Coal mill residual life prediction method and device, electronic equipment and storage medium

Abstract

The invention discloses a coal mill residual life prediction method and device, electronic equipment and a storage medium, and relates to the technical field of coal mills, and the method comprises the steps: collecting acoustic signals and operation state parameters in the operation process of a coal mill; based on the acoustic signals and the operation state parameters, frequency spectrum features reflecting the dynamic evolution process of the material fatigue state are extracted; a material fatigue damage model is constructed, the frequency spectrum features serve as input, model parameters are dynamically calibrated in combination with the actually measured material abrasion loss, and the current fatigue damage degree is calculated; and based on the current fatigue damage degree and the evolution trend thereof, predicting the residual service life of the coal mill, and outputting a residual service life prediction result including a fault type, a wear degree and the residual service life. Through the method, the technical effects of accurately quantifying the fatigue damage degree of the material, improving the prediction accuracy of the residual life of the coal mill and providing a comprehensive and reliable decision basis for equipment operation and maintenance are achieved.

Description

Technical Field

[0001] This application relates to the field of coal mill technology, and in particular to a method, apparatus, electronic device and storage medium for predicting the remaining life of a coal mill. Background Technology

[0002] As a key piece of equipment in the energy production field, the fatigue wear of core components such as grinding rollers and liners in coal mills directly affects operational safety and economy. Remaining life prediction is a core requirement for equipment fault diagnosis. Existing coal mill fault diagnosis technologies largely rely on empirical formulas to estimate wear levels, employing wear models with fixed parameters. Fault judgment is made through static spectral characteristics such as peak frequency and amplitude, with remaining life extrapolated from theoretical wear rates. However, these technologies have significant limitations: they can only qualitatively determine the existence of faults, failing to quantitatively assess the degree of material fatigue damage; they do not consider the impact of coal hardness variations on wear, resulting in poor model adaptability and difficulty in matching different coal types and operating conditions; spectral feature extraction lacks dynamism, failing to accurately reflect the gradual evolution of material fatigue; remaining life prediction is not calibrated with measured data, leading to significant discrepancies between theoretical calculations and actual results; and the lack of a coupling relationship between material fatigue and spectral evolution results in unclear physical meaning of diagnostic results, leading to blind maintenance planning, potential over-maintenance or sudden failures, and an inability to meet the actual needs of precise equipment operation and maintenance. Summary of the Invention

[0003] This application provides a method, apparatus, electronic device, and storage medium for predicting the remaining life of a coal mill. It addresses the problems in related technologies where reliance on a single signal or static feature, lack of measured data calibration in the model, inability to reflect the dynamic evolution of material fatigue, and incomplete output information lead to inaccurate fatigue damage quantification and large deviations in remaining life prediction.

[0004] According to a first aspect of this application, a method for predicting the remaining life of a coal mill is provided, comprising: Acoustic signals and operating status parameters of the coal mill during operation are collected; Based on the acoustic signal and the operating state parameters, spectral features reflecting the dynamic evolution of the material fatigue state are extracted; A material fatigue damage model is constructed, and the spectral features are used as input. The model parameters are dynamically calibrated in combination with the measured material wear to calculate the current fatigue damage degree. Based on the current fatigue damage level and its evolution trend, the remaining service life of the coal mill is predicted, and the remaining service life prediction result including fault type, wear degree and remaining service life is output.

[0005] According to a second aspect of this application, a coal mill remaining life prediction device is provided, comprising: The acquisition module is configured to acquire acoustic signals and operating status parameters during the operation of the coal mill; The extraction module is configured to extract spectral features that reflect the dynamic evolution of the material fatigue state based on the acoustic signal and the operating state parameters. The calculation module is configured to construct a material fatigue damage model and take the spectral features as input, and dynamically calibrate the model parameters in combination with the measured material wear amount to calculate the current fatigue damage degree; The prediction module is configured to predict the remaining service life of the coal mill based on the current fatigue damage level and its evolution trend, and output the remaining service life prediction result including the fault type, wear degree and remaining service life.

[0006] According to a third aspect of this application, an electronic device is provided, comprising: At least one processor; and memory that is communicatively connected to at least one processor; The memory stores instructions that can be executed by at least one processor, which enables the at least one processor to perform the coal mill remaining life prediction method described in the first aspect above.

[0007] According to a fourth aspect of this application, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the coal mill remaining life prediction method described in the first aspect above.

[0008] According to a fifth aspect of this application, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the coal mill remaining life prediction method as described in the first aspect above.

[0009] This application provides a method, apparatus, electronic device, and storage medium for predicting the remaining service life of a coal mill, comprising: acquiring acoustic signals and operating status parameters during the operation of the coal mill; extracting spectral features reflecting the dynamic evolution of material fatigue state based on the acoustic signals and operating status parameters; constructing a material fatigue damage model, and using the spectral features as input, dynamically calibrating the model parameters in conjunction with measured material wear to calculate the current fatigue damage degree; predicting the remaining service life of the coal mill based on the current fatigue damage degree and its evolution trend, and outputting a remaining service life prediction result including fault type, wear degree, and remaining service life. This application addresses the problems in related technologies where reliance on a single signal or static feature, lack of measured data calibration, inability to reflect the dynamic evolution of material fatigue, and incomplete output information lead to inaccurate fatigue damage quantification and large deviations in remaining life prediction. By collecting acoustic signals and operating parameters during the operation of the coal mill, spectral features reflecting the dynamic evolution of material fatigue are extracted, and a material fatigue damage model is constructed. This model is then used as input, combined with measured material wear, to dynamically calibrate model parameters to calculate the current fatigue damage degree. Based on this fatigue damage degree and its evolution trend, the remaining service life is predicted, and a prediction result including fault type, wear degree, and remaining life is output. This achieves the technical effect of accurately quantifying material fatigue damage, improving the accuracy of coal mill remaining life prediction, and providing a comprehensive and reliable decision-making basis for equipment operation and maintenance.

[0010] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0011] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating a method for predicting the remaining life of a coal mill, provided in an embodiment of this application; Figure 2 This is a flowchart illustrating another method for predicting the remaining life of a coal mill provided in an embodiment of this application. Figure 3 This is a flowchart illustrating another method for predicting the remaining life of a coal mill provided in an embodiment of this application. Figure 4 This is a schematic diagram of the structure of a coal mill remaining life prediction device provided in an embodiment of this application. Detailed Implementation

[0013] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0014] The following description, with reference to the accompanying drawings, outlines a method, apparatus, electronic device, and storage medium for predicting the remaining life of a coal mill according to embodiments of this application.

[0015] Figure 1 This is a flowchart illustrating a method for predicting the remaining life of a coal mill, as provided in an embodiment of this application.

[0016] like Figure 1 As shown, the method includes the following steps: Step 101: Collect acoustic signals and operating status parameters during the operation of the coal mill.

[0017] In some embodiments, during the operation of the coal mill, acoustic signals and operating parameters reflecting the equipment's operating status need to be collected simultaneously to provide basic data support for subsequent fatigue damage analysis and remaining life prediction. Acoustic signal acquisition is achieved using a high-precision capacitive acoustic pressure sensor, specifically designed to capture broadband noise generated by the core fault source—the grinding roller bearing—ensuring the targeted and effective signal acquisition. Two sensors are installed in a group, one on each side of the coal mill directly above the bearing housing, approximately 0.5m vertically from the bearing center, and fixed to the housing above the shaft seat. The probe faces the center of the bearing housing and maintains a 30° tilt angle with the housing surface to reduce interference from airflow noise on the acquired signal. This sound pressure sensor has a frequency response range of 10Hz-40kHz and a sensitivity of 10mV / Pa, which can comprehensively cover noise signals generated by faults such as grinding roller wear and bearing defects. During acquisition, two noise signals are synchronously sampled through a 24-bit high-precision AD converter with the sampling frequency set at 40kHz to ensure that high-frequency fault characteristics are not missed. The acquired noise signal is processed by anti-aliasing low-pass filtering with a filter cutoff frequency of 18kHz, and then converted into spectrum data by FFT transformation to complete the acquisition and preliminary processing of acoustic signals. Operating parameters include rotational speed, coal feed rate, and coal hardness. Rotational speed is acquired using a magnetoelectric sensor mounted on the spindle end cover, which measures the mill's rotational speed in real time. Its measurement range is 0-1500 r / min with an accuracy of ±0.1 r / min, accurately reflecting the equipment's operating speed. Coal feed rate is acquired using a weighing sensor installed at the bottom of the feed chute, measuring the mill's real-time load. Its measurement range is 0-50 t / h with an accuracy of ±0.2% FS, ensuring accurate load data. Coal hardness is acquired using a portable hardness tester, measured three times weekly at sampling points on the incoming coal conveyor belt. The maximum and minimum values ​​are discarded, and the average value is used as the coal hardness parameter for that period. This hardness tester has a measurement range of 100-1000 HV with an accuracy of ±5 HV, effectively capturing hardness differences between different coal types. Rotational speed and coal feed rate signals are acquired in real time at a frequency of 1 Hz to ensure reflection of dynamic load changes. All sensor signal lines are shielded and routed separately from the power cable to avoid electromagnetic interference affecting data quality. Data storage employs an hourly automatic storage strategy, with each set containing the average spectrum of two noise signals for that hour (calculated every 10 minutes, averaging after six spectrums), as well as the hourly averages of rotational speed and coal feed rate. Coal hardness parameters are uniformly entered into the system after weekly measurements. This acquisition method comprehensively and accurately obtains key data during the operation of the coal mill, reducing the impact of various interference factors on the data and providing a basis for subsequent analysis.

[0018] Step 102: Based on the acoustic signal and the operating state parameters, extract the spectral features that reflect the dynamic evolution of the material fatigue state.

[0019] In some embodiments, after acquiring the acoustic signals and operating status parameters of the coal mill, a series of targeted processing steps are required to extract spectral features that accurately reflect the dynamic evolution of material fatigue. First, noise spectrum data from the past 72 hours is retrieved from stored historical data. This data is stored in hourly sets, each containing the average spectrum collected and processed by two sound pressure sensors, covering a frequency range of 10Hz-40kHz. To eliminate irrelevant interference, focusing on fault features directly related to roller imbalance and liner wear, the 10Hz-1kHz frequency band is specifically selected. This band effectively avoids the influence of airflow noise above 1kHz on feature extraction, ensuring the relevance of subsequent analysis. Simultaneously, the rotational speed data from the past 72 hours is extracted from the operating status parameters, and the average rotational speed during this period is calculated, providing a basis for determining the characteristic frequencies.

[0020] Next, the amplitude half-life is calculated, a characteristic that directly reflects the signal amplitude attenuation caused by material fatigue. The target characteristic frequency is determined using a specific formula based on the average rotational speed. This frequency is directly related to the roller imbalance fault and is a key frequency point reflecting the material wear state. Amplitude data corresponding to this target frequency are extracted one by one from 72 sets of spectral data, forming 72 continuous amplitude time series. The least squares method is used to curve fit this series, with the fitting function being an exponential decay model. The fitting process involves at least 100 iterations until the sum of squared residuals is no greater than 10. -6 The convergence condition is determined to ensure fitting accuracy. After obtaining the attenuation coefficient through fitting, the amplitude half-life is calculated based on the logarithmic formula. When the half-life is greater than 100 hours, it indicates that the material is in a normal fatigue state. If it drops to 40 hours, it indicates that wear is intensified. This feature clearly shows the gradual change of fatigue state.

[0021] The spectral centroid migration rate was then calculated, as the change in the position of the spectral centroid is closely related to the evolution of material fatigue damage. For hourly spectral data, the spectral centroid for that hour was calculated using a specific formula: the weighted sum of frequency and corresponding spectral amplitude divided by the total amplitude, with the weights being the absolute values ​​of the spectral amplitudes. This yielded 72 spectral centroid values ​​from 72 sets of data. Using a 24-hour time interval, the rate of change of the spectral centroid within adjacent time intervals was calculated, resulting in 49 migration rate values. The average of these values ​​was taken as the final spectral centroid migration rate. Under normal operating conditions, this rate should be less than 0.1 Hz / h. A rate reaching 0.6 Hz / h indicates accelerated wear, enabling dynamic monitoring of the spectral distribution changes caused by material fatigue.

[0022] Finally, the two extracted spectral features were normalized to provide standardized data for subsequent model input. For the amplitude half-life, 30 hours was used as the wear limit half-life and 120 hours as the new device half-life, and they were mapped to the [0,1] interval through linear transformation. For the spectral centroid migration rate, 1.0 Hz / h was used as the wear aggravation threshold, and it was also mapped to the [0,1] interval. If the calculation result exceeded this interval during processing, 0 or 1 was taken as the feature value, respectively, to avoid outliers interfering with subsequent analysis. This feature extraction method can dynamically track the progressive evolution of material fatigue, providing a physically meaningful and reliable input basis for subsequent fatigue damage calculation, effectively making up for the inability of static features to reflect the evolution trend.

[0023] Step 103: Construct a material fatigue damage model and use the spectral features as input. Combine the measured material wear amount to dynamically calibrate the model parameters in order to calculate the current fatigue damage degree.

[0024] In some embodiments, when constructing a material fatigue damage model, the initial configuration of basic parameters must first be completed. Material property information of the grinding roller liner is extracted from the equipment register, and corresponding material constants are determined according to the material type. Different materials have specific values ​​for their material constants to ensure the model accurately reflects the fatigue characteristics of the material itself. Subsequently, standard stress amplitude calibration is performed by running the coal mill under no-load conditions for one hour, collecting the average rotational speed of the main shaft during this period, and calculating the standard stress amplitude using a specific formula, combined with the equipment's specific structural coefficients. This calculated stress amplitude is then stored in the model parameter library to provide a benchmark for subsequent stress amplitude calculations.

[0025] The model construction process must fully consider the impact of coal hardness on material wear. It utilizes the average coal hardness data collected this week, which has been averaged after multiple measurements to remove extreme values, ensuring accuracy. A preset formula correlates the standard stress amplitude with the average coal hardness, calculating the stress amplitude under actual operating conditions. This data is then synchronously updated to the material fatigue damage model, enabling it to adapt to hardness variations in different coal types and improving its operational adaptability.

[0026] The calculation of the number of cycles is a key step in the model operation. The rotational speed sampling data of the past hour is extracted from the operating status parameters, the average value of all sampling values ​​is taken, and then combined with the time parameters to calculate the fatigue cycle number of the material in that period through a formula. At the same time, the timestamp of each calculation and the corresponding number of cycles are recorded to provide data support for the calculation of cumulative damage.

[0027] During the model computation phase, the spectral features extracted and normalized in step 102 are used as input. These spectral features accurately reflect the dynamic evolution of the material's fatigue state and are closely related to the material's fatigue damage mechanism, providing the model with input basis with clear physical meaning. Simultaneously, measured material wear data are introduced to participate in the dynamic calibration of model parameters. The measured material wear data is obtained through dedicated testing equipment and, after outlier removal and averaging, possesses extremely high reliability.

[0028] A specific formula is used to establish a correspondence between the current fatigue damage degree and the wear amount, calculate the wear amount predicted by the model, and then compare it with the actual material wear amount. The damage evolution rate of the model is corrected according to the deviation ratio between the two, and the output layer weights of the neural network in the model are updated synchronously to achieve dynamic calibration of the model parameters, ensuring that the model always remains consistent with the actual operating state of the equipment.

[0029] Based on Miner's cumulative damage theory, the ratio of the calculated number of cycles to the fatigue life under the corresponding stress amplitude is calculated each time. This ratio is continuously accumulated from the start of the model, and the final cumulative value is the current fatigue damage degree. This construction and calibration method ensures that the fatigue damage degree calculation not only conforms to material fatigue theory but also fully integrates measured data and dynamic characteristics, effectively improving the accuracy of damage degree calculation and giving the results stronger physical meaning and interpretability, thus laying a reliable foundation for subsequent remaining life prediction.

[0030] Step 104: Based on the current fatigue damage level and its evolution trend, predict the remaining service life of the coal mill, and output the remaining service life prediction result including the fault type, wear degree and remaining service life.

[0031] In some embodiments, when predicting the remaining service life of a coal mill based on the current fatigue damage level and its evolution trend, it is first necessary to accurately obtain the damage evolution law. By retrieving historical fatigue damage data stored in the system, continuous damage records from the past 30 days are selected, and the difference between the latest damage level and the damage level 30 days ago is calculated. This difference is then divided by 30 days to obtain the damage evolution rate per unit time. This rate directly reflects the speed of material fatigue damage development and provides a core dynamic basis for remaining service life prediction. Subsequently, combining the current fatigue damage level and the damage evolution rate, the theoretical remaining service life is derived through a specific calculation formula. During the calculation process, the design service life boundary of the equipment must be fully considered. If the theoretically calculated remaining service life exceeds the design service life of the equipment, the hours corresponding to the design service life are used as the final remaining service life value to ensure that the prediction result conforms to the objective limitations of actual equipment use.

[0032] When outputting the remaining life prediction results, the system integrates multi-dimensional analysis data to form a complete output including fault type, wear level, and remaining life. The fault type is determined based on the correlation analysis between the current fatigue damage level and the previously extracted spectral dynamic features. By matching the feature patterns corresponding to different faults, specific fault forms such as grinding roller imbalance and liner wear are accurately located. The wear level is calculated by combining the current fatigue damage level with the specific wear coefficient of the grinding roller and liner material, and clearly presents the component's wear status as a percentage, allowing maintenance personnel to intuitively grasp the material fatigue level. The remaining life is clearly output in hours, and the system adopts a dual-axis display mode of "wear level - remaining life" to dynamically display the evolution trend of both, making it convenient for staff to quickly grasp the changes in equipment status.

[0033] The system automatically stores the characteristic parameters and diagnostic results hourly, supporting historical data tracing up to five years, providing data support for subsequent equipment status reviews and design improvements. Simultaneously, the system incorporates a tiered early warning mechanism, triggering corresponding warnings based on specific values ​​of remaining lifespan and wear level, ensuring maintenance personnel can respond promptly to potential fault risks. This prediction and output method fully integrates the dynamic changes in actual equipment operation, providing comprehensive and accurate information to offer a scientific and reliable basis for maintenance planning, effectively avoiding over-maintenance or sudden failures, and improving the safety and economy of equipment operation.

[0034] Compared with related technologies, this embodiment collects acoustic signals and operating status parameters during the operation of the coal mill; based on the acoustic signals and operating status parameters, it extracts spectral features reflecting the dynamic evolution of material fatigue state; it constructs a material fatigue damage model, and uses the spectral features as input, combined with measured material wear, to dynamically calibrate the model parameters to calculate the current fatigue damage degree; based on the current fatigue damage degree and its evolution trend, it predicts the remaining service life of the coal mill and outputs a remaining service life prediction result including fault type, wear degree, and remaining service life. This solves the problems in related technologies where reliance on a single signal or static feature, lack of measured data calibration, inability to reflect the dynamic evolution of material fatigue, and incomplete output information lead to inaccurate fatigue damage quantification and large deviations in remaining service life prediction. It achieves the technical effect of accurately quantifying the degree of material fatigue damage, improving the accuracy of coal mill remaining service life prediction, and providing a comprehensive and reliable decision-making basis for equipment operation and maintenance.

[0035] Figure 2 A flowchart illustrating another method for predicting the remaining life of a coal mill provided in this application embodiment includes the following steps: Step 201: An acoustic sensor is arranged near the bearing housing of the coal mill to collect broadband acoustic signals including the operating noise of the grinding rollers and liners, and to simultaneously collect the real-time rotational speed and coal feed parameters of the coal mill.

[0036] In some embodiments, acoustic sensors are arranged near the bearing housings of the coal mill to collect broadband acoustic signals including operating noise from the grinding rollers and liners, and simultaneously collect the real-time rotational speed and coal feed parameters of the coal mill. High-precision capacitive acoustic sensors are selected and installed in pairs directly above the bearing housings on both sides of the coal mill. The vertical height of the installation position from the center of the bearing is controlled at approximately 0.5m. The sensors are fixed to the housing above the bearing housing, with the probe facing the center of the bearing housing and maintaining a 30° tilt angle with the housing surface. This arrangement minimizes the interference of airflow noise on the collected signals, ensuring direct capture of the noise generated by the core fault source—the grinding roller bearing. This type of acoustic sensor has a wide frequency response range of 10Hz-40kHz and a sensitivity of 10mV / Pa, which can comprehensively cover broadband noise signals generated during operation due to faults such as grinding roller wear and bearing defects. During acquisition, a 24-bit high-precision AD converter is used to synchronously sample the two acoustic signals at a sampling frequency of 40kHz to ensure that high-frequency fault characteristics are not missed. Meanwhile, a magnetoelectric speed sensor is installed at the end cover of the coal mill's main shaft to acquire the rotational speed of the mill disc in real time. This sensor has a measurement range of 0-1500 r / min and an accuracy of ±0.1 r / min, accurately reflecting the equipment's operating speed. A weighing-type coal feed rate sensor is installed at the bottom of the coal chute to measure the real-time load of the coal mill. Its measurement range is 0-50 t / h and its accuracy is ±0.2% FS, ensuring the accuracy of the load data. The speed and coal feed rate sensors are installed according to standard procedures. All sensor signal lines are shielded and routed separately from the power cable to avoid electromagnetic interference affecting data quality. After acoustic signal acquisition, anti-aliasing low-pass filtering is performed with a filter cutoff frequency set to 18 kHz. Then, FFT transformation is used to convert the signals into spectral data. Speed ​​and coal feed rate signals are acquired in real time at a frequency of 1 Hz to ensure dynamic reflection of equipment load changes. All acquired broadband acoustic signals, real-time speed, and coal feed rate data are automatically stored hourly, providing comprehensive and accurate basic data support for subsequent material fatigue state analysis.

[0037] Step 202: Periodically obtain the hardness parameters of the coal fed into the furnace and use them as input conditions for model correction.

[0038] In some embodiments, the hardness parameters of the coal fed into the furnace are periodically acquired and used as input conditions for model correction. A portable coal hardness tester is selected as the measuring tool. This hardness tester has a measurement range of 100-1000 HV and an accuracy of ±5 HV, which can accurately capture the hardness differences of different coal types. The measurement is carried out periodically. At a fixed time each week, maintenance personnel conduct on-site measurements at the coal conveyor belt sampling point. Each measurement is no less than 3 times. After the measurement is completed, the maximum and minimum values ​​in the measurement data are removed, and the remaining valid data are arithmetically averaged to obtain the average hardness parameter of the coal fed into the furnace within that period. The calculated average hardness parameter is then uniformly entered into the system database and correlated with the coal mill operation data collected during the same period, serving as an important correction input condition for the material fatigue model. By periodically acquiring the hardness parameters of the coal fed into the furnace, the hardness changes of different batches of coal fed into the furnace can be captured in a timely manner, providing a key correction basis for the calculation of stress amplitude in the subsequent material fatigue damage model. This allows the model to dynamically adapt to the hardness differences of different coal types, effectively improving the model's adaptability to complex working conditions and avoiding fatigue damage assessment deviations caused by changes in coal hardness.

[0039] Step 203: Identify characteristic frequencies associated with roller imbalance or liner wear from historical spectrum data of a preset time length.

[0040] In some embodiments, when identifying characteristic frequencies associated with roller imbalance or liner wear from historical spectrum data of a preset time period, it is first necessary to clarify the source and range of the historical spectrum data. This data is taken from the acoustic signal processing results collected and stored previously. The preset time period is set to 72 hours, corresponding to 72 sets of spectrum data stored regularly once per hour. Each set of data includes the average spectrum after anti-aliasing low-pass filtering and FFT transformation of two sound pressure sensors, with the original frequency covering 10Hz-40kHz. To eliminate irrelevant interference factors and ensure the relevance of characteristic frequency identification, the historical spectrum data needs to be frequency band filtered, focusing on the 10Hz-1kHz frequency band, which is directly related to roller imbalance and liner wear. This frequency band can effectively avoid the masking of fault characteristics by airflow noise above 1kHz, making subsequent frequency identification more accurate. At the same time, real-time rotational speed data of the past 72 hours is extracted from the operating status parameters. These data are collected at a frequency of 1Hz, and the average of 60 sample values ​​is taken to obtain the hourly average rotational speed. The 72-hour average rotational speed is then further averaged to obtain the overall average rotational speed of the past 72 hours. Based on the frequency characteristics of grinding roller imbalance faults, a specific formula is used to calculate characteristic frequencies. This formula uses average rotational speed as the core parameter and, through the inherent correlation between rotational speed and frequency, directly identifies the target frequencies corresponding to grinding roller imbalance and liner wear. For example, when the average rotational speed is 300 r / min, the calculated characteristic frequency is 5 Hz. This frequency accurately reflects the vibration and noise characteristics of the grinding roller and liner caused by fatigue wear or imbalance during operation, laying the foundation for subsequent extraction of dynamic evolution spectrum features. This identification method, by focusing on key frequency bands and combining them with equipment operating parameters, ensures a strong correlation between characteristic frequencies and target faults, avoids interference from irrelevant frequencies, and improves the targeting and reliability of subsequent feature extraction.

[0041] Step 204: Calculate the attenuation characteristics of the amplitude over time at the characteristic frequency to obtain the amplitude half-life characteristics.

[0042] In some embodiments, when calculating the amplitude decay characteristics at a characteristic frequency over time to obtain the amplitude half-life characteristics, it is necessary to first extract the amplitude data corresponding to the characteristic frequency identified in step 203 from 72 sets of historical spectrum data one by one, forming an amplitude time series containing 72 data points. Each data point corresponds to the characteristic frequency amplitude at a specific time point, completely recording the amplitude change over a period of time. To accurately capture the amplitude decay law, the least squares method is used to curve fit the time series. The function selected for fitting is the exponential decay model, which can accurately reflect the gradual change trend of the characteristic frequency amplitude during material fatigue wear. During the fitting process, a sufficient number of iterations must be ensured, with no less than 100 iterations. At the same time, strict convergence conditions are set, requiring the sum of squared residuals to be no greater than 10. -6 This ensures that the fitted curve closely matches the actual amplitude variation, minimizing fitting errors. After obtaining the attenuation coefficient from the fitted curve, the amplitude half-life is calculated based on the logarithmic principle; this is the time required for the amplitude to decay to half its initial value. The magnitude of the amplitude half-life directly reflects the rate of material fatigue development. When the half-life is greater than 100 hours, it indicates that the grinding roller and liner are in a normal fatigue state with moderate wear. When the half-life drops to around 40 hours, it indicates increased wear and a deterioration in the material fatigue state. This characteristic calculation method dynamically captures the amplitude decay over time, transforming the abstract material fatigue process into a quantifiable time parameter, clearly presenting the gradual evolution trend of the fatigue state. It compensates for the inability of static features to reflect dynamic changes, providing a dynamic indicator with clear physical meaning for subsequent fatigue damage calculations.

[0043] Step 205: Calculate the migration rate of the centroid of the spectrum within the preset frequency band in the historical spectrum data over time.

[0044] In some embodiments, when calculating the migration rate of the spectral centroid within a preset frequency band in historical spectrum data over time, the preset frequency band is first defined as 10Hz-1kHz. This band has been confirmed in previous screening to be closely related to roller imbalance and liner wear, and can effectively reflect the spectral distribution changes caused by material fatigue. For each set of 72 sets of historical spectrum data, the spectral centroid within the preset frequency band is calculated according to a specific formula. This formula uses frequency as a variable and the absolute value of the corresponding spectral amplitude as a weight, and obtains the spectral centroid value of each set of data through a weighted average, ultimately forming a time series containing 72 spectral centroid values, each value corresponding to the core position of the spectral distribution for one hour. To accurately reflect the changing trend of the spectral centroid, a 24-hour time interval is set. Two values ​​from adjacent 24-hour intervals are selected from the 72 spectral centroid values, and the difference is calculated to obtain the change in the spectral centroid within adjacent time periods. This change is then divided by the time interval to obtain a single migration rate value. A total of 49 migration rate values ​​are calculated in this way. Finally, these values ​​are averaged to obtain the final spectral centroid migration rate. The positional change of the spectral centroid is closely related to material fatigue damage. As the wear of the grinding roller and liner intensifies, the spectral distribution undergoes a regular shift, leading to a change in the position of the spectral centroid. The migration rate directly reflects the speed of this shift. Under normal operating conditions, the migration rate should be less than 0.1 Hz / h. When the migration rate reaches 0.6 Hz / h, it indicates that wear is intensifying. This calculation method, by quantifying the migration pattern of the spectral centroid, further supplements the characteristic dimension reflecting the dynamic evolution of material fatigue. This, along with the amplitude half-life characteristic, provides a more comprehensive and accurate characterization of the fatigue state, offering richer dynamic data for subsequent model input.

[0045] Step 206: Periodically measure the actual wear of the grinding roller or liner using a non-contact distance measuring device.

[0046] In some embodiments, when periodically measuring the actual wear of the grinding roller or liner using a non-contact ranging device, a high-precision laser displacement sensor is selected as the core measuring equipment. This device can accurately acquire distance data without contacting the surface of the grinding roller or liner, avoiding interference or damage to operating components, and effectively avoiding wear or errors that may occur with contact measurements. Two laser displacement sensors are fixed on a dedicated bracket, and the installation angle and position are strictly adjusted to ensure that the laser beam is perpendicular to the generatrix of the grinding roller surface, and the measurement distance is controlled within 500mm to ensure measurement accuracy. A proximity switch is installed along the path traversed by the edge of the grinding roller to accurately identify specific reference points on the grinding roller and provide a trigger signal for synchronous sampling. The measurement adopts a periodic manual triggering method. When wear data needs to be acquired, the grinding roller is slowly rotated by manually turning the wheel or jogging the motor. Whenever the proximity switch detects the reference point passing by, it triggers the laser sensor to sample. During one revolution of the grinding roller, the sensor uniformly collects 20 data points. The system preprocesses these data, first removing abrupt abnormal values ​​caused by factors such as slag adhesion, and then performing an arithmetic average on the remaining valid data to obtain the current actual radius of the grinding roller. The current radius is compared with the initial radius recorded in the equipment log or the radius of the last measurement. The difference between the two is the actual wear amount of the grinding roller or liner within this cycle. This data is then stored in the system database, providing direct experimental basis for subsequent model calibration. This non-contact measurement method ensures the safety of the measurement process and improves data accuracy through multi-point sampling, outlier removal, and averaging, ensuring that the actual wear amount truly reflects the wear state of the component.

[0047] Step 207: Compare the actual wear amount with the theoretically predicted wear amount calculated based on the fatigue damage degree.

[0048] In some embodiments, when comparing the actual wear amount with the theoretically predicted wear amount calculated based on fatigue damage degree, the calculation logic of the theoretically predicted wear amount is first clarified. The corresponding wear coefficient is determined according to the material properties of the grinding roller or liner. Different materials have specific values ​​for the wear coefficient; for example, the wear coefficient for high-chromium cast iron is set to 5mm, meaning that when the fatigue damage degree reaches 1, the theoretical wear amount is 5mm. Based on the calculated current fatigue damage degree, the theoretically predicted wear amount is obtained by multiplying the wear coefficient by the fatigue damage degree. Then, the actual wear amount data measured and stored in step 206 is retrieved. This data is the average of the measurement results from two laser displacement sensors and has extremely high reliability after outlier processing. During the comparison process, the focus is on calculating the ratio of the actual wear amount to the theoretically predicted wear amount. This also helps analyze the absolute and relative deviations between the two, clearly quantifying the degree of difference between the theoretical calculation and the actual situation. This comparison process establishes a direct correlation between the theoretical calculation results of fatigue damage and the actual wear state of the component, clarifies the direction and magnitude of the deviation between the model prediction and the actual situation, provides an intuitive and accurate basis for subsequent model parameter correction, avoids model distortion caused by blind correction, and ensures that the correction process is clearly targeted.

[0049] Step 208: Based on the comparison results, adaptively correct the damage evolution rate parameter in the fatigue damage model.

[0050] In some embodiments, when adaptively correcting the damage evolution rate parameter in the fatigue damage model based on the comparison results, the ratio of the actual wear amount to the theoretically predicted wear amount obtained in step 207 is used as the core correction basis to construct a dynamic correction logic. If the actual wear amount is greater than the theoretically predicted wear amount, it indicates that the current damage evolution rate of the model is lower than the actual fatigue development rate of the component, and the damage evolution rate needs to be increased by a multiple of the ratio; if the actual wear amount is less than the theoretically predicted wear amount, the damage evolution rate is reduced accordingly, and the specific correction formula is that the new damage evolution rate is equal to the original damage evolution rate multiplied by the ratio. While correcting the damage evolution rate, the output layer weights of the BP neural network in the fatigue damage model are updated simultaneously, so that the prediction results of the neural network can quickly adapt to the changes in the actual wear situation, forming a closed-loop adaptive mechanism of "actual measurement comparison - parameter correction - model optimization". This correction method does not require manual intervention and can automatically adjust the model parameters according to the periodically measured wear data, ensuring that the damage evolution rate is always consistent with the actual operating conditions of the coal mill and the fatigue state of the material, effectively making up for the defect that fixed parameter models cannot adapt to changes in operating conditions. Through continuous adaptive correction, the model's prediction accuracy is constantly improved, and the deviation between theoretical calculations and actual conditions is controlled within a small range, providing more reliable model support for subsequent remaining life prediction and further enhancing the practicality and accuracy of the entire diagnostic solution.

[0051] Step 209: Calculate the daily average evolution rate of the damage degree based on the current fatigue damage degree and its historical data.

[0052] In some embodiments, calculating the daily average evolution rate of damage severity relies on historical fatigue damage severity data stored in the system over a long period. This data is continuously recorded and categorized at an hourly frequency, supporting historical backtracking up to five years to ensure a sufficient sample size to reflect the true evolution trend. First, current fatigue damage severity data and historical fatigue damage severity data from the past 30 days are retrieved from the database. The 30-day time span is chosen because it covers typical operating conditions of the equipment and effectively avoids interference from short-term data fluctuations in the rate calculation, ensuring the stability of the results. The retrieved 30-day historical data is preprocessed to extract the daily average fatigue damage severity, reducing bias caused by random factors in hourly data. Then, the current fatigue damage severity value is subtracted from the fatigue damage severity value from 30 days ago to obtain the total change in damage severity during that period. Finally, the total change is divided by 30 days to obtain the daily average evolution rate of damage severity. This rate directly reflects the recent development pace of fatigue damage in coal mill materials. The data calculation process fully relies on continuous and reliable historical records, avoiding the limitations of data from a single point in time. It can accurately capture the real trend of damage development and provide stable and realistic core parameter support for the subsequent theoretical calculation of remaining life.

[0053] Step 210: Calculate the theoretical remaining life based on the current fatigue damage degree and damage evolution rate.

[0054] In some embodiments, when calculating the theoretical remaining life based on the current fatigue damage degree and damage evolution rate, the gradual development law of material fatigue damage should be the core basis. First, the fatigue damage degree is defined as ranging from 0 to 1, where 1 represents the material reaching its fatigue limit and unable to continue safe operation. Therefore, the remaining damage amount is the difference between 1 and the current fatigue damage degree, which directly reflects the upper limit of fatigue wear that the coal mill can still withstand. Dividing the remaining damage amount by the previously calculated daily average evolution rate yields the time required for the remaining damage amount, expressed in days. This time is then converted to hours by multiplying by 3600, which better reflects equipment operation and maintenance habits. This is the theoretical remaining life. During the calculation, the current fatigue damage degree is a precise value dynamically calibrated based on measured wear, and the daily average evolution rate is derived from 30 days of historical data. This ensures the reliability of the two core parameters, making the calculation of the theoretical remaining life not merely a theoretical extrapolation, but closely integrated with the fatigue evolution law of actual equipment operation. This effectively reduces the deviation between purely theoretical calculations and actual conditions, making the theoretical results more valuable.

[0055] Step 211: Compare the theoretical remaining life with the design life of the coal mill, and take the smaller value as the final prediction result of the remaining life.

[0056] In some embodiments, when comparing the theoretical remaining life with the design life of the coal mill and taking the smaller value as the final remaining life prediction result, it is first necessary to clarify the core significance of the design life. It is the upper limit of safe use determined by the equipment at the time of manufacture based on factors such as material properties, structural strength, and operating conditions. It is an important reference standard to ensure the long-term stable operation of the equipment. This data can be directly retrieved from the equipment ledger or factory technical documents to ensure its authority and accuracy. The reason for choosing the smaller value between the two is that if the theoretical remaining life exceeds the design life, even if the material fatigue damage has not reached its limit, the equipment may still face potential risks such as structural aging and performance degradation due to long-term operation. Using the equipment beyond the design life will significantly increase the probability of failure, which does not meet the requirements of safe operation and maintenance. On the other hand, when the theoretical remaining life is shorter than the design life, it indicates that the development rate of material fatigue damage is faster than expected, and the operation and maintenance plan should be formulated based on the actual evolution trend. This method of value determination fully considers the actual evolution of material fatigue damage and strictly adheres to the design safety boundaries of the equipment. It avoids the risk of overuse that may result from relying solely on fatigue damage calculations, ensuring that the final remaining life prediction results are both consistent with actual working conditions and have solid safety guarantees, providing a more comprehensive and reliable basis for operation and maintenance decisions.

[0057] Figure 3 A flowchart illustrating another method for predicting the remaining life of a coal mill provided in this application embodiment includes the following steps: Step 301: Collect acoustic signals and operating status parameters during the operation of the coal mill.

[0058] Step 302: Based on the acoustic signal and the operating state parameters, extract the spectral features that reflect the dynamic evolution of the material fatigue state.

[0059] Step 303: Construct a material fatigue damage model, and use the spectral features as input. Combine the measured material wear amount to dynamically calibrate the model parameters in order to calculate the current fatigue damage degree.

[0060] Step 304: Based on the current fatigue damage level and its evolution trend, predict the remaining service life of the coal mill, and output the remaining service life prediction result including the fault type, wear degree and remaining service life.

[0061] For a description of steps 301-304, please refer to the description of steps 101-104 in the above embodiment. This embodiment will not repeat them in detail.

[0062] Step 305: Establish a historical diagnostic database and store the spectral characteristics, fatigue damage degree, and remaining life prediction results in time series.

[0063] In some embodiments, a historical diagnostic database is established to store spectral characteristics, fatigue damage degree, and remaining life prediction results in a time-series manner, providing solid support for subsequent trend analysis and data backtracking. The database is built on high-capacity, high-read / write-speed storage devices and employs a partitioned storage strategy to ensure the orderly management of data. Specifically, dedicated areas are allocated to store various core data: spectral characteristics include the extracted and normalized amplitude half-life, spectral centroid migration rate, and corresponding raw calculation data; fatigue damage degree includes the cumulative damage degree D value obtained from each calculation and key intermediate parameters in the calculation process; and remaining life prediction results fully record the remaining life hours, fault type determination results, and wear percentage for each output. All data is strictly stored in a time-series manner, forming independent data record packages on an hourly basis. Each record package is labeled with a precise timestamp, coal mill number, and corresponding operating condition information (such as coal hardness, average rotational speed, and average coal feed rate), facilitating rapid retrieval by time range, equipment number, and other dimensions. The database supports the retention of historical data for up to five years, and features data backup and recovery capabilities to ensure the security and integrity of data storage. It also employs a standardized data format, ensuring compatibility with subsequent data analysis tools and enabling effective traceability and utilization of each data set. This provides a comprehensive data foundation for research on material fatigue evolution and equipment condition review. The establishment of this database transforms scattered diagnostic data into a systematic resource reserve, preventing data loss and providing a reliable guarantee for subsequent trend curve generation and in-depth analysis.

[0064] Step 306: Based on the historical diagnostic database, generate a material fatigue evolution trend curve to support maintenance decisions and equipment optimization.

[0065] In some embodiments, material fatigue evolution trend curves are generated based on historical diagnostic databases, providing an intuitive and scientific basis for maintenance decisions and equipment optimization and upgrades. During data processing, corresponding spectral characteristics, fatigue damage, wear degree, and remaining life data are extracted from the database according to preset time spans (such as daily, weekly, monthly, or specific operating cycles). Multi-dimensional evolution trend curves are constructed using data visualization technology. Curve types include single-parameter time evolution curves (such as curves showing changes in fatigue damage over operating time, and time-series fluctuation curves of spectral centroid migration rate) and curves comparing related parameters (such as dual-axis linkage curves of wear degree and remaining life), clearly presenting the changing patterns and interrelationships of each core parameter. Maintenance personnel can intuitively observe the gradual development process of material fatigue through the curves, such as determining whether fatigue damage is accelerating or whether there are abnormal abrupt changes in spectral characteristics, thereby accurately grasping the rhythm of changes in equipment health status and providing data support for developing targeted maintenance plans. When the curve shows an accelerated increase in wear degree and a rapid reduction in remaining life, maintenance can be scheduled in advance to avoid sudden failures; when the curve shows a stable evolution trend, the maintenance cycle can be extended to reduce over-maintenance costs. Meanwhile, the long-accumulated evolution trend curves can provide data feedback for equipment optimization. Designers can analyze the curve characteristics under different operating stages and coal quality conditions to summarize the fatigue failure laws of core components such as grinding roller liners, thereby optimizing material selection, structural design, or operating parameter settings, and improving the overall durability and operational stability of the equipment. These trend curves transform abstract data into intuitive visual information, reducing the difficulty of operation and maintenance decisions, and providing strong data analysis support for the entire lifecycle management and optimization of equipment.

[0066] Figure 4 This is a schematic diagram of the structure of a coal mill remaining life prediction device provided in an embodiment of this application, as shown below. Figure 4 As shown, it includes: acquisition module 401, extraction module 402, calculation module 403, and prediction module 404.

[0067] The acquisition module 401 is configured to acquire acoustic signals and operating status parameters during the operation of the coal mill; Extraction module 402 is configured to extract spectral features reflecting the dynamic evolution of material fatigue state based on the acoustic signal and the operating state parameters; The calculation module 403 is configured to construct a material fatigue damage model and take the spectral features as input, and dynamically calibrate the model parameters in combination with the measured material wear amount to calculate the current fatigue damage degree; The prediction module 404 is configured to predict the remaining service life of the coal mill based on the current fatigue damage level and its evolution trend, and output the remaining service life prediction result including the fault type, wear degree and remaining service life.

[0068] In some examples of this embodiment, the acquisition module 401 is specifically configured to arrange acoustic sensors near the bearing seat of the coal mill to acquire broadband acoustic signals including the operating noise of the grinding rollers and liners, and simultaneously acquire the real-time rotational speed and coal feed parameters of the coal mill; periodically acquire the hardness parameters of the coal fed into the furnace, and use them as input conditions for model correction.

[0069] In some examples of this embodiment, the extraction module 402 is specifically configured to identify characteristic frequencies associated with roller imbalance or liner wear from historical spectrum data of a preset time length; calculate the attenuation characteristics of the amplitude at the characteristic frequency over time to obtain the amplitude half-life characteristics; and calculate the migration rate of the spectral centroid within the preset frequency band in the historical spectrum data over time.

[0070] In some examples of this embodiment, the calculation module 403 is specifically configured to periodically measure the actual wear of the grinding roller or liner using a non-contact ranging device; compare the actual wear with the theoretically predicted wear calculated based on the fatigue damage degree; and adaptively correct the damage evolution rate parameter in the fatigue damage model according to the comparison result.

[0071] In some examples of this embodiment, the prediction module 404 is specifically configured to calculate the daily average evolution rate of the damage degree based on the current fatigue damage degree and its historical data; calculate the theoretical remaining life according to the current fatigue damage degree and the damage evolution rate; compare the theoretical remaining life with the design life of the coal mill, and take the smaller value as the final prediction result of the remaining life.

[0072] It should be noted that other corresponding descriptions of the functional units involved in the coal mill remaining life prediction device provided in this embodiment can be found in [reference needed]. Figure 1 , Figure 2 and Figure 3 The corresponding descriptions in [the document] will not be repeated here.

[0073] Based on the above, Figure 1 , Figure 2 and Figure 3 The embodiment of this paper presents a method for predicting the remaining life of a coal mill. Correspondingly, this embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method. Figure 1 , Figure 2 and Figure 3 This paper presents a method for predicting the remaining life of a coal mill.

[0074] Based on the above, Figure 1 , Figure 2 and Figure 3The embodiment illustrates a method for predicting the remaining life of a coal mill. Correspondingly, this embodiment also provides a computer program product storing a computer program that, when executed by a processor, implements the aforementioned method. Figure 1 , Figure 2 and Figure 3 This paper presents a method for predicting the remaining life of a coal mill.

[0075] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0076] Based on the above, Figure 1 , Figure 2 and Figure 3 A method for predicting the remaining life of a coal mill is shown, and Figure 4 To achieve the above objectives, the present application also provides an electronic device, such as a personal computer or a server, in the illustrated virtual device embodiment. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to implement the above-described virtual device. Figure 1 , Figure 2 and Figure 3 This paper presents a method for predicting the remaining life of a coal mill.

[0077] In some embodiments, the aforementioned physical device may further include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, an input unit such as a keyboard, etc., and optionally, a USB interface, a card reader interface, etc. In some embodiments, the network interface may include a standard wired interface, a wireless interface (such as a Wi-Fi interface), etc.

[0078] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0079] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0080] The above are merely specific embodiments of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for predicting the remaining life of a coal mill, characterized in that, include: Acoustic signals and operating status parameters of the coal mill during operation are collected; Based on the acoustic signal and the operating state parameters, spectral features reflecting the dynamic evolution of the material fatigue state are extracted; A material fatigue damage model is constructed, and the spectral features are used as input. The model parameters are dynamically calibrated in combination with the measured material wear to calculate the current fatigue damage degree. Based on the current fatigue damage level and its evolution trend, the remaining service life of the coal mill is predicted, and the remaining service life prediction result including fault type, wear degree and remaining service life is output.

2. The method for predicting the remaining life of a coal mill according to claim 1, characterized in that, The acoustic signals and operating status parameters collected during the operation of the coal mill include: Acoustic sensors are arranged near the bearing housing of the coal mill to collect broadband acoustic signals including the operating noise of the grinding rollers and liners, and to simultaneously collect the real-time speed and coal feed parameters of the coal mill. The hardness parameters of the coal fed into the furnace are periodically acquired and used as input conditions for model correction.

3. The method for predicting the remaining life of a coal mill according to claim 1, characterized in that, The step of extracting spectral features reflecting the dynamic evolution of material fatigue state based on the acoustic signal and the operating state parameters includes: Identify characteristic frequencies associated with roller imbalance or liner wear from historical spectrum data of a preset time period; The amplitude decay characteristics over time at the characteristic frequency are calculated to obtain the amplitude half-life characteristics; Calculate the migration rate of the centroid of the spectrum within the preset frequency band in the historical spectrum data as a function of time.

4. The method for predicting the remaining life of a coal mill according to claim 1, characterized in that, The process involves constructing a material fatigue damage model, using the spectral characteristics as input, and dynamically calibrating the model parameters based on measured material wear to calculate the current fatigue damage level. This includes: The actual wear of the grinding roller or liner is periodically measured using a non-contact distance measuring device. The actual wear amount is compared with the theoretically predicted wear amount calculated based on the fatigue damage degree; Based on the comparison results, the damage evolution rate parameter in the fatigue damage model is adaptively corrected.

5. The method for predicting the remaining life of a coal mill according to claim 1, characterized in that, Based on the current fatigue damage level and its evolution trend, the remaining service life of the coal mill is predicted, and a remaining service life prediction result including fault type, wear degree, and remaining service life is output, including: Based on the current fatigue damage level and its historical data, calculate the daily average evolution rate of the damage level; Calculate the theoretical remaining life based on the current fatigue damage level and damage evolution rate; The theoretical remaining life is compared with the design life of the coal mill, and the smaller value is taken as the final prediction result of the remaining life.

6. The method for predicting the remaining life of a coal mill according to claim 1, characterized in that, Also includes: Establish a historical diagnostic database to store the spectral characteristics, fatigue damage degree, and remaining life prediction results in a time series. Based on the historical diagnostic database, a material fatigue evolution trend curve is generated to support maintenance decisions and equipment optimization.

7. A device for predicting the remaining life of a coal mill, characterized in that, include: The acquisition module is configured to acquire acoustic signals and operating status parameters during the operation of the coal mill; The extraction module is configured to extract spectral features that reflect the dynamic evolution of the material fatigue state based on the acoustic signal and the operating state parameters. The calculation module is configured to construct a material fatigue damage model and take the spectral features as input, and dynamically calibrate the model parameters in combination with the measured material wear amount to calculate the current fatigue damage degree; The prediction module is configured to predict the remaining service life of the coal mill based on the current fatigue damage level and its evolution trend, and output the remaining service life prediction result including the fault type, wear degree and remaining service life.

8. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the coal mill remaining life prediction method according to any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to execute the coal mill remaining life prediction method according to any one of claims 1-6.

10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the method for predicting the remaining life of a coal mill according to any one of claims 1-6.

Images