Cylindrical gear ghost order detection method and detection equipment based on tooth surface waviness analysis
Through the tooth surface waviness analysis method of adaptive sampling interval and iterative sine wave fitting algorithm, accurate detection and prediction of gear ghost steps are achieved, which solves the shortcomings of high-precision waviness analysis in existing technologies. It is suitable for various gear measurement centers and eliminates ghost steps through micro-vibration devices, thereby improving gear processing accuracy.
Patent Information
- Application Number
- CN202510880756.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-27
- Publication Date
- 2025-09-26
AI Technical Summary
Existing technologies lack high-precision and high-efficiency waviness analysis and gear ghost step detection methods, making it difficult for gear manufacturers to break away from their dependence on foreign equipment. They are unable to deeply explore the source and cause of waviness, and the waviness prediction results cannot match those of single meshing instruments or production line terminal test equipment.
A method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis is provided. The tooth surface deviation data is acquired through adaptive sampling intervals. Combined with tilt error correction and deshaping, Gaussian filtering and an iterative sine wave fitting algorithm are used to extract the waviness signal and generate an order spectrum. This method realizes accurate detection of ghost steps and suppresses ghost steps during the gear machining process through a micro-vibration device.
It achieves accurate detection and prediction of gear ghost steps, with the detection results differing by 0.01μm from those of foreign commercial software. It is applicable to all gear measurement centers, can accurately predict the whine noise during meshing transmission, and provides a gear processing method to eliminate ghost steps, thereby improving gear processing accuracy.
Smart Images

Figure CN120702753A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of gear processing detection, and specifically relates to a method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis, and a corresponding computer program product and gear quality detection equipment. Background Art
[0002] In recent years, fields such as electric vehicles and humanoid robots have shown a booming development trend. In the process of technological innovation in these fields, the transmission system has been continuously optimized and upgraded, and the number of transmission levels has been greatly reduced. This change has made the role of gears in the entire power transmission process more critical, and their noise performance at high speeds has received more attention. Therefore, the gear design and manufacturing industry is actively adjusting strategies and improving technologies to meet the challenge of low noise. Among the many factors that affect gear noise, waviness analysis is particularly urgent for solving the problem of ghost-step noise. By conducting in-depth waviness analysis, the microscopic geometric shape characteristics of the gear surface can be accurately grasped, the root cause of the ghost-step noise can be found, and the design and manufacturing process of the gear can be optimized in a targeted manner to meet the strict requirements of low-noise performance in related fields.
[0003] Currently, commercial waviness analysis software developed by Klingelnberg in Germany is widely used in this field. However, the software algorithm is highly protected and requires the use of equipment at the Klingelnberg Gear Testing Center. The high cost of the equipment and licensing fees significantly limits its market value.
[0004] The market still lacks high-precision and efficient waviness analysis and its-based gear ghost step detection methods, and the waviness prediction results do not match the ghost steps measured by single-meshing instruments or end-of-line test equipment. This has led to gear manufacturers currently struggling to break free from their dependence on foreign equipment and to delve into the sources and causes of waviness. Therefore, developing a self-controlled waviness analysis algorithm that can accurately predict gear ghost steps and apply it to general gear measurement centers has become a key task and urgent need to promote industrial upgrading. Summary of the Invention
[0005] In order to solve the problem that the existing technology lacks high-precision and high-efficiency waviness analysis and gear ghost step detection methods, the present invention provides a cylindrical gear ghost step detection method for tooth surface waviness analysis, and its corresponding computer program product and gear quality detection equipment.
[0006] The technical solution provided by the present invention is:
[0007] A method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis, comprising:
[0008] The tooth surface deviation data for each tooth on the target gear is acquired sequentially at adaptive sampling intervals, along the left side of the tooth profile, the right side of the tooth profile, the left side of the tooth direction, and the right side of the tooth direction. After resampling, tilt error correction, and deshaping, the deviation curves for each tooth are flattened based on the tooth sequence and combined into a full tooth deviation curve in four directions at a preset overlap ratio. Waviness data is extracted from each of the four full tooth deviation curves using Gaussian filtering and combined into the corresponding full tooth waviness curve based on the overlap ratio.
[0009] Each full-tooth waviness curve is subjected to RC filtering to obtain a waviness signal. The waviness signal is converted into a frequency domain signal after windowed fast Fourier transform. A frequency range is selected based on the frequency domain signal. Then, the waviness signal is spectrally analyzed within the specified frequency range using an iterative sine wave fitting algorithm to extract all sinusoidal components. The frequencies of the extracted sinusoidal components are converted into orders to generate the corresponding order spectrum.
[0010] When there is a sinusoidal wave component with an amplitude greater than or equal to 0.02μm and a non-integer multiple of the number of teeth in the order spectrum in any direction of the target gear, it means that there is a ghost order in the gear, and the amplitude, frequency and phase of the waviness signal corresponding to the ghost order are output for fault tracing.
[0011] As a further improvement of the present invention, the adaptive sampling interval T of the gear measurement center in the tooth profile direction and the tooth direction is cp and T cl The calculation formula is as follows:
[0012]
[0013] In the above formula, L AE Indicates the effective length of the tooth profile; n old Indicates the number of measured waves under standard amplitude attenuation conditions; n new represents the measured wave number of the new amplitude attenuation applied to adapt to the changes in wave number and tooth width; b represents the tooth width length.
[0014] As a further improvement of the present invention, the resampling of the tooth surface deviation data is used to unify the number of measurement points in the tooth surface deviation data in four directions of each tooth.
[0015] As a further improvement of the present invention, the method of tilt error correction and de-shaping operation is:
[0016] First, the least squares method is used to perform a linear fit on each data point of the collected tooth surface deviation data to generate an error curve. Then, the tilt error correction is completed by subtracting the corresponding error amount in the error curve from the original measurement point value. Next, the values of each measurement point after tilt error correction are fitted to a quadratic parabola to generate the corresponding drum curve. Finally, the de-sculpting operation is completed by subtracting the corresponding point value in the drum curve from the value of each measurement point after tilt error correction.
[0017] As a further improvement of the present invention, the combination method of the full tooth deviation curve in any direction is as follows:
[0018] According to the effective measured length in the tooth profile direction and tooth direction, the overlap rate of adjacent teeth in the corresponding direction is calculated:
[0019]
[0020] In the above formula, L r,z is the rolling path length of a single tooth; P bt is the end face base circle pitch; e p Indicates the tooth profile overlap rate; L x Indicates the effective measuring length of the tooth direction, P x Indicates the axial pitch between consecutive teeth in the axial direction; e l Indicates the tooth overlap ratio.
[0021] Flatten the deviation curves of all teeth to a straight line direction, and combine them into a full tooth deviation line in an overlapping form according to the preset overlap rate; where the position coordinates of the i-th tooth in the tooth profile direction and tooth direction coordinate axis are and The calculation formula is:
[0022]
[0023] In the above formula, z i Indicates the tooth sequence of each tooth in the gear.
[0024] As a further improvement of the present invention, the short-wave cutoff wavelength λc of the Gaussian filter for extracting the waviness data is 1 mm, and the long-wave cutoff wavelength λf is 10 mm.
[0025] As a further improvement of the present invention, the expression of the windowed fast Fourier transform operation of the waviness signal is as follows:
[0026]
[0027] In the above formula, y(n) represents the ripple signal after RC filtering; w(n) represents the window signal of the Hamming window; z(n) represents the windowed ripple signal; Z(k) represents the frequency domain signal after z(n) is transformed by FFT; n represents the discrete time point, n = 0, 1, 2, ..., N-1; N is the signal length; Is an imaginary unit, k represents a discrete frequency point in the frequency domain, and the corresponding frequency is f k :f k =k / NT, T is the sampling period.
[0028] As a further improvement of the present invention, the process of performing spectrum analysis on the waviness signal y(n) by an iterative sine wave fitting algorithm is as follows:
[0029] ⅰ、Calculate the mean μ and standard deviation σ of |Z(k)| and generate the frequency range [f min , f max ], take the ripple signal y(n) as the initial signal y0(n), and let the initial frequency f0 = f min And set the iteration termination condition to f = f max .
[0030] ii. Based on selecting the initial frequency f from the frequency range i , using the sine wave model to calculate y i (n) is fitted to obtain the corresponding sinusoidal wave signal s i (n):
[0031] s i (n) = p i sin(2πf i nT)+q i cos(2πf i nT)
[0032] In the above formula, A i represents the amplitude of the sine wave, T is the sampling period, is the phase of the sine wave.
[0033] iii. Define the error function E(p,q) under any frequency f as:
[0034]
[0035] The error function is calculated using the gradient descent iterative method, and the values of p and q are initialized to: p = 0, q = 0, and the learning rate α (for example, α = 0.01) and the maximum number of iterations M are set.
[0036] Then calculate the partial derivatives of the error function E(p,q) with respect to p and q
[0037]
[0038] In the above formula, x1(n)=sin(2πfnT); x2(n)=cos(2πfnT).
[0039] Update the parameters p and q according to the following formula:
[0040]
[0041] In the above formula, q i and p i and are the sine wave coefficients before updating, q i+1 and p i+1 are the updated sine wave coefficients.
[0042] Repeat the above process until the maximum number of iterations M is reached or the following convergence conditions are met:
[0043] |E(p i+1 ,q i+1 )-E(p i ,q i )|<∈;
[0044] Among them, E(p i+1 ,q i+1 ) represents the updated error function value, E(p i ,q i )| represents the value of the error function before updating; ∈ is a very small positive number;
[0045] Based on the updated values of p and q, the amplitude A and phase φ of the corresponding sine wave signal are calculated using the following formula:
[0046]
[0047] ⅳ、The initial signal y of this round i (n) and the sinusoidal wave signal s fitted in this round i The difference between (n) is used as the next round of initial signal y i+1 (n):
[0048] y i+1 (n) = y i (n)-s i (n)
[0049] ⅴ, from [f min , f max ] to select the next frequency point f i+1 , repeat steps ⅱ to ⅳ to iteratively analyze the waviness signal until the iteration termination condition is reached.
[0050] ⅵ. Convert the frequencies of the various sinusoidal signals obtained through iterative analysis into corresponding orders, and generate corresponding order spectra.
[0051] As a further improvement of the present invention, in step i, the frequency range [f min , f max ] is as follows:
[0052] Define an adaptive threshold τ:
[0053] τ=μ+βσ,
[0054] Among them, β is an adjustable parameter with a value range of 1 to 3;
[0055] By comparing the size of |Z(k)| and τ, all frequency points that satisfy |Z(k)|>τ are determined, and the frequency range corresponding to these frequency points is recorded as [f min , f max ].
[0056] As a further improvement of the present invention, in step ⅵ, the order O t The conversion formula between the sine wave frequency f is as follows:
[0057] O t =f / f r =60·f / n c
[0058] In the above formula, f r is the reference axis frequency; n c is the reference shaft speed.
[0059] The present invention also includes a computer program product, which includes a computer program. When the computer program is executed by a processor, it completes the steps of the cylindrical gear ghost step detection method based on tooth surface waviness analysis as described above, and then detects and locates the ghost steps of the cylindrical gear based on the collected tooth surface deviation data of each tooth in the target gear in four directions.
[0060] The present invention also includes a gear quality inspection device comprising a measuring instrument and a data processing device. The measuring instrument is configured to sequentially measure tooth surface deviation data for each tooth of a target gear in four directions: left side of the tooth profile, right side of the tooth profile, left side of the tooth direction, and right side of the tooth direction. The data processing device includes a computer program product as described above, which detects and locates ghost steps on cylindrical gears based on the collected tooth surface deviation data for each tooth of the target gear in four directions.
[0061] The present invention also includes a gear processing method for eliminating gear ghost steps, which includes the following steps:
[0062] During the gear machining process, a micro-vibration generating device is added to the workpiece or tool axis position.
[0063] By adopting the method of theoretical analysis followed by experimental verification, several correction process parameter mappings that can achieve ghost order suppression are obtained, forming the required sparse process set.
[0064] The data format of each correction process parameter mapping is as follows:
[0065]
[0066] In the above formula, A1, f1 and They represent the amplitude, frequency and phase of the waviness signal corresponding to the gear ghost order, which constitute the error condition of gear machining; A2, f2 and They respectively represent the vibration amplitude, vibration frequency and vibration phase of the micro-vibration generating device when the corresponding gear ghost step suppression can be achieved. The three constitute the correction parameters for dimensional processing.
[0067] A neural network model is trained using a sparse process set. The neural network model is used to train the neural network model according to the input A1, f1, Output corresponding A2, f2,
[0068] The trained neural network model is then used to generate more correction process parameter mappings to form a dense process set.
[0069] After completing the processing of any gear sample, the aforementioned cylindrical gear ghost step detection method based on tooth surface waviness analysis is used to predict the ghost steps of the processed gear, and the corresponding error conditions are obtained when ghost steps are detected.
[0070] According to the known error working conditions of the gear sample, the dense process set is queried to obtain a correction process parameter mapping with the highest working condition similarity S.
[0071] During the batch processing of the gears corresponding to the gear samples, the micro-vibration generating device is started according to the correction parameters in the queried correction process parameter mapping, thereby achieving gear ghost step suppression.
[0072] As a further improvement of the present invention, the micro-vibration device adopts a piezoelectric ceramic driver or an electromagnetic vibrator.
[0073] As a further improvement of the present invention, the calculation formula of the operating condition similarity S is as follows:
[0074]
[0075] In the above formula, A0, f0 and They represent the amplitude, frequency and phase of the error condition of the gear sample with ghost steps respectively; A i , f i and They represent the amplitude, frequency and phase of the error condition of the i-th correction process parameter mapping in the dense process set respectively; α, β and γ represent the influence weights corresponding to the amplitude, frequency and phase in the condition similarity respectively.
[0076] The technical solution provided by the present invention has the following beneficial effects:
[0077] The method for detecting ghost orders of cylindrical gears for tooth surface waviness analysis provided by the present invention can more accurately set the cutoff wavelength in the filter and extract more abundant effective information according to the amplitude attenuation. The iterative sine wave fitting method adopted by this solution can accurately calculate the order and its amplitude information during the waviness order analysis process, and can be used without the gear measurement center. The displayed order amplitude is only 0.01μm different from the foreign Klingelnberg commercial waviness analysis software. After comparing the single meshing instrument noise test results with the waviness analysis results of the present invention, it was found that the solution of the present invention can indeed accurately predict the ghost orders that cause the howling noise during gear meshing transmission.
[0078] The technical solution provided by the present invention can be applied to all gear measurement centers. The waviness measurement and analysis methods adopted by the present invention are universal and can be applied to the analysis and quality inspection of various cylindrical gears. This lays the foundation for the integration of domestically produced waviness analysis software into gear measurement centers.
[0079] In addition, using the analysis results obtained in the ghost step prediction of the present invention, this embodiment also designs a gear processing method for eliminating gear ghost steps. This method actively applies adaptive micro-vibration on the gear processing tool to correct the processing of various gears with ghost steps, so as to eliminate the ghost step problem in the gear batch processing process and improve the gear processing accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0080] Figure 1 This is a flowchart of the steps of the cylindrical gear ghost step detection method based on tooth surface waviness analysis provided in Example 1 of the present invention.
[0081] Figure 2 This is a practical example of combining the full tooth deviation curve in Example 1 of the present invention.
[0082] Figure 3 This is a flow chart of the steps of performing spectrum analysis on a waviness signal by using an iterative sine wave fitting algorithm in Example 1 of the present invention.
[0083] Figure 4 It is the full tooth deviation curve in four directions of the tooth surface of the sample gear in the test experiment.
[0084] Figure 5 This is the waviness order spectrum in four directions of the tooth surface of the sample gear in the test experiment. DETAILED DESCRIPTION
[0085] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0086] Example 1
[0087] This embodiment provides a method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis. This method can utilize the tooth profile direction of the left tooth surface (hereinafter referred to as the tooth profile left side), the tooth direction direction of the left tooth surface (hereinafter referred to as the tooth direction left side), the tooth profile direction of the right tooth surface (hereinafter referred to as the tooth profile right side), and the tooth direction direction of the right tooth surface (hereinafter referred to as the tooth direction right side) of each tooth in the gear measured by a general gear measurement center, and combine it with the newly proposed data analysis direction to realize feature extraction and spectrum analysis of the waviness in each direction of the tooth surface, thereby accurately predicting the possible ghost steps in the gear, and locating the amplitude, frequency, phase and other information of the waviness signal that causes the ghost steps, so as to trace the error of the dimensional processing process later.
[0088] Specifically, if Figure 1 As shown, the method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis provided in this embodiment includes the following steps:
[0089] S1: Acquire tooth surface deviation data of each tooth in the target gear measured sequentially by the gear measurement center according to the adaptive sampling interval in four directions: left side of tooth profile, right side of tooth profile, left side of tooth direction and right side of tooth direction.
[0090] In this embodiment, a gear measurement center measures tooth surface deviation data on both the left and right flanks of each tooth in a gear. During the measurement process on both sides of the tooth surface, the measurement center collects data at equal intervals along a preset trajectory in both the tooth profile and tooth direction. Profile measurement involves sampling at evenly spaced points within the end section plane at the midpoint of the tooth width, while tooth direction measurement involves sampling at evenly spaced points along the pitch diameter along the tooth width.
[0091] To ensure that the collected data can effectively extract more waviness information from the gear tooth surface, it is necessary to properly set the sampling intervals in the tooth direction and tooth profile direction. Unlike traditional solutions that mostly use fixed sampling intervals, this embodiment uses an adaptive sampling interval during the data collection phase. This allows for better extraction of waviness information even when a large number of waves are present within the evaluation range or the gear tooth width b becomes very large.
[0092] Specifically, this embodiment firstly determines the effective length L of the tooth profile. AE The cutoff wavelength λ in the tooth profile direction and tooth direction is calculated by the following formula: cp and λ cl :
[0093] λ cp =L AE / 30,
[0094] λ cl =b / 30.
[0095] If there are a large number of waves within the evaluation range, or if the tooth width b becomes very large, the cutoff wavelength set above will no longer be universally applicable, and the acquired waviness data will contain errors. Therefore, during the measurement process, this embodiment adjusts the wave amplitude attenuation individually by setting an adaptive cutoff wavelength in the filter, as shown in the following formula:
[0096]
[0097] In the above formula, n old Indicates the number of measured waves under standard amplitude attenuation conditions; n new Indicates the measured wave number to which the new amplitude attenuation should be applied in order to adapt to the changes in wave number and tooth width.
[0098] In order to capture the data of the calculated minimum wavelength, considering the Nyquist frequency criterion and preventing the sampling point from missing the peak of the signal surface, the sampling interval must be set to at least one-sixth of the minimum wavelength considered. That is:
[0099]
[0100] In summary, the adaptive sampling interval T of the gear measurement center in the tooth profile direction and tooth direction of this embodiment is cp and T cl The calculation formula is as follows:
[0101]
[0102] S2: After resampling, tilt error correction and deshaping of each tooth surface deviation data of each tooth, the deviation curves of each tooth are flattened according to the tooth sequence and then combined into full tooth deviation curves in four directions according to a preset overlap rate.
[0103] This embodiment uniformly resamples the tooth surface deviation data collected in different directions on both sides of each tooth, corrects for tilt errors, and performs de-shaping. Resampling the tooth surface deviation data is used to unify the number of measurement points in the tooth surface deviation data in four directions for each tooth. Tilt error correction and de-shaping are used to remove tilt errors and crowning from the collected data, which may interfere with ghost step detection accuracy.
[0104] The tooth surface deviation data can be resampled in the following ways: If the number of data points in the tooth direction and the tooth profile direction of any tooth surface is inconsistent, the sample data with fewer data points can be upsampled to the same size as the sample data with more data points, or the sample data with more data points can be downsampled to the same size as the sample data with fewer data points. In practical applications, the latter method is often used, for example, by randomly deleting redundant data points to unify the data size of the tooth surface deviation data in different directions, facilitating subsequent waviness and ghost order analysis.
[0105] In this embodiment, the tilt error correction and de-shaping operation method is:
[0106] First, the least squares method is used to perform linear fitting on each data point of the collected tooth surface deviation data to obtain an error curve. For example, the data point of the collected tooth surface deviation data is recorded as (x, f(x)). For each collected data point, the expression of the error curve after fitting in this embodiment is as follows:
[0107] y(x)=m0x+n0,
[0108] Where m0 represents the slope of the error curve; n0 represents the intercept of the error curve.
[0109] Then, the tilt error correction is completed by subtracting the corresponding error amount in the error curve from the original measurement point value. For example, for a data point (x, f(x)), the result after tilt error correction is recorded as (x, f(x) - y(x)).
[0110] Next, the values of each measurement point after the tilt error correction are fitted into a quadratic parabola to obtain the corresponding drum-shaped curve. The expression of the fitted drum-shaped curve in this embodiment is as follows:
[0111] g(x)=a0+a1x+a2x 2 ,
[0112] In the above formula, a0, a1 and a2 are the coefficients of the constant term, linear term and quadratic term in the drum curve respectively.
[0113] Finally, the de-sculpting operation is completed by subtracting the value of the corresponding point in the drum curve from the value of each measured point after the tilt error correction. For the data point (x, f(x) - y(x)) after the tilt error correction, the data point of the tooth surface deviation data corrected by the de-sculpting operation is recorded as (x, f(x) - y(x) - g(x)).
[0114] After completing the tilt error correction and de-shaping operations, this embodiment further connects the tooth surface deviation data in the four directions of each tooth surface in sequence according to the tooth sequence on the gear, and then combines them into four continuous full tooth deviation curves. Specifically, Figure 2 As shown, the combination method of the full tooth deviation curve in any direction is as follows:
[0115] According to the effective measured length in the tooth profile direction and tooth direction, the overlap rate of adjacent teeth in the corresponding direction is calculated:
[0116]
[0117] In the above formula, L r,z is the rolling path length of a single tooth; P bt is the end face base circle pitch; e p Indicates the tooth profile overlap rate; L x Indicates the effective measuring length of the tooth direction, P x Indicates the axial pitch between consecutive teeth in the axial direction; e l represents the tooth overlap ratio, where for spur gears, e l =0.
[0118] Flatten the deviation curves of all teeth to a straight line direction, and combine them into a full tooth deviation line in an overlapping form according to the preset overlap rate; where the position coordinates of the i-th tooth in the tooth profile direction and tooth direction coordinate axis are and The calculation formula is:
[0119]
[0120] In the above formula, z i Indicates the tooth sequence of each tooth in the gear.
[0121] S3: The waviness data from the four full-tooth deviation curves are extracted respectively through Gaussian filtering, and the waviness data are combined into the corresponding full-tooth waviness curve according to the overlap ratio.
[0122] In this embodiment, the waviness component can be separated from the deviation data of the actual surface profile by a Gaussian filter, combined with a precisely set cutoff wavelength. Specifically, the Gaussian filter used to extract the waviness data in this embodiment has a short-wavelength cutoff wavelength λc = 1 mm and a long-wavelength cutoff wavelength λf = 10 mm.
[0123] After completing the data extraction, this embodiment further combines the waviness data in the four directions of the tooth surface according to the aforementioned overlap rate in the same manner as the deviation data above, thereby obtaining four continuous full-tooth waviness curves corresponding to different directions of the tooth surface.
[0124] S4: Perform RC filtering on each full-tooth waviness curve to obtain a waviness signal.
[0125] In this embodiment, there may still be some noise in the combined full-tooth waviness curve. This embodiment uses an RC filter to perform denoising preprocessing on each full-tooth waviness curve to obtain a noise-reduced waviness signal, which is recorded as y(n) in this embodiment.
[0126] S5: Perform a windowed fast Fourier transform on the corrugation signal and convert it into a frequency domain signal. Select a frequency range based on the frequency domain signal. Then, perform spectrum analysis on the corrugation signal within the specified frequency range using an iterative sine wave fitting algorithm to extract all sinusoidal components. Convert the frequencies of the extracted sinusoidal components into orders to generate the corresponding order spectrum.
[0127] In the technical solution provided in this embodiment, the expression of the windowed fast Fourier transform operation of the waviness signal is as follows:
[0128]
[0129] In the above formula, y(n) represents the ripple signal after RC filtering; w(n) represents the window signal of the Hamming window; z(n) represents the windowed ripple signal; Z(k) represents the frequency domain signal after z(n) is transformed by FFT; n represents the discrete time point, n = 0, 1, 2, ..., N-1; N is the signal length; Is an imaginary unit, k represents a discrete frequency point in the frequency domain, and the corresponding frequency is f k :f k =k / NT, T is the sampling period.
[0130] The process of performing spectrum analysis on the waviness signal y(n) by using an iterative sine wave fitting algorithm in this embodiment is as follows:
[0131] ⅰ、Calculate the mean μ and standard deviation σ of |Z(k)| and generate the frequency range [f min , f max ], take the ripple signal y(n) as the initial signal y0(n), and let the initial frequency f0 = f min And set the iteration termination condition to f = f max .
[0132] In the technical solution provided by this embodiment, the accuracy of the spectrum analysis will be determined by taking into account the calculated frequency range. In order to determine the possible frequency range of the sine wave, an adaptive threshold method is further adopted to finally determine the frequency range corresponding to these frequency points based on the mean μ and standard deviation σ of the frequency domain signal |Z(k)|; specifically, the frequency range [f min , f max ] is generated as follows:
[0133] Define an adaptive threshold τ:
[0134] τ=μ+βσ,
[0135] Among them, β is an adjustable parameter with a value range of 1 to 3;
[0136] By comparing the size of |Z(k)| and τ, all frequency points that satisfy |Z(k)|>τ are determined, and the frequency range corresponding to these frequency points is recorded as [f min , f max ].
[0137] ii. Based on selecting the initial frequency f from the frequency range i , using the sine wave model to calculate y i (n) is fitted to obtain the corresponding sinusoidal wave signal s i (n):
[0138] In this embodiment, the sine wave model used is:
[0139]
[0140] Where A is the amplitude of the sine wave, f is the frequency of the sine wave, and T is the sampling period. is the phase of the sine wave.
[0141] According to the following trigonometric function and angle formula:
[0142] sin(α0+β0)=sinα0cosβ0+cosα0sinβ0.
[0143] In this embodiment, let α0=2πfnT,
[0144] Then we have:
[0145] s(n)=psin(2πfnT)+qcos(2πfnT),
[0146] Where p and q are the sine parameters and cosine coefficients of the sine wave model.
[0147] In summary, the ripple signal y i(n) Sine wave signal s obtained after fitting i (n) as follows:
[0148] s i (n) = p i sin(2πf i nT)+q i cos(2πf i nT)
[0149] In the above formula, A i represents the amplitude of the sine wave, T is the sampling period, is the phase of the sine wave.
[0150] iii. Define the error function E(p,q) under any frequency f as:
[0151]
[0152] The error function is calculated using the gradient descent iterative method, and the values of p and q are initialized to: p = 0, q = 0, and the learning rate α (for example, α = 0.01) and the maximum number of iterations M are set.
[0153] Then calculate the partial derivatives of the error function E(p,q) with respect to p and q
[0154]
[0155] In the above formula, x1(n)=sin(2πfnT); x2(n)=cos(2πfnT);.
[0156] Update the parameters p and q according to the following formula:
[0157]
[0158] In the above formula, q i and p i and are the sine wave coefficients before updating, q i+1 and p i+1 are the updated sine wave coefficients.
[0159] Repeat the above process until the maximum number of iterations M is reached or the following convergence conditions are met:
[0160] |E(p i+1 ,q i+1 )-E(p i ,q i )|<∈;
[0161] Among them, E(p i+1 ,q i+1 ) represents the updated error function value, E(pi ,q i )| represents the value of the error function before updating; ∈ is a very small positive number;
[0162] Based on the updated values of p and q, the corresponding sine wave signal s is calculated by the following formula i The amplitude A and phase of (n)
[0163]
[0164] ⅳ、The initial signal y of this round i (n) and the sinusoidal wave signal s fitted in this round i The difference between (n) is used as the next round of initial signal y i+1 (n):
[0165] y i+1 (n) = y i (n)-s i (n)
[0166] ⅴ, from [f min , f max ] to select the next frequency point f i+1 , repeat steps ⅱ to ⅳ to iteratively analyze the waviness signal until the iteration termination condition is reached.
[0167] ⅵ. Convert the frequencies of the various sinusoidal signals obtained through iterative analysis into corresponding orders, and generate corresponding order spectra.
[0168] Order O t The conversion formula between the sine wave frequency f is as follows:
[0169] O t =f / f r =60·f / n c
[0170] In the above formula, f r is the reference axis frequency; n c is the reference shaft speed.
[0171] Specifically, in this embodiment, the frequency range [f min , f max When selecting each frequency point and iteratively extracting the sinusoidal wave components, the sinusoidal wave frequency corresponding to the first calculated maximum amplitude is considered the first order and is plotted into the order spectrum to output the eigenvalue. Subsequently, through iterative calculation, the sinusoidal waves corresponding to all frequency points can be obtained. In practical applications, it is also possible to extract only the sinusoidal wave components corresponding to some frequency points where the minimum sinusoidal wave amplitude is greater than or equal to 0.02μm. For example, only the orders of the first 10 maximum amplitudes can be recorded.
[0172] S6: Make the following judgments and decisions based on the generated order spectra in the four directions:
[0173] When there is a sinusoidal wave component with an amplitude greater than or equal to 0.02μm and a non-integer multiple of the number of teeth in the order spectrum in any direction of the target gear, it means that there is a ghost order in the gear, and the amplitude, frequency and phase of the waviness signal corresponding to the ghost order are output for fault tracing.
[0174] Example 2
[0175] In practical applications, the method for detecting ghost steps on cylindrical gears based on tooth surface waviness analysis provided in Example 1 can be promoted and applied in the form of a computer program. Therefore, based on the solution in Example 1, this embodiment further provides a computer program product, which includes a computer program. When executed by a processor, the computer program performs the steps of the method for detecting ghost steps on cylindrical gears based on tooth surface waviness analysis in Example 1, thereby detecting and locating ghost steps on cylindrical gears based on the collected tooth surface deviation data of each tooth in four directions of the target gear.
[0176] This embodiment also provides a storage medium storing a computer program. When the computer program is executed by a processor, the steps of the cylindrical gear ghost step detection method based on tooth surface waviness analysis as in Example 1 are completed, thereby detecting and locating the ghost steps of the cylindrical gear based on the collected tooth surface deviation data of each tooth in the target gear in four directions.
[0177] Furthermore, this embodiment provides a gear quality inspection device comprising a measuring instrument and a data processing device. The measuring instrument is configured to sequentially measure tooth surface deviation data for each tooth of a target gear in four directions: left side of the tooth profile, right side of the tooth profile, left side of the tooth direction, and right side of the tooth direction. The data processing device includes a computer program product as described above, thereby enabling detection and location of ghost steps on cylindrical gears based on the collected tooth surface deviation data for each tooth of the target gear in four directions.
[0178] The gear quality inspection equipment provided in this embodiment essentially builds upon existing gear measurement centers by adding ghost step analysis-related programs to the back-end data processing equipment, thereby enabling existing equipment to have ghost step prediction capabilities. This resolves the existing problem of requiring specialized gear measurement center equipment with integrated software and hardware provided by certain equipment vendors for gear ghost step analysis.
[0179] Example 3
[0180] Based on Example 1, this embodiment further provides a gear processing method for eliminating gear ghost steps, which includes the following process:
[0181] 1. Improvement of processing equipment
[0182] During gear machining, a microvibration generator is added to the workpiece or tool axis. This embodiment effectively suppresses the generation of ghost steps in gear tooth surface waviness by precisely controlling the amplitude, frequency, and phase of the microvibration, thereby improving gear machining quality. In practical applications, the microvibration device utilizes a piezoelectric ceramic driver or electromagnetic vibrator.
[0183] 2. Construction of database for correction process parameter mapping
[0184] Firstly, a theoretical analysis followed by experimental verification is adopted to obtain several correction process parameter mappings that can achieve ghost order suppression, forming the required sparse process set.
[0185] The data format of each correction process parameter mapping is as follows:
[0186]
[0187] In the above formula, A1, f1 and They represent the amplitude, frequency and phase of the waviness signal corresponding to the gear ghost order, which constitute the error condition of gear machining; A2, f2 and They respectively represent the vibration amplitude, vibration frequency and vibration phase of the micro-vibration generating device when the corresponding gear ghost step suppression can be achieved. The three constitute the correction parameters for dimensional processing.
[0188] In this embodiment, each correction process parameter mapping can be constructed first through theoretical analysis and then experimentally verified. If the experimental verification process proves that ghost order elimination can indeed be achieved, the data is retained. Otherwise, the correction parameter portion is adjusted until a feasible correction parameter is obtained under the current error condition. It should be noted that in actual applications, the feasible domain of the correction parameters for each error condition is actually distributed in a three-dimensional space composed of amplitude, frequency, and phase. In this embodiment, it is not necessary to accurately solve the spatial range of this feasible domain; instead, only a set of typical feasible solutions is required.
[0189] The process of obtaining a preliminary process parameter mapping through theoretical analysis includes: integrating the principles of elastic mechanics and gear cutting dynamics, constructing a mechanical model of the tool-workpiece system under micro-vibration, and analyzing the relationship between micro-vibration parameters and tooth surface ripple characteristics (frequency, amplitude, and phase). Finite element software is also used to model the gear machining system. Combined with modal testing, the system's natural frequency and various modes are analyzed to understand the gear machining system's response characteristics under different frequency excitations, providing a reference for selecting micro-vibration frequencies and preventing resonance.
[0190] Then, a neural network model is trained using the sparse process set. The neural network model is used to train the neural network model according to the input A1, f1, Output corresponding A2, f2, The neural network of this embodiment can adopt a BP neural network, an MLP model, or other similar network models that can establish a mapping between input data and output data. During the network model training process, the error conditions in the mapping of each correction process parameter are used as data, and the correction parameters are used as corresponding label information.
[0191] Finally, the trained neural network model is used to amplify the sparse process set, generating more correction process parameter mappings to form a dense process set. During the data amplification process, new error conditions are generated from the amplitude, frequency, and phase dimensions at a specified step size. These new error conditions are then input into the trained neural network model, which predicts the corresponding correction parameters and forms a new correction process parameter mapping.
[0192] 3. Adjustment of processing parameters
[0193] After completing the processing of any gear sample, the cylindrical gear ghost step detection method based on tooth surface waviness analysis as in Example 1 is used to predict the ghost steps of the processed gear, and the corresponding error condition is obtained when the ghost steps are detected.
[0194] Based on the known error working conditions of the gear sample, the dense process set is searched to obtain a correction process parameter mapping with the highest working condition similarity S. In this embodiment, the working condition similarity S is calculated as follows:
[0195]
[0196] In the above formula, A0, f0 and They represent the amplitude, frequency and phase of the error condition of the gear sample with ghost steps respectively; A i , f i and They represent the amplitude, frequency and phase of the error condition of the i-th correction process parameter mapping in the dense process set respectively; α, β and γ represent the influence weights corresponding to the amplitude, frequency and phase in the condition similarity respectively.
[0197] During the batch processing of the gears corresponding to the gear samples, the micro-vibration generating device is started according to the correction parameters in the queried correction process parameter mapping, thereby achieving gear ghost step suppression.
[0198] It can be seen that in the solution provided by this embodiment, the enterprise can first generate a sample based on the gear processing tasks of each batch, and then detect whether the sample has ghost steps, and generate corresponding error conditions when ghost steps exist, and then use the error conditions to query the pre-built dense process set to obtain the most suitable correction process parameter mapping, and then use the correction parameters in the correction process parameter mapping to adjust the operating status of the gear processing center to actually induce an adaptive micro-vibration to the workpiece or tool shaft during the gear processing process, thereby improving the accuracy of the processed gears and avoiding the occurrence of ghost steps.
[0199] Performance Testing
[0200] In order to verify the effectiveness of the technical solution provided in this embodiment, technicians tested the solution provided by the present invention on a sample gear (helical gear); the experimental process is as follows:
[0201] 1. Ghost Level Analysis
[0202] First, this experiment uses the gear measurement center to collect the tooth surface deviation data of each tooth in the sample gear. After the deviation data is exported in text form and preprocessed, the following is obtained: Figure 4 The full tooth deviation curve is shown in Figure 5. Then, the waviness signal is extracted and the spectrum is analyzed for the full tooth deviation, and the order spectrum is obtained as shown in Figure 5.
[0203] analyze Figure 5 From the data, we can see that the order spectrum of the gear's tooth profile on the left, tooth direction on the right, and tooth direction on the left has an amplitude greater than or equal to 0.02μm at the 66th order, so the 66th order is the most obvious ghost order.
[0204] 2. Performance Verification
[0205] To verify the effectiveness of the analysis results from the present invention, technicians further used a single-mesh rolling tester to detect gear noise and abnormal order information. The ghost orders calculated using the waviness analysis method were compared with the results from the single-mesh tester. The test results indicate that the waviness analysis-based gear ghost order detection solution provided by the present invention is highly accurate and reliable, making it suitable for widespread application.
[0206] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis, characterized in that: It includes: Acquire tooth surface deviation data of each tooth in the target gear in four directions, namely, left side of the tooth profile, right side of the tooth profile, left side of the tooth direction, and right side of the tooth direction, measured sequentially by a gear measurement center according to an adaptive sampling interval; resample, correct for tilt errors, and perform deshaping on each tooth surface deviation data; flatten the deviation curves of each tooth in accordance with the tooth sequence, and then combine them into full-tooth deviation curves in four directions according to a preset overlap ratio; extract waviness data from each of the four full-tooth deviation curves using Gaussian filtering, and combine the waviness data into a corresponding full-tooth waviness curve according to the overlap ratio; Each full-tooth waviness curve is subjected to RC filtering to obtain a waviness signal. The waviness signal is converted into a frequency domain signal after a windowed fast Fourier transform. A frequency range is selected based on the frequency domain signal. Then, within the specified frequency range, the waviness signal is subjected to spectrum analysis using an iterative sine wave fitting algorithm to extract all sinusoidal components. The frequencies of each extracted sinusoidal component are then converted into orders to generate the corresponding order spectrum. When there is a sinusoidal wave component with an amplitude greater than or equal to 0.02μm and a non-integer multiple of the number of teeth in the order spectrum in any direction of the target gear, it means that there is a ghost order in the gear, and the amplitude, frequency and phase of the waviness signal corresponding to the ghost order are output for fault tracing.
2. The method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis according to claim 1, wherein: The adaptive sampling interval T of the gear measurement center in the tooth profile direction and tooth direction cp and T cl The calculation formula is as follows: In the above formula, L AE Indicates the effective length of the tooth profile; n old Indicates the number of measured waves under standard amplitude attenuation conditions; n new represents the measured wave number of the new amplitude attenuation applied to adapt to the changes in wave number and tooth width; b represents the tooth width length.
3. The method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis according to claim 1, wherein: The resampling of the tooth surface deviation data is used to unify the number of measurement points in the tooth surface deviation data in the four directions of each tooth; and / or, The method for tilt error correction and deshaping operation is: First, the least squares method is used to perform linear fitting on each data point of the collected tooth surface deviation data to obtain an error curve; then, the original value of the measurement point is subtracted from the corresponding error amount in the error curve to complete the tilt error correction; then, the value of each measurement point after the tilt error correction is fitted into a quadratic parabola to obtain the corresponding drum curve; finally, the value of each measurement point after the tilt error correction is subtracted from the value of the corresponding point in the drum curve to complete the de-shaping operation.
4. The method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis according to claim 1, wherein: The combination method of the full tooth deviation curve in any direction is as follows: According to the effective measured length in the tooth profile direction and tooth direction, the overlap rate of adjacent teeth in the corresponding direction is calculated: In the above formula, L r,z is the rolling path length of a single tooth; P bt is the end face base circle pitch; e p Indicates the tooth profile overlap rate; L x Indicates the effective measuring length of the tooth direction, P x Indicates the axial pitch between consecutive teeth in the axial direction; e l Indicates the tooth overlap ratio; Flatten the deviation curves of all teeth to a straight line direction, and combine them into a full tooth deviation line in an overlapping form according to the preset overlap rate; where the position coordinates of the i-th tooth in the tooth profile direction and tooth direction coordinate axis are and The calculation formula is: In the above formula, z i Indicates the tooth sequence of each tooth in the gear.
5. The method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis according to claim 1, wherein: The Gaussian filter for extracting waviness data has a short-wavelength cutoff wavelength λc = 1 mm and a long-wavelength cutoff wavelength λf = 10 mm.
6. The method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis according to claim 1, wherein: The expression of the windowed fast Fourier transform operation of the waviness signal is as follows: In the above formula, y(n) represents the ripple signal after RC filtering; w(n) represents the window signal of the Hamming window; z(n) represents the windowed ripple signal; Z(k) represents the frequency domain signal after z(n) is transformed by FFT; n represents the discrete time point, n = 0, 1, 2, ..., N-1; N is the signal length; Is an imaginary unit, k represents a discrete frequency point in the frequency domain, and the corresponding frequency is f k :f k =k / NT, T is the sampling period.
7. The method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis according to claim 6, wherein: The process of spectrum analysis of the waviness signal y(n) using the iterative sine wave fitting algorithm is as follows: ⅰ、Calculate the mean μ and standard deviation σ of |Z(k)| and generate the frequency range [f min , f max ], take the ripple signal y(n) as the initial signal y0(n), and let the initial frequency f0 = f min And set the iteration termination condition to f = f max ; ii. Based on selecting the initial frequency f from the frequency range i , using the sine wave model to calculate y i (n) is fitted to obtain the corresponding sinusoidal wave signal s i (n): s i (n)=p i sin(2πf i nT)+q i cos(2πf i nT) In the above formula, A i represents the amplitude of the sine wave, T is the sampling period, is the phase of the sine wave; iii. Define the error function E(p,q) under any frequency f as: The error function is calculated using the gradient descent iterative method, the values of p and q are initialized to: p = 0, q = 0, and the learning rate α and the maximum number of iterations M are set; Then calculate the partial derivatives of the error function E(p,q) with respect to p and q In the above formula, x1(n) = sin(2πfnT); x2(n) = cos(2πfnT); Update the parameters p and q according to the following formula: In the above formula, q i and p i and are the sine wave coefficients before updating, q i+1 and p i+1 is the updated sine wave coefficient; Repeat the above process until the maximum number of iterations M is reached or the following convergence conditions are met: |E(p i+1 ,q i+1 )-E(p i ,q i )|<∈; Among them, E(p i+1 ,q i+1 ) represents the updated error function value, E(p i ,q i )| represents the value of the error function before updating; ∈ is a very small positive number; Based on the updated values of p and q, the amplitude A and phase of the corresponding sine wave signal are calculated by the following formula: ⅳ、The initial signal y of this round i (n) and the sinusoidal wave signal s fitted in this round i The difference between (n) is used as the next round of initial signal y i+1 (n): y i+1 (n)=y i (n)-s i (n) ⅴ, from [f min , f max ] to select the next frequency point f i+1 , repeat steps ⅱ to ⅳ to iteratively analyze the waviness signal until the iteration termination condition is reached; ⅵ. Convert the frequencies of the various sinusoidal signals obtained through iterative analysis into corresponding orders, and generate corresponding order spectra.
8. The method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis according to claim 7, wherein: In step i, the frequency range [f min , f max ] is as follows: Define an adaptive threshold τ: τ=μ+βσ, Among them, β is an adjustable parameter with a value range of 1 to 3; By comparing the size of |Z(k)| and τ, all frequency points that satisfy |Z(k)|>τ are determined, and the frequency range corresponding to these frequency points is recorded as [f min , f max ]; and / or, In step ⅵ, order O t The conversion formula between the sine wave frequency f is as follows: O t =f / f r =60·f / n c In the above formula, f r is the reference axis frequency; n c is the reference shaft speed.
9. A gear processing method for eliminating gear ghost steps, characterized in that: It includes: During the gear machining process, a micro-vibration generating device is added to the workpiece or tool axis position; By adopting the method of theoretical analysis followed by experimental verification, several correction process parameter mappings that can achieve ghost order suppression are obtained to form the required sparse process set; the data format of the correction process parameter mapping is as follows: In the above formula, A1, f1 and They represent the amplitude, frequency and phase of the waviness signal corresponding to the gear ghost order, which constitute the error condition of gear machining; A2, f2 and They represent the vibration amplitude, vibration frequency and vibration phase of the micro-vibration generating device when the corresponding gear ghost step suppression can be achieved, and the three constitute the correction parameters of dimensional processing; The sparse process set is used to train a neural network model, which is used to train the neural network model according to the input A1, f1, Output corresponding A2, f2, Then, the trained neural network model is used to generate more correction process parameter mappings to form a dense process set. After completing the processing of any gear sample, the method for detecting ghost steps of cylindrical gears based on tooth surface waviness analysis as described in any one of claims 1 to 8 is used to predict ghost steps of the processed gear, and when ghost steps are detected, the corresponding error condition is obtained; According to the known error working conditions of the gear sample, the dense process set is searched to obtain a correction process parameter mapping with the highest working condition similarity S; During the batch processing task of the gears corresponding to the gear samples, the micro-vibration generating device is started according to the correction parameters in the queried correction process parameter mapping, thereby realizing the gear ghost step suppression.
10. The gear processing method for eliminating gear ghost steps according to claim 9, wherein: The micro-vibration device adopts a piezoelectric ceramic driver or an electromagnetic vibrator; and / or The calculation formula of the working condition similarity S is as follows: In the above formula, A0, f0 and They represent the amplitude, frequency and phase of the error condition of the gear sample with ghost steps respectively; A i , f i and They represent the amplitude, frequency and phase of the error condition of the i-th correction process parameter mapping in the dense process set respectively; α, β and γ represent the influence weights corresponding to the amplitude, frequency and phase in the condition similarity respectively.
Citation Information
Cited By
Gearbox fault shaft positioning method and system
CN121185615A
Gear tooth surface waviness Fourier analysis error interval evaluation method based on forward modeling calculation
CN121858947A
