Fault early warning method for tailing paste filling system

Abstract

The invention relates to the field of fault early warning, in particular to a fault early warning method for a tailing paste filling system. Comprising the steps of monitoring and storing parameters in real time to form a parameter matrix; a sliding time window technology is introduced, standardization processing is carried out on the parameters at each moment, and standardized parameter values are obtained; on the basis of the standardized parameter values, a dynamic feature extraction algorithm is adopted for processing to obtain feature values; based on the characteristic value, a fault probability value is calculated through a dynamic gain adjustment fault probability calculation algorithm, the fault probability value is compared with a fault threshold value, and a fault judgment result is obtained; according to the characteristic value and the fault judgment result, a saliency score is calculated through a fault positioning algorithm, and a fault source is determined according to the saliency score; and performing fault early warning based on the fault source. The technical problems that in a tailing paste filling system, multiple monitoring points exist, so that real-time data processing becomes complex, and unified comparison and analysis are difficult to conduct are solved.

Description

Technical Field

[0001] The invention relates to the field of fault warning, and in particular to a fault warning method for a tailings paste filling system. Background Art

[0002] Tailings paste filling technology is an important means of mine waste treatment, which aims to reduce the pollution of tailings to the environment and realize the reuse of mine resources. However, with the increasing application of tailings paste filling technology, the importance of tailings paste filling system in mine production has become increasingly prominent. The failure of the tailings paste filling system not only affects the efficiency of tailings paste filling, but also may pose a threat to the safety of the mine and may even cause mine accidents. The traditional tailings paste filling system fault warning mostly relies on manual inspection and simple equipment monitoring, and has the disadvantages of monitoring delay and poor accuracy. It is difficult to detect the potential risks of tailings paste filling system failure in real time. Therefore, timely detection and warning of potential failures of the tailings paste filling system can effectively prevent the spread of failures, reduce losses and ensure the safe production of mines.

[0003] However, the existing fault warning methods for tailings paste filling systems have the following technical problems: In the tailings paste filling system, there are multiple monitoring points, which makes real-time data processing complicated and difficult to compare and analyze uniformly. Simple static standardization processing will result in the inability to effectively integrate the parameters of different monitoring points, thereby affecting the accuracy of fault warning; traditional fault detection methods are difficult to effectively capture the sudden changes or non-periodic fluctuations of the tailings paste filling system, especially in the case of multiple monitoring points and multiple parameters. The existing feature extraction method is difficult to accurately reflect the change pattern of the tailings paste filling system, resulting in a lag in fault judgment; when calculating the fault probability, traditional fault detection methods often use fixed weights or linear weighting methods, which are not sensitive enough to the rapidly changing monitoring points, and are prone to ignoring sudden events, affecting the accuracy of fault judgment; fault location methods rely on static parameter differences or simple threshold judgments, and cannot accurately locate the source of the fault, especially when the parameters of multiple monitoring points change dramatically, which is prone to misjudgment; traditional fault warning methods for tailings paste filling systems lack comprehensive evaluation of faults, and cannot display the location and abnormal parameters of abnormal monitoring points in real time, resulting in the inability of management personnel to quickly locate and handle problems when a fault occurs in the tailings paste filling system. Summary of the invention

[0004] The invention provides a tailings paste filling system fault early warning method to solve the problem that there are multiple monitoring points in the tailings paste filling system, which makes real-time data processing complicated and difficult to compare and analyze uniformly. Simple static standardization processing will cause the parameters of different monitoring points to be unable to be effectively integrated, thereby affecting the accuracy of fault early warning; traditional fault detection methods are difficult to effectively capture sudden changes or non-periodic fluctuations of the tailings paste filling system, especially in the case of multiple monitoring points and multiple parameters, the existing feature extraction method is difficult to accurately reflect the change mode of the tailings paste filling system, resulting in a lag in fault judgment; when calculating the fault probability, the traditional fault detection method often adopts a fixed weight or linear weighting method, which is not sensitive enough to the rapidly changing monitoring points, and is easy to ignore sudden events, affecting the accuracy of fault judgment; the fault location method relies on static parameter differences or simple threshold judgment, and cannot accurately locate the fault source, especially in the case of drastic parameter changes at multiple monitoring points, which is prone to misjudgment; the traditional tailings paste filling system fault early warning method lacks comprehensive evaluation of the fault, and cannot display the position and abnormal parameters of abnormal monitoring points in real time, resulting in the technical problem that when the tailings paste filling system fails, the management personnel cannot quickly locate and handle the problem.

[0005] A tailings paste filling system fault early warning method of the present invention specifically includes the following technical solutions: A tailings paste filling system fault early warning method comprises the following steps: S1: Monitor and store parameters in real time to form a parameter matrix; introduce sliding time window technology to standardize the parameters at each moment to obtain the standardized parameter values; S2: Based on the standardized parameter values, the dynamic feature extraction algorithm is used to obtain the feature values; S3: Based on the characteristic value, a fault probability value is calculated by a dynamic gain adjustment fault probability calculation algorithm, and the fault probability value is compared with the fault threshold to obtain a fault judgment result; S4: Based on the characteristic value and the fault judgment result, a significance score is calculated by a fault location algorithm, and the fault source is determined according to the significance score; based on the fault source, a fault warning is performed.

[0006] Preferably, the S1 specifically includes: The sliding time window technology analyzes the parameters monitored in real time by dividing the sliding window into different time periods. By sliding a fixed step, as time goes by, new parameters will automatically enter the sliding time window, and outdated parameters will be removed, so that each sliding time window covers the parameters of the current and next time periods.

[0007] Preferably, the S1 specifically includes: Based on the real-time monitored and stored parameters, the parameters at each moment are standardized through real-time calculation of the mean and standard deviation to obtain the standardized parameter value. The standardization formula is:

[0008] in, Indicates at time Time The monitoring point The standardized parameter values; Indicates at time Time The monitoring point parameter values; Indicates at time Time The monitoring point The mean of the parameters; Indicates at time Time The monitoring point The standard deviation of the parameter.

[0009] Preferably, the S2 specifically includes: The core of the dynamic feature extraction algorithm includes two aspects: dynamic energy characteristics and dynamic change enhancement items; the dynamic energy characteristics are obtained by squaring and integrating the standardized parameter values ​​of the monitoring points, and normalizing the dynamic energy characteristics of the parameters; the dynamic change enhancement items adjust the sensitivity of non-periodic fluctuations by integrating the absolute values ​​of the standardized parameter values ​​of the monitoring points within the sliding time window.

[0010] Preferably, the S2 specifically includes: The calculation formula of the eigenvalue is:

[0011] in, Indicated in The first sliding time window The monitoring point The characteristic value of the parameter; Indicates at time Time The monitoring point The standardized parameter values; Indicated in The first sliding time window The monitoring point Dynamic energy characteristics of parameters; represents the normalization term; Represents the square root operation; Indicates dynamically changing enhancement items; Represents the weight factor.

[0012] Preferably, the S3 specifically includes: The dynamic gain adjustment fault probability calculation algorithm obtains the fault probability value by multiplying the characteristic value of the monitoring point parameter with the gain adjustment factor, adding them up, and combining them with the nonlinear activation function. The specific calculation formula is:

[0013] in, Indicated in Failure probability value within a sliding time window; express Activation function; It means weighted summation of the characteristic values ​​of all parameters of all monitoring points; Indicated in The first sliding time window The monitoring point The characteristic value of a parameter; Indicated in The first sliding time window The monitoring point The gain adjustment factor of a parameter.

[0014] Preferably, the S4 specifically includes: In the implementation of the fault location algorithm, the absolute difference between the characteristic value of the monitoring point parameter and the expected value is calculated. By introducing the standard deviation of the characteristic value of the monitoring point parameter and the dynamic adaptive coupling factor and combining it with the sine function, the significance score is obtained.

[0015] Preferably, the S4 specifically includes: The calculation formula for the significance score is:

[0016] in, Indicated in The first sliding time window The monitoring point The significance score of each parameter; Indicated in The first sliding time window The monitoring point The characteristic value of the parameter; Indicates expected value; represents the parameter deviation, that is, The first sliding time window The monitoring point The absolute difference between the characteristic value and the expected value of a parameter; Indicated in The first sliding time window The monitoring point The standard deviation of the eigenvalues ​​of the parameters; represents a very small constant; represents the normalization term; represents the dynamic adaptive coupling factor.

[0017] Preferably, the S4 specifically includes: A significance score threshold is set and compared with the significance score to determine the fault source; when the significance score exceeds the significance score threshold, the parameter at the monitoring point is considered to be the source of the fault; an early warning is issued based on the fault source, and the location of the abnormal monitoring point and the abnormal parameters are displayed.

[0018] The beneficial effects of the technical solution of the present invention are: 1. By installing sensors at key monitoring points of the tailings paste filling system and monitoring parameters in real time, such as pressure, flow rate, and concentration, and combining the sliding time window technology to process the real-time monitored parameters in real time, the real-time monitoring of the tailings paste filling system can be ensured. It can not only ensure that the tailings paste filling system can obtain timely and accurate monitoring data at all times, but also standardize the various parameters to eliminate the problems of dimensional differences and different numerical ranges, thereby providing a reliable data basis for subsequent dynamic analysis, ensuring that the tailings paste filling system can obtain and process the parameters of each monitoring point in real time, reduce the analysis errors caused by dimensional differences, and improve the accuracy and efficiency of data processing.

[0019] 2. By introducing the dynamic feature extraction algorithm, the fluctuation intensity and sudden changes of the tailings paste filling system state can be effectively evaluated. The dynamic energy characteristics reflect the fluctuation intensity of each parameter in the tailings paste filling system within the sliding time window, and the dynamic change enhancement item can sensitively capture non-periodic fluctuations, providing multi-dimensional data support for fault detection, thereby enhancing the early warning capability of tailings paste filling system failures. Through the extraction of characteristic values, the fluctuation intensity and sudden changes in the tailings paste filling system can be accurately identified, thereby improving the sensitivity and accuracy of fault diagnosis.

[0020] 3. The dynamic gain adjustment fault probability calculation algorithm is adopted. By performing weighted summation on the eigenvalues ​​and combining them with a nonlinear activation function to calculate the fault probability value of the tailings paste filling system, the fault probability value can be quickly responded to the dynamic changes of the tailings paste filling system. In particular, when sudden faults occur, the fault probability value can be sensitively detected and responded to, thereby improving the real-time performance of fault detection and being able to quickly detect potential risks at the early stage of a fault, thus ensuring the efficient operation and safety of the tailings paste filling system.

[0021] 4. After the fault is determined, the significance score is calculated through the fault location algorithm, and the abnormality of each monitoring point is further analyzed. Through accurate fault location, the source of the fault can be effectively determined, the possibility of misjudgment can be reduced, and the downtime and maintenance cost of the tailings paste filling system can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The present invention provides a flow chart of a tailings paste filling system fault warning method. DETAILED DESCRIPTION

[0023] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] 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.

[0025] The specific scheme of the tailings paste filling system fault warning method provided by the present invention is described in detail below with reference to the accompanying drawings.

[0026] See attached Figure 1 , which shows a flow chart of a tailings paste filling system fault warning method provided by an embodiment of the present invention, the method comprising the following steps: S1. Monitor and store parameters in real time to form a parameter matrix; introduce sliding time window technology to standardize the parameters at each moment to obtain standardized parameter values; According to the key processes and monitoring requirements of the tailings paste filling system, key monitoring points are selected and the parameters to be monitored, such as pressure, flow rate, and concentration, are determined; sensors are installed at each monitoring point to monitor and store parameters in real time to form a parameter matrix, which is expressed as follows:

[0027] in, Indicates at time Time The monitoring point parameter values, such as pressure, flow rate, concentration; Indicates the number of monitoring points; Indicates the number of parameters monitored at each monitoring point; In order to effectively process the parameters monitored in real time and perform dynamic analysis, the sliding time window technology is introduced. The sliding time window technology analyzes the parameters monitored in real time by dividing the sliding time window into multiple time periods. Contains data within a certain time range, defined as:

[0028] in, Indicates The start time of a sliding time window; Indicates the length of the sliding time window; Indicates The end time of a sliding time window; The sliding process of the time window in the sliding time window technology is also real-time, by sliding a fixed step size each time. As time goes by, new parameters will automatically enter the sliding time window, and outdated parameters will be removed, so that each sliding time window can cover the parameters of the current and next time periods, ensuring that the tailings paste filling system processes the monitored parameters in real time and ensures the timeliness of dynamic analysis; Based on the parameters monitored and stored in real time, the parameters at each moment can be standardized through real-time calculation of the mean and standard deviation to obtain the standardized parameter value, so as to eliminate the problem of dimension difference of the parameter and different parameter value range; The normalization formula is as follows:

[0029] in, Indicates at time Time The monitoring point The standardized parameter values; Indicates at time Time The monitoring point parameter values, such as pressure, flow rate, concentration; Indicates at time Time The monitoring point The mean of the parameters is calculated as: , Indicates the sliding window length; Indicates at time Time The monitoring point parameter values; Indicates the time from Time The parameters within are summed and averaged; Indicates at time Time The monitoring point The standard deviation of the parameters, i.e. Time The squares of the differences between the parameter value and the mean are summed and the square root of the mean is taken. The calculation formula is: ; The mean and standard deviation are calculated by recursive updating, which can realize real-time data processing. It only needs to update the current mean and standard deviation according to the parameters in the sliding time window, and then perform standardization. It can not only respond to changes in data streams in real time, but also reduce calculation complexity. It is suitable for fault warnings with extremely high real-time requirements. S2, based on the standardized parameter value, a dynamic feature extraction algorithm is used to obtain a feature value; Based on the standardized parameter values, the dynamic feature extraction algorithm is used to obtain the feature values; The core of the dynamic feature extraction algorithm lies in two aspects: dynamic energy characteristics and dynamic change enhancement items; The dynamic energy characteristics measure the energy characteristics by evaluating the fluctuation amplitude of each parameter within the sliding time window. Specifically, the total energy characteristics within the current sliding time window are calculated by squaring and integrating the standardized parameter values ​​of the monitoring points. In order to make the measurement comparable, the dynamic energy characteristics of all parameters are normalized. The normalized dynamic energy characteristics can reflect the fluctuation intensity of each parameter within a specific sliding time window. The stronger the fluctuation, the more instability or potential failure of the tailings paste filling system is reflected. The dynamic change enhancement item integrates the absolute value of the standardized parameter value of each parameter value at each monitoring point within the sliding time window to enhance the sensitivity to non-periodic fluctuations, so as to capture sudden changes or abnormal phenomena in the tailings paste filling system; The calculation formula of the eigenvalue is:

[0030] in, Indicated in The first sliding time window The monitoring point The characteristic value of the parameter; Indicates at time Time The monitoring point The standardized parameter values; Indicated in The first sliding time window The monitoring point The dynamic energy characteristics of each parameter are calculated by integrating the square of the standardized parameter value to calculate the energy distribution of the monitoring point and parameter within the time window, which can reflect the degree of fluctuation or the "intensity" of the change of the parameter; Represents the normalization term, which is used to calculate the The square integral of all standardized parameter values ​​of all monitoring points within a sliding time window ensures that the energy measurement of each eigenvalue is not affected by the overall fluctuation; Represents the square root operation, which is used for scale adjustment to keep it within a suitable scale range; It represents the dynamic change enhancement term, which is to integrate the absolute value of the standardized parameter value within the sliding time window, and is used to capture the non-periodic changes or abnormal fluctuations in the tailings paste filling system; represents the weight factor, which is used to balance the relative importance of the normalized dynamic energy characteristics and the dynamic change enhancement term; S3. Based on the characteristic value, a fault probability value is calculated by a dynamic gain adjustment fault probability calculation algorithm, and the fault probability value is compared with the fault threshold to obtain a fault determination result; Based on the characteristic value, the fault probability is calculated by the dynamic gain adjustment fault probability calculation algorithm; The dynamic gain adjustment fault probability calculation algorithm obtains a fault probability value by weighted summing the characteristic values ​​of all parameters of all monitoring points and combining it with a nonlinear activation function to reflect whether there is a fault in the tailings paste filling system. Specifically, the characteristic value of each parameter at each monitoring point is multiplied by the gain adjustment factor, and the sum is accumulated and input into In the activation function, we get a value between arrive The value between represents the probability of failure, that is, the failure probability value; The gain adjustment factor calculates the dynamic change rate of the characteristic value within the sliding time window, normalizes the dynamic change rate of all characteristic values, ensures the consistency of response to changes in all monitoring points, and can respond quickly to emergencies in the tailings paste filling system by enhancing the sensitivity of the characteristic value to rapid changes; The calculation formula of the failure probability value is:

[0031] in, Indicated in The failure probability value within a sliding time window ranges from ; express The activation function is used to smooth the weighted sum and convert it into a value between and To ensure that the output failure probability value is always within a reasonable range; It represents the weighted sum of the characteristic values ​​of all parameters of all monitoring points. The weighted sum reflects the overall dynamic change of the tailings paste filling system within the sliding time window and is the core of fault judgment. Indicated in The first sliding time window The monitoring point The characteristic value of the parameter; Indicated in The first sliding time window The monitoring point The gain adjustment factor of each parameter is used to weight the fault discrimination according to the characteristic change rate of the monitoring point parameter. The calculation formula is:

[0032] in, represents an exponential function, which amplifies the impact of eigenvalue changes by exponentially processing the dynamic change rate of the eigenvalue; The adjustment coefficient that controls the sensitivity of the gain adjustment factor determines the magnification of the dynamic change rate of the characteristic value to the gain adjustment factor. It can be set according to the specific implementation scenario and is not limited here. Indicated in The first sliding time window The monitoring point The absolute integral of the dynamic change rate of the eigenvalue of each parameter, by taking the absolute value, ignoring the direction of the feature change, emphasizing the amplitude of the feature change, the integral reflects the change amount of the eigenvalue within the time window, and can evaluate the degree of dynamic change of the eigenvalue; Indicates The monitoring point The dynamic change rate of the characteristic value of a parameter; It means calculating the maximum value of the characteristic change rate of all parameters of all monitoring points, ensuring that all gain adjustment factors are normalized according to the maximum characteristic change rate, to prevent some characteristics from changing drastically and affecting the stability of the entire tailings paste filling system; Setting the Failure Threshold , which can be set according to the specific implementation scenario and is not limited here; the fault probability value is compared with the fault threshold to obtain a fault judgment result; based on the fault judgment result, it is determined whether the tailings paste filling system has a fault. When the fault probability value is greater than the fault threshold, it is considered that the tailings paste filling system is at fault. There is a fault within a sliding time window, otherwise, there is no fault; S4. According to the characteristic value and the fault determination result, a significance score is calculated by a fault location algorithm, and the fault source is determined according to the significance score; based on the fault source, a fault warning is performed; According to the characteristic value and the fault judgment result, the significance score is calculated by the fault location algorithm; The fault location algorithm calculates the absolute difference between the characteristic value and the expected value of each parameter at each monitoring point. The absolute difference reflects the difference between the current state and the normal state of the monitoring point. The greater the difference, the higher the degree of abnormality. In order to further improve the calculation accuracy, the standard deviation of the characteristic value of the monitoring point parameter is introduced to measure the fluctuation range of the characteristic value. It can pay more attention to the drastic change of the characteristic value, thereby improving the sensitivity of fault detection. A dynamic adaptive coupling factor is further introduced to reflect the fluctuation of the tailings paste filling system state based on the change of the characteristic value of the monitoring point parameter between adjacent time windows. The nonlinear processing is performed through the sine function, so that when the characteristic value changes greatly, the response is sensitive, which helps to accurately capture the abnormal behavior of the tailings paste filling system. The calculation formula for the significance score is:

[0033] in, Indicated in The first sliding time window The monitoring point The significance score of each parameter; Indicated in The first sliding time window The monitoring point The characteristic value of the parameter; represents the expected value, which can be calculated by a predefined normal value, which is a technical means well known to those skilled in the art and will not be described in detail here; represents the parameter deviation, that is, The first sliding time window The monitoring point The absolute difference between the characteristic value and the expected value of a parameter; Indicated in The first sliding time window The monitoring point The standard deviation of the characteristic value of each parameter is used to measure the volatility of the characteristic value of the monitoring point parameter, which is a technical means well known to those skilled in the art and will not be described in detail here; Indicates a very small constant, which is used to prevent the denominator from being zero and ensure calculation stability. It can be set according to the specific implementation scenario and is not limited here; represents the normalization term, which is used to normalize the significance scores of all parameters of all monitoring points; represents the dynamic adaptive coupling factor, which is calculated as follows:

[0034] in, Indicates The monitoring point The change in the characteristic value of a parameter between adjacent time windows is calculated as follows: ; Indicated in The first sliding time window The monitoring point The characteristic value of the parameter; Indicated in The first sliding time window The monitoring point The characteristic value of the parameter; Represents a mathematical constant, approximately equal to 3.14159; It means that nonlinear processing is performed through a sine function, so that when the characteristic value changes greatly, the response is sensitive, which helps to accurately capture the abnormal behavior of the tailings paste filling system; Represents the coupling coefficient, reflecting the The monitoring point The correlation between the parameters can be set according to the specific implementation scenario and is not limited here; A significance score threshold is set and compared with the significance score to determine the fault source; specifically, when the significance score exceeds the significance score threshold, the parameter at the monitoring point is considered to be the source of the fault; the significance score threshold can be set according to the specific implementation scenario and is not limited here; Early warning is given according to the source of the fault, and the location and abnormal parameters of the abnormal monitoring point are displayed.

[0035] In summary, a tailings paste filling system fault early warning method is completed.

[0036] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0037] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0038] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A tailings paste filling system fault early warning method, characterized in that: The following steps are involved: S1: Monitor and store parameters in real time to form a parameter matrix; introduce sliding time window technology to standardize the parameters at each moment to obtain the standardized parameter values; S2: Based on the standardized parameter values, the dynamic feature extraction algorithm is used to obtain the feature values; S3: Based on the characteristic value, a fault probability value is calculated by a dynamic gain adjustment fault probability calculation algorithm, and the fault probability value is compared with the fault threshold to obtain a fault judgment result; S4: According to the characteristic value and the fault determination result, a significance score is calculated by a fault location algorithm, and the fault source is determined according to the significance score; Based on the fault source, fault warning is carried out.

2. A tailings paste filling system fault early warning method according to claim 1, characterized in that: The S1 specifically includes: The sliding time window technology analyzes the parameters monitored in real time by dividing the sliding window into different time periods. By sliding a fixed step, as time goes by, new parameters will automatically enter the sliding time window, and outdated parameters will be removed, so that each sliding time window covers the parameters of the current and next time periods.

3. A tailings paste filling system fault early warning method according to claim 2, characterized in that: The S1 specifically includes: Based on the real-time monitored and stored parameters, the parameters at each moment are standardized through real-time calculation of the mean and standard deviation to obtain the standardized parameter value. The standardization formula is: ； in, Indicates at time Time The monitoring point The standardized parameter values; Indicates at time Time The monitoring point parameter values; Indicates at time Time The monitoring point The mean of the parameters; Indicates at time Time The monitoring point The standard deviation of the parameter.

4. A tailings paste filling system fault early warning method according to claim 1, characterized in that: The S2 specifically includes: The core of the dynamic feature extraction algorithm includes two aspects: dynamic energy characteristics and dynamic change enhancement items; the dynamic energy characteristics are obtained by squaring and integrating the standardized parameter values ​​of the monitoring points, and normalizing the dynamic energy characteristics of the parameters; the dynamic change enhancement items adjust the sensitivity of non-periodic fluctuations by integrating the absolute values ​​of the standardized parameter values ​​of the monitoring points within the sliding time window.

5. A tailings paste filling system fault early warning method according to claim 4, characterized in that: The S2 specifically includes: The calculation formula of the eigenvalue is: ； in, Indicated in The first sliding time window The monitoring point The characteristic value of a parameter; Indicates at time Time The monitoring point The standardized parameter values; Indicated in The first sliding time window The monitoring point Dynamic energy characteristics of parameters; represents the normalization term; Represents the square root operation; Indicates dynamically changing enhancement items; Represents the weight factor.

6. A tailings paste filling system fault early warning method according to claim 1, characterized in that: The S3 specifically includes: The dynamic gain adjustment fault probability calculation algorithm obtains the fault probability value by multiplying the characteristic value of the monitoring point parameter with the gain adjustment factor, adding them up, and combining them with the nonlinear activation function. The specific calculation formula is: ； in, Indicated in Failure probability value within a sliding time window; express Activation function; It means weighted summation of the characteristic values ​​of all parameters of all monitoring points; Indicated in The first sliding time window The monitoring point The characteristic value of a parameter; Indicated in The first sliding time window The monitoring point The gain adjustment factor of a parameter.

7. A tailings paste filling system fault early warning method according to claim 1, characterized in that: The S4 specifically includes: In the implementation of the fault location algorithm, the absolute difference between the characteristic value of the monitoring point parameter and the expected value is calculated. By introducing the standard deviation of the characteristic value of the monitoring point parameter and the dynamic adaptive coupling factor and combining it with the sine function, the significance score is obtained.

8. A tailings paste filling system fault early warning method according to claim 7, characterized in that: The S4 specifically includes: The calculation formula for the significance score is: ； in, Indicated in The first sliding time window The monitoring point The significance score of each parameter; Indicated in The first sliding time window The monitoring point The characteristic value of a parameter; Indicates expected value; represents the parameter deviation, that is, The first sliding time window The monitoring point The absolute difference between the characteristic value and the expected value of a parameter; Indicated in The first sliding time window The monitoring point The standard deviation of the eigenvalues ​​of the parameters; represents a very small constant; represents the normalization term; represents the dynamic adaptive coupling factor.

9. A tailings paste filling system fault early warning method according to claim 7, characterized in that: The S4 specifically includes: A significance score threshold is set and compared with the significance score to determine the fault source; when the significance score exceeds the significance score threshold, the parameter at the monitoring point is considered to be the source of the fault; an early warning is issued based on the fault source, and the location of the abnormal monitoring point and the abnormal parameters are displayed.

Images