Multi-feature fusion lead screw pair transmission precision estimation method

Abstract

The invention discloses a multi-feature fusion lead screw pair transmission precision estimation method, which comprises the following steps: screening from a precision estimation model library to obtain a precision estimation model M matched with specification information and current working condition information of a lead screw to be detected, and determining a sensitive feature set C corresponding to the precision estimation model M; the method comprises the following steps: placing a to-be-tested screw rod on a loading test bench, carrying out a loading test, and obtaining test data A; after wavelet denoising and standardization processing are carried out on the test data A, 12-dimensional time domain features and 4-dimensional frequency domain features are extracted, and a to-be-input feature set is obtained; sensitive features indicated by the sensitive feature set C are screened from the to-be-input feature set and serve as input of the precision estimation model M, and a precision estimation result of the to-be-detected lead screw is output after the sensitive features pass through the precision estimation model. According to the method, the problems that in the prior art, the measurement accuracy is greatly influenced by assembly errors, the measurement real-time performance is poor, information of a single sensor is incomplete, and the working condition and lead screw specification adaptation capacity is insufficient are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of predictive health management of mechanical systems, and in particular relates to a method for estimating the transmission accuracy of a screw pair by integrating multiple features. Background Art

[0002] A lead screw is a mechanical transmission device that can realize the conversion between rotational motion and linear motion. It has been widely used in CNC machine tools, aerospace, automobile manufacturing and other fields.

[0003] The measurement of screw transmission accuracy is one of the basic methods to evaluate screw reliability. Common screw accuracy measurement methods include three categories:

[0004] The first type is to directly measure the screw accuracy through a special precision measurement test bench. The main problem is that the screw needs to be manually disassembled and assembled when measuring the screw accuracy, and real-time measurement is impossible, and the measurement result is greatly affected by the assembly error.

[0005] The second type is to integrate the precision measuring device into the loading test bench to measure the screw precision. The main problem is that the screw precision measurement must be performed without the magnetic powder brake being activated, that is, the screw precision cannot be measured in real time under the loading state. In addition, the design and tooling of the device are complex, and it is difficult to achieve this through simple modification of the conventional loading bench, which is not conducive to its promotion and application.

[0006] The third type is to use sensor signals to indirectly measure the accuracy of the screw. By monitoring the vibration signal of the ball screw pair under simulated working conditions and mapping the vibration signal to the screw accuracy, the accuracy of the screw pair can be predicted. The main problem is that the vibration signal from a single dimension cannot fully and accurately reflect the decay process of the screw accuracy. At the same time, this method is only suitable for fixed screw specifications and working conditions, and lacks the ability to predict the accuracy of screws of different specifications under various working conditions. Summary of the invention

[0007] The technology of the present invention solves the problem: Overcoming the shortcomings of the prior art, providing a method for estimating the transmission accuracy of a screw pair by fusion of multiple features, so as to solve the problems in the prior art such as the measurement accuracy being greatly affected by assembly errors, poor measurement real-time performance, incomplete information from a single sensor, and insufficient adaptability to working conditions and screw specifications.

[0008] In order to solve the above technical problems, the present invention discloses a multi-feature fusion screw pair transmission accuracy estimation method, comprising:

[0009] According to the specification information and current working condition information of the screw rod to be tested, an accuracy estimation model M matching the specification information and current working condition information of the screw rod to be tested is screened from the accuracy estimation model library, and a sensitive feature set C corresponding to the accuracy estimation model M is determined;

[0010] Place the screw rod to be tested on the loading test bench, perform a loading test, and obtain test data A;

[0011] After wavelet denoising and standardization processing of the test data A, 12-dimensional time domain features and 4-dimensional frequency domain features are extracted to obtain the feature set to be input;

[0012] The sensitive features indicated by the sensitive feature set C are screened from the feature set to be input and used as the input of the accuracy estimation model M. After the accuracy estimation model is used, the accuracy estimation result of the screw rod to be tested is output.

[0013] In the above-mentioned multi-feature fusion screw pair transmission accuracy estimation method, the loading test bench includes: a motor, a motor support seat, a coupling A, a torque sensor, a coupling B, a transmission shaft, a screw tooling bearing seat, a coupling C, an actuating sleeve, a two-force rod, a tension sensor, a loading slide, a slider, a linear guide, a linear grating ruler, a machine base, a spherical bearing, a hydraulic cylinder, a hydraulic cylinder bracket, a control console, a vibration sensor A, a temperature sensor A, a temperature sensor B, a vibration sensor C, a vibration sensor D and a vibration sensor B; wherein, the screw to be tested includes: a screw rod and screw nuts located at both ends of the screw rod;

[0014] The motor is mounted on the machine base through a motor support base;

[0015] The output end of the motor is connected to one end of the torque sensor through coupling A; the other end of the torque sensor is connected to the transmission shaft through coupling B; wherein the torque sensor is used to measure the output torque of the motor;

[0016] The output end of the transmission shaft is connected to the lead screw through the coupling C inside the lead screw tooling bearing seat; the lead screw tooling bearing seat is installed on the machine base;

[0017] The output end of the screw tooling bearing seat is connected to the actuating sleeve through a bolt, the screw rod passes through the actuating sleeve, and the actuating sleeve cooperates with the limiting short shaft on the upper end surface of the screw nut through a bearing; the screw nut is connected to one end of the two-force rod through a thread, the other end of the two-force rod is connected to one end of the tension sensor, and the other end of the tension sensor is connected to one side of the loading slide support plate; wherein the tension sensor is used to measure the output tension pressure of the screw;

[0018] The loading slide is mounted on a linear guide rail through a slider, and the linear guide rail is mounted on the machine base;

[0019] The linear grating scale is used as a displacement sensor to measure the displacement of the lead screw. The linear grating scale includes: a linear grating scale reading head and a linear grating scale main scale. The linear grating scale reading head is installed on one side of the loading slide support plate, and the linear grating scale main scale is installed on the machine base parallel to the linear guide rail.

[0020] The other side of the loading slide is connected to the end of the hydraulic cylinder push rod through a joint bearing; the hydraulic cylinder is installed on the machine base through a hydraulic cylinder bracket;

[0021] The control console is arranged on the side of the machine base and connected to the motor through a wire;

[0022] The torque sensor, the tension sensor, the linear grating ruler, the vibration sensor A, the temperature sensor A, the temperature sensor B, the vibration sensor C, the vibration sensor D and the vibration sensor B are respectively connected to the control console through wires;

[0023] Temperature sensor A is installed on the upper end surface of the motor to measure the temperature of the upper end surface of the motor;

[0024] The temperature sensor B and the vibration sensor C are installed on the upper end surface of the screw nut, and are used to measure the upper end surface temperature of the screw nut and the upper end surface vibration data of the screw nut respectively;

[0025] The vibration sensor D is installed on the side end face of the screw nut to measure the vibration data of the side end face of the screw nut;

[0026] The vibration sensor B is installed on the upper end surface of the actuating sleeve to measure the vibration data of the upper end surface of the actuating sleeve;

[0027] The vibration sensor A is installed on the side end surface of the actuating sleeve and is used to measure the vibration data of the side end surface of the actuating sleeve.

[0028] In the above-mentioned multi-feature fusion screw pair transmission accuracy estimation method, it also includes: constructing an accuracy estimation model library.

[0029] In the above-mentioned multi-feature fusion screw pair transmission accuracy estimation method, the construction process of the accuracy estimation model library is as follows:

[0030] S11, selecting a sample screw for model training and setting a test condition;

[0031] S12, placing the sample lead screw on a loading test bench, and performing a loading test based on the set test conditions, wherein the motor drives the sample lead screw to reciprocate on the loading test bench;

[0032] S13, after the sample lead screw has reciprocated several times, the sample lead screw is disassembled and assembled to the precision measuring table, and the precision of the sample lead screw is measured by the precision measuring table; if the precision level of the sample lead screw reaches G5, step S14 is executed; if the precision level of the sample lead screw does not reach G5, steps S12 to S13 are repeated until the precision level of the sample lead screw reaches G5;

[0033] S14, collecting and obtaining the loading test data of the entire test cycle of the loading test;

[0034] S15, inputting the set test conditions, specification information of the sample lead screw, loading test data of the entire test cycle, and precision measurement data into the model library extension interface, performing model training, and obtaining a precision estimation model corresponding to the sample lead screw;

[0035] S16, repeating steps S11 to S15, obtaining a number of accuracy estimation models under different specifications and different working conditions, and then constructing an accuracy estimation model library.

[0036] In the above-mentioned multi-feature fusion screw pair transmission accuracy estimation method, the loading test data of the entire test cycle include: side end face vibration data of the actuating sleeve, upper end face vibration data of the actuating sleeve, upper end face vibration data of the screw nut, side end face vibration data of the screw nut, upper end face temperature of the motor, upper end face temperature of the screw nut, motor output torque, screw output tension and pressure, and screw displacement.

[0037] In the above-mentioned multi-feature fusion screw pair transmission accuracy estimation method, the set test conditions, the specification information of the sample screw, the loading test data of the whole test cycle and the accuracy measurement data are input into the model library extension interface for model training to obtain the accuracy estimation model corresponding to the sample screw, including:

[0038] S151, after wavelet denoising and standardization processing of the loading test data of the entire test cycle, 12-dimensional time domain features and 4-dimensional frequency domain features of each data are constructed to obtain a feature data set;

[0039] S152, performing trend and correlation analysis on each feature in the feature data set, selecting the sensitive features of screw precision degradation according to the maximum trend and minimum correlation principle, and constructing a sensitive feature set;

[0040] S153, based on the heuristic knowledge that the accuracy degradation trend is strongly monotonic and the degradation threshold fluctuation is weak, a sensitive feature fusion optimization model is constructed, that is, the accuracy estimation model corresponding to the sample lead screw.

[0041] In the above-mentioned multi-feature fusion screw pair transmission accuracy estimation method, trend and correlation analysis is performed on each feature in the feature data set. According to the maximum trend and minimum correlation principle, the screw accuracy degradation sensitive features are selected to construct a sensitive feature set, including:

[0042] According to the principle of including most sensor data, the trend threshold is set to q;

[0043] Calculate the trend of each feature in the feature data set:

[0044]

[0045] Among them, Tre(F k) represents the trend of the kth feature in the feature data set, F k represents the kth feature, F k，t represents the value of the kth feature in the tth test cycle, t = 1, 2, ..., T, T represents the total number of test cycles;

[0046] Select features with a trend greater than q from the feature data set and put them into the set Q;

[0047] Calculate the correlation between the features in set Q:

[0048]

[0049] Among them, MIC(Y:X) represents the correlation between feature Y and feature X in the feature data set, H(Y) represents the information entropy of feature Y, and H(Y|X) represents the conditional entropy of feature Y relative to feature X; p(x i ) indicates that the feature X takes the value x i The probability of x i represents the value of feature X in the i-th test cycle, i = 1, 2, ..., T; p(y j ) indicates that the feature Y takes the value y j The probability of y j represents the value of feature Y in the jth test cycle, j = 1, 2, ..., T; p(x i ,y j ) indicates that the feature X takes the value x i 、The feature Y value is y j probability;

[0050] Features with correlation less than 1 are selected from the set Q as sensitive features of screw precision degradation, and are put into the set W, which is the sensitive feature set.

[0051] In the above-mentioned multi-feature fusion screw pair transmission accuracy estimation method, the accuracy estimation model is expressed as follows:

[0052] S acc =ω1F1+ω2F2+…+ω s F s

[0053] Among them, S acc Represents the accuracy estimation result, ω1, ω2, ..., ω s Represent features F1, F2, ..., F s The corresponding weights, F1, F2, ..., F s Represents each sensitive feature of the sensitive feature set.

[0054] In the above-mentioned multi-feature fusion screw pair transmission accuracy estimation method, the mathematical description of the model optimization process is as follows:

[0055]

[0056] stw′M′1=1,Mw≥0;E τ w τ ≥0

[0057] ε τ，υ ≥0, τ=1,...,m, υ=1,...,T τ -1

[0058] Among them, r represents the regularization coefficient, ε τ，υ represents the slack variable, c τ，υ represents the slack variable ε τ，υ The weight coefficient, T τ represents the maximum observation period of individual τ, m represents the total number of sample screws, Y represents the matrix of the last observation before individual failure, D represents the symmetric matrix, w τ Represents the vector composed of multi-feature fusion weight coefficients and slack variables, E τ represents the matrix that measures the monotonicity of accuracy degradation; w represents the multi-feature fusion weight coefficient vector, w=[ω1ω2…ω s ]; M represents a diagonal matrix. If the corresponding sensor data shows an increasing / decreasing trend, the corresponding diagonal position is 1 / -1; υ=1, ..., T τ -1.

[0059] In the above-mentioned multi-feature fusion screw pair transmission accuracy estimation method, 12-dimensional time domain features include: minimum value, maximum value, rectified mean, root amplitude, root mean square value, variance, waveform factor, peak factor, pulse factor, margin factor, kurtosis and skewness; 4-dimensional frequency domain features include: mean frequency, center of gravity frequency, root mean square frequency and frequency standard deviation.

[0060] The present invention has the following advantages:

[0061] (1) The present invention discloses a method for estimating the transmission accuracy of a screw pair by integrating multiple features, and proposes a model for estimating the transmission accuracy of a screw pair by integrating multiple features, thereby reducing the dependence on a screw accuracy testing device, thereby avoiding the problem that the measurement results are greatly disturbed by assembly errors due to frequent disassembly and assembly.

[0062] (2) The present invention discloses a method for estimating the transmission accuracy of a screw pair by integrating multiple features, and proposes a method for estimating the accuracy of a screw pair that integrates data collection, feature extraction, model optimization, and model calculation. The entire process does not require shutdown or other operations that may hinder the normal operation of the screw, and real-time online estimation of the screw accuracy can be achieved.

[0063] (3) The present invention discloses a method for estimating the transmission accuracy of a screw pair by integrating multiple features, which comprehensively utilizes multiple feature quantities of multiple types of sensor information such as vibration, temperature, torque, displacement, etc. to estimate the transmission accuracy of the screw, thereby avoiding the problem of low credibility of the estimation result of the screw transmission accuracy due to the one-sided information of a single type of sensor.

[0064] (4) The present invention discloses a multi-feature fusion screw pair transmission accuracy prediction method and proposes a multi-working condition scalable accuracy prediction model library. The model library covers accuracy prediction models of screws of various specifications under different working conditions and has an open interface and expansion and update capabilities, thereby improving the universality of the screw accuracy prediction method. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a flow chart of a method for estimating transmission accuracy of a screw pair by integrating multiple features in an embodiment of the present invention;

[0066] Figure 2 This is a front view of a loading test bench in an embodiment of the present invention;

[0067] Figure 3 is a top view of a loading test bench in an embodiment of the present invention;

[0068] Figure 4 It is a flow chart for constructing an accuracy estimation model library in an embodiment of the present invention. DETAILED DESCRIPTION

[0069] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments disclosed in the present invention will be further described in detail below with reference to the accompanying drawings.

[0070] Reference Figure 1 In this embodiment, the multi-feature fusion screw pair transmission accuracy estimation method includes:

[0071] Step 1: According to the specification information and current working condition information of the screw rod to be tested, an accuracy estimation model M matching the specification information and current working condition information of the screw rod to be tested is screened from the accuracy estimation model library, and a sensitive feature set C corresponding to the accuracy estimation model M is determined.

[0072] Step 2, placing the screw rod to be tested on a loading test bench, performing a loading test, and obtaining test data A;

[0073] Step 3: After performing wavelet denoising and standardization on the test data A, 12-dimensional time domain features and 4-dimensional frequency domain features are extracted to obtain the feature set to be input.

[0074] Step 4, the sensitive features indicated by the sensitive feature set C are screened from the feature set to be input, and used as the input of the accuracy estimation model M. After the accuracy estimation model is used, the accuracy estimation result of the screw rod to be tested is output.

[0075] In this embodiment, the loading test bench can apply loads to the lead screw according to various load spectra, including sinusoidal load spectra, step load spectra, and swept frequency load spectra, and is equipped with various sensors such as vibration, temperature, torque, tension, pressure, and displacement. Figures 2-3 As shown, the loading test bench mainly includes: a motor 1, a motor support seat 2, a coupling A3, a torque sensor 4, a coupling B5, a transmission shaft 6, a screw tooling bearing seat 7, a coupling C8, an actuating sleeve 10, a two-force rod 12, a tension sensor 13, a loading slide 14, a slider 15, a linear guide 16, a linear grating ruler, a machine base 18, a spherical bearing 19, a hydraulic cylinder 20, a hydraulic cylinder bracket 21, a control console 22, a vibration sensor A23, a temperature sensor A24, a temperature sensor B25, a vibration sensor C26, a vibration sensor D27 and a vibration sensor B28; wherein, the screw to be tested includes: a screw 9 and screw nuts 11 located at both ends of the screw 9.

[0076] The connection relationship between the various structural components of the loading test bench is as follows:

[0077] The motor 1 is installed on the machine base 18 through the motor support seat 2. The output end of the motor 1 is connected to one end of the torque sensor 4 through the coupling A3; the other end of the torque sensor 4 is connected to the transmission shaft 6 through the coupling B5; the torque sensor 4 is used to measure the output torque of the motor. The output end of the transmission shaft 6 is connected to the lead screw 9 through the coupling C8 inside the lead screw tooling bearing seat 7; the lead screw tooling bearing seat 7 is installed on the machine base 18. The output end of the lead screw tooling bearing seat 7 is connected to the actuating sleeve 10 through a bolt, the lead screw 9 passes through the actuating sleeve 10, and the actuating sleeve 10 cooperates with the upper end surface limiting short shaft of the lead screw nut 11 through a bearing; the lead screw nut 11 is connected to one end of the two-force rod 12 through a thread, the other end of the two-force rod 12 is connected to one end of the tension sensor 13, and the other end of the tension sensor 13 is connected to one side of the support plate of the loading slide 14; the tension sensor 13 is used to measure the output tension pressure of the lead screw. The loading slide 14 is mounted on the linear guide 16 through the slider 15, and the linear guide 16 is mounted on the machine base 18. The linear grating ruler is used as a displacement sensor to measure the displacement of the lead screw; the linear grating ruler includes: a linear grating ruler reading head 171 and a linear grating ruler main scale 172; the linear grating ruler reading head 171 is installed on one side of the support plate of the loading slide 14, and the linear grating ruler main scale 172 is mounted on the machine base 18 parallel to the linear guide 16. The other side of the loading slide 14 is connected to the end of the push rod of the hydraulic cylinder 20 through a joint bearing 19; the hydraulic cylinder 20 is mounted on the machine base 18 through a hydraulic cylinder bracket 21. The control console 22 is set on the side of the machine base 18 and connected to the motor 1 through a wire. The torque sensor 4, the tension sensor 13, the linear grating ruler, the vibration sensor A23, the temperature sensor A24, the temperature sensor B25, the vibration sensor C26, the vibration sensor D27 and the vibration sensor B28 are respectively connected to the control console 22 through wires. The temperature sensor A24 is installed on the upper end face of the motor 1 to measure the temperature of the upper end face of the motor. The temperature sensor B25 and the vibration sensor C26 are installed on the upper end face of the screw nut 11 to measure the temperature of the upper end face of the screw nut and the vibration data of the upper end face of the screw nut, respectively. The vibration sensor D27 is installed on the side end face of the screw nut 11 to measure the vibration data of the side end face of the screw nut. The vibration sensor B28 is installed on the upper end face of the actuating sleeve 10 to measure the vibration data of the upper end face of the actuating sleeve. The vibration sensor A23 is installed on the side end face of the actuating sleeve 10 to measure the vibration data of the side end face of the actuating sleeve.

[0078] In this embodiment, the multi-feature fusion screw pair transmission accuracy estimation method may further include: constructing an accuracy estimation model library. Figure 4 As shown in the figure, the construction process of the accuracy estimation model library is as follows:

[0079] S11, select a sample screw for model training and set the test conditions.

[0080] S12, placing the sample lead screw on the loading test bench, and performing a loading test based on the set test conditions, with the motor driving the sample lead screw to perform reciprocating motion on the loading test bench.

[0081] S13, after the sample screw has reciprocated several times, the sample screw is disassembled and assembled to the precision measuring table, and the accuracy of the sample screw is measured by the precision measuring table; if the accuracy level of the sample screw reaches G5, execute step S14; if the accuracy level of the sample screw does not reach G5, repeat steps S12 to S13 until the accuracy level of the sample screw reaches G5.

[0082] S14, collecting and obtaining the loading test data of the whole test cycle of the loading test. The loading test data of the whole test cycle include: the vibration data of the side end face of the actuating sleeve, the vibration data of the upper end face of the actuating sleeve, the vibration data of the upper end face of the lead screw nut, the vibration data of the side end face of the lead screw nut, the upper end face temperature of the motor, the upper end face temperature of the lead screw nut, the output torque of the motor, the output tension and pressure of the lead screw, and the displacement of the lead screw.

[0083] S15, input the set test conditions, specification information of the sample screw, loading test data of the whole test cycle and precision measurement data into the model library extension interface, perform model training, and obtain the precision estimation model corresponding to the sample screw.

[0084] In this embodiment, the specific implementation of step S15 is as follows:

[0085] S151, after wavelet denoising and standardization processing of the loading test data of the entire test cycle, the 12-dimensional time domain features and 4-dimensional frequency domain features of each data are constructed to obtain a feature data set.

[0086] Preferably, as shown in Table 1 below, the 12-dimensional time domain features mainly include: minimum value, maximum value, rectified mean, root amplitude, root mean square value, variance, waveform factor, peak factor, pulse factor, margin factor, kurtosis and skewness; the 4-dimensional frequency domain features mainly include: mean frequency, center of gravity frequency, root mean square frequency and frequency standard deviation.

[0087]

[0088] Table 1, time domain and frequency domain feature example table

[0089] S152, performing trend and correlation analysis on each feature in the feature data set, selecting the screw precision degradation sensitive features based on the maximum trend and minimum correlation principle, and constructing a sensitive feature set.

[0090] Preferably, the sensitive feature set is constructed as follows:

[0091] According to the principle of including most sensor data, the trend threshold is set to q.

[0092] Calculate the trend of each feature in the feature data set:

[0093]

[0094] Among them, Tre(F k ) represents the trend of the kth feature in the feature data set, Fk represents the kth feature, F k，t represents the value of the kth feature in the tth test cycle, t = 1, 2, ..., T, T represents the total number of test cycles.

[0095] Select features with a trend greater than q from the feature data set and put them into the set Q.

[0096] Calculate the correlation between the features in set Q:

[0097]

[0098] Among them, MIC(Y:X) represents the correlation between feature Y and feature X in the feature data set, H(Y) represents the information entropy of feature Y, and H(Y|X) represents the conditional entropy of feature Y relative to feature X; p(x i ) indicates that the feature X takes the value x i The probability of x i represents the value of feature X in the i-th test cycle, i = 1, 2, ..., T; p(y j ) indicates that the feature Y takes the value y j The probability of y j represents the value of feature Y in the jth test cycle, j = 1, 2, ..., T; p(x i ,y j ) indicates that the feature X takes the value x i 、The feature Y value is y j probability.

[0099] Features with correlation less than 1 are selected from the set Q as sensitive features of screw precision degradation, and are put into the set W, which is the sensitive feature set.

[0100] S153, based on the heuristic knowledge that the accuracy degradation trend is strongly monotonic and the degradation threshold fluctuation is weak, a sensitive feature fusion optimization model is constructed, that is, the accuracy estimation model corresponding to the sample lead screw.

[0101] Preferably, the accuracy estimation model is expressed as follows:

[0102] S acc =ω1F1+ω2F2+…+ω s F s

[0103] Among them, Sacc Represents the accuracy estimation result, ω1, ω2, ..., ω s Represent features F1, F2, ..., F s The corresponding weights, F1, F2, ..., F s Represents each sensitive feature of the sensitive feature set.

[0104] Furthermore, the mathematical description of the model optimization process is as follows:

[0105]

[0106] stw′M′1=1,Mw≥0;E τ w τ ≥0

[0107] ε τ，υ ≥0, τ=1,...,m, υ=1,...,T τ -1

[0108] Among them, r represents the regularization coefficient, ε τ，υ represents the slack variable, c τ，υ represents the slack variable ε τ，υ The weight coefficient, T τ represents the maximum observation period of individual τ, m represents the total number of sample screws, Y represents the matrix of the last observation before individual failure, D represents the symmetric matrix, w τ Represents the vector composed of multi-feature fusion weight coefficients and slack variables, E τ represents the matrix that measures the monotonicity of accuracy degradation; w represents the multi-feature fusion weight coefficient vector, w=[ω1ω2…ω s ]; M represents a diagonal matrix. If the corresponding sensor data shows an increasing / decreasing trend, the corresponding diagonal position is 1 / -1; υ=1, ..., T τ -1.

[0109] S16, repeating steps S11 to S15, obtaining a number of accuracy estimation models under different specifications and different working conditions, and then constructing an accuracy estimation model library.

[0110] Although the present invention has been disclosed as above in the form of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications to the technical solution of the present invention by using the methods and technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope of the technical solution of the present invention.

[0111] The contents not described in detail in the specification of the present invention belong to the common knowledge of the professionals in this field.

Claims

1. A multi-feature fusion ball screw pair transmission accuracy estimation method, characterized in that: include: According to the specification information and current working condition information of the screw rod to be tested, an accuracy estimation model M matching the specification information and current working condition information of the screw rod to be tested is screened from the accuracy estimation model library, and a sensitive feature set C corresponding to the accuracy estimation model M is determined; Place the screw rod to be tested on the loading test bench, perform a loading test, and obtain test data A; After wavelet denoising and standardization processing of the test data A, 12-dimensional time domain features and 4-dimensional frequency domain features are extracted to obtain the feature set to be input; The sensitive features indicated by the sensitive feature set C are screened from the feature set to be input and used as the input of the accuracy estimation model M. After the accuracy estimation model is used, the accuracy estimation result of the screw rod to be tested is output.

2. The multi-feature fusion screw pair transmission accuracy estimation method according to claim 1 is characterized in that: A loading test bench comprises: a motor (1), a motor support seat (2), a coupling A (3), a torque sensor (4), a coupling B (5), a transmission shaft (6), a screw tool bearing seat (7), a coupling C (8), an actuating sleeve (10), a two-force rod (12), a tension sensor (13), a loading slide (14), a slider (15), a linear guide rail (16), a linear grating ruler, a machine base (18), a spherical bearing (19), a hydraulic cylinder (20), a hydraulic cylinder bracket (21), a control console (22), a vibration sensor A (23), a temperature sensor A (24), a temperature sensor B (25), a vibration sensor C (26), a vibration sensor D (27) and a vibration sensor B (28); wherein the screw to be tested comprises: a screw rod (9) and screw nuts (11) located at both ends of the screw rod (9); The motor (1) is mounted on a machine base (18) via a motor support base (2); The output end of the motor (1) is connected to one end of the torque sensor (4) through a coupling A (3); the other end of the torque sensor (4) is connected to the transmission shaft (6) through a coupling B (5); wherein the torque sensor (4) is used to measure the output torque of the motor; The output end of the transmission shaft (6) is connected to the lead screw (9) through a coupling C (8) inside the lead screw tool bearing seat (7); the lead screw tool bearing seat (7) is installed on the machine base (18); The output end of the screw tool bearing seat (7) is connected to the actuating sleeve (10) through a bolt, the screw rod (9) passes through the actuating sleeve (10), and the actuating sleeve (10) cooperates with the limiting short shaft on the upper end surface of the screw nut (11) through a bearing; the screw nut (11) is connected to one end of the two-force rod (12) through a thread, the other end of the two-force rod (12) is connected to one end of the tension sensor (13), and the other end of the tension sensor (13) is connected to one side of the support plate of the loading slide (14); wherein the tension sensor (13) is used to measure the output tension pressure of the screw; The loading slide (14) is mounted on a linear guide rail (16) via a slider (15), and the linear guide rail (16) is mounted on a machine base (18); The linear grating ruler is used as a displacement sensor for measuring the displacement of the lead screw; the linear grating ruler comprises: a linear grating ruler reading head (171) and a linear grating ruler main scale (172); the linear grating ruler reading head (171) is installed on one side of the support plate of the loading slide (14), and the linear grating ruler main scale (172) is installed on the machine base (18) parallel to the linear guide rail (16); The other side of the loading slide (14) is connected to the end of the push rod of the hydraulic cylinder (20) through a joint bearing (19); the hydraulic cylinder (20) is installed on the machine base (18) through a hydraulic cylinder bracket (21); The control console (22) is arranged on the side of the machine base (18) and is connected to the motor (1) via a wire; The torque sensor (4), the tension sensor (13), the linear grating ruler, the vibration sensor A (23), the temperature sensor A (24), the temperature sensor B (25), the vibration sensor C (26), the vibration sensor D (27) and the vibration sensor B (28) are respectively connected to the control console (22) through wires; The temperature sensor A (24) is mounted on the upper end surface of the motor (1) and is used to measure the temperature of the upper end surface of the motor; A temperature sensor B (25) and a vibration sensor C (26) are mounted on the upper end surface of the screw nut (11) and are used to measure the temperature of the upper end surface of the screw nut and the vibration data of the upper end surface of the screw nut respectively; A vibration sensor D (27) is mounted on the side end surface of the screw nut (11) and is used to measure vibration data of the side end surface of the screw nut; A vibration sensor B (28) is mounted on the upper end surface of the actuating sleeve (10) and is used to measure vibration data of the upper end surface of the actuating sleeve; A vibration sensor A (23) is mounted on the side end surface of the actuating sleeve (10) and is used to measure vibration data of the side end surface of the actuating sleeve.

3. The multi-feature fusion screw pair transmission accuracy estimation method according to claim 2 is characterized in that: Also includes: A precision estimation model library is constructed.

4. The multi-feature fusion screw pair transmission accuracy estimation method according to claim 3 is characterized in that: The construction process of the accuracy estimation model library is as follows: S11, selecting a sample screw for model training and setting a test condition; S12, placing the sample lead screw on a loading test bench, and performing a loading test based on the set test conditions, wherein the motor (3) drives the sample lead screw to perform reciprocating motion on the loading test bench; S13, after the sample lead screw has reciprocated several times, the sample lead screw is disassembled and assembled to the precision measuring table, and the precision of the sample lead screw is measured by the precision measuring table; if the precision level of the sample lead screw reaches G5, step S14 is executed; if the precision level of the sample lead screw does not reach G5, steps S12 to S13 are repeated until the precision level of the sample lead screw reaches G5; S14, collecting and obtaining the loading test data of the entire test cycle of the loading test; S15, inputting the set test conditions, specification information of the sample lead screw, loading test data of the entire test cycle, and precision measurement data into the model library extension interface, performing model training, and obtaining a precision estimation model corresponding to the sample lead screw; S16, repeating steps S11 to S15, obtaining a number of accuracy estimation models under different specifications and different working conditions, and then constructing an accuracy estimation model library.

5. The method for estimating transmission accuracy of a screw pair by integrating multiple features according to claim 4 is characterized in that: The loading test data of the whole test cycle include: the vibration data of the side end face of the actuating sleeve, the vibration data of the upper end face of the actuating sleeve, the vibration data of the upper end face of the screw nut, the vibration data of the side end face of the screw nut, the upper end face temperature of the motor, the upper end face temperature of the screw nut, the output torque of the motor, the output tension and pressure of the screw and the displacement of the screw.

6. The multi-feature fusion screw pair transmission accuracy estimation method according to claim 5 is characterized in that: Input the set test conditions, specification information of the sample screw, loading test data of the whole test cycle and accuracy measurement data into the model library extension interface for model training to obtain the accuracy estimation model corresponding to the sample screw, including: S151, after wavelet denoising and standardization processing of the loading test data of the entire test cycle, 12-dimensional time domain features and 4-dimensional frequency domain features of each data are constructed to obtain a feature data set; S152, performing trend and correlation analysis on each feature in the feature data set, selecting the sensitive features of screw precision degradation according to the maximum trend and minimum correlation principle, and constructing a sensitive feature set; S153, based on the heuristic knowledge that the accuracy degradation trend is strongly monotonic and the degradation threshold fluctuation is weak, a sensitive feature fusion optimization model is constructed, that is, the accuracy estimation model corresponding to the sample lead screw.

7. The method for estimating transmission accuracy of a screw pair by integrating multiple features according to claim 6 is characterized in that: The trend and correlation analysis of each feature in the feature data set is performed. According to the maximum trend and minimum correlation principle, the sensitive features of screw precision degradation are selected to construct a sensitive feature set, including: According to the principle of including most sensor data, the trend threshold is set to q; Calculate the trend of each feature in the feature data set: Among them, Tre(F k ) represents the trend of the kth feature in the feature data set, F k represents the kth feature, F k,t represents the value of the kth feature in the tth test cycle, t = 1, 2, ..., T, T represents the total number of test cycles; Select features with a trend greater than q from the feature data set and put them into the set Q; Calculate the correlation between the features in set Q: Among them, MIC(Y:X) represents the correlation between feature Y and feature X in the feature data set, H(Y) represents the information entropy of feature Y, and H(Y|X) represents the conditional entropy of feature Y relative to feature X; p(x i ) indicates that the feature X takes the value x i The probability of x i represents the value of feature X in the i-th test cycle, i = 1, 2, ..., T; p(y j ) indicates that the feature Y takes the value y j The probability of y j represents the value of feature Y in the jth test cycle, j = 1, 2, ..., T; p(x i ,y j ) indicates that the feature X takes the value x i 、The feature Y value is y j probability; Features with correlation less than 1 are selected from the set Q as sensitive features of screw precision degradation, and are put into the set W, which is the sensitive feature set.

8. The method for estimating transmission accuracy of a screw pair by integrating multiple features according to claim 6 is characterized in that: The accuracy estimation model is expressed as follows: S acc =ω1F1+ω2F2+…+ω s F s Among them, S acc Represents the accuracy estimation result, ω1, ω2, ..., ω s Represent features F1, F2, ..., F s The corresponding weights, F1, F2, ..., F s Represents each sensitive feature of the sensitive feature set.

9. The method for estimating transmission accuracy of a screw pair by integrating multiple features according to claim 8 is characterized in that: The mathematical description of the model optimization process is as follows: s.t.w′M′1=1,Mw≥0;E τ w τ ≥0 e τ,υ ≥0,τ=1,...,m,υ=1,...,T τ -1 Among them, r represents the regularization coefficient, ε τ,υ represents the slack variable, c τ,υ represents the slack variable ε τ,υ The weight coefficient, T τ represents the maximum observation period of individual τ, m represents the total number of sample screws, Y represents the matrix of the last observation before individual failure, D represents the symmetric matrix, w τ Represents the vector composed of multi-feature fusion weight coefficients and slack variables, E τ represents the matrix that measures the monotonicity of accuracy degradation; w represents the multi-feature fusion weight coefficient vector, w=[ω1ω2...ω s ]; M represents a diagonal matrix. If the corresponding sensor data shows an increasing / decreasing trend, the corresponding diagonal position is 1 / -1; υ=1,...,T τ -1.

10. The method for estimating transmission accuracy of a screw pair by integrating multiple features according to claim 1 or 6, characterized in that: 12-dimensional time domain features, including: minimum value, maximum value, rectified mean, root amplitude, root mean square value, variance, shape factor, peak factor, pulse factor, margin factor, kurtosis and skewness; 4-dimensional frequency domain features, including: mean frequency, center of gravity frequency, root mean square frequency and frequency standard deviation.

Images