Material blockage early warning method based on calculation of material flow value difference degree of material inlet and outlet
By installing high-speed cameras and line laser generators at the inlet and outlet, and combining them with a dynamic time warping algorithm to analyze the differences in material flow curves, the problems of lag and high false alarm rate in traditional blockage detection methods are solved, achieving early warning and efficient blockage detection.
Patent Information
- Application Number
- CN202510943403.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-09
- Publication Date
- 2025-10-17
AI Technical Summary
Traditional blockage detection methods have problems such as detection lag, high false alarm rate, and poor adaptability, making it difficult to prevent blockage failures in a complex production environment.
By calculating the difference in material flow values at the inlet and outlet, a high-speed camera and a line laser generator are used to obtain material flow data. The dynamic time warping algorithm is used to analyze the trend and numerical difference of the flow curve to achieve early warning of blockage failure.
It achieves early warning of material blockage failure, reduces the false alarm rate by more than 80%, and controls the average delay time to less than 1 second, thus improving the operating efficiency and safety of the material conveying system.
Smart Images
Figure CN120806250A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of industrial automation control, material conveying monitoring and intelligent early warning, and in particular to a material blockage early warning method based on material flow value difference calculation of inlet and outlet ports, which is suitable for material conveying systems in industries such as mining, metallurgy, chemical industry, grain processing and cement production, and can be applied to continuous material conveying systems such as belt conveyors and screw conveyors for real-time monitoring of material conveying state and timely early warning of material blockage failure. BACKGROUND
[0002] In industrial production processes, stable operation of material conveying systems is crucial to ensure production efficiency. Traditional material blockage detection methods mainly rely on mechanical flashboard switches or single-point sensors for monitoring. These methods have problems such as detection lag, high false alarm rate and poor adaptability. Mechanical detection can only trigger an alarm after material blockage occurs, while single-point sensors are easily affected by factors such as uneven material distribution, equipment vibration and electromagnetic interference.
[0003] Material blockage problems in the material conveying process are common types of failure in industrial production, particularly in the mining, metallurgy and chemical industries. Traditional material blockage detection methods mainly rely on single-point sensor monitoring, such as pressure sensors, vibration sensors or acoustic sensors, which set fixed thresholds for reporting. These methods have certain effects in simple working conditions, but have obvious limitations in complex production environments.
[0004] Single-point sensors are easily affected by factors such as uneven material distribution, equipment vibration and electromagnetic interference, resulting in a high false alarm rate. Traditional methods can only trigger an alarm after material blockage occurs, making it difficult to prevent in a timely manner. Fixed thresholds are difficult to adapt to different material characteristics (such as viscosity and particle size) and changes in working conditions.
[0005] The present application proposes a solution to the above problems. SUMMARY
[0006] The present application proposes a material blockage early warning method based on material flow value difference calculation of inlet and outlet ports. To address the problems of detection lag, high false alarm rate and poor adaptability in existing technologies, the method calculates the time offset of inlet and outlet ports and aligns the flow curves on the time axis, then analyzes the trend difference of inlet and outlet port flow curves using dynamic time warping algorithm, and analyzes the numerical difference degree characteristics through the numerical characteristics of flow curves, which can realize early warning of material blockage failure.
[0007] The present application adopts the following technical solutions.
[0008] A kind of blockage early warning method based on the difference degree of material flow value of inlet and outlet, the method is based on the matching degree of material flow value of inlet and outlet, first by the time offset calculation of inlet and outlet, and the alignment of flow curve on time axis, then the trend difference degree of inlet and outlet flow curve is analyzed by dynamic time warping algorithm, the numerical difference degree characteristics are analyzed by the numerical characteristics of flow curve, to realize the early warning of blockage fault.
[0009] The inlet is fed with material by a conveyor belt, and the outlet is fed with material by a conveyor belt, the material fed by the inlet falls to the outlet through the drop path connected with the drop port, and the blockage early warning method is provided with terminal equipment including high-speed camera and line laser generator at the conveyor belt of inlet and outlet.
[0010] The system is as shown in Figure 1 The high-speed camera is used to provide camera picture, and the line laser generator is used as external hardware to provide the surface profile of object in camera picture.
[0011] The blockage early warning method includes the following steps.
[0012] Step one, data statistics and pretreatment.
[0013] Step two, time offset calculation of inlet and outlet material flow curve.
[0014] Step three, trend matching degree calculation of inlet and outlet material flow curve.
[0015] Step four, numerical difference degree calculation of inlet and outlet material flow curve.
[0016] Step five, trend difference degree and numerical difference degree fusion
[0017] Step one is to filter the data, specifically as shown in Figure 2 The upper and lower two colored data lines in the figure respectively represent the inlet and outlet material flow value, with unit of cubic meter / hour, when detecting the material flow value on the belt of inlet and outlet, due to the influence of environment and working condition, the flow value may have certain Gaussian noise and occasional pulse noise, the data fluctuates in time, in order to avoid the influence of occasional abnormal flow value on subsequent processing, the data needs to be filtered,
[0018] Supposing the flow value data of single material port in a time range is f (t), m≤t≤n, the camera of terminal equipment provides about 30 frames of images per second, so the flow value data provided is also about 30 (per second), select every 500 milliseconds (about 15 data points) to do once slice median filtering to filter out the peak noise appearing in a short time, and the filtering calculation expression is as follows:
[0019] y(t) = median(f(t),...,f(t+500))
[0020] The results before and after filtering are shown in Figure 3 The left half of the image is the original material flow curve detected, and the right half of the image is the material flow curve after filtering. It can be seen that the filtering effect is obvious.
[0021] In step two, the material at the feeding port needs to pass through the discharging port to reach the discharging port, so that the appearance time of a certain material at the feeding and discharging ports is not equal, resulting in a time axis offset of the material flow curves at the feeding and discharging ports on the time axis;
[0022] Take the material flow value at the feeding port detection device position as the reference timestamp t0, and the time when the material reaches the discharging port detection device position is t1. The time offset of the material flow value at the feeding port and the material flow value at the discharging port is Δt = t1-t0. Let x be the timestamp, and let the flow data of the material at the feeding port in the time t0-t1 be f in (x), f out (x), and let the random error affected by other factors be ε, which has the following relationship:
[0023] f out (x) = f in (x+Δt) + ε
[0024] For the calculation of this time offset Δt, since the value of t1 cannot be directly observed in practice;
[0025] By using a time window with a size of a reasonable value Size preset by a person, Δt is searched in the upper and lower interval ranges of Size. When the following relationship is satisfied, it indicates that the flow in this window calculation is very small and cannot be used for time offset calculation. This special case needs to be excluded.
[0026]
[0027] Then the window calculates the difference value of the feeding and discharging port flow curves on the flow data Size times, and the step size is 500 milliseconds each time. The calculation expression is as follows:
[0028]
[0029] The D(i) is normalized to obtain norm(D(i)), so as to reduce the degree of influence of the trough (as shown in Figure 4 ) by the scale;
[0030] There are three cases for D(i): there is one obvious trough, there is no trough, and there are multiple obvious troughs; since D(i) has been normalized, a preset threshold can be used to determine whether a trough is obvious;
[0031] The definition of a trough is given as follows: let p be a point in the correlation coefficient curve calculated in a window, and let p i i i . When p i satisfies the following conditions, p i is a trough calculated in the current window:
[0032] p i-1 ≤ p i ≤ p i+1
[0033] A set of troughs P(x) = {p (1) , p (2) ,..., p (n)} can be obtained, where 0 ≤ x ≤ n ≤ Size. To extract obvious troughs from P(x), the following cases are excluded: when p is calculated, it indicates that the correlation coefficient curve forms an abrupt slope rather than a trough; when two troughs p are taken from P(x) continuously, if p indicates that there are multiple continuous troughs in the correlation coefficient curve,
[0034] Since a reasonable Δt will only have one in one window calculation, multiple continuous troughs will affect the judgment, and this calculation should not be considered.
[0035] Threshold1 and Threshold2 are threshold values obtained from a large number of tests on multiple inlet and outlet ports. When Threshold1 = Threshold2 = 0.4, the time offset calculation value is reasonable.
[0036] When the above two special cases are excluded, the vertical coordinate of the smallest trough point in the trough set P(x) is selected as the best time offset calculated at this point.
[0037] For the case of no trough, for the last two cases, the calculation of this time offset does not cover the last calculation to ensure the stability of the offset. The index value corresponding to the result of only one obvious trough is selected as the final calculated time offset Δt, and D(i) is the correlation coefficient function of a small time window in the material flow value curve, which changes with the characteristics of the data in the time window. If the Δt calculated by the continuous two time offset calculations exceeds several time units, the last time offset calculation result is not covered. Thus, the time offset of the material flow curve of the inlet and outlet is obtained.
[0038] The results before and after the time alignment of the inlet and outlet flow curves are shown in Figure 5 . The left half of the image is the filtered material flow curve, and the right half of the image is the material flow curve obtained after time alignment. It can be seen that the time alignment effect is good.
[0039] In step three, on the basis of filtering and time aligning the inlet and outlet material flow curves, in order to meet the demand of rapid blockage early warning, a queue cache of 5 seconds of data points will be used. This queue meets the principle of first-in first-out, ensuring real-time updating of data points. The update frequency is equal to the filtering time consumption, which fluctuates around 0.5 seconds. Thus, about 10 data points form a small sample data point group. Similar shape features exist in similar flow curves. Since the sample points are small enough, the speed of calculating the trend feature difference is fast enough. At the same time, under the premise of meeting the speed, a high-precision method for calculating the flow curve difference is also needed. Here, the dynamic time warping algorithm shown in Figure 6 is used to calculate the trend difference of the two flow curves. The dynamic time warping (DTW) algorithm is abbreviated as DTW algorithm, which calculates the warping path distance of two time sequences to measure the similarity between the two time sequences.
[0040] The specific method of applying the DTW algorithm here is as follows:
[0041] Let the flow rates of the inlet and outlet in the current time window (n = m = data points = 10) be X = {x1, x2, …, x n} and Y = {y1, y2, …, y m} respectively, the distance between any point x i on the X sequence and any point y i on the Y sequence is defined as d(i, j) = |x i -y j | p , where || pis the norm, when p = 2 it is the 2-norm, then d(i,j) = |x i -y j | 2 , which is the Euclidean distance;
[0042] Create an n×m cumulative distance matrix, defined as follows:
[0043]
[0044] Where D(1,1)=d(1,1), As shown in the definition of the cumulative distance matrix above: the cumulative distance D(i, j) is the sum of the distance d(i, j) of the current element and the cumulative distance of the smallest adjacent element that can reach the element;
[0045] Define a regular path W = {w1,w2,...,w K}, where w k =d(i,j) k ,max(n,m)≤K≤n+m-1,
[0046] K is the total length of the regular path; the regular path satisfies the following constraints:
[0047] Boundary conditions: w1=D(1,1),w K =D(n,m), indicating that the first element of the regular path W must be D(1,1) and the last element must be D(n,m), that is, the selected regular path must start at D(1,1) and end at D(n,m);
[0048] Monotonicity: Note w k =D(i k ,j k ),w k+1 =D(i k+1 ,j k+1 ), then i k ≤i k+1 ,j k ≤j k+1 .
[0049] Continuity: Remember w k =D(i k ,j k ),w k+1 =D(i k+1 ,j k+1 ), then i k+1 -i k ≤1,j k+1 -j k ≤1;
[0050] The calculation formula of the DTW distance, i.e., the minimum path length calculated by the DTW, is as follows:
[0051]
[0052] wherein DTW(X, Y) is the trend difference degree of the two flow curves finally calculated.
[0053] When the trends of the two flow curves are similar, the minimum path length value calculated by the dynamic time warping is generally small, and in order to reduce the influence of the data point dimension on the trend difference degree of the flow curves calculated by the dynamic time warping, the two flow curves are normalized in the patent, and then the minimum path length value is calculated by the dynamic time warping algorithm, which is denoted as Trend value . As shown in Figure 7 , the dynamic time warping algorithm is used to calculate the flow curves at the inlet and outlet and the simulated flow curve at the outlet with blockage, and it can be seen that when there is no blockage, the minimum path length of the flow curves at the inlet and outlet is obviously smaller than the minimum path length calculated when the blockage occurs.
[0054] In step four, due to the volatility of the flow curves on the time axis, only the trend difference degree of the flow curves is not enough to provide a reliable blockage early warning, and therefore the difference degree of the flow curves in the numerical characteristics needs to be calculated.
[0055] As shown in Figure 8 , the total difference between the flow curves at the inlet and outlet is directly used as the numerical difference degree, which is denoted as Numberical value , which can fully consider the influence of different material flow conditions, and from Figure 8 it can be seen that when there is no blockage, the value of the numerical difference degree is obviously smaller than the numerical matching degree calculated when the blockage occurs.
[0056] In step five, in order to reduce the occasional false detection of the inlet under the condition of small flow (as shown in Figure 9 ), the trend difference degree caused by frequent data fluctuations is large, and under the condition of large flow, the numerical difference degree caused by large numerical dimension is large, and the multiplication fusion method of the trend difference degree and the numerical difference degree makes it adapt to the variable flow in the production environment, and thus the final difference degree measurement index is as follows:
[0057] Indicator = Trend value *Numerical value .
[0058] When Indicator ≤ Blockage thresholdWhen the value is less than the threshold value, it indicates that no blockage occurs, otherwise, it triggers a blockage warning; wherein Blockage threshold The threshold value is obtained through a large number of tests of a plurality of pairs of inlet and outlet ports.
[0059] The concave-convex characteristics of the laser line of the linear laser generator on the material are not affected by the shape of the material, so that the conveyor belt material of any shape, including granular bodies and solid shapes, can be supported.
[0060] The intensity of the linear laser generator is greater than the threshold value, so that the reflection of common materials such as ore, coal, and sand transported by the conveyor belt can be ignored.
[0061] The present application relates to a blockage early warning method based on the difference degree calculation of the material flow value of the inlet and outlet ports, belonging to the technical field of industrial automation control. The method is mainly applied to continuous material conveying systems such as belt conveyors and screw conveyors, for real-time monitoring of material conveying state and timely early warning of blockage failure.
[0062] The present application realizes early warning of blockage failure by innovatively using the method of calculating the difference degree of the material flow value of the inlet and outlet ports. The method first installs high-precision flow detection devices at the inlet and outlet of the conveying equipment, and real-time collects material flow data. By calculating the time offset of the inlet and outlet flow curves and analyzing the trend difference degree characteristics of the inlet and outlet flow curves through dynamic time warping, combining the numerical characteristics of the two flow curves, fusing the trend difference degree characteristics and the numerical difference degree characteristics, and calculating the difference degree index of the two curves, when the difference degree is higher than the preset threshold value, the system automatically triggers the hierarchical early warning mechanism, and takes different processing measures according to the severity.
[0063] The core innovation of the present application lies in establishing a blockage judgment model based on the difference degree of the flow curve. The model fully considers the time delay characteristics in the material conveying process, and realizes the accurate difference calculation of the trend similarity of the flow curve through the dynamic time warping algorithm. Compared with the traditional method, the present application has the advantages of high detection precision, fast response speed, and strong adaptability. The practical application shows that the method can reduce the false alarm rate of blockage detection by more than 80%, and the average delay time from blockage to alarm is controlled within 1 second.
[0064] In the embodiment of the present application, the system adopts modular design, including a data acquisition module, a data processing module and an early warning output module. The data acquisition module is responsible for real-time acquisition of flow data of the inlet and outlet; the data processing module completes data preprocessing, difference calculation and fault judgment; the early warning output module triggers corresponding alarm and control instructions according to the judgment result; the present application is particularly suitable for production scenes such as mines, cement and grain processing which have the risk of material blockage. In these applications, the system shows good stability and reliability, which can effectively prevent equipment damage and production interruption caused by blockage. Through early warning and automatic processing, the operation efficiency and safety of the material conveying system are significantly improved. BRIEF DESCRIPTION OF DRAWINGS
[0065] The present application will be further described in detail below in combination with the drawings and specific embodiments:
[0066] Figure 1 is a system principle schematic diagram of the present application; Figure 1
[0067] Figure 2 is an inlet and outlet flow statistics schematic diagram of the inlet and outlet of the feeding port in the embodiment; Figure 2
[0068] Figure 3 is a filter before and after effect schematic diagram of step one in the embodiment; Figure 3
[0069] Figure 4 is a size = 20, different time offset correlation coefficient calculation result schematic diagram of step two in the embodiment; Figure 4
[0070] Figure 5 is a time alignment before and after effect schematic diagram of step two in the embodiment; Figure 5
[0071] Figure 6 is a dynamic time warping algorithm schematic diagram of step three in the embodiment; Figure 6
[0072] Figure 7 is a dynamic time warping algorithm trend difference calculation schematic diagram of step three in the embodiment; Figure 7
[0073] Figure 8 is a difference method calculation numerical difference degree schematic diagram of step four in the embodiment; Figure 8
[0074] Figure 9 is a small flow accidental false detection schematic diagram of step five in the embodiment; Figure 9
[0075] Figure 10 is an algorithm flow schematic diagram of the present application. Figure 10 DETAILED DESCRIPTION
[0076] As shown in the figure, a material blockage warning method based on the calculation of the difference in material flow values at the inlet and outlet is described. The method is based on the calculation of the matching degree of the material flow values at the inlet and outlet. First, the time offset of the inlet and outlet is calculated and the flow curves are aligned on the time axis. Then, a dynamic time warping algorithm is used to analyze the trend difference of the flow curves of the inlet and outlet. The numerical characteristics of the flow curves are used to analyze the numerical difference characteristics to achieve early warning of material blockage failure.
[0077] The feed port uses a conveyor belt to feed materials, and the discharge port uses a conveyor belt to feed materials. The materials fed into the feed port fall to the discharge port through a drop path connected to the drop port. The blockage warning method is provided with terminal equipment including a high-speed camera and a line laser generator at the conveyor belts of the feed port and the discharge port;
[0078] System such as Figure 1 As shown, the high-speed camera is used to provide camera images and is a peripheral input source device for the real-time material flow detection equipment. The line laser generator is used as external hardware to provide the surface contour of the object in the camera image.
[0079] The material blocking early warning method comprises the following steps:
[0080] Step 1: Data statistics and preprocessing;
[0081] Step 2: Calculate the time offset of the material flow curve of the inlet and outlet;
[0082] Step 3: Calculate the trend matching degree of the material flow curve of the inlet and outlet;
[0083] Step 4: Calculate the numerical difference of the material flow curves at the inlet and outlet;
[0084] Step 5: Fusion of trend difference and numerical difference
[0085] Step 1 is to perform data filtering on the data, specifically: Figure 2 As shown in the figure, the two colored data lines above and below represent the material flow rate values of the inlet and outlet respectively, in cubic meters per hour. When detecting the material flow value on the inlet and outlet belts, due to the influence of the environment and working conditions, the flow value will have a certain amount of Gaussian noise and occasional pulse noise, and the data will fluctuate in time. In order to avoid the occasional abnormal flow value affecting subsequent processing, data filtering is required.
[0086] Assume that the flow rate data of a single material inlet within a period of time is f(t), m≤t≤n. The camera of the terminal device provides about 30 frames of image per second, so the flow rate data provided is also about 30 (frames / second). Slice median filtering is performed every 500 milliseconds (about 15 data points) to filter out spike noise that occurs in a short period of time. The filter calculation expression is as follows:
[0087] y(t) = median(f(t),...,f(t+500))
[0088] The results before and after filtering are shown in Figure 3 The left half of the image is the original material flow curve detected, and the right half of the image is the material flow curve after filtering. It can be seen that the filtering effect is obvious.
[0089] In step two, the material at the feeding port needs to pass through the discharging port to reach the discharging port, so that the appearance time of a certain material at the feeding and discharging ports is not equal, resulting in a time axis offset of the material flow curves at the feeding and discharging ports on the time axis.
[0090] Take the material flow value at the feeding port detection device position as the reference timestamp t0, and the time when the material reaches the discharging port detection device position is t1. The time offset of the material flow value at the feeding port and the material flow value at the discharging port is Δt = t1-t0. Let x be the timestamp, and let the flow data of the material at the feeding port within t0-t1 be f in (x), f out (x), and let the random error affected by other factors be ε, which has the following relationship:
[0091] f out (x) = f in (x+Δt) + ε
[0092] For the calculation of this time offset Δt, the value of t1 cannot be directly observed in practice.
[0093] By using a time window with a size of a reasonable value Size preset by a person, Δt is found within the upper and lower interval ranges of Size. When the following relationship is satisfied, it indicates that the flow in this window calculation is very small and cannot be used for time offset calculation. This special case needs to be excluded.
[0094]
[0095] Then the window calculates the difference value of the flow data of the feeding and discharging ports for Size times, with a step size of 500 milliseconds each time. The calculation expression is as follows:
[0096]
[0097] Normalize D(i) to obtain norm(D(i)) to reduce the degree of influence of the trough (as shown in Figure 4 ) by the scale;
[0098] There are three possible cases for D(i): one obvious trough, no trough, or multiple obvious troughs. Since D(i) has been normalized, a preset threshold can be used to determine whether the trough is obvious.
[0099] Here we propose the definition of trough: there is a point in the correlation coefficient curve calculated under a certain window, which is recorded as p i =(t i ,c i ). For p i When the following conditions are met, record p i The trough calculated for the current window;
[0100] p i-1 ≤p i ≤p i+1
[0101] From this we can get a trough set P(x) = {p (1) ,p (2) ,…,p (n)}, 0≤x≤n≤Size;
[0102] In order to extract a clear trough from P(x), the following situations are excluded: First, when the calculated When , it indicates that the current correlation coefficient curve forms a steep slope rather than a trough; secondly, two trough points in the trough set P(x) are continuously taken and recorded as like This indicates that the correlation coefficient curve has multiple continuous troughs.
[0103] Because there is only one reasonable Δt in one window calculation, multiple consecutive troughs will affect the judgment and should not be considered in this calculation.
[0104] Threshold1 and Threshold2 are thresholds obtained through a large number of tests at multiple pairs of inlets and outlets. It can be found that when Threshold1 = Threshold2 and the value of time offset calculation is 0.4, it is more reasonable.
[0105] When the above two special cases are excluded, the ordinate of the smallest trough point in the trough set P(x) is selected as the optimal time offset calculated here.
[0106] For the case of no wave trough, for the last two cases, the calculation of this time offset does not cover the last calculation to ensure the stability of the offset. The index value corresponding to the result of only one obvious trough is selected as the final calculated time offset Δt, and D(i) is the correlation coefficient function of a small time window in the material flow value curve, which changes with the data characteristics of the time window. If the Δt calculated by the continuous two time offset calculations exceeds several time units, the last time offset calculation result is not covered. Thus, the time offset of the material flow curve of the inlet and outlet is obtained.
[0107] The results before and after the time alignment of the inlet and outlet flow curves are shown in Figure 5 . The left half of the image is the filtered material flow curve, and the right half of the image is the material flow curve obtained after time alignment. It can be seen that the time alignment effect is good.
[0108] In step three, on the basis of filtering and time aligning the inlet and outlet material flow curves, in order to meet the demand of rapid blockage early warning, a queue cache of 5 seconds of data points will be used. This queue meets the principle of first-in first-out, ensuring real-time updating of data points. The update frequency is equal to the filtering time consumption, which fluctuates around 0.5 seconds. Thus, about 10 data points form a small sample data point group. Similar shape features exist in similar flow curves. Since the sample points are small enough, the speed of calculating the trend feature difference is fast enough. At the same time, under the premise of meeting the speed, a high-precision method for calculating the flow curve difference is also needed. Here, the dynamic time warping algorithm shown in Figure 6 is used to calculate the trend difference of the two flow curves. The dynamic time warping (DTW) algorithm is abbreviated as DTW algorithm, which calculates the warping path distance of two time sequences to measure the similarity between the two time sequences.
[0109] The specific method of applying the DTW algorithm here is as follows:
[0110] Let the flow rates of the inlet and outlet in the current time window (n = m = data points = 10) be X = {x1, x2, …, x n} and Y = {y1, y2, …, y m} respectively, the distance between any point x i on the X sequence and any point y i on the Y sequence is defined as d(i, j) = |x i -y j | p , where || pis the norm, when p = 2 it is the 2-norm, then d(i,j) = |x i -y j | 2 , which is the Euclidean distance;
[0111] Create an n×m cumulative distance matrix, defined as follows:
[0112]
[0113] Where D(1,1)=d(1,1), As shown in the definition of the cumulative distance matrix above: the cumulative distance D(i, j) is the sum of the distance d(i, j) of the current element and the cumulative distance of the smallest adjacent element that can reach the element;
[0114] Define a regular path W = {w1,w2,...,w K}, where w k =d(i,j) k ,max(n,m)≤K≤n+m-1,
[0115] K is the total length of the regular path; the regular path satisfies the following constraints:
[0116] Boundary conditions: w1=D(1,1),w K =D(n,m), indicating that the first element of the regular path W must be D(1,1) and the last element must be D(n,m), that is, the selected regular path must start at D(1,1) and end at D(n,m);
[0117] Monotonicity: Note w k =D(i k ,j k ),w k+1 =D(i k+1 ,j k+1 ), then i k ≤i k+1 ,j k ≤j k+1 .
[0118] Continuity: Remember w k =D(i k ,j k ),w k+1 =D(i k+1 ,j k+1 ), then i k+1 -i k ≤1,j k+1 -j k ≤1;
[0119] The calculation formula of the DTW distance, i.e., the minimum path length calculated by the DTW, is as follows:
[0120]
[0121] wherein DTW(X, Y) is the trend difference degree of the two flow curves finally calculated.
[0122] When the trends of the two flow curves are similar, the minimum path length value calculated by the dynamic time warping is generally small, and in order to reduce the influence of the data point dimension on the trend difference degree of the flow curves calculated by the dynamic time warping, the two flow curves are normalized in the patent, and then the minimum path length value is calculated by the dynamic time warping algorithm, which is denoted as Trend value . As shown in Figure 7 , the dynamic time warping algorithm is used to calculate the flow curves at the inlet and outlet and the simulated flow curve at the outlet with blockage, and it can be seen that when there is no blockage, the minimum path length of the flow curves at the inlet and outlet is obviously smaller than the minimum path length calculated when the blockage occurs.
[0123] In step four, due to the volatility of the flow curves on the time axis, only the trend difference degree of the flow curves is not enough to provide a reliable blockage early warning, and therefore the difference degree of the flow curves in the numerical characteristics needs to be calculated.
[0124] As shown in Figure 8 , the total difference between the flow curves at the inlet and outlet is directly used as the numerical difference degree, which is denoted as Numerical value , which can fully consider the influence of different material flow conditions, and from Figure 8 it can be seen that when there is no blockage, the value of the numerical difference degree is obviously smaller than the numerical matching degree calculated when the blockage occurs.
[0125] In step five, in order to reduce the occasional false detection of the inlet under the condition of small flow (as shown in Figure 9 ), the trend difference degree caused by frequent data fluctuations is large, and under the condition of large flow, the numerical difference degree caused by large numerical dimension is large, and the multiplication fusion method of the trend difference degree and the numerical difference degree makes it adapt to the variable flow in the production environment, and thus the final difference degree measurement index is as follows:
[0126] Indicator = Trend value *Numerical value .
[0127] When Indicator ≤ Blockage thresholdWhen the value is less than the threshold value, it indicates that no blockage occurs, otherwise it triggers a blockage warning; wherein Blockage threshold The threshold value is obtained from a large number of tests of the pairs of inlet and outlet ports.
[0128] The concave-convex characteristics of the laser line generated by the linear laser generator on the material are not affected by the shape of the material, so it can support the conveying of materials of any shape, including granular bodies and solid shapes. Whether the belt of the conveying surface is concave or not has little effect on detection and can be ignored.
[0129] The intensity of the linear laser generator is greater than the threshold value, so the reflection of commonly seen materials such as ore, coal, and sand conveyed by the conveying belt can be ignored.
Claims
1. A material blockage early warning method based on the calculation of the difference in material flow values at the inlet and outlet, characterized by: The method is based on the calculation of the matching degree of material flow values at the inlet and outlet. First, the time offset of the inlet and outlet is calculated and the flow curves are aligned on the time axis. Then, a dynamic time warping algorithm is used to analyze the trend difference of the flow curves of the inlet and outlet. The numerical characteristics of the flow curves are used to analyze the numerical difference characteristics to achieve early warning of blockage failure.
2. The method for early warning of material blockage based on calculation of material flow difference between inlet and outlet according to claim 1, characterized in that: The feed port uses a conveyor belt to feed materials, and the discharge port uses a conveyor belt to feed materials. The materials fed into the feed port fall to the discharge port through a drop path connected to the drop port. The blockage warning method is provided with terminal equipment including a high-speed camera and a line laser generator at the conveyor belts of the feed port and the discharge port; The high-speed camera is used to provide camera images and is the peripheral input source device of the real-time material flow detection equipment. The line laser generator is the external hardware that provides the surface contour of the object in the camera image.
3. The material blockage early warning method based on the calculation of the difference in material flow rate values at the inlet and outlet according to claim 2 is characterized in that: The material blocking early warning method comprises the following steps: Step 1: Data statistics and preprocessing; Step 2: Calculate the time offset of the material flow curve of the inlet and outlet; Step 3: Calculate the trend matching degree of the material flow curve of the inlet and outlet; Step 4: Calculate the numerical difference of the material flow curves at the inlet and outlet; Step 5: Fusion of trend difference and numerical difference.
4. The method for early warning of material blockage based on calculation of material flow difference between inlet and outlet according to claim 3, characterized in that: Step 1: Perform data filtering on the data. Specifically, assume that the flow value data of a single material outlet within a period of time is f(t), m≤t≤n, the camera of the terminal device provides about 30 frames of images per second, and the flow value data provided is 30 / second. Select every 500 milliseconds, that is, about 15 data points, to perform a slice median filter to filter out the spike noise that appears in a short period of time. The filter calculation expression is as follows: y(t)=median(f(t),…,f(t+500)).
5. The method for early warning of material blockage based on calculation of material flow difference between inlet and outlet according to claim 3, characterized in that: In step 2, the material at the feed port needs to pass through the discharge port before reaching the discharge port, so the appearance time of a certain material at the feed port and the discharge port is not equal, resulting in the material flow curve of the feed port and the discharge port on the time axis being offset on the time axis; The material flow rate at the feed inlet detection device is taken as the reference timestamp t0, and the time when the material reaches the discharge detection device is t1. The time offset between the material flow rate at the feed inlet and the material flow rate at the discharge is recorded as Δt = t1-t0. Let x be the timestamp, and the flow data of the material at the feed inlet from t0 to t1 is recorded as f in (x), f out (x), and the random error caused by other factors is ε, which has the following relationship: f out (x)=f in (x+Δt)+ε To calculate the time offset Δt, use a time window with a preset reasonable size (Size). Δt is searched within the range above and below Size. If the following relationship is met, the traffic in this window is too small to be used for time offset calculation. This special case needs to be excluded. Then this window performs a difference calculation of the inlet and outlet flow curves of Size times on the flow data. The calculation expression is as follows: Normalize D(i) to get norm(D(i)) to reduce the degree to which the trough is affected by the scale; D(i) uses a preset threshold to determine whether the trough is obvious; The definition of trough is as follows: there is a point in the correlation coefficient curve calculated under a certain window. Denoted as p i =(t i ,c i ). For p i When the following conditions are met, record p i is the trough calculated in the current window; p i-1 ≤p i ≤p i+1 Thus we get a trough set P(x) = {p (1) ,p (2) ,…,p (n) }, 0≤x≤n≤Size; In order to extract clear valleys from P(x), exclude the following cases: First, when the calculated When , it indicates that the current correlation coefficient curve forms a steep slope rather than a trough; The second is to continuously select two trough points in the trough set P(x), denoted as like This indicates that the correlation coefficient curve has multiple continuous troughs. Threshold 1 and Threshold 2 are threshold values obtained through extensive testing of multiple pairs of inlets and outlets. When the above two cases are excluded, the ordinate of the smallest trough point in the trough set P(x) is selected as the optimal time offset calculated here; For the case where there is no trough, that is, for the latter two cases, the current time offset calculation does not overwrite the previous calculation to ensure the stability of the offset; The index value corresponding to the result with only one obvious trough is selected as the time offset Δt obtained by the final calculation, and D(i) is the correlation coefficient function of a small time window in the material flow value curve. This function changes with the data characteristics of the time window. If the Δt obtained by two consecutive time offset calculations exceeds a certain threshold time unit, the previous time offset calculation result will not be overwritten; thus, the time offset of the material flow curve of the inlet and outlet is obtained.
6. The method for early warning of material blockage based on calculation of material flow difference between inlet and outlet according to claim 3, characterized in that: In step 3, a queue is used to cache data points within 5 seconds. The first-in, first-out principle is used to ensure real-time updates of data points. The update frequency is equal to the filtering time. Small sample data points are obtained. Each group of small sample data points has shape characteristics. Two similar traffic curves have similar shape characteristics. The dynamic time warping algorithm, abbreviated as DTW algorithm, is used to calculate the trend difference between the two traffic curves. The DTW algorithm calculates the warped path distance between two time series to measure the similarity between the two time series. The specific method of applying the DTW algorithm here is as follows: Assume that in the current time window, n=m=number of data points, and the material flow rates at the inlet and outlet are X={x1,x2,…,x n }, Y={y1,y2,…,y m }, any point x on the X sequence i and any point y on the Y sequence i The distance between them is defined as d(i,j)=|x i -y j | p , where || p is the norm, when p = 2 it is the 2-norm, then d(i,j) = |x i -y j | 2 , which is the Euclidean distance; Create an n×m cumulative distance matrix, defined as follows: Where D(1,1)=d(1,1), As shown in the definition of the cumulative distance matrix above: the cumulative distance D(i, j) is the sum of the distance d(i, j) of the current element and the cumulative distance of the smallest adjacent element that can reach the element; Define a regular path W = {w1,w2,...,w K }, where w k =d(i,j) k ,max(n,m)≤K≤n+m-1, K is the total length of the regular path; the regular path satisfies the following constraints: Boundary conditions: w1=D(1,1),w K =D(n,m), indicating that the first element of the regular path W must be D(1,1) and the last element must be D(n,m), that is, the selected regular path must start at D(1,1) and end at D(n,m); Monotonicity: Note w k =D(i k ,j k ),w k+1 =D(i k+1 ,j k+1 ), then i k ≤i k+1 ,j k ≤j k+1 . Continuity: Remember w k =D(i k ,j k ),w k+1 =D(i k+1 ,j k+1 ), then i k+1 -i k ≤1,j k+1 -j k ≤1; The DTW distance, that is, the minimum path length calculated by DTW, is calculated as follows: Among them, DTW(X,Y) is the trend difference between the two flow curves finally calculated. In order to reduce the impact of the dimension of the data point on the trend difference of the flow curve calculated by dynamic time warping, the two flow curves are normalized and then the dynamic time warping algorithm is used to calculate the minimum path length value, which is recorded as Trend value The flow curve of the feed port is compared with the flow curve of the discharge port, and the flow curve of the discharge port simulating the blockage is calculated by dynamic time warping algorithm.
7. The method for early warning of material blockage based on calculation of material flow difference between inlet and outlet according to claim 3, characterized in that: In step 4, the total difference between the inlet and outlet flow curves is directly used as the numerical difference, which is recorded as Numerical value , in order to fully consider the impact of different material flow conditions.
8. The method for early warning of material blockage based on calculation of material flow difference between inlet and outlet according to claim 3, characterized in that: In step 5, the trend difference and the value difference are multiplied and merged to adapt to the variable traffic in the production environment. The final difference measurement index is as follows: Indicator=Trend value *Numerical value ; When Indicator≤Blockage threshold When , it indicates that there is no material blocking, otherwise it triggers the material blocking warning; threshold This is the threshold value obtained through extensive testing at multiple pairs of inlets and outlets.
9. The method for early warning of material blockage based on calculation of material flow difference between inlet and outlet according to claim 2, characterized in that: The concave and convex characteristics reflected by the laser line of the line laser generator on the material are not affected by the shape of the material, and the belt depression on the conveying surface of the conveyor belt can be ignored.
10. The method for early warning of material blockage based on calculation of material flow difference between inlet and outlet according to claim 2, characterized in that: The intensity of the line laser of the line laser generator is greater than the threshold value, so that the reflection effect of the material transported by the conveyor belt can be ignored.