A method for measuring time-varying speed of bearing cage considering the influence of oil pollution

Abstract

The present invention discloses a method for measuring the time-varying speed of a bearing cage considering the influence of oil pollution; a plurality of evenly distributed features of the same size are processed on the bearing cage respectively, an eddy current displacement sensor is aligned with the geometric center of one of the features, and eddy current pulse signals at different rotation frequencies of the inner ring of the bearing are collected; the upper limit of the rotation frequency of the cage is estimated, and the eddy current pulse signals that are 3 times higher than the upper limit frequency are filtered out by a FIR low-pass filter; a short-time Fourier transform is performed on the filtered eddy current pulse signal to obtain a preliminary time-frequency diagram, and the time-frequency ridge is extracted by using a path optimization CrazyClimber algorithm; an adaptive short-time Fourier transform is designed to obtain an accurate time-frequency diagram and time-frequency ridge, and the length of the time-frequency window function is adaptively adjusted in combination with the first-order derivative of the accurate ridge function with respect to time, so as to obtain the time-varying speed of the cage. The present invention can realize accurate measurement of the time-varying speed of the cage under variable speed conditions considering the influence of lubricating oil pollution.

Description

Technical Field

[0001] The invention relates to the field of bearing cage rotation speed measurement, and in particular to a bearing cage time-varying rotation speed measurement method taking into account the influence of oil pollution. Background Art

[0002] As the operating requirements of high-end bearings continue to increase, the working environment of bearings has become more complex and changeable. Among the many failures of bearings, cage slippage failure accounts for the largest proportion and is difficult to monitor. As an important indicator for measuring the degree of slippage and studying the slippage mechanism, the bearing cage slippage rate needs to be accurately measured and calculated. There has been some research on the measurement of bearing cage speed. Traditional photoelectric sensors are easily affected by lubricating oil pollution, so it is difficult to ensure the measurement accuracy of bearing cage speed under the influence of oil pollution. Eddy current sensors measure bearing cage speed based on the eddy current effect. They are not easily affected by lubricating oil pollution and can accurately measure the bearing cage speed in harsh environments.

[0003] The eddy current pulse signal measured by the eddy current sensor at different bearing cage speeds can obtain accurate bearing cage speeds through effective data processing. The essence of the time-frequency analysis method is to convert the one-dimensional time domain information of the signal into two-dimensional time-frequency information, and characterize the signal from both time and frequency aspects. The time-frequency window width of the traditional short-time Fourier transform is fixed, which is not suitable for the complex and changeable instantaneous frequency of the signal in engineering practice, and it is difficult to obtain more accurate analysis and measurement results. In view of the above shortcomings, the present invention adaptively adjusts the length of the time-frequency window function based on the first-order derivative of the ridge function with respect to time, and designs an adaptive short-time Fourier transform in combination with low-pass filtering technology to obtain accurate cage variable speed time-frequency diagrams and time-frequency ridges, thereby realizing accurate measurement of the time-varying speed of the bearing cage under the influence of oil pollution and variable working conditions. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a method for measuring the time-varying speed of a bearing cage taking into account the influence of oil pollution, which can accurately measure the time-varying speed of the cage taking into account the influence of oil pollution and variable working conditions.

[0005] In order to achieve the above object, the present invention adopts the following technical solution:

[0006] A method for measuring the time-varying speed of a bearing cage considering the influence of oil pollution, the method comprising:

[0007] Step S1, processing a plurality of evenly distributed features with the same shape on the bearing retainer, arranging an eddy current displacement sensor, and aligning the eddy current displacement sensor with the geometric center of one of the features, and collecting eddy current pulse signals of the features at different bearing inner ring rotation frequencies;

[0008] Step S2, estimating the upper limit rotation frequency of the bearing cage, and using a filter to filter out the components of the eddy current pulse signal obtained in step S1 that are higher than 3 times the upper limit rotation frequency;

[0009] Step S3, performing short-time Fourier transform on the filtered eddy current pulse signal obtained in step S2 to obtain a preliminary time-frequency diagram, and using a path optimization Crazy Climber algorithm to extract preliminary time-frequency ridges;

[0010] Step S4, combining the first-order derivative of the preliminary time-frequency ridge function with respect to time obtained in step S3 to adaptively adjust the length of the time-frequency window function, design an adaptive short-time Fourier transform to obtain an accurate time-frequency diagram and time-frequency ridge of the bearing cage with respect to the speed, and then obtain the time-varying speed of the bearing cage.

[0011] Preferably, the implementation process of step S1 is:

[0012] Step S101, processing a plurality of evenly distributed features having the same shape on the bearing retainer:

[0013] The depth of the feature d f Satisfies formula (1):

[0014]

[0015] Where: w c is the bearing cage width; r c is the axial width of the bearing cage pocket;

[0016] The width of the feature w f Satisfies formula (2):

[0017] w f =0.6d c (2)

[0018] Where: d c is the bearing cage wall thickness;

[0019] Step S102: arranging an eddy current displacement sensor, and aligning the eddy current displacement sensor with the geometric center of one of the features, and collecting eddy current pulse signals of the features at different bearing inner ring rotation frequencies:

[0020] First, determine the measuring distance range between the eddy current displacement sensor and the bearing cage:

[0021] According to the eddy current sensor model, the maximum value of the effective measurement distance of the eddy current sensor is set as the maximum measurement distance, and the minimum measurement distance of the eddy current sensor should be greater than the axial swing limit displacement η of the bearing cage;

[0022] Then, the signal acquisition frequency of the eddy current sensor is adjusted to be no less than 6 times the maximum value of the bearing inner ring rotation frequency, and the characteristic eddy current pulse signal x(t) at different bearing inner ring rotation frequencies is collected.

[0023] Preferably, the shape of the feature is an equilateral triangle or a circle, and when the shape of the feature is an equilateral triangle, the width w of the feature is f is the side length of an equilateral triangle; when the shape of the feature is a circle, the width of the feature w f is the diameter of the circle.

[0024] Preferably, the implementation process of step S2 is:

[0025] Step S201, estimating the upper limit rotation frequency f of the bearing cage cmax , as shown in formula (3):

[0026]

[0027] In the formula, R r is the bearing roller radius; R m is the bearing pitch radius; ω i is the speed of the inner ring of the bearing;

[0028] Step S202: Use a FIR low-pass filter to filter out the components of the eddy current pulse signal x(t) obtained in step S102 that are higher than three times the upper limit rotation frequency:

[0029] Design the FIR low-pass filter h(n), as shown in equation (4):

[0030] h(n)=h d (n)ω R (n)n=0,1,2,...,N-1 (4)

[0031] Where: h d (n) is the infinite length impulse response; N-1 is the order of the FIR low-pass filter; h(n) is the finite length impulse response after Hamming window truncation; ω R (n) is the Hamming window function, satisfying formula (5):

[0032]

[0033] Where: N is the length of the Hamming window;

[0034] FIR low-pass filter frequency response function H(e jω ) satisfies formula (6):

[0035]

[0036] Where: ω is the frequency of the FIR low-pass filter input signal x(t); e is the natural exponent; j is the imaginary unit;

[0037] The eddy current pulse signal x(t) is input as the input signal to the FIR low-pass filter. The output signal y(t) of the FIR low-pass filter is the filtered eddy current pulse signal. The frequency domain relationship between the input and output signals of the FIR low-pass filter is as follows:

[0038] Y(e jω )=X(e jω )H(e jω ) (7)

[0039] Where: X(e jω ) is the frequency domain transform of the FIR low-pass filter input signal x(t); Y(e jω ) is the frequency domain transform of the FIR low-pass filter output signal y(t);

[0040] According to the upper limit rotation frequency of the bearing cage obtained in step S201, the cut-off frequency f of the FIR low-pass filter is set. t satisfy:

[0041] f t =f cmax (8).

[0042] Preferably, the implementation process of step S3 is:

[0043] Step S301, performing short-time Fourier transform on the filtered eddy current pulse signal y(t) obtained in step S2 to obtain a preliminary time-frequency diagram:

[0044]

[0045] Where g(τ-t) represents the short-time Fourier transform window function; g * represents the conjugate of the variable g; F(t,ω) represents the preliminary time-frequency matrix of the FIR low-pass filter output signal y(t) at time t; τ is the microvariable at time t;

[0046] Perform Fourier transform based on the obtained preliminary time-frequency matrix F(t,ω) to obtain a preliminary time-frequency diagram;

[0047] Step S302: Extract preliminary time-frequency ridges using the preliminary time-frequency graph obtained in step S301 using the path optimization Crazy Climber algorithm:

[0048] The preliminary time-frequency matrix F(t,ω) obtained by short-time Fourier transform in step S301 is used as the input of the path optimization CrazyClimber algorithm; wherein the preliminary time-frequency matrix F(t,ω) is a matrix with M rows and N columns;

[0049] (1) Initialize the metric matrix D with the same size as the initial time-frequency matrix F(t,ω) R , set K movable climbing points of the climber and the initial temperature T(0) = max[F(i,j)]-min[F(i,j)]; where i = 1, 2, ..., M, j = 1, 2, ..., N, max(·) is the maximum value of the variable, and min(·) is the minimum value of the variable;

[0050] (2) Climber can move the climbing point from the position Γ at the tth moment t =(i,j) moves to the position Γ at time t+1 t+1 =(i',j'); where i'=1,2,...,M and j'=1,2,...,N; the climber can move the climbing point first to the left and right, with the probability of moving left and right being 0.5, and the probability of not moving being 0.5; the climber can move the climbing point and then move up and down, with the probability of moving up and down being p λ , the probability of not moving is 1-p λ , where p λ The calculation formula satisfies formula (10):

[0051]

[0052] Where T(t) is the current temperature of the Climber movable climbing point after it moves;

[0053] (3) After the Climber movable climbing point is moved, in the metric matrix D R Increase the metric value 1 at the corresponding position and update the metric matrix D R , as shown in formula (11):

[0054]

[0055] In the formula, is the metric matrix of the kth step; is the metric matrix of the k+1th step;

[0056] (4) Repeat (2) and (3) until the temperature of the Climber movable climbing point reaches the set temperature value T d =T(0) / 50, the iteration ends, and the final metric matrix D' is obtained R ;

[0057] (5) For the final metric matrix D' R Traverse in the time direction, for the final measurement matrix D' R Find the optimal adjacent point of any point in the image to obtain the preliminary time-frequency ridge line.

[0058] Preferably, the implementation process of step S4 is:

[0059] Step S401, using the least squares method to fit the preliminary time-frequency ridge obtained in step S302 to obtain a preliminary ridge function v(t);

[0060] Step S402: adjust the length of the time-frequency window function according to the first-order derivative v'(t) of the time-frequency preliminary ridge function v(t) obtained in step S401, and obtain an accurate time-frequency diagram of the cage with variable speed, the first-order derivative v'(t) and the length of the time-frequency window function T t The relationship is:

[0061]

[0062] Step S403, obtaining the time-varying speed of the bearing cage:

[0063] Based on the precise cage variable speed time-frequency diagram obtained in step S402, the path optimization Crazy Climber algorithm described in step S302 is used to extract the precise time-frequency ridge line, which is the time-varying speed of the bearing cage.

[0064] The beneficial effects of the present invention are:

[0065] 1. Based on the first-order derivative of the ridge function with respect to time, the length of the time-frequency window function is adaptively adjusted, and the adaptive short-time Fourier transform is designed in combination with the low-pass filtering technology to obtain the accurate cage variable speed time-frequency diagram and time-frequency ridge;

[0066] 2. The principle of processing the bearing cage features is summarized as follows: appropriate feature parameters such as shape, depth, number and size should be selected. The feature shape has a greater impact on the measurement accuracy of the bearing cage speed. It is recommended to process the circular hole feature first. When the feature is a circular hole, the depth has the greatest impact, followed by the number, and the diameter has the least impact. When the feature is an equilateral triangle, the side length has the greatest impact, followed by the number, and the depth has the least impact.

[0067] 3. A complete method for measuring the speed of a bearing cage based on an eddy current displacement sensor is proposed, which can effectively overcome the influence of lubricating oil contamination on the measurement accuracy of the bearing cage speed and can realize accurate measurement of the cage's time-varying speed while considering the influence of oil contamination and variable working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 A schematic flow chart of a method for measuring the time-varying speed of a bearing cage taking into account the influence of oil pollution provided in Example 1;

[0069] Figure 2The eddy current pulse signal of the bearing cage speed provided in Example 1;

[0070] Figure 3 is the filtered eddy current pulse signal of the bearing cage speed provided in Example 1, wherein a is the amplitude and phase of the low-pass filter, and b is the eddy current pulse signal after the low-pass filter;

[0071] Figure 4 The preliminary time-frequency diagram of the inner ring of the bearing provided in Example 1;

[0072] Figure 5 is the time-frequency diagram after filtering and denoising provided in Example 1, wherein a is the time-frequency diagram after filtering and denoising of the bearing inner ring speed, and b is the time-frequency diagram after filtering and denoising of the bearing retainer speed;

[0073] Figure 6 is the preliminary time-frequency ridgeline diagram provided in Example 1, wherein a is the preliminary time-frequency ridgeline diagram of the bearing inner ring speed, and b is the preliminary time-frequency ridgeline diagram of the bearing cage speed;

[0074] Figure 7 The schematic diagram of the adaptive short-time Fourier transform provided in Example 1;

[0075] Figure 8 It is the precise time-frequency ridgeline diagram provided in Example 1, wherein a is the precise time-frequency ridgeline diagram of the bearing inner ring speed, and b is the precise time-frequency ridgeline diagram of the bearing cage speed;

[0076] Fig. 9 This is a comparison chart of the rotational speeds of the bearing inner ring and the cage provided in Example 1. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] Example 1

[0079] See also Figure 1-Figure 9 This embodiment provides a method for measuring the time-varying speed of a bearing cage taking into account the influence of oil pollution. The process of the method is as follows: Figure 1 As shown, this embodiment uses the N208 bearing cage speed measurement as an example to verify the algorithm performance. The relevant structural parameters of the N208 bearing are shown in Table 1:

[0080] Table 1N208 bearing related structural parameters

[0081]

[0082] The strategy specifically includes the following steps:

[0083] Step S1, processing three evenly distributed features with the same shape on the bearing cage respectively, arranging eddy current displacement sensors, aligning the eddy current displacement sensors with the geometric center of the features, and collecting eddy current pulse signals of the features at different bearing inner ring rotation frequencies;

[0084] Specifically, in this embodiment, step S1 includes:

[0085] Step S101, processing three evenly distributed features with the same size on the bearing retainer respectively:

[0086] Three evenly distributed features of the same depth and size are machined on the bearing cage, where the feature depth d f Satisfies formula (1):

[0087]

[0088] Where: w c is the bearing cage width; r c is the axial width of the bearing cage pocket;

[0089] Characteristic diameter w f Satisfies formula (2):

[0090] w f =0.6d c =0.6*5.40=3.24mm (2)

[0091] Where: d c is the wall thickness of the bearing cage; in Example 1, various types of features were processed for testing, and the specific feature sizes are shown in Table 2:

[0092] Table 2 Feature dimensions

[0093]

[0094] Step S102: Arrange the eddy current displacement sensor, align the eddy current displacement sensor with the characteristic geometric center, and collect eddy current pulse signals at different bearing inner ring rotation frequencies:

[0095] Arrange the eddy current displacement sensor and align it with the characteristic geometric center. The measuring distance range between the eddy current displacement sensor and the bearing cage is determined according to the following principles:

[0096] According to the model of the eddy current sensor, the maximum value of the effective measurement distance is set as the maximum measurement distance, which is selected as 5mm in the embodiment; the minimum measurement distance is determined according to the lubricating oil flow rate, the rotation speed of the inner ring of the bearing, and the characteristic size and depth; the value of the minimum measurement distance can be increased when the lubricating oil flow rate decreases, the rotation speed of the inner ring of the bearing, and the characteristic size and depth increase; the value of the minimum measurement distance needs to be reduced when the lubricating oil flow rate increases, the rotation speed of the inner ring of the bearing, and the characteristic size and depth decrease; the value of the minimum measurement distance needs to be greater than the axial swing limit displacement η of the bearing retainer; the test condition parameters in Example 1 are shown in Table 3:

[0097] Table 3 Bearing test conditions

[0098]

[0099] The signal acquisition frequency is adjusted to be not less than 6 times the maximum rotation frequency of the bearing inner ring, which is 300 Hz in Example 1. The eddy current pulse signal x(t) at different bearing inner ring rotation frequencies is collected. Three evenly distributed circular features with a diameter of 3 mm and a depth of 3 mm are processed on the bearing retainer. The eddy current pulse signal measured under the condition of a lubricating oil flow rate of 2.5 L / min is as follows: Figure 2 shown.

[0100] Step S2, estimating the upper limit of the cage rotation frequency according to the bearing inner ring speed and the bearing structure, and using a FIR low-pass filter to filter out the components of the eddy current pulse signal obtained in step S1 that are higher than 3 times the upper limit frequency;

[0101] Specifically, in this embodiment, step S2 includes:

[0102] Step S201, estimating the theoretical speed of the inner ring of the cage according to the speed of the inner ring of the bearing and the bearing structure:

[0103]

[0104] Where: R r is the roller radius; R m is the bearing pitch radius; ω i is the inner ring speed; ω cm is the theoretical speed of the bearing cage;

[0105] Step S202: Use a FIR low-pass filter to filter out the components of the eddy current pulse signal obtained in step S102 that are higher than three times the upper limit frequency:

[0106] Design a FIR low-pass filter as follows:

[0107] h(n)=h d (n)ω R (n)n=0,1,2,...,N-1 (4)

[0108] Where: h d (n) is the infinite length impulse response; N-1 is the filter order; h(n) is the finite length impulse response after Hamming window truncation; ω R (n) is the Hamming window function, satisfying formula (5):

[0109]

[0110] Where: N is the length of the Hamming window;

[0111]

[0112] Where: ω is the frequency of the FIR low-pass filter input signal x(t); e is the natural exponent; j is the imaginary unit; H(e jω ) is the filter frequency response function; the input and output frequency domain relationship of the FIR low-pass filter is as follows:

[0113] Y(e jω )=X(e jω )H(e jω ) (7)

[0114] Where: X(e jω ) is the frequency domain transform of the filter input signal x(t); Y(e jω ) is the frequency domain transform of the filter output signal y(t);

[0115] According to the upper limit of the cage rotation frequency obtained in step S201, the cut-off frequency f of the low-pass filter is set. t for:

[0116]

[0117] The filtered eddy current pulse signal is as follows: Figure 3 shown.

[0118] Step S3, performing short-time Fourier transform on the filtered eddy current pulse signal obtained in step S2 to obtain a preliminary time-frequency diagram, and using a path optimization Crazy Climber algorithm to extract preliminary time-frequency ridges;

[0119] Specifically, in this embodiment, step S3 includes:

[0120] Step S301, perform short-time Fourier transform on the filtered eddy current pulse signal obtained in step S2 to obtain a preliminary time-frequency diagram:

[0121]

[0122] Where: g(τ-t) represents the short-time Fourier transform window function; g *(τ-t) means taking the conjugate of g(τ-t); F(t,ω) means the preliminary time-frequency matrix of the eddy current pulse signal to be analyzed at time t; τ is the microvariable at time t; the preliminary time-frequency diagram is as follows Figure 4 As shown in the figure, the time-frequency diagram after filtering and denoising is as follows Figure 5 shown.

[0123] Based on the obtained preliminary time-frequency matrix F(t,ω), Fourier transform is performed to obtain a preliminary time-frequency diagram.

[0124] Step S302: Use the path optimization Crazy Climber algorithm in step S301 to obtain the preliminary time-frequency graph and extract the preliminary time-frequency ridgeline:

[0125] The preliminary time-frequency matrix F(t,ω) obtained by short-time Fourier transform in step S301 is used as the input of the path optimization CrazyClimber algorithm; wherein the time-frequency matrix F(t,ω) is an M-row and N-column matrix; the ridge extraction steps based on the path optimization CrazyClimber algorithm include:

[0126] (1) Initialization and preliminary time-frequency matrix C R The metric zero matrix D with the same dimensions R , K climbers and initial temperature T(0) = max[C R (i,j)]-min[C R (i, j)] = 1000; where i = 1, 2, ..., M, j = 1, 2, ..., N, max(·) is the maximum value of the variable, and min(·) is the minimum value of the variable;

[0127] (2) The climber can move the climbing point from the position Γ at the tth moment t =(i,j) moves to the position Γ at time t+1 t+1 =(i',j'); where i'=1,2,...,M and j'=1,2,...,N; the climber can move the climbing point horizontally first, with the probability of moving left and right being 0.5, and the probability of not moving being 0.5; the climber can move the climbing point and then move up and down, with the probability of moving up and down being p λ , the probability of not moving is 1-p λ , where p λ The calculation formula satisfies:

[0128]

[0129] Where: T(t) is the current temperature updated after the Climber movable climbing point moves;

[0130] (3) After the climber moves, in the metric matrix DR By adding the metric value 1 to the corresponding position of , we can get the following formula:

[0131]

[0132] In the formula, is the metric matrix of the kth step; is the metric matrix of the k+1th step;

[0133] (4) Repeat the climbing process in steps (2) and (3) until the temperature of the movable climbing point of the Climber reaches the set value T d =T(0) / 50=20, the iteration ends, and the final metric matrix D' is obtained R ;

[0134] (5) For the final metric matrix D' R Traverse in the time direction, for the final measurement matrix D' R Find the optimal adjacent point of any point in the image to obtain the preliminary time-frequency ridge line.

[0135] Step S4, combining the first-order derivative of the preliminary time-frequency ridge function obtained in step S3 with respect to time, adaptively adjusting the length of the time-frequency window function, designing an adaptive short-time Fourier transform to obtain an accurate cage variable speed time-frequency diagram and time-frequency ridge, and then obtaining the cage time-varying speed;

[0136] Specifically, in this embodiment, step S4 includes:

[0137] Step S401: Combine the preliminary time-frequency ridge obtained in step S302 with the least squares method to obtain a preliminary ridge function f(t). The preliminary time-frequency ridge function is as follows: Figure 6 shown.

[0138] Step S402: Combine the time-frequency preliminary ridge function v(t) obtained in step S401, adaptively adjust the time-frequency window function length according to its first-order derivative with respect to time, and design an adaptive short-time Fourier transform to obtain an accurate cage variable speed time-frequency diagram:

[0139]

[0140] Where: v'(t) represents the first-order derivative of the ridge function with respect to time; T t Represents the length of the time-frequency window function; its principle is as follows Figure 7 shown.

[0141] Step S403, obtaining the time-varying speed of the cage:

[0142] Based on the precise time-frequency diagram of cage speed variation obtained in step S402, the path optimization Crazy Climber algorithm described in step S302 is used to extract the precise time-frequency ridge line, which is the cage speed variation. The precise time-frequency ridge line of the bearing inner ring speed and cage speed is as follows: Figure 8 As shown. Fig. 9 As shown, it can be found that the theoretical speed of the bearing cage is 0.41 times the speed of the bearing inner ring, which corresponds to the result obtained by formula (3), where the speed is 60 times the rotation frequency.

[0143] The matters not described in detail in the present invention are all known technologies to those skilled in the art.

[0144] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for measuring the time-varying speed of a bearing cage considering the influence of oil pollution, characterized in that: The following steps are involved: Step S1, processing a plurality of evenly distributed features with the same shape on the bearing retainer, arranging an eddy current displacement sensor, and aligning the eddy current displacement sensor with the geometric center of one of the features, and collecting eddy current pulse signals of the features at different bearing inner ring rotation frequencies; Step S2, estimating the upper limit rotation frequency of the bearing cage, and using a filter to filter out the components of the eddy current pulse signal obtained in step S1 that are higher than 3 times the upper limit rotation frequency; Step S3, performing short-time Fourier transform on the filtered eddy current pulse signal obtained in step S2 to obtain a preliminary time-frequency diagram, and using a path optimization Crazy Climber algorithm to extract preliminary time-frequency ridges; Step S4, combining the first-order derivative of the preliminary time-frequency ridge function with respect to time obtained in step S3 to adaptively adjust the length of the time-frequency window function, design an adaptive short-time Fourier transform to obtain an accurate time-frequency diagram and time-frequency ridge of the bearing cage with respect to the speed, and then obtain the time-varying speed of the bearing cage.

2. A method for measuring the time-varying speed of a bearing cage considering the influence of oil pollution according to claim 1, characterized in that: The implementation process of step S1 is: Step S101, processing a plurality of evenly distributed features having the same shape on the bearing retainer: The depth of the feature d f Satisfies formula (1): Where: w c is the bearing cage width; r c is the axial width of the bearing cage pocket; The width of the feature w f Satisfies formula (2): In f =0.6d c (2) Where: d c is the bearing cage wall thickness; Step S102: arranging an eddy current displacement sensor, and aligning the eddy current displacement sensor with the geometric center of one of the features, and collecting eddy current pulse signals of the features at different bearing inner ring rotation frequencies: First, determine the measuring distance range between the eddy current displacement sensor and the bearing cage: According to the eddy current sensor model, the maximum value of the effective measurement distance of the eddy current sensor is set as the maximum measurement distance, and the minimum measurement distance of the eddy current sensor should be greater than the axial swing limit displacement η of the bearing cage; Then, the signal acquisition frequency of the eddy current sensor is adjusted to be no less than 6 times the maximum value of the bearing inner ring rotation frequency, and the characteristic eddy current pulse signal x(t) at different bearing inner ring rotation frequencies is collected.

3. A method for measuring the time-varying speed of a bearing cage considering the influence of oil pollution according to claim 1 or 2, characterized in that: The shape of the feature is an equilateral triangle or a circle, and when the shape of the feature is an equilateral triangle, the width w of the feature is f is the side length of an equilateral triangle; when the shape of the feature is a circle, the width of the feature w f is the diameter of the circle.

4. A method for measuring the time-varying speed of a bearing cage considering the influence of oil pollution according to claim 3, characterized in that: The implementation process of step S2 is: Step S201, estimating the upper limit rotation frequency f of the bearing cage cmax , as shown in formula (3): In the formula, R r is the bearing roller radius; R m is the bearing pitch radius; ω i is the speed of the inner ring of the bearing; Step S202: Use a FIR low-pass filter to filter out the components of the eddy current pulse signal x(t) obtained in step S102 that are higher than three times the upper limit rotation frequency: Design the FIR low-pass filter h(n), as shown in equation (4): h(n)=h d (n)ω R (n)n=0,1,2,...,N-1 (4) Where: h d (n) is the infinite length impulse response; N-1 is the order of the FIR low-pass filter; h(n) is the finite length impulse response after Hamming window truncation; ω R (n) is the Hamming window function, satisfying formula (5): Where: N is the length of the Hamming window; FIR low-pass filter frequency response function H(e jω ) satisfies formula (6): Where: ω is the frequency of the FIR low-pass filter input signal x(t); e is the natural exponent; j is the imaginary unit; The eddy current pulse signal x(t) is input as the input signal to the FIR low-pass filter. The output signal y(t) of the FIR low-pass filter is the filtered eddy current pulse signal. The frequency domain relationship between the input and output signals of the FIR low-pass filter is as follows: Y(e jω )=X(e jω )H(e jω ) (7) Where: X(e jω ) is the frequency domain transform of the FIR low-pass filter input signal x(t); Y(e jω ) is the frequency domain transform of the FIR low-pass filter output signal y(t); According to the upper limit rotation frequency of the bearing cage obtained in step S201, the cut-off frequency f of the FIR low-pass filter is set. t satisfy: f t =f cmax (8)。 5. A method for measuring the time-varying speed of a bearing cage considering the influence of oil pollution according to claim 4, characterized in that: The implementation process of step S3 is: Step S301, performing short-time Fourier transform on the filtered eddy current pulse signal y(t) obtained in step S2 to obtain a preliminary time-frequency diagram: Where g(τ-t) represents the short-time Fourier transform window function; g * represents the conjugate of the variable g; F(t,ω) represents the preliminary time-frequency matrix of the FIR low-pass filter output signal y(t) at time t; τ is the microvariable at time t; Perform Fourier transform based on the obtained preliminary time-frequency matrix F(t,ω) to obtain a preliminary time-frequency diagram; Step S302: Extract preliminary time-frequency ridges using the preliminary time-frequency graph obtained in step S301 using the path optimization Crazy Climber algorithm: The preliminary time-frequency matrix F(t,ω) obtained by short-time Fourier transform in step S301 is used as the input of the path optimization CrazyClimber algorithm; wherein the preliminary time-frequency matrix F(t,ω) is a matrix with M rows and N columns; (1) Initialize the metric matrix D with the same size as the initial time-frequency matrix F(t,ω) R , set K movable climbing points of the climber and the initial temperature T(0) = max[F(i,j)]-min[F(i,j)]; where i = 1, 2, ..., M, j = 1, 2, ..., N, max(·) is the maximum value of the variable, and min(·) is the minimum value of the variable; (2) Climber can move the climbing point from the position Γ at the tth moment t =(i,j) moves to the position Γ at time t+1 t+1 =(i',j'); where i'=1,2,...,M and j'=1,2,...,N...; Climber can move the climbing point first to the left and right, with the probability of moving left and right being 0.5, and the probability of not moving being 0.5; Climber can move the climbing point and then move up and down, with the probability of moving up and down being p λ , the probability of not moving is 1-p λ , where p λ The calculation formula satisfies formula (10): Where, T(t) is the current temperature of the Climber movable climbing point after it moves; (3) After the Climber movable climbing point is moved, in the metric matrix D R Increase the metric value 1 at the corresponding position and update the metric matrix D R , as shown in formula (11): In the formula, is the metric matrix of the kth step; is the metric matrix of the k+1th step; (4) Repeat (2) and (3) until the temperature of the Climber movable climbing point reaches the set temperature value T d =T(0) / 50, the iteration ends, and the final metric matrix D' is obtained R ; (5) For the final metric matrix D' R Traverse in the time direction, for the final measurement matrix D' R Find the optimal adjacent point of any point in the image to obtain the preliminary time-frequency ridge line.

6. A method for measuring the time-varying speed of a bearing cage considering the influence of oil pollution according to claim 5, characterized in that: The implementation process of step S4 is: Step S401, using the least squares method to fit the preliminary time-frequency ridge obtained in step S302 to obtain a preliminary ridge function v(t); Step S402: adjust the length of the time-frequency window function according to the first-order derivative v'(t) of the time-frequency preliminary ridge function v(t) obtained in step S401, and obtain an accurate time-frequency diagram of the cage with variable speed, the first-order derivative v'(t) and the length of the time-frequency window function T t The relationship is: Step S403, obtaining the time-varying speed of the bearing cage: Based on the precise cage variable speed time-frequency diagram obtained in step S402, the path optimization CrazyClimber algorithm described in step S302 is used to extract the precise time-frequency ridge line, which is the time-varying speed of the bearing cage.

Images