Early warning method, system and equipment for raw coal hanging on wall of raw coal bunker and medium

Abstract

The invention discloses a raw coal bunker raw coal wall-hanging early warning method, system and device and a medium, and the method comprises the steps: obtaining and preprocessing a coal flow voiceprint signal, and obtaining a voiceprint feature; calculating a coal flow falling speed through a cross-correlation time delay method based on the voiceprint signal; inputting the voiceprint features, the coal quality parameters and the coal flow parameters into a fusion model, and outputting a wall-hanging risk probability; based on the wall hanging risk probability, combining DCS parameter model output and coal type switching time logic output, and performing probability calibration through a D-S evidence theory; performing joint probability calculation on the calibrated probability through a Dempster combination rule to obtain a synthesized final wall hanging probability; and performing judgment and early warning based on the final wall hanging probability. According to the method, data preprocessing and feature level fusion are carried out on multi-source data, and the multi-source data are input into a bidirectional LSTM network for decision level fusion; the probability is integrated by adopting the Dempster-Shafer evidence theory, so that the problem of sensor confidence conflict can be solved, the accuracy of wall-hanging prediction can be improved, and a wall-hanging early warning effect can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of raw coal wall attachment detection, and in particular to a raw coal wall attachment early warning method, system, equipment and medium for a raw coal bin. Background Art

[0002] To meet economical operational requirements, thermal power plants all use a blended coal combustion method for boiler combustion. Each pulverizer may use a different type of coal, and the same pulverizer typically changes coal types every eight hours. This poses significant challenges to pulverizer equipment safety and operation, as well as boiler combustion and adjustment. This is primarily due to the uncertainty of coal quality. Changing coal quality while maintaining consistent operational adjustments can easily lead to problems such as deflagration, overheating, boiler coking, and adjustment lag. Therefore, it is necessary to accurately calculate when the pulverizer will run on which type of coal. However, the raw coal bunker above the pulverizer, which typically has a capacity of around 500 tons and serves as both a storage and feed system, often experiences a wall-hanging phenomenon. This occurs when a large amount of coal clings to the bunker walls, preventing it from flowing properly into the pulverizer below. Instead, the coal added later can enter the pulverizer through the blocked coal channel gaps in the bunker. This leads to two problems:

[0003] First, the coal type change situation is not easy to obtain through conventional calculation methods, which is likely to interfere with the operation of the pulverizer and the boiler combustion operation, bringing adverse effects on reliability and economy; second, the sudden collapse of the adhered coal type will cause coal flow blockage, coal mill tripping, sudden change of coal type in the pulverizer operation, etc., seriously interfering with the safe operation of the boiler.

[0004] Currently, vibration sensors are placed on the outer walls of raw coal silos to detect changes in coal flow. However, due to the large volume and thickness of raw coal silos, it is difficult to extract fault signatures from vibration signals. This makes it difficult to detect coal sticking to the wall, and even more difficult to provide early warning. In real industrial production environments, equipment is affected by a variety of factors (such as changes in coal type and silo flow), and accurate prediction of current equipment status cannot be achieved based on a single type of data. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a method and system for early warning of raw coal hanging on the wall in a raw coal bin to solve the problem that the current raw coal bin has a large volume and thick wall, it is difficult to extract fault characteristic values ​​from vibration signals, and it is impossible to identify the level of raw coal hanging on the wall, and the equipment is affected by multiple factors. A single type of data cannot accurately predict the status information of the current equipment, let alone provide early warning.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, the present invention provides an early warning method for raw coal sticking to the wall of a raw coal bin, comprising: providing a plurality of signal acquisition devices based on a preset height of a conical section of the raw coal bin;

[0009] Acquire and preprocess the coal flow voiceprint signal to obtain the voiceprint features;

[0010] Calculating the falling velocity of the coal flow by using a cross-correlation time delay method based on the soundprint signal;

[0011] Input voiceprint features, coal quality parameters, and coal flow parameters into the fusion model and output the probability of coal wall risk;

[0012] Based on the wall hanging risk probability, combined with the DCS parameter model output and the coal type switching time logic output, probability calibration is performed through DS evidence theory;

[0013] The calibrated probabilities are combined using the Dempster combination rule to calculate the joint probability, and the conflict factor of each pair of probabilities is combined to obtain the final composite wall hanging probability.

[0014] Make judgments and early warnings based on the final probability of wall hanging.

[0015] As a preferred solution of the method for early warning of raw coal wall accumulation in raw coal bin of the present invention, the pre-processing includes:

[0016] Eliminate background noise of voiceprint signal through LMS filtering algorithm;

[0017] Convert the long-term non-stationary voiceprint signal into multiple short-term stationary signals through framing;

[0018] Windowing is performed on multiple short-term stationary signals after framing;

[0019] Through short-time Fourier transform, each frame signal is expanded from the time domain to the frequency domain;

[0020] Take the energy spectrum of the short-time Fourier transform and calculate the logarithmic power spectrum.

[0021] As a preferred solution of the method for early warning of raw coal sticking to the wall of a raw coal bin of the present invention, it further includes: performing feature enhancement on the pre-processed voiceprint signal;

[0022] Pass the logarithmic power spectrum through the Mel filter bank to obtain the Fbank feature;

[0023] Perform discrete cosine transform based on Fbank features to obtain static MFCC parameters;

[0024] Dynamic features are obtained through differential operation based on static MFCC parameters;

[0025] The static MFCC is combined with the dynamic features and the frame capability to obtain the voiceprint feature, which is a 39-dimensional MFCC map.

[0026] The beneficial effects of the preferred solution of the present invention are: enhancing the frequency domain discrimination of coal flow friction sound through Mel spectrum; and improving the sensitivity to transient changes of coal flow through differential features combined with static MFCC.

[0027] As a preferred solution of the method for early warning of raw coal hanging on the wall of the raw coal bin of the present invention, wherein: the coal flow falling speed is calculated based on the voiceprint signal by the cross-correlation time delay method, including:

[0028] The coal flow falling speed is calculated based on the distance between the upper and lower signal acquisition devices of the tapered section and the cross-correlation peak delay value of the voiceprint signal.

[0029] As a preferred solution of the method for early warning of raw coal wall accumulation in raw coal bin of the present invention, the fusion model includes:

[0030] The voiceprint features, coal quality parameters and coal flow parameters are input into the input layer;

[0031] The hidden layer is a 2-layer bidirectional LSTM, and the activation function is set to tanh;

[0032] The fully connected layer outputs the risk probability of wall hanging.

[0033] As a preferred solution of the raw coal bunker wall early warning method of the present invention, based on the wall hanging risk probability, combined with the DCS parameter model output and the coal type switching time logic output, probability calibration is performed through DS evidence theory, including:

[0034] The wall hanging risk probability output by the fusion model is mapped through the Sigmoid function to achieve probability calibration;

[0035] For the wall hanging risk probability output by the DCS parameter model, ReLU linear normalization is used to map it to achieve probability calibration;

[0036] For the probability of wall hanging risk output by the coal type switching time logic, probability calibration is achieved by setting thresholds and mutation rules.

[0037] The beneficial effects of the preferred embodiment of the present invention are as follows: the physical limitations of the voiceprint characteristics are compensated by coal quality parameters, and data fusion can also reduce the risk of failure of a single sensor.

[0038] As a preferred solution of the raw coal bunker wall early warning method of the present invention, the calibrated probabilities are jointly calculated using the Dempster combination rule, and the conflict factors of each pair of probabilities are combined to obtain the final composite wall probability, including:

[0039] Combine the three probabilities in pairs, calculate the conflict factor for each pair of probabilities, and sum them up to get the total conflict factor;

[0040] If the total conflict factor is greater than the preset threshold, an adaptive weight adjustment is performed to obtain the adjusted corresponding probability;

[0041] The synthesis results are calculated based on the adjusted corresponding probabilities to obtain the final wall hanging probability.

[0042] The beneficial effect of the preferred solution of the present invention is that the confidence level of the risk probability can be reduced in the case of high conflict by combining the Dempster combination rule with the conflict factor, and the synthesis result also takes into account the confidence level and the uncertainty interval.

[0043] In a second aspect, the present invention provides a raw coal bunker wall early warning system, comprising:

[0044] An acquisition module is used to acquire and preprocess the coal flow voiceprint signal to obtain the voiceprint feature;

[0045] a calculation module, configured to calculate the falling velocity of the coal flow based on the voiceprint signal by using a cross-correlation time delay method;

[0046] The fusion module is used to input voiceprint features, coal quality parameters, and coal flow parameters into the fusion model and output the probability of coal wall damage risk;

[0047] A calibration module is used to perform probability calibration based on the wall hanging risk probability, combined with the DCS parameter model output and the coal type switching time logic output through the DS evidence theory;

[0048] The synthesis probability module is used to calculate the joint probability of the calibrated probabilities using the Dempster combination rule, and combine the conflict factors of each pair of probabilities to obtain the final composite wall hanging probability;

[0049] The early warning module is used to make judgments and early warnings based on the final probability of wall hanging.

[0050] In a third aspect, the present invention provides a computer device, comprising:

[0051] memory and processor;

[0052] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method for early warning of raw coal hanging on the wall of the raw coal bin are realized.

[0053] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the method for early warning of raw coal sticking to the wall in a raw coal bin.

[0054] Compared with the existing technology, the beneficial effects of the present invention are as follows: the present invention performs abnormal warning by setting up data collection on site and obtaining multi-dimensional data with a unique time scale, such as coal quality and coal flow characteristics monitoring. The data from these three sources are preprocessed and then fused at the feature level. The processed parameters are input into the bidirectional LSTM network for decision-level fusion to obtain the probability of wall hanging through multi-dimensional analysis. The Dempster-Shafer evidence theory is used to integrate the probabilities output by each subsystem, solve the problem of sensor confidence conflict, improve the accuracy of wall hanging prediction, and achieve wall hanging warning effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 The figure is a schematic diagram of the overall process of a method for early warning of raw coal sticking to the wall in a raw coal bin according to one embodiment of the present invention.

[0057] Figure 2 The figure is a schematic diagram of an arrangement of acoustic sensors in a method for early warning of raw coal wall buildup in a raw coal bin according to an embodiment of the present invention.

[0058] Figure 3 The present invention is a schematic diagram of an acoustic sensor arrangement method for a method for early warning of raw coal wall buildup in a raw coal bin according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0060] Example 1, with reference to Figure 1 , as one embodiment of the present invention, provides a method for early warning of raw coal sticking to the wall of a raw coal bin, comprising:

[0061] S100: multiple signal acquisition devices are set based on a preset height of the conical section of the raw coal bin;

[0062] S200: Acquire and pre-process the coal flow voiceprint signal to obtain the voiceprint feature;

[0063] S300: Calculating the falling velocity of the coal flow by using a cross-correlation time delay method based on the soundprint signal;

[0064] S400: Input the voiceprint features, coal quality parameters, and coal flow parameters into the fusion model and output the probability of coal wall adhesion risk;

[0065] S500: Based on the probability of coal wall hanging risk, combined with the DCS parameter model output and the coal type switching time logic output, probability calibration is performed through DS evidence theory;

[0066] S600: performing joint probability calculation on the calibrated probabilities using the Dempster combination rule, combining the conflict factors of each pair of probabilities, and obtaining the final composite wall hanging probability;

[0067] S700: Make a judgment and early warning based on the final probability of wall hanging.

[0068] It should be noted that in actual operation, the presence of wall hanging can cause a significant deviation between the actual coal type fed into the mill and the planned mix. However, the existing system still calculates combustion parameters based on the planned coal type, resulting in data disparity. Furthermore, blockage in the coal flow channel can cause false coal flow signals, preventing the DCS system from identifying the true coal feed rate and leading to an imbalance in the air-to-coal ratio. The instantaneous collapse of the wall hanging can trigger a sudden step change in coal flow, causing drastic fluctuations in mill output. Furthermore, the vibration signal is attenuated by the thick steel structure of the raw coal bunker, becoming drowned out by structural noise, making it impossible to accurately determine the wall hanging condition based on a single signal.

[0069] Therefore, through steps S100-S700, three different types of data sources, including field sensor data, power plant distributed control system (DCS) and coal mill and coal feeder related operating parameters, and coal mill operating coal type data in the coal blending and combustion system, are used to obtain the final diagnostic results through cross-adaptation verification, data coupling, and feature extraction model calculation.

[0070] Example 2, reference Figure 1-Figure 3 , which is an embodiment of the present invention, provides an early warning method for raw coal hanging on the wall of a raw coal bin based on the above embodiment.

[0071] In an embodiment of the present application, in step S100, multiple signal acquisition devices are set based on the preset height of the conical section of the raw coal bin, including: setting multiple acoustic sensors, and the preset height can be 3 meters and 5 meters from the outlet of the coal bin conical section.

[0072] For example, Figure 2 As shown, 4 explosion-proof annular broadband microphone (20Hz-20kHz) arrays (8 channels, 2m spacing) with a sampling rate of 48kHz are installed at 3 meters and 5 meters away from the outlet of the coal bunker cone section.

[0073] It should be noted that due to the increase in coal flow friction caused by cross-sectional shrinkage, the phenomenon of raw coal bunker wall hanging mostly occurs in the conical section. Therefore, at 3 meters, that is, close to the exit, the acoustic characteristics of the coal flow in the early stage of falling can be monitored, such as: the difference in coal type fluidity; at 5 meters, that is, in the middle of the conical section, the acoustic signal of the coal seam adhering to the coal bunker wall can be captured, such as: the collapse of the coal seam hanging on the wall will produce low-frequency sound waves.

[0074] If a coal block hanging on the wall collapses at a height of 5 meters, the 3-meter sensor may only capture a sudden change in the coal flow, while the 5-meter sensor can simultaneously detect the sound pattern of the collapse impact, thereby improving the accuracy of the early warning through dual-point data comparison.

[0075] In an optional embodiment, as Figure 3 As shown in the overhead view, the preset heights can also be set to install 4 explosion-proof ring-shaped broadband microphones (8 channels, 2m spacing) at 6 meters and 3 meters respectively. At the same time, magnetic suction + mechanical locking can be used to fix them to avoid welding damaging the explosion-proof structure of the coal bunker.

[0076] In an optional embodiment, a dustproof grille and a sound collector may be added to the surface of the sensor to prevent coal dust from clogging the microphone diaphragm and to block a portion of external noise.

[0077] In the embodiment of the present application, the pre-processing in step S200 includes the following steps A1-A5:

[0078] A1: Eliminate background noise of voiceprint signal through LMS filtering algorithm;

[0079] Specifically, the background noise elimination of the time domain signal can be expressed as:

[0080] w(n+1)=w(n)+μ·e(n)·x(n) (1)

[0081] Where w is the filter weight, which is updated iteratively; x(n) is the reference channel noise; the main channel signal d(n) is filtered and outputs the error e(n), where e(n) is the instantaneous error; μ is the step size factor (convergence factor), which can be 0.01.

[0082] A2: Convert the long-term non-stationary voiceprint signal into multiple short-term stationary signals through framing;

[0083] Specifically for a long non-stationary signal, it can be expressed as:

[0084]

[0085] Where n is the total number of signal points.

[0086] A3: Windowing is performed on multiple short-term stationary signals after framing;

[0087] Specifically, the signal sequence after framing can be expressed as:

[0088]

[0089] Among them, N f is the frame length, N s Indicates frame shift, and T is the total number of frames after framing.

[0090] In an optional embodiment, the windowing may be performed by a Hanning window, which is expressed as:

[0091]

[0092] In another optional embodiment, windowing can also be performed using a Hamming window, which can be expressed as:

[0093]

[0094] A4: Through short-time Fourier transform, each frame signal is expanded from the time domain to the frequency domain;

[0095] Specifically, the windowed signal is converted from the time domain signal to the frequency domain through FFT operation, which can be expressed as:

[0096]

[0097] Among them, X(k) is the frequency domain signal after transformation; x(n) is a segment with N points. f The time domain signal, N fff is the number of Fourier transform points, w(n) is the window function;

[0098] The STFT result of the signal can be obtained by stacking the FFT calculation results of all the signals after framing. The STFT result is denoted as F:Rn→Cf×T, as shown in the following formula (7):

[0099]

[0100] Wherein, FFT(·) is the Fourier transform expressed in equation (6).

[0101] A5: Take the energy spectrum of the short-time Fourier transform and calculate the logarithmic power spectrum.

[0102] Specifically, the STFT of a signal is a complex number, including the amplitude spectrum and phase spectrum, and the energy spectrum is the square of the amplitude spectrum. Taking the energy spectrum of the signal and expressing it in dB, we can get the logarithmic power spectrum of the signal:

[0103] 10*log 10 (|F(s)| 2 )∈R f×T (8)

[0104] For example, the above parameters can be selected as frame length Nf =512, the number of Fourier transform points N fft =512, frame shift N s =256, preferably use Hanning window as window function, because FFT has phase symmetry, STFT feature dimension is the number of Fourier transform points N fft Half the size.

[0105] In the embodiment of the present application, step S200 further includes feature enhancement of the pre-processed voiceprint signal, which specifically includes the following steps B1-B4:

[0106] B1: Pass the logarithmic power spectrum through the Mel filter bank to obtain the Fbank feature;

[0107] Specifically, the Mel filter bank parameters are set to 40 triangular filters with a range of 20 Hz to 16 kHz.

[0108] It should be noted that Fbank features are better at characterizing non-stationary signals than STFT. Mapping the linear spectrum to the human ear perception scale can improve the recognition of coal friction sound. The calculation formula can be expressed as:

[0109]

[0110] Where f represents the signal frequency.

[0111] B2: Perform discrete cosine transform based on Fbank features to obtain static MFCC parameters;

[0112] Specifically, the logarithm of the Fbank feature is taken to compress the dynamic range; the discrete cosine transform (DCT) of the logarithmic energy is performed to extract the cepstral coefficients; the first N dimension coefficients (such as N = 13) are retained to obtain the static MFCC.

[0113] B3: Dynamic features are obtained through differential operation based on static MFCC parameters;

[0114] Specifically, input the static MFCC coefficients to the first-order difference and calculate the rate of change of the MFCCs of adjacent frames.

[0115]

[0116] Among them, d t represents the tth first-order difference, C t represents the tth cepstral coefficient, Q represents the order of the cepstral coefficient, and K represents the time difference of the first-order derivative, which can be 1 or 2.

[0117] Differencing the first-order difference again and substituting the result of the above formula can obtain the parameters of the second-order difference.

[0118] B4: Static MFCC is combined with dynamic features and frame capabilities to obtain voiceprint features. The voiceprint features are 39-dimensional MFCC maps.

[0119] It should be noted that based on the above scheme, the entire composition of MFCC is actually composed of: N-dimensional MFCC parameters (N / 3 MFCC coefficients + N / 3 first-order difference parameters + N / 3 second-order difference parameters) + frame energy; frame energy refers to the volume (ie energy) of a frame, which is also an important feature of speech, and the final 39-dimensional MFCC map is obtained.

[0120] It should also be noted that the Mel spectrum was chosen because it can map the perception characteristics of the human ear, thereby enhancing the frequency domain discrimination of the coal flow friction sound; and the differential features combined with the static MFCC can also improve the sensitivity to transient changes in the coal flow, providing more accurate input for subsequent steps.

[0121] In the embodiment of the present application, the step S300 calculates the falling velocity of the coal flow based on the voiceprint signal by using the cross-correlation time delay method, including:

[0122] The coal flow falling velocity is calculated based on the distance d between the upper and lower signal acquisition devices of the conical section and the peak time delay value τ of the cross-correlation of the voiceprint signal.

[0123] Specifically, the velocity can be expressed as: v=d / τ.

[0124] For example, the time delay between the sensors at 3 meters and 5 meters is 0.12 seconds, so the coal flow speed is (5-3) / 0.12=16.67m / s. The abnormal deceleration indicates wall blockage.

[0125] It should be noted that the cross-correlation method can still accurately estimate the time delay under low signal-to-noise ratio conditions, while avoiding the wear problem of mechanical contact speed measurement equipment and is also suitable for the high dust environment of the coal bunker.

[0126] In the embodiment of the present application, the fusion model operation in step S400 includes C1-C3:

[0127] C1: Voiceprint features, coal quality parameters and coal flow parameters are input into the input layer;

[0128] For example, input voiceprint features (39-dimensional MFCC) + coal quality parameters (6 dimensions such as calorific value, moisture, ash content) + coal flow parameters (4 dimensions such as speed and coal feed rate), a total of 49 dimensions, and a time window length of 60 seconds (step length of 1 second) to the input layer.

[0129] C2: The hidden layer is a 2-layer bidirectional LSTM, and the activation function is set to tanh;

[0130] For example, a 2-layer bidirectional LSTM may have 128 units per layer and a dropout rate of 0.3.

[0131] C3: The fully connected layer outputs the wall hanging risk probability;

[0132] For example, the wall hanging risk probability P∈[0,1], the threshold can be set to 0.65 through ROC curve optimization.

[0133] In the embodiment of the present application, in step S500, based on the probability of coal wall hanging risk, combined with the DCS parameter model output and the coal type switching time logic output, probability calibration is performed through the DS evidence theory, including D1-D3:

[0134] D1: The wall hanging risk probability output by the fusion model is mapped using the Sigmoid function to achieve probability calibration;

[0135] For example, the first independent evidence source m1 is the wall hanging risk probability P1 output by the fusion model;

[0136] m1(wall hanging) = Sigmoid(K1(P1-0.5))(K1=4.0)

[0137] D2: For the wall hanging risk probability output by the DCS parameter model, ReLU linear normalization is used to map it to achieve probability calibration;

[0138] For example, the second independent evidence source m2 is the wall hanging risk probability P2 output by the DCS parameter model;

[0139] m2(wall hanging)=ReLU(P2-0.3) / 0.7

[0140] D3: For the probability of wall hanging risk output by the coal type switching time logic, probability calibration is achieved by setting thresholds and mutation rules.

[0141] For example, the third independent evidence source m3 is the wall hanging risk probability P3 output by the coal type switching time logic;

[0142]

[0143] It should be noted that the probability calibration of DS evidence theory can explicitly quantify evidence conflicts, such as the contradiction between DCS parameters and voiceprint features, to avoid misjudgment.

[0144] In the embodiment of the present application, in step S600, the calibrated probabilities are combined using the Dempster combination rule to calculate the joint probability, and the conflict factors of each pair of probabilities are combined to obtain the final composite wall hanging probability, including E1-E3:

[0145] E1: Combine the three probabilities in pairs, calculate the conflict factor for each pair of probabilities, and sum them to obtain the total conflict factor;

[0146] Specifically, the three probabilities are combined in pairs, and the conflict factor of each pair of probabilities is calculated, which is expressed as:

[0147] K ij =m i (B)m j (C)+m i (C)m j (B)

[0148] Among them, B represents the wall hanging state; C represents the normal state; m i represents the i-th probability; m j represents the j-th probability.

[0149] For example, in the basic probability distribution:

[0150] m1 (wall hanging) = 0.7, m1 (normal) = 0.3

[0151] m2 (wall-mounted) = 0.8, m2 (normal) = 0.2

[0152] m3 (wall hanging) = 0.6, m3 (normal) = 0.4

[0153] Calculate the conflict between m1 and m2: K 12 =(0.7·0.2)+(0.3·0.8)=0.26

[0154] Calculate the conflict between m1 and m3: K 13 =(0.7·0.4)+(0.3·0.6)=0.48

[0155] Calculate the conflict between m2 and m3: K 23 =(0.8·0.4)+(0.2·0.6)=0.44

[0156] Total conflict factor K = K 12 +K 13 +K 23 =1.18.

[0157] E2: If the total conflict factor is greater than the preset threshold, adaptive weight adjustment is performed to obtain the adjusted corresponding probability;

[0158] Specifically, the adaptive weight adjustment is expressed as:

[0159]

[0160] For example, the total conflict factor K is 1.18, and the preset threshold is 0.8. Since 1.18 is greater than 0.8, an adaptive weight adjustment is performed, and the adjusted weight is:

[0161] m1 (wall): Weight = 0.7

[0162] m2 (wall): Weight = 1.2

[0163] m3 (wall): Weight = 1.0

[0164] E3: Calculate the synthesis results based on the adjusted corresponding probabilities to obtain the final wall hanging probability;

[0165] Specifically, it can mean:

[0166]

[0167] Among them, K 总 is the adjusted total conflict factor.

[0168] Assuming the adjusted K 总 =0.6, then:

[0169]

[0170] It should be noted that the combination of Dempster combination rules and conflict factors can reduce the confidence level of risk probability in high conflict situations, and can also be extended to multi-sensor evidence fusion in subsequent application scenarios.

[0171] In the embodiment of the present application, the judgment and warning are performed based on the final wall hanging probability in step S700, which can be to judge whether the final wall hanging probability is greater than a preset warning threshold and whether the solid time meets the requirements.

[0172] For example, if Bel(wall hanging)≥0.75 and lasts for 5 minutes, an early warning is triggered.

[0173] In an optional embodiment, the warning judgment based on the final wall hanging probability in step S700 can also dynamically adjust the warning threshold according to the real-time working conditions. For example, when the temperature in the bin increases and the viscosity of the coal decreases, the threshold automatically increases by 0.05 to avoid false alarms.

[0174] In another optional embodiment, the warning based on the final wall hanging probability in step S700 can also trigger the warning through the time series change trend of the wall hanging probability (rather than the instantaneous value). For example, a linear regression is performed on the wall hanging probability sequence in the last N minutes to calculate the slope k. If the slope k is continuously positive and exceeds the critical value, a low-level warning can still be triggered even if the current Bel (wall hanging) is less than 0.75.

[0175] In summary, this invention provides real-time, online monitoring of the raw coal bunker's acoustic signals, coal quality, and coal flow characteristics for anomaly warnings. This replaces the traditional method where a coal bunker blockage causes the coal feeder to trip, requiring manual on-site dredging by knocking on the bunker wall, or even requiring manual access to the bunker itself. Manual methods can only be handled after the problem occurs, and are extremely risky, impacting not only equipment safety but also personal safety.

[0176] Through the synchronous collection and fusion modeling of voiceprint signals and coal quality and coal flow characteristics, it is possible to determine whether the voiceprint signals are abnormal under different equipment load conditions, which is more accurate than the determination of pure voiceprint signals.

[0177] Currently, vibration sensors are used for measurement, but their signals are severely attenuated in thick-walled raw coal silos, making it impossible to capture the early features of wall hanging. The multi-source data fusion method proposed in this invention can make up for the shortcomings of a single sensor. Voiceprint analysis may be more sensitive to changes in coal flow, and combined with coal quality parameters and operating parameters, it can detect signs of wall hanging earlier.

[0178] Example 3. The above is a schematic scheme of a method for early warning of raw coal accumulation on the raw coal bunker wall. It should be noted that the technical scheme of this system for early warning of raw coal accumulation on the raw coal bunker wall is based on the same concept as the technical scheme of the above-mentioned method for early warning of raw coal accumulation on the raw coal bunker wall. For details not described in detail in the technical scheme of the system for early warning of raw coal accumulation on the raw coal bunker wall in this embodiment, please refer to the description of the technical scheme of the above-mentioned method for early warning of raw coal accumulation on the raw coal bunker wall.

[0179] This embodiment also provides a raw coal bunker wall early warning system, comprising:

[0180] An acquisition module is used to acquire and preprocess the coal flow voiceprint signal to obtain the voiceprint feature;

[0181] A calculation module, used to calculate the falling velocity of the coal flow based on the voiceprint signal through a cross-correlation time delay method;

[0182] The fusion module is used to input voiceprint features, coal quality parameters, and coal flow parameters into the fusion model and output the probability of coal wall damage risk;

[0183] The calibration module is used to perform probability calibration based on the wall hanging risk probability, combined with the DCS parameter model output and the coal type switching time logic output through the DS evidence theory;

[0184] The synthesis probability module is used to calculate the joint probability of the calibrated probabilities using the Dempster combination rule, and combine the conflict factors of each pair of probabilities to obtain the final composite wall hanging probability;

[0185] The early warning module is used to make judgments and early warnings based on the final probability of wall hanging.

[0186] This embodiment also provides a computer device suitable for the situation of early warning of raw coal hanging on the wall in a raw coal bin, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the early warning method of raw coal hanging on the wall in a raw coal bin as proposed in the above embodiment.

[0187] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for realizing the early warning method for raw coal sticking to the wall of a raw coal bin as proposed in the above embodiment is implemented.

[0188] The storage medium proposed in this embodiment and the method for realizing the warning of raw coal hanging on the wall in the raw coal bin proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0189] Through the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented with the help of software and necessary general hardware, and of course can also be implemented by hardware. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0190] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for early warning of raw coal sticking to the wall of a raw coal bunker, characterized in that: include: Multiple signal acquisition devices are set based on the preset height of the conical section of the raw coal bin; Acquire and preprocess the coal flow voiceprint signal to obtain the voiceprint features; Calculating the falling velocity of the coal flow by using a cross-correlation time delay method based on the soundprint signal; Input voiceprint features, coal quality parameters, and coal flow parameters into the fusion model and output the probability of coal wall risk; Based on the wall hanging risk probability, combined with the DCS parameter model output and the coal type switching time logic output, probability calibration is performed through DS evidence theory; The calibrated probabilities are combined using the Dempster combination rule to calculate the joint probability, and the conflict factor of each pair of probabilities is combined to obtain the final composite wall hanging probability. Make judgments and early warnings based on the final probability of wall hanging.

2. The method for early warning of raw coal accumulation on the raw coal bunker according to claim 1, characterized in that: The pretreatment includes: Eliminate background noise of voiceprint signal through LMS filtering algorithm; Convert the long-term non-stationary voiceprint signal into multiple short-term stationary signals through framing; Windowing is performed on multiple short-term stationary signals after framing; Through short-time Fourier transform, each frame signal is expanded from the time domain to the frequency domain; Take the energy spectrum of the short-time Fourier transform and calculate the logarithmic power spectrum.

3. The method for early warning of raw coal wall accumulation in a raw coal bunker according to claim 2, characterized in that: It also includes,feature enhancement of the pre-processed voiceprint signal; Pass the logarithmic power spectrum through the Mel filter bank to obtain the Fbank feature; Perform discrete cosine transform based on Fbank features to obtain static MFCC parameters; Dynamic features are obtained through differential operation based on static MFCC parameters; The static MFCC is combined with the dynamic features and the frame capability to obtain the voiceprint feature, which is a 39-dimensional MFCC map.

4. The method for early warning of raw coal accumulation on the wall of a raw coal bunker according to claim 3, characterized in that: Calculating the coal flow falling speed based on the voiceprint signal by using a cross-correlation time delay method includes: The coal flow falling speed is calculated based on the distance between the upper and lower signal acquisition devices of the tapered section and the cross-correlation peak delay value of the voiceprint signal.

5. The method for early warning of raw coal accumulation on the wall of a raw coal bunker according to claim 4, characterized in that: The fusion model includes: The voiceprint features, coal quality parameters and coal flow parameters are input into the input layer; The hidden layer is a 2-layer bidirectional LSTM, and the activation function is set to tanh; The fully connected layer outputs the risk probability of wall hanging.

6. The method for early warning of raw coal accumulation on the wall of a raw coal bunker according to claim 5, characterized in that: Based on the wall hanging risk probability, combined with the DCS parameter model output and the coal type switching time logic output, probability calibration is performed through DS evidence theory, including: The wall hanging risk probability output by the fusion model is mapped through the Sigmoid function to achieve probability calibration; For the wall hanging risk probability output by the DCS parameter model, ReLU linear normalization is used to map it to achieve probability calibration; For the probability of wall hanging risk output by the coal type switching time logic, probability calibration is achieved by setting thresholds and mutation rules.

7. The method for early warning of raw coal accumulation on the wall of a raw coal bunker according to claim 6, characterized in that: The calibrated probabilities are combined using the Dempster combination rule to calculate the joint probability. The conflict factor of each pair of probabilities is combined to obtain the final composite wall hanging probability, including: Combine the three probabilities in pairs, calculate the conflict factor for each pair of probabilities, and sum them up to get the total conflict factor; If the total conflict factor is greater than the preset threshold, an adaptive weight adjustment is performed to obtain the adjusted corresponding probability; The synthesis results are calculated based on the adjusted corresponding probabilities to obtain the final wall hanging probability.

8. A raw coal wall-mounting early warning system for a raw coal bunker, applying the method according to any one of claims 1 to 7, characterized in that: include: An acquisition module is used to acquire and preprocess the coal flow voiceprint signal to obtain the voiceprint feature; a calculation module, configured to calculate the falling velocity of the coal flow based on the voiceprint signal by using a cross-correlation time delay method; The fusion module is used to input voiceprint features, coal quality parameters, and coal flow parameters into the fusion model and output the probability of coal wall damage risk; A calibration module is used to perform probability calibration based on the wall hanging risk probability, combined with the DCS parameter model output and the coal type switching time logic output through the DS evidence theory; The synthesis probability module is used to calculate the joint probability of the calibrated probabilities using the Dempster combination rule, and combine the conflict factors of each pair of probabilities to obtain the final composite wall hanging probability; The early warning module is used to make judgments and early warnings based on the final probability of wall hanging.

9. A computer device comprising: memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method for early warning of raw coal hanging on the wall in the raw coal bin as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the method for early warning of raw coal sticking to the wall in a raw coal bin as described in any one of claims 1 to 7.

Images