Medium-speed coal mill degradation analysis and fault early warning method

Abstract

The application discloses a medium-speed coal mill degradation analysis and fault early warning method, comprising the following steps: collecting the running state parameters of the coal mill, wherein the running state parameters comprise the coal supply amount, the loading force, the outlet temperature, the current and the inlet-outlet differential pressure of the medium-speed coal mill; the Gini formula is used to calculate the coal supply amount and the loading force respectively, and the parameter state of the CART decision tree is determined; based on the outlet temperature, the current and the inlet-outlet differential pressure, the degradation trend decision is calculated, wherein the degradation trend decision comprises a degradation trend line and a threshold control line; based on the parameter state of the CART decision tree, the degradation trend line and the threshold control line, the fault degradation analysis of the medium-speed coal mill is carried out; the application improves the degradation trend analysis accuracy of the medium-speed coal mill, and improves the running life of the medium-speed coal mill to a certain extent.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, specifically to a method for deterioration analysis and fault early warning of a medium-speed coal mill. Background Technology

[0002] Medium-speed coal mills are widely used in power plants in my country. As an important auxiliary machine on the furnace side, the coal mill provides fuel support for combustion in the furnace. Coal mills operate in harsh environments for extended periods, and the deterioration of coal quality leads to less than ideal overall performance, making them prone to malfunctions such as coal shortages and blockages during operation. Any malfunction in the coal mill directly affects the safe and stable operation of the unit. Unplanned shutdowns caused by malfunctions not only result in significant economic losses for power generation companies but also impact the stability of the entire power grid, directly affecting both enterprise profitability and grid security. Therefore, diagnosing malfunctions in medium-speed coal mills is crucial. Although some progress has been made in coal mill condition monitoring and fault diagnosis, research into methods for condition monitoring and fault diagnosis remains of great importance.

[0003] In related technologies, the literature "Research on Condition Monitoring and Fault Diagnosis Methods of Medium-Speed ​​Coal Mills, Liu Haitao, Master's Thesis" proposes a method for diagnosing large vibration faults in coal mills based on wavelet packet analysis and least squares support vector machines. In the coal mill primary air volume fault early warning method proposed in patent application CN116797204A, the relationship between health status and JS divergence is established based on the JS divergence of the model's output value and the actual primary air volume value, according to commonly used index functions in process quantification indicators, to obtain the current health status of the primary air volume. This scheme mainly uses JS divergence to measure the differences between primary air volume parameters. Summary of the Invention

[0004] The technical problem to be solved by this invention is how to improve the accuracy of the analysis of the deterioration trend of medium-speed coal mills and improve the service life of medium-speed coal mills to a certain extent.

[0005] The present invention solves the above-mentioned technical problems through the following technical means:

[0006] A method for deterioration analysis and fault early warning of medium-speed coal mills is proposed, the method comprising:

[0007] The operating status parameters of the coal mill are collected, including the coal feed rate, loading force, outlet temperature, current and inlet-outlet differential pressure of the medium-speed coal mill;

[0008] The coal feed rate and the loading force are calculated using the Gini formula to determine the parameter state of the CART decision tree.

[0009] Based on the outlet temperature, the current, and the inlet-outlet differential pressure, a degradation trend decision is calculated, which includes a degradation trend line and a threshold control line.

[0010] Based on the parameter states of the CART decision tree, the degradation trend line, and the threshold control line, a fault degradation analysis is performed on the medium-speed coal mill.

[0011] Furthermore, the calculation of the coal feed rate and the loading force using the Gini formula to determine the parameter state of the CART decision tree includes:

[0012] Extract the coal feed rate and loading force within a unit cycle period before and after the moment of failure;

[0013] The Gini formula was used to calculate the coal feed rate and loading force before and after the failure, respectively, to obtain the Gini coefficient of the first coal feed rate and the Gini coefficient of the first loading force.

[0014] According to the standard range of coal feed rate and loading force, calculate the second coal feed rate Gini coefficient and the second loading force Gini coefficient under the standard, respectively;

[0015] The smaller of the first coal feed rate Gini coefficient and the second coal feed rate Gini coefficient is selected as the parameter state under the CART decision tree, and the smaller of the first loading force Gini coefficient and the second loading force Gini coefficient is selected as the parameter state under the CART decision tree.

[0016] Furthermore, the Gini formula is expressed as:

[0017]

[0018] In the formula: gi(x i )=1-Σp i 2 p i Indicates fault category i in sample x i The probability of Nx i This represents the number of samples after dividing the data into categories, where Nx = Nx1 + Nx2 + ... + Nx n gi(x i ) represents the Gini coefficient for each sample of coal feed rate or loading force.

[0019] Further, the step of calculating the degradation trend decision based on the outlet temperature, the current, and the inlet / outlet differential pressure includes:

[0020] Based on the outlet temperature, the current, and the inlet-outlet differential pressure, the deterioration trend line is calculated using JS divergence.

[0021] The threshold control line is calculated using the three sigma criterion based on the outlet temperature, the current, and the inlet-outlet differential pressure.

[0022] Further, the step of calculating the degradation trend line using JS divergence based on the outlet temperature, the current, and the inlet / outlet differential pressure includes:

[0023] Calculate the probability density function of the Gaussian distribution of the sample data of the three types of measuring points: the outlet temperature, the current, and the inlet-outlet differential pressure.

[0024] Based on the probability density functions corresponding to the three types of measurement point sample data, the scatter plot of the three types of measurement point sample data is calculated using JS divergence.

[0025] The time-domain signals of each scatter plot are converted into frequency-domain waveform signals by using Fast Fourier Transform, and then fused to obtain a fused waveform.

[0026] The degradation trend line is obtained by windowing the fused waveform.

[0027] Furthermore, based on the probability density functions corresponding to the three types of measurement point sample data, the scatter plot of the three types of measurement point sample data is calculated using JS divergence, and the formula is expressed as:

[0028]

[0029] In the formula, A, B, and C are the probability density functions corresponding to outlet temperature, current, and inlet / outlet differential pressure, respectively; JSD(A||B) represents the JS divergence between the outlet temperature probability density function and the current probability density function, used for state classification between samples; JSD(A||C) represents the JS divergence between the outlet temperature probability density function and the inlet / outlet differential pressure probability density function; JSD(B||C) represents the JS divergence between the current probability density function and the inlet / outlet differential pressure probability density function; KL(A||α) represents the divergence between the outlet temperature probability density function and the A+B / 2 probability density function, used to measure the asymmetry of the difference between the two classes of samples. KL(B||α) represents the divergence between the current probability density function and the A+B / 2 probability density function; KL(A||β) represents the divergence between the outlet temperature probability density function and the A+C / 2 probability density function; KL(C||β) represents the divergence between the probability density function corresponding to the inlet and outlet differential pressure and the A+C / 2 probability density function; KL(B||γ) represents the divergence between the current probability density function and the B+C / 2 probability density function; KL(C||γ) represents the divergence between the probability density function corresponding to the inlet and outlet differential pressure and the B+C / 2 probability density function, respectively; || represents the symbol for the sample probability distribution.

[0030] Further, the step of windowing the fused waveform to obtain the degradation trend line includes:

[0031] An adaptive scaling factor is used to adaptively adjust the Hanning window function. The Hanning window function is then used to window the fused waveform to obtain the degradation trend line. The Hanning window function δHan(n) is expressed as:

[0032]

[0033]

[0034] In the formula: n is the total length of the window function, n = 1, 2, ..., N-1, k is the effective length of the window function, δ is the scaling factor, f is the area of ​​the residual frame in the current window, and F is the area of ​​a frame.

[0035] Further, the calculation of the threshold control line based on the outlet temperature, the current, and the inlet / outlet differential pressure using the three-sigma criterion includes:

[0036] Calculate the mean and variance of the sample data from the three types of measuring points: the outlet temperature, the current, and the inlet-outlet differential pressure.

[0037] Based on the mean and variance, the threshold control line L is calculated using the three sigma criterion:

[0038]

[0039] In the formula, L i =μ i +3ν i , i∈[1,3], the mean values ​​of the outlet temperature, current and inlet-outlet differential pressure measuring point parameters are respectively denoted as μ1, μ2 and μ3; the variances of the outlet temperature, current and inlet-outlet differential pressure measuring point parameters are respectively denoted as ν1, ν2 and ν3.

[0040] Furthermore, the method also includes:

[0041] The threshold control line is corrected using the Bessel sample standard deviation, and the corrected threshold control line L k .

[0042] Furthermore, the fault degradation analysis of the medium-speed coal mill based on the parameter states of the CART decision tree, the degradation trend line, and the threshold control line includes:

[0043] When the coal feed rate and loading force of the medium-speed coal mill are normal, the deterioration trend decision is met, and the outlet temperature and current value decrease, but the inlet and outlet differential pressure increases, it is determined that the coal mill has a coal blockage fault.

[0044] When the coal feed rate and loading force of the medium-speed coal mill are normal, and the deterioration trend decision is met but the conditions including reduced outlet temperature, reduced current value, but increased inlet and outlet differential pressure are not met, it is determined that the coal mill has other abnormalities at this time.

[0045] When the coal feed rate of the medium-speed coal mill is normal, the loading force is abnormal, the deterioration trend decision is met, and the outlet temperature and current value decrease, but the inlet and outlet differential pressure increases, it is determined that the coal mill has a coal blockage fault.

[0046] When the coal feed rate of the medium-speed coal mill is normal, the loading force is abnormal, and the deterioration trend decision is met but the conditions including reduced outlet temperature, reduced current value, and increased inlet and outlet differential pressure are not met, it is determined that the coal mill has an abnormal loading force.

[0047] When the coal feed rate of a medium-speed coal mill is too high, meets the deterioration trend decision and meets the conditions of decreasing outlet temperature, decreasing current value, but increasing inlet and outlet differential pressure, it is determined that the coal mill has a coal blockage fault.

[0048] When the coal feed rate of a medium-speed coal mill is too high, it meets the deterioration trend decision but does not meet the following conditions: the outlet temperature decreases, the current value decreases, but the inlet and outlet differential pressure increases. At this time, it is determined that the coal mill has other abnormalities.

[0049] When the coal feed rate of a medium-speed coal mill is too low, meets the deterioration trend decision, and the outlet temperature rises, the current value drops, and the inlet and outlet differential pressure decreases, it is determined that the coal mill has a coal shortage fault.

[0050] When the coal feed rate of a medium-speed coal mill is too low, meets the deterioration trend decision criteria but does not meet the criteria including increased outlet temperature, decreased current value, and decreased inlet and outlet differential pressure, it is determined that the coal mill has other abnormalities.

[0051] The advantages of this invention are:

[0052] (1) This invention takes data mining methods as its starting point and uses mathematical theory combined with artificial intelligence algorithms to carry out deterioration analysis and model building for medium-speed coal mills in power plants. Mathematical theory is convenient to calculate and highly scalable, and has a good decision-making effect for deterioration analysis of medium-speed coal mills in power plants. By using CART decision tree based on Gini coefficient, the CART decision tree model has excellent characteristics such as simple calculation, high accuracy and small amount of calculation in processing classification decisions compared with the currently widely used information entropy model. Based on CART decision tree, the fault deterioration analysis and alarm of medium-speed coal mill are realized, which improves the service life of medium-speed coal mill and provides reference and guidance for the maintenance plan of medium-speed coal mill.

[0053] (2) This invention uses Gaussian distribution combined with JS divergence to classify the status of parameters at the outlet temperature, current and inlet / outlet differential pressure of a medium-speed coal mill. Compared with existing classification algorithms, JS divergence has excellent characteristics such as good classification effect and high robustness, and can accurately classify data based on data characteristics.

[0054] (3) In this embodiment, the Fast Fourier Transform (FFT) combined with the improved Hanning window is used to realize the fusion and smoothing of the control lines of the trend changes of the outlet temperature, current and inlet and outlet differential pressure measurement points of the medium-speed coal mill. The Hanning window has high accuracy in processing signal jumps and can maximize the mutual cancellation of side lobes, thereby strengthening the main lobe effect and having a good suppression effect on frequency leakage.

[0055] (4) The present invention uses an improved three sigma criterion to complete the setting of the threshold control line. The three sigma criterion has a good convergence effect in threshold division. The three sigma criterion has a relatively accurate division accuracy when performing threshold control in the industrial field, and it objectively exists in the real environment.

[0056] (5) The present invention uses a threshold control line corrected by Bessel standard deviation, which can dynamically improve the accuracy and flexibility of the threshold control line.

[0057] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0058] Figure 1 This is a flowchart illustrating a method for deterioration analysis and fault early warning of a medium-speed coal mill according to an embodiment of the present invention.

[0059] Figure 2 This is a flowchart of a method for calculating a degradation trend line in one embodiment of the present invention;

[0060] Figure 3 This is a flowchart of a threshold control line calculation method in one embodiment of the present invention;

[0061] Figure 4 This is a degradation analysis model diagram in one embodiment of the present invention;

[0062] Figure 5 This is a hierarchical diagram of the overall system structure in one embodiment of the present invention. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] like Figure 1 As shown, an embodiment of the present invention discloses a method for deterioration analysis and fault early warning of a medium-speed coal mill, the method comprising the following steps:

[0065] S10. Collect the operating status parameters of the coal mill, including the coal feed rate, loading force, outlet temperature, current and inlet / outlet differential pressure of the medium-speed coal mill;

[0066] It should be noted that this embodiment can collect real-time and historical data of mirrored traffic in the core switch of the SIS system through the interface module. The real-time operating status data of mirrored traffic includes real-time data of coal feed rate, loading force, outlet temperature, current and inlet and outlet differential pressure measurement points of medium-speed coal mill, and the historical operating status data includes historical data of outlet temperature, current and inlet and outlet differential pressure measurement points of medium-speed coal mill.

[0067] It should be noted that in this embodiment, the selection of parameters mainly considers the correlation and contribution of the parameters. Among them, the selection of three measurement points for historical operating data—outlet temperature, inlet-outlet differential pressure, and current—is mainly because the coal mill failure is strongly correlated with these three parameters. The impact of coal feed rate and loading force on coal mill failure is mainly an indirect factor and has no substantial impact on model training. The selection of two measurement points for real-time operating data—coal feed rate and loading force—is mainly for the purpose of analyzing the deterioration trend of the coal mill.

[0068] S20. The coal feed rate and the loading force are calculated using the Gini formula to determine the parameter state of the CART decision tree.

[0069] S30. Based on the outlet temperature, the current, and the inlet-outlet differential pressure, calculate the degradation trend decision, which includes a degradation trend line and a threshold control line.

[0070] S40. Based on the parameter status of the CART decision tree, the degradation trend line, and the threshold control line, perform fault degradation analysis on the medium-speed coal mill.

[0071] This embodiment uses a CART decision tree based on the Gini coefficient. Compared with the widely used information entropy model, the CART decision tree model has excellent characteristics such as simple calculation, high accuracy and small computational load in processing classification decisions. Based on the CART decision tree, the fault deterioration analysis and alarm of the medium-speed coal mill are realized, which improves the service life of the medium-speed coal mill and provides reference and guidance for the maintenance plan of the medium-speed coal mill.

[0072] In one embodiment, step S20: calculating the coal feed rate and the loading force using the Gini formula to determine the parameter state of the CART decision tree includes the following steps:

[0073] S21. Extract the coal feed rate and loading force within the unit cycle time period before and after the time of the fault occurrence;

[0074] Specifically, data from the period ω before the fault occurred and data from the period ω after the fault occurred are extracted, where ω is the experimental unit period.

[0075] S22. The coal feed rate and loading force before and after the failure time are calculated using the Gini formula to obtain the first coal feed rate Gini coefficient and the first loading force Gini coefficient, which are denoted as Je1 and Jc1 respectively.

[0076] Specifically, the Gini formula is expressed as:

[0077]

[0078] In the formula: gi(x i )=1-∑p i 2 p i Indicates fault category i in sample x i The probability of Nx i This represents the number of samples after dividing the data into categories, where Nx = Nx1 + Nx2 + ... + Nx n gi(x i ) represents the Gini coefficient of each sample.

[0079] S23. Calculate the second coal feed rate Gini coefficient and the second loading force Gini coefficient under the standard, respectively, according to the specified standard range of coal feed rate and loading force.

[0080] It should be noted that this embodiment follows the standard range of coal feed rate and loading force specified in the standard. Here, the Gini coefficient Je of the second coal feed rate under the standard is calculated based on the Gini formula for three ranges: less than the standard range, standard range, and greater than the standard range. 21 Je 22 Je 23The second loading force Gini coefficient Jc under the standard was calculated based on the Gini formula for three ranges: less than the standard range, standard range, and greater than the standard range. 21 Jc 22 Jc 23 .

[0081] S24. Select the smaller of the first coal feed rate Gini coefficient and the second coal feed rate Gini coefficient as the parameter state under the CART decision tree, and select the smaller of the first loading force Gini coefficient and the second loading force Gini coefficient as the parameter state under the CART decision tree.

[0082] It should be noted that, according to the properties of the Gini coefficient: the smaller the Gini coefficient, the lower the degree of disorder in the system, i.e., the higher the predictive discrimination. Here, we compare Je1 and Je... 21 Je 22 Je 23 The smaller value is selected as the state of coal feed rate under different rules; compare Jc1 and Jc 21 Jc 22 Jc 23 The smaller value is selected as the state of the applied force under different rules.

[0083] The medium-speed coal mill deterioration analysis decision model used in this embodiment relies on human experience and standard methods and is established using a CART decision tree based on the Gini coefficient. Compared with the widely used information entropy model, the CART decision tree model has excellent characteristics such as simple calculation, high accuracy and small computational load in processing classification decisions.

[0084] In one embodiment, step S30: Based on the outlet temperature, the current, and the inlet-outlet differential pressure, a degradation trend decision is calculated. The degradation trend decision includes a degradation trend line and a threshold control line, comprising the following steps:

[0085] S31. Based on the outlet temperature, the current, and the inlet-outlet differential pressure, calculate the deterioration trend line using JS divergence.

[0086] S32. Based on the outlet temperature, the current, and the inlet-outlet differential pressure, calculate the threshold control line using the three sigma criterion.

[0087] In one embodiment, such as Figure 2 As shown, in this embodiment, when performing fault deterioration analysis on a medium-speed coal mill, the deterioration trend line is calculated using outlet temperature, current, and inlet / outlet differential pressure measurement points. Accordingly, step S31: based on the outlet temperature, current, and inlet / outlet differential pressure, the deterioration trend line is calculated using JS divergence, including the following steps:

[0088] S311. Calculate the probability density function of the Gaussian distribution of the sample data of the three types of measuring points: the outlet temperature, the current, and the inlet-outlet differential pressure.

[0089] It should be noted that each sample probability distribution is approximately assumed to be a Gaussian distribution when the sample size is sufficiently large; that is, for any random variable X, its probability density function follows a Gaussian distribution. Here, the probability density functions following a Gaussian distribution are calculated for the sample data of three types of measuring points: outlet temperature, current, and inlet / outlet differential pressure, and denoted as A, B, and C, respectively.

[0090] S312. Based on the probability density functions corresponding to the three types of measurement point sample data, use JS divergence to calculate the scatter plot of the three types of measurement point sample data.

[0091] Specifically, the scatter plot of the three types of measurement point sample data is calculated using JS divergence, and the formula is expressed as:

[0092]

[0093] In the formula, A, B, and C are the probability density functions corresponding to outlet temperature, current, and inlet / outlet differential pressure, respectively; JSD(A||B) represents the JS divergence between the outlet temperature probability density function and the current probability density function, used for state classification between samples; JSD(A||C) represents the JS divergence between the outlet temperature probability density function and the inlet / outlet differential pressure probability density function; JSD(B||C) represents the JS divergence between the current probability density function and the inlet / outlet differential pressure probability density function; KL(A||α) represents the divergence between the outlet temperature probability density function and the A+B / 2 probability density function, used to measure the asymmetry of the difference between the two classes of samples. KL(B||α) represents the divergence between the current probability density function and the A+B / 2 probability density function; KL(A||β) represents the divergence between the outlet temperature probability density function and the A+C / 2 probability density function; KL(C||β) represents the divergence between the probability density function corresponding to the inlet and outlet differential pressure and the A+C / 2 probability density function; KL(B||γ) represents the divergence between the current probability density function and the B+C / 2 probability density function; KL(C||γ) represents the divergence between the probability density function corresponding to the inlet and outlet differential pressure and the B+C / 2 probability density function, respectively; || represents the symbol for the sample probability distribution.

[0094] Specifically, The JS divergence, or divergence between probability distributions i and j, represents a measure of the asymmetry of the difference between the two probability distributions i and j. Because when a medium-speed coal mill malfunctions, the outlet temperature, current, and inlet / outlet differential pressure measurement parameters change independently but are interconnected, the influence weight of the three measurement points on the medium-speed coal mill malfunction is considered equal; that is, the mutual JS divergences of the two parameters are the same. This embodiment uses JS divergence to mainly calculate the differences in the three types of parameters—outlet temperature, current, and inlet / outlet differential pressure—and then performs a degradation analysis of the coal mill's basic operating condition.

[0095] S313. After converting the time-domain signals of each scatter plot into frequency-domain waveform signals using Fast Fourier Transform, the signals are fused to obtain a fused waveform.

[0096] It should be noted that, in order to better characterize and describe the relationship and changing trends of the three types of samples, this embodiment uses Fast Fourier Transform (FFT) to convert the time-domain signal of the scatter plot into a frequency-domain waveform signal. FFT has excellent characteristics such as low computational complexity, high coupling, and fast computation speed in the time-frequency domain conversion process, and its overall performance is better than the widely used data representation method. Finally, the waveform diagram of the three measurement points—outlet temperature, current, and inlet / outlet differential pressure—is obtained.

[0097] S314. Window the fused waveform to obtain the degradation trend line.

[0098] Specifically, in this embodiment, an adaptive scaling factor is used to adaptively adjust the Hanning window function. The Hanning window function is used to window the fused waveform to obtain the degradation trend line. The Hanning window function δHan(n) is expressed as:

[0099]

[0100]

[0101] In the formula: n is the total length of the window function, n = 1, 2, ..., N-1, k is the effective length of the window function, δ is the scaling factor, f is the area of ​​the residual frame in the current window, and F is the area of ​​a frame.

[0102] It should be noted that because the FFT algorithm selects a portion of discrete data for conversion during the time-frequency domain transformation, periodic signals require assumption processing. Therefore, the resulting fused waveform exhibits jumps within non-integer periods, leading to distortion and frequency leakage. To obtain a more realistic degradation trend line, this embodiment uses an optimized Hanning window function to window the waveform after the FFT algorithm. The Hanning window is highly accurate in handling energy jumps and effectively suppresses frequency leakage; however, its adaptability to different curves is weak. To overcome this deficiency, the Hanning window is optimized by designing an adaptive scaling factor to adjust it accordingly.

[0103] The optimized window function in this embodiment can be represented as δHan(n), which can be understood as follows: the smaller the residual frame area, the larger the correction portion; and the larger the residual frame area, the smaller the correction portion. This satisfies the requirements for elastic smoothing and repair of the fused degradation trend line. Ultimately, the optimized Hanning window can smooth the degradation trend line, reducing distortion and frequency leakage, resulting in a more accurate degradation trend line.

[0104] In one embodiment, such as Figure 3 As shown, step S32: Based on the outlet temperature, the current, and the inlet-outlet differential pressure, the threshold control line is calculated using the three sigma criterion, including the following steps:

[0105] S321. Calculate the mean and variance of the sample data of the three types of measuring points: the outlet temperature, the current, and the inlet-outlet differential pressure.

[0106] It should be noted that in this embodiment, sample data of outlet temperature, current and inlet / outlet differential pressure measuring point parameters under non-fault conditions are selected from the historical data of medium-speed coal mill. Then, the mean and variance of the outlet temperature, current and inlet / outlet differential pressure measuring point parameters are calculated. Here, the mean is denoted as μ1, μ2 and μ3, and the variance is denoted as ν1, ν2 and ν3, respectively.

[0107] S322. Based on the mean and variance, the threshold control line L is calculated using the three-sigma criterion:

[0108]

[0109] In the formula, L i =μ i +3ν i , i∈[1,3], the mean values ​​of the outlet temperature, current and inlet-outlet differential pressure measuring point parameters are respectively denoted as μ1, μ2 and μ3; the variances of the outlet temperature, current and inlet-outlet differential pressure measuring point parameters are respectively denoted as ν1, ν2 and ν3.

[0110] It should be noted that, considering that the outlet temperature, current and inlet / outlet differential pressure measuring points all exhibit abnormal and independent changes when a fault occurs in the medium-speed coal mill, the weights of the three are considered to be equal, and the threshold control line L is finally obtained.

[0111] This embodiment relies on the measurement parameters of the outlet temperature, current, and inlet / outlet differential pressure of the medium-speed coal mill to select the three sigma criterion to calculate the threshold control line. The three sigma criterion has relatively accurate division accuracy when performing threshold control in the industrial field and objectively exists in the real environment.

[0112] In one embodiment, the method further includes the following steps:

[0113] The threshold control line is corrected using the Bessel sample standard deviation, and the corrected threshold control line L k .

[0114] Specifically, in this embodiment, when using the three-sigma criterion for threshold limit setting, the accuracy reaches 99.74%, indicating that the three-sigma criterion has a high accuracy rate. However, there is still an error range with a probability of 0.26%. To eliminate the algorithm's own error, a dynamic bias ρ is introduced as a correction value for the three-sigma control line. Here, the dynamic bias ρ is calculated using the Bessel sample standard deviation. The Bessel standard deviation, based on mathematical statistics, can meet the accuracy requirements of error correction and has high robustness. The specific calculation method is as follows:

[0115] First, select sample data for outlet temperature, current, and inlet / outlet differential pressure measurement points. Then, calculate the sample standard deviation of the three types of sample populations. Choosing to calculate the standard deviation from the overall sample data here better represents the dispersion of the overall sample. The specific calculation method is as follows:

[0116]

[0117] In the formula: N is the total sample size, x i For sample values, This represents the sample mean. Considering the three sigma criterion has an error range of 0.26% outside the interval, the portion exceeding the threshold will not affect the threshold setting. Therefore, a threshold control line L corrected for Bessel standard deviation is used. k It can be represented as:

[0118] L k =0.9987L + 0.0013ρ

[0119] The final corrected threshold control line is obtained, which dynamically improves the accuracy and flexibility of the threshold control line. ρ is the bias calculated from the standard deviation of the Bessel samples.

[0120] In one embodiment, after the parameter states, degradation trend lines, and threshold control lines of the CART decision tree are determined, a CART decision tree is established for medium-speed coal mills experiencing coal blockage, coal shortage, abnormal loading force, and other abnormal conditions. Specifically, this is manifested as follows:

[0121] When a medium-speed coal mill experiences a coal blockage, the outlet temperature decreases, the current value decreases, but the inlet and outlet differential pressure increases.

[0122] When a medium-speed coal mill experiences a coal shortage, the outlet temperature rises, the current value decreases, and the inlet and outlet differential pressure decreases.

[0123] When an abnormal loading force occurs in a medium-speed coal mill, the outlet temperature and current value decrease, but the inlet and outlet differential pressure increases.

[0124] Therefore, when medium-speed coal mills experience coal blockage, coal shortage, abnormal loading force, or other abnormal situations, the most relevant parameter measurement points are mainly the outlet temperature, current, and inlet-outlet differential pressure.

[0125] like Figure 4 As shown in the figure, this embodiment considers that the failure of a medium-speed coal mill is often closely related to multiple parameters. Therefore, a CART decision tree is established for fault decision-making and alarm. The specific judgment process is as follows:

[0126] When the coal feed rate and loading force of the medium-speed coal mill are normal, meet the deterioration trend decision method, and the outlet temperature and current value decrease, but the inlet and outlet differential pressure increases, it is determined that the coal mill has a coal blockage fault.

[0127] When the medium-speed coal mill has normal coal feed rate, normal loading force, meets the deterioration trend decision method, and does not meet the following conditions, including decreased outlet temperature, decreased current value, but increased inlet and outlet differential pressure, it is determined that the coal mill has other abnormalities at this time.

[0128] When the coal feed rate of the medium-speed coal mill is normal, the loading force is abnormal, the trend decision method is met, and the outlet temperature and current value decrease, but the inlet and outlet differential pressure increases, it is determined that the coal mill has a coal blockage fault.

[0129] When the coal feed rate of a medium-speed coal mill is normal, the loading force is abnormal, and the trend decision method is met but the following conditions are not met: the outlet temperature decreases, the current value decreases, but the inlet and outlet differential pressure increases, it is determined that the coal mill has an abnormal loading force.

[0130] When the coal feed rate of a medium-speed coal mill is too high, meets the trend decision method, and satisfies the conditions of decreasing outlet temperature, decreasing current value, but increasing inlet and outlet differential pressure, it is determined that the coal mill has a coal blockage fault.

[0131] When the coal feed rate of a medium-speed coal mill is too high, and the trend decision method is met but the conditions including a decrease in outlet temperature, a decrease in current value, but an increase in inlet and outlet differential pressure are not met, it is determined that the coal mill has other abnormalities.

[0132] When the coal feed rate of the medium-speed coal mill is too low, meets the trend decision method, and the outlet temperature rises, the current value drops, and the inlet and outlet differential pressure decreases, it is determined that the coal mill has a coal shortage fault.

[0133] When the coal feed rate of a medium-speed coal mill is too low, meets the trend decision method but does not meet the following conditions: increased outlet temperature, decreased current value, and decreased inlet and outlet differential pressure, it is determined that the coal mill has other abnormalities.

[0134] In summary, this allows for the establishment and diagnosis of CART decision trees for common coal blockage, coal shortage, abnormal loading force, and other abnormal faults in medium-speed coal mills.

[0135] Furthermore, such as Figure 5 As shown, this embodiment provides a corresponding medium-speed coal mill deterioration analysis and fault early warning system, including a physical layer 1, a data layer 2, and an application layer 3. The physical layer 1 includes an interface module, the data layer 2 includes a calculation module and a disaster recovery module, and the application layer 3 includes a display module and an alarm module. The data layer 2 communicates and interacts with the physical layer 1 and the application layer 3. The physical layer 1, the data layer 2, and the application layer 3 together form an industrial integrated machine.

[0136] The physical layer 1 includes an interface module, which involves the network port, serial port, debugging port, and spare port of the industrial all-in-one machine. It has the characteristics of flexible debugging and sensitive response. It is used to collect real-time and historical data of mirrored traffic in the core switch of the SIS system. The real-time data of mirrored traffic includes real-time data of coal feed rate, loading force, outlet temperature, current, and inlet and outlet differential pressure measurement points of the medium-speed coal mill. The historical data includes historical data of outlet temperature, current, and inlet and outlet differential pressure measurement points of the medium-speed coal mill. The interface module transmits the collected data to the data layer for logical calculation and analysis decision-making.

[0137] The data layer 2 includes a computing module and a disaster recovery module, both built into the logic computing unit of the industrial integrated machine. The logic computing unit is a micro-CPU computing hardware with computing and storage functions. The computing module is used to model and analyze the medium-speed coal mill data transmitted by the physical layer interface module, and feed the analysis results back to the application layer for display and alarms. The disaster recovery module is used to store and back up the models and key data established by the computing module, and can restore the established models and important data on-site at any time in the event of hardware failure of the industrial integrated machine.

[0138] The application layer 3 includes a display module and an alarm module. The display module is a high-definition visual LCD screen built into the industrial all-in-one machine, which has the characteristics of high resolution and human-computer interaction. It is used to receive the calculation results of the data layer calculation module and display them clearly. The alarm module is an audible and visual alarm device built into the industrial all-in-one machine. It is used to receive the analysis results of the data layer calculation module and alarm for faults in the medium-speed coal mill.

[0139] The hardware of this invention consists of commercially available products based on existing technologies, and the software is developed by those skilled in the art based on automation technology and artificial intelligence algorithm technology, making it easy to implement.

[0140] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0141] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0142] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for deterioration analysis and fault early warning of a medium-speed coal mill, characterized in that, The method includes: The operating status parameters of the coal mill are collected, including the coal feed rate, loading force, outlet temperature, current and inlet-outlet differential pressure of the medium-speed coal mill; The coal feed rate and the loading force are calculated using the Gini formula to determine the parameter state of the CART decision tree. Based on the outlet temperature, the current, and the inlet-outlet differential pressure, a degradation trend decision is calculated, which includes a degradation trend line and a threshold control line. Based on the parameter states of the CART decision tree, the degradation trend line, and the threshold control line, a fault degradation analysis is performed on the medium-speed coal mill. The step of calculating the degradation trend decision based on the outlet temperature, the current, and the inlet / outlet differential pressure includes: Based on the outlet temperature, the current, and the inlet-outlet differential pressure, the deterioration trend line is calculated using JS divergence. Based on the outlet temperature, the current, and the inlet-outlet differential pressure, the threshold control line is calculated using the three sigma criterion. The step of calculating the degradation trend line using JS divergence based on the outlet temperature, the current, and the inlet / outlet differential pressure includes: Calculate the probability density function of the Gaussian distribution of the sample data of the three types of measuring points: the outlet temperature, the current, and the inlet-outlet differential pressure. Based on the probability density functions corresponding to the three types of measurement point sample data, the scatter plot of the three types of measurement point sample data is calculated using JS divergence. The time-domain signals of each scatter plot are converted into frequency-domain waveform signals by using Fast Fourier Transform, and then fused to obtain a fused waveform. The degradation trend line is obtained by windowing the fused waveform.

2. The method for deterioration analysis and fault early warning of a medium-speed coal mill as described in claim 1, characterized in that, The calculation of the coal feed rate and the loading force using the Gini formula to determine the parameter state of the CART decision tree includes: Extract the coal feed rate and loading force within a unit cycle period before and after the moment of failure; The Gini formula was used to calculate the coal feed rate and loading force before and after the failure, respectively, to obtain the Gini coefficient of the first coal feed rate and the Gini coefficient of the first loading force. According to the standard range of coal feed rate and loading force, calculate the second coal feed rate Gini coefficient and the second loading force Gini coefficient under the standard, respectively; The smaller of the first coal feed rate Gini coefficient and the second coal feed rate Gini coefficient is selected as the parameter state under the CART decision tree, and the smaller of the first loading force Gini coefficient and the second loading force Gini coefficient is selected as the parameter state under the CART decision tree.

3. The method for deterioration analysis and fault early warning of a medium-speed coal mill as described in claim 2, characterized in that, The Gini formula is expressed as: In the formula: , Indicates the fault category In the sample The probability in This represents the number of samples after dividing the data into categories, where... , This represents the Gini coefficient for each sample of coal feed rate or loading force.

4. The method for deterioration analysis and fault early warning of a medium-speed coal mill as described in claim 1, characterized in that, The scatter plot of the three types of measurement point sample data is calculated using JS divergence based on the probability density functions corresponding to the three types of measurement point sample data. The formula is as follows: In the formula, , and These are the probability density functions corresponding to the outlet temperature, current, and inlet / outlet differential pressure, respectively. The JS divergence represents the probability density function of outlet temperature and the probability density function of current, which is used for state classification between samples. The JS divergence represents the probability density function of the outlet temperature and the probability density function of the inlet-outlet differential pressure. The JS divergence represents the probability density function of the current and the probability density function of the inlet and outlet differential pressure. The divergence between the outlet temperature probability density function and the A+B / 2 probability density function is used to measure the asymmetry of the difference between the two classes of samples. This represents the divergence between the current probability density function and the A+B / 2 probability density function; This represents the divergence between the probability density function of the outlet temperature and the probability density function of A+C / 2. This represents the divergence between the probability density function corresponding to the inlet-outlet differential pressure and the probability density function of A+C / 2. This represents the divergence between the current probability density function and the B+C / 2 probability density function; These represent the divergence between the probability density function corresponding to the inlet and outlet differential pressure and the probability density function of B+C / 2, respectively. Symbols representing the probability distribution of a sample.

5. The method for deterioration analysis and fault early warning of a medium-speed coal mill as described in claim 1, characterized in that, The step of windowing the fused waveform to obtain the degradation trend line includes: An adaptive scaling factor is used to adaptively adjust the Hanning window function. The Hanning window function is then used to window the fused waveform to obtain the degradation trend line. The Hanning window function... Represented as: In the formula: The total length of the window function. , For the effective length of the window function, This is the scaling factor. The area of ​​the remaining frame in the current window. The area of ​​one frame.

6. The method for deterioration analysis and fault early warning of a medium-speed coal mill as described in claim 1, characterized in that, The calculation of the threshold control line based on the outlet temperature, the current, and the inlet / outlet differential pressure using the three-sigma criterion includes: Calculate the mean and variance of the sample data from the three types of measuring points: the outlet temperature, the current, and the inlet-outlet differential pressure. Based on the mean and variance, the threshold control line is calculated using the three sigma criterion. : In the formula, , The average values ​​of the outlet temperature, current, and inlet / outlet differential pressure measuring points are recorded as follows: , and The variances of the outlet temperature, current, and inlet / outlet differential pressure measuring point parameters are denoted as follows: , and .

7. The method for deterioration analysis and fault early warning of a medium-speed coal mill as described in claim 1 or 6, characterized in that, The method further includes: The threshold control line is corrected using the Bessel sample standard deviation. .

8. The method for deterioration analysis and fault early warning of a medium-speed coal mill as described in claim 1, characterized in that, The fault degradation analysis of the medium-speed coal mill based on the parameter states of the CART decision tree, the degradation trend line, and the threshold control line includes: When the coal feed rate and loading force of the medium-speed coal mill are normal, the deterioration trend decision is met, and the outlet temperature and current value decrease, but the inlet and outlet differential pressure increases, it is determined that the coal mill has a coal blockage fault. When the coal feed rate and loading force of the medium-speed coal mill are normal, and the deterioration trend decision is met but the conditions including reduced outlet temperature, reduced current value, but increased inlet and outlet differential pressure are not met, it is determined that the coal mill has other abnormalities at this time. When the coal feed rate of the medium-speed coal mill is normal, the loading force is abnormal, the deterioration trend decision is met, and the outlet temperature and current value decrease, but the inlet and outlet differential pressure increases, it is determined that the coal mill has a coal blockage fault. When the coal feed rate of the medium-speed coal mill is normal, the loading force is abnormal, and the deterioration trend decision is met but the conditions including reduced outlet temperature, reduced current value, and increased inlet and outlet differential pressure are not met, it is determined that the coal mill has an abnormal loading force. When the coal feed rate of a medium-speed coal mill is too high, meets the deterioration trend decision and meets the conditions of decreasing outlet temperature, decreasing current value, but increasing inlet and outlet differential pressure, it is determined that the coal mill has a coal blockage fault. When the coal feed rate of a medium-speed coal mill is too high, it meets the deterioration trend decision but does not meet the following conditions: the outlet temperature decreases, the current value decreases, but the inlet and outlet differential pressure increases. At this time, it is determined that the coal mill has other abnormalities. When the coal feed rate of a medium-speed coal mill is too low, meets the deterioration trend decision, and the outlet temperature rises, the current value drops, and the inlet and outlet differential pressure decreases, it is determined that the coal mill has a coal shortage fault. When the coal feed rate of a medium-speed coal mill is too low, meets the deterioration trend decision criteria but does not meet the criteria including increased outlet temperature, decreased current value, and decreased inlet and outlet differential pressure, it is determined that the coal mill has other abnormalities.

Images