A method of detecting formation uniformity of a wet paper web material of a sanitary paper

Abstract

The present application belongs to the field of intelligent detection and machine vision technology, and proposes a forming uniformity detection method for wet paper web material of toilet paper, specifically: arranging a CCD camera on a wet paper web conveying belt, then acquiring paper web images through real-time data acquisition of the CCD camera, and pre-processing the paper web images, then counting the number of uniform failure points and estimating the range of failure points from the paper web images, and forming a decision calculation value through speed matching decision analysis of the number of failure points and the estimated range of failure points, and finally judging the current wet paper web uniformity by using the decision value. Recognize the functional decline risk of the forming screen cloth device in the process of uniform distribution of pulp when the forming speed of the wet paper web is too fast in the production process, and timely find the wet paper web with abnormal flatness and surface smoothness quality caused by uneven stress, realize accurate evaluation of the uniformity of toilet paper material in the wet paper web process, and significantly reduce the loss of raw materials and production cost.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of intelligent detection and machine vision, and particularly relates to a method for detecting the forming uniformity of a wet paper web material of toilet paper. BACKGROUND

[0002] In the material detection of toilet paper, the flatness of the paper is a crucial indicator. On the one hand, the flatness is a reflection of the uniformity of the paper material distribution, and on the other hand, it better ensures the comfort during use. In high-end toilet paper products, the risk of breakage caused by uneven thickness is paid more attention to. Usually, the related detection of the flatness of the paper in the production process of toilet paper is usually based on testing of the finished product, but such testing leads to a large amount of raw material loss. The principle causing the flatness problem of the paper is that the matching decision range of the forming speed of the wet paper web and the uniform spreading quality of the pulp always exists dynamic matching. The forming speed of the wet paper web is usually preset according to the material of the toilet paper and the forming net cloth apparatus, so the forming speed of the wet paper web can guarantee a relatively stable output effect within the matching decision range, but the production system always has a high tendency for work efficiency, and a certain degree of abnormality is allowed on this basis. When the forming speed of the wet paper web exceeds the matching decision range, that is, the forming speed of the wet paper web is too fast, the process of the forming net cloth apparatus for uniform distribution of the pulp will appear functional decline, the wet paper web fibers are unevenly arranged or have excessive tendency, and then the paper web after pressing is prone to stress unevenness during drying, resulting in shrinkage difference of the paper during the drying process, which finally affects the flatness and surface smoothness of the paper. Therefore, by detecting the uniform distribution effect of the wet paper web pulp through machine vision, the rationality of the forming speed of the wet paper web can be controlled in advance, so as to guarantee the flatness of the paper in the material detection of the toilet paper. SUMMARY

[0003] The present application aims to provide a method for detecting the forming uniformity of a wet paper web material of toilet paper to solve one or more technical problems existing in the prior art and at least provide a beneficial choice or create conditions.

[0004] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for detecting the forming uniformity of a wet paper web material of toilet paper is provided, which comprises the following steps:

[0005] arranging a CCD camera on the wet paper web conveying belt; obtaining a paper web image through real-time data acquisition by the CCD camera and pre-processing the paper web image; counting the number of uniform failure points and the failure point estimation range from the paper web image; forming a decision calculation value through speed matching decision analysis of the number of failure points and the failure point estimation range; and judging the current uniformity of the wet paper web by using the decision value.

[0006] Further, the method of arranging the CCD camera on the wet paper web conveying belt is: making the imaging direction of the CCD camera perpendicular to the plane of the wet paper web conveying belt, using a ring-shaped or area light source as the light source, using a line-scan CCD camera as the CCD camera, or replacing the CCD camera with a CMOS camera.

[0007] Further, the method of obtaining a paper web image through real-time data acquisition by the CCD camera and pre-processing the paper web image is: using the CCD camera to acquire an image of the wet paper web on the wet paper web conveying belt, taking the obtained image as the paper web image, and performing grayscale processing on the paper web image; using a Gaussian filter to process the paper web image to reduce high-frequency noise, and using a median filter to remove isolated noise points; and enhancing the contrast of the image through histogram equalization.

[0008] Further, the method of counting the number of uniform failure points and estimating the range of the failure points from the paper web image is: using a detection model preset in the VisionPro program to identify a plurality of regions from the paper web image as uniform failure points, the number of the uniform failure points being recorded as the number of uniform failure points, and the diameter of the circumscribed circle of the uniform failure points being recorded as a sub-estimated range, and the average of each sub-estimated range being recorded as the estimated range of the failure points.

[0009] Further, the method of forming a decision calculation value through speed matching decision analysis of the number of failure points and the estimated range of the failure points is: setting a time period as a detection period R_LMT, R_LMT ∈ [0.5, 1.5] hours, and performing detection at equal intervals within the current R_LMT period; at each detection time T i The corresponding detection binary values are the number of uniform failure points Los_num i and the estimated range of the failure points Los_dis i ； The average and the standard deviation of all the estimated ranges of the failure points in the current detection period are recorded as Ave_dis and Std_dis, respectively, and the failure range boundary is set as Los_lin = Ave_dis - 3 × Std_dis; if the estimated range of the failure points Los_dis i at any time is less than the failure range boundary, the time is defined as a range defect time; and the period between any range defect time and the first range defect time in the reverse time direction is defined as a sub-defect interval.

[0010] Within the sub-defect interval, using the uniform number of failure points as the classification variable, the boxplot function is used to search for outliers in the predicted range of failure points for each classification variable. The sequence formed by the time intervals corresponding to each outlier is denoted as the outlier sequence Cov.ls. Within the outlier sequence, the frequency corresponding to each classification variable is defined as the spillover risk Spris. All first outliers under the classification variable with the maximum spillover risk are traversed, and the minimum value in the predicted range of failure points for each traversed element is defined as the failure level. For any element in the outlier sequence, if the predicted range of failure points at its corresponding time interval is less than the failure level, the failure level is updated to the predicted range of failure points at that time interval. The uniform number of failure points and the average value of the predicted range of failure points at each detection time interval are denoted as Ave_num and Ave_fdi, respectively. The value of any detection time T is calculated. i The corresponding codif i = 1 / n × (Los_num) i -Ave_num)(Los_dis i -Ave_fdi);

[0011] Based on each detection time T within any outlier sequence i The calculated decision calculus value Deval is:

[0012] ;

[0013] Codif j This refers to the test time T within an outlier sequence. i The moment before T i-1 And at the next moment T i+1 The corresponding cointegration product, mean is the average function, and ∏ is the product operator.

[0014] The aforementioned decision calculation values ​​are based on the accurate identification of the uniform number of failure points. The frequency of different categorical variables pinpoints the critical moments when wet paper web quality issues arise, thus effectively quantifying the real-time quality fluctuations of the wet paper web. However, the use of box plots to identify outliers can lead to overly coarse classification granularity during the acquisition of failure point numbers, causing some subtle quality differences to be overlooked. This is especially problematic when the frequency of failure point counts needs to be statistically analyzed to assess overflow risk. To address this issue of quality difference identification and improve the accuracy of decision calculation values, this invention proposes a more efficient solution.

[0015] Preferably, the method of forming a decision algorithm value by speed matching decision analysis of the number of failure points and the estimated range of failure points is as follows: a time period is set as a detection period R_LMT, R_LMT∈[0.5, 1.5] hours, and the value of each uniform failure point number is taken as a classification variable;

[0016] Detection is continuously performed within the current R_LMT period, and each detection time T i The corresponding detection binary values are the uniform failure point number Los_num i and the estimated range of failure points Los_dis i , and the frequency of each classification variable within the current R_LMT period is defined as the overflow probability,

[0017] The ratio of the overflow probability and the estimated range of failure points at each detection time is defined as the running density gradient Den_grad, and within the detection period R_LMT, the binary tuple composed of the detection time and the running density gradient constitutes the original time series set Orset, Utilizing The plrep and splev functions perform cubic spline interpolation on the original time series set to obtain a smooth curve, and the detection time corresponding to the maximum value and the minimum value in the obtained curve is the risk time, and the binary tuple composed of the risk time and the corresponding running density gradient constitutes the reconstructed time series set Reset, and the diff function is used to perform high-order difference operation on the non-stationary time series reconstructed time series set to obtain the stationary time series modified time series set;

[0018] Due to the factors of model accuracy and environmental changes, irregular fluctuations and noises may appear in the data acquisition results. Through cubic spline interpolation, these fluctuations can be smoothed, so as to more accurately reflect the actual trend of the wet paper web quality. In addition, in the case of sudden failure or quality decline, the smooth curve generated by cubic spline interpolation can help to clearly identify these change points, i.e. the maximum value and the minimum value, so as to facilitate positioning the key moment when the wet paper web quality problem occurs. After high-order difference processing, the autocorrelation between the data points in the modified time series set will be significantly reduced, because the difference operation removes the correlation components in the sequence to some extent. Sequences with high autocorrelation often contain some long-term trends, which may not be the real reflection of the wet paper web quality change, but the inertia phenomenon in the model or preparation process. By reducing the autocorrelation to remove the trend component, the real-time quality fluctuation can be better focused.

[0019] For any risk time, define the absolute value of the difference between the running density gradient corresponding to the reconstructed time set and the corrected time set as the reconstruction error Rerro, all reconstruction errors constitute a first overflow sequence, take the risk time corresponding to each maximum value in the first overflow sequence as an overflow point, and record the set of each detection time between any overflow point and the first overflow point in the reverse time direction as a peak point interval.

[0020] In any peak point interval, calculate the reconstruction error of any non-risk time by linear interpolation method, record the median of all time reconstruction errors as the error median, and the difference between the reconstruction error of each time and the error median is the median deviation. When the median deviation of any time is less than half of the difference between the reconstruction errors of the two overflow points in the peak point interval to which it belongs, the time is removed from the peak point interval. Record the set of times after screening as the corrected point set of the peak point interval, and define the data retention rate of the corrected point set as the ratio between the number of times not removed and the original number of times. Use the absolute value of the median deviation to replace the reconstruction error corresponding to all times between the two overflow points in the corrected point set, and the failure point estimation range corresponding to any time in the corrected point set constitutes a second overflow sequence Scout.

[0021] For any detection time ReT i in any corrected point set, calculate the sub-decision evaluation value m_Deval:

[0022] ;

[0023] Where FsT and ScT are the start overflow point and the end overflow point in the corrected point set where the detection time ReT i is located, and hs<> is the harmonic mean function.

[0024] Define the maximum value of all sub-decision evaluation values in each corrected point set as the decision evaluation representative value S_Deval, and calculate the weighted average of each decision evaluation representative value in the peak point interval as the decision evaluation value Deval. The weight in the weighted average calculation is the data retention rate.

[0025] Further, the method for judging the current wet paper web uniformity by using the decision value is: the difference between the decision value at the current time and the first maximum value searched in the reverse time direction is the calculation difference, if the calculation difference at the current time is greater than the calculation difference at the previous time and is a positive number, the current time is defined as satisfying the first decision barrier condition; the preset time period is the recent decision section SR_LMT, the value range of which is SR_LMT∈[5, 10] minutes; the SR_LMT time period in the reverse time direction of the current time is defined as the current recent decision section, and the proportion of the time at which the calculation difference value is positive in the current recent decision section is the positive calculation proportion; if the positive calculation proportion at the current time is greater than the upper quartile of the positive calculation proportion at each time in the time period SR_LMT, the current time is defined as satisfying the second decision barrier condition; if the current time satisfies both the first decision barrier condition and the second decision barrier condition, it is determined that the current time does not satisfy the uniformity requirement of the wet paper web.

[0026] Preferably, all the undefined variables in the present application can be manually set thresholds if not defined.

[0027] The present application also provides a forming uniformity detection system for a wet paper web material of toilet paper, which comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps in the forming uniformity detection method for the wet paper web material of toilet paper when executing the computer program, and the forming uniformity detection system for the wet paper web material of toilet paper can be run in a desktop computer, a notebook computer, a palm computer, a cloud data center, and other computing devices, and the executable system can include, but is not limited to, a processor, a memory, a server cluster, and the processor executes the computer program to run in the following system units:

[0028] An image acquisition setting unit is configured to arrange a CCD camera on a wet paper web conveying belt.

[0029] A preprocessing unit is configured to acquire a paper web image through real-time data acquisition by the CCD camera and pre-process the paper web image.

[0030] An image analysis unit is configured to count the number of uniformity failure points and estimate the range of the failure points from the paper web image.

[0031] A decision analysis unit is configured to perform speed matching decision analysis to form a decision value by the number of failure points and the estimated range of the failure points.

[0032] A judgment and screening unit is configured to judge the current wet paper web uniformity by using the decision value.

[0033] The present application provides a forming uniformity detection method for a wet paper web material of toilet paper, which realizes real-time detection and analysis on the wet paper web image through machine vision technology, quantifies the functional decline risk of the forming speed of the wet paper web in the production process, and further finds the wet paper web with abnormal flatness and surface smoothness quality caused by uneven stress in time, so as to realize accurate evaluation on the uniformity of the toilet paper material in the production process. Not only the raw material loss and production cost are significantly reduced, but also the labor cost is reduced by reducing the dependence on manual inspection, the problems of false detection and missed detection caused by human factors are effectively avoided, the production efficiency is significantly improved, and a detection scheme for flatness and smoothness quality guarantee is provided for the toilet paper industry. BRIEF DESCRIPTION OF DRAWINGS

[0034] The above and other features of the present application will become more apparent from the following detailed description of embodiments taken in conjunction with the accompanying drawings, in which like reference characters indicate the same or similar elements throughout the drawings, and in which:

[0035] Figure 1 Fig. 1 shows a flowchart of a forming uniformity detection method for a wet paper web material of toilet paper;

[0036] Figure 2 Fig. 2 shows a forming uniformity detection system structure diagram for a wet paper web material of toilet paper. DETAILED DESCRIPTION

[0037] The concept, specific structure and generated technical effects of the present application will be described clearly and completely in combination with embodiments and drawings, so as to fully understand the purpose, scheme and effects of the present application. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0038] As Figure 1 Fig. 1 shows a flowchart of a forming uniformity detection method for a wet paper web material of toilet paper, and the forming uniformity detection method for a wet paper web material of toilet paper according to the embodiments of the present application will be described below in combination with Figure 1 The method comprises the following steps:

[0039] The CCD camera is arranged on the wet paper web conveying belt; the paper web image is obtained through real-time data acquisition by the CCD camera, and the paper web image is pretreated; the number of uniform failure points and the failure point estimation range are counted from the paper web image; the decision calculation value is formed through speed matching decision analysis by the number of failure points and the failure point estimation range; and the current wet paper web uniformity is judged by using the decision value.

[0040] Further, the method of arranging the CCD camera on the wet paper web conveying belt is that the imaging direction of the CCD camera is perpendicular to the plane of the wet paper web conveying belt, the light source is a ring or a surface light source, the CCD camera is a line scanning CCD camera, or a CMOS camera is used to replace the CCD camera.

[0041] Further, the method of obtaining the paper web image through real-time data acquisition by the CCD camera and pretreating the paper web image is that the image of the wet paper web on the wet paper web conveying belt is collected by using the CCD camera, the obtained image is taken as the paper web image, and the paper web image is grayed; the high-frequency noise is reduced by using a Gaussian filter on the paper web image, and isolated noise points are removed by using a median filter; and the contrast of the image is enhanced by histogram equalization, so that the edge of the paper web is more obvious.

[0042] Further, the method of counting the number of uniform failure points and the failure point estimation range from the paper web image is that a detection model is preset in the VisionPro program, and a plurality of regions are identified as uniform failure points from the paper web image by using the detection model, the number of the uniform failure points is recorded as the number of uniform failure points, the diameter of the circumscribed circle of the uniform failure points is recorded as the sub-estimation range, and the average value of each sub-estimation range is taken as the failure point estimation range.

[0043] In the process of obtaining the uniform failure points, the paper web image is binarized by using the Ostu threshold method, connected component labeling is performed, and a plurality of foreground regions are identified as uniform failure points; the connected component labeling refers to marking the continuous region composed of pixels with the same value of 255 as the region with the same serial number in the binary image.

[0044] Further, the method of forming the decision calculation value through speed matching decision analysis by the number of failure points and the failure point estimation range is that a time period is taken as a detection period R_LMT, R_LMT∈[0.5, 1.5] hours, detection is performed at equal intervals within the current R_LMT period, that is, the number of failure points and the failure point estimation range are obtained once, the detection event scale is recorded as a detection time T i ;T i has a one-to-one correspondence with i; the corresponding detection binary value of each detection time T i is the number of uniform failure points Los_num i and the failure point estimation range Los_dis.i , wherein binary value refers to the element in binary group; Ave_dis and Std_dis are respectively recorded as the average value and standard deviation of all failure point prediction range at current detection period, and the failure range boundary is set as Los_lin = Ave_dis - 3 x Std_dis; if the failure point prediction range Los_dis obtained at any time is less than the failure range boundary, the time is defined as the range defect time; the period between any range defect time and the first range defect time in the reverse time direction thereof is defined as a sub-defect interval; i i i i i

[0045] In the sub-defect interval, the uniform failure point number is taken as the classification variable, the boxplot function is used to search the outliers of the failure point prediction range under each classification variable, and the sequence composed of the time corresponding to each outlier is recorded as the outlier sequence Cov.ls; in the outlier sequence, the frequency corresponding to each classification variable is defined as the overflow risk Spris, wherein the frequency corresponding to each uniform failure point number refers to the ratio of the number of outliers in the same classification variable to the number of elements in the outlier sequence. All the first outliers in the classification variable corresponding to the overflow risk with the maximum value are traversed, and the minimum value in the failure point prediction range corresponding to each traversed element is defined as the failure level; the first outlier refers to the element in the outlier sequence; for any element in the outlier sequence, if the failure point prediction range at the time corresponding thereto is less than the failure level, the failure level is updated to the failure point prediction range at the time, the average value of the uniform failure point number and the failure point prediction range at each detection time is recorded as Ave_num and Ave_fdi respectively, and the consensus product Codif corresponding to any detection time T i i i i is calculated.

[0046] According to each detection time T i in any outlier sequence, the decision algorithm value Deval is calculated.

[0047] ;

[0048] wherein Codif j refers to the consensus product corresponding to the time T i and the time T i-1 and the time T i+1 preceding and succeeding the detection time T i in the outlier sequence; if T i is the initial detection time or the end detection time, the denominator is defined as , the initial detection time or the end detection time refers to the first and last detection time in the R_LMT period, exp() is the exponential function with e as the base number, and the boxplot function is called from the matplotlib library in python;

[0049] Preferably, the method of forming a decision algorithm value by analyzing the speed matching decision through the number of failure points and the estimated range of failure points is as follows: a time period is set as the detection period R_LMT, R_LMT ∈ [0.5, 1.5] hours, and the value of each uniform failure point number is taken as a classification variable;

[0050] Continuous detection is performed in the current R_LMT period, and each detection time T i The corresponding detection binary values are the uniform failure point number Los_num i and the estimated range of failure points Los_dis i , and the frequency corresponding to each classification variable in the current R_LMT period is defined as the overflow probability, wherein the frequency corresponding to any classification variable refers to the proportion of the number of times of occurrence of a classification variable to the total number of detections in the detection period;

[0051] The ratio of the overflow probability to the estimated range of failure points at each detection time is defined as the running density gradient Den_grad, and in the detection period R_LMT, the binary tuple composed of the detection time and the running density gradient constitutes the original time series set Orset, and the smooth curve is obtained by using the plrep and splev functions for cubic spline interpolation of the original time series set. The maximum value and the minimum value in the obtained curve correspond to the risk time, and when the time corresponding to the maximum value and the minimum value is not a detection time, the previous detection time and the next detection time with the shortest interval from the time are defined as the risk time. The binary tuple composed of the risk time and the corresponding running density gradient constitutes the reconstructed time series set Reset, and the stationary time series modified time series set is obtained by using the diff function for high-order difference operation of the non-stationary time series reconstructed time series set;

[0052] For any risk time, the absolute value of the difference of the running density gradient corresponding to the reconstructed time series set and the modified time series set is defined as the reconstruction error Rerro, and all reconstruction errors constitute the first overflow sequence. The risk time corresponding to each maximum value in the first overflow sequence is taken as the overflow point, and the set of each detection time between any overflow point and the first overflow point in the reverse time direction is taken as the peak point interval;

[0053] In any peak interval, the reconstruction error of any non-risk time is calculated by linear interpolation method, and the median of all reconstruction errors is recorded as the error median. The difference between the reconstruction error of each time and the error median is the median deviation. If the median deviation of any time is less than half of the difference between the reconstruction errors of the two overflow points in its peak interval, the time is removed from the peak interval. The set of filtered times is recorded as the modified point set of the peak interval, and the data retention rate of the modified point set is defined as the ratio between the number of times not removed and the original number of times. The absolute value of the median deviation is used instead of the reconstruction error of all times between the two overflow points in the modified point set. The reconstruction errors of the two overflow points are not changed. The failure point estimation range corresponding to any time in the modified point set constitutes the second overflow sequence Scout.

[0054] For any detection time ReT i , the sub-decision evaluation value m_Deval is calculated:

[0055] ;

[0056] where FsT and ScT are the start overflow point and the end overflow point in the modified point set where the detection time ReT i is located, hs<> is the harmonic mean function, exp() is the exponential function with base e.

[0057] The maximum value of all sub-decision evaluation values in each modified point set is defined as the decision evaluation representative value S_Deval, and the weighted average of each decision evaluation representative value in the peak interval is calculated as the decision evaluation value Deval. The weight in the weighted average calculation is the data retention rate.

[0058] The calculation method of the decision evaluation value is as follows:

[0059] ;

[0060] where Num is the number of modified point sets in the current detection period, the weight used is the data retention rate of each modified point set, and S_Deval j , Darat j are the decision evaluation value and the data retention rate of the jth modified point set.

[0061] where the splrep and splev functions in python are called by the matplotlib library, and the diff function is called by the statsmodels library.

[0062] Further, the method for judging the current wet paper web uniformity by using the decision value is: the difference between the decision calculation value at the current time and the first maximum value searched in the reverse time direction is obtained, and the obtained difference value is recorded as a calculation difference; if the calculation difference at the current time is greater than the calculation difference at the previous time and is a positive number, the current time is defined as satisfying the first decision obstacle condition; a preset time period is used as a new decision section SR_LMT, and the value range of the new decision section SR_LMT is SR_LMT [5, 10] minutes; the SR_LMT time period in the reverse time direction of the current time is defined as a current new decision section, and the proportion of the time at which the calculation difference value is a positive value in the current new decision section is a positive calculation proportion; if the positive calculation proportion corresponding to the current time is greater than the upper quartile value of the positive calculation proportion at each time in the time period SR_LMT, the current time is defined as satisfying the second decision obstacle condition; if the current time satisfies the first decision obstacle condition and the second decision obstacle condition at the same time, it is determined that the wet paper web corresponding to the current time does not satisfy the uniformity requirement.

[0063] The embodiment of the present application provides a forming uniformity detection system of a wet paper web material of toilet paper. Figure 2 As shown in FIG. 1, the embodiment of the present application provides a forming uniformity detection system of a wet paper web material of toilet paper, which comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, and the processor realizes the steps in the forming uniformity detection method of the wet paper web material of toilet paper.

[0064] The system comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, and the processor executes the computer program to run in the following units of the system:

[0065] An image acquisition setting unit is configured to arrange a CCD camera on the wet paper web conveying belt.

[0066] A preprocessing unit is configured to acquire a paper web image by real-time data acquisition of the CCD camera and pre-process the paper web image.

[0067] An image analysis unit is configured to count the number of uniformity failure points and estimate the range of the failure points from the paper web image.

[0068] A decision analysis unit is configured to analyze the speed matching decision by the number of failure points and the estimated range of the failure points to form a decision calculation value.

[0069] A judgment and screening unit is configured to judge the current wet paper web uniformity by using the decision value.

[0070] The forming uniformity detection system of the wet paper web material of the sanitary paper can run in a desktop computer, a notebook computer, a palm computer, a cloud server and other computing devices. The forming uniformity detection system of the wet paper web material of the sanitary paper can run a system which can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the example is only an example of the forming uniformity detection system of the wet paper web material of the sanitary paper, and does not constitute a limitation on the forming uniformity detection system of the wet paper web material of the sanitary paper, and can include more or less components, or combine certain components, or different components, for example, the forming uniformity detection system of the wet paper web material of the sanitary paper can also include an input / output device, a network access device, a bus and the like.

[0071] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor or the like. The processor is a control center of the forming uniformity detection system of the wet paper web material of the sanitary paper running system, and connects each part of the forming uniformity detection system of the wet paper web material of the sanitary paper running system through various interfaces and lines.

[0072] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the system for detecting the forming uniformity of the wet paper web material of the sanitary paper by running or executing the computer program and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playing function, an image playing function, etc.), and the like; and the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), and the like. In addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0073] Although the description of the present application has been quite detailed and particularly described with respect to several described embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, but rather effectively covers the intended scope of the present application. Furthermore, the present application has been described above with respect to embodiments that the inventors can foresee, and the purpose is to provide a useful description, and non-essential modifications to the present application that have not been foreseen at present can still represent equivalent modifications to the present application.

Claims

1. A method for detecting the forming uniformity of wet paper web material in toilet paper, characterized in that, The method includes the following steps: arranging a CCD camera on the wet paper web conveyor belt; acquiring paper web images in real time through the CCD camera, and preprocessing the paper web images; counting the number of uniform failure points and the estimated range of failure points from the paper web images; performing speed matching decision analysis based on the number of failure points and the estimated range of failure points to form a decision calculation value; and using the decision value to judge the current uniformity of the wet paper web. The method for generating decision calculation values ​​through speed matching decision analysis based on the number of failure points and the estimated range of failure points is as follows: Let a time period be designated as the detection period R_LMT, where R_LMT ∈ [0.5, 1.5] hours. Detections are performed at equal intervals within the current R_LMT period, with each detection time T... i The corresponding binary values ​​obtained are the uniform number of failure points, Los_num. i And the estimated range of failure points Los_dis i Let Ave_dis and Std_dis be the average and standard deviation of the estimated range of all failure points during the current detection period, respectively, and set the boundary of the failure range as Los_lin = Ave_dis - 3 × Std_dis; if the estimated range of failure points Los_dis at any time is... i If the time is less than the failure range boundary, then the time is defined as the range defect time; the time interval between any range defect time and the first range defect time in the reverse time direction is defined as a sub-defect interval. Within the sub-defect interval, using the uniform number of failure points as the classification variable, the boxplot function is used to search for outliers in the predicted range of failure points under each classification variable. The sequence formed by the time intervals corresponding to each outlier is denoted as the outlier sequence Cov.ls. Within the outlier sequence, the frequency corresponding to each classification variable is defined as the spillover risk Spris. All first outliers under the classification variable with the maximum spillover risk are traversed, and the minimum value in the predicted range of failure points corresponding to each traversed element is defined as the failure level. For any element in the outlier sequence, if the predicted range of failure points at its corresponding time interval is less than the failure level, the failure level is updated to the predicted range of failure points at that time interval. The average number of uniform failure points and the average value of the predicted range of failure points at each detection time interval are denoted as Ave_num and Ave_fdi, respectively. The value of any detection time T is calculated. i The corresponding cointegration product is used to calculate the decision calculus value Deval based on the spillover risk and the cointegration product, where at any detection time T i The corresponding codif product is: Codif i = 1 / n × (Los_num) i -Ave_num)(Los_dis i -Ave_fdi), where the decision calculus value Deval is calculated based on the spillover risk and the cointegration product. The process of calculating the decision calculus value Deval is as follows: ； Codif j This refers to the test time T within an outlier sequence. i The moment before T i-1 And at the next moment T i+1 The corresponding cointegrated product, mean is the average function, and ∏ is the product operator.

2. The method for detecting the forming uniformity of wet paper web material of toilet paper according to claim 1, characterized in that, The method for arranging a CCD camera on the wet paper web conveyor belt is as follows: the imaging direction of the CCD camera is perpendicular to the plane of the wet paper web conveyor belt, a ring or surface light source is used, and the CCD camera is a line scan CCD camera, or a CMOS camera is used instead of a CCD camera.

3. The method for detecting the forming uniformity of wet paper web material of toilet paper according to claim 1, characterized in that, The method for acquiring paper web images in real time using a CCD camera and preprocessing the paper web images is as follows: the CCD camera is used to acquire images of the wet paper web on the wet paper web conveyor belt, and the acquired images are used as paper web images. The paper web images are then converted to grayscale. Gaussian filters are used to process the paper web images to reduce high-frequency noise, and median filters are used to remove isolated noise points. Histogram equalization is used to enhance the contrast of the images.

4. The method for detecting the forming uniformity of wet paper web material of toilet paper according to claim 1, characterized in that... The method for statistically analyzing the uniform number of failure points and the estimated range of failure points from paper images is as follows: By pre-setting a detection model in the VisionPro program, the detection model identifies several regions from the paper image and records them as uniform failure points. The number of these regions is recorded as the uniform number of failure points. The outer diameter of the uniform failure point is the sub-estimated range, and the average value of each sub-estimated range is used as the estimated range of failure points.

5. The method for detecting the forming uniformity of wet paper web material of toilet paper according to claim 1, characterized in that, The method of generating decision calculation values ​​by speed matching decision analysis based on the number of failure points and the estimated range of failure points can be replaced by: setting a time period as the detection period R_LMT, R_LMT∈[0.5,1.5] hours, and using the values ​​of the uniform number of failure points as hierarchical variables respectively; Detection is performed continuously within the current R_LMT time period, with each detection time T... i The corresponding binary values ​​obtained are the uniform number of failure points, Los_num. i And the estimated range of failure points Los_dis i The overflow probability is defined as the frequency of each hierarchical variable within the current R_LMT period. The ratio of the overflow probability to the predicted range of the failure point at each detection time is defined as the running density gradient Den_grad. During the detection period R_LMT, the original time series set Orset is formed by the tuple composed of the detection time and the running density gradient. The original time series set is subjected to cubic spline interpolation using the plcep and splev functions to obtain a smooth curve. The detection times corresponding to the maximum and minimum values ​​in the obtained curve are the risk times. The tuple composed of the risk times and their corresponding running density gradients constitutes the reconstructed time series set Reset. The reconstructed time series set of the non-stationary time series is subjected to high-order difference operation using the diff function to obtain the corrected time series set of the stationary time series. For any risk moment, the absolute value of the difference between the running density gradients corresponding to the reconstructed time series set and the corrected time series set is defined as the reconstruction error Rerro. All reconstruction errors constitute the first overflow sequence. The risk moments corresponding to each maximum value in the first overflow sequence are taken as overflow points. The set of detection moments between any overflow point and the first overflow point in the reverse time direction is denoted as the peak interval. Within any peak interval, the reconstruction error at any non-risk moment is calculated using linear interpolation. The median of the reconstruction errors at all moments is denoted as the median error. The difference between the reconstruction error at each moment and the median error is denoted as the median deviation. If the median deviation corresponding to any moment is less than half the difference in reconstruction errors between the two overflow points within its peak interval, then that moment is removed from the peak interval. The set of filtered moments is denoted as the set of correction points for the peak interval. The data retention rate of the set of correction points is defined as the ratio between the number of moments that were not removed and the original number of moments. The absolute value of the median deviation is used to replace the reconstruction error corresponding to all moments within the set of correction points that lie between the two overflow points. The estimated range of failure points corresponding to any moment within the set of correction points constitutes the second overflow sequence Scout. For any detection time ReT within any set of correction points i The sub-decision calculus value m_Deval is calculated as follows: ; Where FsT and SCT are the detection times ReT i The set of correction points contains the starting and ending overflow points, and hs<> is the harmonic mean function.

6. The method for detecting the forming uniformity of wet paper web material of toilet paper according to claim 1, characterized in that, The method for judging the uniformity of the current wet paper web using decision values ​​is as follows: the difference between the current decision calculation value and the first maximum value searched in the reverse time direction is recorded as the calculation difference. If the calculation difference at the current time is larger than the calculation difference at the previous time and is positive, then the current time is defined as meeting the first decision obstacle condition. A time period is preset as the recent decision segment SR_LMT. The SR_LMT time period in the reverse time direction of the current time is defined as the current recent decision segment. The proportion of times with positive calculation differences in the current recent decision segment is defined as the positive calculation proportion. If the positive calculation proportion corresponding to the current time is greater than the upper quartile of the positive calculation proportion at each time in the time period SR_LMT, then the current time is defined as meeting the second decision obstacle condition. If the current time meets both the first and second decision obstacle conditions, then the wet paper web corresponding to the current time is determined to not meet the uniformity requirements.

7. A system for detecting the forming uniformity of wet paper web material for toilet paper, characterized in that, The forming uniformity detection system for wet paper web material of toilet paper includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the forming uniformity detection method for wet paper web material of toilet paper according to any one of claims 1-6. The forming uniformity detection system for wet paper web material of toilet paper runs on a desktop computer, a laptop computer, a handheld computer, or a computing device in a cloud data center.

Images