A method for online detection and real-time control of grinding media ball diameter in a disc grinder.
By constructing an effective signal time window and depth attenuation compensation in a disc mill, and combining multidimensional characteristic parameters and force-shape coupling response matrix, online detection and real-time control of the media ball diameter were achieved. This solved the problems of insufficient detection accuracy and unstable control, and improved grinding efficiency and product particle size stability.
Patent Information
- Application Number
- CN202611171130.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-04
- Publication Date
- 2026-11-17
AI Technical Summary
In the existing technology, the detection accuracy of the grinding media ball diameter in disc grinders is insufficient, and the control of operating parameters is difficult to stabilize, resulting in poor grinding efficiency and product particle size.
By installing an exciter on the grinding disc or sidewall to emit detection signals, collecting and analyzing time-domain echo signals, constructing an effective signal time window, performing deep attenuation compensation, extracting multi-dimensional feature parameters, and combining them with the force-shape coupling response matrix for coordinated control, online detection and real-time control of the medium sphere diameter can be achieved.
It improves the accuracy and stability of media ball diameter detection, ensures coordinated adjustment of grinding mill operating parameters, avoids parameter oscillation, and achieves continuous stability of ball diameter gradation.
Smart Images

Figure CN123102452A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of grinding and pulverizing control technology, specifically to a method for online detection and real-time control of the diameter of grinding media balls in a disc grinder. Background Technology
[0002] Disc mills are core equipment in the industrial grinding field. They utilize a high-speed rotating grinding disc to drive grinding media (such as grinding balls) in a cascading motion within the grinding chamber, achieving material crushing through the impact and crushing action of the grinding balls. The diameter and gradation of the grinding media balls are key process parameters determining grinding efficiency and product particle size. Currently, the detection of grinding ball diameter mainly relies on manual sampling and sieving after machine shutdown or offline analysis based on laser particle size analyzers. During operation, operators judge the ball wear status based on indirect indicators such as main motor current and output particle size, combined with experience, and replenish or replace the grinding balls according to a fixed cycle. Existing technologies aim to achieve online sensing of ball diameter by installing vibration sensors or acoustic emission sensors on the grinding disc or chamber, collecting the impact signals generated by the collision of grinding balls, and performing spectral analysis.
[0003] However, the amplitude of the rigid reflection signal from the grinding disc / liner is much higher than the weak scattered signal generated by the collision of the grinding media balls. The two signals are severely aliased in the time domain, resulting in the extracted spectral features being mixed with a large amount of non-ball diameter information, leading to insufficient ball diameter detection accuracy. Secondly, the scattered signals at different material layer depths exhibit significant amplitude distortion due to differences in propagation distance. Existing methods do not effectively compensate for this, misinterpreting propagation attenuation as a change in ball diameter, further reducing detection reliability. Thirdly, when a ball diameter deviation is detected, there is a strong coupling relationship between operating parameters such as grinding disc rotation speed, auxiliary airflow, and grinding pressure. Adjusting any one parameter alone will trigger a chain reaction of drift in other parameters, leading to oscillations and difficulty in convergence during the control process, making it difficult to maintain the continuous stability of the optimal ball diameter distribution. Therefore, this invention studies and designs an online detection and real-time control method for the ball diameter of grinding media in a disc grinder. Summary of the Invention
[0004] Therefore, the technical problem to be solved by the present invention is to overcome the defects in the prior art where rigid reflection signals and medium sphere scattering signals are severely mixed in the time domain, resulting in the extracted spectral features being mixed with a large amount of non-sphere diameter interference information and insufficient sphere diameter detection accuracy. Thus, an online detection and real-time control method for the sphere diameter of grinding media in a disc grinder is provided.
[0005] To address the above problems, this invention provides a method for online detection and real-time control of the diameter of grinding media balls in a disc grinder, comprising the following steps: S1: Control the actuator installed on the grinding disc or sidewall to emit a detection signal to the grinding medium layer, and collect the time-domain echo signal after propagation through the grinding medium layer; perform waveform analysis on the time-domain echo signal to determine the time of the first reflection wave and the time of the second reflection wave; construct an effective signal time window based on the time of the first reflection wave and the time of the second reflection wave, and extract the original echo signal within the effective signal time window as the signal within the window; calculate the equivalent attenuation coefficient based on the amplitude corresponding to the time of the first reflection wave and the amplitude corresponding to the time of the second reflection wave, as well as the preset reference propagation distance; perform depth attenuation compensation on the signal within the window based on the equivalent attenuation coefficient to obtain the effective scattered signal; S2: Extract time-domain feature parameters, including at least root mean square amplitude and peak factor, from the effective scattering signal. Combine the equivalent attenuation coefficient and the propagation speed of the detection signal in the grinding medium layer to construct a multi-dimensional feature vector to characterize the ball diameter state of the medium. Perform dimensionality reduction and fusion processing on the multi-dimensional feature vector to generate a ball diameter evaluation index to characterize the current ball diameter state of the medium. S3: Compare the ball diameter evaluation index with the preset target ball diameter parameter and calculate the ball diameter deviation; calculate the cooperative execution vector based on the ball diameter deviation and the preset force-shape coupling response matrix; generate and send control commands to the actuator of the grinding machine according to the cooperative execution vector to adjust the operating parameters of the grinding machine so that the ball diameter evaluation index converges to the target ball diameter parameter; S4: Repeat steps S1 to S3 to form a deviation ablation cycle; calculate the weighted energy function of the ball diameter deviation in each deviation ablation cycle; when the calculated value of the weighted energy function is less than the preset convergence threshold, it is determined that a steady state has been reached and the current operating parameters are locked.
[0006] Preferably, in step S1, the detection signal includes a low-frequency detection signal and a high-frequency detection signal; Step S1 further includes: controlling the exciter to alternately transmit the low-frequency detection signal and the high-frequency detection signal based on the time-division multiplexing mechanism, and collecting the corresponding low-frequency time-domain echo signal and high-frequency time-domain echo signal respectively; The effective signal time window is truncated and the depth attenuation compensation is performed on the low-frequency time-domain echo signal and the high-frequency time-domain echo signal, respectively, to obtain the low-frequency effective scattering signal and the high-frequency effective scattering signal.
[0007] Preferably, in step S1, the construction of the effective signal time window is as follows: after taking the absolute value of the time-domain echo signal, envelope detection is performed, and the first peak position exceeding a preset threshold is searched in the envelope signal. The time corresponding to the peak position is marked as the first reflection wave time; the time corresponding to the next peak position of the first reflection wave time is marked as the second reflection wave time. The pulse main lobe time width is determined based on the number of oscillation cycles and center frequency of the detected signal. The sum of the first reflection wave time and the time width is taken as the lower limit of the effective signal time window, and the difference between the second reflection wave time and the time width is taken as the upper limit of the effective signal time window. The equivalent attenuation coefficient is calculated by dividing the natural logarithm of the ratio of the amplitude corresponding to the first reflection wave moment to the amplitude corresponding to the second reflection wave moment by twice the preset reference propagation distance.
[0008] Preferably, in step S1, the method for performing depth attenuation compensation includes: The propagation speed of the detection signal in the abrasive medium layer is determined based on the ratio of the preset reference propagation distance to the moment of the first reflection wave. For each moment within the effective signal time window, calculate the time difference between that moment and the moment of the first reflected wave, and take half of the product of the time difference and the propagation speed as the scattering depth; An exponential function is constructed with the natural constant as the base and the product of the equivalent attenuation coefficient and the scattering depth as the exponent; the exponential function is multiplied by the signal within the window to obtain the effective scattering signal.
[0009] Preferably, in step S2, the multidimensional feature vector includes the ratio of the root mean square amplitude of the low-frequency effective scattering signal and the high-frequency effective scattering signal, and the ratio of the root mean square amplitude serves as a size sensitivity index.
[0010] Preferably, in step S2, the method for performing dimensionality reduction and fusion processing on the multidimensional feature vector includes: Each feature in the multidimensional feature vector is normalized to obtain a normalized feature vector; Using the unit circle as the mapping space, a corresponding number of anchor points are determined on the circumference of the unit circle according to the number of dimensions of the normalized feature vector, and each anchor point corresponds to a one-dimensional feature of the normalized feature vector. Each feature value is projected proportionally onto the line connecting the anchor point and the origin to obtain the projection point; Connect all the projection points in sequence to form a projection polygon, calculate the perimeter and area of the projection polygon, and use the perimeter and area as second-order feature quantities; The second-order feature quantity is mapped to the sphere diameter evaluation index according to the preset sphere diameter evaluation model.
[0011] Preferably, the sphere diameter evaluation model is a high-order polynomial model. The model uses the perimeter and area of the projected polygon as independent variables, and obtains polynomial coefficients by fitting multiple sets of known sphere diameter calibration data. The polynomial combination value of the perimeter and area is used as the sphere diameter evaluation index.
[0012] Preferably, in step S3, the method for calculating the cooperative execution vector includes: Obtain the force-shape coupling response matrix, where each element of the force-shape coupling response matrix represents the influence coefficient of an operating parameter on the ball diameter evaluation index; Performing a generalized inverse operation on the force-shape coupling response matrix yields the inverse coupling deconstruction matrix; The ball diameter deviation is constructed into a deviation vector. The inverse coupling deconstruction matrix and the deviation vector are subjected to matrix operation. An anti-oscillation gain coefficient is introduced to scale the operation result, resulting in the cooperative execution vector containing multiple operating parameter adjustment amounts.
[0013] Preferably, the method further includes: obtaining the cumulative operating time or cumulative processing volume of the grinding media layer, and calculating the theoretical attenuation coefficient based on the cumulative operating time or cumulative processing volume; Obtain the actual ball diameter data under the preset standard test conditions, and compare it with the theoretical ball diameter of the target ball diameter parameter under the standard test conditions to calculate the measured offset coefficient. The theoretical attenuation coefficient and the measured offset coefficient are weighted and summed to generate the performance attenuation factor; The force-shape coupling response matrix is scaled and corrected using the performance degradation factor to obtain a time-varying force-shape coupling response matrix.
[0014] Preferably, step S1 further includes: synchronously collecting the deposition thickness distribution data of the grinding media layer, and calculating the porosity correction factor based on the deposition thickness distribution data; The effective scattering signal is divided using the porosity correction factor to obtain the density-corrected effective scattering signal. Step S3 further includes: identifying abnormal areas where the ball diameter deviation exceeds a preset threshold based on the spatial distribution of the stacking thickness distribution data and the ball diameter evaluation index; Based on the location and deviation of the abnormal area, a fixed-point targeted adjustment command is generated, which includes target position parameters and action intensity parameters.
[0015] The online detection and real-time control method for the diameter of grinding media balls in a disc grinder provided by this invention has the following beneficial effects: 1. This invention identifies the time of the first and second reflection waves in the time domain, and constructs an effective signal time window accordingly. The window starts after the end of the main lobe of the first reflection wave and ends before the start of the main lobe of the second reflection wave. This eliminates the strong energy interference from the rigid reflection of the grinding disc / liner plate within the window, retaining only the scattered signal from the dielectric sphere and the material layer. Simultaneously, based on the amplitude ratio of the reflected waves, an equivalent attenuation coefficient is calculated, and the signal at different times within the window is exponentially compensated according to the scattering depth, eliminating amplitude distortion caused by differences in propagation distance. Therefore, the extracted effective scattered signal can truly reflect the scattering characteristics of the dielectric sphere itself, rather than aliasing interference or attenuation artifacts, avoiding the problem of low sphere diameter inversion accuracy caused by time-domain aliasing of rigid reflection and scattered signals in passive impact detection. 2. This invention also extracts multi-dimensional time-domain feature parameters such as root mean square amplitude and peak factor from the effective scattering signal, constructs a multi-dimensional feature vector by combining the equivalent attenuation coefficient and propagation velocity, and then normalizes and projects each feature onto a unit circle to form a projected polygon using a circle-like mapping method. The perimeter and area of the polygon are used as second-order feature quantities that comprehensively reflect all feature information, and the second-order feature quantities are mapped into a sphere diameter evaluation index through a pre-calibrated sphere diameter evaluation model. The dimensionality reduction and fusion processing integrates multiple heterogeneous physical features into a single sphere diameter evaluation index, avoiding the one-sidedness of single feature evaluation. 3. This invention also constructs an inverse coupling deconstruction matrix by performing a generalized inverse operation on the force-shape coupling response matrix. The ball diameter deviation is mapped into a collaborative execution vector containing the adjustment of multiple operating parameters in one go through this inverse coupling deconstruction matrix. This allows multiple operating parameters, such as the grinding disc speed, auxiliary air volume, and grinding pressure, to be synchronously and collaboratively adjusted according to their respective influence weights. Then, by iteratively executing the deviation ablation cycle and determining convergence with a weighted energy function, the ball diameter deviation gradually converges to the target value within a finite number of iterations. The collaborative control mechanism mathematically decouples the multivariate coupling influence of multiple operating parameters on the ball diameter, avoiding the ebb and flow parameter oscillations caused by the sequential adjustment of single variables. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the feature extraction and spherical diameter evaluation process of the present invention; Figure 3 This is a schematic diagram of the closed-loop control and iterative convergence process of the present invention. Detailed Implementation
[0017] like Figure 1-3 As shown, the present invention provides an online detection and real-time control method for the diameter of grinding media balls in a disc grinder, which includes the following steps: S1: Control the exciter installed on the grinding disc or sidewall to emit a detection signal to the grinding medium layer, and collect the time-domain echo signal after propagation through the grinding medium layer; perform waveform analysis on the time-domain echo signal to determine the time of the first reflection wave and the time of the second reflection wave; construct an effective signal time window based on the time of the first reflection wave and the time of the second reflection wave, and extract the original echo signal within the effective signal time window as the signal within the window; calculate the equivalent attenuation coefficient based on the amplitude corresponding to the time of the first reflection wave and the amplitude corresponding to the time of the second reflection wave, as well as the preset reference propagation distance; perform depth attenuation compensation on the signal within the window based on the equivalent attenuation coefficient to obtain the effective scattered signal.
[0018] The actuator employs a piezoelectric ceramic transducer, installed at the bottom of the grinding disc or on the side wall of the grinding chamber, and fixed by adhesive or threaded connection. A broadband mechanical sensor is installed near the actuator as a receiver; the sensor is a piezoelectric accelerometer with a frequency response range covering 0.1 to 10 times the center frequency of the actuator. When the actuator receives an electrical pulse signal, it converts electrical energy into mechanical vibration and transmits a detection signal to the grinding media layer. During propagation, the detection signal is reflected, refracted, and scattered when it encounters grinding media balls, material particles, and the surface of the grinding disc, with some signal returning to the receiver, forming a time-domain echo signal. The signal acquisition system uses a high-speed data acquisition card with a sampling frequency of not less than 10 times the center frequency of the actuator, and the sampling duration covers at least the arrival time of the second reflected wave.
[0019] Specifically, the waveform analysis process is as follows: Bandpass filtering is applied to the time-domain echo signal, with the passband covering the main frequency range of the detection signal; the absolute value of the filtered signal is taken, and then envelope detection is performed using a low-pass filter to extract the signal envelope; the global maximum value is searched in the envelope signal, and if the maximum value exceeds a preset threshold, it is marked as the moment of the first reflection wave; the preset threshold is set as follows: background noise signals are collected when the grinding machine is unloaded and the grinding disc surface is clean, and three times the peak value of the background noise is used as the threshold; the first reflection wave corresponds to the reflection of the detection signal on the upper surface of the grinding disc, with the highest amplitude and earliest arrival time; the next local maximum value is searched in the envelope signal after the first reflection wave, and if the local maximum value exceeds 10% of the amplitude of the first reflection wave, it is marked as the moment of the second reflection wave; the second reflection wave corresponds to the signal of the detection signal completing one round trip reflection between the upper and lower surfaces of the grinding disc.
[0020] Furthermore, the effective signal time window is constructed as follows: the number of oscillation cycles of the excitation signal is obtained, which is determined by the pulse width of the exciter; the number of oscillation cycles is divided by the center frequency of the exciter to obtain the time width of the pulse main lobe; the lower limit of the effective signal time window is the sum of the time width of the first reflection wave and the second reflection wave, and the upper limit is the difference between the time width of the second reflection wave and the second reflection wave; the interval between the lower and upper limits of the window is the effective signal time window; the window starts after the end of the first reflection wave main lobe and ends before the start of the second reflection wave main lobe, thereby excluding the rigid reflection signal on the surface of the grinding disk and retaining only the scattering signal between the dielectric sphere and the material layer.
[0021] The equivalent attenuation coefficient is calculated as follows: the natural logarithm of the ratio of the amplitude corresponding to the first reflection wave moment to the amplitude corresponding to the second reflection wave moment is divided by twice the preset reference propagation distance; where the preset reference propagation distance is the path length of the detection signal in a single propagation in the grinding medium layer, which is obtained by a calibration measurement performed under no-load conditions of the grinding machine; the equivalent attenuation coefficient characterizes the energy loss rate of the detection signal when it propagates in the grinding medium layer.
[0022] The method for depth attenuation compensation is as follows: the propagation speed is obtained by dividing the preset reference propagation distance by the time of the first reflection wave; for each moment within the effective signal time window, the corresponding scattering depth is calculated based on the time difference between that moment and the time of the first reflection wave and the propagation speed, and then an exponential compensation function with the natural constant as the base is constructed. The compensation function is multiplied by the signal within the window to obtain the effective scattered signal; this compensation eliminates the amplitude difference caused by different propagation distances, so that the scattered signals at different depths are restored to the same amplitude reference.
[0023] S2: Extract time-domain feature parameters, including at least root mean square amplitude and peak factor, from the effective scattering signal. Combine the equivalent attenuation coefficient and the propagation speed of the detection signal in the grinding medium layer to construct a multi-dimensional feature vector to characterize the ball diameter state of the medium. Perform dimensionality reduction and fusion processing on the multi-dimensional feature vector to generate a ball diameter evaluation index to characterize the current ball diameter state of the medium.
[0024] Among them, the root mean square amplitude of the effective scattered signal is obtained by integrating the instantaneous amplitude of the effective scattered signal with the square of the time window, dividing by the time window length, and then taking the square root. The root mean square amplitude reflects the overall energy level of the scattered signal and is related to the overall scattering cross section of the dielectric sphere. The peak factor is defined as the ratio of the peak amplitude of the effective scattered signal to the root mean square amplitude, reflecting the waveform peak of the scattered signal and is related to the uniformity of the size distribution of the dielectric sphere. The propagation speed is obtained by dividing the reference propagation distance by the time of the first reflection wave. The above four characteristic parameters constitute a four-dimensional characteristic vector, and each characteristic parameter reflects the state of the dielectric sphere from different physical dimensions.
[0025] Specifically, the dimensionality reduction and fusion processing method is as follows: The four-dimensional feature vector is normalized to its maximum and minimum values, mapping each feature dimension to the 0-1 interval to eliminate dimensional differences; using the unit circle as the mapping space, four anchor points are uniformly set on the circumference, corresponding to the four feature dimensions respectively; the normalized feature values are projected proportionally from the origin onto the line connecting the corresponding anchor points to obtain four projection points; the four projection points are connected sequentially to form a projection quadrilateral, and the perimeter and area of the quadrilateral are calculated, with the perimeter and area used as second-order feature quantities; the second-order feature quantities are mapped to sphere diameter evaluation indicators through a pre-established sphere diameter evaluation model; the sphere diameter evaluation model is a high-order polynomial model, determined through calibration experiments: multiple sets of standard medium sphere samples with known diameters are prepared, run in a grinding mill and signals are collected, and after extracting the second-order feature quantities, polynomial fitting is performed.
[0026] S3: Compare the ball diameter evaluation index with the preset target ball diameter parameter and calculate the ball diameter deviation; calculate the cooperative execution vector based on the ball diameter deviation and the preset force-shape coupling response matrix; generate and send control commands to the actuator of the grinding machine according to the cooperative execution vector to adjust the operating parameters of the grinding machine so that the ball diameter evaluation index converges to the target ball diameter parameter.
[0027] The target ball diameter parameter is the desired ball diameter value set by the grinding process, which is preset by the operator according to the material characteristics and product requirements. The ball diameter deviation is equal to the ball diameter evaluation index minus the target ball diameter parameter. A positive deviation indicates that the current ball diameter is too large, and a negative deviation indicates that the current ball diameter is too small.
[0028] Specifically, the force-shape coupling response matrix is a 2x3 matrix, with the rows corresponding to the two geometric dimensions of the ball diameter evaluation index and the columns corresponding to the three operating parameters of the grinding machine. Each element of the matrix is obtained by calibrating the corresponding operating parameter through a step disturbance experiment. The calculation method of the cooperative execution vector is as follows: perform a generalized inverse operation on the force-shape coupling response matrix to obtain an inverse coupling deconstruction matrix; construct the ball diameter deviation as a deviation vector; perform matrix operations on the inverse coupling deconstruction matrix and the deviation vector, and introduce an anti-oscillation gain coefficient to scale the calculation result to obtain a cooperative execution vector containing the adjustment amounts of multiple operating parameters. The role of the anti-oscillation gain coefficient is to limit the single adjustment amplitude and prevent the system from over-adjustment oscillation.
[0029] S4: Repeat steps S1 to S3 to form a deviation ablation cycle; calculate the weighted energy function of the ball diameter deviation in each deviation ablation cycle; when the calculated value of the weighted energy function is less than the preset convergence threshold, it is determined that a steady state has been reached and the current operating parameters are locked.
[0030] Each deviation ablation cycle includes a complete signal acquisition, ball diameter evaluation, deviation calculation, and adjustment. The weighted energy function is defined as the product of the square of the ball diameter deviation and the preset weight coefficient. When the weighted energy function value is less than the convergence threshold, it indicates that the ball diameter deviation has been controlled within an acceptable range, the system has reached a steady state, the current operating parameters are locked, and the iteration stops. The convergence threshold is set according to the process accuracy requirements.
[0031] In some embodiments, in step S1, the detection signal includes a low-frequency detection signal and a high-frequency detection signal; step S1 further includes: controlling the exciter to alternately transmit the low-frequency detection signal and the high-frequency detection signal based on a time-division multiplexing mechanism, and acquiring the corresponding low-frequency time-domain echo signal and high-frequency time-domain echo signal respectively; performing the effective signal time window truncation and the depth attenuation compensation on the low-frequency time-domain echo signal and the high-frequency time-domain echo signal respectively to obtain the low-frequency effective scattering signal and the high-frequency effective scattering signal.
[0032] The low-frequency detection signal has a frequency range of 15kHz to 30kHz, while the high-frequency detection signal has a frequency range of 80kHz to 150kHz. The low-frequency signal has a longer wavelength and stronger penetration ability, and is sensitive to changes in the overall density of the medium layer but relatively insensitive to changes in the size of individual particles. The high-frequency signal has a shorter wavelength, is sensitive to changes in particle size, but has relatively weaker penetration ability. The alternating transmission of the two frequency signals is achieved through a time-division multiplexing mechanism: within a complete detection cycle, the low-frequency signal is transmitted first and then received, and after a certain interval, the high-frequency signal is transmitted and received, ensuring that the two frequency signals do not overlap in the time domain.
[0033] In this process, two independent exciters are used to transmit signals of different frequencies. In practical applications, a single exciter can also be used to respond to excitation pulses of two frequencies simultaneously. The time-domain echo signal collected by the receiver is separated into a low-frequency time-domain echo signal and a high-frequency time-domain echo signal according to the transmission sequence. The effective signal time window truncation and depth attenuation compensation described in step S1 are performed independently on these two sets of signals to obtain a low-frequency effective scattered signal and a high-frequency effective scattered signal. The low-frequency effective scattered signal reflects the overall density state of the dielectric layer, and the high-frequency effective scattered signal reflects the size state of the dielectric sphere.
[0034] In some implementations, in step S1, constructing the effective signal time window involves: taking the absolute value of the time-domain echo signal and performing envelope detection; searching for the first peak position exceeding a preset threshold in the envelope signal; marking the time corresponding to the peak position as the first reflection wave time; marking the time corresponding to the next peak position after the first reflection wave time as the second reflection wave time; determining the pulse main lobe time width based on the oscillation period number and center frequency of the detected signal; using the sum of the first reflection wave time and the time width as the lower limit of the effective signal time window; and using the difference between the second reflection wave time and the time width as the upper limit of the effective signal time window. The equivalent attenuation coefficient is calculated by dividing the natural logarithm of the ratio of the amplitude corresponding to the first reflection wave time to the amplitude corresponding to the second reflection wave time by twice the preset reference propagation distance.
[0035] Specifically, the envelope detection is implemented as follows: after taking the absolute value of the time-domain echo signal, it is filtered by a low-pass filter with a cutoff frequency of 0.1 times the center frequency of the exciter to output the signal envelope; the preset threshold is set as follows: before the grinder is started and the grinding disc is unloaded, the background noise signal is collected, the peak value of the noise signal envelope is calculated, and three times that value is used as the threshold; the first peak position in the envelope signal that exceeds the threshold corresponds to the first reflected wave of the detection signal after it propagates from the exciter to the upper surface of the grinding disc and is reflected back to the receiver, which has the highest amplitude and the earliest arrival time; the number of oscillation cycles is the number of complete waveform cycles contained in the detection signal emitted by the exciter, which is determined by the excitation pulse width.
[0036] In some embodiments, the method for performing depth attenuation compensation in step S1 includes: determining the propagation speed of the detected signal in the abrasive medium layer based on the ratio of the preset reference propagation distance to the moment of the first reflection wave; calculating the time difference between the moment and the moment of the first reflection wave for each moment within the effective signal time window, and taking half of the product of the time difference and the propagation speed as the scattering depth; constructing an exponential function with the natural constant as the base and the product of the equivalent attenuation coefficient and the scattering depth as the exponent; and multiplying the exponential function by the signal within the window to obtain the effective scattered signal.
[0037] Specifically, the propagation speed is calculated as follows: the preset reference propagation distance is divided by the moment of the first reflection wave. Specifically, the propagation speed represents the average propagation speed of the detection signal in the abrasive medium layer, with the unit being meters per second. Since the abrasive medium layer is a porous medium composed of particles, its equivalent sound speed is less than the sound speed in a dense solid. The meaning of the scattering depth is: the signal at a certain moment within the window originates from the scattering of the detection signal at a certain depth position on the propagation path. The propagation distance of the signal returning from the scattering point to the receiver is equal to the time difference between that moment and the moment of the first reflection wave multiplied by half the propagation speed. The scattering depth is calculated from the receiver end face and corresponds to the physical position where scattering occurs in the abrasive medium layer. After compensation, the scattering signals at different depths are restored to the same amplitude reference, eliminating the interference of propagation attenuation on the extraction of scattering features.
[0038] In some implementations, in step S2, the multidimensional feature vector includes the ratio of the root mean square amplitude of the low-frequency effective scattered signal and the high-frequency effective scattered signal, and the ratio of the root mean square amplitude serves as a size sensitivity index.
[0039] Specifically, the root mean square amplitude of the low-frequency effective scattered signal is extracted from the low-frequency effective scattered signal, and the root mean square amplitude of the high-frequency effective scattered signal is extracted from the high-frequency effective scattered signal. The ratio between the two is calculated as: the low-frequency root mean square amplitude divided by the high-frequency root mean square amplitude. The physical meaning of the ratio is as follows: the low-frequency signal is more affected by the change in the density of the medium layer but less affected by the change in particle size, while the high-frequency signal is more affected by the change in particle size but less affected by the change in density. When the density of the medium layer increases, the root mean square amplitudes of both high-frequency and low-frequency signals decrease by similar proportions, and the ratio remains basically unchanged. When the diameter of the medium spheres changes, the change amplitude of the high-frequency signal is significantly greater than that of the low-frequency signal, and the ratio changes significantly. The size sensitivity index, together with other features in the multidimensional feature vector, forms a complete multidimensional feature vector, which participates in the dimensionality reduction and fusion processing.
[0040] In some embodiments, step S2, the method for performing dimensionality reduction and fusion processing on the multidimensional feature vector, includes: normalizing each dimension of the multidimensional feature vector to obtain a normalized feature vector; using a unit circle as the mapping space, determining a corresponding number of anchor points on the circumference of the unit circle according to the number of dimensions of the normalized feature vector, each anchor point corresponding to a one-dimensional feature of the normalized feature vector; projecting each feature value proportionally onto the line connecting the anchor point and the origin to obtain a projection point; sequentially connecting all the projection points to form a projection polygon, calculating the perimeter and area of the projection polygon, and using the perimeter and area as second-order feature quantities; and mapping the second-order feature quantities to the sphere diameter evaluation index according to a preset sphere diameter evaluation model.
[0041] Specifically, the normalization process employs the maximum-minimum normalization method: for each feature dimension of the multidimensional feature vector, the maximum and minimum values of that feature are found in the calibration dataset. The current value is subtracted from the minimum value and then divided by the range, mapping it to the interval between 0 and 1. The normalization method does not change the relative size relationship of each feature dimension and eliminates dimensional differences. The anchor points are set as follows: they are evenly distributed on the circumference of the unit circle, with each anchor point corresponding to a fixed one-dimensional feature. The projection point is located on the line connecting the origin and the anchor point, and the distance from the origin is equal to the normalized feature value multiplied by the radius of the unit circle. The perimeter of the projected polygon is obtained by accumulating the Euclidean distance between adjacent projection points, and the area is obtained by decomposing the polygon into multiple triangles with the origin as the common vertex and calculating the sum of the areas of each triangle. The perimeter and area comprehensively reflect the value range and balance of all feature dimensions.
[0042] In some embodiments, the sphere diameter evaluation model is a high-order polynomial model. The model uses the perimeter and area of the projected polygon as independent variables, and obtains polynomial coefficients by fitting multiple sets of known sphere diameter calibration data. The polynomial combination value of the perimeter and area is used as the sphere diameter evaluation index.
[0043] Specifically, the order of the high-order polynomial model is determined based on the amount of calibration data and the required fitting accuracy. The model includes constant terms and various power combinations of perimeter and area. These coefficients are obtained by fitting the calibration data using the least squares method. The calibration data is obtained as follows: prepare multiple sets of standard medium ball samples with different diameters, covering the range that may occur in the grinding process; load each set of samples into a grinding machine and run it under the same conditions as actual production, collecting signals, extracting features, and calculating second-order feature quantities according to the aforementioned method, while accurately measuring the actual ball diameter using calipers or image analysis; input multiple sets of data into a polynomial regression algorithm to calculate the polynomial coefficients; after the model is established, for online detection samples with unknown ball diameters, extract second-order feature quantities according to the same process, substitute them into the polynomial model, and calculate the output value, which is the ball diameter evaluation index.
[0044] In some embodiments, the method for calculating the collaborative execution vector in step S3 includes: obtaining the force-shape coupling response matrix, where each element of the force-shape coupling response matrix represents the influence coefficient of an operating parameter on the sphere diameter evaluation index; performing a generalized inverse operation on the force-shape coupling response matrix to obtain an inverse coupling deconstruction matrix; constructing the sphere diameter deviation as a deviation vector; performing a matrix operation between the inverse coupling deconstruction matrix and the deviation vector; and introducing an anti-oscillation gain coefficient to scale the operation result to obtain the collaborative execution vector containing multiple operating parameter adjustment amounts.
[0045] Specifically, the force-shape coupling response matrix is a 2x3 matrix: row 1 corresponds to the average ball diameter dimension in the ball diameter evaluation index, and row 2 corresponds to the ball diameter distribution width dimension in the ball diameter evaluation index; column 1 corresponds to the grinding disc rotation speed, column 2 corresponds to the auxiliary air volume, and column 3 corresponds to the grinding pressure; the matrix elements represent the influence of a unit change in the j-th operating parameter on the i-th geometric dimension; the values of each element are calibrated through a step response experiment: keeping other operating parameters constant, the j-th operating parameter is increased by a small increment, the change in the i-th geometric dimension is measured, and the ratio of the change to the increment is calculated; the generalized inverse operation maps the deviation vector from the 2-dimensional geometric space to the 3-dimensional control space; the deviation vector is a 2x1 matrix, with the first component being the average ball diameter deviation and the second component being the distribution width deviation; multiplying the generalized inverse matrix with the deviation vector yields the initial adjustment vector; an anti-oscillation gain coefficient is introduced to scale all adjustment values to obtain the collaborative execution vector; the three components of the collaborative execution vector correspond to the adjustment values of each operating parameter, with signed values indicating increases or decreases.
[0046] In some embodiments, the method further includes: obtaining the cumulative operating time or cumulative processing volume of the grinding media layer, and calculating the theoretical attenuation coefficient based on the cumulative operating time or cumulative processing volume; obtaining the actual ball diameter data under a preset standard test condition, and comparing it with the theoretical ball diameter of the target ball diameter parameter under the standard test condition to calculate the measured offset coefficient; performing a weighted summation of the theoretical attenuation coefficient and the measured offset coefficient to generate a performance attenuation factor; and using the performance attenuation factor to scale and correct the force-shape coupling response matrix to obtain a time-varying corrected force-shape coupling response matrix.
[0047] Specifically, the cumulative running time is obtained through the grinding mill's running timer, and the cumulative throughput is obtained through the cumulative flow meter of the feeding system. The theoretical attenuation coefficient is calculated based on Archard's wear law: the wear of the media balls is directly proportional to the cumulative throughput and inversely proportional to the hardness of the media ball material; the theoretical attenuation coefficient starts from 1 and decreases monotonically with running time, and its variation curve is obtained through previous wear experiments; the standard test condition is defined as a specific combination of operating parameters, under which the grinding mill's operating state is relatively stable; at certain time intervals, the control system automatically switches the grinding mill to the standard test condition, and after running for a preset time, it collects signals and calculates the actual ball diameter evaluation index; at the same time, it queries the theoretical ball diameter value under the standard test condition from the stored model; the ratio of the actual value to the theoretical value is the measured offset coefficient; the weight coefficient of the weighted summation is dynamically adjusted according to the reliability of the measured data; the force-shape coupling response matrix is scaled and corrected using a performance attenuation factor: each element of the matrix is multiplied by the performance attenuation factor.
[0048] In some embodiments, step S1 further includes: synchronously acquiring the stacked thickness distribution data of the grinding media layer, calculating a porosity correction factor based on the stacked thickness distribution data; performing a division operation on the effective scattering signal using the porosity correction factor to obtain a density-corrected effective scattering signal; step S3 further includes: identifying abnormal regions where the ball diameter deviation exceeds a preset threshold based on the spatial distribution of the stacked thickness distribution data and the ball diameter evaluation index; generating a fixed-point targeted adjustment command based on the location and deviation of the abnormal region, wherein the fixed-point targeted adjustment command includes target position parameters and action intensity parameters.
[0049] Specifically, the method for collecting the stacking thickness distribution data is as follows: multiple sets of exciter-receiver pairs are arranged at different radial positions of the grinding disk, with each set spaced apart in space; using the reflected wave time measured by each set, combined with the known installation height of the exciter, the vertical distance from the surface of the grinding disk to the end face of the sensor at each position is calculated, and then the stacking thickness at each position is obtained through differential calculation; the data from multiple positions are summarized to form the stacking thickness distribution data. The porosity correction factor is calculated as follows: For each spatial location, the accumulated thickness at that location is divided by the standard reference loose thickness to obtain the geometric compression ratio; the geometric compression ratio is raised to a preset power exponent and then multiplied by the geometric attenuation reference coefficient to obtain the porosity correction factor at that location; the porosity correction factor reflects the degree of pore closure caused by compression in the medium layer; dividing the effective scattered signal by the porosity correction factor removes the additional mechanical attenuation caused by excessively dense accumulation; the abnormal area identification method is as follows: the spatial distribution of accumulated thickness data and ball diameter evaluation index is correlated; if the accumulated thickness of a certain area exceeds a preset threshold and the ball diameter deviation of that area exceeds a preset deviation threshold, then the area is determined to be an abnormal area; the method for generating the fixed-point targeted adjustment command is as follows: the position coordinates of the abnormal area are converted into the moving target position of the actuator, the accumulated thickness value is combined with the preset mechanical safety margin to calculate the target tooth penetration depth, and the local ball diameter deviation is combined with the global working condition constraints to calculate the stirring or ball replenishment intensity.
[0050] This application identifies the times of the first and second reflection waves in the time domain, and constructs an effective signal time window accordingly. The window starts after the end of the main lobe of the first reflection wave and ends before the start of the main lobe of the second reflection wave. This eliminates the strong energy interference from the rigid reflection of the grinding disc / liner plate within the window, retaining only the scattered signal from the dielectric sphere and the material layer. Simultaneously, based on the amplitude ratio of the reflected waves, an equivalent attenuation coefficient is calculated, and the signal at different times within the window is exponentially compensated according to the scattering depth, eliminating amplitude distortion caused by differences in propagation distance. As a result, the extracted effective scattered signal can truly reflect the scattering characteristics of the dielectric sphere itself, rather than aliasing interference or attenuation artifacts, avoiding the problem of low sphere diameter inversion accuracy caused by time-domain aliasing of rigid reflection and scattered signals in passive impact detection.
[0051] This application extracts multi-dimensional time-domain feature parameters such as root mean square amplitude and peak factor from the effective scattering signal, constructs a multi-dimensional feature vector by combining the equivalent attenuation coefficient and propagation velocity, and then normalizes and projects each feature onto a unit circle to form a projected polygon using a circle-like mapping method. The perimeter and area of the polygon are used as second-order feature quantities that comprehensively reflect all feature information, and the second-order feature quantities are mapped into a sphere diameter evaluation index through a pre-calibrated sphere diameter evaluation model. Dimensionality reduction and fusion processing integrates multiple heterogeneous physical features into a single sphere diameter evaluation index, avoiding the one-sidedness of single feature evaluation.
[0052] This application constructs an inverse coupling deconstruction matrix by performing a generalized inverse operation on the force-shape coupling response matrix. The ball diameter deviation is mapped into a collaborative execution vector containing the adjustment of multiple operating parameters through this inverse coupling deconstruction matrix. This allows multiple operating parameters, such as grinding disc speed, auxiliary air volume, and grinding pressure, to be synchronously and collaboratively adjusted according to their respective influence weights. Then, by iteratively executing the deviation ablation cycle and determining convergence with a weighted energy function, the ball diameter deviation gradually converges to the target value within a finite number of iterations. The collaborative control mechanism mathematically decouples the multivariate coupling influence of multiple operating parameters on the ball diameter, avoiding the ebb and flow parameter oscillations caused by sequential adjustment of single variables.
[0053] This application also provides an online detection and real-time control system for the diameter of grinding media balls in a disc grinder, including the online detection and real-time control method for the diameter of grinding media balls in a disc grinder as described in any of the preceding claims, comprising: a signal acquisition and preprocessing module, used to control an exciter to emit a detection signal and acquire a time-domain echo signal, perform waveform analysis on the time-domain echo signal to determine the time of the first reflection wave and the time of the second reflection wave, construct an effective signal time window based on the time of the first reflection wave and the time of the second reflection wave and extract the signal within the window, calculate the equivalent attenuation coefficient based on the amplitude corresponding to the time of the first reflection wave and the amplitude corresponding to the time of the second reflection wave and a preset reference propagation distance, and based on the equivalent attenuation coefficient... The effective attenuation coefficient performs deep attenuation compensation on the signal within the window, outputting an effective scattered signal; the feature extraction and spherical diameter evaluation module is used to extract time-domain feature parameters from the effective scattered signal and construct a multi-dimensional feature vector, perform dimensionality reduction and fusion processing on the multi-dimensional feature vector, and output a spherical diameter evaluation index; the closed-loop control module is used to compare the spherical diameter evaluation index with the target spherical diameter parameter to calculate the spherical diameter deviation, calculate the cooperative execution vector based on the spherical diameter deviation and the force-shape coupling response matrix, generate and send control commands to the actuator according to the cooperative execution vector; the iterative convergence control module is used to execute the deviation ablation loop, calculate the weighted energy function and determine whether the steady-state convergence condition has been reached.
[0054] Specifically, the signal acquisition and preprocessing module includes an excitation drive unit, a signal conditioning unit, a high-speed acquisition unit, and a digital signal processing unit. The excitation drive unit generates a high-voltage pulse signal to drive the exciter, outputting a pulse amplitude of 100V to 500V with adjustable pulse width. The signal conditioning unit pre-amplifies and bandpass-filters the weak electrical signal output from the receiver, with an amplification gain of 40dB to 80dB. The high-speed acquisition unit uses an analog-to-digital converter with a sampling rate no less than 10 times the center frequency of the exciter and a resolution of 12-bit to 16-bit. The digital signal processing unit is implemented using a field-programmable gate array or a digital signal processor, performing real-time calculations such as envelope detection, peak detection, window truncation, and attenuation compensation. The feature extraction and spherical diameter evaluation module runs on an embedded processor or industrial computer, receiving the effective scattered signal, calculating time-domain characteristic parameters such as root mean square amplitude and peak factor, and performing normalization. The system processes and projects circular mappings, calculates the perimeter and area of the projected polygon, and then calculates the spherical diameter evaluation index using a pre-stored spherical diameter evaluation model. The closed-loop control module receives the spherical diameter evaluation index, combines it with the stored target spherical diameter parameters and force-shape coupling response matrix, performs generalized inverse matrix operations and anti-oscillation gain scaling, and generates a collaborative execution vector. The closed-loop control module communicates with the drivers of each actuator through an industrial Ethernet interface and sends control commands. The iterative convergence control module monitors the weighted energy function value of each deviation ablation cycle. When the weighted energy function value is less than the convergence threshold, it sends a steady-state signal to the system and locks the current operating parameters.
[0055] Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus necessary hardware platform, or it can be implemented entirely by hardware; based on this understanding, all or part of the technical solution of the present invention that contributes to the background technology can be embodied in the form of a software product. The computer software product can be stored in a storage medium and includes several instructions to cause a computer device to execute some of the methods described in various embodiments or examples of the present invention.
[0056] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention. The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the protection scope of the present invention.
Claims
1. A method for on-line detection and real-time control of the ball diameter of a disc mill grinding medium, characterized in that, Includes the following steps: S1: Control the actuator installed on the grinding disc or sidewall to emit a detection signal to the grinding medium layer, and collect the time-domain echo signal after propagation through the grinding medium layer; perform waveform analysis on the time-domain echo signal to determine the time of the first reflection wave and the time of the second reflection wave; construct an effective signal time window based on the time of the first reflection wave and the time of the second reflection wave, and extract the original echo signal within the effective signal time window as the signal within the window; Based on the amplitude corresponding to the first reflection wave moment and the amplitude corresponding to the second reflection wave moment, as well as the preset reference propagation distance, the equivalent attenuation coefficient is calculated; based on the equivalent attenuation coefficient, the signal within the window is subjected to depth attenuation compensation to obtain an effective scattered signal; S2: Extract time-domain feature parameters, including at least the root mean square amplitude and peak factor, from the effective scattered signal, and combine them with the equivalent attenuation coefficient and the propagation speed of the detection signal in the grinding medium layer to construct a multi-dimensional feature vector to characterize the diameter state of the medium spheres. The multidimensional feature vector is subjected to dimensionality reduction and fusion processing to generate a sphere diameter evaluation index for characterizing the current sphere diameter state of the medium; S3: Compare the ball diameter evaluation index with the preset target ball diameter parameters and calculate the ball diameter deviation; calculate the cooperative execution vector based on the ball diameter deviation and the preset force-shape coupling response matrix; Control commands are generated and sent to the grinding machine's actuators based on the cooperative execution vector to adjust the grinding machine's operating parameters so that the ball diameter evaluation index converges to the target ball diameter parameter. S4: Repeat steps S1 to S3 to form a deviation ablation cycle; calculate the weighted energy function of the ball diameter deviation in each deviation ablation cycle; when the calculated value of the weighted energy function is less than the preset convergence threshold, it is determined that a steady state has been reached and the current operating parameters are locked.
2. The method according to claim 1, wherein the method is characterized by: In step S1, the detection signal includes a low-frequency detection signal and a high-frequency detection signal; Step S1 further includes: controlling the exciter to alternately transmit the low-frequency detection signal and the high-frequency detection signal based on the time-division multiplexing mechanism, and collecting the corresponding low-frequency time-domain echo signal and high-frequency time-domain echo signal respectively; The effective signal time window is truncated and the depth attenuation compensation is performed on the low-frequency time-domain echo signal and the high-frequency time-domain echo signal, respectively, to obtain the low-frequency effective scattering signal and the high-frequency effective scattering signal.
3. The method of claim 1, wherein the method further comprises: determining the diameter of the grinding media balls in the mill; and adjusting the diameter of the grinding media balls in the mill in real time. In step S1, the construction of the effective signal time window is as follows: after taking the absolute value of the time-domain echo signal, envelope detection is performed, and the first peak position exceeding a preset threshold is searched in the envelope signal. The time corresponding to the peak position is marked as the first reflection wave time; the time corresponding to the next peak position of the first reflection wave time is marked as the second reflection wave time. The pulse main lobe time width is determined based on the number of oscillation cycles and center frequency of the detected signal. The sum of the first reflection wave time and the time width is taken as the lower limit of the effective signal time window, and the difference between the second reflection wave time and the time width is taken as the upper limit of the effective signal time window. The equivalent attenuation coefficient is calculated by dividing the natural logarithm of the ratio of the amplitude corresponding to the first reflection wave moment to the amplitude corresponding to the second reflection wave moment by twice the preset reference propagation distance.
4. The method for online detection and real-time control of the grinding media ball diameter in a disc grinder according to claim 3, characterized in that, In step S1, the method for performing depth attenuation compensation includes: The propagation speed of the detection signal in the abrasive medium layer is determined based on the ratio of the preset reference propagation distance to the moment of the first reflection wave. For each moment within the effective signal time window, calculate the time difference between that moment and the moment of the first reflected wave, and take half of the product of the time difference and the propagation speed as the scattering depth; An exponential function is constructed with the natural constant as the base and the product of the equivalent attenuation coefficient and the scattering depth as the exponent; the exponential function is multiplied by the signal within the window to obtain the effective scattering signal.
5. The method for online detection and real-time control of the diameter of grinding media balls in a disc grinder according to claim 1, characterized in that, In step S2, the multidimensional feature vector includes the ratio of the root mean square amplitude of the low-frequency effective scattering signal and the high-frequency effective scattering signal, and the ratio of the root mean square amplitude serves as a size sensitivity index.
6. The method for online detection and real-time control of the diameter of grinding media balls in a disc grinder according to claim 1, characterized in that, In step S2, the method for performing dimensionality reduction and fusion processing on the multidimensional feature vector includes: Each feature in the multidimensional feature vector is normalized to obtain a normalized feature vector; Using the unit circle as the mapping space, a corresponding number of anchor points are determined on the circumference of the unit circle according to the number of dimensions of the normalized feature vector, and each anchor point corresponds to a one-dimensional feature of the normalized feature vector. Each feature value is projected proportionally onto the line connecting the anchor point and the origin to obtain the projection point; Connect all the projection points in sequence to form a projection polygon, calculate the perimeter and area of the projection polygon, and use the perimeter and area as second-order feature quantities. The second-order feature quantity is mapped to the sphere diameter evaluation index according to the preset sphere diameter evaluation model.
7. The method for online detection and real-time control of the ball diameter of the grinding media in a disc grinder according to claim 6, characterized in that, The sphere diameter evaluation model is a high-order polynomial model. The model uses the perimeter and area of the projected polygon as independent variables, and obtains polynomial coefficients by fitting multiple sets of known sphere diameter calibration data. The polynomial combination value of the perimeter and area is used as the sphere diameter evaluation index.
8. The method for online detection and real-time control of the diameter of grinding media balls in a disc grinder according to claim 1, characterized in that, In step S3, the method for calculating the cooperative execution vector includes: Obtain the force-shape coupling response matrix, where each element of the force-shape coupling response matrix represents the influence coefficient of an operating parameter on the ball diameter evaluation index; Performing a generalized inverse operation on the force-shape coupling response matrix yields the inverse coupling deconstruction matrix; The ball diameter deviation is constructed into a deviation vector. The inverse coupling deconstruction matrix and the deviation vector are subjected to matrix operation. An anti-oscillation gain coefficient is introduced to scale the operation result, resulting in the cooperative execution vector containing multiple operating parameter adjustment amounts.
9. The method for online detection and real-time control of the diameter of grinding media balls in a disc grinder according to claim 8, characterized in that, Also includes: Obtain the cumulative operating time or cumulative processing volume of the grinding media layer, and calculate the theoretical attenuation coefficient based on the cumulative operating time or processing volume; Obtain the actual ball diameter data under the preset standard test conditions, and compare it with the theoretical ball diameter of the target ball diameter parameter under the standard test conditions to calculate the measured offset coefficient. The theoretical attenuation coefficient and the measured offset coefficient are weighted and summed to generate the performance attenuation factor; The force-shape coupling response matrix is scaled and corrected using the performance degradation factor to obtain a time-varying corrected force-shape coupling response matrix.
10. The method for online detection and real-time control of the ball diameter of the grinding media in a disc grinder according to claim 6, characterized in that, Step S1 further includes: Simultaneously collect the deposition thickness distribution data of the grinding media layer, and calculate the porosity correction factor based on the deposition thickness distribution data; The effective scattering signal is divided using the porosity correction factor to obtain the density-corrected effective scattering signal. Step S3 further includes: Based on the spatial distribution of the stacking thickness data and the ball diameter evaluation index, identify abnormal areas where the ball diameter deviation exceeds a preset threshold; Based on the location and deviation of the abnormal area, a fixed-point targeting adjustment command is generated, which includes target position parameters and action intensity parameters.
