Friction signal based running-in state recognition method without threshold recursive analysis

Abstract

The application discloses a running-in state identification method based on a friction signal threshold-free recursive analysis result, extracts nonlinear components of a friction signal time sequence output in a friction and wear process by collecting the friction signal time sequence; constructs a threshold-free recursive matrix by combining a sliding window and an avoidance embedding phase space reconstruction method and realizes visualization of the threshold-free recursive matrix in a two-dimensional plane; divides a lower triangular matrix of the threshold-free recursive matrix into several regions by a group of equidistant line segments parallel to a main diagonal line of the recursive matrix, extracts mean values of recursive elements in the regions, analyzes a change rate of element values of the threshold-free recursive matrix when the threshold-free recursive matrix transitions along the main diagonal line to edges and corners as a recursive measure, analyzes variation rules of the recursive measure along with a running-in and wear process, and further realizes running-in state online identification driven by the friction signal threshold-free recursive analysis result. The application optimizes a recursive characteristic representation and a measure extraction method of the friction signal time sequence, and realizes running-in state online identification from a perspective of dynamics-friction coupling.

Description

TECHNICAL FIELD

[0001] The present application relates to tribology, in particular a running-in state recognition method based on a threshold-free recursive analysis result of a friction signal. BACKGROUND

[0002] Friction and wear of mechanical friction pairs is one of the main reasons for material loss of parts and failure of equipment. However, the strong system dependence and time-varying characteristics exhibited by friction and wear behavior make it difficult to construct accurate mathematical or physical models for analysis and judgment, and usually require the use of friction signals, wear surface and wear particle characteristics output during the wear process to realize wear mechanism, state recognition and fault diagnosis. Therefore, establishing an efficient, automated, data-driven friction and wear behavior characterization method is an important basis and theoretical foundation for evaluating friction and wear performance and analyzing friction and wear mechanism.

[0003] The initial value sensitivity of friction and wear behavior with respect to control parameters such as contact load, relative motion speed and lubrication conditions determines its typical nonlinear and non-stationary characteristics. Chaos analysis methods represented by attractor theory and feature parameter extraction techniques can comprehensively reveal the evolution law and system dependence of friction and wear behavior from the perspective of dynamics and tribology coupling. Recursive analysis technology, as a development of chaos theory and nonlinear time series analysis methods, is a fine feature extraction method suitable for short non-stationary time series. It can realize recursive characteristic extraction of friction signals and reveal its time-varying characteristics with respect to wear state and system dependence with respect to friction control parameters through visualization of reconstructed attractor recursive matrix (recurrence plot) and feature parameter extraction (quantitative recursive analysis), thereby providing quantitative parameters for friction and wear state recognition and performance evaluation. It is an important application of nonlinear dynamics methods in friction and wear problem research.

[0004] The recursive matrix constructed according to the relative distance between each point of the reconstructed attractor is the basis for carrying out recursive characteristic analysis. In standard recursive analysis, each element in the recursive matrix is defined as: R i,j = Θ(ε-||X i -X j ||), where X h = [x h ,x h+τ ,x h+2τ ,…,x h+(m-1)τ ], (h = i, j) is a univariate time series signal x = {x h| h = 1,2, …, t} reconstructs the corresponding phase point coordinates of the high-dimensional phase space trajectory X at each time; the remaining parameters are defined as: ε - the preset neighborhood radius, m - the embedding dimension of the reconstructed phase space, τ - the time delay of the reconstructed phase space, Θ(·) - the Heaviside step function [Θ(k) = 1|k≥0; Θ(k) = 0|k<0]. On the one hand, when the sliding window method is used to extract the evolution law of the RQA measure with the running-in process, the optimal embedding dimension, time delay and neighborhood radius of the friction signal at different wear stages are significantly different; on the other hand, the extraction process of the above parameters has not yet formed a universal calculation method, which leads to the strong subjectivity and empiricism of the recursive analysis results. SUMMARY

[0005] The purpose of the application is to provide a running-in state recognition method based on a friction signal threshold-free recursive analysis result without threshold and non-embedding, extract a quantitative measure to characterize the evolution law thereof with the running-in process, and further realize the construction of a running-in state online recognition system.

[0006] Technical scheme: the running-in state recognition method based on the friction signal threshold-free recursive analysis result, comprising the following steps:

[0007] S1, using the acquisition device to collect the friction signal output in the friction and wear process in real time, and storing it in time sequence format to the computer; the acquisition device comprises a sensor network and a data acquisition card.

[0008] S2, the nonlinear component of the real-time collected friction signal is extracted.

[0009] S2.1, the collected sequence is decomposed into a plurality of modal functions IMF and a residual component r by using empirical mode decomposition.

[0010] S2.2, the power spectrum of each order IMF is analyzed, and the modal functions corresponding to the periodic noise component, the random noise component, and the modal function with continuous bandwidth and exponential decay nonlinear chaotic characteristics are identified according to the power spectrum distribution characteristics. c The periodic noise component includes the vibration of the test machine motor and the periodic interference of other associated devices in the friction system, and the random noise component includes the background interference in the data acquisition process.

[0011] S2.3, for the part of the modal function which is difficult to identify by using the power spectrum distribution characteristics, residual variance and waveform similarity parameters are introduced for further analysis.

[0012] S2.4, reconstructing the modal function and residual component with nonlinear characteristics, completing the nonlinear component extraction of the friction signal, and providing a data basis for the subsequent recursive characteristic analysis of the friction signal.

[0013] S3, using the sliding window method to divide the friction signal with nonlinear characteristics into continuous but non-overlapping rectangular calculation windows, reconstructing the threshold-free recursive matrix corresponding to the time sequence in each rectangular calculation window, and realizing its visualization.

[0014] S3.1, using the avoidance embedding phase space reconstruction method to obtain the attractor corresponding to the single variable time sequence.

[0015] In the delay coordinate method, the high-dimensional phase space trajectory X = [X1, X2, …, X i ,…,X n} of the single variable time sequence {x1, x2, …, x i ,…,x N} is reconstructed. T , where X i = [x(i), x(i+τ), x(i+2τ), …, x(i+(m-1)τ)], (i = 1, 2, …, n-(m-1)τ), and the parameters m and τ are set to 1.

[0016] S3.2, using the membership function to construct the threshold-free recursive matrix of the attractor.

[0017] The expression of the membership function is: R i,j = cos [π•d i,j / (2max(d i,j ))], where d i,j = ||X i -X j || represents the relative distance between the phase points X i and X j , and the function value is inversely proportional to the distance between the phase points, which can effectively represent the recursive characteristics of the phase trajectory.

[0018] S3.3, applying a gray-scale image to realize the visualization of the threshold-free recursive matrix in a two-dimensional plane, and the deeper the color of the pixel point position, the closer the matrix element value at that position to 1, and vice versa, the lighter the color of the pixel point, the closer the matrix element value to 0.

[0019] S4, extracting the recursive measure of the threshold-free recursive matrix in each rectangular calculation window, identifying the wear stage corresponding to each rectangular calculation window according to the change rule of the recursive measure during the running-in wear process, and realizing the running-in state online identification driven by the threshold-free recursive analysis result of the friction signal.

[0020] S4.1, adopt a set of parallel to the main diagonal of the recursive matrix, and equal interval 45 ° line segment will matrix right down triangle area is divided into several areas.

[0021] S4.2, calculate the average value GR of each area in the threshold-free recursive matrix element value a (k).

[0022] S4.3, draw GR a (k) along the main diagonal to the corner transition direction of change trend, and fitting to obtain the linear relationship, the slope of linear function is the threshold-free recursive measure RT of friction signal G .

[0023] S4.4, analysis measure RT G With the evolution law of rectangular calculation window, namely the running-in process, the measure RT G Enter the stationary stage as the system enters the stable wear stage, and complete the running-in state online identification driven by the threshold-free recursive analysis result of friction signal.

[0024] A computer storage medium, which stores a computer program, the computer program is executed by a processor to realize the above-mentioned running-in state identification method based on the threshold-free recursive analysis result of friction signal.

[0025] A computer device, comprising a storage, a processor and a computer program stored on the storage and executable on the processor, wherein the processor executes the computer program to realize the above-mentioned running-in state identification method based on the threshold-free recursive analysis result of friction signal.

[0026] Advantages: compared with the prior art, the present application has the following advantages:

[0027] 1. The present application starts from the coupling of dynamics and tribology, and applies recursive analysis technology to realize fine feature extraction of friction coefficient signal in friction and wear process, and provides a new perspective for quantitative characterization of friction and wear behavior.

[0028] 2. The present application constructs a recursive characteristic evaluation method based on avoidance embedded phase space reconstruction and membership function, aiming at the influence of input parameters such as phase space reconstruction, neighborhood radius and norm on traditional recursive analysis result, so as to improve the universality and accuracy of recursive analysis result. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 Flow chart for running-in state online identification method driven by threshold-free recursive analysis result of friction signal;

[0030] Figure 2 Flow chart for friction signal nonlinear component extraction method based on empirical mode decomposition;

[0031] Figure 3 The friction coefficient time-domain signal diagram is shown in the example, where Figure 3 (a) is a time-domain signal diagram of the original sequence of friction coefficients in the embodiment. Figure 3 (b) is a time-domain signal diagram of the nonlinear sequence extracted from the original friction coefficient sequence of the embodiment after empirical mode decomposition;

[0032] Figure 4 for Figure 3 The projection of a thresholdless recursive matrix onto a two-dimensional plane (thresholdless recursive graph) is obtained by constructing four rectangular computation window subsequences of a time-domain signal.

[0033] Figure 5 for Figure 4 The GR was extracted from the thresholdless recursive graph of the subsequence within Window 30 shown. a (k)-k parameter variation trajectory and linear fitting results;

[0034] Figure 6 for Figure 3 The time-domain signal of the friction coefficient shown is extracted to obtain the evolution law of recursive analysis results with the running-in and wear process. Detailed Implementation

[0035] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0036] like Figure 1 As shown, an online identification method for running-in status based on thresholdless recursive analysis of friction signals includes the following steps:

[0037] Step 1: Friction and Wear Test, Friction Signal Acquisition, and Nonlinear Component Extraction. A sliding wear test is conducted on a general-purpose / self-made friction and wear testing machine. The friction signals output during the test are acquired using sensors. This embodiment uses the friction coefficient signal as an example. Empirical mode decomposition is applied to extract the nonlinear components from the acquired raw sequence. The specific process is as follows: Figure 2 As shown. By reconstructing the modal functions and residual components that satisfy continuous bandwidth and exponential decay criteria, a time series of the friction signal with nonlinear characteristics is obtained, as shown in the figure. Figure 3 As shown.

[0038] Step 2: Extracting the evolution law of the friction coefficient signal without threshold recursion during the break-in process based on the sliding window. The specific steps are further divided into:

[0039] (1) The friction coefficient signal extracted above is divided into several equal-length, non-overlapping rectangular calculation windows using the sliding window method;

[0040] (2) The friction coefficient signal in each rectangular calculation window is reconstructed by avoiding embedding in phase space, and a threshold-free recursive matrix is obtained according to a membership function, and a threshold-free recurrence plot is applied on a two-dimensional plane to realize visualization;

[0041] (3) The lower triangular matrix of the threshold-free recursive matrix is divided into several equally spaced regions, and the average value GR a (k) of the matrix elements in each region k is calculated;

[0042] (4) The GR a (k)-k function relationship is linearly fitted, and the slope of the fitted straight line is taken as the recursive measure RT G of the friction coefficient in the rectangular calculation window.

[0043] Step three, identification of the running-in state based on the threshold-free recursive analysis result of the friction signal. The recursive measures RT G obtained in the above-mentioned rectangular calculation windows are connected in turn to obtain their variation law with the wear process, so as to construct a running-in state identification method based on the recursive analysis result of the friction coefficient signal. A rotating sliding friction and wear test is carried out on a self-made friction and wear tester using a ball-disc friction pair, and the time domain signal of the generated friction coefficient is collected and nonlinear component extraction is carried out, and the results are shown in Figure 3 (a) and Figure 3 (b), Figure 3 (b) The experimental time corresponding to the four rectangular calculation windows is 20-30 min, 100-110 min, 240-250 min, and 300-310 min, respectively.

[0044] A rectangular calculation window containing 6000 data points (corresponding to an experimental time of 10 min) is used to divide the original time sequence into several subsequences, and four rectangular calculation windows with experimental times of 20-30 min, 100-110 min, 240-250 min, and 300-310 min are taken as examples. The threshold-free recursive matrix of the corresponding time sequence is reconstructed by avoiding embedding phase space reconstruction and membership function reconstruction, and a two-dimensional plane mapping is constructed, and the results are shown in Figure 4 .

[0045] A set of parallel to the main diagonal direction and equally spaced line segments are used to divide the right lower triangular region of the threshold-free recursive matrix into several regions (10 regions in this embodiment), and the average value GR a (k) of all elements of the threshold-free recursive matrix in each region is calculated, and a scatter plot is drawn and the GR a (k)-k function relationship is linearly fitted, and the variation law of the area division and the parameter GR a (k) of the threshold-free recurrence plot of the subsequence in the above-mentioned rectangular calculation window Window 30 is shown inFigure 5 is shown.

[0046] The above steps of phase space reconstruction, threshold-free recurrence matrix reconstruction, and recurrence measure extraction are repeated for the time series within all rectangular calculation windows, and the measure RT G The results are shown in Fig. 2 as the evolution of the running-in process. Figure 6 It can be seen from the figure that, from the rectangular calculation window 1 to the rectangular calculation window 11, the measure RT G shows an increasing trend of oscillation, indicating that the friction coefficient signal at this stage contains a significant trend component, but gradually tends to be stable, corresponding to the running-in wear stage; from the rectangular calculation window 11, the measure RT G fluctuates stably within a small range, indicating that the amplitude of the friction coefficient signal at this stage is stable, and the system enters the stable wear stage, thus completing the online identification of the running-in state driven by the results of the threshold-free recurrence analysis of the friction signal.

Claims

1. A method for identifying the running-in state based on thresholdless recursive analysis results of friction signals, characterized in that, Includes the following steps: S1. Use a data acquisition device to collect the friction signal output during the friction and wear process in real time, and store it in a time series format in a computer. S2. Extract nonlinear components from the real-time acquired friction signals; S3. The extracted friction signal with nonlinear characteristics is divided into continuous but non-overlapping rectangular calculation windows using the sliding window method. The attractor corresponding to the univariate time series is obtained using the avoidance embedding phase space reconstruction method. A thresholdless recursive matrix of the attractor is constructed based on the membership function, and the thresholdless recursive matrix is ​​visualized on a two-dimensional plane using a grayscale image. The darker the pixel, the closer the matrix element value is to 1; conversely, the lighter the pixel, the closer the matrix element value is to 0. The avoidance embedding phase space reconstruction method uses the univariate time series { x 1, x 2,…, x i ,…, x n The phase point coordinates are obtained by reconstruction. X i =[ x ( i ), x ( i + τ ), x ( i +2 τ ), …, x ( i +( m -1) τ )], i =1, 2, …, n -( m -1) τ When setting parameters m and τ The values ​​are all 1; the membership function is R i,j =cos[ π • d i,j / (2max( d i,j ))], d i,j =|| X i - X j || indicates phase point X i and X j The relative distance between phase points is such that the value of the function is inversely proportional to the distance between the phase points. S4. Using a set of 45° line segments parallel to the main diagonal of the recursive matrix and equally spaced, divide the lower right triangular region of the matrix into several regions, and calculate the mean value GR of the thresholdless recursive matrix elements in each region. a (k), plot GR a (k) The trend of change along the main diagonal to the corner transition direction is obtained by fitting a linear relationship. The slope of the linear function is extracted as the thresholdless recursive measure RT of the friction signal. G Analytical measure RT G The wear stage corresponding to each matrix calculation window is identified based on the changing pattern of the rectangular calculation window, i.e., the break-in process, and the wear stage is determined online.

2. The method for identifying the running-in state based on thresholdless recursive analysis results of friction signals according to claim 1, characterized in that, The acquisition device mentioned in step S1 includes a sensor network and a data acquisition card.

3. The method for identifying the running-in state based on thresholdless recursive analysis results of friction signals according to claim 1, characterized in that, Step S2 includes the following steps: S2.

1. Applying Empirical Mode Decomposition (EMD), the acquired sequence is decomposed into several Mode Functions (IMFs) and a residual component. r ; S2.2 Perform power spectrum analysis on IMFs of each order, and identify the mode functions corresponding to periodic noise components, random noise components, and IMFs with continuous bandwidth and exponentially decaying nonlinear chaotic characteristics based on their power spectrum distribution characteristics. c ; S2.3 For some modal functions that are difficult to identify using the characteristics of power spectrum distribution, residual variance and waveform similarity parameters will be introduced for further analysis. S2.4 Reconstruct the mode function and residual components with nonlinear characteristics to complete the extraction of nonlinear components of the friction signal and provide a data basis for subsequent analysis of the recursive characteristics of the friction signal.

4. The method for identifying the running-in state based on thresholdless recursive analysis results of friction signals according to claim 3, characterized in that, The periodic noise components include vibrations from the testing machine motor and periodic interference from other related equipment within the friction system; the random noise components include background interference during the data acquisition process.

5. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a running-in state identification method based on the results of thresholdless recursive analysis of friction signals as described in any one of claims 1-4.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a break-in state identification method based on the results of thresholdless recursive analysis of friction signals as described in any one of claims 1-4.

Images