Online oil detection system and method for cement ball milling

Abstract

The invention discloses an on-line oil liquid detection system and method for cement ball milling, and relates to the technical field of industrial equipment state monitoring and fault diagnosis, the system comprises a data acquisition module, a data processing module, a data processing module and a data processing module, the data acquisition module adopts a multi-channel vibration sensor array to perform synchronous data acquisition on positions of a ball mill bearing seat, a gear box and a transmission shaft; an original vibration signal matrix containing multi-component coupling information is obtained through high-frequency sampling, if the sampling frequency is lower than ten times of the wear characteristic frequency, sampling parameters are automatically adjusted until the Nyquist theorem requirement is met, and the fault early warning and model self-adaption module adapts to the change of the equipment operation condition by updating the model parameters in real time; according to the online oil detection system and method for the cement ball mill, the equipment operation reliability and maintenance efficiency are effectively improved, the service life is prolonged, the maintenance cost is reduced, and data support is provided for preventive maintenance decision making of the ball mill.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial equipment condition monitoring and fault diagnosis, and particularly relates to an on-line oil detection system and method for cement ball mills. Background Art

[0002] As a core equipment in heavy industries such as mineral processing, cement production, and chemical engineering, the stable operation of a ball mill is directly related to the efficiency and safety of the entire production line. Monitoring the wear condition of key components such as bearings and gears is of decisive significance for preventing equipment failures and reducing maintenance costs. With the continuous improvement of industrial automation, the demand for real-time control of equipment health status is becoming increasingly urgent. Traditional wear monitoring methods mainly rely on regular shutdown inspections and manual experience judgments. This passive maintenance mode not only causes production interruption losses but also makes it difficult to accurately grasp the actual wear process of components. Although some existing on-line monitoring technologies can collect basic signals such as vibration and temperature, their signal analysis capabilities are limited under complex working conditions and often cannot provide accurate wear degree evaluations and reliable fault warnings. The composite vibration signals generated by multiple key components operating simultaneously inside the ball mill are coupled with each other, forming an extremely complex signal environment. When multiple wear modes coexist, such as bearing rolling element wear, gear tooth surface spalling, and transmission shaft imbalance, the characteristic signals of each component show severe aliasing in the frequency domain and time domain. This signal aliasing directly leads to difficulties in extracting wear characteristics, making it difficult for traditional analysis methods to accurately separate the wear information of specific components from the composite signals. More critically, the signal characteristics in different wear stages show non-linear variation laws. Weak early wear signals are often masked by the normal operation noise of the equipment, and when the wear signals are obvious enough, the components may be close to the failure critical point. The non-linear characteristics of this wear evolution process make it extremely difficult to establish an accurate wear state assessment model, and existing methods cannot achieve quantitative evaluation of wear degree and reliable prediction of development trends. Summary of the Invention

[0003] The purpose of the present invention is to provide an on-line oil detection system and method for cement ball mills to solve the problems existing in the prior art.

[0004] To achieve the above purpose, the present invention provides the following technical solution: An on-line oil detection system for cement ball mills, the system includes: A data acquisition module that synchronously acquires data at the positions of the ball mill bearing housing, gearbox, and transmission shaft using a multi-channel vibration sensor array, and obtains an original vibration signal matrix containing multi-component coupling information through high-frequency sampling. If the sampling frequency is lower than ten times the wear characteristic frequency, the sampling parameters are automatically adjusted until the Nyquist theorem requirements are met; The wear assessment and trend analysis module obtains real-time wear assessment results for condition monitoring based on the original vibration signal matrix. It performs trend analysis on the wear assessment results within a continuous time period through the sliding time window technique, fits the wear evolution curve using the least squares method, calculates the wear development rate and acceleration based on the curve slope and second derivative, and obtains wear evolution law parameters for prediction analysis; The fault warning and model adaptation module establishes a fault warning mechanism based on the wear evolution law parameters. If the wear development rate exceeds the safety threshold or the predicted remaining life is lower than the maintenance cycle, it triggers a warning signal and generates maintenance suggestions, and adapts to changes in the equipment operating conditions by real-time updating the model parameters.

[0005] Preferably, the wear assessment and trend analysis module obtaining real-time wear assessment results for condition monitoring based on the original vibration signal matrix includes: Performing preprocessing operations on the original vibration signal matrix, using a band-pass filter to remove power frequency interference and high-frequency noise, and eliminating random noise components through the time-domain synchronous averaging algorithm to obtain a purified multi-channel vibration signal dataset for subsequent analysis and processing.

[0006] Preferably, the wear assessment and trend analysis module obtaining real-time wear assessment results for condition monitoring based on the original vibration signal matrix further includes: Performing blind source separation processing on the purified multi-channel vibration signal dataset through the independent component analysis algorithm, decomposing the coupled composite signal into mutually independent source signal components according to the statistical independence principle. If the number of separated signal components exceeds the preset threshold, the kurtosis criterion is used to screen out the main components related to wear.

[0007] Preferably, the wear assessment and trend analysis module obtaining real-time wear assessment results for condition monitoring based on the original vibration signal matrix further includes: Performing time-frequency domain feature extraction on the independent source signal components, using the wavelet packet decomposition algorithm to obtain the energy distribution characteristics of different frequency bands, and obtaining a multi-dimensional feature vector set reflecting the wear states of different components by calculating the energy entropy values and relative energy ratios of each frequency band.

[0008] Preferably, the wear assessment and trend analysis module obtaining real-time wear assessment results for condition monitoring based on the original vibration signal matrix further includes: Establishing a support vector regression model based on the multi-dimensional feature vector set, mapping the feature space using the radial basis kernel function, training the regression parameters with historical wear data. If the training error exceeds the preset accuracy requirement, adjust the kernel function parameters and regularization coefficients until the model converges.

[0009] Preferably, the wear assessment and trend analysis module, according to the original vibration signal matrix, obtaining the real-time wear assessment result for condition monitoring further includes: Using a support vector regression model to perform wear degree quantification calculation on the current feature vector, determining the wear grades of the bearing and the gear according to the regression output value, judging the health state of the component by comparing with the preset wear threshold, and obtaining the real-time wear assessment result for condition monitoring.

[0010] Preferably, the multi-channel vibration sensor array includes at least three channels, which are respectively arranged at the bearing seat, the gearbox and the transmission shaft of the ball mill.

[0011] Preferably, the fault warning and model adaptation module has a warning level classification mechanism, sets multi-level thresholds according to the wear rate and the remaining life prediction result, and outputs warning instructions and maintenance suggestions of different levels.

[0012] Preferably, it further includes a data communication module for real-time transmitting the wear assessment result, the trend data and the warning information to the upper monitoring platform or the remote expert system through industrial Ethernet or wireless communication.

[0013] An on-line oil detection method for cement ball mills, using the on-line oil detection system for cement ball mills as described above, the method includes: Using a multi-channel vibration sensor array to synchronously collect data at the bearing seat, the gearbox and the transmission shaft of the ball mill, obtaining the original vibration signal matrix containing multi-component coupling information through high-frequency sampling. If the sampling frequency is lower than ten times the wear characteristic frequency, automatically adjust the sampling parameters until the Nyquist theorem requirements are met; According to the original vibration signal matrix, obtaining the real-time wear assessment result for condition monitoring, performing trend analysis on the wear assessment results within a continuous time period through the sliding time window technology, using the least squares method to fit the wear evolution curve, and calculating the wear development rate and acceleration according to the curve slope and the second derivative, obtaining the wear evolution law parameters for prediction analysis; Establishing a fault warning mechanism according to the wear evolution law parameters. If the wear development rate exceeds the safety threshold or the predicted remaining life is lower than the maintenance cycle, trigger a warning signal and generate a maintenance suggestion, and adapt to the change of the equipment operation condition by real-time updating the model parameters.

[0014] It can be seen from the above technical solutions that the present invention has the following beneficial effects: The on-line oil detection system for cement ball mills collects the original vibration signals of bearing seats, gearboxes and transmission shafts through a multi-channel vibration sensor array, separates the signals through preprocessing and independent component analysis, extracts time-frequency domain features to construct a support vector regression model, and quantitatively evaluates the wear degrees of bearings and gears. The sliding time window is used to analyze the wear evolution trend, establish an early warning mechanism and generate maintenance suggestions. The present invention can realize real-time monitoring, trend analysis and fault early warning of the wear states of key components of ball mills, effectively improve the operation reliability and maintenance efficiency of equipment, extend the service life, reduce the maintenance cost, and provide data support for the preventive maintenance decision-making of ball mills. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a connection diagram of system modules of the present invention; Figure 2 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0016] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0017] As Figure 1 shown, the present invention provides a technical solution: an on-line oil detection system for cement ball mills, and the system includes: A data acquisition module synchronously acquires data on the positions of the bearing seats, gearboxes and transmission shafts of the ball mill by using a multi-channel vibration sensor array, and obtains an original vibration signal matrix containing multi-component coupling information through high-frequency sampling. If the sampling frequency is lower than ten times the wear characteristic frequency, the sampling parameters are automatically adjusted until the Nyquist theorem requirements are met; A wear evaluation and trend analysis module obtains real-time wear evaluation results for condition monitoring according to the original vibration signal matrix, analyzes the trend of the wear evaluation results within a continuous time period through the sliding time window technology, fits the wear evolution curve by using the least square method, calculates the wear development rate and acceleration according to the curve slope and second derivative, and obtains wear evolution law parameters for prediction analysis; A fault early warning and model adaptation module establishes a fault early warning mechanism according to the wear evolution law parameters. If the wear development rate exceeds the safety threshold or the predicted remaining life is lower than the maintenance period, an early warning signal is triggered and maintenance suggestions are generated, and the model parameters are updated in real time to adapt to the changes in the operating conditions of the equipment.

[0018] The system realizes the synchronous condition monitoring of multiple key components of the ball mill by integrating a multi-channel vibration sensor array. Each vibration sensor is respectively fixed around the bearing housing, gearbox and transmission shaft, and captures minute mechanical vibration changes through high-frequency sampling. During the acquisition process, the system automatically monitors whether the sampling frequency meets the requirement of being more than ten times the wear characteristic frequency. If not, it triggers a dynamic sampling adjustment mechanism to ensure that the data sampling meets the Nyquist sampling theorem, thus avoiding spectrum aliasing and information loss.

[0019] The original vibration signals are organized into a high-dimensional matrix and input into the wear assessment and trend analysis module. First, typical frequency components are extracted through time-frequency domain analysis techniques such as fast Fourier transform or wavelet transform, and potential abnormal frequency characteristics are identified. Subsequently, the system uses methods based on empirical mode decomposition or ensemble empirical mode decomposition to extract intrinsic mode functions and further analyze the wear patterns in non-stationary signals. With the support of the sliding time window technique, the system can dynamically extract the wear state parameters in each time period during continuous operation. By constructing a time series and using the least squares method for regression fitting, a wear evolution curve is formed.

[0020] By calculating the first derivative and the second derivative of this curve respectively, the system estimates the current wear rate and acceleration in real time, extracts the trend change parameters of wear development, and provides reliable inputs for predictive analysis. Next, the fault warning and model adaptation module conducts threshold comparison analysis on the wear evolution law parameters. When it detects that the wear rate rises sharply or the predicted remaining life is lower than the set safety boundary, it automatically triggers the warning mechanism and gives corresponding maintenance suggestions. To cope with the continuous changes in the equipment operating state, this module also integrates a model adaptation function, using self-updating parameter mechanisms or online learning strategies, such as recursive least squares method or incremental training algorithms, to ensure that the model always reflects the latest operating characteristics of the equipment and guarantees the accuracy and robustness of system prediction and evaluation.

[0021] The system adopts a technical path based on the analysis of multi-source high-frequency vibration signals and is applicable to the monitoring of ball mills under complex coupling structures. The data acquisition module is configured with a three-channel acceleration sensor array, which is respectively installed at key parts such as the bearing housing, gearbox and transmission shaft of the ball mill to simultaneously and synchronously capture the original vibration signals generated during the equipment operation. The system sets the lower limit of the initial sampling frequency to be more than ten times a reference frequency, and this reference frequency corresponds to the highest wear characteristic frequency that the equipment may generate. If the sampling frequency during actual operation is insufficient to meet this condition, the system will automatically increase the sampling rate to ensure that the collected signals completely reflect the real working conditions and avoid spectrum aliasing or loss of key information.

[0022] In the signal processing stage, the system first conducts a segmented analysis on the collected vibration signals, splitting the entire time-domain signal into multiple time segments and separately extracting the frequency characteristics of each segment to observe how the vibration intensity and frequency change over time. To further extract non-linear or non-stationary characteristics, the system uses a technical method called ensemble empirical mode decomposition to decompose the complex signal into multiple independent frequency modes and extract the key information contained in each frequency band.

[0023] Subsequently, the system sets a continuously sliding time window, which can be ten seconds or a complete operating cycle of the ball mill. A characteristic value for measuring the degree of equipment wear is calculated within each time window. As time goes by, the system accumulates the characteristic values within multiple time windows and arranges them in chronological order to form a data sequence that can be used for trend fitting.

[0024] The system conducts trend modeling on this data sequence, using a mathematical fitting method to approximate it as a quadratic curve that changes over time. In this curve, each time point corresponds to a predicted wear value; the slope at this point is used to represent how fast the wear degree increases at this moment, that is, the wear rate; and the degree of curvature of the curve at this point is used to characterize the trend of wear intensification, that is, the wear acceleration.

[0025] The obtained wear rate and acceleration are input into the next stage as important parameters for evaluating the health status of the equipment. The system sets a safety limit value to define the upper limit of the wear rate of the equipment under normal operating conditions. If the actually measured wear rate exceeds this safety limit, or the remaining service time of the equipment calculated based on the wear trend is less than the preset maintenance cycle time, the system will actively issue a warning and generate maintenance suggestions.

[0026] To ensure that the system remains effective under the background of equipment load changes, operating state fluctuations, and even equipment aging, the system is also equipped with an adaptive model update module. This module will automatically update the parameters of the wear trend model according to the newly collected real-time data, making the system's analysis ability consistent with the current actual operating state of the equipment, thereby ensuring the continuous accuracy and real-time nature of the warning mechanism.

[0027] The present invention realizes real-time status monitoring and accurate wear prediction of each key component of the ball mill through the synchronous acquisition and analysis of high-frequency vibration coupling signals. The multi-channel array design enhances the ability to express signal spatial information, and the EEMD analysis improves the modeling accuracy of non-linear and non-stationary signals. The system has an adaptive sampling and modeling mechanism, significantly improving the timeliness of fault identification and prediction accuracy, effectively reducing the incidence of sudden failures, extending the service life of the equipment, and enhancing the scientific nature of maintenance decisions. Compared with traditional single-point detection, this system has obvious improvements in performance, stability, and intelligent level.

[0028] Wear assessment and trend analysis module, based on the original vibration signal matrix, obtains real-time wear assessment results for condition monitoring, including performing preprocessing operations on the original vibration signal matrix, using a band-pass filter to remove power frequency interference and high-frequency noise, and eliminating random noise components through the time-domain synchronous averaging algorithm to obtain a purified multi-channel vibration signal dataset for subsequent analysis and processing.

[0029] First, the system synchronously acquires the original vibration signals of multiple channels from the multi-channel vibration sensor array of the ball mill. These signals include the real state information during the operation of the equipment, but are also mixed with environmental interference and noise. To improve the accuracy of subsequent analysis, the system first performs band-pass filtering on these original signals. The lower limit frequency set by the band-pass filter is generally 10 Hz, which is used to remove the 50 Hz power frequency interference from the power grid and the low-frequency background signals caused by mechanical low-frequency shaking, etc.; while the upper limit frequency is generally set within 5 kHz, which is used to remove electromagnetic interference and high-frequency electronic noise. The goal of this filtering process is to retain the effective signals in the middle frequency band that are highly correlated with wear characteristics.

[0030] The filtered signals may still contain some non-periodic interference caused by operation disturbances. Therefore, the system further introduces synchronous averaging technology, that is, aligning the vibration signals of multiple consecutive cycles in time according to the position of each sensor, and then averaging the data at these same time points. This method can effectively eliminate the random components, because the non-periodic noise approaches zero after multiple averaging; while the periodic signal part will be superimposed and enhanced, so that the stable operating state of the equipment can be characterized more clearly.

[0031] To implement the above synchronous averaging operation, the system first measures the spindle speed of the ball mill and calculates the time period required for each rotation; then, extracts multiple complete cycle data segments from the original signals and performs corresponding superposition and mean value calculation. Generally, the system will select 5 to 10 cycles of data to participate in the averaging, and the specific quantity can be adjusted according to the signal stability and processing accuracy requirements.

[0032] After the above filtering and synchronous averaging processing, the system obtains a set of purified multi-channel vibration signals. These signals have significantly reduced the interference of environmental and system noise and have higher stability and interpretability. This set of purified signals will then be sent to the subsequent state variable extraction module, and key characteristic parameters for judging the wear degree will be further extracted through methods such as envelope demodulation, Hilbert transform, wavelet decomposition, etc., and used as the input for equipment condition monitoring and wear trend prediction analysis.

[0033] By introducing two signal preprocessing techniques, band-pass filtering and time-domain synchronous averaging, the purification ability and feature retention degree of vibration signals are significantly improved. The band-pass filter can effectively filter out power frequency interference and high-frequency noise, ensuring that only the frequency components closely related to mechanical wear are retained in the signal, thereby improving the accuracy and stability of subsequent analysis results. The synchronous averaging technique strengthens the extraction effect of periodic wear signals during equipment operation through multi-period overlapping averaging, while suppressing the random noise introduced by irregular disturbances and occasional interferences, significantly improving the signal-to-noise ratio during the feature extraction process. Compared with the traditional single time-domain or frequency-domain filtering processing method, the multi-step signal purification process provided by the present invention has stronger robustness and is applicable to the state monitoring scenarios of cement ball mills under various complex operating conditions. As a result, the wear assessment process is more sensitive and accurate, effectively avoiding false alarms and missed alarms, laying a reliable data foundation for subsequent trend analysis and fault warning, improving the prediction ability and operating stability of the system, and extending the trouble-free operation time of the equipment.

[0034] The wear assessment and trend analysis module obtains real-time wear assessment results for state monitoring based on the original vibration signal matrix. It also includes performing blind source separation processing on the purified multi-channel vibration signal dataset through the independent component analysis algorithm, decomposing the coupled composite signal into independent source signal components according to the statistical independence principle. If the number of separated signal components exceeds the preset threshold, the kurtosis criterion is used to screen out the main components related to wear.

[0035] After the signal preprocessing is completed, the independent component analysis algorithm is introduced to perform blind source separation operations to identify the intrinsic components closely related to mechanical wear from multi-channel coupled signals.

[0036] At this stage, the purified multi-channel vibration signal dataset can be represented as an observed signal matrix, which consists of time series recorded by multiple vibration sensors and is denoted as a set of mixed signals. According to the blind source separation theory, this set of mixed signals can be regarded as a linear combination of several source signals through an unknown mixing process. Its basic model expression is: the mixed signal is equal to the product of an unknown mixing matrix and several independent source signals.

[0037] The core objective of the independent component analysis algorithm is to recover these source signals, that is, to obtain a set of separation matrices such that the recovered signals have the strongest statistical independence. Common independence metrics include strategies such as minimizing information entropy, maximizing negentropy, and enhancing non-Gaussianity. In this system, the FastICA algorithm is used as the implementation path, and its optimization objective is to maximize the negentropy function to extract source components.

[0038] The approximate form of negentropy is that the negentropy value is equal to the statistical difference between the observed signal and the Gaussian signal under the mapping of a certain non - linear function. Its specific form is the difference in expected values constructed using non - linear functions (such as hyperbolic tangent, exponential function, etc.), which is used to evaluate independence.

[0039] During the FastICA iteration process, by maximizing this approximate negentropy function, the separation vectors corresponding to each independent component are gradually estimated, and finally the source signal sequence is obtained.

[0040] After obtaining all independent components, to identify the components significantly related to mechanical wear from them, the system introduces the kurtosis criterion as a screening index. Kurtosis is a statistic that describes the "sharpness" of the signal probability distribution and is used to quantify the degree to which the signal deviates from the Gaussian distribution. Its definition is: kurtosis is equal to the fourth - order central moment of the signal divided by the square of the square of the second - order central moment, minus the constant three.

[0041] The higher the kurtosis value, the sharper the signal peaks and the stronger the impact, which is common in mechanical impact - type wear signals; When the kurtosis value is close to three, it means the signal is close to the Gaussian distribution and the information contribution is low.

[0042] The system presets a kurtosis threshold, such as ten or fifteen. If the kurtosis value of an independent component is higher than this threshold, it indicates that it may contain sudden and strongly non - linear wear information, and then this component is retained as the main analysis object; the remaining components are excluded to avoid redundant data interfering with the analysis.

[0043] Finally, the selected high - kurtosis source signals are sent to the wear trend modeling module to complete subsequent processes such as curve fitting, rate calculation, and early warning generation.

[0044] By introducing the independent component analysis algorithm, this system can effectively identify potential independent sources in vibration signals, that is, the intrinsic responses generated by multiple coupled components. Compared with traditional methods based on single - channel analysis or frequency filtering, the blind source separation technology has the ability to automatically recover wear - related components from complex mixed signals, without relying on pre - known signal models or physical assumptions, greatly improving the sensitivity and generality of fault diagnosis. Especially combined with the kurtosis screening mechanism, the present invention can effectively eliminate background - stable signals and only retain abnormal vibration modes with impact or non - linear enhancement characteristics, thus focusing on the key signals truly related to the wear state. This processing method improves the accuracy of wear trend modeling, reduces the misjudgment rate and false - alarm risk, and enables the system to cope with the actual industrial scenarios with complex equipment structures and strong coupling interference, having stronger adaptability and technical practicability.

[0045] The wear assessment and trend analysis module obtains real-time wear assessment results for condition monitoring based on the original vibration signal matrix. It also includes time-frequency domain feature extraction for independent source signal components, using the wavelet packet decomposition algorithm to obtain the energy distribution characteristics of different frequency bands. By calculating the energy entropy values and relative energy ratios of each frequency band, a multi-dimensional feature vector set reflecting the wear states of different components is obtained.

[0046] After completing the purification, independent component separation, and screening of the original vibration signal, the system further performs feature extraction operations on the selected independent source signal components to comprehensively reflect different wear types and component states. This process uses the wavelet packet decomposition algorithm for multi-scale time-frequency analysis of the signal, extracting frequency band energy features and entropy information for constructing a multi-dimensional state feature vector.

[0047] The specific steps are as follows: Wavelet packet decomposition processing. For each independent source signal, wavelet packet decomposition is used to perform multi-level frequency band division. Different from traditional wavelet decomposition that only further decomposes low-frequency signals, wavelet packet technology can completely decompose both high-frequency and low-frequency parts simultaneously, thus achieving uniform coverage of the entire frequency range. Set the decomposition level of wavelet packet decomposition to L, and the original signal is divided into 2 L sub-frequency bands, and each sub-frequency band represents the energy components within a specific frequency range. Usually, the decomposition level is set according to the system sampling rate and the expected fault frequency band. For example, when the sampling rate is 10,000 Hz, choosing to decompose three layers can obtain eight frequency bands, and the width of each frequency band is approximately 1,250 Hz.

[0048] Frequency band energy calculation. The energy of each sub-frequency band signal is based on the result of calculating its two-norm, specifically the sum of the squares of the amplitudes of all sample points. Let the signal corresponding to the jth frequency band be d j , and its energy E j The calculation method is: square the vibration amplitudes of all sampling points in this frequency band and then sum them to obtain the energy value. The sum of the energies of all frequency bands is E total . By calculating the proportion P j = E j / E total , the relative energy distribution is obtained, which is used to judge the concentration degree of vibration energy and the distribution of fault frequency bands.

[0049] Energy entropy calculation. To further measure the degree of chaos in frequency distribution and the degree of spectrum expansion, the energy entropy index is introduced. The calculation method of energy entropy is: multiply the energy ratio of each frequency band by its logarithm value, then accumulate the obtained results and take the negative sign to get the total energy entropy. The larger this value, the more dispersed the energy distribution, indicating a higher complexity of the vibration signal; a lower entropy value means that the energy is concentrated in a few frequency bands, often corresponding to local structural wear or impact faults.

[0050] Construct a multi-dimensional feature vector by combining the relative energy values of each frequency band with the total energy entropy. The dimension of this vector is equal to the number of frequency bands plus one, i.e., a complete wear feature set including the energy ratios of all frequency bands and the energy entropy, forming an input sample for training the model or identifying the state.

[0051] The number of wavelet decomposition levels controls the number of frequency bands and the analysis accuracy, usually set to three to four layers, determined according to the sampling rate and the frequency spectrum range; The sub-band energy is the total vibration energy within each frequency band, calculated by summing the squared amplitudes; The relative energy ratio is the proportion of the energy of each frequency band to the total energy, obtained by dividing the sub-band energy by the total energy; The energy entropy represents the entropy value of the frequency band energy distribution, using the Shannon entropy calculation logic to evaluate the complexity and concentration degree.

[0052] By introducing the wavelet packet decomposition and frequency band energy feature extraction methods, the system can conduct refined analysis on the vibration responses in different frequency ranges, significantly enhancing the recognition ability for multi-component and multi-type wear behaviors. Especially the introduction of the energy entropy index enables the system not only to judge the energy concentration interval but also to identify the change in signal complexity, effectively distinguishing local faults from overall vibration anomalies. Compared with the traditional single-frequency band analysis method, the multi-dimensional feature vector provided by the solution described in the present invention covers a wider range of frequency information and has stronger discrimination and expression ability, suitable for supporting the training of machine learning models or threshold-driven judgment. Its processing process does not rely on manual rule setting, has a highly automated feature, is applicable to the long-term and unattended equipment status monitoring scenario in the actual production environment, and significantly improves the diagnostic accuracy and system practicality.

[0053] The wear assessment and trend analysis module, based on the original vibration signal matrix, obtains real-time wear assessment results for condition monitoring. It also includes establishing a support vector regression model according to the multi-dimensional feature vector set, using the radial basis kernel function to map the feature space, training the regression parameters with historical wear data. If the training error exceeds the preset accuracy requirement, the kernel function parameters and regularization coefficients are adjusted until the model converges.

[0054] On the basis of the multi-dimensional wear feature vector extracted by wavelet packet decomposition mentioned above, this system further introduces a machine learning-based non-linear regression model for predicting the future wear trend of the equipment. This model uses the support vector regression method, and its core idea is to learn the corresponding relationship between the wear features and the real wear state in the historical samples, so as to construct a function model that can be used for future prediction.

[0055] Specifically, the support vector regression model maps the input multi-dimensional feature vectors into a higher-dimensional feature space, where an optimal hyperplane is sought to make the predicted output as close as possible to the true value, and the error is kept within an allowable tolerance range. To achieve this goal, the system adopts a kernel function called radial basis to map the original input into a high-dimensional space. This kernel function constructs a similarity index based on the distance relationship between input features, thereby enhancing the model's ability to model non-linear relationships.

[0056] During the model training process, the system first prepares a set of historical data, which includes multi-dimensional feature vectors extracted at different time periods and the corresponding actual wear degree at that time as labels. The system uses this data to train the regression model to construct a function in the high-dimensional space that can estimate the wear value with the minimum error. During the training process, the system sets an error tolerance range to ensure that a certain degree of deviation is allowed during prediction, and at the same time introduces a penalty coefficient to control the balance between model complexity and training error.

[0057] When the model is initially trained, if the evaluation results show that its prediction error exceeds the preset accuracy requirement, for example, the average error is greater than five percentage points, the system will activate the automatic parameter adjustment mechanism to adjust the width parameter of the kernel function and the error penalty coefficient. The width of the kernel function affects the sensitivity of the model to input changes. A larger width makes the prediction curve smoother and is suitable for situations with slow trend changes; while a smaller width can enhance the model's response ability to mutations. The penalty coefficient is used to adjust the tolerance for errors. The larger the value, the more the model tends to reduce the error of each sample point.

[0058] The system gradually tries different parameter combinations, and uses methods such as cross-validation to compare the prediction accuracy under each combination, and finally selects the parameter combination with the smallest prediction error as the final configuration of the model.

[0059] The trained model can be used to predict the wear trend corresponding to the current state of the device in real time, so as to achieve the purpose of early detection of abnormal wear and optimization of maintenance strategies. By continuously inputting new feature vectors, the model can continuously output updated prediction results to support the system to make dynamic decisions.

[0060] By introducing a support vector regression model into the wear assessment and trend analysis module, the present invention establishes a non-linear function framework with high-precision prediction ability. Through training with historical data, the model can accurately capture the complex relationship between feature vectors and actual wear states, achieving forward-looking prediction of future wear trends. Compared with traditional linear regression methods, SVR combined with a radial basis kernel function has stronger non-linear modeling ability and can still maintain stable prediction accuracy in scenarios with large fluctuations in equipment operating states and frequent load changes. By automatically adjusting the kernel function parameters and regularization coefficients, the system can avoid overfitting while achieving rapid convergence of training errors, enhancing its adaptive ability to different working conditions and data scales. Finally, the integration of this model significantly improves the reliability of equipment operating state prediction, provides a solid data foundation for preventive maintenance and intelligent decision-making, and enhances the overall operating efficiency and safety of the system.

[0061] The wear assessment and trend analysis module obtains real-time wear assessment results for condition monitoring based on the original vibration signal matrix. It also includes using a support vector regression model to quantitatively calculate the wear degree of the current feature vector, determining the wear levels of bearings and gears based on the regression output values, and judging the health status of components by comparing with preset wear thresholds to obtain real-time wear assessment results for condition monitoring.

[0062] After the wear trend modeling phase is completed, the system uses the support vector regression model to perform real-time wear prediction on the currently collected feature vectors to quantify the wear degree of each key component of the current equipment. The specific process is as follows: the system first processes the vibration data within a certain period of time through preprocessing, wavelet decomposition and feature extraction to form a feature vector containing multiple dimensions. The feature vector is then sent to the trained support vector regression model for calculation. The model has learned the mapping relationship between the input features and the degree of wear during the previous training process, and can output a continuous value as the current predicted wear level. The predicted value represents the severity of wear of a certain component of the current equipment (such as a bearing or gear). The system then compares this predicted value with a set of pre-set wear level intervals. Usually, these levels are divided into several ranges, for example: a predicted value between zero and 0.3 indicates slight wear, between 0.3 and 0.6 indicates moderate wear, between 0.6 and 0.9 indicates heavy wear, and more than 0.9 indicates severe wear. These wear level thresholds are set based on a large amount of actual operation and maintenance experience and historical data analysis results, and have clear physical meanings. For example, in the bearing scenario, slight wear may correspond to surface microcracks or poor lubrication, moderate wear may be the early stage of raceway spalling, and severe wear usually means imminent failure. Once the system completes the wear level judgment, it will automatically judge the health status of the equipment according to the corresponding interval. If it is judged to be slight wear, the system only records the current status; if it is moderate or severe, maintenance recommendations are generated; if it is a serious level, an alarm is triggered immediately, and the control system can be linked to achieve shutdown protection or send remote notifications. The entire process does not rely on human intervention. All judgment criteria come from data-driven models and threshold comparison logic, which are highly automated and standardized, ensuring that the system can stably and accurately provide real-time wear assessment results under various operating conditions, serving equipment status monitoring and proactive maintenance management.

[0063] The present invention not only realizes the real-time evaluation of components such as bearings and gears under complex operating conditions by applying the support vector regression model to the continuous quantitative calculation of the degree of wear, but also establishes a standardized wear grade classification system, so that the system has a clear and executable basis for intelligent judgment. Compared with the traditional extensive monitoring mode based on empirical thresholds, the present invention uses a data-driven approach to continuously model the wear trend and convert it into a grade classification output, which has higher accuracy, sensitivity and engineering interpretability. It can detect minor changes at an early stage, accurately identify the direction of trend development, and trigger corresponding strategies according to preset logic, such as alarm, recording or linkage maintenance, greatly improving the automation and intelligence level of equipment monitoring. At the same time, the grade output method is concise and clear, adapting to the industrial site's operation and maintenance management requirements for the "red, yellow, and green" classification status, and improving the system's operability and promotion value in actual engineering scenarios.

[0064] The multi-channel vibration sensor array includes at least three channels, which are respectively arranged at the bearing seat, gearbox and transmission shaft of the ball mill.

[0065] To comprehensively monitor the vibration status of key parts of the ball mill, at least three channels of vibration sensor arrays are configured in the data acquisition module of this system. These sensors are respectively arranged at three positions of the bearing seat, gearbox and transmission shaft of the ball mill, corresponding to typical mechanical vibration sources respectively, aiming to synchronously collect the coupled vibration signals of multiple components.

[0066] The sensors of the first channel are installed near the bearing seat, mainly used to detect medium and low frequency vibration signals caused by problems such as fatigue wear of rolling elements or raceways, insufficient lubrication or hardening of grease. The second channel is set at the position of the gearbox housing, used to collect high frequency or impact vibration information caused by possible periodic impacts, changes in meshing clearances, gear fractures, etc. during the gear meshing process. The third channel is installed near the transmission shaft, responsible for sensing the dynamic change signals caused by torque fluctuations, axis offsets or coupling looseness, reflecting the torsional and eccentric characteristics of the system.

[0067] All three channels are configured in high frequency sampling mode and are controlled by a unified trigger signal to ensure that all signals are synchronized in the time dimension. This means that the sampling time points of each signal are strictly the same, which can avoid coupling errors caused by data asynchronization.

[0068] The sampling frequency is set according to the target analysis frequency. In principle, it must be much higher than the most significant wear characteristic frequency of the equipment. In practical applications, the system usually sets the number of sampling points per second between five thousand and twenty-five thousand to capture high frequency impacts or signs of local faults. For example, when analyzing a signal with a maximum vibration frequency of one thousand hertz, the sampling frequency should be set at least above ten thousand hertz to meet the information restoration requirements.

[0069] The vibration signals of the three channels collected will be respectively saved in chronological order and form a structured data set, where each signal corresponds to an independent time series, and the three groups of data are aligned on the time axis. This data structure can be directly used as the input basis for subsequent processing modules such as filtering, feature extraction, blind source separation and trend modeling.

[0070] To match the mechanical rigidity and expected vibration amplitudes of different installation parts, the system differentially sets the gain of each channel during configuration. For example, the vibration amplitude at the transmission shaft part is relatively small, and it is suitable to set a higher gain; while there may be strong impact signals at the gearbox, so a medium gain is set to avoid signal saturation.

[0071] In summary, the reasonable layout and parameter configuration of the multi-channel array ensure that the system can obtain stable, comprehensive, and high-precision original vibration signals under different working conditions, providing a solid data support for subsequent intelligent diagnosis and prediction.

[0072] By deploying a multi-channel vibration sensor array at three core parts, namely the bearing housing, gearbox, and transmission shaft, the system can obtain comprehensive dynamic data representing the operating states of different components. This structure is significantly superior to the single-point monitoring method, and can effectively capture the coupled signals from different sources of the equipment, thus supporting subsequent blind source separation, wavelet energy analysis, and trend modeling. The multi-channel design improves the spatial resolution of the data, helps to analyze the correlation and time-series characteristics between vibration signals, and provides a solid foundation for fault source localization. Especially in the structure jointly driven by the bearing and gearbox, this design can identify the composite vibration modes generated by mutual influence, significantly improving the diagnostic accuracy and response ability. In addition, this array deployment scheme has structural flexibility and compatibility, is applicable to ball mill systems of different specifications and operating conditions, and has good on-site adaptability and popularization value.

[0073] The fault warning and model adaptive module has a warning level classification mechanism, which sets multiple thresholds based on the wear rate and remaining life prediction results, and outputs warning instructions and maintenance suggestions of different levels.

[0074] In the fault warning and model adaptive module, to enhance the sensitivity of the system to the operating state of the equipment and the pertinence of the response, the present invention sets a multi-level warning classification mechanism based on the wear rate and remaining life prediction results. This mechanism aims to automatically output warning instructions of different levels and corresponding maintenance suggestions according to the dual judgments of the current state evolution speed and available remaining time of the equipment.

[0075] First, during the operation of the system, it continuously monitors the vibration characteristics of the equipment, and calculates the current wear rate through a trend analysis model. The wear rate refers to the change speed of the wear degree per unit time, and its calculation method is to compare the change values of the wear indicators of the equipment in adjacent time periods and divide by the time interval to obtain the average change speed during this period. At the same time, the system will also use the current trend curve to extrapolate into the future to estimate the time required for the equipment to reach the set termination wear threshold under the premise that the current trend remains unchanged, which is the remaining life.

[0076] The system makes a level judgment based on the above two key indicators, namely the wear rate and the remaining life. To achieve refined management, the system divides the warning into three levels: the first level is the warning level, the second level is the serious level, and the third level is the critical level. The judgment logic is as follows: When the wear rate of the equipment is at a low level and the predicted remaining life is long, the system determines that the current equipment state is good, and only gives a low-level reminder or records the log; When the wear rate increases, or the remaining life enters the medium- and short-term range, the system determines that there is a certain risk in the current state, issues a medium-level warning, and recommends arranging maintenance. If the wear rate of the equipment has increased significantly, or the remaining life is already lower than the minimum tolerance time threshold, the system immediately determines it as a high-level alarm, activates the linkage response mechanism, and outputs complete emergency measures including shutdown instructions, emergency maintenance suggestions, and personnel notifications.

[0077] The judgment process of the above warning levels is based on multiple preset numerical thresholds, which are set according to the statistical analysis results of historical operation data. For example, the grading thresholds of the wear rate may be 1% and 3% per hour respectively, and the time limits of the remaining life can be set to 48 hours and 24 hours. The system compares the current calculated values with these thresholds during operation and automatically matches the corresponding warning levels.

[0078] Finally, the system will output a warning instruction containing information such as level identification, text suggestions, and color tags (such as green, orange, red) according to the judgment result, so that the operator can clearly understand the current state and take appropriate measures accordingly. All warning results will also be recorded in the log for subsequent retrospective analysis and model optimization.

[0079] The present invention introduces a multi-level warning mechanism based on the combined judgment of wear rate and remaining life, enabling the system to not only identify abnormal trends in the equipment operation state, but also achieve dynamic hierarchical response according to the fault evolution speed and expected remaining life. Compared with the traditional single-threshold trigger method, this mechanism has higher response sensitivity and fault discrimination accuracy. By classifying and managing the wear rate and life prediction results, the system can effectively distinguish between mild fluctuations and serious failures, avoiding false alarms or missed alarms. At the same time, the multi-level response strategy can support the deployment of step-by-step control measures, realizing the hierarchical execution of alarms, suggestions, and emergency measures, which helps the scientific scheduling of production plans and the optimal allocation of operation and maintenance resources. This mechanism improves the forward-looking and initiative of equipment fault handling, providing powerful intelligent decision-making support and safety guarantee for the operation of ball mills under complex working conditions.

[0080] It also includes a data communication module for real-time transmitting the wear assessment results, trend data, and warning information to the upper monitoring platform or remote expert system through industrial Ethernet or wireless communication.

[0081] The on-line oil detection system for cement ball mills of the present invention is provided with a complete data encapsulation, transmission and remote integration mechanism in the data communication module, ensuring that the wear assessment results, trend data and warning information can be stably and real-timely transmitted to the upper monitoring platform or remote expert system, realizing data closed-loop and intelligent collaboration. Specifically, the system first structurally encapsulates the data from each module, including equipment number, acquisition time, current wear assessment value, remaining life prediction value, warning level, wear trend parameters and operation status flag, etc., and performs standardized processing using common industrial communication formats such as Modbus TCP, OPCUA or MQTT to ensure data compatibility and resolvability. In terms of communication method, the system can select industrial Ethernet for high-speed and stable transmission according to the actual on-site requirements, which is suitable for areas with mature wiring; it can also adopt wireless communication methods such as Wi-Fi, 4G or LoRa to adapt to application environments with long distances, difficult wiring or strong equipment mobility. The default data push frequency of the system is once every thirty seconds, and at the same time, it supports the event trigger mode based on state changes and has flexible configuration capabilities. To ensure communication real-time and reliability, the system controls the total communication delay within two seconds and builds in a cyclic redundancy check and retransmission mechanism to ensure data integrity and accuracy. In addition, the system supports establishing a stable connection through TCP / IP and allows the remote receiving end to be configured as a visualization monitoring platform or an AI expert system, so as to provide diagnostic suggestions and automatically trigger the maintenance process in the early stage of the fault. The integration of this module realizes the efficient circulation of the operation status information of the ball mill from edge perception to remote collaboration, significantly improving the intelligent level, response speed and management efficiency of the system, and meeting the actual needs of large industrial equipment for reliable remote monitoring and intelligent maintenance.

[0082] By integrating the data communication module, the present invention can achieve cross-layer transmission and remote sharing of system-level information, not only enhancing the real-time and response capabilities of the local monitoring system, but also enabling the equipment operation status to be transparently presented on the remote platform, providing a real-time data basis for operation and maintenance scheduling, expert consultation and production management. Compared with the traditional "local data acquisition and manual reading" mode, the communication mechanism of this system can achieve high-frequency, low-latency and structured data push, and cooperate with standard protocols to achieve seamless docking with mainstream industrial software systems. It is especially suitable for cement production lines with distributed deployment, significantly improving the equipment status collaborative perception ability, and supporting providing intelligent judgment and maintenance suggestions through cloud algorithm models in the early stage of fault development to avoid fault spread. This function significantly improves the overall intelligent level, response speed and management efficiency of the system, and is especially suitable for continuous production scenarios with high requirements for equipment reliability and short maintenance windows.

[0083] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An on-line oil liquid detection system for cement ball mills, characterized in that, The system includes: A data acquisition module that synchronously acquires data on the positions of the ball mill bearing housing, gearbox, and transmission shaft using a multi-channel vibration sensor array, obtaining an original vibration signal matrix containing multi-component coupling information through high-frequency sampling. If the sampling frequency is less than ten times the wear characteristic frequency, the sampling parameters are automatically adjusted until the Nyquist theorem requirements are met. A wear assessment and trend analysis module that, based on the original vibration signal matrix, obtains real-time wear assessment results for condition monitoring, performs trend analysis on the wear assessment results within a continuous time period through a sliding time window technique, fits a wear evolution curve using the least squares method, calculates the wear development rate and acceleration based on the curve slope and second derivative, and obtains wear evolution law parameters for predictive analysis. A fault warning and model adaptation module that establishes a fault warning mechanism based on the wear evolution law parameters. If the wear development rate exceeds the safety threshold or the predicted remaining life is less than the maintenance cycle, a warning signal is triggered and maintenance suggestions are generated, and the model parameters are updated in real time to adapt to changes in the equipment operating conditions.

2. The online oil detection system for cement ball mills according to claim 1, wherein: The wear assessment and trend analysis module, based on the original vibration signal matrix, obtaining real-time wear assessment results for condition monitoring includes: Performing preprocessing operations on the original vibration signal matrix, using a band-pass filter to remove power frequency interference and high-frequency noise, and eliminating random noise components through a time-domain synchronous averaging algorithm to obtain a purified multi-channel vibration signal dataset for subsequent analysis and processing.

3. The on-line oil detection system for cement ball mills according to claim 2, characterized in that: The wear assessment and trend analysis module, based on the original vibration signal matrix, obtaining real-time wear assessment results for condition monitoring further includes: Performing blind source separation processing on the purified multi-channel vibration signal dataset through an independent component analysis algorithm, decomposing the coupled composite signal into mutually independent source signal components according to the principle of statistical independence. If the number of separated signal components exceeds a preset threshold, the kurtosis criterion is used to screen out the main components related to wear.

4. The on-line oil detection system for cement ball mills according to claim 3, characterized in that: The wear assessment and trend analysis module, based on the original vibration signal matrix, obtaining real-time wear assessment results for condition monitoring further includes: Performing time-frequency domain feature extraction on the independent source signal components, obtaining the energy distribution characteristics of different frequency bands using a wavelet packet decomposition algorithm, and calculating the energy entropy value and relative energy ratio of each frequency band to obtain a multi-dimensional feature vector set reflecting the wear states of different components.

5. The on-line oil detection system for cement ball mill according to claim 4, characterized in that: The wear assessment and trend analysis module, based on the original vibration signal matrix, obtaining real-time wear assessment results for condition monitoring further includes: Establishing a support vector regression model based on the multi-dimensional feature vector set, mapping the feature space using a radial basis kernel function, training the regression parameters with historical wear data. If the training error exceeds the preset accuracy requirement, the kernel function parameters and regularization coefficients are adjusted until the model converges.

6. The online oil detection system for cement ball mills according to claim 5, characterized in that: The wear assessment and trend analysis module, based on the original vibration signal matrix, obtaining real-time wear assessment results for condition monitoring further includes: The support vector regression model is used to quantitatively calculate the wear degree of the current feature vector. The wear grades of the bearing and the gear are determined according to the regression output value. By comparing with the preset wear threshold, the health state of the component is judged, and the real-time wear evaluation result is obtained for condition monitoring.

7. The on-line oil detection system for cement ball mill according to claim 1, wherein: The multi-channel vibration sensor array includes at least three channels, which are respectively arranged at the positions of the bearing seat, the gearbox and the transmission shaft of the ball mill.

8. The on-line oil detection system for cement ball mill according to claim 1, characterized in that: The fault warning and model adaptation module has a warning level classification mechanism. According to the wear rate and the prediction result of the remaining life, multiple thresholds are set, and warning instructions and maintenance suggestions of different levels are output.

9. The online oil detection system for cement ball mills according to claim 1, wherein It also includes a data communication module, which is used to transmit the wear evaluation result, the trend data and the warning information to the upper monitoring platform or the remote expert system in real time through industrial Ethernet or wireless communication.

10. The on-line oil fluid detection method for cement ball mills, which uses the on-line oil fluid detection system for cement ball mills described in any one of claims 1-9, is characterized in that, The method includes: The multi-channel vibration sensor array is used to synchronously collect data at the positions of the bearing seat, the gearbox and the transmission shaft of the ball mill. The original vibration signal matrix containing multi-component coupling information is obtained through high-frequency sampling. If the sampling frequency is lower than ten times the wear characteristic frequency, the sampling parameters are automatically adjusted until the Nyquist theorem requirements are met. According to the original vibration signal matrix, the real-time wear evaluation result is obtained for condition monitoring. Through the sliding time window technology, the trend analysis of the wear evaluation results in a continuous time period is carried out. The least squares method is used to fit the wear evolution curve, and the wear development rate and acceleration are calculated according to the curve slope and the second derivative, and the wear evolution law parameters are obtained for prediction analysis. According to the wear evolution law parameters, a fault warning mechanism is established. If the wear development rate exceeds the safety threshold or the predicted remaining life is lower than the maintenance period, a warning signal is triggered and a maintenance suggestion is generated, and the model parameters are updated in real time to adapt to the changes in the operating conditions of the equipment.

Images