A method for identifying based on electrostatic signals of lubricating oil abrasive particles

Abstract

The application relates to the technical field of abrasive particle electrostatic signal monitoring, and discloses a method and system for identifying abrasive particles based on lubricating oil electrostatic signals, which comprise a sensor, a metal probe, an insulating tube, a shielding cover and a collection circuit and the like modules; the internal insulating tube of the sensor is a thin ceramic tube, oil passes through the ceramic tube, and the sensor spiral probe is attached to the surface of the ceramic tube; the application provides a real-time abrasive particle identification method based on variational mode decomposition and Hausdorff distance measurement; firstly, a signal enhancement model based on variational mode decomposition and envelope signal correlation criteria is constructed, the possible abrasive particle signal pulses are enhanced through mode component optimization reconstruction; then, the abrasive particle signals are further confirmed based on the double-channel probe signal and the Hausdorff distance measurement method; finally, verification experiments are carried out through an oil charging experiment device and a gear reducer lubricating oil online monitoring platform, and the diagnosis of fault components is realized.

Description

Technical Field

[0001] This invention belongs to the field of abrasive electrostatic signal monitoring technology, specifically a method and system for identifying abrasive electrostatic signals based on lubricating oil. Background Technology

[0002] Heavy machinery such as scraper conveyors and coal mining machines are core equipment in coal mining. Due to their complex structure and heavy workload, they require high-quality lubricating oil in key transmission components such as gear reducers to reduce wear on friction pairs and remove abrasive particles, ensuring healthy equipment operation. Therefore, lubricating oil typically carries abrasive particles, which contain information about the wear level or early failure of transmission components. Real-time monitoring of these abrasive particles allows for the detection of wear conditions, which is crucial for equipment maintenance.

[0003] Currently, the main methods for abrasive particle analysis include offline sampling based on spectroscopy and ferrography, and online monitoring based on inductance and capacitance principles. Offline sampling methods offer high accuracy but rely heavily on experience and cannot provide real-time feedback on abrasive particle information. Inductance and capacitance monitoring technologies are used in various industries, but because they require active voltage excitation of sensitive components, their instantaneous voltage amplitude can easily exceed safety standards, making them difficult to apply in coal mine environments.

[0004] In recent years, electrostatic monitoring technology for lubricating oil abrasive particles has become a research hotspot. Based on the principle of electrostatic induction, this technology can detect abrasive particles in the early stage of wear in real time, and has higher sensitivity and fault early warning capabilities.

[0005] The University of Southampton has conducted research on abrasive electrostatic monitoring technology from aspects such as abrasive charging mechanism, sensor design, and bench experiments; Harvey discovered that abrasive particles generated during bearing scuffing wear carry a positive charge, and the magnitude of the charge is directly related to the total volume loss; Wood studied the phenomenon of friction charging of lubricating oil and found that the flow of lubricating oil carries a certain charge; Morris et al. studied the influence of friction materials on the charge mechanism and explained the main reason why electrostatic signals appear earlier than scuffing failures. Regarding sensing models, Xu et al. established a finite element model of a ring-shaped electrostatic sensor and studied its spatial sensitivity, spatial filtering effect, and time-frequency response characteristics. Yan et al. performed mathematical analysis on the spatial distribution of the electrostatic field of point charges, revealing that the sensor's sensitivity is related to the magnitude of the charge and the particle position. Based on the ring-shaped electrode charge model, Rahmat et al. revealed the relationship between the particle size and the signal. ZUO et al. conducted research on sensing mechanisms and sensing circuits, performed model simulation and optimization for electrostatic sensors, and developed a hardware and software prototype system. MAO conducted exploratory experiments in a laboratory environment using a pin-disc wear tester, initially demonstrating a positive correlation between wear state and electrostatic signal amplitude, and preliminarily verifying the feasibility of electrostatic monitoring technology.

[0006] Previous research has largely focused on the electrostatic induction mechanism and sensing models. How to apply this technology to the monitoring of abrasive particles in gear lubricating oil of mining equipment to achieve accurate abrasive particle statistics still requires further research. Furthermore, since electrostatic signals are easily interfered with by noise, finding effective methods to reduce external noise interference, extract the signals formed by abrasive particles, and further accurately identify the signals to achieve real-time tracking of abrasive particles is another challenge. To address this difficulty, this invention proposes a real-time abrasive particle identification method based on variational mode decomposition and Hösdorf distance metric. First, a signal enhancement model based on variational mode decomposition and envelope signal correlation criteria is constructed. Modal component optimization and reconstruction enhance any potential abrasive particle signal pulses. Subsequently, the abrasive particle signals are further confirmed using a dual-channel probe signal and the Hösdorf distance metric method. Finally, verification experiments are conducted using an oil charging experimental device and a gear reducer lubricating oil online monitoring platform to diagnose faulty components. Summary of the Invention

[0007] In view of the above situation and to overcome the defects of the prior art, the present invention provides a method and system for identifying electrostatic signals of lubricating oil abrasive particles, which effectively solves the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying wear particles based on electrostatic signals of lubricating oil, comprising the following steps: Step 1: Denoising and enhancing the possible abrasive signals. Considering the noise characteristics and computation time, an abrasive signal enhancement model based on variational mode decomposition (VMD) and envelope signal correlation is proposed. Step 2: The VMD decomposition results show that the pulse waveform caused by abrasive particles has an obvious impact pattern. Typical abrasive particle waveforms can be seen in the original signal. Based on this characteristic, when reconstructing the IMF, it is necessary to first consider screening out the VIMF components with impact patterns. Step 3: Use the kurtosis criterion to filter and select the two IMF components with the highest kurtosis as the initial screening results; Step 4: Calculate the correlation coefficient between the envelope signal and the IMF obtained from VMD decomposition.

[0009] Preferably, step one further includes: the VMD algorithm can adaptively decompose the electrostatic signal, find several optimal IMFs through decomposition, and hope that the sum of the spectral bandwidths of these IMFs will be minimized, with the constraint that the sum of the time domains of each IMF is equal to the input signal. The VMD algorithm is well-suited for processing signals containing several main modes, and wear-particle electrostatic signals with multiple types of noise also have similar characteristics.

[0010] Preferably, step three further includes: kurtosis is sensitive to impact signals, the kurtosis coefficient of random white noise is about 3, and the kurtosis value increases rapidly when the abrasive pulses increase. In order to avoid missing small-amplitude impact components, the present invention sorts the 6 VIMF components. In order to enhance the abrasive signal by leveraging the morphological similarity, a second screening is performed using the correlation between the envelope signal of the original signal and the IMF as a criterion.

[0011] Preferably, step four further includes: if the absolute value of the correlation coefficient is greater than 0.7, then the sequences are considered to be strongly correlated; The correlation coefficients between the two IMFs and the envelope of the original signal are calculated, and the IMFs with a value greater than 0.7 are selected as the reconstructed components. If all of them are greater than 0.7, they are all selected as the reconstructed components. The correlation coefficients for IMF5 and IMF6 were calculated to be greater than 0.7, so they were ultimately used as reconstructed components, and the possible abrasive signals were significantly enhanced.

[0012] Preferably, step one further includes: after acquiring the enhanced signal, the wear particle signal needs to be identified using a model. A wear particle identification model based on sliding window cross-correlation and peak protrusion is used. First, based on the sensor structural parameters and oil flow velocity, a simulated sliding window signal for identification is constructed. Considering that the wear particles may have different polarities, a charge signal is used.

[0013] Preferably, after obtaining the enhanced signal, the method further includes: integrating the obtained enhanced signal to obtain the original charge signal, setting the sliding window to step by 10 sampling points each time, and calculating the cross-correlation sequence between the sliding window signal and the original charge signal within the sliding window range each time. The cross-correlation function can effectively measure the similarity between two time series waveforms. The cross-correlation maximum value obtained after the sliding window traverses the entire signal, after normalization, yields four large peaks in the cross-correlation function. Obviously, the four large peaks are largely caused by the four possible abrasive pulses. Some small peaks can still be seen, so the number of abrasive particles cannot be determined solely by identifying peaks. A specific method is needed to filter out the four larger peaks. Observation revealed that, due to the abrasive particles having a certain velocity, approximately equal to the velocity of the oil flow, the peaks of the cross-correlation maximum value image caused by the abrasive particles have a significantly steeper characteristic. Therefore, this invention selects peak significance as the overall discrimination criterion. Peak prominence can be used to filter out those peaks with relatively flat shapes to avoid interference. Based on MATLAB software, the peak finding results of the cross-correlation function maximum value curve are obtained. It is necessary to consider setting a threshold for judging abrasive particles. When the significance of a peak is higher than the threshold, it is considered that it is very likely caused by abrasive particles, while if it is lower than the threshold, it is not identified. The threshold is set to 1.5 times the interquartile range of the significance of all peaks. Points higher than the threshold are identified as outliers. Through outliers, the specific location of the peak in the cross-correlation function graph can be located.

[0014] This invention also provides an abrasive electrostatic monitoring system based on the aforementioned method for identifying abrasive electrostatic signals from lubricating oil particles. The system includes modules such as a sensor, a metal probe, an insulating tube, a shielding cover, and a data acquisition circuit. The internal insulating tube of the sensor is a thin ceramic tube through which oil flows, and the sensor's spiral probe is attached to the surface of the ceramic tube. The shielding cover is partially grounded to protect the sensor from external electromagnetic interference. The induction probe is connected to a signal conditioning circuit, which converts the charge signal into a voltage signal for acquisition. After data analysis, deeper information is obtained.

[0015] Preferably, the metal probe adopts a dual-probe front and rear array design.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. The algorithm proposed in this invention can effectively identify electrostatic signal pulses caused by abrasive particles, providing a powerful means for monitoring abrasive particles in the lubricating oil of mining equipment reducers.

[0017] 2. This invention proposes a real-time abrasive particle identification method based on variational mode decomposition and Hösdorf distance metric. First, a signal enhancement model based on variational mode decomposition and envelope signal correlation criterion is constructed. The possible abrasive particle signal pulses are enhanced through modal component optimization and reconstruction. Subsequently, the abrasive particle signal is further confirmed based on dual-channel probe signal and Hösdorf distance metric method. Finally, verification experiments are carried out through oil charging experimental device and gear reducer lubricating oil online monitoring platform to realize the diagnosis of faulty parts.

[0018] 3. This invention is a passive monitoring technology that does not require voltage excitation, which meets the limitations of its use in gearbox components of underground mining equipment and has great potential. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0020] In the attached diagram: Figure 1 This is a schematic diagram of the gearbox acquiring signals; Figure 2 This is a schematic diagram of the VMD decomposition results; Figure 3 This is a schematic diagram of the kurtosis results; Figure 4 This is a schematic diagram of the original signal envelope; Figure 5 This is a schematic diagram showing the correlation coefficient between the envelope signal of the original signal and each IMF; Figure 6 This is a schematic diagram of the enhanced signal; Figure 7 This is a schematic diagram of the calculation results of the overall cross-correlation function of the signal; Figure 8 This is a schematic diagram of the significance of the peaks in the cross-correlation function curve; Figure 9 This is a schematic diagram illustrating outliers in peak significance. Figure 10 This is a schematic diagram of the front and rear installation of a classic sensor; Figure 11 This is a schematic diagram of the abrasive signal enhancement algorithm of the present invention; Figure 12 This is a flowchart illustrating the fragment signal identification algorithm of the present invention. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] Depend on Figure 1-12 The present invention provides a method for identifying electrostatic signals of lubricating oil wear particles, which can passively monitor without voltage excitation and is highly sensitive to early-stage wear failure products. The debris generated by friction wear often carries a certain amount of charge in the lubricating oil. When the wear particles pass through the inside of the electrostatic probe, an electrostatic field is formed between the wear particle charge and the sensor probe, and they interact. Electrons on the probe redistribute as the charged object moves to balance the nearby electrostatic field, resulting in an induced current inside the probe. This induced current is converted into an output voltage by a conditioning circuit, further filtered and amplified, and then output as an electrostatic signal. Since electrostatic signals are easily affected by complex noise interference, such as… Figure 1 Directly identifying abrasive particles from the raw signal acquired by the gearbox would reduce accuracy. This invention first denoises and enhances the potential abrasive particle signals. Considering noise characteristics and computation time, a abrasive particle signal enhancement model based on Variational Mode Decomposition (VMD) and envelope signal correlation is proposed. The VMD algorithm adaptively decomposes the electrostatic signal, finding several optimal Integrated Mode Factors (IMFs) and aiming to minimize the sum of the spectral bandwidths of these IMFs. The constraint is that the time-domain superposition of the IMFs equals the input signal. This algorithm is well-suited for processing signals containing several main modes, and abrasive particle electrostatic signals with multiple types of noise also exhibit similar characteristics.

[0023] like Figure 2 The VMD decomposition results show that the pulse waveform caused by abrasive particles has an obvious impact morphology. Typical abrasive particle waveform morphology can be seen in the original signal. Based on this characteristic, when reconstructing the IMF, it is necessary to first consider screening out the VIMF components with impact morphology. This invention uses the kurtosis criterion for screening. Kurtosis is sensitive to impact signals. The kurtosis coefficient of random white noise is approximately 3, and the kurtosis value increases rapidly with the increase of abrasive pulses. To avoid missing small-amplitude impact components, this invention sorts the six VIMF components, and the analysis results are as follows. Figure 3 ; The two IMF components with the highest kurtosis were selected as the initial screening results, namely IMF5 and IMF6. To enhance the abrasive signal through morphological similarity, a second screening was conducted using the correlation between the envelope signal of the original signal and the IMF components as a criterion. Figure 4 The envelope signal is given; The correlation coefficients between the envelope signal and the IMF obtained from VMD decomposition were calculated sequentially, and the results are as follows: Figure 5 : If the absolute value of the correlation coefficient is greater than 0.7, the sequences are considered strongly correlated. The correlation coefficients between the two IMFs and the envelope of the original signal are calculated, and the IMFs with a correlation coefficient greater than 0.7 are selected as reconstructed components. If all IMFs are greater than 0.7, they are all selected as reconstructed components. In this case, the correlation coefficients for IMF5 and IMF6 are both greater than 0.7, therefore they are both ultimately selected as reconstructed components. Figure 6 As shown, the possible abrasive signals are significantly enhanced.

[0024] After acquiring the enhanced signal, the abrasive particle signal needs to be identified using a model. This invention proposes an abrasive particle identification model based on sliding window cross-correlation and peak protrusion. First, based on the sensor structural parameters and oil flow velocity, a simulated sliding window signal for identification is constructed. Considering that abrasive particles may have different polarities, a charge signal is used.

[0025] Will Figure 4 The enhanced signal is integrated to obtain the original charge signal. The sliding window is set to step by 10 sampling points each time. The cross-correlation sequence between the sliding window signal and the original charge signal within the sliding window range is calculated each time. The cross-correlation function can effectively measure the similarity between two time series waveforms.

[0026] The calculation result of the maximum cross-correlation value obtained after the sliding window traverses the entire signal is as follows: Figure 7 After normalization, four peaks with large cross-correlation functions were obtained. Obviously, the four large peaks were largely caused by the four possible abrasive pulses.

[0027] Some small peaks are still visible, so the number of abrasive grains cannot be determined solely by identifying peaks; a specific method is needed to filter out the four larger peaks. Observation revealed that, due to the abrasive grains' velocity, approximately equal to the oil flow velocity, the peaks of the cross-correlation maximum image caused by the abrasive grains exhibit a noticeably steep characteristic. Therefore, this invention selects peak salience as the overall criterion, using peak prominence to filter out relatively flat peaks and avoid interference. Based on MATLAB software, the following results were obtained: Figure 8 Peak finding results of the maximum value curve of the cross-correlation function.

[0028] It is necessary to consider setting a threshold for abrasive particle identification. When the peak significance is higher than this threshold, it is considered highly likely to be caused by abrasive particles, while points below the threshold are not identified. The threshold is set to 1.5 times the interquartile range of the significance of all peaks. Points above the threshold are identified as outliers. Through outliers, the specific location of peaks in the cross-correlation function graph can be located, such as... Figure 9 .

[0029] This invention also provides an electrostatic monitoring system for abrasive particles, based on the aforementioned method for identifying electrostatic signals from lubricating oil abrasive particles. The system includes modules such as a sensor, a metal probe, an insulating tube, a shielding cover, and a data acquisition circuit. The internal insulating tube of the sensor is typically a thin ceramic tube through which oil flows, and the sensor's spiral probe is attached to the surface of the ceramic tube. The shielding cover is partially grounded to protect the sensor from external electromagnetic interference. The inductive probe is connected to a signal conditioning circuit, which converts the charge signal into a voltage signal for acquisition. After data analysis, deeper information is obtained. In this invention, the electrostatic sensor employs a dual-probe array design, such as... Figure 10 When abrasive particles flow through the sensor with lubricating oil, abrasive pulses will be generated successively on the two probes. There will be a certain time delay between the two abrasive characteristic pulses. The magnitude of the time delay is related to the distance between the probes and the oil flow rate.

[0030] Therefore, based on this structural characteristic, it can be further confirmed whether the previously identified large peaks are caused by abrasive particles. The same calculations need to be performed on the second channel signal, and compared with the first channel signal. If a peak point of the cross-correlation function of the second channel signal is found near the lag time t1 of the peak point of the first channel signal's cross-correlation function, where t1 = L1 / v, it can be further considered a signal caused by abrasive particles. The number of cases meeting the above conditions is then statistically analyzed and matched with the peak significance anomaly statistics of the two channel signals for confirmation, thus completing the abrasive particle quantity identification process. When the electrostatic monitoring sensor is installed using a dual-probe array, it can effectively identify abrasive particles. The method proposed in this invention has a high identification accuracy.

[0031] The steps and procedures are as follows: Figure 11 and Figure 12 (1) Determine the electrostatic signal y of the first input channel, remove the DC component, and eliminate zero drift; (2) Perform VMD decomposition on y, with a penalty factor of 2000 and a decomposition mode number of 6; (3) Calculate the kurtosis value for each IMF, sort the kurtosis values, and select the top two IMFs as the reconstructed components; (4) Calculate the envelope signal of the original signal y, and calculate the correlation coefficient between the envelope signal and the VIMF obtained in the first screening in turn. Find the IMF with a correlation coefficient greater than 0.7 from the results of step (3); (5) Reconstruct the IMF obtained in (4) to obtain the abrasive enhancement signal; (6) Integrate the enhanced voltage signal and normalize it to obtain the charge signal sequence; (7) Normalized abrasive simulation pulse waveforms are constructed based on flow velocity and probe structure parameters.

[0032] (8) Set the simulation waveform as a sliding window signal and start sliding calculation on the reconstructed signal obtained in (5) to obtain the cross-correlation function sequence matched by each window and normalize it.

[0033] (9) Find the peak of the normalized cross-correlation function sequence, calculate the peak significance, and generate the sequence; (10) Use 1.5 times the interquartile range of the peak significance sequence as the threshold. Points above the threshold can be considered as corresponding abrasive particles. Count the number of abnormal points. (11) Input the signal y1 of the second channel during the same time period, and repeat steps (1)-(8); (12) Determine whether the peak value can be found in the cross-correlation function sequence of the second channel at the time lag t1 of the peak value of the cross-correlation sequence of the first channel signal, and count the number of times the condition is met.

[0034] (13) Repeat (9)-(10) the cross-correlation function sequence of the second channel signal and count the number of outliers; (14) Compare the statistical results of (10), (12), and (13) to confirm the number of abrasive particles and complete the overall identification process.

[0035] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0036] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for identifying abrasive particles in lubricating oil based on electrostatic signals, characterized in that: The steps include: Step 1: Denoise and enhance the possible abrasive electrostatic signals by constructing an abrasive signal enhancement model based on variational mode decomposition (VMD) and envelope signal correlation. The VMD algorithm can adaptively decompose the electrostatic signal, find several optimal IMFs through decomposition, and minimize the sum of the spectral bandwidths of these IMFs. The constraint is that the sum of the time domains of each IMF equals the input signal. Step 2: In the VMD decomposition results, the pulse waveform caused by abrasive particles has a clear impact morphology. Based on this characteristic, when reconstructing the IMF, the VIMF components with impact morphology are first screened out. Step 3: Use the kurtosis criterion for screening. Kurtosis is sensitive to impact signals. The kurtosis coefficient of random white noise is about 3. When the number of abrasive pulses increases, the kurtosis value increases rapidly. To avoid missing small-amplitude impact components, the 6 VIMF components are sorted, and the two IMF components with the highest kurtosis are selected as the initial screening results. Further screening is carried out using the correlation between the envelope signal of the original signal and the IMF as a criterion. Step 4: Calculate the correlation coefficient between the envelope signal and the IMF obtained from VMD decomposition. If the absolute value of the correlation coefficient is greater than 0.7, the sequences are considered to be strongly correlated. Select the IMF with a correlation coefficient greater than 0.7 as the reconstructed component. If all are greater than 0.7, they are all used as the reconstructed components. Step 5: After acquiring the enhanced signal, the abrasive signal is identified using an abrasive identification model based on sliding window cross-correlation and peak protrusion. Based on the sensor structural parameters and oil flow velocity, a simulated sliding window signal for identification is constructed. Considering that abrasive particles may have different polarities, a charge signal is used. The obtained enhanced signal is integrated to obtain the original charge signal. The sliding window is set to step by 10 sampling points each time, and the cross-correlation sequence between the sliding window signal and the original charge signal within the sliding window range is calculated each time. Step 6: After the sliding window traverses the entire signal, normalize the calculated maximum cross-correlation value and filter out steep peaks by peak significance to avoid interference from flat peaks; A threshold for judging abrasive particles is set, which is 1.5 times the interquartile range of the significance of all peaks. Points above the threshold are identified as outliers. The specific location of the peaks in the cross-correlation function graph is located by the outliers, thus completing the identification of abrasive particle signals. The method for identifying electrostatic signals of lubricating oil abrasive particles is applied to an abrasive particle electrostatic monitoring system, which includes modules such as a sensor, a metal probe, an insulating tube, a shielding cover, and a data acquisition circuit. The sensor's internal insulating tube is a thin ceramic tube, through which oil flows. The sensor's spiral probe is attached to the surface of the ceramic tube, and the shielding shell is partially grounded to protect the sensor from external electromagnetic interference. The sensing probe is connected to the signal conditioning circuit, which converts the charge signal into a voltage signal for acquisition. After data analysis, in-depth information is obtained. The metal probe adopts a dual-probe front and rear array design.

Images