An internet of things fault evolution analysis and maintenance decision method for papermaking equipment

Abstract

The application discloses a papermaking equipment-oriented Internet of Things fault evolution analysis and maintenance decision method, relates to the technical field of monitoring and analysis, and comprises the following steps: acquiring pressure pulsation data collected by a pressure sensor arranged at the back of a suction box panel and the inlet of a vacuum pump; acquiring acoustic emission data collected by an acoustic emission sensor arranged synchronously with the pressure pulsation data; calculating ultrasonic intensity data according to the acoustic emission data; when the ultrasonic intensity data exceeds a preset intensity threshold, performing phase analysis on the pressure pulsation data to obtain pressure phase angle data; matching the pressure phase angle data with a pre-stored cavitation source phase feature library, if the pressure phase angle data falls within a first predetermined phase angle range, generating first maintenance instruction data, and if the pressure phase angle data falls within a second predetermined phase angle range, generating cleaning instruction data and performing a cleaning operation on the suction box panel. The application has the effect of improving fault evolution analysis efficiency.

Description

Technical Field

[0001] This application relates to the field of monitoring and analysis technology, and in particular to an Internet of Things (IoT) fault evolution analysis and maintenance decision-making method for papermaking equipment. Background Technology

[0002] In the continuous production process of the paper industry, the stable operation of the vacuum system is directly related to the forming and dewatering efficiency of the paper sheet, thus affecting the quality and production capacity of the finished paper. A vacuum system typically consists of a vacuum pump, suction tank, pipelines, and related valves. Its core function is to rapidly remove moisture from the paper web using negative pressure. However, vacuum systems operate under complex fluid conditions of high flow rate and low pressure for extended periods, making them highly susceptible to cavitation. Cavitation not only leads to increased equipment vibration and excessive noise, but also causes damage to the pump impeller, corrosion of the suction tank panel, and erosion of the pipeline inner wall. In severe cases, it can trigger unplanned shutdowns of the entire paper production line. Currently, paper companies mainly rely on regular maintenance or vibration monitoring to address cavitation problems. However, these methods are insufficient for effective identification in the early stages of cavitation, and cannot accurately pinpoint the specific source of cavitation, resulting in delayed and insufficiently targeted maintenance decisions.

[0003] Most existing monitoring technologies for cavitation in vacuum systems are based on threshold judgments of a single physical quantity. For example, vibration sensors are installed on the vacuum pump casing or the outer wall of the pipe to collect vibration amplitude data and compare it with preset alarm values. When the vibration exceeds the limit, an abnormality is determined to exist in the equipment. However, vibration signals are easily interfered with by the operation of surrounding machinery, and the vibration characteristics in the early stages of cavitation are often submerged in background noise, making accurate extraction difficult. Another common approach is to use acoustic emission sensors to collect ultrasonic signals and determine whether cavitation has occurred by monitoring the ultrasonic intensity. However, ultrasonic intensity only reflects the severity of the cavitation phenomenon and cannot distinguish whether cavitation occurs in the impeller area inside the vacuum pump or in the tiny channels of the suction tank panel. Since the downtime and maintenance costs involved in vacuum pump repair and suction tank panel cleaning differ greatly, if the source of the fault cannot be accurately identified, maintenance personnel often have to adopt a conservative strategy and prioritize a comprehensive shutdown inspection. This not only causes unnecessary production losses but may also lead to a problem that could be solved by simple cleaning being escalated into a complex disassembly and repair due to misjudgment.

[0004] Furthermore, existing technical solutions typically employ linear judgment logic when processing monitoring data. This means that an alarm or maintenance command is triggered as soon as a parameter exceeds a certain limit, lacking a dynamic tracking and verification mechanism for the fault evolution process. During vacuum system operation, cavitation characteristics may evolve non-linearly due to fluctuations in operating conditions and changes in fluid properties. For example, minor panel blockage may resolve itself due to changes in slurry properties, and pump cavitation may temporarily disappear due to load adjustments. Traditional single-threshold alarms cannot capture these dynamic changes, easily leading to false alarms or missed alarms, making maintenance decisions lack a reliable basis. More critically, existing systems often assume the fault is resolved after performing cleaning or adjustment operations, lacking closed-loop data verification of the operation's effectiveness. This results in the same fault potentially recurring and failing to be fundamentally resolved.

[0005] To address the aforementioned issues, there is an urgent need in this field for a fault evolution analysis and maintenance decision-making method that can integrate multi-source sensor data, deeply mine fault characteristic phase information, and introduce a nonlinear cyclic verification mechanism. This method would enable precise location of cavitation sources in the vacuum system of papermaking equipment and ensure the effectiveness of maintenance operations through a clean-verify closed-loop logic, thereby reducing the frequency of unplanned downtime and extending the service life of the equipment. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this application provides an IoT-based fault evolution analysis and maintenance decision-making method for papermaking equipment.

[0007] In a first aspect, this application provides an IoT-based fault evolution analysis and maintenance decision-making method for papermaking equipment, applied to the vacuum system of the papermaking equipment. The vacuum system includes a vacuum pump and a water suction tank. The method includes the following steps: Acquire pressure pulsation data from pressure sensors deployed at the vacuum pump inlet and on the back of the suction tank panel; Acquire acoustic emission data collected by an acoustic emission sensor deployed synchronously with the pressure pulsation data; Calculate the ultrasonic intensity data based on the acoustic emission data; When the ultrasonic intensity data exceeds a preset intensity threshold, phase analysis is performed on the pressure pulsation data to obtain pressure phase angle data corresponding to the acoustic emission data; The pressure phase angle data is matched with a pre-stored cavitation source phase feature library, which includes a first predetermined phase angle range corresponding to pump cavitation and a second predetermined phase angle range corresponding to panel blockage cavitation. If the pressure phase angle data falls within the first predetermined phase angle range, then first maintenance instruction data is generated, which is used to instruct the maintenance of the vacuum pump. If the pressure phase angle data falls within the second predetermined phase angle range, cleaning instruction data is generated, and the cleaning operation of the water tank panel is executed. After the cleaning operation is completed, the updated pressure pulsation data and the updated acoustic emission data are reacquired, and the updated ultrasonic intensity data and the updated pressure phase angle data are calculated based on the updated pressure pulsation data and the updated acoustic emission data. If the updated ultrasonic intensity data does not exceed the preset intensity threshold, then cleaning success confirmation data is generated; If the updated ultrasonic intensity data still exceeds the preset intensity threshold, then based on the matching result between the updated pressure phase angle data and the cavitation source phase feature library, a second maintenance instruction data is generated or the cleaning operation is repeated.

[0008] Preferably, the calculation of ultrasonic intensity data based on the acoustic emission data specifically includes: The acoustic emission data is subjected to bandpass filtering to obtain filtered acoustic emission data; The filtered acoustic emission data is subjected to envelope detection processing to extract the envelope waveform data of the acoustic emission signal; The envelope waveform data is subjected to root mean square calculation of a sliding window according to a preset time window length to generate ultrasonic intensity time series data; The ultrasonic intensity time series data is subjected to first-order hysteresis filtering to obtain ultrasonic intensity data.

[0009] Preferably, phase analysis is performed on the pressure pulsation data to obtain pressure phase angle data corresponding to the acoustic emission data, specifically including: The pressure pulsation data is subjected to a fast Fourier transform to extract the dominant frequency component data, and the dominant frequency of the pressure fluctuation is determined based on the dominant frequency component data. Perform a Hilbert transform on the pressure pulsation data to obtain the instantaneous phase time series data of the pressure signal; The acoustic emission data is subjected to a short-time Fourier transform to generate time-spectrum data, and the peak moments in which the energy density exceeds a preset energy threshold are identified from the time-spectrum data to obtain energy peak moment data. Verify whether the energy peak moment data is within the time period when the ultrasonic intensity data exceeds the preset intensity threshold. If so, the energy peak moment data is taken as valid peak moment data. From the instantaneous phase time series data, the instantaneous phase value corresponding to the effective peak time data is extracted, and the instantaneous phase value is normalized to obtain the pressure phase angle data.

[0010] Preferably, the construction methods for the cavitation source phase feature library include: Acquire historical fault data, which includes historical pressure pulsation data and acoustic emission data corresponding to known fault types. The known fault types include at least pump cavitation type and panel blockage cavitation type. Phase analysis was performed on the historical pressure pulsation data corresponding to each known fault type to obtain the historical pressure phase angle dataset for each fault type. Cluster analysis was performed on the historical pressure phase angle dataset for each fault type to obtain the core phase angle distribution interval for each fault type; Based on the core phase angle distribution range, the first predetermined phase angle range corresponding to cavitation inside the pump and the second predetermined phase angle range corresponding to cavitation due to blockage of the panel are determined respectively. The first predetermined phase angle range and the second predetermined phase angle range are associated and stored in the cavitation source phase feature library.

[0011] Preferably, if the pressure phase angle data falls within the second predetermined phase angle range, cleaning instruction data is generated, specifically including: When the pressure phase angle data exhibits a random distribution characteristic in multiple consecutive monitoring cycles and the ultrasonic intensity data continuously exceeds a preset intensity threshold, the sliding standard deviation of the pressure phase angle data is calculated. Determine whether the sliding standard deviation data exceeds a preset randomness threshold; if so, determine that the pressure phase angle data falls within the second predetermined phase angle range. Based on the current location of the suction tank panel where cavitation is occurring, clean instruction data is generated that includes the target suction tank panel identification and cleaning parameter data, including spray pressure value, spray duration and spray angle. The cleaning instruction data is sent to the high-pressure spray actuator corresponding to the marking on the target water tank panel.

[0012] Preferably, based on the matching results between the updated pressure phase angle data and the cavitation source phase feature library, a second maintenance instruction data is generated or the cleaning operation is repeated, specifically including: Obtain the updated pressure phase angle data and the updated ultrasonic intensity data; Determine whether the updated pressure phase angle data still falls within the second predetermined phase angle range, and whether the updated ultrasonic intensity data exceeds a preset intensity threshold; If both are true, then accumulate the cleaning count data and determine whether the cleaning count data exceeds the preset count threshold; If the preset number of times threshold is not exceeded, the cleaning parameter data is adjusted according to the data change trend after this cleaning operation, updated cleaning instruction data is generated, and the cleaning operation is executed again. If the preset number of times is exceeded, a third maintenance instruction data is generated, and the pressure phase angle data and ultrasonic intensity data before and after each cleaning operation are packaged as attachment data.

[0013] Preferred options also include: When the updated pressure phase angle data shifts from the second predetermined phase angle range to the first predetermined phase angle range, the updated ultrasonic intensity data is acquired; Determine whether the updated ultrasonic intensity data exceeds a preset intensity threshold; if so, generate fourth maintenance instruction data. If the updated ultrasonic intensity data does not exceed the preset intensity threshold, cleaning success confirmation data is generated, and the offset of the pressure phase angle data is recorded as reference data for changes in equipment status.

[0014] Preferred options also include: The ultrasonic intensity data, the pressure phase angle data, the first maintenance command data, the cleaning command data, the cleaning success confirmation data, and the execution time data of each cleaning operation are associated and stored in the IoT cloud platform to generate local device health record data. Based on the local equipment health record data and combined with historical fault data of similar papermaking equipment, cavitation evolution trend comparison data is generated. The cavitation evolution trend comparison data includes phase angle offset rate, ultrasonic intensity growth rate and cleaning operation frequency. When any item in the cavitation evolution trend comparison data exceeds a preset trend threshold, predictive maintenance early warning data is generated.

[0015] Preferred options also include: Obtain successful verification data after the cleaning operation. The successful verification data includes the updated ultrasonic intensity data and the updated pressure phase angle data, as well as the rate of change of ultrasonic intensity and the rate of change of pressure phase angle before and after the cleaning operation. Based on the successful verification data, calculate the actual phase angle offset data of the cavitation characteristics under the current operating conditions; The actual phase angle offset data is compared with the historical phase angle range in the cavitation source phase feature library. If the deviation exceeds the preset deviation threshold, the feature library correction program is triggered. In the feature library correction program, the actual phase angle offset data corresponding to multiple successful verifications within a preset time period are obtained, and a weighted average calculation is performed to obtain the corrected phase angle range. The corrected phase angle range is then used to update the first predetermined phase angle range or the second predetermined phase angle range in the cavitation source phase feature library.

[0016] Secondly, this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform any of the above-described IoT fault evolution analysis and maintenance decision-making methods for papermaking equipment.

[0017] In summary, this application includes at least one of the following beneficial technical effects: 1. This application provides an IoT-based fault evolution analysis and maintenance decision-making method for papermaking equipment. By introducing phase analysis technology, it can correlate the ultrasonic signal generated by cavitation with its corresponding pressure fluctuation phase angle, thereby constructing a cavitation source phase feature library. Furthermore, it can accurately distinguish between costly pump cavitation and panel blockage cavitation that can be solved by simple cleaning. This effectively solves the pain point of not being able to distinguish the fault source, avoids a complete shutdown inspection due to misjudgment, significantly reduces unnecessary equipment disassembly and repair costs, and ensures that the production line only needs in-depth maintenance when necessary. 2. By continuously tracking the pressure phase angle data, the dynamic evolution of cavitation characteristics can be captured. When the pressure phase angle data shifts from the second predetermined phase angle range to the first predetermined phase angle range, the migration of the fault source can be keenly identified. This ability to analyze the fault evolution process enables maintenance personnel to gain insight into the deep changes in equipment status, providing a scientific basis for formulating forward-looking maintenance strategies, rather than passively responding to a single fault point. 3. This invention associates and stores each monitoring data, maintenance instruction, operation effect and time information to the Internet of Things cloud platform, generating a dynamic local equipment health record. By combining historical fault data of similar equipment, the system can generate cavitation evolution trend comparison data. When the trend index exceeds the preset threshold, the system can issue a predictive maintenance warning in advance, upgrading traditional post-maintenance or periodic maintenance to predictive maintenance, minimizing unplanned downtime. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a method for IoT-based fault evolution analysis and maintenance decision-making for papermaking equipment, as described in this application. Detailed Implementation

[0020] The following description, in conjunction with the implementation of this invention, is merely an example and illustration of the concept of this invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the inventive concept or exceed the scope defined in these claims, all of which should fall within the protection scope of this invention.

[0021] Application Overview: In existing technologies, cavitation fault detection in vacuum systems of papermaking equipment often relies on threshold judgments of a single physical quantity, making it difficult to balance the accuracy of fault source localization with the targeted nature of maintenance decisions. Traditional methods typically employ vibration monitoring or acoustic emission energy detection, triggering an alarm when the vibration amplitude or ultrasonic intensity exceeds a preset threshold. However, vibration signals are easily affected by surrounding mechanical interference, and while acoustic emission energy can reflect the severity of cavitation, it cannot distinguish whether cavitation occurs inside the vacuum pump or at the suction tank panel. Existing equipment cannot simultaneously sense the phase correlation between pressure fluctuation patterns and acoustic emission energy. Especially in the early stages of cavitation when characteristics are weak or multiple faults are coupled, single-parameter detection models may exhibit systematic misjudgments, leading maintenance personnel to often adopt a conservative strategy, prioritizing a complete shutdown for inspection, resulting in unnecessary production losses. This detection method, which relies on a single threshold and lacks a fault source differentiation mechanism, is insufficient to meet the precise maintenance requirements of continuous papermaking production.

[0022] To address the aforementioned issues, the inventors, through extensive experimental research, discovered a strong correlation between the occurrence of cavitation events and the instantaneous phase of pressure fluctuations, with cavitation sources at different locations exhibiting differentiated characteristic phase angles. The research revealed that cavitation within vacuum pumps typically occurs within a specific phase interval after a pressure trough, and this phase angle is relatively stable and repeatable. In contrast, cavitation caused by blockage of the suction tank panel exhibits no fixed phase relationship with pressure fluctuations, displaying a random distribution. Based on this finding, the inventors proposed a method for dynamically switching maintenance strategies according to pressure phase angle characteristics, upgrading simple energy threshold judgment to precise positioning through phase feature matching. Further experimental verification incorporated the mapping relationship between pressure phase angle and acoustic emission energy into the fault diagnosis model, and a data closed-loop verification mechanism was introduced after cleaning operations, forming a maintenance decision-making system with self-learning and adaptive capabilities.

[0023] Specifically, the detection system first simultaneously collects high-frequency pressure pulsation data and acoustic emission data from the vacuum pump inlet and the back of the suction tank panel. The ultrasonic intensity is calculated using the acoustic emission data as a cavitation energy indicator; when the energy exceeds a preset threshold, phase analysis is initiated. A Hilbert transform is performed on the pressure pulsation data to extract instantaneous phase time-series data, while a short-time Fourier transform is performed on the acoustic emission data to identify the energy peak moment, extracting the corresponding pressure phase angle as a feature quantity. The real-time pressure phase angle is matched with a pre-stored cavitation source phase feature library: if the phase angle falls within the first predetermined range corresponding to cavitation within the pump, a vacuum pump maintenance command is generated; if the phase angle exhibits a random distribution and persists for more than a set duration, it is determined to be panel blockage cavitation, triggering a high-pressure spray cleaning operation on the corresponding suction tank panel. After cleaning, the system automatically re-collects data and calculates updated ultrasonic intensity and phase angle. If the energy drops below the threshold, cleaning is confirmed as successful. If the energy still exceeds the limit and the phase angle characteristics remain unchanged, the cleaning parameters are dynamically adjusted based on the number of cleaning cycles (e.g., increasing spray pressure or extending duration), and the cleaning is repeated until the preset maximum number of cycles is reached, at which point a shutdown inspection command is issued. During continuous monitoring, the system corrects the phase angle range in the cavitation source phase characteristic library in real time based on feedback data from multiple successful cleaning verifications, forming a dynamic optimization closed loop. In special cases where the phase angle characteristics migrate after the first cleaning (shifting from panel blockage characteristics to pump cavitation characteristics), the system generates corresponding warning or confirmation information based on whether the energy exceeds the limit.

[0024] Compared to existing technologies, traditional methods rely on a single energy threshold and lack the ability to differentiate fault sources, making it difficult to accurately locate the root cause of cavitation problems and easily leading to misjudgments and over-maintenance. This solution innovatively integrates pressure pulsation and acoustic emission dual-mode sensing data, establishing a phase correlation between cavitation events and pressure fluctuations through phase synchronization analysis, achieving precise fault source localization based on characteristic phase angles. Unlike the linear logic of existing technologies that default to fault resolution after cleaning operations, this solution introduces a closed-loop cleaning-verification mechanism. Data retesting after cleaning ensures operational effectiveness, and subsequent strategies are dynamically adjusted based on cleaning results, avoiding repeated ineffective operations. The dynamic wavelength switching mechanism in this solution corresponds to maintenance strategy switching based on phase angle characteristics, combining the high sensitivity of energy threshold detection with the accuracy of phase angle localization. The adaptive adjustment function of model parameters ensures that the feature library can be dynamically corrected as equipment ages or operating conditions change, guaranteeing long-term detection reliability.

[0025] Through the above technical solutions, this application effectively overcomes the technical deficiency that single physical quantity detection cannot distinguish fault sources, and significantly improves the accuracy of cavitation fault location while ensuring real-time performance. The maintenance strategy switching mechanism based on phase angle characteristics balances the speed of energy threshold detection with the accuracy of phase matching, the cleaning-verification closed-loop feedback function ensures the effectiveness of maintenance operations, and the feature library adaptive adjustment function maintains the long-term applicability of the diagnostic model as the equipment condition evolves. This method provides a reliable technical means for fault evolution analysis and precise maintenance of vacuum systems in papermaking equipment, and is particularly suitable for the effective control of unplanned shutdowns in continuous papermaking production scenarios.

[0026] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0027] Example This application discloses an Internet of Things (IoT) fault evolution analysis and maintenance decision-making method for papermaking equipment.

[0028] Reference Figure 1 A method for IoT-based fault evolution analysis and maintenance decision-making in papermaking equipment includes the following steps: A vacuum system for use in papermaking equipment, the vacuum system comprising a vacuum pump and a suction tank, the method comprising the following steps: Acquire pressure pulsation data from pressure sensors deployed at the vacuum pump inlet and on the back of the suction tank panel; Acquire acoustic emission data collected by an acoustic emission sensor deployed synchronously with the pressure pulsation data; Calculate the ultrasonic intensity data based on the acoustic emission data; When the ultrasonic intensity data exceeds a preset intensity threshold, phase analysis is performed on the pressure pulsation data to obtain pressure phase angle data corresponding to the acoustic emission data; The pressure phase angle data is matched with a pre-stored cavitation source phase feature library, which includes a first predetermined phase angle range corresponding to pump cavitation and a second predetermined phase angle range corresponding to panel blockage cavitation. If the pressure phase angle data falls within the first predetermined phase angle range, then first maintenance instruction data is generated, which is used to instruct the maintenance of the vacuum pump. If the pressure phase angle data falls within the second predetermined phase angle range, cleaning instruction data is generated, and the cleaning operation of the water tank panel is executed. After the cleaning operation is completed, the updated pressure pulsation data and the updated acoustic emission data are reacquired, and the updated ultrasonic intensity data and the updated pressure phase angle data are calculated based on the updated pressure pulsation data and the updated acoustic emission data. If the updated ultrasonic intensity data does not exceed the preset intensity threshold, then cleaning success confirmation data is generated; If the updated ultrasonic intensity data still exceeds the preset intensity threshold, then based on the matching result between the updated pressure phase angle data and the cavitation source phase feature library, a second maintenance instruction data is generated or the cleaning operation is repeated.

[0029] In this embodiment, pressure pulsation data refers to the instantaneous fluctuation signal of fluid pressure in the vacuum system acquired by a high-frequency pressure sensor. Specifically, it can be implemented using a piezoelectric pressure sensor deployed on the inlet pipe of the vacuum pump and the back of the suction tank panel, with a sampling frequency of not less than 10kHz, used to capture the pressure shock wave characteristics generated by bubble collapse during cavitation. Acoustic emission data refers to the ultrasonic frequency signal synchronously acquired by an acoustic emission sensor. Specifically, it can be implemented using a broadband acoustic emission sensor deployed at the same location as the pressure sensor, with a sampling frequency of not less than 1MHz, used to capture the high-frequency energy information released by bubble collapse during cavitation.

[0030] Among them, ultrasonic intensity data refers to the quantitative index obtained after energy extraction of the original acoustic emission signal. Specifically, it can be achieved by bandpass filtering, envelope detection and sliding window root mean square calculation. It is used to characterize the severity of cavitation events and serve as the threshold for triggering subsequent phase analysis. The preset intensity threshold is a trigger threshold set according to the statistical characteristics of ultrasonic intensity under normal operating conditions. Specifically, it can be determined by 3 to 5 times the standard deviation of the mean ultrasonic intensity under normal operating conditions. It is used to distinguish between normal fluctuations and abnormal cavitation events.

[0031] Among them, the pressure phase angle data refers to the instantaneous phase value of pressure fluctuation corresponding to the acoustic emission event extracted through phase synchronization analysis. Specifically, the instantaneous phase time series data of the pressure signal can be obtained by using Hilbert transform, and the peak moment of acoustic emission energy can be identified by combining short-time Fourier transform. The unit is degrees. It is used to establish the phase correlation characteristics between cavitation events and pressure fluctuations. The cavitation source phase feature library refers to the pre-stored database of characteristic phase angle ranges corresponding to different types of cavitation faults. Specifically, it can be constructed by collecting historical fault data and performing cluster analysis. It includes the first predetermined phase angle range corresponding to pump cavitation and the second predetermined phase angle range corresponding to panel blockage cavitation, which is used as a benchmark reference for real-time phase angle data matching and judgment.

[0032] The first predetermined phase angle range refers to the characteristic phase angle interval of the cavitation event inside the vacuum pump determined by statistical analysis. Specifically, it can be represented as a continuous angle interval, such as 280 degrees to 340 degrees, used to identify cavitation faults inside the pump. The second predetermined phase angle range refers to the characteristic description method of the cavitation event of the suction tank panel blockage determined by statistical analysis. Since the phase angle of the panel blockage cavitation exhibits a random distribution characteristic, the second predetermined phase angle range does not adopt a specific angle interval, but is defined as the random distribution characteristic of the phase angle data. The judgment basis is that the sliding standard deviation of the phase angle data exceeds a preset randomness threshold, such as 90 degrees, within multiple consecutive monitoring cycles.

[0033] The first maintenance instruction data refers to the maintenance instruction information generated when cavitation inside the pump is detected. Specifically, it may include fields such as equipment identification, fault type, suggested maintenance content, and urgency level, used to notify maintenance personnel to shut down and inspect the vacuum pump. The cleaning instruction data refers to the automatic cleaning operation instruction generated when panel cavitation is detected. Specifically, it may include parameters such as the target suction tank panel identification, spray pressure value, spray duration, and spray angle, used to drive the high-pressure spray actuator to perform automated cleaning.

[0034] The updated pressure pulsation data and updated acoustic emission data refer to real-time monitoring data that are re-collected after the cleaning operation. The specific collection method is exactly the same as the initial data, and they are used to evaluate the actual effect of the cleaning operation.

[0035] The cleaning success confirmation data refers to the confirmation information generated when the ultrasonic intensity drops below a preset threshold after cleaning. This information may include the cleaning operation time, equipment identification, and a comparison of data before and after cleaning, used to record a successful maintenance event. The second maintenance instruction data refers to the upgraded maintenance instruction generated when the ultrasonic intensity still exceeds the limit after cleaning and multiple cleaning attempts have failed. This instruction may include a requirement to stop the machine and inspect the suction tank panel, and includes a package of data from each cleaning operation before and after as an attachment, used to guide manual in-depth maintenance.

[0036] The core innovation of this application lies in constructing a cavitation fault evolution analysis and maintenance decision-making mechanism for papermaking vacuum systems based on phase angle feature matching and a cleaning-verification closed loop. By simultaneously collecting pressure pulsation and acoustic emission data, the pressure phase angle features of cavitation events are extracted and matched with a pre-stored feature library to accurately locate the fault source. At the same time, data retesting and cyclic verification after cleaning operations are introduced. Based on the verification results, cleaning parameters are adaptively adjusted or maintenance strategies are upgraded, thereby solving the technical bottleneck of traditional single threshold detection being unable to distinguish fault sources and lacking effectiveness verification.

[0037] The working process and principle of this application are as follows: First, pressure pulsation data is acquired from pressure sensors deployed at the vacuum pump inlet and on the back of the suction tank panel. Simultaneously, acoustic emission data is acquired from acoustic emission sensors deployed synchronously with the pressure pulsation data. Then, ultrasonic intensity data is calculated based on the acoustic emission data using bandpass filtering, envelope detection, and sliding window root mean square calculation. When the ultrasonic intensity data exceeds a preset intensity threshold, phase analysis is performed on the pressure pulsation data. The instantaneous phase is extracted using Hilbert transform and combined with the peak acoustic emission energy time identified by short-time Fourier transform to obtain the pressure phase angle data corresponding to the acoustic emission data. Next, the pressure phase angle data is matched with a pre-stored cavitation source phase feature library. This feature library includes a first predetermined phase angle range corresponding to cavitation inside the pump and a second predetermined phase angle range corresponding to cavitation caused by panel blockage. If the pressure phase angle data falls within the first predetermined phase angle range... The system generates a first maintenance instruction to instruct the vacuum pump to be repaired. If the pressure phase angle data falls within the second predetermined phase angle range, a cleaning instruction is generated and the cleaning operation of the suction tank panel is executed. After the cleaning operation is completed, updated pressure pulsation data and updated acoustic emission data are reacquired, and updated ultrasonic intensity data and updated pressure phase angle data are calculated based on these updated data. If the updated ultrasonic intensity data does not exceed the preset intensity threshold, cleaning success confirmation data is generated. If the updated ultrasonic intensity data still exceeds the preset intensity threshold, a second maintenance instruction is generated based on the matching result of the updated pressure phase angle data and the cavitation source phase feature library, or the cleaning operation is repeated after adjusting the cleaning parameters according to the number of cleanings and the effect. In this way, a closed-loop control for the precise location, automated processing, and effect verification of cavitation faults in the papermaking vacuum system is achieved.

[0038] Furthermore, based on the acoustic emission data, the ultrasonic intensity data is calculated, specifically including: The acoustic emission data is subjected to bandpass filtering to obtain filtered acoustic emission data; The filtered acoustic emission data is subjected to envelope detection processing to extract the envelope waveform data of the acoustic emission signal; The envelope waveform data is subjected to root mean square calculation of a sliding window according to a preset time window length to generate ultrasonic intensity time series data; The ultrasonic intensity time series data is subjected to first-order hysteresis filtering to obtain ultrasonic intensity data.

[0039] In one specific embodiment, ultrasonic intensity calculation is performed on the raw acoustic emission time-series data collected by the acoustic emission sensor. This process first involves bandpass filtering of the raw acoustic emission signal to remove mechanical vibration noise and electrical interference unrelated to cavitation characteristics. Considering the complex electromagnetic environment of the paper mill and the broadband background noise generated during vacuum system operation, the passband frequency range of the bandpass filter needs to be specifically designed based on the typical spectral characteristics of the cavitation signal. In this embodiment, through prior spectral analysis of historical cavitation event data, it was found that the energy of the acoustic emission signal generated by cavitation is mainly concentrated in the 100kHz to 500kHz frequency band, while motor electromagnetic interference is usually below 1kHz, mechanical vibration frequencies are mostly below 10kHz, and fluid turbulence noise, although having a wide frequency band, has a low energy density. Therefore, the lower passband limit frequency of the bandpass filter is set to 100kHz, the upper passband limit frequency is set to 500kHz, the transition band width is set to 50kHz, and the stopband attenuation is greater than 60dB. A fourth-order Butterworth bandpass filter is selected, whose amplitude-frequency characteristics within the passband are flat, which can avoid distortion of the cavitation signal amplitude. Let the original acoustic emission timing data be , where is the sampling point number and the sampling frequency is 1MHz (i.e., one sampling point per microsecond). After bandpass filtering, the filtered acoustic emission data is obtained. This data retains the cavitation characteristic components in the frequency band from 100kHz to 500kHz, while significantly suppressing out-of-band noise interference.

[0040] Next, envelope detection processing is performed on the filtered acoustic emission data to extract the envelope waveform data of the acoustic emission signal. The purpose is to extract the low-frequency envelope line reflecting the signal amplitude change trend from the high-frequency oscillation signal, which is convenient for subsequent energy calculation. Envelope detection is implemented using the Hilbert transform method, which can accurately obtain the analytical form of the narrowband signal and thus obtain the instantaneous amplitude envelope. In the specific implementation, the Hilbert transform is performed on the signal to obtain its Hilbert transform sequence, and the analytical signal is constructed, where is the imaginary unit. The envelope waveform data is the modulus of the analytical signal, and the calculation formula is . The Hilbert transform is implemented through Fast Fourier Transform: the frequency domain sequence is obtained by performing FFT on the signal, multiplying the positive frequency part by -j and the negative frequency part by j, and then performing inverse FFT. The advantage of this method is that it can extract the envelope without distortion and is not sensitive to the signal phase, making it particularly suitable for processing transient signals such as cavitation. In this embodiment, the sequence length obtained by envelope detection is the same as that of the original signal, and the sampling frequency is still 1MHz, but it reflects the instantaneous amplitude envelope of the signal, with the unit being volts (V), consistent with the original signal.

[0041] Then, the envelope waveform data is subjected to a sliding window root mean square (RMS) calculation according to a preset time window length to generate ultrasonic intensity time-series data. Its function is to convert instantaneous envelope values ​​into physically meaningful short-term energy indicators. The sliding window RMS calculation involves sliding a fixed-length window along a continuous time axis and calculating the root mean square of the envelope values ​​within the window to obtain the energy sequence changing over time. The preset time window length is a key parameter, requiring a trade-off between response speed and smoothing effect: a window that is too short will lead to drastic energy fluctuations, making it difficult to set a stable threshold; a window that is too long will smooth out short-term cavitation pulses, reducing detection sensitivity. In this embodiment, by analyzing the typical duration of cavitation events and combining the dynamic response characteristics of the papermaking vacuum system, the time window length is set to 0.1 seconds, corresponding to 100,000 sampling points (sampling frequency 1MHz). Let the starting point of the current window be the m-th sampling point, and the envelope value sequence within the window be e(m), e(m+1), ..., e(m+L-1), where L is the window length of 100,000 points. Then, the formula for calculating the RMS value of this window is... The sliding step size is set to 1 sampling point. That is, for each new sampling point acquired, the window slides forward 1 point to calculate a new RMS value, thus generating ultrasonic intensity time-series data corresponding one-to-one with the original sampling points (for easy indexing, the RMS value corresponding to the center time of the window is assigned to that time). The unit is the same as volts (V), but its physical meaning is the average energy within a 0.1-second window. A specific numerical demonstration: Assuming the sampling values ​​of a certain envelope data within the window are [0.2V, 0.5V, 0.8V, 1.2V, 0.9V, 0.4V, ...], taking the first 6 points to calculate the RMS value when the window length L=6, it is 0.746V. In practical applications, L=100000, the computational load is large, but it can be efficiently implemented through a recursive algorithm: only the two points entering and exiting the window need to be updated.

[0042] Finally, a first-order lag filter is applied to the ultrasonic intensity time series data to obtain smoothed ultrasonic intensity data. The purpose is to further eliminate glitches caused by random noise or transient fluctuations, making the threshold judgment more stable and reliable. The first-order lag filter is essentially a digital low-pass filter, and its recursive formula is: where is the smoothed output value at the current moment, is the smoothed output value at the previous moment, and is the filter coefficient, with a value range of 0 << 1. The value of determines the filter's response speed and smoothness: the closer to 1, the greater the weight of the current input, resulting in a faster response but poorer smoothing; the closer to 0, the greater the weight of historical values, resulting in better smoothing but a slower response. In this embodiment, it is necessary to balance rapid response to cavitation events and effective smoothing of energy fluctuations. Through experimental comparison, is set to 0.2, meaning the current ultrasonic intensity value is determined by 20% of the latest measured value and 80% of the historical smoothed value. This not only reflects the rising trend of cavitation energy in a timely manner but also effectively suppresses transient spike interference. A specific numerical demonstration is provided: Assuming the smoothed value s(n-1) = 0.426V at the previous moment and the measured value u(n) = 0.53V at the current moment, then the current smoothed value s(n) = 0.2 * 0.53 + 0.8 * 0.426 = 0.4468V. This value serves as the final ultrasonic intensity data output, used for subsequent comparison and judgment with the preset intensity threshold. The preset intensity threshold is set based on the statistical characteristics of the equipment during normal operation: 72 hours of continuous normal operation data are collected, the mean and standard deviation of the ultrasonic intensity are calculated, and the threshold is set to +5 to ensure that normal fluctuations will not trigger false alarms, while early cavitation energy reaching the +3 level can be detected in time. In this embodiment, through statistics, we obtain = 0.15V and = 0.03V, so the preset intensity threshold Th = 0.15 + 5 * 0.03 = 0.30V. The above processing flow gradually transforms the original acoustic emission signal into a smooth and stable ultrasonic intensity index with clear physical meaning, providing a reliable data foundation for subsequent cavitation detection and phase analysis. The entire calculation process is completed in real time in the edge computing nodes. Each time a new acoustic emission sampling point is acquired, the ultrasonic intensity output can be updated within microseconds to meet the real-time monitoring requirements.

[0043] Furthermore, phase analysis is performed on the pressure pulsation data to obtain pressure phase angle data corresponding to the acoustic emission data, specifically including: The pressure pulsation data is subjected to a fast Fourier transform to extract the dominant frequency component data, and the dominant frequency of the pressure fluctuation is determined based on the dominant frequency component data. Perform a Hilbert transform on the pressure pulsation data to obtain the instantaneous phase time series data of the pressure signal; The acoustic emission data is subjected to a short-time Fourier transform to generate time-spectrum data, and the peak moments in which the energy density exceeds a preset energy threshold are identified from the time-spectrum data to obtain energy peak moment data. Verify whether the energy peak moment data is within the time period when the ultrasonic intensity data exceeds the preset intensity threshold. If so, the energy peak moment data is taken as valid peak moment data. From the instantaneous phase time series data, the instantaneous phase value corresponding to the effective peak time data is extracted, and the instantaneous phase value is normalized to obtain the pressure phase angle data.

[0044] In one specific embodiment, phase analysis is performed on synchronously acquired pressure pulsation data and acoustic emission data to extract the pressure phase angle characteristics corresponding to cavitation events. This process first performs a Fast Fourier Transform (FFT) on the pressure pulsation time series data to extract the dominant frequency component data and determine the dominant frequency of the pressure fluctuation. Considering the rotating mechanical characteristics of the vacuum pump and the fluid pulsation characteristics of the pipeline system, pressure fluctuations typically contain multiple frequency components, among which the fundamental frequency component related to the equipment's operating speed has the strongest energy and is the reference frequency for phase analysis. In this embodiment, the pressure pulsation data sampling frequency is set to 20kHz, and each analysis takes a continuous 2-second data segment (i.e., 40,000 sampling points). After applying a Hanning window to this data segment, a 2048-point Fast Fourier Transform is performed. Let the pressure pulsation time series data be denoted as , where n is the sampling point number, and its FFT result yields a complex sequence in the frequency domain, k is the frequency index, and the corresponding frequency value is , where is the frequency resolution. The dominant frequency is then determined by searching for the frequency index corresponding to the maximum value in the amplitude spectrum of . To ensure the accuracy of the dominant frequency extraction, an amplitude threshold condition is set: only when |P(k_max)| exceeds 1.5 times the second largest peak value is the frequency used as the dominant frequency; if multiple close peak values ​​exist (amplitude difference less than 20%), the energy-weighted average of these peak values ​​is taken as the dominant frequency, calculated as follows: where is the index of all frequencies whose amplitude exceeds 80% of the largest amplitude. In this embodiment, historical data analysis revealed that the dominant frequency of pressure fluctuations in a certain type of vacuum pump under normal operating conditions is concentrated between 24Hz and 26Hz (corresponding to pump speeds of 1440-1560rpm). If the extracted dominant frequency deviates from this range by more than 5Hz, an anomaly flag is triggered, indicating possible speed fluctuations or sensor malfunction. The determination of the dominant frequency will provide a reference for subsequent bandpass filtering of the Hilbert transform and will also be used to verify the validity of the pressure pulsation data.

[0045] Next, a Hilbert transform is performed on the pressure pulsation data to obtain the instantaneous phase time series data of the pressure signal. The purpose is to convert the real signal into an analytic signal, thereby obtaining the instantaneous phase value at each moment. The Hilbert transform is implemented in the frequency domain: an FFT is performed on the pressure pulsation data to construct the frequency domain representation of the analytic signal: when k=0 or k=N / 2, ...; when k=1kN / 2-1, ...; when k=N / 2+1kN-1, ..., ... . An inverse FFT is performed on the signal to obtain the analytic signal, where ... is the Hilbert transform of ... The instantaneous phase is the principal argument of the analytic signal, calculated using the formula: where atan2 is the arctangent function in the four quadrants, with a return value range of (-, ). Since atan2 has a jump at a point, phase unwinding is required: when the absolute value of the phase difference between two adjacent points is greater than a certain value, add or subtract an integer multiple of 2 to the subsequent phase, making the phase sequence continuously and monotonically change. Let the continuous phase after unwinding be , with units in radians. To facilitate subsequent correspondence with the acoustic emission event time, the phase also needs to be normalized according to the dominant frequency to eliminate the influence of frequency changes on the physical meaning of the phase angle: the instantaneous phase is converted into a normalized phase relative to the dominant frequency period, with a value range of [0, 1), representing the relative position within one period. In this embodiment, a specific numerical demonstration is used: assuming that the real part Re = 0.8 and the imaginary part Im = 0.6 corresponding to a certain time t, the instantaneous phase is rad. After normalization, this means that the instant is located at 10.24% of the pressure fluctuation period. Through the above processing, the instantaneous phase time series data corresponding to each pressure sampling point is obtained. This data has the same time index as p(n) and serves as the lookup table for subsequent phase angle extraction.

[0046] Subsequently, a short-time Fourier transform (SFT) is performed on the acoustic emission data to generate time-frequency spectrum data. Peak moments where the energy density exceeds a preset energy threshold are then identified from the time-frequency spectrum data, yielding the energy peak moment data. The SFT is performed by sliding a window along the time axis, revealing the energy distribution of the signal in both time and frequency. In this embodiment, the acoustic emission data sampling frequency is 1MHz, and the time resolution requirement is millisecond-level. Therefore, the window length is set to 1024 points (corresponding to 1.024ms), the Hanning window type is selected to reduce spectral leakage, the overlap rate is 75% (i.e., 256 points are slid each time), and the frequency resolution is [value missing]. Let the acoustic emission time-series data be a(m), where m is the sampling point number. An FFT is performed on the data of the [number missing] window to obtain the spectrum of that window, where [value missing] is the window index, k is the frequency index, [value missing] is the sliding step size of 256 points, and w(m) is the Hanning window function. The spectra of each window are arranged by time to obtain a two-dimensional time-frequency spectrum matrix, with units of V / Hz, representing the power spectral density. To identify the energy peak moment corresponding to a cavitation event, it is necessary to extract local maxima points where the energy density exceeds a preset energy threshold from the time spectrum. The preset energy threshold is set based on background noise statistics: the mean and standard deviation of the energy from the time spectrum data for one consecutive hour under normal equipment operation are taken and set. For each frequency point k, a local maximum of energy density is searched along the time axis. This maximum is required to be greater than a certain value and to be the maximum value within three windows before and after it. The center time and frequency of the window corresponding to each maximum value that meets the conditions are recorded to obtain a candidate energy peak moment dataset. Considering that cavitation events usually excite energy simultaneously in a relatively wide frequency band, this embodiment further adopts a frequency band aggregation strategy: the 100kHz-500kHz range is divided into 8 sub-bands (each with a bandwidth of 50kHz). The distribution of energy peak moments is statistically analyzed within each sub-band. If multiple sub-bands simultaneously exhibit energy peaks at close intervals (time difference less than 2ms), the energy of these peak moments is summed, and the moment with the largest energy sum is taken as the representative moment of the cavitation event. Through the above processing, the energy peak time data is obtained, in seconds (s), which corresponds to the precise time point when the cavitation event occurs.

[0047] Then, it is verified whether the energy peak moment data falls within the period when the ultrasonic intensity data exceeds the preset intensity threshold. If so, the energy peak moment data is taken as valid peak moment data. The purpose of this dual verification step is to ensure that the energy peak identified by the short-time Fourier transform does indeed correspond to a significant cavitation event, rather than an isolated noise spike. The ultrasonic intensity data is a smoothed energy index calculated, and its preset intensity threshold is set to 0.30V in this embodiment. A verification window is formed by extending 0.2 seconds forward and backward from the energy peak moment, and it is checked whether there is a continuous period of not less than 50ms within this window. If it exists, the energy peak moment is considered to be confirmed by the ultrasonic intensity data and is marked as a valid peak moment; if it does not exist, the peak moment is discarded, as it may be a brief electromagnetic interference or random noise. In this embodiment, a specific numerical demonstration is used: assuming that an energy peak is identified through short-time Fourier transform, querying the ultrasonic intensity data reveals that the energy level fluctuates continuously between 0.32V and 0.38V during the period from 12.140s to 12.210s, exceeding the threshold of 0.30V. Furthermore, this period covers the entire timeframe; therefore, 12.345s is confirmed as the valid peak time. This verification mechanism effectively reduces the false trigger rate, ensuring high reliability of the event time used for subsequent phase angle extraction.

[0048] Finally, the instantaneous phase value corresponding to the effective peak moment data is extracted from the instantaneous phase time series data, and this instantaneous phase value is normalized to obtain the pressure phase angle data. Since the instantaneous phase time series data and the pressure pulsation data have the same sampling frequency (20kHz), and the accuracy of the effective peak moment can reach the microsecond level, it is necessary to obtain the phase value of the precise moment through interpolation. This embodiment adopts the linear interpolation method: find the two most recent pressure sampling moments and their corresponding normalized phases and, respectively, the phase of moment is . The premise of linear interpolation is that the phase change between adjacent sampling points is approximately linear. Considering that the pressure fluctuation frequency is usually below 100Hz, the phase change within a 50s sampling interval does not exceed 0.018rad (about 1), and the interpolation error can be ignored. After obtaining the normalized value, since it was originally a normalized value in the range of [0,1), it needs to be converted into the angle representation commonly used in engineering. The conversion formula is = 360, the unit is degrees (), and the value range is 0~360. This is the final pressure phase angle data, representing the relative position of the cavitation event in the pressure fluctuation cycle. Continuing the example with specific numerical demonstrations: =12.345s, the pressure sampling times before and after it are =12.34495s corresponding to 0.352, =12.34500s corresponding to 0.358, therefore =0.358, which is converted to an angle of 0.358360 = 128.88. This phase angle data will be used for subsequent matching with the cavitation source phase feature library: if it is stable within the range of 300~340, it is determined to be pump-internal cavitation; if multiple measurements of the phase angle show a random distribution (standard deviation greater than 90), it is determined to be panel blockage cavitation. The entire phase analysis process deeply integrates pressure pulsation and acoustic emission data, and through progressive processing in the frequency domain, time-frequency domain, and phase domain, finally extracts fault feature quantities with clear physical meaning, providing a reliable basis for accurately locating the cavitation source. The entire calculation process is completed in real time in the edge computing node, and the delay from acquiring raw data to outputting the pressure phase angle is controlled within 100ms, meeting the real-time monitoring needs of industrial sites.

[0049] Furthermore, the construction methods for the cavitation source phase feature library include: Acquire historical fault data, which includes historical pressure pulsation data and acoustic emission data corresponding to known fault types. The known fault types include at least pump cavitation type and panel blockage cavitation type. Phase analysis was performed on the historical pressure pulsation data corresponding to each known fault type to obtain the historical pressure phase angle dataset for each fault type. Cluster analysis was performed on the historical pressure phase angle dataset for each fault type to obtain the core phase angle distribution interval for each fault type; Based on the core phase angle distribution range, the first predetermined phase angle range corresponding to cavitation inside the pump and the second predetermined phase angle range corresponding to cavitation due to blockage of the panel are determined respectively. The first predetermined phase angle range and the second predetermined phase angle range are associated and stored in the cavitation source phase feature library.

[0050] In one specific embodiment, a cavitation source phase feature library is constructed by collecting historical fault data and performing system feature extraction and statistical analysis. First, historical fault data was acquired. Data collection was conducted on the vacuum system of a paper mill over a period of six months, collecting 47 pump cavitation events and 83 panel blockage cavitation events confirmed by shutdown inspection. For each fault event, pressure pulsation data (sampling frequency 20kHz) and acoustic emission data (sampling frequency 1MHz) were extracted for 5 minutes before and after the fault time, and stored according to fault type.

[0051] Next, phase analysis was performed on the historical pressure pulsation data for each type of fault to obtain a historical pressure phase angle dataset. The phase analysis was strictly performed according to the above method: Hilbert transform was performed on the pressure pulsation data to obtain the instantaneous phase, and short-time Fourier transform was performed on the acoustic emission data to identify the energy peak moment, and the pressure phase angle corresponding to that moment was extracted. A total of 312 effective cavitation pulses were extracted from 47 pump cavitation events to form a pump cavitation dataset; 546 effective cavitation pulses were extracted from 83 panel blockage events to form a panel blockage dataset.

[0052] Subsequently, cluster analysis was performed on the historical pressure phase angle datasets for each fault type. Since the phase angles are circumferential data, the DBSCAN density clustering algorithm based on circumferential statistics was used to convert the angles to unit circle coordinates. The neighborhood radius was set to 0.261 Euclidean distance corresponding to an angle difference of 15, and the minimum number of points (MinPts) was set to 5. Clustering the pump cavitation dataset yielded a main cluster containing 286 points, ranging from 282 to 341, with a mean of 312.5 and a standard deviation of 8.7. Clustering the panel blockage dataset failed to identify significant clusters. The calculated standard deviation of all points was 103.7 (the theoretical uniform distribution standard deviation is 104), and the Rayleigh test statistic Z=4.13 was less than the critical value of 13.8, confirming that the phase angles were randomly distributed.

[0053] Next, the first and second predetermined phase angle ranges are determined based on the core phase angle distribution interval. For pump cavitation, based on the quartiles of the main clusters and the actual data distribution, the first predetermined phase angle range is determined to be [280, 340], which covers 95.4% of the historical samples. For panel blockage cavitation, due to its random distribution characteristics, the second predetermined phase angle range is defined as having random distribution characteristics, and the judgment criteria are: the sliding standard deviation of the pressure phase angle data exceeds the preset randomness threshold of 90 within three consecutive monitoring cycles.

[0054] Finally, the determined phase angle range is associated and stored in the cavitation source phase feature library. The feature library is stored in the form of a structured data table. The pump cavitation record is (Type=Pump Cavitation, Lower Limit=280, Upper Limit=340, Judgment Method=Interval Matching), and the panel blockage record is (Type=Panel Blockage, Judgment Method=Random Matching, Random Threshold=90). This feature library provides executable judgment rules for subsequent real-time fault diagnosis, transforming historical fault experience into quantitative matching criteria, ensuring the accuracy and consistency of phase angle identification.

[0055] Furthermore, if the pressure phase angle data falls within the second predetermined phase angle range, cleaning instruction data is generated, specifically including: When the pressure phase angle data exhibits a random distribution characteristic in multiple consecutive monitoring cycles and the ultrasonic intensity data continuously exceeds a preset intensity threshold, the sliding standard deviation of the pressure phase angle data is calculated. Determine whether the sliding standard deviation data exceeds a preset randomness threshold; if so, determine that the pressure phase angle data falls within the second predetermined phase angle range. Based on the current location of the suction tank panel where cavitation is occurring, clean instruction data is generated that includes the target suction tank panel identification and cleaning parameter data, including spray pressure value, spray duration and spray angle. The cleaning instruction data is sent to the high-pressure spray actuator corresponding to the marking on the target water tank panel.

[0056] In a specific embodiment, when the system determines that the pressure phase angle data exhibits the random distribution characteristics of panel cavitation, a cleaning instruction generation and execution process is triggered. First, it is confirmed that the pressure phase angle data shows a random distribution and the ultrasonic intensity continuously exceeds the standard over multiple consecutive monitoring cycles. In this embodiment, each monitoring cycle is set to 5 minutes, and the effective cavitation phase angle sample set extracted within the cycle is denoted as N3; simultaneously, the duration of ultrasonic intensity u(t) exceeding a preset threshold of 0.30V within the cycle is not less than 80%, and any consecutive time below the threshold does not exceed 30 seconds. When the above conditions are met for three consecutive cycles, the sliding standard deviation of all M phase angles within these three cycles is calculated. A linear standard deviation formula is used, with units in degrees (°). If the standard deviation exceeds a preset randomness threshold of 90°, the phase angle is determined to fall within a second predetermined phase angle range.

[0057] Assuming 12 phase angles are acquired over three consecutive cycles: 15, 78, 162, 245, 310, 355, 22, 105, 188, 270, 330, 40, the mean is calculated, and the sum of squared deviations is approximately 210,000. The standard deviation is confirmed to be random distribution. Then, the current cavitation-affected water tank panel is located. Based on the sensor channel identifier extracted from the triggered phase angle, the associated panel number and actuator ID are read. Cleaning command parameters are generated: spray pressure = 12MPa as the base value, dynamically adjusted according to the current average ultrasonic intensity, calculated as follows: if , then 14.5MPa; spray duration is linearly interpolated based on the panel area, if = 1.5m, then = 65 seconds; spray angle is set according to the panel width, if 1.2m, then = 33. The command containing the actuator ID, spray pressure, duration, and angle is encapsulated into a message and sent to the corresponding PLC via industrial Ethernet to drive the high-pressure spray actuator for automatic cleaning.

[0058] Furthermore, based on the matching results between the updated pressure phase angle data and the cavitation source phase feature library, a second maintenance instruction data is generated or the cleaning operation is repeated, specifically including: Obtain the updated pressure phase angle data and the updated ultrasonic intensity data; Determine whether the updated pressure phase angle data still falls within the second predetermined phase angle range, and whether the updated ultrasonic intensity data exceeds a preset intensity threshold; If both are true, then accumulate the cleaning count data and determine whether the cleaning count data exceeds the preset count threshold; If the preset number of times threshold is not exceeded, the cleaning parameter data is adjusted according to the data change trend after this cleaning operation, updated cleaning instruction data is generated, and the cleaning operation is executed again. If the preset number of times threshold is exceeded, a third maintenance instruction data is generated. The third maintenance instruction data is used to instruct the machine to stop and check the water tank panel, and the pressure phase angle data and ultrasonic intensity data before and after each cleaning operation are packaged as attachment data.

[0059] In one specific embodiment, after the system completes the first cleaning operation of the suction tank panel, it enters a closed-loop verification and cyclic processing flow. First, it acquires the updated pressure phase angle dataset extracted within the first complete monitoring cycle (5 minutes) after cleaning, along with the average ultrasonic intensity within that cycle. The calculated linear sample standard deviation is then used. If it is >90 and >0.30V, the updated data is determined to still fall within the second predetermined phase angle range, indicating that the cleaning has not completely resolved the problem, and cyclic processing is required.

[0060] The system maintains a cleaning frequency counter, `count`, which starts at 0 and automatically increments by 1 after each cleaning. The preset frequency threshold is 3. When `count` is less than 3, the system adjusts the cleaning parameters based on the data trends before and after the cleaning. The ultrasonic intensity decrease rate is calculated as follows: where is the average ultrasonic intensity of the cycle before cleaning, and is the average ultrasonic intensity of the current cycle after cleaning. The system also includes formulas for adjusting spray pressure and spray duration. For example, if the initial voltage before the first cleaning is 0.55V, and the voltage after cleaning is 0.45V, and the initial voltage is 0.182, with an original pressure of 14.5MPa and a duration of 65 seconds, then the second cleaning will have a pressure and duration of [value missing]. After the second cleaning, data is collected and analyzed again; if the values ​​still exceed the limits, the cycle continues.

[0061] When the count reaches 3 (i.e., the cumulative cleaning count is 3), cleaning will not be repeated, and a third maintenance instruction will be generated, instructing the machine to stop and inspect the suction tank panel. Simultaneously, key data from each cleaning operation, including the phase angle dataset before each cleaning, the average ultrasonic intensity, cleaning parameters, corresponding data after cleaning, and calculated values, are packaged as attachments and sent in JSON format along with the instruction to the maintenance terminal and cloud database, providing a complete basis for manual maintenance. This closed-loop process enables intelligent decision-making from cleaning effect evaluation and adaptive parameter adjustment to final upgrade maintenance.

[0062] Furthermore, it also includes: When the updated pressure phase angle data shifts from the second predetermined phase angle range to the first predetermined phase angle range, the updated ultrasonic intensity data is acquired; Determine whether the updated ultrasonic intensity data exceeds a preset intensity threshold. If so, generate fourth maintenance instruction data. The fourth maintenance instruction data is used to indicate the inspection of the vacuum pump's operating status and includes a warning message that the cleaning operation may have masked signs of cavitation inside the pump. If the updated ultrasonic intensity data does not exceed the preset intensity threshold, cleaning success confirmation data is generated, and the offset of the pressure phase angle data is recorded as reference data for changes in equipment status.

[0063] In one specific embodiment, after the system completes the cleaning operation of the suction tank panel, a fundamental shift in the pressure phase angle characteristics may be detected during the data retesting phase. This shift manifests as the updated phase angle no longer exhibiting the chaotic random distribution seen before cleaning, but rather showing a clear concentration trend, i.e., a shift from characteristics representing panel blockage to characteristics representing pump cavitation. This phenomenon has important physical implications. While the cleaning operation may reduce the degree of panel blockage, it alters the flow characteristics within the vacuum system, making previously masked pump cavitation problems more prominent, or the cleaning process itself may induce changes in pump cavitation conditions.

[0064] The system first acquires valid cavitation phase angle data within the first complete monitoring cycle after cleaning. To ensure the reliability of the statistical results, at least three valid cavitation events need to be collected within this cycle. These phase angle data are then statistically analyzed, calculating their arithmetic mean to understand the central location of the overall distribution, and their standard deviation to assess the dispersion of the data. Based on previous analysis of a large amount of historical fault data, pump cavitation events exhibit stable phase angle characteristics, with a standard deviation typically small, generally within ten degrees; while panel blockage events show a random distribution of phase angles, with a standard deviation often exceeding ninety degrees. Therefore, when the standard deviation of the phase angles significantly decreases after cleaning, from nearly 100 degrees to below 30 degrees, it can be determined that the phase angles have changed from a random distribution to a concentrated distribution. Simultaneously, it is necessary to verify whether these concentrated phase angles fall within the characteristic range of pump cavitation, i.e., whether their average value is close to the typical pump cavitation angle area obtained from previous historical data statistics. When both conditions are met—the standard deviation being below the set clustering threshold and the average value falling within the characteristic range of pump cavitation—the system can confirm that a characteristic shift has occurred.

[0065] After confirming the feature deviation, the system immediately obtains the updated average ultrasonic intensity within the same monitoring period, which reflects the energy level of the current cavitation event. If the ultrasonic intensity still exceeds the normal range during normal operation, it indicates that although the phase angle feature has changed to the pump cavitation mode, the cavitation phenomenon itself has not been eliminated, and the energy is still excessive. At this time, the system generates a fourth maintenance instruction, instructing maintenance personnel to pay attention to the operating status of the vacuum pump. This instruction not only includes basic equipment identification and fault prompts, but also adds different levels of warning information based on the degree of agreement between the current phase angle and typical pump cavitation characteristics. If the deviation between the current average phase angle and the typical value is very small, it indicates a high degree of agreement with the characteristics of pump cavitation, and the warning level is high, suggesting priority inspection of the vacuum pump impeller and inlet filter; if the deviation is slightly larger, the warning level is lowered accordingly, suggesting inspection of the vacuum pump operating status and attention to inlet pressure fluctuations; if the deviation is large but still within an acceptable range, it indicates the need for a comprehensive inspection of the connection between the pump body and the pipeline.

[0066] If the updated ultrasonic intensity does not exceed the normal range, it indicates that the cavitation energy has decreased to a normal level after the cleaning operation. Although the phase angle characteristic has shifted, no actual cavitation has occurred. At this point, the system generates cleaning success confirmation data, confirming that the cleaning operation effectively resolved the panel blockage problem. Simultaneously, the system records this phase angle shift as important equipment status change information. The records include the cleaning operation time, the dispersion of the phase angle before cleaning, the average and dispersion of the phase angle after cleaning, the deviation of the current phase angle from the typical value, and the current ultrasonic intensity value. This information is stored in the equipment health record database as a historical basis for subsequent analysis of equipment status evolution trends. When similar shifts occur repeatedly in the future with gradually increasing energy, these records will help the system provide early warnings of potential in-pump cavitation development trends, enabling a shift from reactive maintenance to predictive maintenance.

[0067] Furthermore, it also includes: The ultrasonic intensity data, the pressure phase angle data, the first maintenance command data, the cleaning command data, the cleaning success confirmation data, and the execution time data of each cleaning operation are associated and stored in the IoT cloud platform to generate local device health record data. Based on the local equipment health record data and combined with historical fault data of similar papermaking equipment, cavitation evolution trend comparison data is generated. The cavitation evolution trend comparison data includes phase angle offset rate, ultrasonic intensity growth rate and cleaning operation frequency. When any item in the cavitation evolution trend comparison data exceeds a preset trend threshold, predictive maintenance early warning data is generated.

[0068] In one specific embodiment, the system associates and stores all monitoring data and maintenance records on a cloud platform, and periodically performs trend analysis to generate predictive maintenance alerts. First, the system encapsulates the data for each fault event's entire lifecycle into a standard data package and reports it to the cloud. This package includes the fault time, device number, fault type, ultrasonic intensity time-series data, pressure phase angle event list, generated maintenance command type, parameters and times of each cleaning operation, before-and-after cleaning comparison data, and the final processing result. The cloud then establishes a database indexed by device number and time, creating a full lifecycle health record for each device.

[0069] The system performs a monthly trend analysis on equipment health records, generating comparative data on cavitation evolution trends. The trend analysis primarily includes the calculation of three core indicators: The first indicator is the phase angle shift rate, which reflects the long-term drift trend of the characteristic cavitation angle within the pump. The system extracts the average phase angle of all cavitation events within the pump from historical data to form a time series. Linear regression is then performed on this series to obtain the slope as the phase angle shift rate, measured in degrees per month. When this rate exceeds five degrees per month, it indicates that the equipment condition is deteriorating rapidly.

[0070] The second indicator is the ultrasonic intensity growth rate, reflecting the evolution trend of cavitation energy. The system extracts the average ultrasonic intensity of all fault events to construct a time series, and linear regression is used to obtain the slope as the energy growth rate, expressed in volts per month. When this rate exceeds 0.03 volts per month, it indicates that the cavitation degree is intensifying.

[0071] The third indicator is the frequency of cleaning operations, reflecting the equipment's dependence on maintenance intervention. The system calculates the average monthly cleaning frequency over the past three months, compares it with the baseline frequency at the beginning of the equipment's operation, and determines the frequency growth rate. When the growth rate exceeds 30%, it indicates a significant deterioration in the equipment's health condition.

[0072] When any of the above indicators exceeds a preset trend threshold, the system generates predictive maintenance early warning data. The warning information includes the name of the exceeding indicator, its current value, the threshold, and suggested maintenance actions. This information is pushed to maintenance management personnel via the IoT platform and stored in the equipment health record as a basis for subsequent maintenance plans. Through this trend analysis and early warning mechanism, the system achieves an upgrade from passively responding to faults to proactively predicting risks.

[0073] Furthermore, it also includes: Obtain successful verification data after the cleaning operation. The successful verification data includes the updated ultrasonic intensity data and the updated pressure phase angle data, as well as the rate of change of ultrasonic intensity and the rate of change of pressure phase angle before and after the cleaning operation. Based on the successful verification data, calculate the actual phase angle offset data of the cavitation characteristics under the current operating conditions; The actual phase angle offset data is compared with the historical phase angle range in the cavitation source phase feature library. If the deviation exceeds the preset deviation threshold, the feature library correction program is triggered. In the feature library correction program, the actual phase angle offset data corresponding to multiple successful verifications within a preset time period are obtained, and a weighted average calculation is performed to obtain the corrected phase angle range. The corrected phase angle range is then used to update the first predetermined phase angle range or the second predetermined phase angle range in the cavitation source phase feature library.

[0074] In one specific embodiment, after each successful cleaning operation, the system automatically records the relevant data as verification success data and periodically uses this data to dynamically correct the cavitation source phase feature library to ensure that the feature library remains accurate as equipment ages or operating conditions change. Verification success data is acquired after each cleaning operation when data retesting confirms that the ultrasonic intensity has dropped below the threshold and the fault has been resolved. The data collected at this time represents the true state of the fault characteristics under the current operating conditions. Each verification success data set includes the average ultrasonic intensity during the monitoring cycle before cleaning, the average ultrasonic intensity after cleaning, the effective cavitation phase angle dataset extracted before cleaning and its mean and standard deviation, and the phase angle dataset extracted after cleaning and its mean and standard deviation.

[0075] The system first determines the type of fault based on the standard deviation of the phase angle before cleaning: if the standard deviation exceeds the randomness threshold of 90 degrees, it is determined to be a panel blockage fault; if the standard deviation is less than 30 degrees and the average phase angle falls near the cavitation characteristic range inside the pump, it is determined to be an internal cavitation fault. For internal cavitation faults, the actual phase angle offset is calculated, which is the absolute value of the difference between the average phase angle before cleaning and the center value of the first predetermined phase angle range currently stored in the feature library, in degrees. For example, if the average phase angle before cleaning is 320 degrees and the current center value in the feature library is 312 degrees, then the offset is 8 degrees. The system records each successfully verified event in the feature offset observation table. When the offset of an event exceeds the preset deviation threshold of 15 degrees, the feature library correction procedure is immediately triggered; if a single event does not exceed the threshold but multiple consecutive events show an increasing trend, correction can also be triggered after accumulating a certain number.

[0076] After the feature library correction procedure is initiated, the system obtains the average pre-cleaning phase angle of all successful cavitation verification events in the pumps of the equipment within the past three months, and calculates a weighted average of these angle values. The weighting follows the principle that newer data has a higher weight; for example, the current event has the highest weight, followed by events from one month ago, then two months ago, and the lowest weight from three months ago. After obtaining the corrected center value through weighted averaging, the original range width is used to determine a new first predetermined phase angle range. For example, if there were four events in the past three months with average phase angles of 308 degrees, 312 degrees, 315 degrees, and 318 degrees respectively, the new center value obtained after weighted averaging is approximately 314 degrees. Therefore, the corrected first predetermined phase angle range is updated to 284 degrees to 344 degrees. Finally, the system updates the corresponding record for the equipment in the cavitation source phase feature library with the corrected range and retains historical versions for traceability. Through this self-learning mechanism, the feature library can automatically correct itself as the equipment operates, avoiding misjudgments caused by equipment aging or operating condition drift, and ensuring that the diagnostic model remains synchronized with the actual state of the equipment.

[0077] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention, they should all fall within the protection scope of the present invention.

[0078] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0079] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for Internet of Things failure evolution analysis and maintenance decision-making for papermaking equipment, characterized in that, A vacuum system for use in papermaking equipment, the vacuum system comprising a vacuum pump and a suction tank, the method comprising the following steps: Acquire pressure pulsation data from pressure sensors deployed at the vacuum pump inlet and on the back of the suction tank panel; Acquire acoustic emission data collected by an acoustic emission sensor deployed synchronously with the pressure pulsation data; Calculate the ultrasonic intensity data based on the acoustic emission data; When the ultrasonic intensity data exceeds a preset intensity threshold, phase analysis is performed on the pressure pulsation data to obtain pressure phase angle data corresponding to the acoustic emission data; The pressure phase angle data is matched with a pre-stored cavitation source phase feature library, which includes a first predetermined phase angle range corresponding to pump cavitation and a second predetermined phase angle range corresponding to panel blockage cavitation. If the pressure phase angle data falls within the first predetermined phase angle range, then first maintenance instruction data is generated, which is used to instruct the maintenance of the vacuum pump. If the pressure phase angle data falls within the second predetermined phase angle range, cleaning instruction data is generated, and the cleaning operation of the water tank panel is executed. After the cleaning operation is completed, the updated pressure pulsation data and the updated acoustic emission data are reacquired, and the updated ultrasonic intensity data and the updated pressure phase angle data are calculated based on the updated pressure pulsation data and the updated acoustic emission data. If the updated ultrasonic intensity data does not exceed the preset intensity threshold, then cleaning success confirmation data is generated; If the updated ultrasonic intensity data still exceeds the preset intensity threshold, then based on the matching result between the updated pressure phase angle data and the cavitation source phase feature library, a second maintenance instruction data is generated or the cleaning operation is repeated.

2. The method of claim 1, wherein, Based on the acoustic emission data, the ultrasonic intensity data is calculated, specifically including: The acoustic emission data is subjected to bandpass filtering to obtain filtered acoustic emission data; The filtered acoustic emission data is subjected to envelope detection processing to extract the envelope waveform data of the acoustic emission signal; The envelope waveform data is subjected to root mean square calculation of a sliding window according to a preset time window length to generate ultrasonic intensity time series data; The ultrasonic intensity time series data is subjected to first-order hysteresis filtering to obtain ultrasonic intensity data.

3. The method of claim 1, wherein, Phase analysis is performed on the pressure pulsation data to obtain pressure phase angle data corresponding to the acoustic emission data, specifically including: The pressure pulsation data is subjected to a fast Fourier transform to extract the dominant frequency component data, and the dominant frequency of the pressure fluctuation is determined based on the dominant frequency component data. Perform a Hilbert transform on the pressure pulsation data to obtain the instantaneous phase time series data of the pressure signal; The acoustic emission data is subjected to a short-time Fourier transform to generate time-spectrum data, and the peak moments in which the energy density exceeds a preset energy threshold are identified from the time-spectrum data to obtain energy peak moment data; Verify whether the energy peak moment data is within the time period when the ultrasonic intensity data exceeds the preset intensity threshold. If so, the energy peak moment data is taken as valid peak moment data. From the instantaneous phase time series data, the instantaneous phase value corresponding to the effective peak time data is extracted, and the instantaneous phase value is normalized to obtain the pressure phase angle data.

4. The IoT-based fault evolution analysis and maintenance decision-making method for papermaking equipment according to claim 1, characterized in that, The methods for constructing the cavitation source phase feature library include: Acquire historical fault data, which includes historical pressure pulsation data and acoustic emission data corresponding to known fault types. The known fault types include at least pump cavitation type and panel blockage cavitation type. Phase analysis was performed on the historical pressure pulsation data corresponding to each known fault type to obtain the historical pressure phase angle dataset for each fault type. Cluster analysis was performed on the historical pressure phase angle dataset for each fault type to obtain the core phase angle distribution interval for each fault type; Based on the core phase angle distribution range, the first predetermined phase angle range corresponding to cavitation inside the pump and the second predetermined phase angle range corresponding to cavitation due to blockage of the panel are determined respectively. The first predetermined phase angle range and the second predetermined phase angle range are associated and stored in the cavitation source phase feature library.

5. The IoT-based fault evolution analysis and maintenance decision-making method for papermaking equipment according to claim 1, characterized in that, If the pressure phase angle data falls within the second predetermined phase angle range, cleaning instruction data is generated, specifically including: When the pressure phase angle data exhibits a random distribution characteristic in multiple consecutive monitoring cycles and the ultrasonic intensity data continuously exceeds a preset intensity threshold, the sliding standard deviation of the pressure phase angle data is calculated. Determine whether the sliding standard deviation data exceeds a preset randomness threshold; if so, determine that the pressure phase angle data falls within the second predetermined phase angle range. Based on the current location of the suction tank panel where cavitation is occurring, clean instruction data is generated that includes the target suction tank panel identification and cleaning parameter data, including spray pressure value, spray duration and spray angle. The cleaning instruction data is sent to the high-pressure spray actuator corresponding to the marking on the target water tank panel.

6. The method for IoT-based fault evolution analysis and maintenance decision-making for papermaking equipment according to claim 1, characterized in that, Based on the matching results between the updated pressure phase angle data and the cavitation source phase feature library, a second maintenance instruction data is generated or the cleaning operation is repeated, specifically including: Obtain the updated pressure phase angle data and the updated ultrasonic intensity data; Determine whether the updated pressure phase angle data still falls within the second predetermined phase angle range, and whether the updated ultrasonic intensity data exceeds a preset intensity threshold; If both are true, then accumulate the cleaning count data and determine whether the cleaning count data exceeds the preset count threshold; If the preset number of times threshold is not exceeded, the cleaning parameter data is adjusted according to the data change trend after this cleaning operation, updated cleaning instruction data is generated, and the cleaning operation is executed again. If the preset number of times is exceeded, a third maintenance instruction data is generated, and the pressure phase angle data and ultrasonic intensity data before and after each cleaning operation are packaged as attachment data.

7. The method for IoT-based fault evolution analysis and maintenance decision-making for papermaking equipment according to claim 6, characterized in that, Also includes: When the updated pressure phase angle data shifts from the second predetermined phase angle range to the first predetermined phase angle range, the updated ultrasonic intensity data is acquired; Determine whether the updated ultrasonic intensity data exceeds a preset intensity threshold; if so, generate fourth maintenance instruction data. If the updated ultrasonic intensity data does not exceed the preset intensity threshold, cleaning success confirmation data is generated, and the offset of the pressure phase angle data is recorded as reference data for changes in equipment status.

8. The IoT-based fault evolution analysis and maintenance decision-making method for papermaking equipment according to claim 1, characterized in that, Also includes: The ultrasonic intensity data, the pressure phase angle data, the first maintenance command data, the cleaning command data, the cleaning success confirmation data, and the execution time data of each cleaning operation are associated and stored in the IoT cloud platform to generate local device health record data. Based on the local equipment health record data and combined with historical fault data of similar papermaking equipment, cavitation evolution trend comparison data is generated. The cavitation evolution trend comparison data includes phase angle offset rate, ultrasonic intensity growth rate and cleaning operation frequency. When any item in the cavitation evolution trend comparison data exceeds a preset trend threshold, predictive maintenance early warning data is generated.

9. The IoT-based fault evolution analysis and maintenance decision-making method for papermaking equipment according to claim 1, characterized in that, Also includes: Obtain successful verification data after the cleaning operation. The successful verification data includes the updated ultrasonic intensity data and the updated pressure phase angle data, as well as the rate of change of ultrasonic intensity and the rate of change of pressure phase angle before and after the cleaning operation. Based on the successful verification data, calculate the actual phase angle offset data of the cavitation characteristics under the current operating conditions; The actual phase angle offset data is compared with the historical phase angle range in the cavitation source phase feature library. If the deviation exceeds the preset deviation threshold, the feature library correction program is triggered. In the feature library correction program, the actual phase angle offset data corresponding to multiple successful verifications within a preset time period are obtained, and a weighted average calculation is performed to obtain the corrected phase angle range. The corrected phase angle range is then used to update the first predetermined phase angle range or the second predetermined phase angle range in the cavitation source phase feature library.

10. A computer-readable storage medium, characterized in that: The device stores instructions that, when executed on a computer, cause the computer to perform an Internet of Things (IoT) fault evolution analysis and maintenance decision-making method for papermaking equipment as described in any one of claims 1 to 9.

Images