Ore crushing equipment operation monitoring method

Abstract

The invention relates to an ore crushing equipment operation monitoring method. Comprising the following steps: establishing a physical field distribution equation including a vibration wave propagation path and sound wave attenuation characteristics in a crushing cavity in the ore crushing process; collecting multi-physical field data in real time by using a sensor, and converting the data into a characteristic field distribution diagram comprising a vibration intensity distribution diagram; calculating a real-time change rate of a working load through historical data and current data of equipment operation; the vibration intensity and the sound wave energy are combined, and three working conditions of a low-load area, a normal-load area and an overload area are defined in a partitioned mode; on the time scale, the signals are divided into short-time fluctuation and long-time trend, and two types of tracks are extracted through wavelet decomposition and a sliding window method; on the characteristic scale, analyzing the main frequency deviation of the vibration signal, the high-frequency energy peak value distribution of the sound wave signal and the gradient of temperature change; and combining a working condition partitioning result with a dynamic characteristic track, and establishing a working condition and characteristic linkage matrix.

Description

Technical Field

[0001] The invention relates to a method for monitoring equipment operation, in particular to a method for monitoring ore crushing equipment operation. Background Art

[0002] Among the existing technologies for ore crushing equipment operation monitoring methods, the Chinese invention with patent number 201910551256.4 is the representative. Its clustering-based equipment operation monitoring method has certain advancement, but also has obvious shortcomings and disadvantages, which are specifically manifested as follows:

[0003] First, this method relies on the quality and comprehensiveness of the training data in terms of data processing. The key steps adopted are to group the training data set through correlation analysis and generate the first clustering model using the K-means clustering algorithm. However, this method has limited performance when there are complex nonlinear relationships and high-dimensional features in the equipment operation data. Especially for ore crushing equipment, its operation process is usually accompanied by strong nonlinear changes in multiple signals (such as sudden changes in vibration frequency, drastic fluctuations in load, and dynamic changes in temperature). These complex signals are difficult to fully reflect the operation rules and abnormal characteristics of the equipment only through linear correlation analysis and K-means clustering. At the same time, the K-means clustering algorithm is essentially a linear partitioning method, assuming that the data distribution is convex, which does not match the complex and asymmetric signal feature distribution during the operation of ore crushing equipment, which may lead to deviations and misjudgments in the clustering results. Secondly, this method is highly dependent on the accuracy and representativeness of the initial training data set, but the operating environment and material characteristics of the equipment operation are highly dynamic and diverse. If the training data cannot fully cover all operating conditions of the equipment (such as subdivided states of light load, medium load, and high load) or does not contain certain abnormal patterns, the first clustering model will be difficult to effectively reflect the actual operating status of the equipment. This will cause the real-time operation judgment point to not match the clustering model, thereby increasing the probability of false alarms and missed reports. In addition, the operating state of the ore crushing equipment will change over time (such as wear, fatigue effects, etc.), but this method does not provide a mechanism for dynamically updating the first clustering model, which is difficult to adapt to the long-term operating state changes of the equipment, and the risk of the model gradually failing is high. Third, the real-time monitoring process of this method is relatively simple, and it only determines whether the equipment operating state is normal by the position of the judgment point in the clustering model, lacking the ability to deeply analyze the dynamic characteristics of the signal. Short-term anomalies (such as instantaneous vibration shock) and long-term trend changes (such as gradual increase in load or slow increase in temperature) in the operation of ore crushing equipment may have an important impact on the health status of the equipment, and this method only relies on the position of the static judgment point, ignoring the time series characteristics of the signal and the law of multi-dimensional trend changes. In addition, this method does not introduce any dynamic threshold adjustment mechanism. When the working environment changes (such as ore hardness and particle size fluctuations), the judgment criteria cannot be flexibly adjusted, making it difficult to meet the complex and changing needs of the operating environment of ore crushing equipment.

[0004] Fourth, this method does not provide further fault classification and early warning grading capabilities for abnormal judgment. When the judgment point falls into the abnormal area, it only prompts that the equipment is operating abnormally, but does not clarify the type of abnormality (such as vibration abnormality, temperature abnormality or load abnormality) and its possible cause of failure, which makes it difficult for the operator to make response decisions. At the same time, this method does not provide a graded early warning mechanism and cannot distinguish between minor abnormalities and major failures, which may lead to potential risks of excessive alarms or ignoring early abnormal signals. In addition, this method has not established an abnormal sample library or historical data learning mechanism, and cannot improve the judgment model through historical abnormal data, and lacks the ability to optimize the failure mode in the long term. Summary of the invention

[0005] The purpose of the present invention is to provide a method for monitoring the operation of ore crushing equipment, so as to solve some of the disadvantages and shortcomings pointed out in the background technology.

[0006] The present invention solves the above-mentioned technical problems by adopting the following technical solutions:

[0007] S1. Physical field mapping analysis of operation behavior:

[0008] S1.1. Establish the physical field distribution equations of the ore crushing process, including the vibration wave propagation path and sound wave attenuation characteristics in the crushing chamber;

[0009] S1.2, using sensors to collect multi-physical field data in real time, and converting the data into characteristic field distribution maps including vibration intensity distribution maps;

[0010] S2. Work condition partitioning based on dynamic workload:

[0011] S2.1. Calculate the real-time change rate of workload through the historical and current data of equipment operation;

[0012] S2.2, combining vibration intensity and sound wave energy, define three working conditions: low load area, normal load area, and overload area;

[0013] S3. Multi-scale analysis of dynamic feature trajectories:

[0014] S3.1. In terms of time scale, the signal is divided into short-term fluctuations and long-term trends, and two types of trajectories are extracted by wavelet decomposition and sliding window method;

[0015] S3.2. Analyze the main frequency deviation of the vibration signal, the high-frequency energy peak distribution of the acoustic signal, and the gradient of temperature change at the characteristic scale;

[0016] S4. Abnormal pattern recognition based on working condition and feature linkage

[0017] S4.1. Combine the working condition partitioning results with the dynamic feature trajectory to establish a working condition and feature linkage matrix;

[0018] S4.2, apply anomaly detection algorithms to the linkage matrix to identify potential abnormal patterns;

[0019] S4.3, extract typical abnormal patterns including vibration frequency mutation caused by overload, and establish an abnormal sample library;

[0020] S5. Adaptive dynamic reconstruction of running status

[0021] S5.1. When an abnormal mode is identified, the operating parameters including the rotation speed and feed rate are adaptively adjusted through the state model;

[0022] S5.2, using recursive neural network RNN ​​to learn historical patterns and dynamically optimize the operation status prediction;

[0023] S6. Early warning generation based on multi-dimensional trend coupling:

[0024] S6.1. Conduct coupling analysis on the load change trend, vibration characteristic trend and temperature gradient change trend of equipment operation to construct a comprehensive risk assessment index;

[0025] S6.2. When the evaluation index exceeds the threshold, a graded warning including minor abnormalities and major failures is triggered.

[0026] Furthermore, the multi-scale analysis method of the dynamic feature trajectory includes:

[0027] Through the dynamic wavelet basis function selection algorithm, a dynamically adapted wavelet basis function is constructed according to the transient and non-stationary characteristics of the signal; the dynamically selected wavelet basis function is defined as:

[0028]

[0029] Where: ψ(t) is the dynamic wavelet basis function used for signal decomposition; t represents the time variable; a controls the time domain expansion range of the wavelet basis, and a larger value is suitable for capturing low-frequency components; b is the frequency modulation factor, which determines the center frequency of the wavelet basis function; c is the amplitude factor of the high-frequency component; the signal x(t) is decomposed by the wavelet basis to obtain the short-term fluctuation x of the high-frequency component. h (t) and the long-term trend of the low-frequency component x l (t).

[0030] Furthermore, the multi-scale analysis method of the dynamic feature trajectory includes:

[0031] The dual-scale sliding window algorithm is introduced; the transient fluctuation characteristics are captured through the short window, and the trend characteristics are extracted through the long window; on this basis, the trend adjustment module is added, and the local mean of the signal and nonlinear fitting are used to purify the trend characteristics; the trend function f(t) is defined as:

[0032]

[0033] Where: f(t) represents the trend value at time t; w is the width of the sliding window, a short window is used to quickly capture signal fluctuations, and a long window is used to extract smooth trends; x(τ) is the original signal value at time τ; Indicates the average value of the signal within the window; αt 2 is the nonlinear trend compensation term, where α is the adjustment coefficient used to eliminate the baseline drift in the signal.

[0034] Furthermore, the multi-scale analysis method of the dynamic feature trajectory includes:

[0035] Based on the joint analysis of frequency domain and time domain, key features are extracted dynamically. The main frequency deviation trajectory of vibration signal is extracted by short-time Fourier transform (STFT). The energy peak of acoustic signal is extracted by high-frequency energy clustering technology. The dynamic change gradient of temperature signal is analyzed by first-order derivative and second-order derivative to construct temperature gradient matrix for abnormal detection of equipment thermal load distribution. The temperature gradient matrix is ​​defined as:

[0036]

[0037] Where: G is the temperature gradient matrix; Indicates time point t i The rate of temperature change at Indicates time point t i The temperature change acceleration at the location; n is the number of time points, which is used to describe the resolution of the time series; through cluster analysis of the matrix G, the abnormal areas of the equipment thermal load distribution are identified to locate the potential fault source.

[0038] Furthermore, the multi-dimensional trend coupling early warning generation method includes:

[0039] Extract the short-term fluctuation and long-term trend characteristics of the signal from the time series; calculate the rate of change of the signal in different time periods by the sliding window method, and separate the high-frequency disturbance and low-frequency trend of the signal; normalize the extracted signal trend parameters; the mathematical expression of normalization is:

[0040]

[0041] Where: x iis the original eigenvalue of the ith signal; min(X) is the minimum value of the signal feature sequence X; max(X) is the maximum value of the signal feature sequence X; y i is the dimensionless eigenvalue after normalization.

[0042] Furthermore, the multi-dimensional trend coupling early warning generation method includes:

[0043] Through time series correlation analysis, the synergistic effects between trends are quantified to form a comprehensive representation of multi-dimensional trends; the trend coupling intensity is expressed by the following formula:

[0044]

[0045] in:

[0046] C(t) is the trend coupling strength at time t, which is used to measure the comprehensive impact of the coordinated changes of multiple signals; L(τ) is the load trend function at time τ; V(τ) is the vibration characteristic trend function at time τ; G(τ) is the temperature gradient trend function at time τ; w1, w2, w3 are the weights of load, vibration and temperature signals, respectively, reflecting the importance of each signal in the comprehensive evaluation; Represent the time change rate of load, vibration and temperature signals respectively.

[0047] Furthermore, the multi-dimensional trend coupling early warning generation method includes:

[0048] The operation status of the equipment is quantified and a comprehensive risk assessment index is constructed based on the trend coupling results. The comprehensive risk index integrates the trend characteristics of multi-dimensional signals and reflects the contribution of different signals to the equipment status through weight distribution. The calculation formula of the comprehensive risk assessment index is:

[0049]

[0050] Where: R is the comprehensive risk assessment index, which is used to quantify the current comprehensive health status of the equipment; T i is the dimensionless eigenvalue of the ith signal trend; W i is the weight of the ith signal, n is the number of signals involved in the evaluation; the numerator in the formula is the weighted sum of all signal eigenvalues ​​and their corresponding weights, and the denominator is the sum of all weights; the weight W i The design is based on historical equipment operation data.

[0051] Furthermore, the multi-dimensional trend coupling early warning generation method includes:

[0052] The dynamic value of the comprehensive risk assessment index R, the threshold setting is adjusted according to the equipment type and operating conditions; the dynamic threshold determination mechanism is divided into the following two levels: Slight abnormality: When R exceeds the slight abnormality threshold Rl When R exceeds the major fault threshold R h When the device is in serious risk, it is recommended to stop the machine for inspection immediately.

[0053] The present invention provides a method for monitoring the operation of ore crushing equipment. By organically combining technologies such as multi-dimensional trend coupling, dynamic feature trajectory analysis, real-time risk assessment and graded early warning, the method realizes comprehensive, accurate and intelligent monitoring of the equipment operation status, and has the following beneficial effects:

[0054] Ming has constructed a multi-dimensional feature extraction and trend coupling model through joint analysis of multiple signals such as load, vibration, and temperature, which can fully reflect the operating status and potential risks of the equipment and avoid omissions and misjudgments caused by traditional single signal monitoring. Combined with dynamic wavelet decomposition, sliding window analysis, and real-time risk assessment, it can quickly respond to short-term fluctuations and long-term trend changes in equipment operation, realize real-time capture of signal dynamic characteristics, and adapt to the needs of rapid monitoring under complex working conditions.

[0055] Through normalization processing, multi-dimensional trend coupling analysis and calculation of comprehensive risk assessment index, this method significantly improves the accuracy of identifying abnormal conditions and potential faults, especially in the early stages, it can accurately identify minor abnormalities and reduce the risk of fault expansion. The dynamic threshold setting and adaptive adjustment mechanism can automatically optimize the early warning strategy based on the equipment operation history data and current working conditions, provide intelligent early warning and decision support, and reduce the dependence of operators.

[0056] Flexible parameter settings (such as weight allocation, dynamic adjustment of thresholds, etc.) are provided for different equipment types and working conditions, which can be widely applied to the operation monitoring needs of various complex ore crushing equipment. : By establishing an abnormal sample library and a historical data learning mechanism, this method supports the continuous optimization and improvement of the equipment operation mode, providing a reliable technical basis for the full life cycle management of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 The present invention is a flow chart of a method for monitoring the operation of ore crushing equipment.

[0058] Figure 2 This is a flow chart of the multi-scale analysis method of dynamic feature trajectories of the present invention.

[0059] Figure 3 This is a flow chart of the multi-dimensional trend coupling early warning generation method of the present invention. DETAILED DESCRIPTION

[0060] The specific implementation modes of the present invention will be described in detail below in conjunction with the accompanying drawings.

[0061] A method for monitoring the operation of ore crushing equipment: S1. Physical field mapping analysis of operation behavior: By constructing a multi-physical field distribution model of the ore crushing process, the vibration and acoustic wave characteristics of the equipment during operation are fully analyzed. Specifically, the following steps are included: First, the distribution equation of the physical field during the operation of the ore crushing equipment is established. The distribution equation of the vibration wave propagation path is based on wave theory and structural dynamics model, describing the transmission and attenuation characteristics of vibration in various areas inside the crushing chamber; the acoustic wave attenuation characteristics are modeled through the acoustic propagation model, reflecting the reflection, scattering and absorption behavior of the acoustic wave on the air medium and solid wall. Secondly, multimodal sensors arranged at key positions are used to collect the vibration signals and acoustic wave signals generated during the operation of the ore crushing equipment in real time. These sensors include acceleration sensors, pressure sensors and acoustic sensors. The collected data is processed by signal synchronization to ensure the temporal consistency and spatial distribution correlation between the multi-physical field signals. Then, the collected physical field data is input into the feature extraction module, and the data is jointly analyzed in the time and frequency domains through the improved wavelet transform algorithm and high-frequency feature enhancement technology. Through this process, the key characteristic parameters of vibration and sound wave signals are extracted, such as the amplitude, main frequency and propagation direction of the vibration wave, as well as the spectrum energy distribution and attenuation coefficient of the sound wave. Finally, based on the extracted characteristic parameters, characteristic field distribution diagrams are generated. These distribution diagrams intuitively reflect the vibration intensity distribution at different positions in the crushing chamber, the sound wave energy distribution and its change trend over time. This characteristic field distribution diagram not only provides a comprehensive visual analysis of the equipment's operating status, but also provides basic data support for subsequent fault diagnosis and operation optimization.

[0062] S2. Working condition partitioning based on dynamic workload: Using the historical data and current real-time data of equipment operation, dynamically analyze the changing trend of workload, and combine multi-dimensional signals such as vibration intensity and acoustic wave energy to scientifically divide the equipment operating condition area. First, obtain the historical load data of the equipment's long-term operation, which includes parameters such as feed volume, pressure in the crushing chamber, and power consumption of the transmission system, and collect the current real-time operation data of the equipment. Through time series analysis, calculate the real-time change rate of the workload, use the improved differential method to obtain the incremental change of the load over time, and combine the sliding average algorithm to eliminate the noise caused by short-term fluctuations. Secondly, combine the equipment vibration intensity and acoustic wave energy signals to establish a multi-dimensional working condition feature space. In this feature space, vibration intensity characterizes the dynamic characteristics of mechanical motion, while acoustic wave energy reflects the material state and equipment stress during ore crushing. Through data fusion, the real-time load change rate, vibration intensity and acoustic wave energy parameters are jointly modeled, and the characteristic pattern of the equipment operating state is automatically identified using the cluster analysis algorithm. Based on the characteristic pattern, three working conditions are further defined: low load zone, normal load zone and overload zone: the low load zone corresponds to a state with low load change rate, small vibration intensity fluctuation and low sound wave energy, which usually indicates that the equipment is running at no load or light load; the normal load zone corresponds to a state with stable load change rate, vibration intensity and sound wave energy both within the normal range, indicating that the equipment runs smoothly and the material handling efficiency is high; the overload zone corresponds to a state with a sudden increase in load change rate, significantly increased vibration intensity and sound wave energy exceeding the set threshold, which usually reflects that the equipment is at risk of overload operation.

[0063] S3. Multi-scale analysis of dynamic feature trajectories: By constructing a dual analysis framework of time scale and feature scale, the operating signals of ore crushing equipment are comprehensively analyzed to accurately capture the dynamic characteristics of equipment state changes. First, on the time scale, in order to simultaneously obtain short-term fluctuation characteristics and long-term trend information, the collected multimodal signals (such as vibration signals, acoustic signals, and temperature signals) are decomposed at multiple resolutions. The wavelet decomposition method is used to decompose the signal into high-frequency components and low-frequency components. The high-frequency components are used to characterize short-term fluctuation characteristics, and the low-frequency components are used to extract long-term trends. In order to further ensure the accuracy of signal decomposition, the sliding window technology is combined to localize the signal by setting the window length. The short window (such as 100ms) is used to capture instantaneous changes, and the long window (such as 10s) is used to extract stable trends. Secondly, on the feature scale, feature extraction methods are designed according to the characteristics of different signals. For vibration signals, we analyze the deviation trajectory of the main frequency over time, and track the dynamic changes of the main frequency position in the spectrum through Fourier transform, so as to identify the non-stationary characteristics of mechanical parts; for sound wave signals, we focus on the distribution characteristics of high-frequency energy, and use high-frequency energy clustering analysis technology to extract the energy peak in the sound wave signal and quantify its distribution characteristics. This process can keenly capture the abnormal acoustic characteristics of materials during the crushing process; for temperature signals, we extract the gradient information of temperature changes by calculating the first-order derivative and the second-order derivative, and dynamically reflect the changes in the heat load inside the equipment.

[0064] S4. Abnormal pattern recognition based on working condition and feature linkage: Through the linkage analysis of working condition partitioning and dynamic feature trajectory, a complete set of abnormal pattern recognition methods is constructed to achieve accurate positioning and early warning of potential equipment failures. First, according to the working condition partitioning method, the equipment operating state is divided into low load zone, normal load zone and overload zone, and the partition results of the current operating condition are obtained through real-time monitoring. At the same time, the dynamic feature trajectory analysis method is used to extract the key features of the signal from the time scale and feature scale, including the main frequency offset of the vibration signal, the high-frequency energy peak distribution of the acoustic signal, and the gradient information of the temperature change. These dynamic features can describe the multi-dimensional performance of the equipment operating state in detail. Then, the working condition partitioning results are linked with the dynamic feature trajectory to construct the working condition and feature linkage matrix. The rows of the matrix represent the categories of working condition partitions (low load zone, normal load zone, overload zone), and the columns represent the extracted dynamic feature parameters (such as vibration frequency, acoustic wave energy, temperature gradient). Each element in the matrix is ​​the statistical value or change trend of the feature parameter under a specific working condition. By analyzing the linkage matrix, the intrinsic connection between specific working conditions and characteristic changes can be revealed. For example, the overload area is usually accompanied by a significant shift in the main vibration frequency, a sudden increase in high-frequency energy, and an abnormal change in the temperature gradient. Then, the anomaly detection algorithm is applied to the working condition and characteristic linkage matrix to identify potential abnormal patterns. The isolated forest algorithm or the density peak clustering algorithm is used to mine abnormal points that do not conform to the normal operating rules according to the distribution of characteristic parameters in the linkage matrix, and they are marked as potential abnormal patterns. Further extraction of typical abnormal patterns including sudden changes in vibration frequency caused by overload, and establishment of an abnormal sample library, which records the characteristic description of the abnormal pattern, the conditions for occurrence, and the corresponding possible causes of failure, provides empirical data support for subsequent abnormal pattern recognition.

[0065] S5. Adaptive dynamic reconstruction of operating status: By real-time identification and dynamic response to abnormal modes of equipment operating status, an adaptive operating status reconstruction mechanism is established to achieve intelligent optimization and stable control of equipment operation. First, when the abnormal mode recognition module detects problems such as overload, abnormal vibration, abnormal temperature, etc. during equipment operation, the system will trigger an adaptive adjustment mechanism based on the current working conditions and abnormal characteristics. This mechanism dynamically adjusts key operating parameters, including speed and feed rate, based on the relationship between the historical data corresponding to the abnormal mode and the equipment operating parameters. For example, when a sudden change in vibration frequency caused by overload is detected, the system will automatically reduce the speed of the equipment to reduce the load, while optimizing the feed rate to ensure the operating stability of the equipment. Then, in order to further improve the accuracy of the operating status prediction, a recursive neural network (RNN) model is used to perform deep learning on the historical data of equipment operation. Through its time series processing capability, the RNN model can capture the long-term dependency between equipment operating parameters, abnormal characteristics and operating condition changes, and form a dynamic prediction capability for the equipment operating status. During the model training process, historical operating data is input, including equipment load changes, vibration characteristics, temperature gradients, and corresponding operating parameter adjustment records. The model weights are optimized through the back-propagation algorithm to ensure that the model can accurately predict future operating conditions and possible abnormal trends. In real-time operation, the RNN model predicts possible future operating condition changes and abnormal risks based on the currently collected multi-dimensional signals and identified abnormal patterns, and gives targeted operation adjustment suggestions. For example, when the model predicts that the equipment is about to enter an overload state, the feed rate can be adjusted in advance or the speed can be reduced to avoid abnormalities.

[0066] S6. Early warning generation by multi-dimensional trend coupling: By coupling and analyzing the load change trend, vibration characteristic trend and temperature gradient change trend of equipment operation, a comprehensive risk assessment index is constructed to achieve real-time monitoring and graded early warning of equipment operation status. First, multi-dimensional dynamic data is collected from the equipment operation process, including time series of load, vibration and temperature signals. The load change trend reflects the working pressure of the equipment when processing ore, the vibration characteristic trend describes the dynamic stability of the mechanical structure of the equipment during operation, and the temperature gradient change trend reveals the thermal energy distribution state of the equipment under high load. After the time series of these signals are normalized, the short-term fluctuations and long-term trends are extracted by the sliding window method to form the trend curves of each signal. Secondly, through the coupling analysis method, these trend data are multi-dimensionally correlated and modeled to construct comprehensive features reflecting the synergy between signals. For example, when the load trend increases significantly, the vibration amplitude increases and the temperature gradient changes accelerate. This signal coupling usually indicates that the equipment may be in an overloaded operation state. Based on these coupling characteristics, a comprehensive risk assessment index is defined, and each trend signal is comprehensively calculated using a weighted model, where the weight value is determined by historical data and expert experience, reflecting the degree of influence of each signal on the health status of the equipment. The value range of the risk assessment index is usually set between 0 and 1, where 0 means that the operation is completely safe and 1 means that the equipment has serious risks. When the risk assessment index reaches or exceeds the set threshold, the system automatically triggers a graded warning mechanism to provide multi-level warnings from minor abnormalities to major faults. Minor abnormality warnings are triggered when the assessment index slightly exceeds the low threshold, usually indicating short-term fluctuations or small-scale fault risks. At this time, the system will recommend the operator to adjust parameters (such as load reduction or speed reduction) to avoid further deterioration; major fault warnings are triggered when the assessment index significantly exceeds the high threshold, indicating that the equipment has entered a high-risk state, and the system will recommend immediate shutdown inspection and maintenance measures. Through this multi-dimensional trend coupling warning generation method, the accuracy and sensitivity of equipment operation monitoring can be effectively improved, while achieving differentiated responses to minor abnormalities and major faults, providing strong technical guarantees for the safety and stability of equipment operation.

[0067] Embodiment 1:

[0068] In a certain ore crushing plant, a main crushing equipment often malfunctions and stops due to unstable operation, which seriously affects production efficiency. In order to improve the operational reliability of the equipment, the ore crushing plant decided to use a multi-scale analytical method of dynamic feature trajectories to monitor the operation of the equipment. The main signals generated during the operation of the equipment include vibration signals x(t), whose frequency range is between 0.1Hz and 500Hz, and there are short-term vibration shocks and long-term trend drifts. Through this method, the signal can be decomposed and analyzed, short-term anomalies can be identified, and long-term trends can be analyzed, ultimately achieving comprehensive monitoring of the equipment's operating status.

[0069] 1. Dynamic wavelet basis function selection: According to the transient and non-stationary characteristics of the signal, a dynamic wavelet basis function selection algorithm is used to construct a wavelet basis function that adapts to the current signal characteristics. The dynamic wavelet basis function is defined as:

[0070]

[0071] in:

[0072] ψ(t): dynamic wavelet basis function;

[0073] t: time variable;

[0074] a: time domain extension range control parameter, value range 0.1≤a≤5.0;

[0075] b: frequency modulation factor, value range 10≤b≤500Hz;

[0076] c: high-frequency component amplitude factor, ranging from 0.5≤c≤10.0.

[0077] Actual parameter selection: Assuming that the main feature of the current equipment vibration signal is medium-frequency vibration with rapid transient changes, set a=1.0, b=100Hz, and c=2.0.

[0078] 2. Signal decomposition: Input the vibration signal x(t) collected during the operation of the equipment into the wavelet decomposition module. The dynamic wavelet basis function performs multi-resolution decomposition on the signal by matching the current signal characteristics and decomposing it into high-frequency components x h (t) and low-frequency component x l (t). The sampling frequency is 1kHz and the signal duration is 10 seconds.

[0079] Calculation results: Decompose the collected x(t) (assuming that its main feature is periodic fluctuations superimposed on instantaneous shocks) to obtain: high-frequency component x h (t): Short-term fluctuations in the vibration frequency range of 80–200 Hz were captured, and the analysis results showed a significant impact feature at t = 3.5 seconds (the amplitude was approximately 3.2 times the baseline value).

[0080] Low frequency component x l (t): reflects the long-term trend of the device at 0.1–10 Hz. It is found that the signal amplitude gradually increases after t = 7 seconds, which may be related to the gradual overload of the device.

[0081] 3. Multi-scale analysis results: Short-term fluctuation analysis: By hFurther feature extraction of (t) identifies transient impact features, including impact amplitude, frequency range, and impact duration. It is found that the impact of the equipment at t = 3.5 seconds may be related to abnormal ore jamming in the crushing chamber. Long-term trend analysis: l (t) Trend fitting was performed and it was found that the trend of gradually increasing signal amplitude was consistent with the characteristics of equipment overload operation, which was consistent with the load monitoring data, further verifying the potential risks of the equipment.

[0082] In order to further accurately capture the transient fluctuations and long-term trend characteristics of the equipment operation status, it is necessary to introduce a dual-scale sliding window algorithm and a trend adjustment module. This method can extract the local features of the signal while eliminating the pseudo features caused by the signal baseline drift, ensuring a more accurate comprehensive analysis of the equipment operation status.

[0083] According to the sampling frequency (1kHz) and signal characteristics of the equipment operation signal, set the short window and long window. The short window width w s The selection is 100ms (100 sampling points) to capture transient fluctuation characteristics; the long window width w l 10s (10,000 sampling points) is selected to extract the smooth trend of the signal. The vibration signal x(τ) of the equipment operation is the actual data in a certain period of time, and the sampling value range is [-3,3]. The signal presents periodic fluctuations superimposed with instantaneous shocks.

[0084] The trend function f(t) is calculated using the following formula:

[0085]

[0086] in:

[0087] f(t): trend value at time t;

[0088] w: sliding window width, short window w s =100ms, long window w l =10s;

[0089] x(τ): original signal value at time τ;

[0090] The average value of the signal within the window range;

[0091] αt 2 : Nonlinear trend compensation term, where the adjustment coefficient α=0.005 (value range 0.001≤α≤0.01), is used to correct the trend deviation caused by baseline drift.

[0092] Assume that at t = 3 seconds, the signal value in the sliding window is [1.2, 1.4, 1.3, 1.5, 1.6].

[0093] Short window w s = Within 100ms, the signal average value is:

[0094]

[0095] Long window w l = Within 10s, the signal average value is:

[0096]

[0097] Nonlinear compensation term:

[0098] αt 2 =0.005·(3) 2 =0.045.

[0099] Comprehensive trend value:

[0100] f(3)=1.4+0.045=1.445.

[0101] The trend value calculated by the short window reflects the transient fluctuation of the signal and locates the signal anomaly in a short period of time. For example, the significant fluctuation at t = 3 seconds reflects the impact phenomenon that may occur in the crushing chamber. The long window result provides smooth trend information of the signal. For example, it is found that the trend value of t∈[7,10] seconds gradually increases, indicating that the equipment may enter a high-load operation state. The nonlinear compensation term corrects the influence of the signal baseline drift, making the trend value more accurate.

[0102] Through the dual-scale sliding window and trend adjustment module, the operating signal of the ore crushing equipment is decomposed into two parts: transient fluctuations and long-term trends. Transient fluctuations are used to capture abnormal behaviors of the equipment (such as shocks and vibration abnormalities), and long-term trends are used to evaluate the overall operating status of the equipment (such as load changes and fatigue accumulation). In the example, the system successfully detected the instantaneous impact of t = 3 seconds and predicted the high load trend of t = 7-10 seconds, providing operators with a decision-making basis for load reduction and maintenance. By dynamically adjusting the window width w and the compensation coefficient α, this method adapts to different signal characteristics and working conditions, verifying its practicality and reliability in equipment operation monitoring.

[0103] In order to gain a deeper understanding of the equipment's operating status, it is necessary to combine the frequency domain and time domain joint analysis techniques to extract dynamic features of vibration, sound waves, and temperature signals, identify key signal characteristics, and locate abnormal areas. In this example, we will show how to use short-time Fourier transform (STFT), high-frequency energy clustering analysis, and temperature gradient matrix to monitor the equipment's operating status and diagnose faults.

[0104] The vibration signal x(t) collected during the operation of the equipment is subjected to short-time Fourier transform (STFT) to analyze the change trajectory of the signal's main frequency over time. Assuming that the signal sampling frequency is 1kHz and the signal duration is 10 seconds, a Hanning window with a window width of 100ms is used for segmented analysis. Through calculation, the spectrum distribution is obtained, in which the main frequency trajectory is expressed as a curve that changes over time.

[0105] At t = 3.5 seconds, the main frequency quickly shifted from 120Hz to 200Hz for 0.3 seconds. This sudden change may reflect abnormal vibration in the crushing chamber, which may be caused by ore blockage or abnormal mechanical parts. The acoustic wave signals generated during the operation of the crushing equipment are collected, and the energy peaks of the key frequency bands are extracted using the high-frequency energy clustering analysis method. The high frequency band of the signal is defined as 2kHz to 10kHz, and the energy center value in each time period is found through spectrum analysis.

[0106] At t = 7 seconds, the high-frequency energy is concentrated around 6.5kHz and is significantly higher than the energy level in other periods, with the energy value reaching 20% ​​higher than the baseline level. Combined with the equipment operation status, this may point to high-frequency friction caused by excessively large ore particles or abnormal clearance between mechanical parts.

[0107] Temperature sensors are placed on key parts of the equipment (such as motors and bearings) with a sampling frequency of 1 Hz to record the temperature data T(t) at each time point. and the second derivative Obtain the temperature change rate and acceleration, and construct the temperature gradient matrix G:

[0108]

[0109] Wherein: n=10 means the resolution of 10 time points.

[0110] Assume that the temperature data is [40,42,45,50,60,62,63,65,67,70] (unit: °C).

[0111] First-order derivative calculation: like

[0112] Second order derivative calculation: like

[0113] The resulting matrix G is:

[0114]

[0115] Through cluster analysis of the temperature gradient matrix G, it is found that the temperature change rate and acceleration at t = 4 seconds deviate significantly from other time points (rate is 10℃ / s, acceleration is -8℃ / s2 ), indicating that the device may have excessive heat load or heat dissipation failure.

[0116] Embodiment 2:

[0117] In a certain ore crushing plant, a main crushing equipment had intermittent faults during long-term operation. However, since the fault symptoms were brief and difficult to detect, the equipment frequently stopped, causing a significant impact on production. The plant decided to introduce a multi-dimensional trend coupling early warning generation method, which dynamically analyzes the load, vibration and temperature signals during equipment operation, identifies short-term fluctuations and long-term trend characteristics, generates a comprehensive risk assessment index, and provides real-time early warning support for the equipment.

[0118] Install load sensors, vibration sensors and temperature sensors at key locations of the equipment to collect load changes x L (t), vibration intensity x V (t) and temperature x T (t) time series signal. Assuming the sampling frequency is 1Hz and the signal acquisition time is 10 seconds, the following data is obtained:

[0119] Load signal x L (t)=[500,520,530,550,560,580,600,610,620,640]kg;

[0120] Vibration signal x V (t)=[0.8,0.85,0.95,1.0,1.1,1.2,1.25,1.3,1.4,1.5]g;

[0121] Temperature signal x T (t)=[40,42,44,47,50,53,56,60,63,67]℃.

[0122] The sliding window method is used to calculate the rate of change of each signal in different time periods, separating the high-frequency disturbance (short-term fluctuation) and low-frequency trend (long-term trend) of the signal. The window width w = 3 seconds.

[0123] Calculate the load change rate: For example, at t = 3 seconds,

[0124]

[0125] According to this calculation, the load change rate sequence is [10, 15, 15, 15, 20, 20, 15, 20] kg / s.

[0126] The rates of change of the vibration and temperature signals are calculated in the same way;

[0127] The vibration change rate sequence is [0.03, 0.05, 0.05, 0.07, 0.05, 0.03, 0.07, 0.1] gg / s and the temperature change rate sequence is [1.0, 1.0, 1.5, 1.5, 2.0, 2.0, 2.5, 3.0] ℃ / s.

[0128] The change rate sequence is normalized to map the eigenvalues ​​to the dimensionless range of . The normalization formula is:

[0129]

[0130] in:

[0131] x i : the original eigenvalue of the ith signal; min(X): the minimum value of the feature sequence X; max(X): the maximum value of the feature sequence X; y i : Normalized dimensionless eigenvalue.

[0132] Actual calculation: Taking the load change rate as an example,

[0133] min(X)=10, max(X)=20;

[0134] For the first value x1=10,

[0135]

[0136] For the second value x2=15,

[0137]

[0138] The normalized load change rate sequence is [0.0, 0.5, 0.5, 0.5, 1.0, 1.0, 0.5, 1.0].

[0139] The same processing is performed on the vibration and temperature change rates, and the normalized sequence of the vibration change rate is [0.0, 0.4, 0.4, 0.6, 0.4, 0.0, 0.6, 1.0], and the normalized sequence of the temperature change rate is [0.0, 0.2, 0.4, 0.4, 0.6, 0.6, 0.8, 1.0].

[0140] The normalized characteristic values ​​of comprehensive load, vibration and temperature are combined to construct a comprehensive risk assessment index using the weighted average method:

[0141]

[0142] in:

[0143] R: comprehensive risk assessment index;

[0144] W L ,WV ,W T : are the weights of load, vibration and temperature signals respectively. Assume that W L =0.5,W V =0.3,W T =0.2.

[0145] Actual calculation: At t = 5 seconds,

[0146] Load normalized value y L =1.0, vibration normalized value y V =0.4, temperature normalized value y T =0.6;

[0147]

[0148] Set the threshold R for slight anomalies l =0.6, the threshold value of major failure R h =0.8. At t=5 seconds, R=0.74 exceeded the minor abnormality threshold, triggering a minor abnormality warning. The system prompted the operator to check the crushing chamber load and optimize the feed rate. At t=8 seconds, R=0.85 exceeded the major fault threshold, indicating that the equipment may have serious risks, and it is recommended to stop the machine immediately to check the mechanical parts.

[0149] In the above analysis, the load, vibration and temperature trends of the ore crushing equipment have been accurately characterized by normalization and extraction of short-term fluctuations and long-term trends. However, these single signals do not fully reflect the overall status of the equipment. In order to achieve a comprehensive analysis of the coordinated changes of signals and quantify their comprehensive impact on the health of the equipment, this stage uses time series correlation analysis to construct a comprehensive characterization model of multi-dimensional trend coupling and generate early warnings.

[0150] Extract the load trend L(τ), vibration characteristics V(τ) and temperature gradient G(τ) from the time series and calculate their time change rate. Assume that the time series data is as follows (sampling interval is 1 second):

[0151] Load signal: L(τ) = [500, 520, 530, 550, 560, 580, 600, 610, 620, 640] kg;

[0152] Vibration signal: V(τ)=[0.8, 0.85, 0.95, 1.0, 1.1, 1.2, 1.25, 1.3, 1.4, 1.5]g;

[0153] Temperature signal: G(τ)=[40,42,44,47,50,53,56,60,63,67]℃.

[0154] Actual calculation: Take the load change rate as an example:

[0155]

[0156] For example, at τ = 1 second:

[0157]

[0158] Similar calculations yield:

[0159]

[0160] The trend coupling strength C(t) is expressed by the following formula:

[0161]

[0162] in:

[0163] w1, w2, w3: are the weights of load, vibration and temperature signals respectively, assuming the value range is 0.1≤w i ≤1.0, normalized to w1=0.5,w2=0.3,w3=0.2.

[0164] Actual calculation: At t = 5 seconds:

[0165]

[0166] Substitute item by item:

[0167] At τ = 1 second:

[0168] 0.5 20 + 0.3 0.05 + 0.2 2.0 = 10 + 0.015 + 0.4 = 10.415.

[0169] At τ = 2 seconds:

[0170] 0.5·10+0.3·0.10+0.2·2.0=5+0.03+0.4=5.43.

[0171] Calculate and add them one by one, and we get:

[0172] C(5)=10.415+5.43+10.6+5.6+11.0=43.045.

[0173] Set the warning threshold according to the value of C(t): Slight abnormality threshold C l =40, major failure threshold C h=60. At t=5 seconds, C(5)=43.045, exceeding the minor abnormality threshold, triggering a minor abnormality warning, prompting the operator to check the equipment operating status, and suggesting appropriate load reduction. At t=8 seconds, if C(8)=62.5, exceeding the major fault threshold, triggering a major fault warning, it is recommended to immediately stop the machine to check whether there is ore blockage in the crushing chamber or abnormal mechanical parts.

[0174] In order to further quantify the operating status of the equipment and comprehensively consider the characteristic trends of multi-dimensional signals, the factory decided to construct a comprehensive risk assessment index R to quantify the health status of the equipment and provide a clear warning level. Combining the historical data of the equipment with the current real-time signal, the weight distribution method is used to accurately reflect the contribution of multiple signals.

[0175] The eigenvalues ​​of load, vibration and temperature signals are extracted from the real-time operation of the equipment and normalized. The dimensionless range of the eigenvalues ​​is [0,1]. The analysis of the equipment historical data shows that the load contributes the most to the failure risk, so the weight W is assigned. L =0.5; vibration and temperature are next, and weights W are assigned to each V =0.3 and W T =0.2.

[0176] Input data: Assume that the dimensionless eigenvalue of the signal at a certain moment is:

[0177] T L =0.8 (load characteristic value); T_V = 0.6 (vibration characteristic value); T T =0.7 (temperature characteristic value).

[0178] The comprehensive risk assessment index R is expressed by the following formula:

[0179]

[0180] Where n=3, indicating the number of signals involved in the evaluation.

[0181] Calculate the numerator:

[0182]

[0183] Calculate the denominator:

[0184]

[0185] Calculate R:

[0186]

[0187] According to the historical data of the equipment, set the warning threshold of the comprehensive risk assessment index:

[0188] R l=0.6: slight abnormality threshold;

[0189] R h =0.8: Major failure threshold.

[0190] Early warning judgment: Current R = 0.72, exceeding the slight abnormal threshold R l But below the major failure threshold R h , the system triggers a minor abnormality warning. The warning message prompts the operator: "The equipment load is too high. It is recommended to reduce the feed rate and check the mechanical vibration status." The system continuously monitors the equipment signal characteristics and updates the comprehensive risk assessment index in real time. For example, in the next 5 seconds, the signal changes as follows:

[0191] T L =0.9, T V =0.7, T T =0.8.

[0192] Recalculate R:

[0193]

[0194] At this time, R>R h , the system triggers a major fault warning and recommends immediate shutdown to check whether there is ore blockage in the crushing chamber or abnormal wear of mechanical parts.

[0195] Through the weight allocation and real-time calculation of multiple signal features, the comprehensive risk assessment index R quantifies the health status of the equipment with clear numerical values, realizing the combination of real-time monitoring and early warning. The example verifies the system's ability to respond to minor abnormalities when R = 0.72, and the ability to quickly warn of major faults when R = 0.82. Combined with the actual operating environment and equipment characteristics, the weight W i The flexible adjustment further improves the applicability and accuracy of this method, providing reliable guarantee for the efficient operation and maintenance of ore crushing equipment.

[0196] In this stage, a dynamic threshold mechanism is introduced to adjust the slight abnormal threshold R in real time. l and major failure threshold R h , optimize the early warning response strategy for different equipment types and working conditions, and further improve the adaptability and accuracy of the system.

[0197] Assume that the health status of the equipment under different operating conditions is distributed as follows: light load (30%), medium load (50%), and high load (20%). Based on historical data analysis, the distribution range of the comprehensive risk assessment index R value under each operating condition is as follows:

[0198] Light load: R∈[0.3,0.5], with an average value of 0.4;

[0199] Medium load: R∈[0.5,0.7], with an average value of 0.6;

[0200] High load: R∈[0.7,0.9], with an average of 0.8.

[0201] Set the initial threshold based on the equipment operation requirements:

[0202] Slightly abnormal threshold R l =μ+σ, where μ is the average value of the current operating condition R, and σ=0.1 is the standard deviation of the historical fluctuation range;

[0203] Major failure threshold R h =μ+2σ.

[0204] Actual setting: Assume that the equipment is currently operating under medium load conditions:

[0205] R l =0.6+0.1=0.7;

[0206] R h =0.6+2·0.1=0.8.

[0207] Calculate the changes in the comprehensive risk assessment index R in real time and compare it with the dynamic threshold. Assume that the following data are monitored:

[0208] Time t = 5 seconds, R = 0.68;

[0209] Time t = 10 seconds, R = 0.75;

[0210] Time t = 15 seconds, R = 0.82.

[0211] Determination mechanism:

[0212] At t = 5 seconds, R = 0.68 <R l =0.7, the equipment operates normally and no warning is needed.

[0213] At t = 10 seconds, R l =0.7 <R=0.75<R h =0.8, triggering a minor abnormal warning, prompting the operator: "The equipment load is too high. It is recommended to reduce the load appropriately to reduce the operating pressure."

[0214] At t = 15 seconds, R = 0.82 > R h =0.8, triggering a major fault warning, prompting the operator: "The equipment is at serious risk. It is recommended to stop the machine immediately to check the status of the crushing chamber and transmission components."

[0215] When the equipment operating conditions change, the warning threshold is adjusted dynamically. For example, the operator responds to a minor abnormality warning at t = 10 seconds and reduces the equipment feed rate to adjust it from a medium load condition to a light load condition. Reset the threshold according to the new operating condition characteristics:

[0216] R l =0.4+0.1=0.5;

[0217] R h =0.4+2·0.1=0.6.

[0218] After the major fault warning was triggered at t=15 seconds, the operator stopped the machine for inspection as recommended and found that there were large pieces of ore in the crushing chamber that caused the equipment to jam. After cleaning, the equipment resumed normal operation, and the system recorded the abnormal event as a historical sample for subsequent optimization of the dynamic threshold strategy. Through the setting and real-time adjustment of dynamic thresholds, both minor abnormalities and major faults of the equipment are effectively identified. In the example, the system triggered a minor abnormality warning and a major fault warning at t=10 seconds and t=15 seconds, respectively, providing the operator with timely and accurate decision support. This method combines the equipment operating conditions and historical data to dynamically adjust the threshold, which significantly improves the sensitivity and adaptability of monitoring, and helps to ensure the long-term stable operation of the equipment.

Claims

1. A method for monitoring the operation of ore crushing equipment, characterized in that The following steps are involved: S1. Physical field mapping analysis of operation behavior: S1.

1. Establish the physical field distribution equations of the ore crushing process, including the vibration wave propagation path and sound wave attenuation characteristics in the crushing chamber; S1.2, using sensors to collect multi-physical field data in real time, and converting the data into characteristic field distribution maps including vibration intensity distribution maps; S2. Work condition partitioning based on dynamic workload: S2.

1. Calculate the real-time change rate of workload through the historical and current data of equipment operation; S2.2, combining vibration intensity and sound wave energy, define three working conditions: low load area, normal load area, and overload area; S3. Multi-scale analysis of dynamic feature trajectories: S3.

1. In terms of time scale, the signal is divided into short-term fluctuations and long-term trends, and two types of trajectories are extracted by wavelet decomposition and sliding window method; S3.

2. Analyze the main frequency deviation of the vibration signal, the high-frequency energy peak distribution of the acoustic signal, and the gradient of temperature change at the characteristic scale; S4. Abnormal pattern recognition based on working condition and feature linkage: S4.

1. Combine the working condition partitioning results with the dynamic feature trajectory to establish a working condition and feature linkage matrix; S4.2, apply anomaly detection algorithms to the linkage matrix to identify potential abnormal patterns; S4.3, extract typical abnormal patterns including vibration frequency mutation caused by overload, and establish an abnormal sample library; S5. Adaptive dynamic reconstruction of running status: S5.

1. When an abnormal mode is identified, the operating parameters including the rotation speed and feed rate are adaptively adjusted through the state model; S5.2, using recursive neural network RNN ​​to learn historical patterns and dynamically optimize the operation status prediction; S6. Early warning generation based on multi-dimensional trend coupling: S6.

1. Conduct coupling analysis on the load change trend, vibration characteristic trend and temperature gradient change trend of equipment operation to construct a comprehensive risk assessment index; S6.

2. When the evaluation index exceeds the threshold, a graded warning including minor abnormalities and major failures is triggered.

2. A method for monitoring the operation of ore crushing equipment according to claim 1, characterized in that The multi-scale analysis method of the dynamic feature trajectory includes: Through the dynamic wavelet basis function selection algorithm, a dynamically adapted wavelet basis function is constructed according to the transient and non-stationary characteristics of the signal; the dynamically selected wavelet basis function is defined as: Where: ψ(t) is the dynamic wavelet basis function used for signal decomposition; t represents the time variable; a controls the time domain expansion range of the wavelet basis, and a larger value is suitable for capturing low-frequency components; b is the frequency modulation factor, which determines the center frequency of the wavelet basis function; c is the amplitude factor of the high-frequency component; the signal x(t) is decomposed by the wavelet basis to obtain the short-term fluctuation x of the high-frequency component. h (t) and the long-term trend of the low-frequency component x l (t).

3. A method for monitoring the operation of ore crushing equipment according to claim 2, characterized in that The multi-scale analysis method of the dynamic feature trajectory includes: introducing a dual-scale sliding window algorithm; capturing transient fluctuation characteristics through a short window and extracting trend characteristics through a long window; on this basis, adding a trend adjustment module and using the local mean and nonlinear fitting of the signal to purify the trend characteristics.

4. A method for monitoring the operation of ore crushing equipment according to claim 3, characterized in that The multi-scale analysis method of the dynamic feature trajectory includes: Based on the joint analysis of frequency domain and time domain, key features are extracted dynamically. The main frequency deviation trajectory of vibration signal is extracted by short-time Fourier transform (STFT). The energy peak of acoustic signal is extracted by high-frequency energy clustering technology. The dynamic change gradient of temperature signal is analyzed by first-order derivative and second-order derivative to construct temperature gradient matrix for abnormal detection of equipment thermal load distribution. The temperature gradient matrix is ​​defined as: Where: G is the temperature gradient matrix; Indicates time point t i The rate of temperature change at Indicates time point t i The temperature change acceleration at the location; n is the number of time points, which is used to describe the resolution of the time series; through cluster analysis of the matrix G, the abnormal areas of the equipment thermal load distribution are identified to locate the potential fault source.

5. The method for monitoring the operation of ore crushing equipment according to claim 1, characterized in that The multi-dimensional trend coupling early warning generation method comprises: Extract the short-term fluctuation and long-term trend characteristics of the signal from the time series; calculate the rate of change of the signal in different time periods through the sliding window method, separate the high-frequency disturbance and low-frequency trend of the signal; normalize the extracted signal trend parameters.

6. A method for monitoring the operation of ore crushing equipment according to claim 5, characterized in that The multi-dimensional trend coupling early warning generation method comprises: Through time series correlation analysis, the synergistic effects between trends are quantified to form a comprehensive representation of multi-dimensional trends; the trend coupling intensity is expressed by the following formula: in: C(t) is the trend coupling strength at time t, which is used to measure the comprehensive impact of the coordinated changes of multiple signals; L(τ) is the load trend function at time τ; V(τ) is the vibration characteristic trend function at time τ; G(τ) is the temperature gradient trend function at time τ; w1, w2, w3 are the weights of load, vibration and temperature signals, respectively, reflecting the importance of each signal in the comprehensive evaluation; Represent the time change rate of load, vibration and temperature signals respectively.

7. A method for monitoring the operation of ore crushing equipment according to claim 6, characterized in that The multi-dimensional trend coupling early warning generation method comprises: The operation status of the equipment is quantified and a comprehensive risk assessment index is constructed based on the trend coupling results. The comprehensive risk index integrates the trend characteristics of multi-dimensional signals and reflects the contribution of different signals to the equipment status through weight distribution. The calculation formula of the comprehensive risk assessment index is: Where: R is the comprehensive risk assessment index, which is used to quantify the current comprehensive health status of the equipment; T i is the dimensionless eigenvalue of the ith signal trend; W i is the weight of the ith signal, n is the number of signals involved in the evaluation; the numerator in the formula is the weighted sum of all signal eigenvalues ​​and their corresponding weights, and the denominator is the sum of all weights; the weight W i The design is based on historical equipment operation data.

8. A method for monitoring the operation of ore crushing equipment according to claim 7, characterized in that The multi-dimensional trend coupling early warning generation method comprises: The value of the comprehensive risk assessment index R changes dynamically. The threshold setting is adjusted according to the equipment type and operating conditions. The dynamic threshold determination mechanism is divided into the following two levels: Slight abnormality: When R exceeds the slight abnormality threshold R l When R exceeds the major fault threshold R h When the device is in serious risk, it is recommended to stop the machine for inspection immediately.

Images