Ball mill transmission component fault prediction method based on sound-vibration combined frequency spectrum characteristics

Abstract

The present application relates to a ball mill transmission component fault prediction method based on acoustic vibration combined spectrum characteristics, and belongs to the technical field of ball mill transmission component fault prediction, comprising: synchronously collecting acoustic signals and vibration signals of the ball mill transmission component, obtaining acoustic spectrum and vibration spectrum, calculating meshing center frequency and shaft rotation frequency according to transmission ratio, tooth number and shaft rotation speed; extracting candidate side frequencies on both sides of the same meshing center frequency respectively, calculating the frequency interval of the candidate side frequencies and the meshing center frequency, and combining the shaft rotation frequency and the preset frequency difference range to determine the combined sideband formed by the same mechanical modulation; tracking the combined sideband within the continuous sampling period, when the combined sideband changes from single side to both sides within the preset number of continuous sampling periods, determining the transmission component according to the meshing center frequency and the shaft rotation frequency and outputting the fault prediction result. The present application can reduce the interference of random spectrum peaks and single signal anomalies, and improve the reliability of fault feature recognition and fault prediction.

Description

Technical Field

[0001] This invention relates to the field of ball mill transmission component fault prediction technology, specifically a ball mill transmission component fault prediction method based on combined acoustic and vibration spectrum characteristics. Background Technology

[0002] In existing technologies, ball mill transmission systems operate under complex conditions such as heavy load, low speed, impact, and dust. Gears, drive shafts, bearings, and other components often experience wear, eccentricity, cracks, or loosening, leading to changes in meshing frequency and its sideband structure, which manifest as structural vibration and airborne sound radiation. Current fault diagnosis typically relies on the statistical and time-frequency characteristics of a single vibration signal, sound signal, or multi-source signal, combined with a classification model to output the fault category. For example, patent publication CN117556344A discloses a fault diagnosis method and system for ball mill transmission systems based on multi-source information fusion. This method collects gearbox vibration data, ball mill noise data, and motor current data, and performs fault diagnosis after frequency band feature extraction, feature fusion, and feature selection. While this approach can improve fault identification capabilities using multi-modal information, its focus remains on multi-source feature fusion and classification, without establishing a joint determination mechanism based on the meshing center frequency, shaft rotation frequency, and frequency interval relationship for the corresponding sidebands in the sound and vibration spectra under the same mechanical modulation.

[0003] For example, patent publication CN112729817A discloses a gearbox fault diagnosis method, system, and test bench based on multi-source data fusion. This method improves fault location and complex fault diagnosis capabilities by acquiring and fusing multi-source data such as vibration, noise, temperature, displacement, and oil levels. While this approach increases information sources, the relationships between multi-source data are primarily complementary at the feature or decision levels, lacking direct physical constraints on the frequency structures generated by the same meshing modulation in acoustic and vibration signals. When the ball mill transmission components are in the early stages of a fault, the modulation energy generated by local defects is weak, and sidebands may only appear intermittently on one side of a spectrum or meshing center frequency. This can easily be confused with spectral peaks formed by random impacts, background noise, or operating condition fluctuations. Therefore, it is difficult to reliably distinguish between true fault evolution and transient interference based solely on amplitude, statistics, or single-time-period spectral characteristics.

[0004] Furthermore, the rotational speed and load of the ball mill fluctuate during operation, causing the meshing center frequency and sideband positions to shift with changing operating conditions. Existing methods, lacking continuous tracking of the frequency interval relationship between the sidebands and shaft rotation frequency, are prone to feature mismatch due to frequency drift. Simultaneously, without utilizing the combined response of the acoustic and vibration spectra to the same mechanical modulation, it is difficult to reliably identify transmission components when fault symptoms are weak. Therefore, existing technologies still require a solution that can simultaneously utilize both acoustic and vibration spectral information, using the meshing center frequency and shaft rotation frequency as mechanical constraints, to confirm the correspondence of candidate sidebands on both sides, and to identify the fault development process by observing the change in the joint sideband from single-sided occurrence to continuous occurrence on both sides within a continuous sampling period, thereby achieving early prediction and targeted positioning of transmission components. Summary of the Invention

[0005] The purpose of this invention is to provide a method for predicting the faults of ball mill transmission components based on the combined acoustic and vibration spectral characteristics, thereby addressing some of the drawbacks and shortcomings pointed out in the background art.

[0006] The present invention adopts the following technical solution to solve the above-mentioned technical problems:

[0007] The sound and vibration signals of the ball mill transmission components during operation are collected synchronously to obtain the sound spectrum and vibration spectrum. The meshing center frequency and the rotational frequency of each shaft are calculated based on the transmission ratio, number of teeth and rotational speed of each shaft.

[0008] Candidate sidebands located on both sides of the same meshing center frequency are extracted from the acoustic spectrum and vibration spectrum respectively. The frequency interval between the candidate sidebands and the meshing center frequency is calculated. The acoustic spectrum candidate sidebands and vibration spectrum candidate sidebands located on both sides of the meshing center frequency, whose absolute difference between each frequency interval and the same shaft rotation frequency is within a preset frequency difference range, and whose absolute difference between two frequency intervals is within the preset frequency difference range, are determined as joint sidebands formed by the same mechanical modulation.

[0009] The joint sideband is tracked during a continuous sampling period. When the joint sideband changes from appearing at one side frequency position to appearing at two side frequency positions within a preset number of continuous sampling periods, the transmission component is determined based on the meshing center frequency and shaft rotation frequency that form the joint sideband, and the fault prediction result of the transmission component is output.

[0010] Furthermore, the candidate sidebands are tracked during the continuous sampling period when the shaft speed changes; when the meshing center frequency shifts with the shaft speed, the candidate sidebands whose frequency shift direction is the same as the frequency shift direction of the meshing center frequency and whose absolute difference between the frequency interval change and the shaft speed change is within the preset frequency difference range are retained, and the joint sideband is formed by the retained candidate sidebands.

[0011] Furthermore, the order of appearance of the candidate sidebands in the acoustic spectrum and vibration spectrum is tracked according to the continuous sampling period. When the absolute difference between the frequency interval of the candidate sideband and the meshing center frequency and the rotational frequency of the same shaft is within the preset frequency difference range, and the candidate sideband first appears in one spectrum within a preset number of continuous sampling periods, and then appears at a frequency position in another spectrum, and the absolute difference between the frequency interval of the frequency position and the meshing center frequency and the frequency interval of the first appearing candidate sideband is within the preset frequency difference range, the appearance process in the two spectra is determined as successive manifestation, and the successive manifestation is taken as a sign of fault development.

[0012] Furthermore, for two-stage gear meshing involving the same drive shaft, a first meshing center frequency and a second meshing center frequency are determined respectively; when a joint sideband is formed at the first meshing center frequency and the second meshing center frequency, and the absolute difference between the frequency interval of the two joint sidebands and the rotational frequency of the same shaft is within the preset frequency difference range, and the absolute difference between the two frequency intervals is within the preset frequency difference range, the drive shaft and the transmission components associated with the two-stage gear meshing are identified as fault prediction objects.

[0013] Furthermore, when the candidate sideband appears consecutively in another spectrum, the frequency interval change of the candidate sideband relative to the meshing center frequency before and after the consecutive appearance is compared with the shaft rotation frequency change; when the absolute difference between the two changes is within the preset frequency difference range, the consecutive appearance is determined as a sign of fault development.

[0014] Furthermore, after the candidate sidebands are displayed consecutively, the amplitude and frequency position of the candidate sidebands in the two spectra are continuously tracked in subsequent continuous sampling periods. When the amplitude decrease of the sideband in the spectrum where the candidate sideband first appears reaches a preset amplitude change threshold, and the candidate sidebands in the other spectrum appear in a preset number of continuous sampling periods, and the absolute difference between the frequency intervals of the two is within the preset frequency difference range, the change process is determined as a cross-spectrum response migration and is used as a sign of fault development.

[0015] Furthermore, when the candidate sidebands appear consecutively between the acoustic spectrum and the vibration spectrum, the starting time when the candidate sidebands in both spectra first appear within a preset number of consecutive sampling periods is recorded; within a preset number of consecutive sampling periods, when the order of the starting time periods is the same, the order of the starting time periods is used to distinguish between the continuous mechanical response caused by a local fault in the transmission component and the transient spectral response formed by random impact.

[0016] Furthermore, when the rotational speed of the drive shaft changes, the frequency shift and frequency interval change of the joint sideband at the first meshing center frequency and the second meshing center frequency are tracked; when the absolute difference between the two frequency interval changes and the same shaft rotational frequency change is within the preset frequency difference range, and the absolute difference between the two frequency intervals is within the preset frequency difference range, the frequency shift is determined as a fault symptom of the drive shaft generating periodic disturbances and acting on the meshing of the two-stage gears.

[0017] Furthermore, the formation time of the joint sideband at the first meshing center frequency and the second meshing center frequency is recorded according to the continuous sampling period. When the frequency interval first appears in one meshing center frequency within a preset number of continuous sampling periods, and then appears in another meshing center frequency within a preset number of continuous sampling periods, and the absolute difference between the two frequency intervals and the rotational frequency of the same shaft is within the preset frequency difference range, the fault propagation direction of the transmission components associated with the transmission shaft is determined according to the order of the formation time, and the formation process is used as the basis for predicting the expansion of the fault along the transmission path.

[0018] Furthermore, the time difference formed by the joint sideband at the first meshing center frequency and the second meshing center frequency is recorded, and the time difference is converted into the number of rotation cycles according to the shaft rotation frequency of the transmission shaft; the difference in the number of rotation cycles in adjacent continuous sampling periods is calculated, and when the difference is within the range of the preset number of cycle difference, the number of rotation cycles is determined as the basis for judging whether the fault propagation direction is maintained in the continuous sampling period.

[0019] This invention synchronously acquires acoustic and vibration signals from the ball mill's transmission components during operation, obtaining acoustic and vibration spectra respectively. Based on the transmission ratio, number of teeth, and rotational speed of each shaft, the meshing center frequency and shaft rotational frequency are calculated, establishing a correspondence between the spectrum analysis results and the actual kinematic relationship of the transmission system. Furthermore, candidate sidebands located on either side of the same meshing center frequency are extracted from the two types of spectra. Joint screening is performed using the frequency interval between the candidate sidebands and the meshing center frequency, the shaft rotational frequency, and the difference between the frequency intervals on both sides. This allows the spectral responses formed by the same mechanical modulation to be correlated as joint sidebands, reducing interference from random spectral peaks, environmental noise, and single signal anomalies on fault diagnosis, and improving the reliability of fault feature identification.

[0020] This invention tracks the combined sideband during continuous sampling periods and uses the transformation of the combined sideband appearing at one side frequency position to appearing at two side frequency positions within a preset number of continuous sampling periods as a fault development criterion. Compared to diagnosis based solely on single-period amplitude or single spectral characteristics, this invention utilizes the continuous evolution of the sideband structure to characterize the development process of transmission component faults from weak to strong, which is beneficial for forming prediction results before obvious failure occurs. Furthermore, by identifying the corresponding transmission component through the meshing center frequency and shaft rotation frequency of the combined sideband, the correspondence between fault prediction results and specific transmission components can be enhanced, providing a basis for condition maintenance of the ball mill transmission system. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the fault prediction process for ball mill transmission components based on the combined acoustic and vibration spectral characteristics of this invention.

[0022] Figure 2 This is the candidate sideband matching diagram of the acoustic-vibration joint sideband in Embodiment 1 of the present invention.

[0023] Figure 3 This is a frequency migration tracking diagram of candidate sidebands under variable speed conditions in Embodiment 1 of the present invention.

[0024] Figure 4 This is a candidate sideband continuation display and cross-spectral response migration diagram in Embodiment 1 of the present invention.

[0025] Figure 5 This is a correlation diagram of the joint sideband frequency interval of the double meshing position in Embodiment 2 of the present invention.

[0026] Figure 6 This is a frequency interval migration diagram of the dual meshing position under variable speed conditions in Embodiment 2 of the present invention.

[0027] Figure 7 This is a continuous verification diagram of the number of rotation cycles for fault propagation in Embodiment 2 of the present invention. Detailed Implementation

[0028] As attached Figure 1As shown, in this embodiment, acoustic and vibration signals during the operation of the ball mill transmission components are synchronously acquired to ensure the temporal correspondence between the two types of signals. Acoustic signals are acquired by acoustic sensors fixed in position near the ball mill transmission components, while vibration signals are acquired by vibration sensors fixed in the bearing housing, gearbox housing, or mechanically coupled position of the transmission components. Shaft speed is synchronously recorded by a speed sensor. The acoustic, vibration, and speed sensors are synchronously acquired using the same acquisition clock or a unified trigger signal, ensuring that the acoustic, vibration, and shaft speeds correspond to the same operating period. The sampling frequency is set to cover the frequency range of the highest meshing center frequency to be analyzed and its candidate side frequencies, and remains consistent throughout continuous sampling. The acquired acoustic signals are processed in the frequency domain to obtain the acoustic spectrum, and the acquired vibration signals are processed in the frequency domain to obtain the vibration spectrum. Specifically, each continuous sampling period uses the same sampling duration, data length, and frequency domain processing method. A Fast Fourier Transform can be used to obtain the acoustic and vibration spectra, and the frequency position and amplitude corresponding to the spectral peaks are used as subsequent analysis data. Simultaneously, based on the transmission ratios, gear teeth counts, and shaft speeds of each stage of the ball mill transmission system, the meshing center frequencies corresponding to different gear meshing positions are calculated, and the shaft rotational frequencies of each transmission shaft are also calculated. A pre-established correspondence between each meshing position, transmission shaft, and connected gears, bearings, and other transmission components is established according to the actual transmission connections of the ball mill transmission system. This establishes a correspondence between the meshing center frequencies, shaft rotational frequencies, and specific transmission components, providing a frequency reference for subsequent candidate sidefrequency identification and fault component determination.

[0029] After obtaining the acoustic and vibration spectra, candidate sidebands located on both sides of the same meshing center frequency are extracted from the acoustic and vibration spectra, using the corresponding meshing center frequency as the frequency analysis benchmark. Specifically, a candidate sideband search band is set on both sides of the meshing center frequency. Local spectral peaks are identified within the search band, and local spectral peaks with amplitudes higher than a preset spectrum identification threshold are selected as candidate sidebands. The search band is determined based on the shaft rotation frequency of the corresponding drive shaft and the allowable range of shaft rotation speed variation during ball mill operation, ensuring that the search band covers the possible sideband positions corresponding to the shaft rotation frequency. The preset spectrum identification threshold is predetermined based on the amplitude fluctuations of spectral noise and non-fault spectral peaks within the same frequency range when the ball mill is in normal operation, thereby excluding spectral components that cannot be stably identified. For each extracted candidate sideband, the frequency interval between the candidate sideband and the meshing center frequency is calculated, and each frequency interval is compared with the corresponding shaft rotation frequency. When the absolute difference between the frequency intervals of candidate sidebands in the acoustic spectrum and candidate sidebands in the vibration spectrum, located on the same side of the same meshing center frequency, and the absolute difference between the frequency intervals of the two candidate sidebands is within a preset frequency difference range, it indicates that the two candidate sidebands have modulation characteristics corresponding to the rotation period of the same shaft. In this case, the two candidate sidebands are identified as a joint sideband formed by the same mechanical modulation action. The preset frequency difference range is determined in advance based on the frequency resolution of the spectrum, the rotational speed measurement error, and the frequency variation range caused by shaft speed fluctuations under normal operating conditions, and remains unchanged during the judgment process of the same set of continuously sampled data. This utilizes the common modulation characteristics of acoustic and vibration signals in the frequency domain to reduce the influence of non-target frequency components in a single spectrum on fault judgment.

[0030] During continuous sampling, the identified joint sidebands are continuously tracked, and their appearance on both sides of the meshing center frequency is recorded. Continuous sampling periods are sampling windows set sequentially according to preset sampling durations. Each continuous sampling period uses the same sampling duration and frequency domain processing parameters, and the time interval between adjacent continuous sampling periods remains consistent within the same monitoring process. Initially, when a joint sideband appears only at a candidate sideband position on one side of the meshing center frequency, the candidate sideband position and its corresponding frequency interval are recorded. As the continuous sampling process progresses, when the joint sideband changes from appearing at a candidate sideband position on one side to appearing at candidate sideband positions on both sides of the meshing center frequency, and both candidate sideband positions can be identified within a preset number of continuous sampling periods, this change is considered a spectral representation of the joint sideband evolving from a single-sided feature to a double-sided feature. The preset number is used to exclude occasional spectral peaks that appear only in a single sampling period. The specific number can be predetermined based on the duration of transient spectral components under normal ball mill operation and the length of the continuous sampling period, and remains consistent within the same judgment process. The corresponding gear meshing position is determined based on the meshing center frequency that forms the combined sideband, and the relevant transmission shaft is determined by combining the shaft rotation frequency corresponding to the frequency interval of the combined sideband. Then, based on the pre-established correspondence between the meshing position, transmission shaft, and transmission components, the transmission component associated with the meshing position and transmission shaft is determined, and the fault prediction result corresponding to that transmission component is output. The fault prediction result includes at least the determined transmission component, the corresponding meshing position, and the relevant transmission shaft, and records the state of the combined sideband evolving from a single-sided feature to a double-sided feature as the basis for fault prediction.

[0031] In this embodiment, continuous sampling periods are set during the operation of the ball mill drive components, and the shaft rotation speed and candidate sidebands in the acoustic and vibration spectra are recorded synchronously within each continuous sampling period. The shaft rotation speed is compared for adjacent continuous sampling periods. When the shaft rotation speed changes, the corresponding shaft rotation frequency is recalculated based on the changed shaft rotation speed, and the frequency position change of the meshing center frequency is determined synchronously. Subsequently, the candidate sidebands identified in the previous continuous sampling period are tracked, and the frequency position of each candidate sideband in the subsequent continuous sampling period is recorded. The tracking of candidate sidebands uses their corresponding meshing center frequency, location on the same side of the meshing center frequency, and frequency interval matching the corresponding shaft rotation frequency as association conditions, thereby avoiding misidentifying adjacent spectral peaks as the same candidate sideband. The frequency offset direction of the meshing center frequency and the frequency offset direction of the candidate sidebands are determined respectively. If the frequency offset direction of the candidate sideband is the same as the frequency offset direction of the meshing center frequency, the frequency interval of the candidate sideband relative to the meshing center frequency is further calculated, and the change in frequency interval between adjacent continuous sampling periods is obtained. Simultaneously, the change in shaft rotation frequency is calculated based on the shaft rotation frequency corresponding to adjacent consecutive sampling periods, and the change in frequency interval is compared with the change in shaft rotation frequency. When the absolute difference between the two is within a preset frequency difference range, it indicates that the frequency change of the candidate sideband can maintain a change relationship corresponding to the mechanical rotation characteristics with the change in shaft rotation speed, and therefore the candidate sideband is retained. For candidate sidebands whose frequency offset direction is different from the meshing center frequency, or whose absolute difference between the change in frequency interval and the change in shaft rotation frequency exceeds the preset frequency difference range, they are not used for joint sideband formation. After completing the candidate sideband screening, the acoustic spectrum candidate sidebands and vibration spectrum candidate sidebands that meet the frequency offset direction and frequency interval change relationship are associated according to the corresponding meshing center frequency and located on the same side of the meshing center frequency, and the joint sideband is determined according to the aforementioned joint sideband formation conditions, thereby improving the reliability of candidate sideband identification by utilizing the continuous change relationship of frequency characteristics during shaft rotation speed change.

[0032] In this embodiment, candidate sidebands in the acoustic and vibration spectra are continuously tracked according to continuous sampling periods, and the occurrence time and order of the candidate sidebands in the two spectra are recorded. For each continuous sampling period, the frequency interval between the candidate sideband and the corresponding meshing center frequency is calculated, and the frequency interval is compared with the corresponding shaft rotation frequency. When the absolute difference between the frequency interval and the same shaft rotation frequency is within a preset frequency difference range, the candidate sideband is regarded as a tracking object with mechanical rotation correlation characteristics. If the candidate sideband first appears continuously in one of the acoustic or vibration spectra and remains identifiable within a preset number of continuous sampling periods, the frequency position and corresponding frequency interval of the candidate sideband are recorded. Subsequently, when a candidate sideband appears in the other spectrum, the frequency interval between the candidate sideband and the same meshing center frequency is calculated and compared with the frequency interval corresponding to the candidate sideband that appeared earlier. When the absolute difference between two frequency intervals is within a preset frequency difference range, and the two candidate sidebands are located on the same side of the same meshing center frequency, the candidate sidebands appearing in the two spectra in sequence are identified as the successive manifestations of the same mechanical response in the acoustic and vibration spectra, and the successive manifestations are used as spectral signs of transmission component failure development. The occurrence period refers to the continuous sampling period during which the candidate sidebands meet the aforementioned spectral identification conditions, and the order of occurrence is determined according to the starting period when each candidate sideband first continuously meets the identification conditions.

[0033] After confirming the successive manifestation, frequency tracking continues for the candidate sidebands before and after the manifestation. The frequency interval change of the candidate sidebands relative to the meshing center frequency is calculated, and the shaft rotation frequency change is calculated based on the shaft rotation frequency within the corresponding sampling period. The frequency interval change and the shaft rotation frequency change are compared. When the absolute difference between the two is within a preset frequency difference range, it indicates a correspondence between the successive changes of the candidate sidebands in the two spectra and the drive shaft rotation state, thus further confirming the successive manifestation as a sign of fault development. By verifying the relationship between the frequency interval change and the shaft rotation frequency change, the possibility of occasional spectral components being identified as successive manifestations can be reduced.

[0034] After the candidate sidebands are successively displayed, the amplitude and frequency position changes of the corresponding candidate sidebands in the acoustic and vibration spectra are continuously tracked. The amplitude of the candidate sideband in the spectrum where the candidate sideband first appears is recorded and compared with the amplitude of the candidate sideband in subsequent continuous sampling periods. When the decrease in the amplitude of the candidate sideband reaches a preset amplitude change threshold, and the corresponding candidate sideband in another spectrum can be identified within a preset number of continuous sampling periods, and the absolute difference between the corresponding frequency intervals of the candidate sidebands in the two spectra is within a preset frequency difference range, the change process is determined as a cross-spectral response migration. The preset amplitude change threshold is predetermined based on the amplitude fluctuation range of the corresponding frequency position during the normal operation of the ball mill or the stable operation phase before the fault occurs, so that the amplitude decrease reaching this threshold can be distinguished from normal amplitude fluctuations; the same preset amplitude change threshold is used in the cross-spectral response migration judgment process of the same set of continuous sampling data. Cross-spectral response migration indicates that the spectral response related to the same mechanical modulation gradually shifts from a significant manifestation in one signal type to another signal type, and therefore is regarded as a further sign of the continued development of the fault.

[0035] When candidate sidebands appear consecutively between the acoustic and vibration spectra, the starting time corresponding to the first occurrence of the candidate sidebands in both spectra within a preset number of consecutive sampling periods is also recorded. The chronological relationship of the starting time periods obtained in multiple consecutive sampling periods is compared. If the chronological relationship of the starting time periods of the candidate sidebands in the acoustic and vibration spectra repeats and remains the same during continuous sampling, it indicates that there is a repeatable temporal evolution sequence between the two spectra. Based on this chronological relationship, spectral responses with continuous repetition characteristics are identified as mechanical responses related to local faults in transmission components, and spectral responses lacking a fixed chronological relationship and appearing only in individual consecutive sampling periods are distinguished from them. This enables the identification of continuous mechanical responses and transient spectral responses formed by random impacts.

[0036] In this embodiment, for a transmission structure in which the same drive shaft participates in the meshing of two stages of gears, the first meshing center frequency and the second meshing center frequency corresponding to the meshing of the two stages of gears are determined according to the number of teeth, transmission relationship, and rotational speed of the drive shaft. During continuous sampling, the acoustic spectrum and vibration spectrum near the first meshing center frequency and the second meshing center frequency are analyzed, and the joint sidebands corresponding to the two meshing center frequencies are identified according to the aforementioned method for determining joint sidebands. The frequency intervals between the candidate sidebands in the two joint sidebands and the corresponding meshing center frequencies are calculated, and the two frequency intervals are compared with the shaft rotation frequency of the same drive shaft. When the absolute difference between the two frequency intervals and the shaft rotation frequency is within a preset frequency difference range, it indicates that the sideband modulation characteristics generated at the first meshing position and the second meshing position are related to the rotational characteristics of the same drive shaft. Thus, the drive shaft and the transmission components associated with the meshing of the two stages of gears are identified as fault prediction objects, so as to limit the fault range by utilizing the associated spectral characteristics formed by the shaft rotation frequency at the two meshing positions. When specifically identifying transmission components, based on the pre-established transmission connection relationship, gears, bearings, and transmission shafts that have mechanical connection relationships with both the transmission shaft and the corresponding meshing position are taken as the corresponding fault prediction objects, and the fault prediction range is further limited by combining the specific meshing position of the combined side belt.

[0037] During continuous sampling as the drive shaft speed changes, the frequency position and frequency interval changes of the joint sideband at the first and second meshing center frequencies are continuously tracked. As the drive shaft speed changes, the frequency shift process of the two joint sidebands is recorded, and the frequency interval change between adjacent continuous sampling periods is calculated. Simultaneously, the shaft rotation frequency change is calculated based on the shaft rotation frequency within the corresponding sampling period. The frequency interval changes of the joint sidebands at the first and second meshing center frequencies are compared with the same shaft rotation frequency change. When the absolute differences obtained from both comparisons are within the aforementioned preset frequency difference range, and the absolute difference between the frequency intervals of the two joint sidebands is also within the preset frequency difference range, it indicates that the frequency shifts at both meshing positions change with the rotational state of the same drive shaft. In this case, the frequency shifts of the two joint sidebands are identified as a fault symptom of periodic disturbances in the drive shaft acting on the meshing of the two gears, thus further verifying the spectral changes generated at a single meshing position through the correlation response between the two meshing positions.

[0038] The formation time of the joint sideband at the first and second meshing center frequencies is recorded according to continuous sampling periods, and the order of formation of the two joint sidebands is compared. The formation time of the joint sideband refers to the time corresponding to the first continuous sampling period when the joint sideband first meets the aforementioned joint sideband determination conditions and remains identifiable within a preset number of continuous sampling periods. When the frequency interval corresponding to the rotational frequency of the same shaft first appears within a preset number of continuous sampling periods at one meshing center frequency, and then appears within a preset number of continuous sampling periods at another meshing center frequency, the two frequency intervals are correlated. If the absolute difference between the two frequency intervals and the rotational frequency of the same shaft is within the aforementioned preset frequency difference range, and the absolute difference between the two frequency intervals is also within the preset frequency difference range, then the joint sideband formed at the two meshing positions is determined as a continuous spectral response related to the periodic action of the same drive shaft. Based on the order of formation time of the joint sideband at the first and second meshing center frequencies, and combined with the predetermined connection sequence of the two-stage gear meshing in the ball mill transmission path, the fault propagation direction between the transmission components associated with the drive shaft is determined. Specifically, the meshing position where the joint sideband is formed first is taken as the position where the fault response occurs earlier, and the meshing position where the joint sideband is formed later is taken as the position where the fault response occurs subsequently. The direction of fault response propagation from the first meshing position to the second meshing position is determined according to the positions of the two meshing positions in the actual transmission connection relationship. The process of the joint sideband being formed sequentially from one meshing position to another is used as the basis for predicting the spread of the fault along the transmission path.

[0039] Furthermore, the time difference between the formation of the combined sideband at the first and second meshing center frequencies is recorded, and the time difference is converted into the number of rotation cycles of the drive shaft based on the shaft rotation frequency of the corresponding drive shaft. For adjacent consecutive sampling periods, the corresponding number of rotation cycles is obtained, and the difference between adjacent rotation cycle numbers is calculated. When the difference is within the preset range of the number of cycle differences, it indicates that the time relationship of the formation of the combined sideband between the two meshing positions has corresponding repeatability in the continuous sampling process. The preset range of the number of cycle differences is predetermined based on the time resolution of the continuous sampling period, the speed measurement error, and the range of rotation cycle changes caused by speed fluctuations under normal operating conditions, and remains consistent in the continuous judgment of the same fault propagation process. Thus, the number of rotation cycles is used as the basis for judging whether the fault propagation direction is maintained in the continuous sampling period, and combined with the aforementioned sequential relationship of the formation time of the combined sideband, the fault propagation process of the drive shaft and related transmission components is continuously judged.

[0040] Example 1:

[0041] This embodiment provides a method for predicting faults in ball mill transmission components based on combined acoustic and vibration spectral characteristics. Acoustic signals, vibration signals, and shaft speed are synchronously acquired during the operation of the ball mill transmission components. Acoustic signals are acquired by acoustic sensors located near the transmission components to be monitored, vibration signals are acquired by vibration sensors located on the bearing housing or gearbox housing, and shaft speed is synchronously acquired by a speed sensor. All sensors use a unified acquisition clock, ensuring that the acoustic signals, vibration signals, and shaft speed are time-correlated. During continuous sampling, the same sampling frequency and duration are used, dividing the sequentially acquired data into continuous sampling periods. The acoustic and vibration signals within each continuous sampling period are processed in the frequency domain to obtain the acoustic and vibration spectra, and the frequency positions and amplitudes corresponding to each spectral peak are extracted.

[0042] Based on the transmission ratio, number of gear teeth, and rotational speed of each shaft in the ball mill transmission system, the meshing center frequency corresponding to different gear meshing positions and the shaft rotational frequency corresponding to each transmission shaft are determined, and the correspondence between meshing center frequency, shaft rotational frequency, gear meshing position, transmission shaft, and transmission components is established in advance. Taking a set of exemplary sampling data as an example, the rotational speed of the intermediate transmission shaft is 240 r / min, its shaft rotational frequency is 4.00 Hz, the number of teeth of the gears participating in the monitored gear meshing on this transmission shaft is 32, and the corresponding meshing center frequency is 128.00 Hz. Using 128.00 Hz as the frequency analysis benchmark, local spectral peaks in the sound spectrum and vibration spectrum are extracted within the preset search frequency bands on both sides. Local spectral peaks with amplitudes reaching the preset spectrum recognition threshold are identified as candidate sidebands. The preset spectrum recognition threshold is pre-determined based on the background spectrum amplitude and its fluctuation range within the same frequency range under normal ball mill operation, to exclude spectral components with excessively low amplitudes and unstable identification.

[0043] For the extracted candidate sidebands, the frequency interval between them and the corresponding meshing center frequency is determined and compared with the shaft rotation frequency of the corresponding transmission shaft. To enable the acoustic and vibration spectrum candidate sidebands located on either side of the meshing center frequency to be jointly judged under unified conditions, this embodiment sets a joint sideband matching deviation index. .in, This represents the frequency interval between candidate sidebands of the acoustic spectrum and their corresponding meshing center frequency, in Hz. This represents the frequency interval between candidate sidebands of the vibration spectrum and the corresponding meshing center frequency, in Hz. The transmission shaft rotation frequency corresponding to the candidate sideband is expressed in Hz. This represents the preset frequency difference threshold, in Hz. Since the determination of the joint sideband requires that the frequency deviations between the candidate sidebands in the acoustic spectrum and the shaft rotation frequency, the frequency deviations between the candidate sidebands in the vibration spectrum and the shaft rotation frequency, and the deviation between the frequency intervals of the two candidate sidebands all be within acceptable limits, the maximum value of these three deviations is used as the joint sideband matching deviation index, and its calculation formula is:

[0044]

[0045] Using the maximum deviation as The reason is that if any one of the three deviations exceeds the allowable range, the corresponding candidate sidebands of the acoustic spectrum and the candidate sidebands of the vibration spectrum cannot simultaneously satisfy the conditions corresponding to the same mechanical rotation characteristic. Therefore, when Not greater than the preset frequency difference threshold This indicates that the three frequency relationships mentioned above are simultaneously within a preset frequency difference range. This embodiment, based on the spectral frequency resolution, rotational speed measurement error, and the rotational speed fluctuation range under stable operating conditions, will... The value is set to 0.10 Hz. For a set of candidate sidebands near the meshing center frequency of 128.00 Hz, the candidate sidebands in the acoustic spectrum are located on the lower frequency side of the meshing center frequency, with a frequency position of 123.98 Hz, corresponding to a frequency interval of... The frequency is 4.02 Hz; the candidate sideband of the vibration spectrum is located on the high-frequency side of the meshing center frequency, with a frequency position of 132.04 Hz, corresponding to a frequency interval of 4.02 Hz. It is 4.04 Hz; corresponding to the axis rotation frequency. The frequency is 4.00 Hz. The deviation between the candidate sidebands of the acoustic spectrum and the shaft rotation frequency is 0.02 Hz, the deviation between the candidate sidebands of the vibration spectrum and the shaft rotation frequency is 0.04 Hz, and the deviation between the frequency intervals of the two candidate sidebands is 0.02 Hz. Therefore, we obtain... Since the frequency is 0.04 Hz, which is less than 0.10 Hz, the candidate sidebands of the acoustic spectrum and the candidate sidebands of the vibration spectrum located on both sides of the meshing center frequency are determined as a joint sideband candidate combination related to the same mechanical modulation effect. For example... Figure 2 As shown, the spectral peak at 128.00 Hz is used as the meshing center frequency reference, the acoustic spectrum candidate sideband at 123.98 Hz is located on its low-frequency side, and the vibration spectrum candidate sideband at 132.04 Hz is located on its high-frequency side, with frequency intervals of 4.02 Hz and 4.04 Hz on the two sides, respectively. Figure 2 By visually representing the correspondence between the acoustic spectrum, vibration spectrum, and meshing center frequency on the same frequency coordinate, it can be shown that the two candidate sidebands are located on both sides of the meshing center frequency and are close to the shaft rotation frequency of 4.00 Hz, thus reflecting the common constraint relationship between frequency position and frequency interval in the joint sideband identification process.

[0046] The joint sideband is continuously tracked during continuous sampling periods. Initially, when a candidate sideband satisfying the frequency matching condition can be stably identified only at one sideband position of the meshing center frequency, the sideband position, frequency interval, and occurrence period are recorded. As continuous sampling progresses, when a candidate sideband satisfying the judgment condition also appears at another sideband position corresponding to the same shaft rotation frequency, and the candidate sidebands on both sides of the meshing center frequency remain identifiable within a preset number of continuous sampling periods, this process is defined as the joint sideband evolving from a single-sided feature to a double-sided feature. The preset number is determined based on the duration of transient spectral peaks during normal operation. In this embodiment, the ability to identify the sideband within three consecutive sampling periods is used as a stable occurrence condition to reduce the impact of random spectral peaks in a single sampling period on the fault prediction results. The corresponding gear meshing position is determined based on the meshing center frequency forming the joint sideband, the relevant transmission shaft is determined based on the shaft rotation frequency corresponding to the joint sideband frequency interval, and the associated transmission components are determined based on the pre-established mechanical connection relationship.

[0047] When the shaft speed changes during ball mill operation, candidate sidebands are further tracked to determine whether they can migrate synchronously with the rotation of the drive shaft. Let the frequency interval of the candidate sidebands in the acoustic spectrum during the previous continuous sampling period be denoted as . The frequency interval in the subsequent consecutive sampling period is The frequency interval of the candidate sidebands of the vibration spectrum in the previous continuous sampling period is: The frequency interval in the subsequent consecutive sampling period is The axis rotation frequencies in the corresponding two consecutive sampling periods are respectively and The units for the above frequency parameters are all in Hz. Set the speed tracking deviation index. This is used to reflect the consistency between the frequency spacing changes of candidate sidebands in the sound spectrum and vibration spectrum and the frequency changes of the same transmission shaft. Its calculation formula is:

[0048]

[0049] The calculation relationship is derived from the correspondence between the candidate sideband frequency interval change and the shaft rotation frequency change. For the acoustic spectrum, the difference in frequency interval between the subsequent continuous sampling period and the previous continuous sampling period is taken as the frequency interval change of the candidate sideband, and then compared with the shaft rotation frequency change during the same period; the same processing is applied to the vibration spectrum. Larger deviations in the two comparison results constitute... ,therefore The smaller the value, the higher the degree to which candidate sidebands in the sound and vibration spectra maintain mechanical synchronization with changes in shaft rotational speed. Only when... Only when the candidate sideband is within the preset frequency difference range and the frequency offset direction of the candidate sideband is consistent with the frequency offset direction of the meshing center frequency, will the corresponding candidate sideband be retained for subsequent joint sideband tracking.

[0050] Continuing with the aforementioned exemplary sampling data, when the intermediate drive shaft speed increases from 240 r / min to 246 r / min, the shaft frequency increases from 4.00 Hz to 4.10 Hz, and the meshing center frequency of the 32-tooth gear increases from 128.00 Hz to 131.20 Hz. In the previous continuous sampling period, It is 4.02 Hz. It is 4.04 Hz. The frequency interval of the candidate sidebands of the acoustic spectrum is 4.00 Hz; in the subsequent continuous sampling period, the frequency interval of the candidate sidebands of the acoustic spectrum is 4.00 Hz. The frequency spacing of the candidate sidebands of the vibration spectrum is 4.11 Hz. The shaft rotation frequency is 4.13 Hz. The frequency is 4.10 Hz. At this time, the frequency position of the candidate sideband of the acoustic spectrum on the low-frequency side is 127.09 Hz, and the frequency position of the candidate sideband of the vibration spectrum on the high-frequency side is 135.33 Hz. Compared with the previous continuous sampling period, the frequency position of the candidate sideband of the acoustic spectrum increased from 123.98 Hz to 127.09 Hz, and the frequency position of the candidate sideband of the vibration spectrum increased from 132.04 Hz to 135.33 Hz. Both have the same frequency offset direction from the meshing center frequency. The frequency interval of the candidate sideband of the acoustic spectrum increased by 0.09 Hz, the frequency interval of the candidate sideband of the vibration spectrum increased by 0.09 Hz, and the shaft rotation frequency increased by 0.10 Hz. Substituting the above data into... The calculation formula shows that both deviations are 0.01 Hz, therefore The value is 0.01 Hz, which is less than 0.10 Hz. Therefore, the corresponding candidate sidebands are retained and used for joint sideband formation and fault development tracking. Figure 3 The paper further illustrates the continuous migration of three frequency positions as the drive shaft speed increases sequentially from 240 r / min to 242 r / min, 244 r / min, and 246 r / min. The meshing center frequency gradually increases from 128.00 Hz to 131.20 Hz, and the candidate sidebands on the low side of the acoustic spectrum and the candidate sidebands on the high side of the vibration spectrum also continuously shift along the same frequency-increasing direction. Figure 3 The continuity of the three frequency trajectories shows that the candidate sidebands do not remain at fixed frequency positions, but migrate with the meshing center frequency while maintaining a frequency interval matching the corresponding shaft rotation frequency. Therefore, it is possible to distinguish such spectral peaks with mechanical synchronous migration characteristics from fixed-frequency interference or random spectral peaks. For frequency offset directions inconsistent with the meshing center frequency, or... Candidate sidebands that exceed the preset frequency difference range will not be used for subsequent joint sideband formation.

[0051] After completing the rotational speed tracking of the candidate sidebands, the order of their appearance in the acoustic and vibration spectra is recorded according to continuous sampling periods. When a candidate sideband corresponding to the same meshing center frequency and the same shaft rotation frequency remains identifiable in the acoustic spectrum for three consecutive sampling periods, and the frequency interval between it and the corresponding meshing center frequency always maintains a preset frequency difference range with the same shaft rotation frequency, and subsequently, the corresponding candidate sideband appears in the vibration spectrum, and the absolute difference between the frequency interval of the later-appearing candidate sideband and the frequency interval of the earlier-appearing candidate sideband is still within a preset frequency difference range of 0.10 Hz, the appearance process in both spectra is determined as successive appearance. If the candidate sideband first appears stably in the vibration spectrum and then in the acoustic spectrum, the same judgment principle is applied. After successive appearance is determined, the aforementioned rotational speed tracking relationship is used for verification. When the frequency interval change of the candidate sideband before and after successive appearance can maintain a correspondence with the shaft rotation frequency change, the successive appearance is determined as a sign of fault development, reducing the possibility of occasional spectral components being incorrectly judged as continuous mechanical responses.

[0052] After the candidate sidebands appear sequentially, the amplitude and frequency position of the candidate sidebands in both spectra are recorded. Using the stable amplitude of the spectrum where the candidate sideband first appears before the sequential appearance is formed as a comparison benchmark, a preset amplitude change threshold is determined based on the amplitude fluctuation range at the same frequency position under normal operating conditions. For example, if the amplitude fluctuation of the corresponding spectrum peak is within 6% under normal operating conditions, 12% can be set as the preset amplitude change threshold in this embodiment. To further illustrate the continuous process of sequential appearance and cross-spectral response migration, Figure 4 The change in the relative amplitude of candidate sidebands over 10 consecutive sampling periods is used to represent the change. The candidate sidebands in the acoustic spectrum remain near the baseline amplitude in the first three sampling periods. The candidate sidebands in the vibration spectrum become identifiable from the fourth sampling period onwards and remain continuously present from the fourth to the sixth sampling periods, thus forming a continuous display that meets the preset requirement for the number of consecutive samples. In subsequent sampling periods, the relative amplitude of the candidate sidebands in the acoustic spectrum gradually decreases, while the relative amplitude of the candidate sidebands in the vibration spectrum gradually increases. In the eighth sampling period, the relative amplitude of the candidate sidebands in the acoustic spectrum drops to 88%, a 12% decrease relative to the 100% baseline, reaching the preset amplitude change threshold. Simultaneously, the corresponding candidate sidebands in the vibration spectrum remain identifiable for multiple consecutive sampling periods. Therefore, the cross-spectral response migration judgment stage begins from the eighth sampling period. When the frequency difference between the candidate sidebands in the two spectra remains within the preset frequency difference range of 0.10 Hz, this change process is defined as cross-spectral response migration. Figure 4The decreasing amplitude of the candidate sidebands in the acoustic spectrum and the increasing amplitude of the candidate sidebands in the vibration spectrum form a mutually connected trend, indicating that the spectral response related to the same mechanical modulation effect can gradually shift from a more obvious manifestation in one signal type to a more obvious manifestation in another signal type. Therefore, this can be regarded as a further sign of the continued development of the fault.

[0053] To distinguish between continuous mechanical responses caused by local faults and transient spectral responses caused by random impacts, after the successive appearance of candidate sidebands in the acoustic and vibration spectra, the starting time periods corresponding to the first three consecutive sampling periods in each spectrum are recorded. The order of the starting time periods of the candidate sidebands in the two spectra is then compared over multiple consecutive sampling periods. If the order of the acoustic candidate sideband appearing first and the vibration candidate sideband appearing later remains consistent across multiple consecutive sampling periods, or if the order of the vibration candidate sideband appearing first and the acoustic candidate sideband appearing later remains consistent, the spectral response is determined to have a repeatable temporal evolution relationship and is identified as a continuous mechanical response related to a local fault in the transmission component. Spectral responses that appear only in a few consecutive sampling periods, do not meet the stable identification condition of three consecutive sampling periods, or do not have a fixed order of appearance between the two spectra are distinguished from the continuous mechanical responses to reduce the interference of transient spectral responses caused by random impacts on fault diagnosis.

[0054] comprehensive Figure 2 The joint matching relationship between the candidate sidebands of acoustic vibration and the meshing center frequency is shown. Figure 3 The candidate sideband synchronization migration relationship shown under variable speed conditions and Figure 4 The sequential manifestation and cross-spectral response migration process shown in the diagram continuously judges the fault development state of the corresponding transmission components. When the combined sideband evolves from a single-sided feature to a double-sided feature, the candidate sideband can maintain synchronous frequency migration with the shaft speed change, and when a sequential manifestation with a stable order and a cross-spectral response migration satisfying the amplitude change condition further appear between the acoustic spectrum and the vibration spectrum, the gear meshing position is determined according to the meshing center frequency corresponding to the combined sideband, the relevant transmission shaft is determined according to the shaft rotation frequency corresponding to the frequency interval of the combined sideband, and the fault prediction object is determined according to the meshing position, transmission shaft and mechanical connection relationship between related transmission components in the ball mill transmission system. The fault prediction result of the corresponding transmission component is output, thereby improving the reliability of fault prediction of ball mill transmission components by utilizing the common modulation characteristics, speed correlation characteristics and continuous evolution relationship of acoustic signal and vibration signal in the frequency domain.

[0055] Example 2:

[0056] This embodiment provides a fault prediction method for ball mill transmission components based on the combined acoustic and vibration spectral characteristics. For a transmission structure where the same drive shaft simultaneously participates in two stages of gear meshing, the associated spectral characteristics generated at the two meshing positions are continuously tracked. Acoustic signals, vibration signals, and drive shaft rotation speed are simultaneously acquired during ball mill operation, obtaining acoustic and vibration spectra respectively. The corresponding meshing center frequency and shaft rotation frequency are determined based on the number of teeth in each gear stage, the transmission relationship, and the drive shaft rotation speed. For each meshing center frequency, following the combined sideband identification principle described in Embodiment 1, the combined sideband corresponding to the same shaft rotation frequency is determined from the acoustic and vibration spectra. This embodiment does not repeat the judgment process of continuous manifestation and cross-spectral response migration.

[0057] Using a set of exemplary sampling data, the rotational speed of a certain intermediate drive shaft is 240 r / min, corresponding to a shaft frequency of 4.00 Hz. The first gear meshing position on this drive shaft corresponds to a gear with 32 teeth and a first meshing center frequency of 128.00 Hz; the second gear meshing position corresponds to a gear with 40 teeth and a second meshing center frequency of 160.00 Hz. 240 r / min is converted to 4.00 Hz. The meshing center frequency corresponding to 32 teeth and a 4.00 Hz shaft frequency is 128.00 Hz, and the meshing center frequency corresponding to 40 teeth and a 4.00 Hz shaft frequency is 160.00 Hz. Based on the actual mechanical connection relationship of the ball mill transmission system, the connection order of the first and second gear meshing positions in the transmission path is predetermined, and the correspondence between the two meshing positions, the intermediate drive shaft, and the associated gears and bearings is established.

[0058] Joint sidebands are identified near the first and second meshing center frequencies, and the frequency spacing between candidate sidebands in each joint sideband and the corresponding meshing center frequency is calculated. In this embodiment, the frequency spacing of the joint sidebands at the first meshing center frequency... The frequency spacing of the joint sideband at the second meshing center frequency is 4.03 Hz. The shaft rotation frequency of the same drive shaft is 3.97 Hz. The frequency is 4.00 Hz. The preset frequency difference threshold is 0.10 Hz. and The absolute difference between them is 0.03 Hz. and The absolute difference between them is 0.03 Hz. and The absolute difference between them is 0.06 Hz, and the frequency differences of all three terms are less than 0.10 Hz. Therefore, the joint sideband at the two meshing positions is determined as the associated spectral response related to the periodic rotation characteristics of the same drive shaft. Figure 5As shown, using the same shaft rotation frequency of 4.00 Hz as the frequency reference, the joint sideband frequency intervals of the first and second meshing positions are distributed at 4.03 Hz and 3.97 Hz, respectively, both within a preset frequency difference range centered on the shaft rotation frequency. The two frequency intervals in the figure are located on either side of the shaft rotation frequency and maintain a small deviation from it, while the frequency difference between them still satisfies the preset frequency difference condition. This directly reflects the frequency correlation between the two meshing positions when subjected to the periodic action of the same drive shaft.

[0059] After satisfying the above three frequency difference conditions, in order to further quantify the overall correlation and consistency of the rotational characteristics of the two meshing positions relative to the same transmission shaft, this embodiment sets a dual meshing position correlation deviation index. .in, This represents the overall correlation deviation of the joint sidebands at two meshing positions, expressed in Hz. The frequency interval of the joint sidebands at the first engagement center frequency is expressed in Hz. This indicates the frequency interval of the joint sideband at the second meshing center frequency, in Hz. This represents the rotational frequency of the same transmission shaft involved in the two-stage gear meshing, measured in Hz. The average of the squares of the deviations of the first frequency interval relative to the shaft rotational frequency, the deviations of the second frequency interval relative to the shaft rotational frequency, and the deviation between the two frequency intervals is calculated. The square root of this average is then taken to ensure that the final correlation deviation index retains the dimension of Hz. The formula is:

[0060]

[0061] The This does not replace the fundamental judgment of whether the aforementioned three absolute differences meet the preset frequency difference condition, but rather characterizes the degree of consistency of the two correlated spectral responses after the fundamental condition has been met. 4.03 Hz 3.97 Hz and Substituting 4.00 Hz, the three deviations are 0.03 Hz, 0.03 Hz, and 0.06 Hz, with corresponding squared values ​​of 0.0009 Hz², 0.0009 Hz², and 0.0036 Hz². The sum of the three squared values ​​is 0.0054 Hz², and the average value is 0.0018 Hz². Taking the square root of this average value yields... It is approximately 0.0424 Hz. Figure 5The correlation deviation result is also given. The value is less than the preset frequency difference threshold and is a further quantification result after all three original frequency differences meet the judgment conditions. Based on this, the intermediate transmission shaft and the transmission components associated with the meshing positions of the first gear and the second gear can be included in the scope of fault prediction.

[0062] During the continuous sampling process of the transmission shaft speed change, the frequency shift and frequency interval change of the joint sideband at the first and second meshing center frequencies were continuously tracked. As the intermediate transmission shaft speed gradually increased from 240 r / min to 242 r / min, 244 r / min, and 246 r / min, the corresponding shaft frequency increased continuously. When the transmission shaft speed was 240 r / min, the shaft frequency was 4.00 Hz, the frequency interval at the first meshing position was 4.03 Hz, and the frequency interval at the second meshing position was 3.97 Hz. When the transmission shaft speed increased to 246 r / min, the shaft frequency was 4.10 Hz, the frequency interval at the first meshing position was 4.12 Hz, and the frequency interval at the second meshing position was 4.08 Hz. Correspondingly, the first meshing center frequency increased from 128.00 Hz to 131.20 Hz, and the second meshing center frequency increased from 160.00 Hz to 164.00 Hz. In the subsequent continuous sampling period, the absolute difference between the first frequency interval and the 4.10 Hz axis rotation frequency was 0.02 Hz, the absolute difference between the second frequency interval and the 4.10 Hz axis rotation frequency was 0.02 Hz, and the absolute difference between the two frequency intervals was 0.04 Hz. All three results were less than the preset frequency difference threshold of 0.10 Hz.

[0063] like Figure 6 As shown, during the continuous increase of the drive shaft speed from 240 r / min to 246 r / min, the rotational frequency of the same shaft, the frequency interval of the first meshing position, and the frequency interval of the second meshing position all show a continuous upward trend. In the figure, the frequency intervals of the first meshing position are successively 4.03 Hz, 4.06 Hz, 4.09 Hz, and 4.12 Hz; the frequency intervals of the second meshing position are successively 3.97 Hz, 4.01 Hz, 4.04 Hz, and 4.08 Hz; and the rotational frequency of the same shaft gradually increases from 4.00 Hz to 4.10 Hz. The three frequency change trajectories maintain similar migration directions and amplitudes during the continuous speed change, indicating that the frequency intervals of the two joint sidebands are not independent random changes, but rather maintain a continuous correspondence with the rotational state of the same drive shaft.

[0064] Compared to the sampling results at a drive shaft speed of 240 r / min, the frequency interval at the first engagement position changed from 4.03 Hz to 4.12 Hz, a change of 0.09 Hz; the frequency interval at the second engagement position changed from 3.97 Hz to 4.08 Hz, a change of 0.11 Hz; and the shaft rotation frequency changed from 4.00 Hz to 4.10 Hz, a change of 0.10 Hz. Therefore, the absolute difference between the change in frequency interval at the first engagement position and the change in shaft rotation frequency is 0.01 Hz, and the absolute difference between the change in frequency interval at the second engagement position and the change in shaft rotation frequency is also 0.01 Hz. Both results are within the preset frequency difference range of 0.10 Hz. Figure 6 The continuous trajectory of the frequency interval between the two points in the middle migrating synchronously with the shaft rotation frequency determines the frequency migration as a fault symptom caused by the simultaneous action of periodic disturbances of the same transmission shaft on the meshing of two gears.

[0065] The formation time of the joint sideband at the first and second meshing center frequencies is recorded according to continuous sampling periods. The joint sideband formation time is determined as the time corresponding to the first sampling period in which the joint sideband at the corresponding meshing position first satisfies the frequency difference condition and is continuously maintained for a preset number of sampling periods. This determination method can reduce the influence of occasional spectral responses that only occur in a single sampling period on the formation time judgment. In this embodiment, the condition is maintained for three consecutive sampling periods as the stable formation condition of the joint sideband. Taking the relative sampling time in the continuous sampling process as an example, the formation time of the joint sideband at the first meshing center frequency is 42.0 s, the formation time of the joint sideband at the second meshing center frequency is 43.5 s, and the time difference between the two joint sideband formation times is 1.5 s. According to the pre-established mechanical connection relationship, the first gear meshing position is located at the stage before the transmission path, and the second gear meshing position is located at the stage after it. When the joint sideband of the first meshing position is stably formed first, followed by the stable formation of the joint sideband of the second meshing position, and the frequency interval at both positions maintains the frequency difference relationship with the rotational frequency of the same shaft, the first meshing position is determined as the position where the associated fault response occurs earlier, and the second meshing position is determined as the position where the associated fault response occurs later. The formation time relationship is combined with the mechanical connection order of the two meshing positions in the actual transmission path to determine that the fault response propagates from the first meshing position to the second meshing position.

[0066] To further verify the repeatability of the time relationship between the formation of the joint sideband at the two meshing positions, the time difference between the formation of the joint sideband at the two locations is converted into the number of rotation cycles of the drive shaft. This embodiment sets the number of rotation cycles. ,in, This indicates the number of rotational cycles the drive shaft undergoes from the formation of one joint sideband to the formation of another joint sideband; The time difference between the formation time of the joint sideband at the first meshing center frequency and the second meshing center frequency is expressed in seconds. This represents the shaft rotation frequency of the corresponding drive shaft, measured in Hz. The physical meaning of shaft rotation frequency is the number of rotation cycles the drive shaft completes per unit time. Therefore, multiplying the time difference by the shaft rotation frequency yields the number of rotation cycles the drive shaft experiences within that time range. The calculation formula is:

[0067]

[0068] Will For 1.5 s and Substituting 4.00 Hz, we get The value is 6, indicating that the transmission shaft undergoes 6 rotation cycles from the stable formation of the joint sideband at the first engagement position to the stable formation of the joint sideband at the second engagement position. This indicates the number of rotation cycles of the drive shaft, but does not indicate the number of gear meshing times, the number of consecutive samples, or the fault level. For example... Figure 7 As shown, the initial 6.0 rotation cycles are used as a reference for continuous verification. Subsequent continuous sampling processes yield 5.9, 6.1, and 6.0 rotation cycles respectively. The figure also shows the sequential direction of the response from the first engagement position to the second engagement position. To facilitate the display of the fluctuation range of the number of cycles during continuous sampling, the figure shows the corresponding allowable fluctuation area centered on 6.0 cycles. However, the actual judgment is still based on whether the difference between the number of rotation cycles obtained from adjacent continuous sampling processes meets the preset range of cycle difference.

[0069] In subsequent continuous sampling, the difference between 6.0 and 5.9 is 0.1 cycles, the difference between 5.9 and 6.1 is 0.2 cycles, and the difference between 6.1 and 6.0 is 0.1 cycles. If, based on the time resolution of the continuous sampling period, the speed measurement error, and the change in the number of cycles caused by speed fluctuations under normal operating conditions, the preset range of cycle number differences is determined to be no greater than 0.2 cycles, then the above-mentioned differences between adjacent cycle numbers all satisfy the preset range of cycle number differences. Combined with... Figure 7 The continuous sampling results all remained around 6 rotation cycles, which determined that the time relationship between the formation of the joint sideband between the first and second meshing positions was repeatable. Therefore, the order of the joint sideband formation time was used as the basis for judging the fault propagation direction during continuous sampling.

[0070] Using the above method, under the condition that the same transmission shaft participates in the meshing of two stages of gears, by utilizing Figure 5 The frequency correlation between the two meshing positions and the rotational frequency of the same shaft defines the associated transmission components, utilizing... Figure 6 The synchronous migration relationship between the frequency interval at two points and the shaft rotation frequency under the shown variable speed condition verifies the same periodic mechanical disturbance, and then uses... Figure 7 The continuous repetition relationship between the joint sideband formation time and the number of rotation cycles shown verifies the stability of the fault response sequence. Combined with the mechanical connection sequence of the two meshing positions in the actual transmission path, the fault propagation direction and the trend of expansion along the transmission path of the transmission shaft-related transmission components are continuously predicted.

Claims

1. A method for predicting faults in ball mill transmission components based on combined acoustic and vibration spectral characteristics, characterized in that: The sound and vibration signals of the ball mill transmission components during operation are collected synchronously to obtain the sound spectrum and vibration spectrum. The meshing center frequency and the rotational frequency of each shaft are calculated based on the transmission ratio, number of teeth and rotational speed of each shaft. Candidate sidebands located on both sides of the same meshing center frequency are extracted from the acoustic spectrum and vibration spectrum respectively. The frequency interval between the candidate sidebands and the meshing center frequency is calculated. The acoustic spectrum candidate sidebands and vibration spectrum candidate sidebands located on both sides of the meshing center frequency, whose absolute difference between each frequency interval and the same shaft rotation frequency is within a preset frequency difference range, and whose absolute difference between two frequency intervals is within the preset frequency difference range, are determined as joint sidebands formed by the same mechanical modulation. The joint sideband is tracked during a continuous sampling period. When the joint sideband changes from appearing at one side frequency position to appearing at two side frequency positions within a preset number of continuous sampling periods, the transmission component is determined based on the meshing center frequency and shaft rotation frequency that form the joint sideband, and the fault prediction result of the transmission component is output.

2. The prediction method according to claim 1, characterized in that: The candidate sidebands are tracked during continuous sampling periods when the shaft rotation speed changes; when the meshing center frequency shifts with the shaft rotation speed, the candidate sidebands whose frequency shift direction is the same as that of the meshing center frequency and whose absolute difference between the frequency interval change and the shaft rotation frequency change is within the preset frequency difference range are retained, and the joint sideband is formed by the retained candidate sidebands.

3. The prediction method according to claim 1, characterized in that: The candidate sidebands are tracked in the order of appearance in the acoustic and vibration spectra according to continuous sampling periods. When the absolute difference between the frequency interval of the candidate sideband and the meshing center frequency and the rotational frequency of the same shaft is within the preset frequency difference range, and the candidate sideband first appears in one spectrum within a preset number of continuous sampling periods, and then appears at a frequency position in another spectrum, and the absolute difference between the frequency interval of the frequency position and the meshing center frequency and the frequency interval of the first candidate sideband is within the preset frequency difference range, the appearance process in the two spectra is determined as successive appearance, and the successive appearance is taken as a sign of fault development.

4. The prediction method according to claim 1, characterized in that: For two-stage gear meshing involving the same drive shaft, the first meshing center frequency and the second meshing center frequency are determined respectively. When a joint sideband is formed at the first meshing center frequency and the second meshing center frequency, and the absolute difference between the frequency interval of the two joint sidebands and the rotational frequency of the same shaft is within the preset frequency difference range, and the absolute difference between the two frequency intervals is within the preset frequency difference range, the drive shaft and the transmission components associated with the two-stage gear meshing are identified as fault prediction objects.

5. The prediction method according to claim 3, characterized in that: When the candidate sideband appears consecutively in another spectrum, the frequency interval change of the candidate sideband relative to the meshing center frequency before and after the consecutive appearance is compared with the shaft rotation frequency change; when the absolute difference between the two changes is within the preset frequency difference range, the consecutive appearance is determined as a sign of fault development.

6. The prediction method according to claim 3, characterized in that: After the candidate sidebands are displayed sequentially, the amplitude and frequency position of the candidate sidebands in the two spectra are tracked in subsequent continuous sampling periods. When the amplitude decrease of the sideband in the spectrum where the candidate sideband appears first reaches a preset amplitude change threshold, and the candidate sidebands in the other spectrum appear in a preset number of continuous sampling periods, and the absolute difference between the frequency intervals of the two is within the preset frequency difference range, the change process is determined as a cross-spectrum response migration and is used as a sign of fault development.

7. The prediction method according to claim 3, characterized in that: When the candidate sideband appears successively between the acoustic spectrum and the vibration spectrum, the starting time when the candidate sideband first appears in both spectra within a preset number of consecutive sampling periods is recorded. Within a preset number of consecutive sampling periods, when the order of the starting time periods is the same, the order of the starting time periods is used to distinguish between the continuous mechanical response caused by a local fault in the transmission component and the transient spectral response formed by random impact.

8. The prediction method according to claim 4, characterized in that: When the rotational speed of the drive shaft changes, the frequency shift and frequency interval change of the joint sideband at the first meshing center frequency and the second meshing center frequency are tracked. When the absolute difference between the change in frequency interval at the two locations and the change in the rotational frequency of the same shaft are respectively within the preset frequency difference range, and the absolute difference between the two frequency intervals is within the preset frequency difference range, the frequency shift is determined as a fault symptom of the drive shaft generating periodic disturbances and acting on the meshing of the two-stage gears.

9. The prediction method according to claim 4, characterized in that: The formation time of the joint sideband at the first meshing center frequency and the second meshing center frequency is recorded according to the continuous sampling period. When the frequency interval first appears at one meshing center frequency within a preset number of continuous sampling periods, and then appears at another meshing center frequency within a preset number of continuous sampling periods, and the absolute difference between the two frequency intervals and the rotational frequency of the same shaft is within the preset frequency difference range, the fault propagation direction of the transmission components associated with the transmission shaft is determined according to the order of the formation time, and the formation process is used as the basis for predicting the expansion of the fault along the transmission path.

10. The prediction method according to claim 9, characterized in that: Record the time difference formed by the combined sideband at the first meshing center frequency and the second meshing center frequency, and convert the time difference into the number of rotation cycles according to the shaft rotation frequency of the transmission shaft; Calculate the difference in the number of rotation cycles within adjacent consecutive sampling periods. When the difference is within a preset range of cycle number differences, the number of rotation cycles is determined as the criterion for determining whether the fault propagation direction remains within the consecutive sampling period.

Images