Data monitoring method for high-strength refractory brick forming process

Abstract

The invention relates to the technical field of refractory brick forming monitoring, in particular to a data monitoring method for a high-strength refractory brick forming process. The method comprises the following steps: segmenting a pressure sequence in a refractory brick forming process to obtain data segments; obtaining a fall value of the number of data in the horizontal coordinate of the PACF graph of the data segment, and screening to obtain the number of data to be analyzed; calculating the optimization rate of the number of the to-be-analyzed data to obtain a target lag order of the data segment; obtaining the trend change degree of a curve point according to the curvature of the curve point on the trend change curve of the data segment and the left and right adjacent curve points, and further obtaining the reference lag order of the data segment; respectively acquiring two average trend change rates of the last two data segments, and further calculating an average trend change rate difference coefficient to obtain a final lag order of the last data segment; and predicting pressure data at the next moment by using the final lag order and combining with an ARIMA algorithm. The refractory brick forming quality can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of refractory brick forming monitoring, and in particular to a data monitoring method for a high-strength refractory brick forming process. Background Art

[0002] The molding process for high-strength refractory bricks typically includes raw material preparation, molding, drying, and sintering. Molding is a critical step. During the molding process, the molding pressure affects the final properties of the refractory brick. Insufficient pressure can lead to loose bonding between particles and high porosity. Excessive pressure can cause delamination or mold wear. This results in low density, insufficient strength, or internal cracks in the final refractory brick, reducing the production yield.

[0003] Traditional methods often use machine pressing to produce refractory bricks. The pressure during molding is monitored in real time through pressure sensors. When the pressure is not within the optimal pressure range, the pressure is adjusted to ensure the molding effect of the refractory bricks. However, since the pressure adjustment has a delay, that is, the pressure adjustment is only performed when it is detected that the pressure data is not within the optimal pressure range, the adjustment is often affected by inaccurate pressure. Therefore, real-time data can be predicted. When problems with the predicted pressure occur, pressure adjustments can be made in advance to reduce the delay of pressure adjustment.

[0004] Traditional prediction methods generally use the ARIMA algorithm to calculate the lag order based on all existing data for prediction. The lag order is generally obtained through the PACF diagram. However, since the moisture content of the clay will change during the molding of refractory bricks, the change in moisture content will cause the friction of the clay to change, thereby causing the pressure to change. This change is relatively slow and staged. The staged change may cause multiple peaks with similar characteristics to appear in the PACF diagram, and the slowness may result in no obvious truncation in the PACF diagram. Therefore, it is difficult to obtain a suitable lag order. It is necessary to obtain an adaptive lag order based on the analysis of the data's own characteristics. Summary of the Invention

[0005] In order to solve the above technical problems, the purpose of the present invention is to provide a data monitoring method for the high-strength refractory brick forming process. The technical solutions adopted are as follows:

[0006] One embodiment of the present invention provides a data monitoring method for a high-strength refractory brick forming process, the method comprising:

[0007] Obtain pressure data during the refractory brick forming process to form a pressure sequence; use a sliding window to slide on the pressure sequence to segment the pressure sequence to obtain data segments;

[0008] According to the partial autocorrelation coefficient of each data number on the horizontal axis in the PACF diagram of the data segment, the drop value of each data number is obtained, and each data number is screened to obtain the number of data to be analyzed; the preferred rate of the number of data to be analyzed is calculated, and the number of data to be analyzed with the largest preferred rate is the target lag order of the data segment;

[0009] A trend change curve is obtained by fitting the difference sequence of the data segment; the trend change degree of a curve point on the trend change curve is obtained based on the curvature of the curve point and the curve points adjacent to it on the left and right; and the reference lag order of the data segment is obtained based on the curve point with the largest trend change degree.

[0010] Calculate the mean of the difference series of the last two data segments respectively to obtain two average trend change rates; calculate the average trend change rate difference coefficient based on the average trend change rates of the last two data segments; obtain the final lag order of the last data segment based on the average trend change rate difference coefficient, the target lag order of the last data segment and the reference lag order; use the final lag order combined with the ARIMA algorithm to predict the pressure data at the next moment.

[0011] Preferably, before using a sliding window to slide on the pressure sequence to segment the pressure sequence to obtain data segments, the method further includes:

[0012] Obtain the PACF diagram of the pressure sequence, where the horizontal axis represents the number of data and the vertical axis represents the partial autocorrelation coefficient; obtain the partial autocorrelation coefficient of each data number on the horizontal axis of the PACF diagram, arrange the absolute values ​​of the partial autocorrelation coefficients in ascending order, obtain the upper limit threshold of the absolute value of the partial autocorrelation coefficient through the box plot, and record the data points corresponding to the number of data with the absolute value of the partial autocorrelation coefficient greater than the upper limit threshold as the initial stage points; use the initial stage points to segment the pressure sequence to obtain the initial data segments; set a sliding window, and slide the sliding window within an initial data segment to obtain the fitting curve of the data points in each sliding window, calculate the similarity of the fitting curves of every two adjacent sliding windows, and record it as the fitting curve similarity; based on the similarity of the fitting curves corresponding to every two adjacent sliding windows in each initial data segment, use mathematical statistics to obtain the similarity threshold.

[0013] Preferably, using a sliding window to slide on the pressure sequence to segment the pressure sequence to obtain data segments includes:

[0014] A sliding window is used to slide on the pressure sequence, and the similarity of the fitting curves between the current sliding window and the previous sliding window is calculated. If the similarity of the fitting curves is greater than or equal to the similarity threshold, the two adjacent sliding windows are the same data segment. Otherwise, the current sliding window is the starting point of a new data segment. Taking the current sliding window as the starting point, the similarity of the fitting curves between the next sliding window and the current sliding window is calculated. If the similarity of the fitting curves is greater than or equal to the similarity threshold, the next sliding window and the current sliding window are the same data segment. Similarly, the pressure sequence is divided to obtain data segments of the pressure sequence. If the similarity of the fitting curves between the current sliding window and the previous sliding window is less than the similarity threshold, the similarity of the fitting curves between the next sliding window and the previous sliding window is calculated. If the similarity of the fitting curves is greater than or equal to the similarity threshold, the current sliding window is a suspected interference window. The data value in the current sliding window is replaced by the mean of the sliding windows adjacent to the current sliding window, and the current sliding window is added to the data segment to which the previous sliding window belongs.

[0015] Preferably, the drop value of each data number is obtained according to the partial autocorrelation coefficient of each data number on the horizontal axis in the PACF diagram of the data segment, and each data number is screened to obtain the data number to be analyzed, including:

[0016] The difference in partial autocorrelation coefficients between the previous data number and the next data number in every two adjacent data numbers in the data segment is the drop value of the previous data number, and the drop value of the last data number in the data segment is the partial autocorrelation coefficient corresponding to itself; the drop values ​​corresponding to each data number in the data segment are arranged from small to large, and the upper limit threshold of the drop value is obtained by using the box plot based on the arranged drop values, which is recorded as the drop threshold; the number of data with a drop value greater than or equal to the drop threshold is obtained, which is recorded as the number of data to be analyzed.

[0017] Preferably, calculating the optimal ratio of the number of data to be analyzed includes:

[0018] Obtain the difference between the preset value and the reciprocal of the value of the number of data to be analyzed, which is recorded as the first difference; perform weighted summation on the absolute value of the partial autocorrelation coefficient of the number of data to be analyzed and the first difference to obtain the optimization rate of the number of data to be analyzed.

[0019] Preferably, obtaining the trend change degree of a curve point on the trend change curve according to the curvature of the curve point and the left and right adjacent curve points includes:

[0020] Get the curve points adjacent to the left and right of a curve point, connect these three curve points to form a triangle, get the minimum circumscribed circle of the triangle, and calculate the curvature of the curve point based on the minimum circumscribed circle;

[0021] Obtain a first preset number of curve points adjacent to the left side of a curve point and the slopes between every two adjacent curve points in the curve point to form a left slope sequence; obtain a second preset number of curve points adjacent to the right side of a curve point and the slopes between every two adjacent curve points in the curve point to form a right slope sequence; calculate the average of the absolute values ​​of the differences between every two adjacent slopes in the left slope sequence, and record it as the left slope change. Similarly, use the right slope sequence to obtain the right slope change; calculate the average of the left slope change and the right slope change, and record it as the trend change isolation degree; calculate the difference between the preset value and the inverse of the curve point, and multiply it by the inverse of the trend change isolation degree of the curve point to obtain the trend change degree of the curve point.

[0022] Preferably, obtaining a reference lag order of a data segment based on a curve point with the greatest degree of trend change comprises:

[0023] The sum of the horizontal coordinate of the curve point with the largest trend change and the preset value is used as the reference lag order of the data segment.

[0024] Preferably, the average trend change rate difference coefficient is calculated based on the average trend change rates of the last two data segments, including:

[0025] Calculate the absolute value of the difference between the average trend change rate of the last data segment and the previous data segment and add it to the preset value to obtain the addition result; take the difference between the preset value and the reciprocal of the addition result to obtain the average trend change rate difference coefficient.

[0026] Preferably, obtaining the final lag order of the last data segment includes:

[0027] The target lag coefficient and reference lag coefficient of the last data segment are used as boundary values ​​to form the lag order range; the median of the lag order range is obtained; if the average trend change rate of the last data segment is less than the average trend change rate of the previous data segment, the difference between the median of the lag order range and the minimum value in the lag order range is calculated, recorded as the first change basis, the median is subtracted from the product of the average trend change rate difference coefficient and the first change basis and rounded up to obtain the final lag order of the last data segment; if the average trend change rate of the last data segment is greater than or equal to the average trend change rate of the previous data segment, the difference between the maximum value in the lag order range and the median of the lag order range is calculated, recorded as the second change basis, the median is added to the product of the average trend change rate difference coefficient and the second change basis and rounded down to obtain the final lag order of the last data segment.

[0028] The embodiments of the present invention have at least the following beneficial effects: the present application utilizes a sliding window to slide on the corresponding pressure sequence in the refractory brick forming process, segments the pressure sequence according to similarity, combines similar and continuous data in the window into one data segment, reduces the influence of the stage on the acquisition of the lag order, and facilitates obtaining a more appropriate lag order; further analyzes the partial autocorrelation coefficient of each data number on the horizontal axis of the PACF graph corresponding to the data segment, thereby obtaining a suitable target lag order, and then fits the difference sequence of the data segment to obtain a trend change curve, uses the characteristics of the curve points on the trend change curve to analyze the trend change degree of each curve point, obtains the reference lag order of the data segment based on the curve point with the largest trend change degree, and then obtains the average trend change rate difference coefficient according to the characteristics of the difference sequence of the last two data segments, and then obtains the final lag order of the last data segment. This scheme obtains an adaptive lag order according to the characteristics of the pressure data itself in the refractory brick forming process, thereby improving the accuracy of the predicted data. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0030] Figure 1 A method flow chart of a data monitoring method for a high-strength refractory brick forming process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0031] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail a data monitoring method for a high-strength refractory brick forming process according to the present invention, including its specific implementation, structure, features, and effectiveness. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0032] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0033] The following describes in detail a specific scheme of a data monitoring method for a high-strength refractory brick forming process provided by the present invention with reference to the accompanying drawings.

[0034] Example:

[0035] The main application scenario of the present invention is: when making high-strength refractory bricks, the molding pressure is monitored in real time during the molding process, and then the adaptive lag order is obtained according to the inherent characteristics of the pressure data during the molding process of the refractory bricks to perform real-time pressure data prediction, adjust the pressure size in advance, and avoid the impact of the delay in data adjustment.

[0036] See also Figure 1 , which shows a method flow chart of a data monitoring method for a high-strength refractory brick forming process provided by an embodiment of the present invention, the method comprising the following steps:

[0037] Step S1, obtaining pressure data during the refractory brick forming process to form a pressure sequence; using a sliding window to slide on the pressure sequence to segment the pressure sequence to obtain data segments.

[0038] The main purpose of this application is to obtain an adaptive hysteresis order based on the inherent characteristics of pressure data during the refractory brick forming process to predict real-time pressure data, adjust the pressure level in advance, and avoid the impact of data adjustment delays. Furthermore, a hydraulic brick press is used to press refractory bricks. The pressure sensor uses a strain gauge sensor installed on the piston rod of the hydraulic cylinder. During the pressing process, pressure data is collected. The reference value of the pressure data collection frequency is set to once every 2 seconds, and the implementer can adjust it according to actual conditions.

[0039] Refractory brick forming involves placing a processed material in a mold and pressing it through a brick press to form a green body of a certain shape and strength. This pressing slowly removes air from the clay, causing the clay particles to shift and rearrange to form a dense body. At this point, the moisture in the clay slowly evaporates over time, causing the friction between the clay particles to change and the pressure to gradually decay during transmission. Because this change is slow and phased, the phased changes may result in multiple peaks with similar characteristics in the PACF plot. This slowness may result in a lack of clear truncation in the PACF plot, making it difficult to determine an appropriate lag order. To reduce the impact of phases, the data series can be simply segmented based on the peaks in the PACF plot and the similarity of the data itself.

[0040] Furthermore, before segmentation, a PACF plot of the pressure series is obtained. The horizontal axis represents the number of data points, and the vertical axis represents the partial autocorrelation coefficient. The partial autocorrelation coefficient for each data point on the horizontal axis of the PACF plot is obtained. The absolute values ​​of the partial autocorrelation coefficients are arranged in ascending order, and the upper threshold of the absolute value of the partial autocorrelation coefficients is obtained using a boxplot. Because stage characteristics are reflected as obvious peaks in the PACF plot, affecting the judgment of lag order, stage division is necessary. Furthermore, since the data within a stage are mostly similar, the data can be segmented based on obvious stage characteristics to obtain a similarity threshold for subsequent data segmentation. The data points corresponding to the number of data points with an absolute value of the partial autocorrelation coefficient greater than the upper threshold are marked as initial stage points. For example, if the number of data points is 4, the corresponding partial autocorrelation coefficient is greater than the upper threshold, and the fourth data point is marked as the initial stage point. If the number of data points is 20, the corresponding partial autocorrelation coefficient is greater than the upper threshold, and the 20th data point is marked as the initial stage point. The pressure series is segmented using the initial stage points to obtain initial data segments.

[0041] After obtaining the initial data segment, set the sliding window to 1*10, with a sliding step equal to the window length. Starting from the first data point, the fitting curve of the data points within the window is calculated. Starting from the second sliding window, the similarity of the fitting curves of the current window and the previous window is calculated. In other words, the sliding window slides within an initial data segment, calculating the similarity of the fitting curves of every two adjacent sliding windows to obtain the fitting curve similarity. This allows us to obtain the fitting curve similarity corresponding to every two adjacent sliding windows in each initial data segment. Furthermore, based on the fitting curve similarity corresponding to every two adjacent sliding windows in each initial data segment, a similarity threshold XY is obtained using mathematical statistics. Multiple similarity values ​​are obtained within the same data segment, and similarity is not calculated between different initial data segments. This is because the similarity values ​​between different data segments may be small due to stage-specific similarity, so these similarity values ​​are not considered and are meaningless. When obtaining the similarity of the fitting curves, Euclidean distance or the DTW algorithm can be used for calculation. The size of the sliding window can be adjusted according to actual conditions.

[0042] Next, a sliding window is used to slide over the pressure sequence, and the similarity of the fitting curves between the current sliding window and the previous sliding window is calculated. If the similarity of the fitting curves is greater than or equal to the similarity threshold XY, the two adjacent sliding windows are the same data segment. Otherwise, the current sliding window is the starting point of a new data segment. Starting from the current sliding window, the similarity of the fitting curves between the next sliding window and the current sliding window is calculated. If the similarity of the fitting curves is greater than or equal to the similarity threshold XY, the next sliding window and the current sliding window are the same data segment. Similarly, the pressure sequence is segmented to obtain data segments of the pressure sequence. If the similarity of the fitting curves between the current sliding window and the previous sliding window is less than the similarity threshold XY, the similarity of the fitting curves between the next sliding window and the previous sliding window is calculated. If the similarity of the fitting curves is greater than or equal to the similarity threshold XY, the current sliding window is suspected to be an interference window. The data of the current sliding window is replaced using the mean method, that is, the data values ​​in the current sliding window are replaced with the mean of the sliding windows adjacent to the current sliding window, and the current sliding window is added to the data segment to which the previous sliding window belongs. This completes the segmentation of the pressure sequence and obtains data segments of the pressure sequence.

[0043] Step S2, obtain the drop value of each data number according to the partial autocorrelation coefficient of each data number on the horizontal axis in the PACF diagram of the data segment, and screen each data number to obtain the data number to be analyzed; calculate the preferred rate of the data number to be analyzed, and the data number to be analyzed with the largest preferred rate is the target lag order of the data segment.

[0044] When the traditional PACF diagram obtains the lag order, the number of data with the last obvious peak is selected as the lag order. However, since the water evaporates slowly during the refractory brick forming process, the pressure data also changes slowly, so there may be no obvious peak in the PACF diagram. Therefore, it is necessary to obtain the gap threshold based on the difference between the partial autocorrelation coefficient of different data numbers and the next data number in the data segment. Then, the optimization rate of different data numbers is calculated based on the partial autocorrelation coefficient and position relationship corresponding to the number of data greater than the gap threshold, and the target lag order is obtained according to the optimization rate.

[0045] Furthermore, the drop value of each data number is obtained according to the partial autocorrelation coefficient of each data number on the horizontal axis in the PACF diagram of the data segment. Specifically, the difference between the partial autocorrelation coefficients of the previous data number and the next data number in every two adjacent data numbers in the data segment is the drop value of the previous data number, and the drop value of the last data number in the data segment is the partial autocorrelation coefficient corresponding to itself.

[0046] Next, the drop values ​​corresponding to each data number in the data segment are arranged from small to large, and the upper limit threshold of the drop value is obtained based on the arranged drop values ​​using the box plot, which is recorded as the drop threshold; the number of data with a drop value greater than or equal to the drop threshold is obtained, which is recorded as the number of data to be analyzed.

[0047] Then, the number of data to be analyzed in the data segment is analyzed to obtain its corresponding preference rate, and the preference rate of the number of data to be analyzed is calculated. The number of data to be analyzed with the largest preference rate is the target lag order of the data segment.

[0048] Specifically, the difference between the preset value and the reciprocal of the value of the number of data to be analyzed is obtained, which is recorded as the first difference; the absolute value of the partial autocorrelation coefficient of the number of data to be analyzed and the first difference are weighted and summed to obtain the preferred ratio of the number of data to be analyzed, and the preset value is 1; the specific calculation formula is:

[0049]

[0050] Where α represents the preference ratio of the data to be analyzed, |pacf| represents the absolute value of the partial autocorrelation coefficient of the data to be analyzed, and d represents the position of the data to be analyzed, that is, the value of the data to be analyzed. The larger the partial autocorrelation coefficient of the data to be analyzed and the later the position, the more likely the data to be analyzed is a lag order. w1 and w2 represent weights. Since the value of the partial autocorrelation coefficient is more important, w1 = 0.7 and w2 = 0.3 can be set. The data to be analyzed with the highest preference ratio is then selected as the target lag order for the data segment.

[0051] Step S3, fitting the difference sequence of the data segment to obtain a trend change curve; obtaining the trend change degree of a curve point on the trend change curve based on the curvature of the curve point and the curve points adjacent to it on the left and right; and obtaining the reference lag order of the data segment based on the curve point with the largest trend change degree.

[0052] Since the lag order indicates that the current data is affected by the previous data, the partial autocorrelation coefficient of data with similar trends will be larger. If there is a trend change, the number of data before and after the change is more likely to be the lag order, so the trend change curve of the data segment can be obtained. If there is a significant trend change, the curvature of the trend curve at this point will be larger. However, since strain sensors are generally resistive sensors, the production environment of refractory bricks may have temperature changes, or electromagnetic interference may affect the sensor's collected data, so there may be interference data. Since the trend of interference data is relatively isolated, and the normally changing data is continuous, the degree of isolation of the trend change of each point on the trend change curve can be calculated. If the degree of isolation of the trend change is high, that is, the difference from the trend change of the surrounding data is large, then the credibility of the point is low, and it may be a noise point or an interference point. The degree of trend change of each point is obtained according to the curvature and the degree of isolation of the trend change. The point with the largest degree of trend change is found and recorded as the reference lag order.

[0053] For a data segment, the difference between every two adjacent pressure data is obtained to form a difference sequence; the difference sequence is used for fitting to obtain a trend change curve, and each point on the trend change curve is recorded as a curve point; the number of data represented by the points of the trend change curve needs to be +1, that is, the first point of the curve represents two original data points, the second point represents three original data points, and so on.

[0054] Furthermore, the curvature of each curve point is calculated. Specifically, the curve points adjacent to the left and right of a curve point are obtained, and these three curve points are connected to form a triangle. The minimum circumscribed circle of the triangle is obtained, and the curvature of the curve point is calculated based on the minimum circumscribed circle. When calculating the curvature, if a curve point has only one adjacent point, the curvature calculation is performed symmetrically using the adjacent point on the other side.

[0055] Obtain a first preset number of curve points adjacent to the left side of a curve point and the slopes between every two adjacent curve points in the curve point to form a left slope sequence; obtain a second preset number of curve points adjacent to the right side of a curve point and the slopes between every two adjacent curve points in the curve point to form a right slope sequence; calculate the average of the absolute values ​​of the differences between every two adjacent slopes in the left slope sequence, and record it as the left slope change. Similarly, use the right slope sequence to obtain the right slope change; calculate the average of the left slope change and the right slope change, and record it as the trend change isolation degree; calculate the difference between the preset value and the inverse of the curve point, and multiply it by the inverse of the trend change isolation degree of the curve point to obtain the trend change degree of the curve point.

[0056] The specific calculation model is:

[0057]

[0058] Among them, β represents the degree of trend change of a curve point, and ρ represents the curvature of the curve point. The larger the curvature, the more curved the curve of the curve point, that is, the more obvious the trend change here is.

[0059] n + Indicates the first preset number of curve points adjacent to the left of the selected curve point, where the first preset number n + Greater than or equal to 2, k i and k i+1 represents the i-th slope and the i+1-th slope in the left slope sequence, |k i -k i+1 | represents the difference between two adjacent slopes. If the difference is small, it means that the two adjacent slopes are similar, and the three points that constitute these two slopes have similar change trends; It represents the mean of the slope differences between this curve point and the adjacent curve point on the left. If this value is small, it means that there is a similar change trend between this curve point and the curve point on the left.

[0060] n - Indicates the first preset number of curve points adjacent to the right of the selected curve point, where the second preset number n - Greater than or equal to 2, k j and k j+1 represents the jth slope and j+1th slope in the slope sequence on the right, |k j -k j+1 | represents the difference between two adjacent slopes. If the difference is small, it means that the two adjacent slopes are similar, and the three points that constitute these two slopes have similar change trends; It represents the average of the slope differences between this curve point and the adjacent curve point on the right. If this value is small, it means that there is a similar change trend between this curve point and the curve point on the right.

[0061] It indicates the difference in trend changes between the curve point and the surrounding curve points, that is, the degree of isolation of the trend change. If the value is small, it means that the trend change of the curve point is similar to that of the surrounding curve points, and the probability that the trend change of the selected curve point is an isolated change is smaller. Indicates the credibility of the curve point. The less isolated the trend change, the greater the credibility of the curve point. If only one side of the selected point can be calculated, the result of the calculation on that side will prevail. This gives the degree of trend change for each curve point.

[0062] The greater the curvature and credibility of a curve point, the greater the degree of trend change at that curve point and the more reliable it is. This curve point is more likely to be the location where the original data trend changes, and the lag order at this time is most likely to be appropriate. Therefore, the abscissa of the curve point is added to the preset value to obtain the reference lag order for that data segment.

[0063] Step S4, respectively calculate the mean of the difference series of the last two data segments to obtain two average trend change rates; calculate the average trend change rate difference coefficient based on the average trend change rates of the last two data segments; obtain the final lag order of the last data segment based on the average trend change rate difference coefficient, the target lag order of the last data segment, and the reference lag order; use the final lag order combined with the ARIMA algorithm to predict the pressure data at the next moment.

[0064] Since the pressure data at the next moment is predicted, the lag order of the last data segment of the pressure data needs to be obtained. When the target lag coefficient and the reference lag coefficient of the last data segment are the same, the final lag order of the last data segment is one of the corresponding target lag coefficient and the reference lag coefficient.

[0065] If the target lag coefficient and reference lag coefficient for the last data segment differ, analyze the data in conjunction with the previous data segment. Calculate the average trend change rate and the average trend change rate difference coefficient for the last and previous data segments. If the average trend change rate for the last data segment is lower than that for the previous data segment, increase the lag order based on the long-term advantage of historical data. If the average trend change rate for the last data segment is greater than or equal to that for the previous data segment, reduce reliance on historical data and thus the lag order. This results in the final lag order for the last data segment.

[0066] Furthermore, the average values ​​of the difference series of the last two data segments are calculated to obtain the average trend change rate of the last data segment and the previous data segment.

[0067] Next, the average trend change rate difference coefficient is calculated based on the average trend change rate of the last two data segments. Specifically, the absolute value of the difference between the average trend change rate of the last data segment and the previous data segment is calculated and added to the preset value to obtain the addition result; the difference between the preset value and the reciprocal of the addition result is calculated to obtain the average trend change rate difference coefficient. Its specific calculation model is:

[0068]

[0069] Among them, γ represents the coefficient of difference between the average trend change rate of the last data segment and the previous data segment, N represents the number of differences in the difference sequence of the last data segment, and x l Indicates the size of the lth difference in the difference sequence of the last data segment, M indicates the number of differences in the difference sequence of the data segment before the last data segment, y l Indicates the size of the lth difference in the difference sequence of the previous data segment, Indicates the average trend change rate of the last data segment, Indicates the average trend change rate of the data segment before the last data segment. It indicates the difference in the data trend changes of two data segments. If the difference is larger, it means that the difference in the data trend changes of the two data segments is greater, and the difference coefficient is larger.

[0070] Furthermore, the final lag order of the last data segment is obtained based on the average trend change rate difference coefficient, the target lag order of the last data segment and the reference lag order. Specifically, the target lag coefficient and the reference lag coefficient of the last data segment are used as boundary values ​​to form a lag order range; the median of the lag order range is obtained; if the average trend change rate of the last data segment is less than the average trend change rate of the previous data segment, the difference between the median of the lag order range and the minimum value in the lag order range is calculated, recorded as the first change basis, the median is subtracted from the product of the average trend change rate difference coefficient and the first change basis and rounded up to obtain the final lag order of the last data segment; if the average trend change rate of the last data segment is greater than or equal to the average trend change rate of the previous data segment, the difference between the maximum value in the lag order range and the median of the lag order range is calculated, recorded as the second change basis, the median is added to the product of the average trend change rate difference coefficient and the second change basis and rounded down to obtain the final lag order of the last data segment.

[0071] The specific calculation model is:

[0072]

[0073] Where P represents the final lag order of the last data segment, mid(p) represents the median of the lag order range (can be a decimal), γ represents the coefficient of difference between the average trend change rate of the last data segment and the previous data segment, and p mid-min It represents the difference between the median of the lag order range and the minimum of the lag order range, which is also the first change basis, p max-mid It represents the difference between the maximum and median values ​​of the lag order range, which is also the second basic amount of change. Indicates the average trend change rate of the last data segment, Indicates the average trend change rate of the data segment before the last data segment. Indicates rounding up. Indicates rounding down.

[0074] It should be noted that if there is only one data segment in the extreme case when dividing the pressure series into data segments, after obtaining the lag order range, the lag orders can be taken in sequence within this range, and then the ARIMA algorithm can be used to predict the pressure data at the last moment of the data segment. Based on the accuracy of the prediction, the lag order with the highest accuracy is taken as the final lag order.

[0075] The final lag order is combined with the ARIMA algorithm to predict the pressure data at the next moment. Specifically, the ARIMA algorithm is used in combination with the final lag order of the last data segment for prediction, where the final lag order is the lag order obtained by this scheme, and the MA order and difference order are obtained through the ACF diagram and ADF test respectively.

[0076] After obtaining the next moment's pressure data, if the pressure data is not within the specified range, adjustments are made. The pressure required by the refractory brick press varies depending on product specifications and production processes, and the pressure range typically required is between 200 tons and 600 tons. This ensures the quality of the refractory bricks.

[0077] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0078] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0079] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A data monitoring method for a high-strength refractory brick forming process, characterized in that: The method includes: Obtain pressure data during the refractory brick forming process to form a pressure sequence; use a sliding window to slide on the pressure sequence to segment the pressure sequence to obtain data segments; According to the partial autocorrelation coefficient of each data number on the horizontal axis in the PACF diagram of the data segment, the drop value of each data number is obtained, and each data number is screened to obtain the number of data to be analyzed; the preferred rate of the number of data to be analyzed is calculated, and the number of data to be analyzed with the largest preferred rate is the target lag order of the data segment; A trend change curve is obtained by fitting the difference sequence of the data segment; the trend change degree of a curve point on the trend change curve is obtained based on the curvature of the curve point and the curve points adjacent to it on the left and right; and the reference lag order of the data segment is obtained based on the curve point with the largest trend change degree. Calculate the mean of the difference series of the last two data segments respectively to obtain two average trend change rates; calculate the average trend change rate difference coefficient based on the average trend change rates of the last two data segments; obtain the final lag order of the last data segment based on the average trend change rate difference coefficient, the target lag order of the last data segment and the reference lag order; use the final lag order combined with the ARIMA algorithm to predict the pressure data at the next moment.

2. The data monitoring method for a high-strength refractory brick forming process according to claim 1, characterized in that: Before using the sliding window to slide on the pressure sequence to segment the pressure sequence to obtain data segments, the method further includes: Obtain the PACF diagram of the pressure sequence, where the horizontal axis represents the number of data and the vertical axis represents the partial autocorrelation coefficient; obtain the partial autocorrelation coefficient of each data number on the horizontal axis of the PACF diagram, arrange the absolute values ​​of the partial autocorrelation coefficients in ascending order, obtain the upper limit threshold of the absolute value of the partial autocorrelation coefficient through the box plot, and record the data points corresponding to the number of data with the absolute value of the partial autocorrelation coefficient greater than the upper limit threshold as the initial stage points; use the initial stage points to segment the pressure sequence to obtain the initial data segments; set a sliding window, and slide the sliding window within an initial data segment to obtain the fitting curve of the data points in each sliding window, calculate the similarity of the fitting curves of every two adjacent sliding windows, and record it as the fitting curve similarity; based on the similarity of the fitting curves corresponding to every two adjacent sliding windows in each initial data segment, use mathematical statistics to obtain the similarity threshold.

3. The data monitoring method for a high-strength refractory brick forming process according to claim 1 is characterized in that: The step of using a sliding window to slide on the pressure sequence to segment the pressure sequence to obtain data segments includes: A sliding window is used to slide on the pressure sequence, and the similarity of the fitting curves between the current sliding window and the previous sliding window is calculated. If the similarity of the fitting curves is greater than or equal to the similarity threshold, the two adjacent sliding windows are the same data segment. Otherwise, the current sliding window is the starting point of a new data segment. Taking the current sliding window as the starting point, the similarity of the fitting curves between the next sliding window and the current sliding window is calculated. If the similarity of the fitting curves is greater than or equal to the similarity threshold, the next sliding window and the current sliding window are the same data segment. Similarly, the pressure sequence is divided to obtain data segments of the pressure sequence. If the similarity of the fitting curves between the current sliding window and the previous sliding window is less than the similarity threshold, the similarity of the fitting curves between the next sliding window and the previous sliding window is calculated. If the similarity of the fitting curves is greater than or equal to the similarity threshold, the current sliding window is a suspected interference window. The data value in the current sliding window is replaced by the mean of the sliding windows adjacent to the current sliding window, and the current sliding window is added to the data segment to which the previous sliding window belongs.

4. The data monitoring method for a high-strength refractory brick forming process according to claim 1, characterized in that: The method of obtaining the drop value of each data number according to the partial autocorrelation coefficient of each data number on the horizontal axis in the PACF diagram of the data segment and screening each data number to obtain the data number to be analyzed includes: The difference in partial autocorrelation coefficients between the previous data number and the next data number in every two adjacent data numbers in the data segment is the drop value of the previous data number, and the drop value of the last data number in the data segment is the partial autocorrelation coefficient corresponding to itself; the drop values ​​corresponding to each data number in the data segment are arranged from small to large, and the upper limit threshold of the drop value is obtained by using the box plot based on the arranged drop values, which is recorded as the drop threshold; the number of data with a drop value greater than or equal to the drop threshold is obtained, which is recorded as the number of data to be analyzed.

5. The data monitoring method for a high-strength refractory brick forming process according to claim 1 is characterized in that: The calculating of the optimal ratio of the number of data to be analyzed includes: Obtain the difference between the preset value and the reciprocal of the value of the number of data to be analyzed, which is recorded as the first difference; perform weighted summation on the absolute value of the partial autocorrelation coefficient of the number of data to be analyzed and the first difference to obtain the optimization rate of the number of data to be analyzed.

6. The data monitoring method for a high-strength refractory brick forming process according to claim 1, characterized in that: The step of obtaining the trend change degree of a curve point on the trend change curve based on the curvature of the curve point and the curve points adjacent to the left and right of the curve point comprises: Get the curve points adjacent to the left and right of a curve point, connect these three curve points to form a triangle, get the minimum circumscribed circle of the triangle, and calculate the curvature of the curve point based on the minimum circumscribed circle; Obtain a first preset number of curve points adjacent to the left side of a curve point and the slopes between every two adjacent curve points in the curve point to form a left slope sequence; obtain a second preset number of curve points adjacent to the right side of a curve point and the slopes between every two adjacent curve points in the curve point to form a right slope sequence; calculate the average of the absolute values ​​of the differences between every two adjacent slopes in the left slope sequence, and record it as the left slope change. Similarly, use the right slope sequence to obtain the right slope change; calculate the average of the left slope change and the right slope change, and record it as the trend change isolation degree; calculate the difference between the preset value and the inverse of the curve point, and multiply it by the inverse of the trend change isolation degree of the curve point to obtain the trend change degree of the curve point.

7. The data monitoring method for a high-strength refractory brick forming process according to claim 1 is characterized in that: The step of obtaining a reference lag order of a data segment based on a curve point with the largest trend change degree includes: The sum of the horizontal coordinate of the curve point with the largest trend change and the preset value is used as the reference lag order of the data segment.

8. The data monitoring method for a high-strength refractory brick forming process according to claim 1, characterized in that: Calculating the average trend change rate difference coefficient based on the average trend change rates of the last two data segments includes: Calculate the absolute value of the difference between the average trend change rate of the last data segment and the previous data segment and add it to the preset value to obtain the addition result; take the difference between the preset value and the reciprocal of the addition result to obtain the average trend change rate difference coefficient.

9. The data monitoring method for a high-strength refractory brick forming process according to claim 1, characterized in that: The method of obtaining the final lag order of the last data segment includes: The target lag coefficient and reference lag coefficient of the last data segment are used as boundary values ​​to form the lag order range; the median of the lag order range is obtained; if the average trend change rate of the last data segment is less than the average trend change rate of the previous data segment, the difference between the median of the lag order range and the minimum value in the lag order range is calculated, recorded as the first change basis, the median is subtracted from the product of the average trend change rate difference coefficient and the first change basis and rounded up to obtain the final lag order of the last data segment; if the average trend change rate of the last data segment is greater than or equal to the average trend change rate of the previous data segment, the difference between the maximum value in the lag order range and the median of the lag order range is calculated, recorded as the second change basis, the median is added to the product of the average trend change rate difference coefficient and the second change basis and rounded down to obtain the final lag order of the last data segment.

Images