under the current roll material and current tension state.

Claims

1. A laser detection method for the uniformity of coating thickness on tarpaulin, characterized in that, include: S1. At the upstream and downstream of the tarpaulin coating station, laser profile measurement stations with opposite directions are used to collect the thickness of the base fabric before coating, the total thickness of the tarpaulin after coating, tension, width edge and laser echo quality according to the material length pulse of the encoder, and after compensation, the full-width thickness matrix before coating and the full-width total thickness matrix after coating are formed. S2. Based on the equal material length pulse, tension and width edge, map the two thickness matrices to a unified material coordinate system represented by the cumulative strain-free travel length and the lateral material position to obtain the same material correspondence between the post-coating mesh and the pre-coating base fabric unit. S3. Based on the material correspondence and the relationship between longitudinal and transverse strain and thickness calibration of the base fabric, the thickness of the base fabric unit before coating is converted into the predicted base fabric thickness under tension after coating. S4. Map the measured thickness of the base fabric in the uncoated reference area after coating to the same material element, select the grids whose laser echo quality meets the preset effective conditions as reference anchor points, take the difference between the measured and predicted base fabric thickness at the anchor point as the boundary residual, and use the transverse tension or transverse width strain partition as the propagation constraint to solve the base fabric thickness correction field and uncertainty field. S5. Correct the predicted base fabric thickness using the base fabric thickness correction field, subtract the corrected predicted base fabric thickness from the total thickness matrix after coating to obtain the net coating thickness map, calculate the transverse uniformity index and longitudinal uniformity index, merge the coating abnormal areas, and output the measurement status and material coordinate information based on the uncertainty field.

2. The laser detection method for the uniformity of tarpaulin coating thickness according to claim 1, characterized in that, S1 includes: At each laser profile measurement station, the upper laser profile probe and the lower laser profile probe measure the upper and lower surface profiles of the tarpaulin respectively, and the thickness is obtained based on the upper and lower surface profiles at the same lateral position. In the absence of strip material, the reference distance and optical axis tilt angle between the upper and lower laser profile probes are calibrated. Temperature compensation is performed using the calibration relationship between the temperature of the measuring station and the drift of the reference distance. Roll body runout compensation is performed using the periodic displacement sequence corresponding to the angular position of the roll body. The compensated thickness, tension distribution, width edge, belt walking posture and laser echo quality are associated and saved with the pulse identifier of the same material length.

3. The laser detection method for the uniformity of tarpaulin coating thickness according to claim 1, characterized in that, S2 includes: reading the tension and longitudinal strain calibration relationship and the strain-free reference width from the detection parameter file associated with the base fabric type; determining the pre-coating longitudinal strain and post-coating longitudinal strain based on the tension collected by the pre-coating laser profilometry station and the post-coating laser profilometry station, respectively; determining the pre-coating transverse width strain and post-coating transverse width strain based on the change of the measured width defined by the two sides of the width edge detected by each station relative to the strain-free reference width; calculating the strain-free length increment segment by segment based on the tape length increment corresponding to adjacent equal material length pulses and the longitudinal strain of the corresponding measurement station, and accumulating them to obtain the cumulative strain-free tape length; converting the transverse sampling position into a transverse material position according to its position ratio relative to the two sides of the width edge of the corresponding measurement station; and establishing the same material correspondence between the pre-coating base fabric unit and the post-coating mesh with the same cumulative strain-free tape length and the same transverse material position.

4. The laser detection method for the uniformity of tarpaulin coating thickness according to claim 1, characterized in that, S3 includes: applying multiple longitudinal and transverse strain states to a calibration sample of the same type as the base fabric to be tested, recording the change in thickness relative to the unstrained thickness under each strain state, and forming the relationship between longitudinal and transverse strain and thickness of the base fabric; determining the thickness conversion amount from the relationship between longitudinal and transverse strain and thickness of the base fabric based on the pre-coating longitudinal strain and pre-coating transverse width strain corresponding to the pre-coating base fabric unit determined in step S2, and the post-coating longitudinal strain and post-coating transverse width strain corresponding to the same material post-coating grid, and converting the compensated thickness of the pre-coating base fabric unit into the predicted base fabric thickness according to the thickness conversion amount.

5. The laser detection method for the uniformity of tarpaulin coating thickness according to claim 1, characterized in that, S4 includes: obtaining the uncoated reference area from a continuous uncoated edge strip set at least one side of the coating edge or an uncoated reference window spaced apart along the tape-running direction; for each candidate grid in the uncoated reference area, determining whether the laser echo quality meets the preset effective conditions based on the comparison results of echo intensity, light stripe width, saturated pixel ratio, thickness residual of upper and lower laser profile probes, and thickness variation of adjacent frames with the corresponding effective interval or effective upper limit; determining the candidate grid that meets the preset effective conditions and completes the same material mapping as the reference anchor point, and inputting the boundary residual and its material coordinates at the reference anchor point into the solution process of the base fabric thickness correction field and uncertainty field.

6. The laser detection method for the uniformity of tarpaulin coating thickness according to claim 1, characterized in that, S5 includes: determining the lateral uniformity index based on the dispersion of net coating thickness of each lateral grid at the same cumulative strain-free travel length; determining the longitudinal uniformity index based on the dispersion of net coating thickness of each longitudinal grid within a preset roll length interval at the same lateral material position; reading the upper limit of coating thickness, lower limit of coating thickness, minimum physical length, and minimum physical area from the coating process formula; merging connected components of grids that are edge-adjacent or corner-adjacent in a unified material coordinate system and have the same direction of exceeding the limit; and determining the connected components whose longitudinal physical length reaches the minimum physical length or whose area reaches the minimum physical area as the coating. Abnormal regions: When the laser echo quality of the mesh does not meet the preset valid conditions, an invalid state is output; when the laser echo quality of the mesh meets the preset valid conditions and there is no available reference anchor point, an invalid state is output; when the laser echo quality of the mesh meets the preset valid conditions, there is a available reference anchor point, and the uncertainty does not exceed the preset uncertainty upper limit, an valid state is output; when the laser echo quality of the mesh meets the preset valid conditions, there is a available reference anchor point, and the uncertainty exceeds the preset uncertainty upper limit, a retest state is output, and meshes in the retest state or invalid state are prohibited from being judged as too thin anomalies.

7. The laser detection method for the uniformity of tarpaulin coating thickness according to claim 5, characterized in that, Each effective range and effective upper limit is determined by a base fabric calibration sample with no coating and qualified thickness; the distribution of echo intensity, light stripe width, saturated pixel ratio, thickness residual of upper and lower laser profile probes and thickness variation of adjacent frames of the base fabric calibration sample are statistically analyzed respectively, and the range of values ​​covering the preset sample ratio is determined as the corresponding effective range, and the upper limit of the residual or the upper limit of the variation covering the preset sample ratio is determined as the corresponding effective upper limit; The preset sample ratio is recorded in a detection parameter file associated with the base fabric type.

8. The laser detection method for the uniformity of tarpaulin coating thickness according to claim 5, characterized in that, The solution process for the base fabric thickness correction field and uncertainty field includes: dividing the propagation zone according to the boundary position of the transverse tension change or the boundary position of the transverse width strain change; within each propagation zone, minimizing the weighted sum of the boundary residual fitting error of the reference anchor point, the difference in correction amount between adjacent grids, and the difference in correction amount across zones is the objective; the weight of the boundary residual fitting error is determined by the laser echo quality of the reference anchor point, and the correction amount of each grid is determined under the constraint of maintaining the continuity of the correction amount across zones; the uncertainty of each grid is determined according to the material coordinate distance from each grid to the reference anchor point, the degree of dispersion of the boundary residual within the propagation zone, and the laser echo quality of the reference anchor point.

9. The laser detection method for the uniformity of tarpaulin coating thickness according to claim 6, characterized in that, The material coordinate information of the coating anomaly region includes the starting position of the width, the ending position of the width, the starting position of the cumulative unstrained roll length, the ending position of the cumulative unstrained roll length, the area of ​​the region, the extreme value of the net coating thickness, and the measurement confidence value. The measurement confidence value is determined based on the laser echo quality and uncertainty of the effective state grid in the region, and the measurement confidence value does not decrease when the number of effective items corresponding to the laser echo quality increases, and does not increase when the uncertainty increases.

10. The laser detection method for the uniformity of tarpaulin coating thickness according to claim 9, characterized in that, The material coordinate information, lateral uniformity index, longitudinal uniformity index, and valid, retested, or invalid measurement status of the abnormal coating area are associated with the corresponding equal material length pulse markers to form a detection record; a roll material verification positioning interval is generated based on the starting position and ending position of the cumulative strain-free roll length, and the material space interval between adjacent detection records is determined by equal material length pulses after the production speed changes.

Images