A parameter anomaly diagnosis and early warning method for a curved tape-based tail production process

Abstract

This invention provides a parameter anomaly diagnosis and early warning method for the material head production process based on a curve band. It utilizes historical data to screen normal batches; calculates the maximum and minimum values ​​of parameters in the screened historical data; performs curve fitting using a linear regression model for all maximum and minimum values; and determines in real time whether the parameter falls within the curve band range. The method proposed in this invention first filters historical data to identify abnormal batches, and then uses the curve band as the anomaly judgment criterion, effectively avoiding the shortcomings of existing technologies.

Description

Technical Field

[0001] This invention belongs to the field of tobacco processing technology, specifically relating to a method for abnormal parameter diagnosis and early warning in the feed head production process based on a curved strip. Background Technology

[0002] Cigarette manufacturing involves multiple production processes, with almost every stage including the initial, middle, and final stages of production. Since the initial and final stages account for a relatively small proportion of the total production time, most current tobacco production early warning systems only issue alerts for anomalies occurring during the middle stage. For example, patent application CN113837608A discloses a tobacco processing control and early warning platform based on big data analysis, consisting of four modules: a cleaning system, an operating terminal, a control system, and an early warning platform. However, this method fails to consider the possibility of abnormal batches in historical data during the early warning process. It relies solely on the highest and lowest points of historical data as the real-time early warning standard, which may include outliers. This results in an insufficiently rigorous early warning standard, leading to ineffective warnings, such as failing to issue an alert when abnormal parameters occur.

[0003] The inventors discovered that, in actual production processes, apart from abnormal batches causing ineffective early warnings, abnormalities occurring at the material head stage usually have a certain impact on production quality. Due to the special nature of the material head stage, when diagnosing abnormalities in parameters, parameter setting standards cannot be simply used as the basis for diagnosing material head abnormalities. Summary of the Invention

[0004] To avoid the aforementioned problems, this invention proposes an anomaly detection and early warning method based on a curve band for inlet and outlet moisture, cylinder wall temperature, steam flow rate, and material flow rate. The curve band consists of two curves, one with an upper boundary and one with a lower boundary. Early warning based on the curve band means that when the data falls outside the upper and lower boundaries of the curve band, an anomaly is considered to have occurred and an early warning is issued; otherwise, no warning is issued. This curve band-based method can detect and issue early warnings for several important indicators in the inlet and outlet stage, thereby effectively ensuring the production quality of cigarettes.

[0005] To achieve the above objectives, the present invention provides a method for abnormal parameter diagnosis and early warning in the material head production process based on curved strips, comprising the following steps:

[0006] s1. Use historical data to filter normal batches;

[0007] s2. Find the maximum and minimum values ​​of the parameters for the filtered historical data;

[0008] s3. Apply a linear regression model to fit curves for all maximum and minimum values ​​respectively;

[0009] s4. Determine in real time whether the parameter is within the curve range.

[0010] Preferably, in step s1, normal batches are filtered using historical data in the following way:

[0011] First, obtain n batches of historical production data, and then filter out the data for the material head section;

[0012] Next, the average moisture content at the outlet of the n batches of historical feedstock is calculated according to different points. When there are k points in the feedstock, there are k average values.

[0013] Then, a mean standard is formed by k means, and the correlation coefficient is calculated between the moisture data of the feed head section outlet of n batches and the mean standard. When the absolute value of the correlation coefficient is less than x, the batch is regarded as an abnormal batch and removed. Finally, m batches of normal batches are retained (n≥m).

[0014] Preferably, in step s2, the maximum and minimum values ​​of the parameters of the filtered historical data are calculated using the following method:

[0015] For the m batches of normal batches that are retained, the maximum and minimum values ​​are calculated based on the points. Since there are k points, k maximum values ​​and k minimum values ​​can be obtained.

[0016] Preferably, in step s3, two curves are fitted to all maximum and minimum values ​​using a linear regression model, and the two curves serve as the upper and lower bounds of the curve band, respectively.

[0017] Preferably, in step s4, the parameter is determined in real time whether it is within the curve band range using the following method:

[0018] After obtaining the range of the curve zone, the sensor transmits the actual value of the parameter to the system in real time. The system automatically matches whether the actual value falls within the range of the curve zone. When it exceeds the range of the curve zone, an early warning is issued to remind the worker that an abnormality has occurred.

[0019] More preferably, the method for obtaining the mean standard is as follows: plot the outlet moisture data of n batches of feed head stage on the same coordinate axis according to the points, with the horizontal axis representing the point and the vertical axis representing the outlet moisture value, and draw n outlet moisture change curves for the n batches of feed head stage. Select point t, obtain the n outlet moisture values ​​corresponding to point t in the n outlet moisture change curves, and calculate the mean of the n outlet moisture values ​​as the mean standard for outlet moisture of feed head stage.

[0020] The method proposed in this invention first filters historical data in batches to identify abnormal batches, and then uses curves as an anomaly judgment criterion, which can effectively avoid the defects of existing technologies. Attached Figure Description

[0021] Figure 1 This is a flowchart of a parameter anomaly diagnosis and early warning method for a feedstock production process based on a curved strip.

[0022] Figure 2-1 , 2-2 Tables 2-2, 2-3, and 2-4 show the data for 20 batches of material head section.

[0023] Figure 3 The table below shows the correlation coefficients between the moisture content at the outlet of the feedstock section for 20 batches and the mean.

[0024] Figure 4 The image shows the upper and lower limits of the curve band.

[0025] Figure 5 The image shown is a comparison of the moisture content curves at the beginning of a certain batch of exported materials. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of this invention.

[0027] like Figure 1 The diagram shows a flowchart of a parameter anomaly diagnosis and early warning method for the feed head production process based on a curve band. Multiple batches of historical data are selected; the average value of the feed head section outlet moisture data for each point in the multiple batches of historical data is calculated to obtain the standard average outlet moisture value; the correlation coefficient between the outlet moisture data of each batch of feed head section and the standard average outlet moisture value is calculated; when the absolute value of the correlation coefficient is less than a certain preset value, the batch of data is considered an abnormal batch and removed; the maximum and minimum values ​​of the feed head parameters for the retained normal batches are calculated for each point; the maximum and minimum values ​​of the corresponding parameters are fitted using linear regression to obtain the upper and lower bounds of the corresponding curve band; the actual value of the input parameter is judged in real time whether it exceeds the range of the curve band, and an early warning is issued for those exceeding the range.

[0028] The following section uses 20 batches of loosely ventilated export products as an example to detail the abnormality diagnosis process.

[0029] S1. Obtain 20 batches of loose historical data from the system database, see Table 1.

[0030] batch DZ101 DZ102 DZ103 DZ104 DZ105 DZ106 DZ107 DZ108 DZ109 DZ110 DZ111 DZ112 DZ113 DZ114 DZ115 DZ116 DZ117 DZ118 DZ119 DZ120

[0031] Table 1

[0032] S2. The method for removing abnormal batches from these 20 batches of data is as follows:

[0033] Data from these 20 batches of material head section were filtered out, such as Figures 2-1 to 2-8 As shown;

[0034] The average moisture content of the 20 batches of historical raw material at the export outlet was calculated based on different locations. Figure 2-1 The average moisture content at the outlet of the first batch of raw materials for the 20 historical batches shown in the first point (first row of data) is -2.9875, the average moisture content for the second point (second row of data) is -2.9885, and so on.

[0035] These mean values ​​form the standard mean for the moisture content at the feed head outlet. Then, the Pearson correlation coefficient is calculated between the moisture content data at the feed head outlet of these 20 batches and the standard mean. (See [link to relevant documentation]). Figure 3 The table shows the correlation coefficients between the moisture content data at the outlet of the feed head section for the 20 batches and the mean.

[0036] When the absolute value of the correlation coefficient is less than 0.7, the batch is treated as an abnormal batch (DZ120) and removed, and the remaining normal batches are retained.

[0037] S3. For the retained normal batches, filter out their export moisture data, calculate the maximum and minimum values ​​for each data point, and fit two curves using a linear regression model for each maximum and minimum value. These two curves serve as the upper and lower bounds of the curve band, respectively. See [link / reference]. Figure 4 The upper and lower limits of the curve shown are plotted.

[0038] S4. After obtaining the curve range, the sensor collects the outlet moisture data of the current production batch in real time. (See below) Figure 5 The system compares the moisture content curve of a certain batch of material at the outlet with the curve band. The system determines in real time whether the moisture content at the outlet is within the range of the curve band. When it exceeds the range of the curve band, an early warning is issued to remind the workers that there is an abnormality.

[0039] The method of fitting curves to the maximum and minimum values ​​can be achieved by setting a sliding window c, fitting the curve once for every c points, which is equivalent to the upper and lower boundaries of the curve band being composed of multiple curve segments.

[0040] This invention is not only applicable to the diagnosis of material head abnormalities in the ultra-loosening process, but can also be applied to the diagnosis of material head abnormalities in other processes, including: leaf feeding, thin plate drying, and flavoring. The specific embodiments described above are merely one selected example of this invention and are not intended to limit the invention. Any modifications, supplements, and equivalent substitutions made within the scope of the principles of this invention should be included within the protection scope of this invention.

[0041] Cigarette production involves multiple processes, including the initial production stage, the middle production stage, and the final production stage. The initial and final production stages account for a relatively small proportion of the total production time, while the middle production stage accounts for a larger proportion. Current cigarette production early warning systems typically use process standards or set standards to provide early warnings for indicators and parameters in the middle production stage. Due to the special nature of the initial production stage and its short duration, there are fewer detection and early warning systems for the initial production stage.

[0042] This invention discloses a parameter anomaly diagnosis and early warning method for the feedstock production process based on curve strips. It has been found that anomalies in the feedstock stage have a significant impact on the quality of subsequently produced cigarettes. By averaging the outlet moisture data of the feedstock segment from historical data points within the same process, a standard average outlet moisture value is obtained. Then, the correlation coefficient between the outlet moisture value of each batch of feedstock and the standard average value is calculated, and batches with excessively low correlation coefficients are eliminated, improving the accuracy and objectivity of the feedstock production process parameters. Subsequently, the maximum and minimum values ​​of the corresponding parameter indicators are calculated for each point, and all maximum and minimum values ​​are connected to form a curve. A linear regression is used to fit the curve, serving as the upper and lower bounds of the curve strip. The derived curve strip is then used to diagnose and issue early warnings for anomalies in the feedstock data.

[0043] The curve-based method proposed in this invention effectively solves the problem that it is impossible to simply use parameter setting standards to diagnose and warn of anomalies in several important indicators at the feed stage, thereby ensuring the production quality of cigarettes.

Claims

1. A method for abnormal parameter diagnosis and early warning in the material head production process based on curved strips, characterized in that, It includes the following steps: s1. Use historical data to filter normal batches: Obtain n batches of historical production data, and filter out the data for the material head section respectively; The average moisture content at the outlet of the n batches of historical raw materials is calculated based on different points. When there are k points in the raw material head section, there are k average values. Using k means to form a standard mean, the correlation coefficients of the moisture content data at the outlet of the feed head section of n batches are calculated with the standard mean. When the absolute value of the correlation coefficient is less than the preset value x, the batch is regarded as an abnormal batch and removed. Finally, m batches of normal batches are retained. s2. Find the maximum and minimum values ​​of the parameters for the filtered historical data; For s3 and k points, curves are fitted using a linear regression model for all maximum and minimum values ​​respectively, to obtain the upper and lower bounds of the corresponding curve band, so that the curve band is composed of multiple curve segments. s4. Determine in real time whether the parameter is within the curve range.

2. The parameter anomaly diagnosis and early warning method for the material head production process based on curved strips as described in claim 1, characterized in that, In step s2, the maximum and minimum values ​​of the parameters of the filtered historical data are calculated using the following method: For the m batches of normal batches that are retained, the maximum and minimum values ​​are calculated based on the points. Since there are k points, k maximum values ​​and k minimum values ​​can be obtained.

3. A method for abnormal parameter diagnosis and early warning in a material head production process based on a curved strip, as described in claim 1 or 2, characterized in that, The method for obtaining the mean standard is as follows: Plot the outlet moisture data of n batches of feed head section on the same coordinate axis according to the points, with the horizontal axis representing the point and the vertical axis representing the outlet moisture value. Plot n outlet moisture change curves for the n batches of feed head section. Select point t and obtain n outlet moisture values ​​corresponding to point t in the n outlet moisture change curves. Calculate the mean of the n outlet moisture values ​​and use it as the mean standard for outlet moisture of feed head section.

4. A method for abnormal parameter diagnosis and early warning in a material head production process based on a curved strip, as described in any one of claims 1 to 3, characterized in that, In step s4, the parameter is determined in real time whether it is within the curve band range using the following method: After obtaining the range of the curve zone, the sensor transmits the actual value of the parameter to the system in real time. The system automatically matches whether the actual value falls within the range of the curve zone. When it exceeds the range of the curve zone, an early warning is issued to remind the worker that an abnormality has occurred.

Images