A rubber lining board intelligent maintenance management method and system

Abstract

The application discloses a kind of rubber lining intelligent maintenance management method and system, it is related to industrial equipment maintenance technical field, method includes: obtaining the current production process parameter and current sensing data of ball mill;According to the current production process parameter and the current sensing data, target data confidence is calculated;According to the target data confidence, maintenance plan adjustment is carried out to rubber lining, and the maintenance plan adjustment includes remaining service life adjustment and priority adjustment;According to the target data confidence, inventory management strategy adjustment is carried out to rubber lining, and the inventory management strategy adjustment includes safety inventory amount adjustment and procurement strategy adjustment.The application can calculate target data confidence in combination with current production process parameter and current sensing data, and carry out maintenance plan adjustment and inventory management strategy adjustment, to realize rubber lining intelligent maintenance management, improve the accuracy of rubber lining wear identification and maintenance management reliability.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment maintenance technology, and in particular to an intelligent maintenance management method and system for rubber liners. Background Technology

[0002] In heavy industries such as mineral processing and cement production, ball mills, as core grinding equipment, play a crucial role in protecting the metal cylinder from direct impact and wear by the grinding media through their internal rubber liners. With the development of intelligent technologies, existing maintenance systems collect real-time equipment operating status data by deploying vibration and acoustic sensors on the outer surface of the cylinder. This data, combined with production process parameters such as mill speed and material flow rate, is used to construct wear prediction models to assess the remaining lifespan of the liners and formulate maintenance plans. However, in the long-term, complex industrial environment, this maintenance model, reliant on sensor data quality, has significant hidden dangers. Due to the long-term accumulation of fine dust, localized high-temperature radiation, and continuous mechanical vibration on the surface of the ball mill cylinder, the sensors installed on it are highly susceptible to performance degradation, manifesting as a slow drift in the amplitude or frequency response of the output signal, rather than a complete failure. Existing sensor health monitoring modules typically only have the ability to identify obvious anomalies such as open circuits, short circuits, or values ​​exceeding limits. They are unable to identify gradual signal drift and accurately identify the wear of rubber liners. This leads to local perforation of the liner when the actual wear far exceeds the predicted value, resulting in damage to the cylinder steel plate, unplanned downtime, and high maintenance costs. The accuracy of rubber liner wear identification is low, and the reliability of maintenance management is low.

[0003] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the Invention

[0004] The main objective of this invention is to propose an intelligent maintenance management method and system for rubber liners. This system can calculate the confidence level of target data by combining current production process parameters and current sensor data, and adjust maintenance plans and inventory management strategies to achieve intelligent maintenance management of rubber liners, thereby improving the accuracy of rubber liner wear identification and the reliability of maintenance management.

[0005] On one hand, embodiments of the present invention provide an intelligent maintenance and management method for rubber liners, comprising the following steps: The current production process parameters and current sensing data of the ball mill are obtained. The current production process parameters include the mill speed and material flow rate, and the current sensing data includes vibration sensing data and acoustic sensing data. Based on the current production process parameters and the current sensing data, the target data confidence level is calculated, which reflects the measurement accuracy of the sensor. Based on the target data confidence level, the maintenance plan for the rubber liner is adjusted, including adjustments to the remaining service life and priorities. Based on the target data confidence level, the inventory management strategy for rubber liners is adjusted, including adjustments to safety stock levels and procurement strategies.

[0006] In some embodiments, calculating the target data confidence level based on the current production process parameters and the current sensor data includes: Obtain the initial data confidence level from the sensor; Based on the current production process parameters and the current sensing data, a first deviation is calculated, which reflects the measurement deviation caused by sensor aging. Based on the current sensing data, a second deviation is calculated, which is used to reflect the correlation between data from different types of sensors; Based on the current production process parameters and the current sensing data, a third deviation is calculated, which reflects the sensor's response deviation to changes in the production process parameters. The target deviation is obtained by averaging the first deviation, the second deviation, and the third deviation. The target data confidence level is calculated based on the initial data confidence level and the target deviation.

[0007] In some embodiments, calculating the first deviation based on the current production process parameters and the current sensing data includes: Acquire the first historical sensor data under the current production process parameters; Based on the first historical sensing data, calculate the historical average and historical standard deviation; The first deviation is calculated based on the historical average, historical standard deviation, and current sensing data.

[0008] In some embodiments, calculating the second deviation based on the current sensing data includes: Based on the vibration sensing data, extract the frequency components of the vibration signal; Based on the acoustic sensing data, extract the acoustic signal energy; Calculate the Pearson correlation coefficient based on the frequency components of the vibration signal and the energy of the acoustic signal; The second deviation is calculated based on the Pearson correlation coefficient and the preset coefficient threshold.

[0009] In some embodiments, calculating the third deviation based on the current production process parameters and the current sensing data includes: Obtain historical production process parameters; Acquire second historical sensor data under the historical production process parameters; Calculate the change in the current sensing data based on the second historical sensing data and the current sensing data; Calculate the change in production process parameters based on the historical production process parameters and the current production process parameters; Based on the changes in the production process parameters, the target changes in sensor data are determined by querying the mapping table between the changes in production process parameters and the changes in sensor data. The third deviation is calculated based on the change in the current sensing data and the change in the target sensing data.

[0010] In some embodiments, the step of averaging the first deviation, the second deviation, and the third deviation to obtain the target deviation includes: An ultrasonic pulse is applied to the physical connection layer between the vibration sensing device and the ball mill cylinder. The vibration sensing device is disposed on the surface of the ball mill cylinder. The physical connection layer includes a fixing adhesive or a mounting base. The first echo signal generated by the reflection of the ultrasonic pulse through the physical connection layer is acquired. Extract the first echo features of the first echo signal, which include peak amplitude, total energy, propagation time and spectral composition. The first echo feature is compared with the reference echo feature to calculate the fourth deviation. The fourth deviation is used to reflect the tightness of the physical connection between the vibration sensing device and the ball mill cylinder. The reference echo feature includes the peak amplitude reference value, the total energy reference value, the propagation time reference value, and the reference spectrum. The target deviation is obtained by averaging the first deviation, the second deviation, the third deviation, and the fourth deviation.

[0011] In some embodiments, the step of averaging the first deviation, the second deviation, and the third deviation to obtain the target deviation includes: A broadband ultrasonic pulse is applied to a rubber liner, and the second echo signal generated by the reflection of the broadband ultrasonic pulse through the inside of the rubber liner is acquired. Extract the second echo feature of the second echo signal, the second echo feature including the current resonant frequency and the peak half-width at half-maximum, the peak half-width at half-maximum being used to represent the width from the highest point of the peak to half the height of the peak; The current quality factor is calculated based on the current resonant frequency and the peak half-width at half-maximum. The current quality factor is used to reflect the energy attenuation in the damped vibration system. Calculate the resonant frequency drift based on the current resonant frequency and the reference resonant frequency; Calculate the damping variation based on the current quality factor and the reference quality factor; Based on the resonant frequency drift and the damping variation, a fifth deviation is calculated, which reflects the degree of microstructural change in the internal material of the rubber liner caused by fatigue accumulation. The target deviation is obtained by averaging the first deviation, the second deviation, the third deviation, and the fifth deviation.

[0012] In some embodiments, when the maintenance plan is adjusted to a remaining service life adjustment, adjusting the maintenance plan for the rubber liner based on the target data confidence level includes: Obtain the initial data confidence level and remaining lifespan of the sensor; Calculate the confidence ratio based on the initial data confidence level and the target data confidence level; The service life reduction factor is calculated based on the confidence ratio and the preset maximum reduction ratio. The remaining service life is adjusted according to the service life reduction factor.

[0013] In some embodiments, when the inventory management strategy is adjusted to a safety stock level adjustment, adjusting the inventory management strategy for the rubber liner based on the target data confidence level includes: Obtain the current available storage space, the initial data confidence level of the sensors, and the safety stock level of the rubber liner; Calculate the confidence decay coefficient based on the initial data confidence level and the target data confidence level; Based on the safety stock level, the confidence decay coefficient, and the inventory risk coefficient, the inventory adjustment amount is calculated. The inventory risk coefficient is used to reflect the degree of impact of sensor data uncertainty on inventory demand. The safety stock level is adjusted based on the inventory adjustment amount and the currently available storage space.

[0014] On the other hand, embodiments of the present invention provide an intelligent maintenance management system for rubber liners, comprising: The data acquisition module is used to acquire the current production process parameters and current sensing data of the ball mill. The current production process parameters include the mill speed and material flow rate, and the current sensing data includes vibration sensing data and acoustic sensing data. The data confidence calculation module is used to calculate the target data confidence based on the current production process parameters and the current sensing data. The target data confidence reflects the measurement accuracy of the sensor. The maintenance plan adjustment module is used to adjust the maintenance plan of the rubber liner according to the target data confidence level. The maintenance plan adjustment includes remaining service life adjustment and priority adjustment. The inventory management strategy adjustment module is used to adjust the inventory management strategy for rubber liners based on the target data confidence level. The inventory management strategy adjustment includes safety stock adjustment and procurement strategy adjustment.

[0015] The embodiments of this application include at least the following beneficial effects: First, the current production process parameters and current sensor data of the ball mill are obtained. Then, based on the current production process parameters and current sensor data, the target data confidence level is calculated. Then, based on the target data confidence level, the maintenance plan and inventory management strategy of the rubber liner are adjusted. Thus, the target data confidence level can be calculated by combining the current production process parameters and current sensor data, and the maintenance plan and inventory management strategy can be adjusted to achieve intelligent maintenance management of the rubber liner, thereby improving the accuracy of rubber liner wear identification and the reliability of maintenance management.

[0016] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description and the drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0018] Figure 1 This is a flowchart of an intelligent maintenance and management method for rubber liners according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an intelligent maintenance and management system for rubber lining plates according to an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments.

[0020] In related technologies, in large-scale industrial production environments, such as ball mills in mineral processing plants or vertical mills in cement plants, rubber liners are critical components protecting the equipment cylinder from wear and impact. These liners, as consumables, require regular inspection and replacement. Traditional maintenance management typically relies on experienced technicians conducting regular inspections, judging the wear condition of the liners through visual inspection and sound tapping, and performing planned shutdowns for replacement according to preset replacement cycles. This approach has significant limitations: firstly, manual judgment is highly subjective, easily leading to misjudgments of the actual lifespan of the liners, potentially resulting in premature replacement and resource waste, or delayed replacement leading to liner failure and equipment damage; secondly, spare parts inventory management often faces challenges, with excessive spare parts tying up capital and space, while insufficient spare parts may result in untimely supply during emergencies, causing prolonged production line shutdowns. To address these issues, existing systems collect real-time equipment operating status data by installing various sensors externally, such as vibration sensors, acoustic sensors, and temperature sensors. Simultaneously, the system also integrates production process parameters, such as mill speed, material flow rate, and material hardness. These data are fed into a predictive model trained on historical operating data. This model analyzes the wear trend of the liner, predicts its remaining service life, and automatically generates maintenance recommendations and spare parts procurement plans based on the prediction results. In the initial stage of system operation, due to stable production conditions and minimal changes in material characteristics, the system's prediction results closely match the actual wear situation, reducing unplanned downtime, optimizing spare parts inventory, and improving equipment operating efficiency and the level of precision in maintenance management.

[0021] However, with prolonged system operation and the equipment operating in a complex industrial environment, some subtle changes began to appear. Specifically, in a particular area outside the ball mill cylinder, near the discharge end, several sets of vibration and acoustic sensors installed in this area began to exhibit slight signal drift due to long-term exposure to fine dust accumulation, localized high temperatures, and slight mechanical stress. For example, this drift did not indicate complete sensor failure, but rather a slow, gradual deviation in the amplitude or frequency response of their output signals. For instance, with the actual vibration intensity remaining constant, the vibration signal output by the sensors might be slightly lower, or the response within a specific frequency range might become sluggish.

[0022] Existing integrated sensor health monitoring modules often focus on detecting "hard" faults such as complete sensor disconnection or abnormally high or low output values, or identifying "soft" faults by comparing significant differences with other sensors. For slow, gradual, and small-amplitude localized sensor signal drift, the system struggles to recognize it as an anomaly. Furthermore, when processing multi-channel sensor data, systems often employ data fusion or averaging to improve robustness. This dilutes or masks minor drifts in localized areas with overall normal data, failing to trigger system alarms.

[0023] Meanwhile, the wear pattern inside a ball mill is not always uniform. For example, due to the complexity of material flow, differences in the trajectory of steel balls, and potential microscopic inhomogeneities in the liner material itself, liner wear can be faster in certain specific areas than in others. The area where the sensor signal drifts corresponds precisely to a liner area within the ball mill where wear is relatively rapid. Due to the sensor signal deviation, the vibration and acoustic information received by the system about this area is misinterpreted. For instance, when the liner wear in this area actually accelerates, certain specific frequencies of vibration or sound signals should be amplified, but due to sensor drift, these amplified signals are weakened or distorted, causing the system to fail to accurately capture the true extent of accelerated wear in that area.

[0024] Therefore, when the existing system predicts the overall wear of the liner, the implicit bias in the input data for key local areas leads to a significant deviation in its wear prediction for that specific area. The system still reports that the overall condition of the liner is good, with sufficient remaining lifespan and no need for advance maintenance. After several days of continuous operation, one rubber liner in that specific area has become extremely thin due to actual wear far exceeding the system's prediction. Under a normal impact load, a local perforation suddenly occurs in the weak area of ​​this liner, causing the grinding media (steel balls and ore) inside the mill to directly contact the mill's steel plate. For example, this local perforation still has serious consequences: the high-speed steel balls directly impact and rub against the mill's steel plate, quickly creating a dent in the mill and potentially causing localized overheating. Furthermore, the perforation may cause the grinding media to become stuck, further aggravating local wear and even affecting the material flow and the balance of steel ball movement inside the mill.

[0025] After the emergency shutdown, maintenance personnel discovered that in addition to replacing the perforated liner and the surrounding damaged liner, the ball mill cylinder steel plate also suffered localized damage, requiring specialized welding repair. For example, although the accident did not result in catastrophic damage to core components such as the main bearing, the complexity and time-consuming nature of the localized cylinder repair, as well as the disruption to production plans, still caused several days of unplanned downtime and high maintenance costs. Subsequent analysis revealed that the system's predictive model itself was not flawed; the problem lay in the quality of the input data. Specifically, some sensors experienced subtle signal drift during long-term operation, and the system struggled to effectively identify this implicit data quality issue.

[0026] The embodiments of this application will be explained in detail below with reference to the accompanying drawings: Figure 1 This is an optional flowchart of an intelligent maintenance and management method for rubber liners provided in an embodiment of this application. Figure 1 The method may include, but is not limited to, steps S101 to S104.

[0027] Step S101: Obtain the current production process parameters and current sensing data of the ball mill. The current production process parameters include the mill speed and material flow rate, and the current sensing data includes vibration sensing data and acoustic sensing data. Step S102: Calculate the target data confidence level based on the current production process parameters and current sensor data. The target data confidence level is used to reflect the measurement accuracy of the sensor. Step S103: Adjust the maintenance plan for the rubber liner based on the target data confidence level. The maintenance plan adjustment includes adjustment of remaining service life and priority adjustment. Step S104: Adjust the inventory management strategy for rubber liners based on the target data confidence level. The inventory management strategy adjustment includes adjusting the safety stock level and adjusting the procurement strategy.

[0028] Steps S101 to S104 as shown in the embodiments of this application can calculate the confidence level of the target data by combining the current production process parameters and the current sensor data, and adjust the maintenance plan and inventory management strategy to realize intelligent maintenance management of rubber liners, thereby improving the accuracy of rubber liner wear identification and the reliability of maintenance management.

[0029] In some embodiments, steps S101-S104 can first acquire the current production process parameters and current sensing data of the ball mill. The current production process parameters include the mill speed and material flow rate, and the current sensing data includes vibration sensing data and acoustic sensing data. The acquisition of production process parameters can be achieved in various ways. For example, a speed encoder signal acquisition module can be installed in the inverter control cabinet of the ball mill drive motor. The motor speed pulse signal is read through an opto-isolation circuit, and the actual speed of the mill cylinder is calculated, typically ranging from 10 to 20 revolutions per minute. Specifically, an incremental photoelectric encoder generating 1024 pulses per revolution can be installed at the motor shaft end, and a high-speed counter module can be used to count the pulses. The real-time speed is obtained from the count per second, with a resolution of 0.01 revolutions per minute. The material flow rate can be acquired by installing a nuclear belt scale or electronic belt scale on the feed belt, collecting the mass of ore passing through the belt per unit time in real time, typically in tons per hour, with a value ranging from 100 to 300 tons. The electronic belt scale detects the load per unit length of the belt using strain gauge load cells, calculates the instantaneous flow rate using a speed encoder, and outputs a standard current signal of 4 to 20 milliamps, which is then acquired and converted into a digital quantity by an analog input module.

[0030] For sensor data acquisition, vibration sensing data can be obtained through piezoelectric accelerometers mounted on the outer wall of the ball mill cylinder. Industrial-grade accelerometers with a sensitivity of 100 mV / g can be used. The sensors are fixed to the cylinder surface via magnetic bases or adhesive bonding to capture the high-frequency vibration characteristics generated by the impact of steel balls. During data acquisition, the analog signal is first filtered by an anti-aliasing filter, then digitized by a 16-bit analog-to-digital converter. The data is transmitted to the industrial computer's memory buffer via a PCI-E bus or Ethernet. Acoustic sensing data can be acquired through an array of acoustic sensors installed within a 0.5-1 meter range of the cylinder. Capacitive microphones or piezoelectric acoustic emission sensors can be used to capture friction noise and impact sounds generated by liner wear. Acoustic signal acquisition also requires pre-amplification and anti-aliasing filtering to ensure signal quality.

[0031] Then, based on the current production process parameters and current sensor data, the target data confidence score is calculated. This score reflects the accuracy of the sensor measurements. A confidence score evaluation method based on historical statistics can be used. The system maintains a sliding time window, storing sensor data under normal operating conditions over the past 30 days, and calculates the mean and standard deviation of this dataset as a benchmark. The deviation of the current sensor data from this benchmark is quantified using Mahalanobis distance or standardized residuals; the greater the deviation, the lower the confidence score. For example, when the current mean of the vibration signal deviates from the historical benchmark by more than two standard deviations, the confidence score coefficient decreases from the initial value of 1.0 to 0.8. Alternatively, a multi-sensor cross-validation mechanism can be introduced. The system simultaneously collects data from multiple vibration sensors at adjacent locations and calculates the Pearson correlation coefficient within the same time window. When the output of a sensor is significantly inconsistent with other sensors, its confidence score weight is reduced. For example, using the Pearson correlation coefficient to calculate the correlation of vibration data from three adjacent measuring points, if the correlation coefficient is below 0.7, the sensor is considered to be drifting, and its confidence score is lowered to 0.6.

[0032] Based on the confidence level of the target data, the maintenance plan for the rubber liners is adjusted. This adjustment includes adjustments to remaining service life and priority adjustments. Regarding the remaining service life adjustment, a mapping relationship can be established between confidence level and service life reduction factor. When the confidence level of the target data is higher than 0.9, the original service life prediction remains unchanged; when the confidence level is between 0.7 and 0.9, the predicted service life is shortened linearly, for example, a confidence level of 0.8 corresponds to a 10% service life reduction; when the confidence level is lower than 0.7, a conservative estimate is used, and preventative maintenance instructions are triggered directly. Regarding priority adjustments, the system maintains a queue of liners to be maintained, sorted according to the remaining service life value after confidence level adjustment. Liners with smaller service life values ​​have higher maintenance priority, ensuring that areas with poor data quality are inspected first.

[0033] Finally, based on the target data confidence level, the inventory management strategy for rubber liners is adjusted. This adjustment includes adjusting safety stock levels and procurement strategies. For safety stock adjustments, buffer stock can be dynamically increased based on the degree of confidence decay. When the sensor confidence level in a certain area consistently falls below a threshold, the system automatically increases the safety stock of the corresponding specification liners in that area from the usual two sets to three or four sets to cope with potential sudden early replacement needs. Regarding procurement strategy adjustments, when the confidence level falls below 0.8, the system switches from the planned procurement mode to an emergency procurement mode, shortening the supplier delivery cycle from the usual 30-day delivery period to 15 days, while also increasing alternative supplier options to ensure timely spare parts supply.

[0034] Through the above technical solution, this embodiment can identify the implicit quality degradation of sensor data through software algorithms without increasing additional hardware costs, avoid maintenance decision errors caused by data drift, and significantly improve the safety and reliability of maintenance plans.

[0035] In some embodiments, step S102, calculating the target data confidence level based on the current production process parameters and current sensor data, may include, but is not limited to, the following steps: Step S201: Obtain the initial data confidence level of the sensor; Step S202: Calculate the first deviation based on the current production process parameters and current sensing data. The first deviation is used to reflect the measurement deviation caused by sensor aging. Step S203: Calculate the second deviation based on the current sensing data. The second deviation is used to reflect the correlation between data from different types of sensors. Step S204: Calculate the third deviation based on the current production process parameters and current sensing data. The third deviation is used to reflect the sensor's response deviation to changes in production process parameters. Step S205: Calculate the target deviation by averaging the first deviation, the second deviation, and the third deviation. Step S206: Calculate the target data confidence level based on the initial data confidence level and the target deviation level.

[0036] In some embodiments, the initial data confidence level of the sensor can be obtained first. The initial data confidence level can be preset, for example, the initial data confidence level of the sensor at the time of manufacture is 1.0. Based on the current production process parameters and current sensing data, a first deviation is calculated. The first deviation reflects the measurement deviation caused by sensor aging. In engineering practice, sensors exhibit aging phenomena such as zero-point drift and decreased sensitivity as operating time increases. The calculation of the first deviation focuses on the degradation of the sensor's own performance. A time series analysis-based method can be used to establish a regression model between the sensor output and operating time, and residual analysis can be used to identify aging trends. For example, linear fitting can be performed on the data of a vibration sensor over the past six months. If a monotonically decreasing trend in output amplitude is found and the slope exceeds a threshold, a higher first deviation value is calculated.

[0037] Then, based on the current sensing data, a second deviation is calculated. This second deviation reflects the correlation between data from different types of sensors. The calculation of the second deviation is based on the principle of physical correlation. Since both vibration and acoustic signals originate from the impact of the steel ball on the liner, they should be correlated in the frequency domain. The energy proportion of the 1kHz to 5kHz frequency band in the vibration signal can be extracted, and the sound pressure level of the acoustic signal in the same frequency band can be calculated. The consistency between the two is quantified using the Pearson correlation coefficient. When the correlation coefficient is below 0.6, at least one type of sensor is considered abnormal, and a higher second deviation value is assigned.

[0038] Then, based on the current production process parameters and current sensor data, the third deviation is calculated. The third deviation reflects the sensor's response deviation to changes in production process parameters. The calculation of the third deviation focuses on the sensor's sensitivity to changes in operating conditions. Changes in ball mill speed or material flow rate should cause corresponding changes in sensor data. A mapping table between operating parameters and sensor data can be established to record the normal sensor data range under different combinations of speed and flow rate. When the actual change in sensor data does not match the expected change, the third deviation is calculated. For example, if the vibration amplitude does not increase by the expected 15% to 20% when the speed increases from 15 rpm to 18 rpm, the sensor is considered to be sluggish.

[0039] Finally, the target deviation is obtained by averaging the first, second, and third deviations. After obtaining the three deviations, a weighted average or arithmetic average can be used to calculate the target deviation. For example, the first deviation can be weighted at 0.4, the second at 0.3, and the third at 0.3, and the weighted sum can yield a target deviation value between 0 and 1. The initial data confidence level is typically set to 1.0 or, according to the sensor's factory calibration certificate, a value between 0.95 and 1.0. The target data confidence level is then calculated based on the initial data confidence level and the target deviation. The target data confidence level can be obtained by subtracting the target deviation from the initial data confidence level, ensuring that the larger the deviation, the more significant the decrease in confidence.

[0040] Through the above technical solution, this embodiment can effectively distinguish between anomalies of different natures such as sensor aging, data inconsistency, and slow response, avoid misjudgment based on a single indicator, and improve the accuracy and robustness of data quality assessment.

[0041] In some embodiments, in step S202, calculating the first deviation based on the current production process parameters and current sensing data may include, but is not limited to, the following steps: Acquire the first historical sensor data under the current production process parameters; Calculate the historical average and historical standard deviation based on the first historical sensor data; The first deviation is calculated based on the historical average, historical standard deviation, and current sensor data.

[0042] In some embodiments, the first historical sensor data under the current production process parameters can be acquired first. In specific implementations, the selection of historical data needs to ensure consistency of operating conditions. The system maintains an operating condition classification database, dividing the mill speed into multiple intervals such as 10-12 rpm, 12-15 rpm, and 15-18 rpm, and the material flow rate into intervals such as 100-150 tons per hour and 150-200 tons per hour. When it is necessary to calculate the first deviation, the system extracts historical sensor data belonging to the same operating condition interval within the past thirty days from the database based on the current speed and flow rate values ​​as the first historical sensor data. The database construction process is as follows: a two-dimensional operating condition grid is established, with the horizontal axis representing the speed interval and the vertical axis representing the flow rate interval, and each grid cell corresponding to a specific operating condition state; the system continuously records operating data, and classifies the sensor data at each sampling moment into the corresponding grid cell according to its corresponding speed and flow rate values; a circular buffer is maintained within each cell to store the data of the most recent thirty days, and when the data volume exceeds the storage limit, the earliest data is discarded using the first-in-first-out principle.

[0043] Then, based on the first historical sensor data, calculate the historical mean and historical standard deviation. The historical mean can be obtained by averaging the first historical sensor data using the arithmetic mean method. The historical standard deviation can be calculated using the Bessel correction formula to ensure the unbiasedness of the statistic.

[0044] Then, based on the historical average, historical standard deviation, and current sensor data, calculate the first deviation. The first deviation can be calculated using the standardized residual method. The specific calculation formula is as follows: In the formula, The first deviation degree, For the current sensor data, This is the historical average. This represents the historical standard deviation.

[0045] Through the above technical solution, this embodiment can effectively identify the gradual drift of the sensor output. Even if the drift amplitude is small in the early stage, as long as it continues to deviate from the historical benchmark, it can be detected and quantified in time, providing a reliable basis for subsequent confidence adjustment.

[0046] In some embodiments, in step S203, calculating the second deviation based on the current sensing data may include, but is not limited to, the following steps: Based on vibration sensing data, extract the frequency components of the vibration signal; Extract acoustic signal energy from acoustic sensor data; The Pearson correlation coefficient is calculated based on the frequency components of the vibration signal and the energy of the acoustic signal. The second deviation is calculated based on the Pearson correlation coefficient and the preset coefficient threshold.

[0047] In some embodiments, the frequency components of the vibration signal can be extracted first from the vibration sensing data. At the signal processing level, the extraction of the frequency components of the vibration signal can employ Fast Fourier Transform (FFT) or Wavelet Transform (WFT) methods. FFT can be performed on the vibration sensing data to extract the peak frequencies and their corresponding amplitudes in the 1 kHz to 8 kHz range of the spectrum, with a focus on the characteristic frequencies generated by the steel ball impacting the liner. The specific processing flow is as follows: a Hanning window is applied to the acquired time-domain vibration signal to reduce spectral leakage; the window length is 1024 points with an overlap rate of 50%; a FFT is performed to obtain the spectrum; then, the peak values ​​within the 1 kHz to 8 kHz frequency band are located in the spectrum, and the total energy of this frequency band is calculated as the characteristic value of the vibration signal frequency component.

[0048] Then, based on the acoustic sensing data, the acoustic signal energy is extracted. The band energy integration method can be used to calculate the sound pressure level of the acoustic signal within the same frequency band, which is then taken as the acoustic signal energy. Specifically, after windowing and fast Fourier transform processing of the acoustic signal, the sum of squares of the amplitudes of each frequency component within the 1kHz to 8kHz frequency band is calculated, then divided by the number of frequency points to obtain the average energy, and finally, the sound pressure level is obtained through a 20x logarithmic transformation.

[0049] Then, based on the frequency components of the vibration signal and the energy of the acoustic signal, the Pearson correlation coefficient is calculated. The Pearson correlation coefficient is calculated based on aligned time-series data. A time window can be set, such as 10 seconds or 30 seconds. Based on the frequency components of the vibration signal and the energy of the acoustic signal, the vibration frequency band energy sequence and sound pressure level sequence within this window are extracted, and the covariance of the two is calculated divided by the product of their respective standard deviations. The specific calculation formula is as follows: In the formula, The Pearson correlation coefficient is used. The total number of samples within the time window. This represents the i-th sample value of the vibration frequency band energy sequence. The mean of the vibration frequency band energy sequence. For the i-th sample value in the sound pressure level sequence, The mean of the sound pressure level sequence. The standard deviation of the vibration frequency band energy sequence. This represents the standard deviation of the sound pressure level sequence. The Pearson correlation coefficient ranges from -1 to +1, with positive values ​​indicating positive correlation and negative values ​​indicating negative correlation. Under normal operating conditions, vibration and acoustic signals should exhibit a strong positive correlation, and the Pearson correlation coefficient is typically higher than 0.8.

[0050] Finally, the second deviation is calculated based on the Pearson correlation coefficient and a preset threshold. The preset threshold can be dynamically adjusted according to the equipment's operating phase. The preset threshold is determined based on historical statistical data from normal operation: the correlation coefficients of vibration and acoustic signals during the past three months of normal operation are collected, and the 5th percentile of their distribution is calculated as the lower limit threshold, typically falling between 0.7 and 0.75. A baseline threshold of 0.7 can be set. When the Pearson correlation coefficient is higher than 0.8, the second deviation is 0; when the Pearson correlation coefficient is between 0.7 and 0.8, the second deviation is equal to 0.8 minus the Pearson correlation coefficient divided by 0.1; when the correlation coefficient is lower than 0.7, the second deviation is 1. Alternatively, time lag compensation can be introduced. Considering that the speed of sound propagation is lower than the speed of vibration propagation, the acoustic signal is time-shifted before calculating the correlation to ensure the accuracy of the correlation calculation. Specifically, the time delay corresponding to the maximum cross-correlation coefficient between vibration and acoustic signals is calculated using a cross-correlation function, and then the acoustic signal is advanced by this time delay to eliminate the phase difference caused by the difference in propagation paths.

[0051] Through the above technical solution, this embodiment can effectively identify local faults or drift of a single type of sensor by monitoring the correlation between vibration and acoustic signals. When the vibration sensor drifts while the acoustic sensor is normal, the correlation between the two decreases, triggering a second deviation increase, prompting maintenance personnel to check specific sensors and avoiding incorrect judgments based on a single data source.

[0052] In some embodiments, in step S204, calculating the third deviation based on the current production process parameters and current sensing data may include, but is not limited to, the following steps: Obtain historical production process parameters; Acquire second historical sensor data under historical production process parameters; Calculate the change in current sensor data based on the second historical sensor data and the current sensor data; Calculate the changes in production process parameters based on historical and current production process parameters; Based on the changes in production process parameters, the target changes in sensor data are determined by querying the mapping table between the changes in production process parameters and the changes in sensor data. The third deviation is calculated based on the change in current sensor data and the change in target sensor data.

[0053] In some embodiments, historical production process parameters can be acquired first. The acquisition time point for historical production process parameters is typically selected as the moment of the last significant adjustment of operating conditions or the recording point of the previous shift. For example, the rotational speed was 15 revolutions per minute at the previous moment, and it has now been increased to 18 revolutions per minute; the flow rate was 200 tons per hour at the previous moment, and it has now been reduced to 180 tons per hour. The system records the points of change in operating conditions through an event-triggered mechanism. When the change in rotational speed exceeds one revolution per minute or the change in flow rate exceeds ten tons per hour, that moment is automatically marked as a historical reference point.

[0054] Then, acquire the second historical sensor data under the historical production process parameters. Based on the second historical sensor data and the current sensor data, calculate the change in the current sensor data. The change in the current sensor data can be obtained by subtracting the value of the second historical sensor data from the value of the current sensor data. Simultaneously, calculate the change in the production process parameters based on the historical production process parameters and the current production process parameters. The change in the production process parameters can be obtained by subtracting the values ​​of the historical production process parameters from the values ​​of the current production process parameters.

[0055] Then, based on the changes in production process parameters, the target sensor data change is determined by querying the mapping table between changes in production process parameters and changes in sensor data. The mapping table is constructed based on historical data statistics. The specific process is as follows: collect the ball mill's operating data over the past three months and record the speed change before and after each adjustment of operating conditions. Flow rate change and the corresponding change in vibration amplitude. Change in sound pressure level The coefficient matrix was obtained by fitting the data using multiple linear regression, and a system was established. and The relation, in which These are the regression coefficients obtained through least squares fitting; these coefficients are stored as a lookup table or analyzed using analytical expressions for real-time calculation. The mapping relationship is stored in system memory as a lookup table, supporting linear interpolation. For example, when the actual rotational speed changes by two revolutions per minute, if the table only shows data corresponding to a 5% vibration change at one revolution per minute and a 15% vibration change at three revolutions per minute, then the target vibration change corresponding to two revolutions per minute is determined to be 10% through linear interpolation. The target sensor data change is determined by looking up the theoretical expected change value from the mapping table based on the actual changes in production process parameters. For example, an increase of three revolutions per minute in rotational speed is expected to increase the vibration amplitude by 18%; if the actual vibration amplitude only increases by 5%, there is a significant response deviation.

[0056] Finally, the third deviation is calculated based on the changes in the current and target sensor data. This can be achieved by subtracting the absolute value of the change in the target sensor data from the change in the current sensor data, and then dividing by the change in the target sensor data. Furthermore, response delay compensation can be introduced. Considering the inertia of sensor responses, a certain delay is allowed before comparing the changes. Specifically, an observation window is set, for example, waiting 30 to 60 seconds after a change in operating conditions, until the system enters a new steady state before collecting the current sensor data, thus avoiding misjudgments caused by transient processes.

[0057] Through the above technical solution, this embodiment can effectively identify the problems of sensor sensitivity degradation and slow dynamic response, ensuring that the sensor can still accurately track changes in equipment status in a production environment where the mill operating conditions are frequently adjusted, and providing reliable dynamic data support for wear prediction.

[0058] In some embodiments, in step S205, the target deviation is obtained by averaging the first deviation, the second deviation, and the third deviation. This may include, but is not limited to, the following steps: An ultrasonic pulse is applied to the physical connection layer between the vibration sensing device and the ball mill cylinder. The vibration sensing device is set on the surface of the ball mill cylinder. The physical connection layer includes a fixing adhesive or a mounting base. Acquire the first echo signal generated by the reflection of the ultrasonic pulse through the physical connection layer; Extract the first echo features of the first echo signal. The first echo features include peak amplitude, total energy, propagation time and spectral composition. The first echo feature is compared with the reference echo feature to calculate the fourth deviation. The fourth deviation is used to reflect the tightness of the physical connection between the vibration sensing device and the ball mill cylinder. The reference echo features include the peak amplitude reference value, the total energy reference value, the propagation time reference value, and the reference spectrum. The target deviation is obtained by averaging the first deviation, second deviation, third deviation, and fourth deviation.

[0059] In some embodiments, sensor measurement deviations may stem not only from aging electronic components or algorithmic lag but also from deterioration of mechanical connections. Specifically, vibration sensing devices are fixed to the outer wall of the ball mill cylinder via adhesive or bolt bases, enduring long-term severe vibrations ranging from a few hertz to several kilohertz, high-temperature cycling at 80 to 120 degrees Celsius, and erosion from corrosive slurry vapors. The adhesives may crack or detach, and the mounting bases may experience bolt loosening or oxidation of contact surfaces. This degradation of physical connections leads to attenuation or distortion of the vibration transmission path, causing the mechanical vibration signal received by the sensor to be distorted at its source. Such mechanical deviations cannot be effectively identified using the aforementioned electrical signal analysis methods.

[0060] To address this, ultrasonic pulses can be applied to the physical connection layer between the vibration sensor and the ball mill cylinder. The vibration sensor is mounted on the surface of the ball mill cylinder, and the physical connection layer includes a fixing adhesive or mounting base. In practical applications, applying ultrasonic pulses to the physical connection layer requires a dedicated ultrasonic pulse generation and receiving system. A digital ultrasonic flaw detector can be used to apply the ultrasonic pulses. Its built-in pulse transmitting circuit can generate narrowband ultrasonic pulses with a center frequency of 2.5 MHz or 5 MHz, a pulse repetition frequency of 500 Hz to 2000 Hz, a pulse width of 0.5 microseconds to 2 microseconds, and a voltage amplitude of 100 V to 400 V. A piezoelectric transducer with a diameter of 10 mm or 20 mm and a crystal material of lead zirconate titanate ceramic is used. It is fixed to the cylinder surface 10 mm to 20 mm beside the sensor mounting base using magnetic clamps or spring clamping devices to ensure that the sound waves are incident perpendicularly to the connection layer interface. During regular maintenance, the probe is coupled to the top of the sensor housing, and the status of the connection layer is indirectly detected through the metal housing of the sensor. Meanwhile, considering the oxide scale and rough texture that typically exist on the surface of the ball mill cylinder, industrial ultrasonic coupling paste or high-temperature silicone grease can be used, with the coupling layer thickness strictly controlled within 0.1 mm to 0.2 mm. It is understood that ultrasonic pulses refer to mechanical vibration waves generated by piezoelectric transducers, with frequencies higher than 20 kHz, capable of propagating in metals and adhesive materials and being reflected at the interface. The physical bonding layer is the critical path for vibration transmission.

[0061] Then, the first echo signal generated by the reflection of the ultrasonic pulse after passing through the physical connection layer is acquired. The acquisition of the first echo signal is completed using the same probe's receiving mode, employing a high-speed data acquisition card with a sampling frequency set to 50 MHz to 100 MHz and a vertical resolution of at least twelve bits to capture the details of the transient waveform. The acquired raw time-domain signal is first processed by bandpass filtering, with the passband range set to one to ten MHz, using a Bessel filter or Butterworth filter of order four to eight to remove high-frequency noise and low-frequency structural vibration interference. It can be understood that the first echo signal refers to the electrical signal returned to the receiver after the emitted ultrasonic pulse is reflected at the physical connection layer interface; its characteristics reflect the acoustic impedance matching state of the interface.

[0062] Next, the first echo characteristics of the first echo signal are extracted. These characteristics include peak amplitude, total energy, propagation time, and spectral composition. Peak amplitude refers to the maximum voltage amplitude of the echo signal, which can be determined by detecting the maximum value of the waveform envelope. Total energy refers to the integrated energy of the echo signal within a certain time window, which can be obtained by numerically integrating the square of the filtered signal within a 50-microsecond time window. Propagation time refers to the time interval from transmission to reception, which can be determined by identifying the time difference between the zero point of the transmitted pulse and the leading edge of the echo pulse, with an accuracy of 0.01 microseconds. Spectral composition refers to the energy distribution of the echo signal in the frequency domain, which can be obtained through Fast Fourier Transform. The Hanning window is selected as the window function, with an analysis bandwidth of 0 to 20 MHz.

[0063] Finally, the first echo characteristic is compared with the reference echo characteristic to calculate the fourth deviation. The fourth deviation reflects the tightness of the physical connection between the vibration sensing device and the ball mill cylinder. The reference echo characteristics include the peak amplitude reference value, the total energy reference value, the propagation time reference value, and the reference spectrum. The establishment of the reference echo characteristics needs to be calibrated when the connection is confirmed to be good during the initial installation of the sensor or after maintenance. The specific calibration process is as follows: After the sensor is newly installed or re-tightened, wait for the adhesive to fully cure (usually 24 to 72 hours), repeat the measurement at the same location more than three times, and take the average value as the reference. For example, when using a 5 MHz probe, an epoxy structural adhesive, and a thickness of 0.5 mm, the following reference values ​​are recorded: the peak amplitude reference value is set to 40% to 60% of the emitted pulse amplitude; the total energy reference value is positively correlated with the peak amplitude reference value and can be determined through a calibration curve; the propagation time reference value depends on the propagation speed of sound waves in the adhesive (the longitudinal wave speed of epoxy materials is approximately 2,500 m / s) and the adhesive thickness; for a thickness of 0.5 mm, the theoretical propagation time is 0.4 microseconds, and the measured value is usually in the range of 0.38 to 0.42 microseconds; the reference spectrum records the amplitude ratio between the dominant frequency of 5.0 MHz and the first harmonic of 10.0 MHz. These reference data are stored in the system's non-volatile memory and are bound to the sensor number. The normalized deviation of each feature can be calculated. The peak amplitude deviation is obtained by subtracting the reference peak amplitude from the actual peak amplitude and then dividing by the reference peak amplitude. The total energy deviation is calculated similarly. The propagation time deviation is obtained by subtracting the actual propagation time from the reference propagation time and then dividing by the reference propagation time. The spectral similarity deviation is obtained by calculating the cross-correlation coefficient between the actual spectrum and the reference spectrum and then subtracting this coefficient from one. Weights are assigned: peak amplitude deviation is weighted at 0.3, total energy deviation at 0.3, propagation time deviation at 0.2, and spectral similarity deviation at 0.2. A weighted sum is then used to obtain the fourth deviation. The target deviation is obtained by averaging the first, second, third, and fourth deviations. The fourth deviation can be calculated using an arithmetic mean method.

[0064] Through the above technical solution, this embodiment can penetrate the sensor shell to directly assess the microscopic state of mechanical connections, identify the physical root causes of vibration transmission distortion such as adhesive aging and base loosening, fill the gap in mechanical structure health assessment by purely electrical monitoring methods, and ensure that confidence assessment covers the entire link from signal source to data output.

[0065] In some embodiments, in step S205, the target deviation is obtained by averaging the first deviation, the second deviation, and the third deviation. This may include, but is not limited to, the following steps: A broadband ultrasonic pulse is applied to the rubber liner, and the second echo signal generated by the reflection of the broadband ultrasonic pulse through the inside of the rubber liner is acquired. Extract the second echo features of the second echo signal. The second echo features include the current resonant frequency and the peak half-width at half-maximum (HWHM). The HWHM is used to represent the width from the highest point of the peak to half the height of the peak. Calculate the current quality factor based on the current resonant frequency and peak half-width at half-maximum. The current quality factor is used to reflect the energy decay in the damped vibration system. Calculate the resonant frequency drift based on the current resonant frequency and the reference resonant frequency; Calculate the damping variation based on the current quality factor and the baseline quality factor; Based on the resonant frequency drift and damping variation, the fifth deviation is calculated. The fifth deviation is used to reflect the degree of microstructural change of the internal material of the rubber liner caused by fatigue accumulation. The target deviation is obtained by averaging the first deviation, second deviation, third deviation, and fifth deviation.

[0066] In some embodiments, a broadband ultrasonic pulse can be applied to the rubber liner first, and the second echo signal generated by the reflection of the broadband ultrasonic pulse through the interior of the rubber liner can be acquired. In engineering implementation, applying a broadband ultrasonic pulse to the rubber liner requires a broadband ultrasonic transducer, such as a probe using polymer piezoelectric material or composite wafers, with a frequency range of 50 kHz to 500 kHz to cover the typical thickness resonant frequency of the rubber liner. A dual-probe arrangement with one transmitter and one receiver can be used, with the transmitting and receiving probes positioned on opposite sides of the outer surface of the liner (opposite to the cylinder contact side), spaced 20 mm to 50 mm apart, ensuring that the received signal includes bulk wave information penetrating the liner. A single-probe pulse-echo mode can also be used, suitable for situations where the outer surface of the liner is accessible. Due to the high acoustic attenuation characteristics of rubber material (attenuation coefficient of approximately 10 to 50 decibels per centimeter per megahertz), the pulse parameters need special optimization. The pulse width is set to 5 microseconds to 20 microseconds, the voltage amplitude to 50 volts to 200 volts, and the pulse repetition frequency to 100 Hz to 500 Hz to ensure sufficient penetration depth and signal-to-noise ratio. The coupling between the probe and the rubber surface requires a special coupling agent, such as high-viscosity silicone oil (viscosity 1000 to 10000 centistokes) or a dedicated rubber ultrasonic coupling gel, to avoid the absorption of traditional water-based coupling agents by the rubber or causing surface swelling. The coupling layer thickness should be controlled within 0.5 mm, and the applied pressure should be 1 to 3 Newtons. The acquisition of the second echo signal requires attention to the characteristics of the resonance peak. When a broadband ultrasonic pulse enters the rubber liner, it reflects multiple times between the upper and lower surfaces, forming standing waves and generating resonance. The acquired time-domain signal is converted into a spectrum using a fast Fourier transform, with a frequency resolution within 100 Hz. Broadband ultrasonic pulses refer to ultrasonic signals with a wide frequency range that can excite multiple resonant modes in the rubber material. The second echo signal refers to the echo sequence formed after the ultrasonic pulse reflects multiple times between the upper and lower surfaces of the rubber liner, containing information about the acoustic characteristics of the material's interior.

[0067] Then, the second echo features of the second echo signal are extracted. These features include the current resonant frequency and the peak half-width at half-maximum (HWHM). The HWHM represents the width of a peak from its highest point to half its height. The current resonant frequency is determined by searching for local maxima in the spectrum, typically focusing on the first-order thickness resonant peak (fundamental frequency) and the second-order resonant peak. The HWHM is measured by finding two frequency points on the determined resonant peak where the amplitude drops to half (i.e., the negative 3 dB point) from the highest point downwards, and calculating the difference between them. It can be understood that the current resonant frequency refers to the characteristic frequency of concentrated energy in the echo signal spectrum, corresponding to the thickness vibration mode of the rubber liner; the HWHM refers to the frequency width of the resonant peak at half its maximum height, reflecting the sharpness of the resonance.

[0068] Next, based on the current resonant frequency and peak half-width at half-maximum (FWHM), the current quality factor is calculated. The current quality factor reflects the energy attenuation in the damped vibration system. The formula for calculating the current quality factor Q is Q = f0 / Δf, where f0 is the current resonant frequency and Δf is the peak FWHM. For newly installed rubber liners (natural rubber or styrene-butadiene rubber matrix with carbon black filler), the typical Q value is between 20 and 50; when internal fatigue or microcrack propagation occurs, the Q value drops below 10, and in severe cases, it falls below 5. It can be understood that the current quality factor Q is an important parameter for measuring the energy loss of a resonant system, defined as the ratio of the resonant frequency to the FWHM. A higher Q value indicates lower energy loss and a more intact material; a lower Q value indicates the presence of scatterers (such as microcracks) within the material, leading to increased energy dissipation.

[0069] Calculate the resonant frequency drift based on the current resonant frequency and the reference resonant frequency. This can be achieved by subtracting the current resonant frequency from the reference resonant frequency and then dividing by the reference resonant frequency. Similarly, calculate the damping variation based on the current quality factor and the reference quality factor. This can also be achieved by subtracting the current quality factor from the reference quality factor and then dividing by the reference quality factor. The determination of the reference resonant frequency and reference quality factor needs to be completed during the initial installation of the liner or through standard sample testing in a laboratory. Specifically, before installing the new liner, a standard sample (100 mm * 100 mm * actual thickness) is cut from the same batch of material and measured using an ultrasonic system with the same parameters under laboratory conditions (temperature 23 degrees Celsius, relative humidity 50%), and the reference resonant frequency and reference quality factor are recorded.

[0070] Finally, based on the resonant frequency drift and damping variation, the fifth deviation is calculated. The fifth deviation reflects the degree of microstructural change caused by fatigue accumulation within the rubber liner. As a comprehensive indicator, the fifth deviation can be calculated using a weighted average method, for example: Fifth Deviation = 0.6 * Resonant Frequency Drift + 0.4 * Damping Variation. The weights can be set according to the actual application. For instance, when the resonant frequency drift exceeds 15% (i.e., drift greater than 0.15) and the quality factor Q decreases by more than 40% (i.e., variation greater than 0.4), it is determined that there is severe fatigue within the liner, and the fifth deviation is close to 1, indicating that the liner needs to be replaced rather than the sensor calibrated. The first, second, third, and fifth deviations are then averaged to obtain the target deviation. This can be calculated using an arithmetic mean method.

[0071] Through the above technical solution, this embodiment can non-destructively detect the microstructure degradation inside the rubber liner, effectively distinguish the material's physical state from the sensor's electrical drift, avoid misjudgment caused by the aging of the liner material, and provide an independent verification dimension derived from the tested object itself for the confidence assessment of target data, significantly improving the fault differentiation capability and decision reliability in complex scenarios where material aging and sensor drift coexist.

[0072] In some embodiments, in step S103, when the maintenance plan is adjusted to a remaining service life adjustment, the maintenance plan for the rubber liner is adjusted according to the target data confidence level, which may include, but is not limited to, the following steps: Obtain the initial data confidence level and remaining lifespan of the sensor; Calculate the confidence ratio based on the initial data confidence level and the target data confidence level; The service life reduction factor is calculated based on the confidence ratio and the preset maximum reduction ratio. Adjust the remaining service life according to the service life reduction factor.

[0073] In some embodiments, the initial data confidence level and remaining service life of the sensor can be obtained first. In practical applications, the initial data confidence level refers to the data reliability benchmark value confirmed by the sensor in its initial calibration state or at the initial stage of installation. It can be set to 1.0, or a value between 0.95 and 1.0 based on the accuracy level provided by the sensor's factory calibration certificate. This value is stored in the system configuration database and bound to the sensor's unique identifier. The remaining service life can be calculated by the wear prediction model based on current sensor data, historical wear trends, and material characteristic parameters, in hours or days, representing the expected duration for which the liner can continue to operate safely under current operating conditions. The wear prediction model can be selected and trained according to actual needs, without specific limitations.

[0074] Then, based on the initial data confidence level and the target data confidence level, the confidence ratio is calculated. The confidence ratio can be obtained by dividing the target data confidence level by the initial data confidence level. This ratio reflects the degree of degradation of the current data quality relative to the initial state of the sensor, and its value ranges from 0 to 1.

[0075] Then, the useful life reduction factor is calculated based on the confidence ratio and the preset maximum reduction ratio. The useful life reduction factor is calculated using a piecewise linear mapping method: when the confidence ratio is greater than or equal to 0.9, the data quality is considered good, and the reduction factor is set to 1, meaning no reduction. When the confidence ratio is between 0.3 and 0.9, the reduction factor = 1 - (1 - confidence ratio) / (1 - 0.3 * preset maximum reduction ratio). When the confidence ratio is less than 0.3, the reduction factor is 1 - preset maximum reduction ratio. The preset maximum reduction ratio is a configuration parameter set according to the enterprise's risk tolerance, representing the upper limit of the predicted useful life allowed to be reduced in the worst-case scenario. A typical value is set to 0.3 to 0.5, meaning a maximum reduction of 30% to 50% of the predicted useful life, to prevent excessive conservatism and wasted maintenance resources.

[0076] Finally, the remaining service life is adjusted based on the service life reduction factor. The service life reduction factor can be multiplied by the remaining service life to obtain the adjusted remaining service life. For example, if the remaining service life of a liner at a certain location is predicted to be 500 hours based on a wear model, the initial data confidence level of the sensor at that location is 1.0, the currently calculated target data confidence level is 0.7, and the system's preset maximum reduction ratio is set to 0.4, then the confidence ratio is 0.7. Substituting this into the formula, the service life reduction factor is approximately 0.828, and the adjusted remaining service life is 414 hours. Based on this, the system advances the maintenance plan from the original 500 hours to 414 hours and increases the inspection priority for that area.

[0077] Through the above technical solution, this embodiment can automatically adopt a more conservative lifetime estimation strategy when the sensor data quality deteriorates by using a dynamic lifetime reduction mechanism based on confidence ratio. This avoids overly optimistic predictions caused by data drift, significantly reduces the risk of perforation failure of the liner when the actual wear exceeds expectations, and prevents unnecessary maintenance cost increases caused by overly conservative approach by setting a maximum reduction ratio.

[0078] In some embodiments, in step S104, when the inventory management strategy is adjusted to safety stock level adjustment, the inventory management strategy for the rubber liner is adjusted according to the target data confidence level, which may include, but is not limited to, the following steps: Obtain the current available storage space, the initial data confidence level of the sensors, and the safety stock level of the rubber liner; Calculate the confidence decay coefficient based on the initial data confidence level and the target data confidence level; The inventory adjustment is calculated based on the safety stock level, confidence decay coefficient, and inventory risk coefficient. The inventory risk coefficient is used to reflect the degree of impact of sensor data uncertainty on inventory demand. Adjust safety stock levels based on inventory adjustments and current available storage space.

[0079] In some embodiments, the currently available storage space, initial data confidence levels from sensors, and safety stock levels of rubber liners can be obtained first. In practical applications, the currently available storage space can be obtained in real time through the warehouse management system interface, including constraints such as the number of vacant shelf locations, load-bearing limits, and environmental conditions (e.g., moisture-proof and high-temperature-proof areas), in cubic meters or the number of standard storage locations. The safety stock level is a minimum inventory level pre-set based on historical consumption rates, supply cycles, and service level targets, and is stored in the material master data of the enterprise resource planning system.

[0080] Then, based on the initial data confidence level and the target data confidence level, the confidence decay coefficient is calculated. The formula for the confidence decay coefficient is: Confidence decay coefficient = 1 - Target data confidence level / Initial data confidence level. Next, based on the safety stock level, the confidence decay coefficient, and the inventory risk coefficient, the inventory adjustment is calculated. The inventory risk coefficient reflects the degree of impact of sensor data uncertainty on inventory demand. The formula for the inventory adjustment is: Inventory adjustment = Safety stock level * Confidence decay coefficient * Inventory risk coefficient. The inventory risk coefficient is a weighted parameter set according to the criticality of the area where the liner is located and the severity of the consequences of failure. The inventory risk coefficient can be calibrated, with a value range of 0.5 to 2. For critical high-wear areas such as the ball mill discharge end, it is set to 1.5 to 2, and for non-critical areas such as the feed end, it is set to 0.5 to 1.

[0081] Finally, adjust the safety stock based on the inventory adjustment amount and the current available storage space. The safety stock can be increased by adding the inventory adjustment amount to obtain the adjusted safety stock. Furthermore, a constraint must be met: the adjusted safety stock multiplied by the volume of a single liner must be less than or equal to the current available storage space. If the calculated result exceeds the available space, the system triggers a storage space alarm and suggests activating a backup warehouse or adjusting the inventory layout of other materials. For example, if the safety stock of a certain specification of rubber liner is set to 10 pieces, the corresponding initial sensor data confidence level is 1.0, the current target data confidence level is 0.6, the inventory risk coefficient is set to 1.5 (this area is a critical wear zone), the single-piece volume is 0.5 cubic meters, and the current available storage space is 15 cubic meters. The calculated confidence decay coefficient is 0.4, the inventory adjustment amount is 6 pieces, the adjusted safety stock is 16 pieces, and the required storage space is 8 cubic meters, which is less than the available space of 15 cubic meters, making the adjustment feasible. The system updates the inventory limit threshold accordingly and triggers a replenishment command to replenish the inventory to the new safety stock level.

[0082] Through the above technical solution, this embodiment establishes a dynamic quantitative correlation between data quality and inventory level, which can automatically increase safety buffer inventory to cope with increased uncertainty when sensor reliability declines, and maintain an economic inventory level when data quality is good. This achieves an adaptive balance between data uncertainty and inventory risk, avoiding the risk of spare parts shortage caused by data drift, and preventing the capital occupation and warehousing cost waste caused by a fixed high inventory strategy.

[0083] The beneficial effects of implementing the embodiments of the present invention include: the embodiments of this application first obtain the current production process parameters and current sensor data of the ball mill, then calculate the target data confidence level based on the current production process parameters and current sensor data, and then adjust the maintenance plan and inventory management strategy of the rubber liner based on the target data confidence level. Thus, it is possible to combine the current production process parameters and current sensor data to calculate the target data confidence level and adjust the maintenance plan and inventory management strategy to achieve intelligent maintenance management of the rubber liner, thereby improving the accuracy of rubber liner wear identification and the reliability of maintenance management.

[0084] like Figure 2 As shown, this embodiment of the invention also provides an intelligent maintenance and management system for rubber liners, comprising: The data acquisition module 301 is used to acquire the current production process parameters and current sensing data of the ball mill. The current production process parameters include the mill speed and material flow rate, and the current sensing data includes vibration sensing data and acoustic sensing data. The data confidence calculation module 302 is used to calculate the target data confidence based on the current production process parameters and the current sensor data. The target data confidence reflects the measurement accuracy of the sensor. The maintenance plan adjustment module 303 is used to adjust the maintenance plan of the rubber liner according to the target data confidence level. The maintenance plan adjustment includes the remaining service life adjustment and the priority adjustment. The inventory management strategy adjustment module 304 is used to adjust the inventory management strategy for rubber liners based on the target data confidence level. The inventory management strategy adjustment includes safety stock adjustment and procurement strategy adjustment.

[0085] The content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0086] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

Claims

1. A method for intelligent maintenance and management of rubber lining plates, characterized in that, Includes the following steps: The current production process parameters and current sensing data of the ball mill are obtained. The current production process parameters include the mill speed and material flow rate, and the current sensing data includes vibration sensing data and acoustic sensing data. Based on the current production process parameters and the current sensing data, the target data confidence level is calculated, which reflects the measurement accuracy of the sensor. Based on the target data confidence level, the maintenance plan for the rubber liner is adjusted, including adjustments to the remaining service life and priorities. Based on the target data confidence level, the inventory management strategy for rubber liners is adjusted, including adjustments to safety stock levels and procurement strategies.

2. The method according to claim 1, characterized in that, The step of calculating the target data confidence level based on the current production process parameters and the current sensor data includes: Obtain the initial data confidence level from the sensor; Based on the current production process parameters and the current sensing data, a first deviation is calculated, which reflects the measurement deviation caused by sensor aging. Based on the current sensing data, a second deviation is calculated, which is used to reflect the correlation between data from different types of sensors; Based on the current production process parameters and the current sensing data, a third deviation is calculated, which reflects the sensor's response deviation to changes in the production process parameters. The target deviation is obtained by averaging the first deviation, the second deviation, and the third deviation. The target data confidence level is calculated based on the initial data confidence level and the target deviation.

3. The method according to claim 2, characterized in that, The step of calculating the first deviation based on the current production process parameters and the current sensor data includes: Acquire the first historical sensor data under the current production process parameters; Based on the first historical sensing data, calculate the historical average and historical standard deviation; The first deviation is calculated based on the historical average, historical standard deviation, and current sensing data.

4. The method according to claim 2, characterized in that, The step of calculating the second deviation based on the current sensing data includes: Based on the vibration sensing data, extract the frequency components of the vibration signal; Based on the acoustic sensing data, extract the acoustic signal energy; Calculate the Pearson correlation coefficient based on the frequency components of the vibration signal and the energy of the acoustic signal; The second deviation is calculated based on the Pearson correlation coefficient and the preset coefficient threshold.

5. The method according to claim 2, characterized in that, The step of calculating the third deviation based on the current production process parameters and the current sensor data includes: Obtain historical production process parameters; Acquire second historical sensor data under the historical production process parameters; Calculate the change in the current sensing data based on the second historical sensing data and the current sensing data; Calculate the change in production process parameters based on the historical production process parameters and the current production process parameters; Based on the changes in the production process parameters, the target changes in sensor data are determined by querying the mapping table between the changes in production process parameters and the changes in sensor data. The third deviation is calculated based on the change in the current sensing data and the change in the target sensing data.

6. The method according to claim 2, characterized in that, The step of averaging the first deviation, the second deviation, and the third deviation to obtain the target deviation includes: An ultrasonic pulse is applied to the physical connection layer between the vibration sensing device and the ball mill cylinder. The vibration sensing device is disposed on the surface of the ball mill cylinder. The physical connection layer includes a fixing adhesive or a mounting base. The first echo signal generated by the reflection of the ultrasonic pulse through the physical connection layer is acquired. Extract the first echo features of the first echo signal, which include peak amplitude, total energy, propagation time and spectral composition. The first echo feature is compared with the reference echo feature to calculate the fourth deviation. The fourth deviation is used to reflect the tightness of the physical connection between the vibration sensing device and the ball mill cylinder. The reference echo feature includes the peak amplitude reference value, the total energy reference value, the propagation time reference value, and the reference spectrum. The target deviation is obtained by averaging the first deviation, the second deviation, the third deviation, and the fourth deviation.

7. The method according to claim 2, characterized in that, The step of averaging the first deviation, the second deviation, and the third deviation to obtain the target deviation includes: A broadband ultrasonic pulse is applied to a rubber liner, and the second echo signal generated by the reflection of the broadband ultrasonic pulse through the inside of the rubber liner is acquired. Extract the second echo feature of the second echo signal, the second echo feature including the current resonant frequency and the peak half-width at half-maximum, the peak half-width at half-maximum being used to represent the width from the highest point of the peak to half the height of the peak; The current quality factor is calculated based on the current resonant frequency and the peak half-width at half-maximum. The current quality factor is used to reflect the energy attenuation in the damped vibration system. Calculate the resonant frequency drift based on the current resonant frequency and the reference resonant frequency; Calculate the damping variation based on the current quality factor and the reference quality factor; Based on the resonant frequency drift and the damping variation, a fifth deviation is calculated, which reflects the degree of microstructural change in the internal material of the rubber liner caused by fatigue accumulation. The target deviation is obtained by averaging the first deviation, the second deviation, the third deviation, and the fifth deviation.

8. The method according to claim 1, characterized in that, When the maintenance plan is adjusted to a remaining service life adjustment, the step of adjusting the maintenance plan for the rubber liner based on the target data confidence level includes: Obtain the initial data confidence level and remaining lifespan of the sensor; Calculate the confidence ratio based on the initial data confidence level and the target data confidence level; The service life reduction factor is calculated based on the confidence ratio and the preset maximum reduction ratio. The remaining service life is adjusted according to the service life reduction factor.

9. The method according to claim 1, characterized in that, When the inventory management strategy is adjusted to safety stock level adjustment, the step of adjusting the inventory management strategy for rubber liners based on the target data confidence level includes: Obtain the current available storage space, the initial data confidence level of the sensors, and the safety stock level of the rubber liner; Calculate the confidence decay coefficient based on the initial data confidence level and the target data confidence level; Based on the safety stock level, the confidence decay coefficient, and the inventory risk coefficient, the inventory adjustment amount is calculated. The inventory risk coefficient is used to reflect the degree of impact of sensor data uncertainty on inventory demand. The safety stock level is adjusted based on the inventory adjustment amount and the currently available storage space.

10. An intelligent maintenance and management system for rubber liners, characterized in that, include: The data acquisition module is used to acquire the current production process parameters and current sensing data of the ball mill. The current production process parameters include the mill speed and material flow rate, and the current sensing data includes vibration sensing data and acoustic sensing data. The data confidence calculation module is used to calculate the target data confidence based on the current production process parameters and the current sensing data. The target data confidence reflects the measurement accuracy of the sensor. The maintenance plan adjustment module is used to adjust the maintenance plan of the rubber liner according to the target data confidence level. The maintenance plan adjustment includes remaining service life adjustment and priority adjustment. The inventory management strategy adjustment module is used to adjust the inventory management strategy for rubber liners based on the target data confidence level. The inventory management strategy adjustment includes safety stock adjustment and procurement strategy adjustment.

Images