Intelligent operation and maintenance method and system for vibrating screen based on multi-source fusion

Abstract

The application relates to the technical field of intelligent operation and maintenance of vibrating screens, and discloses a vibrating screen intelligent operation and maintenance method and system based on multi-source fusion. The method comprises the following steps: establishing an individualized balance baseline fingerprint in the no-load stage, calculating an asymmetric drift index in the full-load stage, triggering audio voiceprint double cross-validation to confirm a hidden co-evolution window, substituting the hidden co-evolution window with a fatigue equation to obtain a window closing time limit, and outputting a frequency fine-tuning instruction to the exciter to extend the hidden co-evolution window. The application improves the early prediction accuracy of vibrating screen spring failure and the time margin of planned maintenance.

Description

Technical Field

[0001] This application relates to the field of intelligent operation and maintenance technology for vibrating screens, and in particular to an intelligent operation and maintenance method and system for vibrating screens based on multi-source fusion. Background Technology

[0002] Vibrating screens are used in material grading production lines in industries such as building materials, mining, and coal. Their screens and supporting springs operate under conditions of strong vibration, heavy load, and high dust levels for extended periods. Screen breakage and spring failure are the two most common failure modes. Current technologies for monitoring the operating status of vibrating screens primarily employ a single vibration acceleration sensor to collect data at key locations on the screen frame. By extracting statistical characteristics such as time-domain kurtosis and root mean square (RMS) of the vibration signal, or the frequency-domain amplitude spectrum, and comparing them with preset fixed thresholds, alarms or shutdown commands are triggered. Some solutions involve manually applying static loads to the springs while the machine is stopped, using displacement sensors to measure spring compression deformation to diagnose spring failures. Other solutions install temperature sensors at the vibrator bearing housing to monitor whether the bearing temperature rise exceeds a set upper limit. These solutions have been applied to some extent in engineering practice and possess basic single-fault detection capabilities.

[0003] However, existing technologies have the following significant shortcomings. Existing solutions generally use single-type sensor signals for fault diagnosis. Vibration, audio, and temperature signals are independent of each other, lacking a multi-source fusion mechanism. Single-channel signals have a high false alarm rate in environments with strong vibration and background noise. Existing solutions rely on whether the absolute amplitude of a single measuring point exceeds a fixed hard alarm threshold, ignoring the impact of individual installation differences on the amplitude distribution of each measuring point within the same vibrating screen, and failing to detect the subcritical imbalance state of the force ratio between the spring support points of the screen frame. Existing solutions perform static spring diagnosis in a stopped state, failing to reflect the dynamic force changes of the springs in real time during normal operation. More importantly, existing solutions treat screen breakage and spring failure as two independent faults, diagnosing them separately, lacking an understanding of the physical causal relationship between the two. This results in the maintenance system only being able to passively respond to faults that have already occurred, unable to predict or proactively intervene in advance. Summary of the Invention

[0004] This application provides a method and system for intelligent operation and maintenance of vibrating screens based on multi-source fusion. It solves the problems in existing vibrating screen operation and maintenance technologies, such as the inability to perceive the cross-fault causal evolution relationship between screen damage and spring failure, and the fact that the operation and maintenance system can only passively alarm and cannot actively intervene to extend the time window for safe handling. It improves the accuracy of early prediction of spring failure of vibrating screens and the time margin for planned maintenance.

[0005] Firstly, this application provides a method for intelligent operation and maintenance of vibrating screens based on multi-source fusion, the method comprising: Step S1: During the no-load operation of the vibrating screen, the multi-directional vibration signals of each spring support point of the vibrating screen frame are normalized to obtain an individualized balance baseline fingerprint. Step S2: During the full-load operation of the vibrating screen, the real-time vibration signal of each spring support point of the screen frame is compared with the individualized balance baseline fingerprint to obtain the asymmetric drift index. Step S3: When the asymmetric drift index exceeds the primary judgment threshold, the audio signal of the vibrating screen is subjected to voiceprint feature extraction to obtain the voiceprint feature of the screen breakage. The voiceprint feature of the screen breakage is then subjected to double cross-validation with the asymmetric drift index to confirm the hidden co-evolution window. Step S4: Based on the dynamic load increment of the drift-side spring corresponding to the asymmetric drift index in the implicit co-evolution window, substitute it into the fatigue equation to calculate the remaining fatigue cycle number of the drift-side spring, obtain the window closing time limit, and output the excitation frequency fine-tuning command to the vibrating screen exciter according to the window closing time limit, so that the asymmetric drift index falls back to below the primary judgment threshold, and the implicit co-evolution window is extended.

[0006] Secondly, this application provides a multi-source fusion-based intelligent operation and maintenance system for vibrating screens, the multi-source fusion-based intelligent operation and maintenance system for vibrating screens comprising: The processing module is used to normalize the multi-directional vibration signals of each spring support point of the vibrating screen frame during the no-load operation phase of the vibrating screen to obtain an individualized balance baseline fingerprint. The comparison module is used to compare the real-time vibration signal of each spring support point of the screen frame with the individualized balance baseline fingerprint during the full-load operation of the vibrating screen to obtain the asymmetric drift index. The verification module is used to extract the acoustic features of the audio signal of the vibrating screen mesh when the asymmetric drift index exceeds the primary judgment threshold, obtain the acoustic features of the screen mesh damage, and perform double cross-validation between the acoustic features of the screen mesh damage and the asymmetric drift index to confirm the hidden co-evolution window. The calculation module is used to calculate the remaining fatigue cycle number of the drift-side spring based on the dynamic load increment of the drift-side spring corresponding to the asymmetric drift index in the implicit co-evolution window, and obtain the window closing time limit. Based on the window closing time limit, the module outputs a fine-tuning command of the excitation frequency to the vibrating screen exciter so that the asymmetric drift index falls back to below the primary judgment threshold, thereby extending the implicit co-evolution window.

[0007] Thirdly, a multi-source fusion-based intelligent operation and maintenance device for vibrating screens is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the multi-source fusion-based intelligent operation and maintenance device for vibrating screens to execute the aforementioned multi-source fusion-based intelligent operation and maintenance method for vibrating screens.

[0008] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, which, when executed on a computer, cause the computer to perform the above-described intelligent operation and maintenance method for vibrating screens based on multi-source fusion.

[0009] The technical solution provided in this application overcomes the fundamental deficiency of existing technologies that use a general fixed threshold to normalize the multi-directional vibration signals of each spring support point of the screen frame during the no-load operation phase and establish an individualized balance baseline fingerprint. This overcomes the limitation that existing technologies, which use a general fixed threshold, cannot reflect the individualized installation differences and spring preload distribution differences of a single vibrating screen. This allows for a quantitative benchmark for comparing real-time vibration signals during the subsequent full-load operation phase, tailored to the structural characteristics of the equipment itself. Furthermore, the technical feature of comparing real-time vibration signals with the individualized balance baseline fingerprint and calculating the asymmetric drift index upgrades the isolated single-point absolute amplitude judgment in existing technologies to a multi-point collaborative quantitative index reflecting the overall force distribution balance of the screen frame. This allows the subcritical imbalance state, where the amplitude of any single point does not exceed the hard alarm threshold, to be perceived and quantified, fundamentally filling the blind spot in the perception of latent faults in existing technologies. Furthermore, when the asymmetric drift index exceeds the primary judgment threshold, an audio signal voiceprint feature extraction mechanism is introduced, and the voiceprint features of screen breakage are cross-validated with the asymmetric drift index. This mechanism integrates two types of sensor signals with different physical mechanisms, the vibration domain and the acoustic domain, at the decision level, effectively suppressing misjudgments caused by false alarms of single signals. The sensitivity of the Mel frequency cepstral coefficients to the voiceprint of screen wire breakage in the characteristic frequency band of 3kHz to 8kHz is mutually corroborated with the structural evidence of amplitude asymmetric drift, together constituting a high-confidence confirmation basis for the implicit co-evolution window.

[0010] After confirming the implicit co-evolution window, the technical feature of calculating the remaining fatigue cycles and outputting the window closing time limit by substituting the dynamic load increment of the spring on the drift side corresponding to the asymmetric drift index into the fatigue equation, transforms the multi-source fusion sensing results into a life prediction quantity with clear physical meaning. The role of the Basquin fatigue equation here is not only a calculation tool, but also a core bridge to quantify the physical causal relationship between the dimensionless signal quantity of the asymmetric drift index and the fatigue mechanical parameters of the spring material, so that the operation and maintenance decision-making changes from the qualitative "there is a fault" to the quantitative "how much time is left". Crucially, the technology of outputting a fine-tuning command for the exciter based on the window closing time limit, so as to bring the asymmetric drift index back below the primary judgment threshold and extend the implicit co-evolution window, upgrades the operation and maintenance system from a one-way passive monitoring and alarm mode to an intelligent operation and maintenance mode with closed-loop active intervention capability. The fine-tuning of the exciter frequency restores the force balance of the screen frame by changing the dynamic load ratio coefficient of each spring group, thereby delaying the rate of spring fatigue accumulation and actively extending the safe handling time window without stopping the machine. This is a breakthrough in capability that no existing vibrating screen operation and maintenance technology has achieved. Attached Figure Description

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

[0012] Figure 1 This is a schematic diagram of an embodiment of the intelligent operation and maintenance method for vibrating screens based on multi-source fusion in this application. Figure 2 This is a schematic diagram comparing the short-time Fourier transform time-frequency matrix of the audio signal under normal operating conditions and screen breakage conditions in an embodiment of this application. Figure 3 This is a schematic diagram comparing the characteristic matrices of Mel frequency cepstral coefficients under normal operating conditions and screen breakage conditions in an embodiment of this application. Detailed Implementation

[0013] This application provides an intelligent operation and maintenance method and system for vibrating screens based on multi-source fusion. The terms "first," "second," "third," "fourth," etc. (if present)," in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0014] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the intelligent operation and maintenance method for vibrating screens based on multi-source fusion in this application includes: Step S1: During the no-load operation of the vibrating screen, the multi-directional vibration signals of each spring support point of the vibrating screen frame are normalized to obtain an individualized balance baseline fingerprint. Specifically, during the no-load operation phase of the vibrating screen, multi-directional vibration signals from each spring support point of the screen frame are collected and normalized to obtain an individualized balance baseline fingerprint. The individualized balance baseline fingerprint refers to the normalized proportional distribution vector of the vibration amplitude of each spring support point of the vibrating screen frame relative to the mean of all measuring points under no-load steady-state conditions. It reflects the inherent proportional relationship of the forces on each support point under the current installation state, spring preload distribution, and excitation parameters of the vibrating screen. This fingerprint is established during the no-load phase because the no-load state eliminates the interference of material load; the amplitude fluctuations at each measuring point are determined only by the spring stiffness distribution and excitation force, accurately reflecting the structural characteristics of the individual equipment. Normalization converts the absolute amplitude of each measuring point into a relative proportion, eliminating individual deviations caused by differences in the absolute amount of excitation force between different sets of vibrating screens, ensuring that the baseline fingerprint only carries structural information about the force distribution at each support point. The initial judgment threshold is set at 3 times the standard deviation of the natural amplitude fluctuation during the baseline establishment period. This multiple is determined based on the principle of statistical significance. 3 times the standard deviation corresponds to a confidence level of about 99.7% under normal distribution, ensuring that the drift judgment excludes the interference of normal operating condition fluctuations.

[0015] Step S2: During the full-load operation of the vibrating screen, the real-time vibration signal of each spring support point of the screen frame is compared with the individualized balance baseline fingerprint to obtain the asymmetric drift index. Specifically, during full-load operation, the real-time vibration signals of each spring support point of the screen frame are compared with the individualized balance baseline fingerprint to obtain the asymmetric drift index. The asymmetric drift index is a scalar quantity synthesized by subtracting the proportions of the real-time amplitude deviations of the eight measuring points from the baseline in the left-right and front-back directions. Its physical meaning is the degree and direction of the screen frame amplitude distribution deviating from the normal force balance state. When a small damage occurs in a local area of ​​the screen, the material throughput in that area increases and the local load decreases, resulting in a decrease in the dynamic force of the corresponding spring and an increase in amplitude. This non-uniformity is reflected in the asymmetric drift index as a larger value. Existing technology only uses whether the absolute amplitude of a single measuring point exceeds the hard alarm threshold as the judgment criterion, which cannot identify the subcritical state of imbalance in the force ratio between the support points. The asymmetric drift index, through grouped difference calculation, transforms this imbalance into a single monitorable indicator, so that abnormal changes in the overall force distribution of the screen frame can still be detected even when the amplitude of any single point does not exceed the hard alarm threshold.

[0016] Step S3: When the asymmetric drift index exceeds the primary judgment threshold, the audio signal of the vibrating screen is extracted to obtain the screen damage audio feature. The screen damage audio feature and the asymmetric drift index are then subjected to double cross-validation to confirm the hidden co-evolution window. Specifically, when the asymmetric drift index exceeds the initial judgment threshold, acoustic signature features are extracted from the audio signal of the vibrating screen mesh to obtain the acoustic signature features of the screen mesh damage. These acoustic signature features are then subjected to dual cross-validation with the asymmetric drift index to confirm the latent co-evolution window. The acoustic signature features of the screen mesh damage refer to the Mel-frequency cepstral coefficient sequence extracted from the screen mesh audio signal. Mel-frequency cepstral coefficients are feature vectors obtained by converting the audio signal to the time-frequency domain via short-time Fourier transform, calculating the filter bank energy on the Mel-frequency scale, and compressing it via discrete cosine transform. This effectively characterizes the specific acoustic patterns generated by the screen mesh under different damage states. Dual cross-validation requires both acoustic signature anomalies and the asymmetric drift index exceeding the threshold to simultaneously confirm the latent co-evolution window. A single abnormal signal does not trigger confirmation. This mechanism eliminates misjudgments caused by operating condition interference such as sudden changes in material particle size. The latent co-evolution window is a concept proposed in this invention to describe the time interval in which screen damage and spring failure mutually accelerate each other. Within this window, the device has not yet triggered any existing hard alarm thresholds, but screen damage and spring stress redistribution are developing synchronously.

[0017] Step S4: Based on the dynamic load increment of the drift side spring corresponding to the asymmetric drift index in the implicit co-evolution window, substitute it into the fatigue equation to calculate the remaining fatigue cycle number of the drift side spring, obtain the window closing time limit, and output the excitation frequency fine adjustment command to the vibrating screen exciter according to the window closing time limit, so that the asymmetric drift index falls back to below the primary judgment threshold, and the implicit co-evolution window is extended.

[0018] Specifically, based on the dynamic load increment of the drift-side spring corresponding to the asymmetric drift index within the implicit co-evolution window, the remaining fatigue cycle number of the drift-side spring is calculated using the fatigue equation, yielding the window closing time limit. Based on this window closing time limit, a fine-tuning command for the excitation frequency is output to the vibrating screen's exciter to bring the asymmetric drift index back below the primary judgment threshold, thus extending the implicit co-evolution window. The dynamic load increment of the drift-side spring is calculated by multiplying the drift direction component of the asymmetric drift index by the rated dynamic load and the lateral coupling coefficient. The lateral coupling coefficient is taken as 1.15, reflecting the load amplification effect caused by the stiffness transmission of the screen frame. This coefficient is determined by the equipment's finite element calibration experiment. The window closing time limit is the time quantity obtained by dividing the remaining fatigue cycle number of the drift-side spring by the excitation frequency, output in hours. Its physical meaning is the remaining time for the spring to reach its fatigue limit under the condition that the current drift state remains unchanged. The excitation frequency fine-tuning command selects the target frequency within ±2 Hz of the current excitation frequency so that the load ratio coefficient of the drift side spring is closest to the baseline rating. It gradually changes to the target frequency at a slope of 0.5 Hz per second. By changing the magnitude of the excitation force, it actively redistributes the dynamic force ratio of each spring, causing the asymmetric drift index to fall back, thereby extending the window closing time limit and buying time for planned maintenance.

[0019] In one specific embodiment, step S1 includes: A triaxial accelerometer is placed at each of the eight locations: the four corners of the vibrating screen frame, the front and rear ends of the longitudinal central axis, and the middle of the two side plates. The vibration signals of each sensor are synchronously acquired at a sampling frequency of 2000Hz to obtain eight vibration acquisition signals. The 8-channel vibration acquisition signals were divided into frames with 2048 sampling points per frame and a frame overlap rate of 50%. A fast Fourier transform was performed on each frame of data to extract the amplitude spectrum lines at the fundamental frequency, second harmonic frequency and third harmonic frequency of the vibrating screen, and the frequency domain amplitude sequence of each measurement point was obtained. The mean value of each frame amplitude in the frequency domain amplitude sequence of each measurement point is taken to obtain the baseline amplitude matrix; based on the ratio of the fundamental frequency amplitude of each measurement point to the mean value of the fundamental frequency amplitude of the 8 measurement points in the baseline amplitude matrix, the fundamental frequency amplitude of each measurement point is normalized to obtain the normalized baseline fingerprint vector. The normalized baseline fingerprint vector is combined and stored with the excitation frequency, ambient temperature and fundamental frequency phase vector of each measurement point during the no-load phase to obtain the individualized balanced baseline fingerprint.

[0020] Specifically, sensors P1 to P4 are placed at the four corners of the screen frame because these are the mounting points directly above the four sets of support springs of the vibrating screen, and the amplitude at these positions directly reflects the dynamic stress state of the corresponding springs. Sensors P5 and P6 are placed at the front and rear ends of the longitudinal central axis to capture the difference in amplitude between the front and rear of the screen frame along the material conveying direction. Sensors P7 and P8 are placed in the middle of the side plates to sense the lateral amplitude distribution in the width direction of the screen frame. The sampling frequency is set to 2000Hz based on the principle that the excitation frequency of the vibrating screen is usually in the range of 10Hz to 50Hz, and that the Nyquist theorem requires the sampling frequency to be no less than twice the highest analysis frequency. 2000Hz can cover at least 20th order harmonic analysis requirements, far exceeding the minimum sampling rate required for third harmonic analysis, ensuring the integrity of frequency domain feature extraction. A frame length of 2048 points corresponds to a 1.024-second time window at a sampling rate of 2000Hz. This window length includes approximately 35 complete vibration cycles at a 35Hz excitation frequency, sufficient to achieve an FFT frequency resolution of approximately 0.98Hz. This allows for clear differentiation of the fundamental, second, and third harmonic spectral lines without overlap. A 50% frame overlap rate, meaning adjacent frames share 1024 sampling points, reduces signal truncation at frame boundaries while maintaining temporal resolution. The fundamental, second, and third harmonic frequencies are extracted because the energy of the vibrating screen is mainly concentrated in these three harmonic components during normal operation. Changes in spring stiffness distribution preferentially alter the amplitude ratios of these three components. Other higher harmonics have lower energy and are more susceptible to noise interference, thus they are not included in the baseline feature extraction range.

[0021] The baseline amplitude matrix is ​​obtained by averaging the amplitude values ​​of each frame in the frequency domain amplitude sequence of each measuring point. This matrix is ​​a two-dimensional array with 8 rows and 3 columns. The rows correspond to the 8 measuring points P1 to P8, and the columns correspond to the three frequency components: fundamental frequency, second harmonic, and third harmonic. The matrix elements are the steady-state mean amplitude of the measuring point at the corresponding frequency component. Normalization is performed by taking the ratio of the fundamental frequency amplitude of each measuring point to the mean of the fundamental frequency amplitudes of the 8 measuring points. This converts the absolute amplitude into a relative proportion, eliminating individual deviations caused by differences in the absolute amount of excitation force due to differences in motor power and eccentric block mass among different sets of vibrating screens. This ensures that the normalized baseline fingerprint vector only reflects the relative distribution of forces on each support point of the vibrating screen. The excitation frequency and ambient temperature during the no-load phase are stored together with the normalized baseline fingerprint vector as auxiliary parameters because the spring stiffness is affected by the ambient temperature. Temperature changes will cause changes in the elastic modulus of the spring material, thereby changing the amplitude ratio of each measuring point. Storing the ambient temperature at the time of baseline establishment can correct for temperature drift during full-load comparison. The fundamental frequency phase vector of each measuring point records the phase distribution of the vibration signal of each measuring point under the baseline state. It is used to distinguish between amplitude asymmetry caused by uneven spring preload and phase anomaly caused by resonance of the screen frame structure. The two may show similar characteristics in amplitude but different phase distribution patterns. Storing the phase vector can eliminate resonance interference in the judgment.

[0022] In one specific embodiment, step S2 includes: The vibration signals of eight channels at each spring support point of the screen frame were continuously and synchronously acquired at a sampling frequency of 2000Hz. With an update cycle of 30 seconds, the eight vibration signals in each update cycle were divided into frames with 2048 sampling points per frame and a frame overlap rate of 50%. The data of each frame was subjected to fast Fourier transform to extract the amplitude spectrum at the excitation fundamental frequency and obtain the real-time fundamental frequency amplitude matrix of each measuring point. The fundamental frequency baseline amplitude of each measurement point is extracted from the individualized balanced baseline fingerprint. Based on the difference between the fundamental frequency amplitude of each measurement point and the fundamental frequency baseline amplitude of each measurement point in the real-time fundamental frequency amplitude matrix of each measurement point, the normalized deviation of the fundamental frequency amplitude of each measurement point is calculated to obtain the instantaneous drift of each measurement point. The instantaneous drift of each measuring point is averaged according to the measuring point group on the left side and the measuring point group on the right side of the sieve frame, respectively, to obtain the average drift on the left side and the average drift on the right side. The difference between the average drift on the left side and the average drift on the right side is obtained to obtain the left-right asymmetric drift component. The average drift of the front side and the average drift on the rear side of the sieve frame are averaged according to the measuring point group on the front side and the measuring point group on the rear side, respectively, to obtain the average drift on the front side and the average drift on the rear side. The difference between the average drift on the front side and the average drift on the rear side is obtained to obtain the front-rear asymmetric drift component. The asymmetric drift index is obtained by performing a square root operation on the sum of squares of the left-right asymmetric drift components and the front-back asymmetric drift components.

[0023] Specifically, during the full-load operation phase, eight vibration signals are continuously acquired at a sampling frequency of 2000Hz, maintaining the same sampling parameters as in step S1. This ensures consistency between the real-time signal and the baseline signal in the time domain sampling conditions, allowing for direct amplitude comparison. The update cycle is set to 30 seconds, a balance between real-time monitoring and computational resource consumption. Within 30 seconds, 60,000 sampling points can be obtained at a sampling rate of 2000Hz. With a frame length of 2048 points and a 50% overlap rate, approximately 58 frames of valid data can be generated. The stability of the statistical mean at this frame count meets the engineering accuracy requirements. Shorter update cycles would lead to unstable frequency domain estimation due to insufficient frame counts, while longer cycles would delay the timely detection of subcritical drift states. During the full-load phase, only the amplitude spectrum at the fundamental frequency of the excitation is extracted, omitting the second and third harmonics. This is because the calculation of the asymmetric drift index depends on the relative change in the fundamental frequency amplitude between each measuring point. The fundamental frequency component carries the main information about the change in the spring force distribution. The second and third harmonics are more affected by material impact excitation under full-load conditions, and their introduction would actually reduce the stability of the drift calculation. The fundamental frequency baseline amplitude of each measuring point is extracted from the individualized equilibrium baseline fingerprint by reading the element values ​​of the corresponding fundamental frequency column in each row of the baseline amplitude matrix, i.e., the normalized baseline amplitude of the eight measuring points P1 to P8 on the fundamental frequency component, which serves as the reference benchmark for calculating the instantaneous drift of each measuring point. The specific method for calculating the normalized deviation is to divide the difference between the real-time fundamental frequency amplitude of each measuring point and the corresponding fundamental frequency baseline amplitude of the measuring point by the fundamental frequency baseline amplitude, obtaining the relative deviation expressed as a percentage. This relative deviation is the instantaneous drift of each measuring point, and its physical meaning is the degree of deviation of the current amplitude of the measuring point from the no-load baseline state.

[0024] The left measuring point group consists of P1, P3, P5, and P7, corresponding to the measuring points directly above the two springs on the left side of the screen frame and the left side plate. The right measuring point group consists of P2, P4, P6, and P8, corresponding to the measuring points directly above the two springs on the right side of the screen frame and the right side plate. The front measuring point group consists of P1, P2, P5, and P6, corresponding to the measuring points above the spring at the feed end of the screen frame and the front end of the central axis. The rear measuring point group consists of P3, P4, P7, and P8, corresponding to the measuring points above the spring at the discharge end of the screen frame and the rear end of the central axis. The purpose of grouping and averaging is to compress the instantaneous drift of the eight discrete measuring points into two scalar quantities representing the degree of force imbalance in the left-right and front-back directions of the screen frame as a whole, eliminating occasional fluctuations caused by uneven local material distribution at a single measuring point. The left-right asymmetric drift component is calculated by subtracting the right-right drift average from the left-side drift average. A positive value indicates that the load on the left spring increases while that on the right decreases, and a negative value indicates the opposite. The front-back asymmetric drift component is calculated by subtracting the rear-back drift average from the front-side drift average. A positive value indicates that the load on the feed end spring increases. Performing a square root operation on the sum of the squares of the left-right and front-back asymmetric drift components combines the imbalance components in the two directions into a non-directional scalar. This operation is equivalent to calculating the modulus of a two-dimensional vector. The resulting asymmetric drift index comprehensively reflects the degree of force imbalance of the screen frame in any direction. This allows the same threshold judgment logic to uniformly handle asymmetric imbalances caused by screen breakage at any location, without the need to set separate thresholds for the left-right and front-back directions.

[0025] In one specific embodiment, step S3 includes: Based on the directions of the left-right asymmetric drift components and the front-back asymmetric drift components in the asymmetric drift index, the position of the side screen frame with the largest drift is determined. The audio signal on the outer surface of the side plate of the side screen frame with the largest drift is continuously collected at a sampling frequency of 44100Hz, with each collection lasting 30 seconds, and the collection is repeated 5 times. The audio signal collected each time is subjected to short-time Fourier transform with parameters of Hamming window length of 1024 points and frame shift of 256 points to obtain 5 sets of time-frequency matrices. The 13th-order Mel frequency cepstral coefficients were extracted from the five time-frequency matrices in the frequency band from 3000Hz to 8000Hz to obtain the characteristic sequence of the screen damage acoustic signature. A normal voiceprint baseline model is constructed based on the mean vector and covariance matrix of the Mel frequency cepstral coefficients of the audio signal during the no-load phase. The Mahalanobis distance is calculated between the Mel frequency cepstral coefficient vectors corresponding to each acquisition in the screen breakage voiceprint feature sequence and the normal voiceprint baseline model. The Mahalanobis distance calculation results are compared one by one with the 95% confidence interval threshold of the 13-DOF chi-square distribution to obtain the voiceprint Mahalanobis distance decision result. The number of times the Mahalanobis distance in the voiceprint decision result exceeds the threshold and the duration of the asymmetric drift index continuously exceeding the primary decision threshold are used for double cross-validation. When the number of times the threshold is exceeded is not less than 3 and the duration of the asymmetric drift index continuously exceeding the primary decision threshold is not less than 5 minutes, the recessive co-evolution window is confirmed.

[0026] Specifically, the location of the side screen frame with the largest drift is determined by taking the direction corresponding to the larger absolute value of the left-right asymmetrical drift component and the front-back asymmetrical drift component. If the absolute value of the left-right asymmetrical drift component is larger and positive, the left side of the screen frame is selected; if it is negative, the right side is selected. The same applies to the front-back direction. This side is where the screen frame amplitude increases most significantly relative to the baseline, and it is also the area with the greatest screen tension loss and the strongest sound signature signal of the breakage. The sampling frequency is set to 44100Hz, which is the standard sampling rate in the field of audio acquisition. According to the Nyquist theorem, it can cover the audio frequency range up to 22050Hz, far exceeding the 3000Hz to 8000Hz frequency band where the sound signature of the screen breakage feature is located, ensuring that the sound signature feature in the target frequency band is not distorted due to insufficient sampling rate. The design of five consecutive 30-second sampling sessions strikes a balance between the duration of each sampling session and the number of sampling sessions. Within 30 seconds, at a sampling rate of 44100Hz, 1,323,000 sampling points can be obtained. With a Hamming window of 1024 points and a frame shift of 256 points, approximately 5167 frames of time-frequency data can be generated, which is sufficient to stably extract Mel frequency cepstral coefficient features. The five consecutive sampling sessions, with at least three exceeding the threshold as the judgment condition, introduce a majority voting mechanism to eliminate interference from occasional impact noise in a single sampling session. The Hamming window of 1024 points corresponds to a time window of approximately 23 milliseconds at a sampling rate of 44100Hz. This window length matches the duration of the transient impact sound generated by the screen breakage. A window that is too short will truncate the acoustic features, while a window that is too long will alias multiple breakage events. The frame shift of 256 points means that the interval between adjacent frames is approximately 5.8 milliseconds, with a frame overlap rate of 75%. The high overlap rate ensures that the transient signal of the breakage will not fall into the frame boundary and be weakened. The extraction of Mel frequency cepstral coefficients in the 3000Hz to 8000Hz frequency band is based on the physical characteristic that the main energy of the metal impact sound generated by the breakage of the screen wire and the tearing of the mesh is concentrated in this frequency band. The frequency band below 3000Hz is dominated by the vibration noise of the vibrator and the material sliding noise, and the frequency band above 8000Hz has too low a signal-to-noise ratio and is not included in the scope of voiceprint feature extraction.

[0027] The calculation process of Mel frequency cepstral coefficients involves compressing the frequency resolution of each frame's time-frequency matrix within the target frequency band using a Mel filter bank. The Mel filter bank transforms the linear frequency axis into a Mel frequency axis that simulates the perceptual characteristics of the human ear, achieving high resolution in the low-frequency band and low resolution in the high-frequency band, allowing feature extraction to focus more on the low-frequency contour features of the damaged voiceprint. After passing through the Mel filter bank, the logarithm of the energy of each sub-band is taken, followed by a discrete cosine transform, resulting in a Mel frequency cepstral coefficient vector composed of 13 coefficients. The 13th order is an empirical value that strikes a balance between feature dimension and computational complexity. Lower-order coefficients carry the main information of the voiceprint contour, while coefficients added above the 13th order carry progressively less information. The normal voiceprint baseline model is constructed by extracting the Mel frequency cepstral coefficients from the audio signal collected from the outer surface of the same side of the screen frame during the unloaded phase, and then calculating the mean vector and covariance matrix. The mean vector reflects the central position of the voiceprint features under normal operating conditions, and the covariance matrix reflects the correlation structure between coefficients of different orders. Mahalanobis distance is calculated by left-multiplying the difference between the current Mel-frequency cepstral coefficient vector and the mean vector by the inverse of the covariance matrix, then right-multiplying by the transpose of the difference vector and taking the square root. The result comprehensively considers the variance of coefficients at each order and the correlation between coefficients, making it more effective than Euclidean distance in distinguishing voiceprint differences in multidimensional feature space. The 95% confidence interval threshold for the 13-DOF chi-square distribution is set at 22.36, corresponding to 95% of sample points falling within this range under a 13-dimensional normal distribution. Exceeding this threshold indicates that the current voiceprint deviates from the normal voiceprint baseline model with a 95% confidence level. The criterion of exceeding the threshold at least 3 times is based on the principle of majority judgment in 5 collections. A single exceedance of the threshold is considered an occasional interference, while 3 or more exceedances indicate that the abnormal voiceprint is continuous in time. The asymmetric drift index continuously exceeds the primary judgment threshold for at least 5 minutes, corresponding to approximately 10 consecutive exceedance states of 30-second update cycles. This excludes the instantaneous amplitude shift caused by brief material impacts. Both conditions must be met simultaneously to confirm the latent co-evolution window and ensure the reliability of the judgment conclusion.

[0028] Figure 2 This is a schematic diagram comparing the short-time Fourier transform time-frequency matrix of the audio signal under normal operating conditions and screen breakage conditions in an embodiment of this application. Figure 2 The left subplot shows the time-frequency matrix obtained by short-time Fourier transform of the audio signal on the outer surface of the screen frame side plate under normal operating conditions of the vibrating screen. The frequency range covers 500Hz to 10000Hz, and the time domain range is a typical segment in a single 30-second acquisition. The energy distribution is uniform with no significant concentrated areas. The right subplot shows the time-frequency matrix corresponding to the audio signal acquired at the same location when the screen is partially damaged. A significant transient high-energy accumulation response appears in the frequency band of 3kHz to 8kHz (the range marked by the white dashed line in the figure). This energy accumulation area corresponds to the characteristic sound pattern of metal impact generated at the moment the screen wire breaks. The significant difference in energy distribution between the two subplots is the physical basis for subsequent Mel frequency cepstral coefficient extraction and Mahalanobis distance determination.

[0029] Figure 3 This is a schematic diagram comparing the characteristic matrices of Mel frequency cepstral coefficients under normal operating conditions and screen breakage conditions in an embodiment of this application. Figure 3 The left subplot shows the 13th-order Mel frequency cepstral coefficient feature matrix extracted from the audio signal time-frequency matrix in the 3kHz to 8kHz frequency band under normal operating conditions of the vibrating screen. The matrix rows correspond to the coefficients from the 1st to the 13th order, and the columns correspond to the time frame numbers. The coefficient values ​​are evenly distributed and there is no obvious time-domain concentration feature. The right subplot shows the Mel frequency cepstral coefficient feature matrix corresponding to the screen damage condition. In the interval from the 30th to the 48th frame (the range marked by the white dashed line in the figure), there is a significantly abnormally large cluster of higher-order coefficients. This abnormal cluster reflects the specific distribution of the energy of the damaged acoustic signature excitation segment on the Mel frequency scale. By calculating the Mahalanobis distance between the 13-dimensional Mel frequency cepstral coefficient vector corresponding to each frame and the normal acoustic signature baseline model, the degree of deviation of the acoustic signature from the normal distribution in that frame can be quantified and used to generate the subsequent acoustic signature Mahalanobis distance decision results.

[0030] In one specific embodiment, step S4, based on the drift-side spring dynamic load increment corresponding to the asymmetric drift index within the implicit co-evolution window, includes: Based on the sign direction of the left-right asymmetric drift component and the front-back asymmetric drift component in the asymmetric drift index, determine the drift-side spring group, and obtain the rated dynamic load of the drift-side spring group from the equipment nameplate parameters. Based on the left and right asymmetric drift components and the rated dynamic load, the product of half of the rated dynamic load and the lateral coupling coefficient of the screen frame is calculated to obtain the dynamic load increment of the spring on the drift side. The actual dynamic load of the drift-side spring is obtained by summing the dynamic load increment of the drift-side spring with the rated dynamic load.

[0031] Specifically, the rule for determining the drift side spring group is to take the direction corresponding to the larger absolute value of the left-right asymmetrical drift component and the front-back asymmetrical drift component. When the absolute value of the left-right asymmetrical drift component is greater than that of the front-back asymmetrical drift component, the drift side is determined by the sign of the left-right component, with positive values ​​corresponding to the left spring group and negative values ​​corresponding to the right spring group. When the absolute value of the front-back asymmetrical drift component is greater, the drift side is determined by the sign of the front and back components, with positive values ​​corresponding to the front spring group and negative values ​​corresponding to the rear spring group. The rated dynamic load refers to the design dynamic load value that a single spring group can withstand when running at full load under the design conditions of the vibrating screen. This parameter is marked on the equipment nameplate or in the factory technical documents, and is expressed in Newtons. It is the reference load value for calculating the fatigue life of the spring. The 1 / 2 coefficient is introduced in the calculation of the dynamic load increment of the drift-side spring because the asymmetric drift component reflects the relative difference between the drift side and the opposite side. This difference is generated by the superposition of the deviations of both sides from the baseline. The actual increment borne by the drift-side spring is approximately half of the total difference. Taking the 1 / 2 coefficient converts the overall difference into a single-side increment. The lateral coupling coefficient of the screen frame is taken as 1.15, reflecting the load amplification effect caused by the overall stiffness transmission of the screen frame. When the load on a certain side of the spring increases, the rigid connection of the screen frame beam will transfer some of the additional load to the adjacent spring and form a feedback amplification. The amplification coefficient of 1.15 was determined by finite element calibration experiments, that is, the actual load increment is 15% higher than the theoretical calculation value. If this coefficient is not introduced, the actual stress level of the drift-side spring will be underestimated.

[0032] The complete calculation process for the dynamic load increment of the drift-side spring involves multiplying the rated dynamic load by half of the left and right asymmetric drift components, then multiplying by the lateral coupling coefficient of the screen frame (1.15). The result of multiplying these three factors is the additional dynamic load that the spring on this side bears under the current asymmetric drift state compared to the rated operating condition, expressed in Newtons. The actual dynamic load of the drift-side spring is obtained by adding the dynamic load increment of the drift-side spring to the rated dynamic load. Its physical meaning is the total dynamic load actually borne by the drift-side spring under the current implicit co-evolution window state. This value is higher than the rated dynamic load under normal operating conditions, and the excess is caused by the force imbalance of the screen frame due to local damage to the screen mesh. The rated dynamic load appears twice in the calculation as a reference value: once as a proportional reference for the increment calculation, and once as the addend in the summation. The physical meanings of the two uses are different. The former is used to convert the dimensionless drift component into a dimensional load increment, and the latter is used to restore the increment to the actual total load. The actual dynamic load of the drift-side spring obtained after the combined calculation is the direct input for subsequent calculation of the remaining fatigue cycle number in the fatigue equation.

[0033] In one specific embodiment, step S4 involves substituting the fatigue equation to calculate the remaining fatigue cycle number of the drift-side spring, thereby obtaining the window closing time limit, including: Based on the actual dynamic load of the drift side spring, the stress amplitude and average stress of the drift side spring are converted to obtain the stress amplitude and average stress of the drift side spring; the stress amplitude and average stress of the drift side spring are substituted into the Goodman correction criterion to obtain the equivalent fatigue limit stress amplitude of the drift side spring under the current working condition. Substituting the equivalent fatigue limit stress amplitude into the Basquin fatigue equation, the total number of fatigue cycles is obtained; based on the product of the excitation frequency of the vibrating screen and the number of operating hours, the number of fatigue cycles consumed by the drift side spring is calculated, and the number of fatigue cycles consumed is obtained. The remaining fatigue cycle count is obtained by subtracting the total fatigue cycle count from the number of fatigue cycles already consumed; the window closing time limit is obtained by converting the ratio of the remaining fatigue cycle count to the excitation frequency.

[0034] Specifically, the method of converting the stress amplitude and mean stress from the actual dynamic load of the drift-side spring is based on the mechanical formula of a helical compression spring, which converts the load into shear stress on the cross-section of the spring wire. The vibrating screen support spring bears a periodic compressive load during operation, with the load cyclically changing between its maximum and minimum values. The stress amplitude is defined as half the difference between the maximum and minimum shear stress, and the mean stress is defined as half the sum of the maximum and minimum shear stresses. Both are calculated by substituting the spring's mean diameter, spring wire diameter, and curvature correction factor into the spring shear stress formula. The Goodman correction criterion is a linear equation where the sum of the stress amplitude divided by the material's symmetrical cyclic fatigue limit and the mean stress divided by the material's tensile strength equals 1. By substituting the drift-side spring stress amplitude and mean stress, the equivalent fatigue limit stress amplitude that the material can withstand at the current mean stress level is determined. The symmetrical cyclic fatigue limit of the spring material 60Si2Mn is taken as 620 MPa, and the tensile strength as 1274 MPa. These two material parameters are determined from the spring steel material handbook. The physical meaning of the equivalent fatigue limit stress amplitude is the critical stress amplitude at which a material will fail due to fatigue under the current mean stress conditions. This value decreases as the mean stress increases, reflecting the adverse effect of mean stress on fatigue life.

[0035] The Basquin fatigue equation substitutes the equivalent fatigue limit stress amplitude into a power law relationship to calculate the total number of fatigue cycles. This equation describes the power law relationship between the number of cycles a material can withstand to fail under a given stress amplitude and the stress amplitude. The material constant C in the equation is taken as 1180 MPa, and the power exponent m is taken as 8.6. These two parameters are determined by fitting the standard SN curve of 60Si2Mn spring steel. The physical meaning of the total number of fatigue cycles is the total number of cycles the drift-side spring can withstand from its initial state to fatigue failure under the current actual dynamic load level. The number of fatigue cycles consumed is obtained by converting the excitation frequency to the product of the operating hours. The unit of the excitation frequency is Hertz (cycles per second). The operating hours, converted to seconds, are multiplied by the excitation frequency to obtain the total number of fatigue cycles the spring has undergone since it was put into use. This calculation implicitly assumes that the spring is always subjected to the current load level throughout its operation. The remaining fatigue cycle count is the difference between the total fatigue cycle count and the fatigue cycle count already consumed. Dividing this by the excitation frequency converts the cycle count to seconds, and then dividing by 3600 converts it to hours to obtain the window closing time limit. Its physical meaning is the remaining running time of the drift side spring from the fatigue limit under the condition that the current asymmetric drift state remains unchanged. It is output in hours for maintenance personnel to refer to when formulating maintenance plans.

[0036] In one specific embodiment, step S4, based on the window closing time limit, outputs a fine-tuning command for the exciter of the vibrating screen to adjust the excitatory frequency, so that the asymmetric drift index falls back below the primary judgment threshold, extending the latent co-evolution window, including: Based on the stiffness coefficient of the drift side spring group and the natural frequency of the corresponding subsystem of the drift side spring group, the dynamic load proportion coefficient of each side spring group under the current excitation frequency is calculated to obtain the load proportion coefficient of each side spring group. Within ±2Hz of the current excitation frequency, the difference between the load ratio coefficient of the drift side spring group and the corresponding rated load ratio coefficient in the individualized balance baseline fingerprint is calculated one by one at each candidate excitation frequency. The difference between the candidate excitation frequency and the current excitation frequency corresponding to the smallest difference is determined as the optimal fine-tuning amount of the excitation frequency. Based on the optimal fine-tuning amount of the excitation frequency, the excitation frequency fine-tuning command is output to the vibrating screen exciter to gradually change the excitation frequency to the target excitation frequency with a slope of 0.5Hz / s. After the excitation frequency gradually changes, the real-time vibration signal of each spring support point of the sieve frame is re-compared with the individualized balance baseline fingerprint to obtain the updated asymmetric drift index. When the updated asymmetric drift index falls below the primary judgment threshold, the window closing time limit is recalculated to obtain the latent co-evolution window.

[0037] Specifically, the stiffness coefficient of the drift-side spring assembly refers to the equivalent stiffness of the two supporting springs connected in parallel on the drift side, measured in Newtons per millimeter. It is calculated using the spring stiffness formula based on the spring material shear modulus, spring wire diameter, spring mean diameter, and effective number of coils. This parameter is provided in the equipment's technical documentation. The natural frequency of the drift-side spring assembly subsystem refers to the natural vibration frequency of the spring-mass subsystem formed by the drift-side spring assembly and the mass of the screen frame it supports. It is calculated by taking the square root of the ratio of the equivalent stiffness of the spring assembly to the corresponding screen frame mass and dividing by 2π. This natural frequency determines the sensitivity of the excitation frequency to the dynamic load on this side of the springs. The calculation of the dynamic load proportionality coefficient is based on the steady-state response theory of spring-mass systems. When the excitation frequency is much lower than the natural frequency, the system is in the quasi-static response region, and the dynamic load borne by each spring is proportional to its stiffness. When the excitation frequency is close to the natural frequency, the dynamic amplification effect is significant, and the load proportionality coefficient is sensitive to changes in the excitation frequency. By substituting the stiffness coefficients of each spring group, the natural frequency, and the current excitation frequency into the steady-state response equation, the proportion of the load borne by each spring group at the current excitation frequency to the total load can be calculated. The candidate excitation frequency traversal range is limited to within ±2Hz of the current excitation frequency. This is determined based on the tolerance of the vibrating screen's screening efficiency to the excitation frequency deviation. When the excitation frequency deviates from the rated value by more than 2Hz, the particle size distribution of the screened particles changes significantly, affecting product quality. The change in excitation force within ±2Hz does not exceed 0.66% of the rated value, and its impact on screening efficiency is negligible. The rated load ratio coefficient in the individualized balance baseline fingerprint is derived from the normalized ratio of the amplitude of each measuring point when establishing the baseline fingerprint in the no-load stage of step S1. This ratio directly reflects the relative proportion of the load borne by each side spring under normal force distribution under no-load steady-state conditions, and serves as the target reference value for fine-tuning the excitation frequency.

[0038] The physical basis for using the minimum difference as the criterion for selecting the optimal fine-tuning amount is to make the actual load proportional coefficient of the drift-side spring group as close as possible to the rated load proportional coefficient under the baseline state, thereby reducing the excessive stress on the drift-side spring and lowering the fatigue accumulation rate. During the traversal process, the absolute value of the difference between the load proportional coefficient of the drift-side spring group and the rated load proportional coefficient corresponding to each candidate excitation frequency is calculated in a step of 0.1Hz within a range of ±2Hz. The difference between the candidate excitation frequency with the smallest absolute value and the current excitation frequency is taken as the optimal fine-tuning amount for the excitation frequency. The excitation frequency gradient slope is set to 0.5Hz / s to avoid transient impact vibration of the screen frame caused by abrupt changes in the excitation frequency. The gradient rate of 0.5Hz / s allows the excitation frequency to smoothly transition from the current value to the target excitation frequency, and the continuous change in the screen frame vibration state during the gradient process does not produce abrupt load changes. The target excitation frequency is the sum of the current excitation frequency and the optimal fine-tuning amount for the excitation frequency, and is the endpoint value of the excitation frequency gradient. After the excitation frequency is gradually changed, the 8 vibration signals are reacquired and the updated asymmetric drift index is calculated using the same method as in step S2. If the index falls back below the primary judgment threshold, it indicates that the fine adjustment of the excitation frequency has made the spring loads on each side tend to be balanced again. At this time, the updated asymmetric drift index is substituted into the previous calculation in step S4 to obtain the actual dynamic load of the drift side spring. The remaining fatigue cycle number is then calculated by substituting the Goodman correction criterion and the Basquin fatigue equation, and the extended window closing time limit is obtained. This time limit is extended compared to before the fine adjustment, which buys time for planned maintenance.

[0039] The above describes the intelligent operation and maintenance method for vibrating screens based on multi-source fusion in the embodiments of this application. The following describes the intelligent operation and maintenance system for vibrating screens based on multi-source fusion in the embodiments of this application. One embodiment of the intelligent operation and maintenance system for vibrating screens based on multi-source fusion in the embodiments of this application includes: The processing module is used to normalize the multi-directional vibration signals of each spring support point of the vibrating screen frame during the no-load operation phase of the vibrating screen to obtain an individualized balance baseline fingerprint. The comparison module is used to compare the real-time vibration signal of each spring support point of the screen frame with the individualized balance baseline fingerprint during the full-load operation of the vibrating screen to obtain the asymmetric drift index. The verification module is used to extract the acoustic features of the audio signal of the vibrating screen mesh when the asymmetric drift index exceeds the primary judgment threshold, obtain the acoustic features of the screen mesh damage, and perform double cross-validation between the acoustic features of the screen mesh damage and the asymmetric drift index to confirm the hidden co-evolution window. The calculation module is used to calculate the remaining fatigue cycle number of the drift-side spring based on the dynamic load increment of the drift-side spring corresponding to the asymmetric drift index in the implicit co-evolution window, and obtain the window closing time limit. Based on the window closing time limit, the module outputs a fine-tuning command of the excitation frequency to the vibrating screen exciter so that the asymmetric drift index falls back to below the primary judgment threshold, thereby extending the implicit co-evolution window.

[0040] This invention also provides a multi-source fusion-based intelligent operation and maintenance device for vibrating screens, which can be a server. This multi-source fusion-based intelligent operation and maintenance device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor, designed as a computer, provides computing and control capabilities. The memory of the multi-source fusion-based intelligent operation and maintenance device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the multi-source fusion-based intelligent operation and maintenance device stores the data corresponding to this embodiment. The network interface of the multi-source fusion-based intelligent operation and maintenance device is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0041] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the intelligent operation and maintenance method for vibrating screens based on multi-source fusion.

[0042] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0043] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a multi-source fusion-based intelligent operation and maintenance device for vibrating screens (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0044] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent operation and maintenance of vibrating screens based on multi-source fusion, characterized in that, The method includes: Step S1: During the no-load operation of the vibrating screen, the multi-directional vibration signals of each spring support point of the vibrating screen frame are normalized to obtain an individualized balance baseline fingerprint. Step S2: During the full-load operation of the vibrating screen, the real-time vibration signal of each spring support point of the screen frame is compared with the individualized balance baseline fingerprint to obtain the asymmetric drift index. Step S3: When the asymmetric drift index exceeds the primary judgment threshold, the audio signal of the vibrating screen is subjected to voiceprint feature extraction to obtain the voiceprint feature of the screen breakage. The voiceprint feature of the screen breakage is then subjected to double cross-validation with the asymmetric drift index to confirm the hidden co-evolution window. Step S4: Based on the dynamic load increment of the drift-side spring corresponding to the asymmetric drift index in the implicit co-evolution window, substitute it into the fatigue equation to calculate the remaining fatigue cycle number of the drift-side spring, obtain the window closing time limit, and output the excitation frequency fine-tuning command to the vibrating screen exciter according to the window closing time limit, so that the asymmetric drift index falls back to below the primary judgment threshold, and the implicit co-evolution window is extended.

2. The intelligent operation and maintenance method for vibrating screens based on multi-source fusion according to claim 1, characterized in that, Step S1 includes: A triaxial accelerometer is placed at each of the eight locations: the four corners of the vibrating screen frame, the front and rear ends of the longitudinal central axis, and the middle of the two side plates. The vibration signals of each sensor are synchronously acquired at a sampling frequency of 2000Hz to obtain eight vibration acquisition signals. The 8 vibration acquisition signals are divided into frames with 2048 sampling points per frame, and the frame overlap rate is 50%. A fast Fourier transform is performed on each frame of data to extract the amplitude spectrum lines at the fundamental frequency, second harmonic frequency and third harmonic frequency of the vibrating screen, so as to obtain the frequency domain amplitude sequence of each measurement point. The mean value of each frame amplitude in the frequency domain amplitude sequence of each measurement point is taken to obtain the baseline amplitude matrix; based on the ratio of the fundamental frequency amplitude of each measurement point to the mean value of the fundamental frequency amplitude of the eight measurement points in the baseline amplitude matrix, the fundamental frequency amplitude of each measurement point is normalized to obtain the normalized baseline fingerprint vector. The normalized baseline fingerprint vector is combined and stored with the excitation frequency, ambient temperature and fundamental frequency phase vector of each measurement point during the no-load phase to obtain the individualized balanced baseline fingerprint.

3. The intelligent operation and maintenance method for vibrating screens based on multi-source fusion according to claim 1, characterized in that, Step S2 includes: The vibration signals of eight channels at each spring support point of the screen frame are continuously and synchronously acquired at a sampling frequency of 2000Hz. With an update cycle of 30 seconds, the eight vibration signals in each update cycle are divided into frames with 2048 sampling points per frame and a frame overlap rate of 50%. The data of each frame is subjected to fast Fourier transform to extract the amplitude spectrum at the excitation fundamental frequency and obtain the real-time fundamental frequency amplitude matrix of each measuring point. The fundamental frequency baseline amplitude of each measurement point is extracted from the individualized balanced baseline fingerprint. Based on the difference between the fundamental frequency amplitude of each measurement point and the fundamental frequency baseline amplitude of each measurement point in the real-time fundamental frequency amplitude matrix of each measurement point, the normalized deviation of the fundamental frequency amplitude of each measurement point is calculated to obtain the instantaneous drift of each measurement point. The instantaneous drift of each measuring point is averaged according to the left and right measuring point groups of the sieve frame to obtain the left drift average and the right drift average. The difference between the left drift average and the right drift average is obtained to obtain the left-right asymmetric drift component. The average of the front and rear measuring point groups of the sieve frame is averaged to obtain the front drift average and the rear drift average. The difference between the front drift average and the rear drift average is obtained to obtain the front-rear asymmetric drift component. The asymmetric drift index is obtained by performing a square root operation on the sum of squares of the left-right asymmetric drift components and the front-back asymmetric drift components.

4. The intelligent operation and maintenance method for vibrating screens based on multi-source fusion according to claim 1, characterized in that, Step S3 includes: Based on the directions of the left-right asymmetric drift components and the front-back asymmetric drift components in the asymmetric drift index, the position of the side screen frame with the largest drift is determined. The audio signal on the outer surface of the side plate of the side screen frame with the largest drift is continuously collected at a sampling frequency of 44100Hz, with each collection lasting 30 seconds, and the collection is repeated 5 times. The audio signal collected each time is subjected to short-time Fourier transform with parameters of Hamming window length of 1024 points and frame shift of 256 points to obtain 5 sets of time-frequency matrices. The 13th-order Mel frequency cepstral coefficients were extracted from the five time-frequency matrices in the frequency band from 3000Hz to 8000Hz to obtain the screen breakage acoustic signature sequence. A normal voiceprint baseline model is constructed based on the mean vector and covariance matrix of the Mel frequency cepstral coefficients of the audio signal during the no-load phase. The Mahalanobis distance is calculated between the Mel frequency cepstral coefficient vectors corresponding to each acquisition in the screen breakage voiceprint feature sequence and the normal voiceprint baseline model. The Mahalanobis distance calculation results are compared one by one with the 95% confidence interval threshold of the 13-DOF chi-square distribution to obtain the voiceprint Mahalanobis distance decision result. The number of times the Mahalanobis distance in the voiceprint decision result exceeds the threshold is double-cross-validated with the duration for which the asymmetric drift index continuously exceeds the primary decision threshold. When the number of times the threshold is exceeded is not less than 3 and the duration for which the asymmetric drift index continuously exceeds the primary decision threshold is not less than 5 minutes, the latent co-evolution window is confirmed.

5. The intelligent operation and maintenance method for vibrating screens based on multi-source fusion according to claim 1, characterized in that, Step S4, based on the dynamic load increment of the drift-side spring corresponding to the asymmetric drift index in the implicit co-evolution window, includes: Based on the sign direction of the left-right asymmetric drift component and the front-back asymmetric drift component in the asymmetric drift index, the drift-side spring group is determined, and the rated dynamic load of the drift-side spring group is obtained from the equipment nameplate parameters. Based on the left and right asymmetric drift components and the rated dynamic load, the product of half of the rated dynamic load and the lateral coupling coefficient of the screen frame is calculated to obtain the dynamic load increment of the drift side spring. The actual dynamic load of the drift-side spring is obtained by summing the dynamic load increment of the drift-side spring with the rated dynamic load.

6. The intelligent operation and maintenance method for vibrating screens based on multi-source fusion according to claim 5, characterized in that, In step S4, the remaining fatigue cycle number of the drift-side spring is calculated by substituting into the fatigue equation to obtain the window closing time limit, including: Based on the actual dynamic load of the drift-side spring, the stress amplitude and average stress of the drift-side spring are converted to obtain the stress amplitude and average stress of the drift-side spring; the stress amplitude and average stress of the drift-side spring are substituted into the Goodman correction criterion to obtain the equivalent fatigue limit stress amplitude of the drift-side spring under the current working condition. Substituting the equivalent fatigue limit stress amplitude into the Basquin fatigue equation, the total number of fatigue cycles is obtained; based on the product of the excitation frequency of the vibrating screen and the number of operating hours, the number of fatigue cycles consumed by the drift side spring is calculated to obtain the number of fatigue cycles consumed. The remaining fatigue cycle count is obtained by subtracting the total number of fatigue cycles from the number of fatigue cycles already consumed; the window closing time limit is obtained by converting the ratio of the remaining fatigue cycle count to the excitation frequency.

7. The intelligent operation and maintenance method for vibrating screens based on multi-source fusion according to claim 6, characterized in that, In step S4, a fine-tuning command for the excitation frequency is output to the vibrating screen's exciter according to the window closing time limit, so that the asymmetric drift index falls back below the primary judgment threshold, extending the latent co-evolution window, including: Based on the stiffness coefficient of the drift side spring group and the natural frequency of the corresponding subsystem of the drift side spring group, the dynamic load ratio coefficient of each side spring group under the current excitation frequency is calculated to obtain the load ratio coefficient of each side spring group. Within ±2Hz of the current excitation frequency, the difference between the load ratio coefficient of the drift side spring group and the corresponding rated load ratio coefficient in the individualized balance baseline fingerprint at each candidate excitation frequency is calculated one by one. The difference between the candidate excitation frequency corresponding to the smallest difference and the current excitation frequency is determined as the optimal fine-tuning amount of the excitation frequency. Based on the optimal fine-tuning amount of the excitation frequency, an excitation frequency fine-tuning command is output to the vibrator of the vibrating screen to gradually change the excitation frequency to the target excitation frequency with a slope of 0.5 Hz / s. After the excitation frequency gradually changes, the real-time vibration signal of each spring support point of the sieve frame is re-compared with the individualized balance baseline fingerprint to obtain the updated asymmetric drift index. When the updated asymmetric drift index falls below the primary judgment threshold, the window closing time limit is recalculated to obtain the latent co-evolution window.

8. A smart operation and maintenance system for vibrating screens based on multi-source fusion, characterized in that, For implementing the intelligent operation and maintenance method for vibrating screens based on multi-source fusion as described in any one of claims 1-7, the intelligent operation and maintenance system for vibrating screens based on multi-source fusion includes: The processing module is used to normalize the multi-directional vibration signals of each spring support point of the vibrating screen frame during the no-load operation phase of the vibrating screen to obtain an individualized balance baseline fingerprint. The comparison module is used to compare the real-time vibration signal of each spring support point of the screen frame with the individualized balance baseline fingerprint during the full-load operation of the vibrating screen to obtain the asymmetric drift index. The verification module is used to extract the acoustic features of the audio signal of the vibrating screen mesh when the asymmetric drift index exceeds the primary judgment threshold, obtain the acoustic features of the screen mesh damage, and perform double cross-validation between the acoustic features of the screen mesh damage and the asymmetric drift index to confirm the hidden co-evolution window. The calculation module is used to calculate the remaining fatigue cycle number of the drift-side spring based on the dynamic load increment of the drift-side spring corresponding to the asymmetric drift index in the implicit co-evolution window, and obtain the window closing time limit. Based on the window closing time limit, the module outputs a fine-tuning command of the excitation frequency to the vibrating screen exciter so that the asymmetric drift index falls back to below the primary judgment threshold, thereby extending the implicit co-evolution window.

9. A smart operation and maintenance device for vibrating screens based on multi-source fusion, characterized in that, It includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the intelligent operation and maintenance method for vibrating screen based on multi-source fusion as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the intelligent operation and maintenance method for vibrating screen based on multi-source fusion as described in any one of claims 1 to 7.

Images