Friction and wear state online monitoring method and system based on multi-physical-quantity synchronous acquisition

Abstract

The invention relates to the technical field of frictional wear state monitoring, and particularly discloses a frictional wear state online monitoring method and system based on multi-physical-quantity synchronous acquisition, and the method comprises the following steps: 1, synchronously acquiring a plurality of physical quantity signals of a friction pair in an operation process, the plurality of physical quantity signals at least comprise an acoustic emission signal, a vibration signal and a friction force signal; step 2, performing synchronous preprocessing on the plurality of acquired physical quantity signals to obtain multi-channel data with aligned time domains; 3, extracting multi-dimensional characteristic parameters related to the frictional wear state from the multi-channel data, and constructing a characteristic parameter set; according to the frictional wear state online monitoring method and system based on multi-physical-quantity synchronous acquisition, the defects of one-sidedness and hysteresis of traditional single signal monitoring are overcome, and the purpose of accurately and reliably monitoring the whole life cycle of frictional wear is achieved.

Description

Technical Field

[0001] This invention relates to the field of friction and wear condition monitoring technology, specifically to an online monitoring method and system for friction and wear condition based on the synchronous acquisition of multiple physical quantities. Background Technology

[0002] Friction and wear are key factors leading to performance degradation and failure of mechanical equipment, and online monitoring of them is the core of predictive maintenance. Existing monitoring technologies have significant limitations: methods based on oil analysis have strong lag and cannot meet real-time requirements; while methods based on single physical signals such as vibration or acoustic emission, although capable of online monitoring, often suffer from incomplete characterization of complex wear processes due to limited signal features, leading to missed detections or false alarms of early wear. Although some research attempts to integrate multi-sensor information, these efforts mostly remain at the level of simple information parallelism, lacking in-depth exploration of the strict synchronization and intrinsic coupling mechanisms between multiple physical quantities, and thus failing to fundamentally improve the accuracy and reliability of condition identification. Therefore, developing an online monitoring method capable of simultaneous acquisition and deep fusion of multiple physical quantities has become a pressing technical bottleneck in this field. Summary of the Invention

[0003] To solve the above technical problems, the present invention provides the following technical solution: an online monitoring method for friction and wear state based on simultaneous acquisition of multiple physical quantities, comprising the following steps: The first step is to synchronously collect multiple physical quantity signals of the friction pair during operation, including at least acoustic emission signals, vibration signals, and friction force signals. The second step is to perform synchronous preprocessing on the acquired multiple physical quantity signals to obtain time-domain aligned multi-channel data. The third step is to extract multi-dimensional feature parameters related to the friction and wear state from the multi-channel data and construct a feature parameter set. The fourth step is to input the set of feature parameters into the pre-trained wear state recognition model and output the recognition result of the current friction and wear state. The wear state recognition model is trained based on historical data and can establish a mapping relationship between multi-dimensional feature parameters and specific wear states. Fifth step: Output and display the identification results to achieve online monitoring and early warning of friction and wear conditions.

[0004] Preferably, in the first step, a unified clock source is used to provide a synchronous trigger signal for the acquisition channels of all physical quantity signals, synchronizing the acquisition start time and timestamp of all signals.

[0005] Preferably, the specific steps of the synchronous preprocessing in the second step are as follows: Signal conditioning and calibration: Converting the acquired raw voltage signal into engineering unit data with accurate physical meaning, and correcting the error of the sensor itself; Noise suppression and filtering: Preserve the effective frequency bands related to friction and wear, and suppress interference noise from the environment and decoupling; Timestamp alignment and resampling: unify the time base for each data point across all channels; Data segmentation and standardization: prepare for subsequent real-time and continuous analysis and eliminate the influence of different feature units.

[0006] Preferably, the specific steps for constructing the feature parameter set in the third step are as follows: Signal domain partitioning and initial feature extraction: Feature mining is performed on the multi-channel data after synchronous preprocessing in the time domain, frequency domain, and time-frequency domain, respectively; Construction of cross-physical quantity coupling features: By utilizing the synchronous acquisition of multiple physical quantities, we can explore the intrinsic correlation and interaction between signals of different physical quantities and construct cross-physical quantity coupling features; Feature selection and dimensionality reduction: After the first two steps, an initial feature set is obtained, which contains redundant and irrelevant features. The most effective and compact feature subset is selected from these features. Construct a feature parameter set: Organize the final selected, most discriminative features in a predetermined order and format to form a standardized feature vector.

[0007] Preferably, after outputting the recognition result in the fifth step, the method further includes: The current set of feature parameters and the corresponding recognition results are stored in the historical database for incremental learning and online updating of the wear state recognition model.

[0008] An online monitoring system for friction and wear conditions based on simultaneous acquisition of multiple physical quantities includes: A multi-sensor array, arranged on the friction pair or its supporting structure, is used to sense the multiple physical quantity signals; A synchronous data acquisition module, connected to the multi-sensor array, is used to synchronously acquire and digitize the multiple physical quantity signals; The signal processing and feature extraction module is used to perform synchronous preprocessing on the digitized signal and extract the multi-dimensional feature parameters. The state recognition and analysis module has the wear state recognition model built in, which is used to receive the feature parameter set and output the recognition result; The human-computer interaction and early warning module is used to display the recognition results and issue early warning information when abnormal wear is detected.

[0009] Preferably, the synchronous data acquisition module includes a multi-channel acquisition card and a unified clock source, wherein all channels of the multi-channel acquisition card share the unified clock source.

[0010] Preferably, the multi-sensor array includes: Acoustic emission sensor, used to collect acoustic emission signals; An accelerometer is used to collect vibration signals; Force sensor, used to collect friction force signals.

[0011] Preferably, the system also integrates a device control interface. When the status identification and analysis module identifies a severe wear or failure state, it sends a shutdown or load reduction command to the monitored device through the device control interface.

[0012] It has the following beneficial effects: By synchronously acquiring and deeply fusing multiple physical quantity signals, the limitations and lag of traditional single-signal monitoring are overcome, enabling accurate and reliable monitoring of the entire life cycle of friction and wear. Employing a multi-scale time-series fusion model with an attention mechanism, it can not only output discrete state classifications but also provide continuous wear process scores, achieving a leap from threshold alarms to trend predictions. The model's decision-making process is visible, significantly improving the reliability of the results and the maintainability of the system. Ultimately, a closed-loop solution integrating real-time monitoring, intelligent diagnosis, and proactive early warning is formed, providing strong support for predictive maintenance of equipment. Detailed Implementation

[0013] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] This invention provides a technical solution: an online monitoring system for friction and wear conditions based on simultaneous acquisition of multiple physical quantities, comprising: A multi-sensor array, arranged on the friction pair or its supporting structure, is used to sense the multiple physical quantity signals. The multi-sensor array includes: an acoustic emission sensor for acquiring acoustic emission signals; an acceleration sensor for acquiring vibration signals; and a force sensor for acquiring friction force signals. A synchronous data acquisition module is connected to the multi-sensor array and is used to synchronously acquire and digitize the multiple physical quantity signals. The synchronous data acquisition module includes a multi-channel acquisition card and a unified clock source. All channels of the multi-channel acquisition card share the unified clock source. The signal processing and feature extraction module is used to perform synchronous preprocessing on the digitized signal and extract the multi-dimensional feature parameters. The state recognition and analysis module has the wear state recognition model built in, which is used to receive the feature parameter set and output the recognition result; The human-computer interaction and early warning module is used to display the recognition results and issue early warning information when abnormal wear is detected; The system also integrates a device control interface. When the status identification and analysis module identifies a severe wear or failure state, it sends a shutdown or load reduction command to the monitored device through the device control interface.

[0015] A method for online monitoring of friction and wear conditions based on simultaneous acquisition of multiple physical quantities includes the following steps: The first step is to synchronously acquire multiple physical quantity signals of the friction pair during operation. These multiple physical quantity signals include at least acoustic emission signals, vibration signals, and friction force signals. A unified clock source is used to provide a synchronous trigger signal for the acquisition channels of all physical quantity signals, synchronizing the acquisition start time and timestamp of all signals. The second step involves synchronously preprocessing the acquired multiple physical quantity signals to obtain time-domain aligned multi-channel data. The specific steps are as follows: Signal conditioning and calibration: This step aims to convert the acquired raw voltage signal into engineering unit data with accurate physical meaning and to correct the errors of the sensor itself. Unit conversion of physical quantities: Operation: Apply the conversion formula according to the calibration certificate or datasheet of each sensor, for example: Vibration value (m / s²) = Data acquisition card reading (V) / Accelerometer sensitivity (mV / g) * Gravitational acceleration (m / s²); Acoustic emission value (dB) = 20 * log10 (acquisition card reading (V) / reference voltage (V)); Temperature value (°C) = (Thermocouple voltage (mV) - Cold junction compensation voltage (mV)) / Thermocouple coefficient (mV / °C); Objective: To unify all signals to international standard units, ensuring comparability between different physical quantities and facilitating subsequent coupling analysis; Sensor sensitivity calibration and DC bias removal: Operation: Before system startup or each data acquisition, record the average signal value under frictionless or no-load conditions as the DC bias value, and subtract it from subsequent data. At the same time, apply the sensitivity coefficient for amplitude calibration. Objective: To eliminate the inherent zero drift of sensors and acquisition circuits, and ensure the accuracy of the signal's "zero point"; Noise suppression and filtering: This step aims to preserve the effective frequency bands associated with friction and wear, and suppress interference noise from the environment and decoupling. Power frequency notch filtering: Operation: Apply one or more narrowband stop filters to accurately filter out the 50Hz power grid frequency and its main harmonic interference; Objective: To remove the most common environmental electromagnetic interference, which severely affects the low-frequency components of vibration and acoustic emission signals; Bandpass filtering: Operation: Apply different bandpass filters according to the characteristics of each physical quantity: Acoustic emission signals: Usually, the focus is on the high-frequency band. High-pass filters are used to remove low-frequency noise such as mechanical vibrations, and low-pass filters are used to prevent high-frequency aliasing. Vibration signals: Focus on the mid-to-high frequency range to capture the natural frequencies and impact responses of bearings and gears; Friction force signal: usually a low-frequency slowly varying signal, high-frequency noise is removed using a low-pass filter; Objective: To focus the signal of each channel onto its most effective analysis band, thereby improving the signal-to-noise ratio; Timestamp alignment and resampling: This is the most critical step in achieving "time domain alignment," ensuring that every data point in all channels has a unified and accurate time reference. Uniform timestamp: Operation: At the hardware level, the synchronous data acquisition module uses a unified hardware clock and trigger line to ensure that the sampling clocks of all channels are strictly synchronized. At the software level, the sample sequences of all channels are stamped with a precise timestamp based on this clock. Objective: To ensure that all signals are synchronized at the moment of acquisition, which is the most ideal and accurate method. Sampling rate unification and data alignment: Operation: If different physical quantities use different sampling rates due to hardware limitations, sampling rate conversion is required. For high-frequency signals: resample them to a uniform, sufficiently high target sampling rate; For low-frequency signals: interpolation is usually used to upsample them to the same sampling rate as high-frequency signals, and new data points are generated based on a unified timestamp. Objective: To ensure that the data length of each channel is consistent and that the data point corresponding to each index number i represents the physical quantity at the same moment, forming a strictly aligned multi-channel data block; Data segmentation and standardization: This step prepares for subsequent real-time, continuous analysis and eliminates the influence of different characteristic dimensions; Data window splitting: Operation: The continuously acquired, aligned multi-channel data stream is divided into segments of fixed time lengths. Overlapping sliding windows are typically used to increase the amount of data and avoid missing transient events. Objective: To transform the continuous monitoring process into discrete, independently analyzable "data samples" to facilitate real-time processing and status tracking; Data standardization: Operation: Standardize the data for each channel within each data window. Common methods include: Z-Score standardization: x_normalized=(x-μ) / σ, where μ is the mean of the data in this window and σ is the standard deviation; Maximum and minimum value normalization: x_normalized=(x-min(x)) / (max(x)-min(x)); Purpose: Dimensionless elimination: This brings the values ​​of different physical quantities to the same order of magnitude, facilitating subsequent fusion model processing; Enhance model stability: Reduce the impact of overall signal amplitude fluctuations caused by changes in operating conditions on the model, and make the model focus more on changes in the "morphology" of the signal; After the systematic processing of the above four steps, the original and coarse multi-physical quantity signals are transformed into "time-domain aligned multi-channel data". These data blocks are clean, time-synchronized, and dimensionally unified, laying a solid foundation for subsequent high-precision multi-dimensional feature extraction and intelligent diagnosis. The third step is to extract multi-dimensional feature parameters related to the friction and wear state from the multi-channel data and construct a feature parameter set. The specific steps are as follows: Signal domain partitioning and initial feature extraction: This step involves feature mining in the time domain, frequency domain, and time-frequency domain for the multi-channel data (assuming acoustic emission, vibration, and friction force channels) after synchronous preprocessing. Temporal feature extraction: Objective: To characterize the amplitude, energy, distribution pattern, and impact characteristics of signals, and to be sensitive to early wear and sudden wear; Extraction method: Calculate the following features for each data channel: Dimensional indicators: mean, absolute mean, root mean square, peak value, peak-to-peak value, variance; Dimensionless index: Kurtosis: Extremely sensitive to the impact component in the signal, it is an excellent indicator of early wear; Skewness: describes the asymmetry of the signal probability density distribution, which helps to identify changes in wear state; Waveform factor, peak factor, and pulse factor: These are classic wear monitoring indicators, commonly used for fault diagnosis of bearings and gears; Statistical characteristics: quartiles, range, standard deviation; Frequency domain feature extraction: Objective: To reveal the frequency components and energy distribution of a signal, in order to identify specific frequency components related to wear; Extraction method: Perform a Fast Fourier Transform on each data channel to convert the signal from the time domain to the frequency domain, obtaining the power spectrum or amplitude spectrum; Extract the following features from the spectrum: Centroid frequency, mean square frequency, and frequency variance: describe the location and changes in the overall distribution of the frequency spectrum; Spectral kurtosis: The frequency band in which non-stationary impulse components are located in a positioning signal, which is very suitable for early fault detection and diagnosis; Energy in a specific frequency band: Based on the inherent characteristics of the friction pair or prior knowledge, several key frequency bands are divided, and the energy proportion in each frequency band is calculated; Harmonic component analysis: Identifying harmonic components and their sideband variations related to frequency conversion, meshing frequency, etc. Time-frequency domain feature extraction: Objective: To simultaneously capture the localization characteristics of signals in time and frequency, suitable for the analysis of non-stationary, time-varying friction and wear processes; Extraction method: Wavelet transform: Select appropriate wavelet basis functions to perform multi-resolution analysis of the signal; The energy of the wavelet decomposition coefficients at each level is extracted as a feature to form a "wavelet energy spectrum"; Calculate the statistical characteristics of wavelet coefficients at each level; Empirical Mode Decomposition: The signal is adaptively decomposed into a series of intrinsic mode functions; Calculate the energy, correlation coefficient, or sample entropy of the first few IMF components containing the main information as features; Construction of features coupled across physical quantities: Objective: This is the key innovation of this method. By utilizing the advantage of simultaneous acquisition of multiple physical quantities, we can explore the intrinsic correlation and interaction between signals of different physical quantities. Changes in this correlation often reveal the wear mechanism more clearly than the characteristics of a single signal. Construction method: Cross-correlation analysis: Calculate the cross-correlation function of two signals from different channels, and take the maximum value or the time lag of reaching the maximum value as a feature to reflect the causal time relationship between the two physical phenomena; Transfer function / coherence analysis: Analyzes the changes in transfer characteristics from one physical quantity to another, and can be used to identify changes in system stiffness and damping; Joint time-frequency analysis: Compare the energy distribution of different signals on the time-frequency plane, for example, observe whether the high-frequency acoustic emission energy burst is precisely synchronized in time with a specific impact event in the vibration signal; Ratio or combined features: Create composite indices such as "root mean square value of acoustic emission / root mean square value of vibration" and "coefficient of variation of friction / rate of temperature change", which can amplify the trend of wear condition changes; Feature selection and dimensionality reduction: Objective: After the first two steps, we obtain an initial feature set with extremely high dimensionality, containing a large number of redundant and irrelevant features. This step aims to filter out the most effective and compact subset of features to improve model efficiency and generalization ability. Implementation method: Filtering method: Screening is based on statistical indicators, such as calculating the correlation coefficient between each feature and the wear status label, chi-square test, mutual information, etc., and retaining the features with the highest correlation; Wrap-up method: Uses a specific machine learning model to sort and select features based on their importance scores; Embedding: Automatic feature selection during model training, such as using LASSO regression or decision trees; Principal Component Analysis (PCA): The original feature space is linearly transformed to a new orthogonal space, and the first few principal components are taken as new features, thus achieving dimensionality reduction while retaining most of the information. Constructing the feature parameter set: The final selected, most discriminative features are organized in a predetermined order and format to form a standardized feature vector. This vector is the final "feature parameter set" that will be used as the input to the wear condition recognition model; The fourth step is to input the set of feature parameters into the pre-trained wear state recognition model and output the recognition result of the current friction and wear state. The wear state recognition model is trained based on historical data and can establish a mapping relationship between multi-dimensional feature parameters and specific wear states. Step 5: Output and display the recognition results to achieve online monitoring and early warning of friction and wear conditions. Store the current set of feature parameters and the corresponding recognition results in the historical database for incremental learning and online updating of the wear condition recognition model.

[0016] Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art and related fields based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described and explained in the present invention, unless otherwise specified or limited, shall be implemented according to conventional means in the art.

Claims

1. A method for online monitoring of friction and wear conditions based on simultaneous acquisition of multiple physical quantities, characterized in that, Includes the following steps: The first step is to synchronously collect multiple physical quantity signals of the friction pair during operation, including at least acoustic emission signals, vibration signals, and friction force signals. The second step is to perform synchronous preprocessing on the acquired multiple physical quantity signals to obtain time-domain aligned multi-channel data. The third step is to extract multi-dimensional feature parameters related to the friction and wear state from the multi-channel data and construct a feature parameter set. The fourth step is to input the set of feature parameters into the pre-trained wear state recognition model and output the recognition result of the current friction and wear state. The wear state recognition model is trained based on historical data and can establish a mapping relationship between multi-dimensional feature parameters and specific wear states. Fifth step: Output and display the identification results to achieve online monitoring and early warning of friction and wear conditions.

2. The method for online monitoring of friction and wear state based on simultaneous acquisition of multiple physical quantities according to claim 1, characterized in that, In the first step, a unified clock source is used to provide a synchronous trigger signal for the acquisition channels of all physical quantity signals, synchronizing the acquisition start time and timestamp of all signals.

3. The method for online monitoring of friction and wear state based on simultaneous acquisition of multiple physical quantities according to claim 1, characterized in that, The specific steps of the synchronous preprocessing in the second step are as follows: Signal conditioning and calibration: Converting the acquired raw voltage signal into engineering unit data with accurate physical meaning, and correcting the error of the sensor itself; Noise suppression and filtering: Preserve the effective frequency bands related to friction and wear, and suppress interference noise from the environment and decoupling; Timestamp alignment and resampling: unify the time base for each data point across all channels; Data segmentation and standardization: prepare for subsequent real-time and continuous analysis and eliminate the influence of different feature units.

4. The method for online monitoring of friction and wear state based on simultaneous acquisition of multiple physical quantities according to claim 1, characterized in that, The specific steps for constructing the feature parameter set in the third step are as follows: Signal domain partitioning and initial feature extraction: Feature mining is performed on the multi-channel data after synchronous preprocessing in the time domain, frequency domain, and time-frequency domain, respectively; Construction of cross-physical quantity coupling features: By utilizing the synchronous acquisition of multiple physical quantities, we can explore the intrinsic correlation and interaction between signals of different physical quantities and construct cross-physical quantity coupling features; Feature selection and dimensionality reduction: After the first two steps, an initial feature set is obtained, which contains redundant and irrelevant features. The most effective and compact feature subset is selected from these features. Construct a feature parameter set: Organize the final selected, most discriminative features in a predetermined order and format to form a standardized feature vector.

5. The method for online monitoring of friction and wear state based on simultaneous acquisition of multiple physical quantities according to claim 1, characterized in that, After outputting the recognition result in the fifth step, the process also includes: The current set of feature parameters and the corresponding recognition results are stored in the historical database for incremental learning and online updating of the wear state recognition model.

6. An online monitoring system for friction and wear conditions based on simultaneous acquisition of multiple physical quantities, characterized in that, include: A multi-sensor array, arranged on the friction pair or its supporting structure, is used to sense the multiple physical quantity signals; A synchronous data acquisition module, connected to the multi-sensor array, is used to synchronously acquire and digitize the multiple physical quantity signals; The signal processing and feature extraction module is used to perform synchronous preprocessing on the digitized signal and extract the multi-dimensional feature parameters. The state recognition and analysis module has the wear state recognition model built in, which is used to receive the feature parameter set and output the recognition result; The human-computer interaction and early warning module is used to display the recognition results and issue early warning information when abnormal wear is detected.

7. The online monitoring system for friction and wear condition based on simultaneous acquisition of multiple physical quantities according to claim 6, characterized in that, The synchronous data acquisition module includes a multi-channel acquisition card and a unified clock source, and all channels of the multi-channel acquisition card share the unified clock source.

8. The online monitoring system for friction and wear state based on simultaneous acquisition of multiple physical quantities according to claim 6, characterized in that, The multi-sensor array includes: Acoustic emission sensor, used to collect acoustic emission signals; An accelerometer is used to collect vibration signals; Force sensor, used to collect friction force signals.

9. The online monitoring system for friction and wear state based on simultaneous acquisition of multiple physical quantities according to claim 6, characterized in that, The system also integrates a device control interface. When the status identification and analysis module identifies a severe wear or failure state, it sends a shutdown or load reduction command to the monitored device through the device control interface.