Roundness error separation signal noise reduction method based on differential spectrum judgment singular value decomposition

By using the method of singular value decomposition based on differential spectrum, the problem of noise coupling in the roundness error separation process is solved, achieving effective noise reduction and accuracy improvement of the signal, which is applicable to the field of roundness measurement.

CN118896582APending Publication Date: 2024-11-05HARBIN INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411140154.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-11-05

AI Technical Summary

Technical Problem

During the roundness error separation process, random noise is coupled with the spindle rotation error and workpiece roundness error signals, resulting in reduced measurement accuracy. Existing technologies are unable to effectively separate and reduce the impact of noise.

Method used

A singular value decomposition method based on differential spectrum determination is adopted. The signal is filtered by singular value decomposition theory, the useful signal is retained and the noise interference singular values ​​are zeroed. The Hankel matrix and the differential spectrum determination singular value threshold are used for signal denoising.

Benefits of technology

It significantly improves the mean square error and signal-to-noise ratio of the signal, reduces the impact of random noise on roundness error separation, and enhances measurement accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118896582A_ABST
    Figure CN118896582A_ABST
Patent Text Reader

Abstract

The invention relates to a roundness error separation signal noise reduction method, in particular to a roundness error separation signal noise reduction method based on differential spectrum judgment singular value decomposition. The invention aims to solve the problem that random noise is mixed into a sensor acquisition signal in a roundness error separation process and is coupled with a to-be-separated circle contour error and a main shaft rotation error, so that the roundness error separation precision is influenced. The invention provides a roundness error separation signal noise reduction method based on differential spectrum judgment singular value decomposition, aiming at the situation that random noise is mixed into a sensor acquisition signal in a roundness error separation process and is coupled with a circle contour error to be separated and a main shaft rotation error to influence the roundness error separation precision. And the influence of random noise on the roundness error separation precision is reduced. The invention belongs to the technical field of ultra-precise geometric quantity measurement.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a roundness error separation signal noise reduction method, belonging to the technical field of ultra-precision geometric quantity measurement. Background Art

[0002] With the development of emerging science and technology such as atomic energy, aerospace, microelectronics, information and bioengineering, higher and higher requirements are placed on the measurement accuracy of the rotation error of the precision spindle system and the roundness error of the workpiece. For example, in the process of ultra-precision machining and measurement, the roundness error of the specified circular section of the quartz hemisphere, which is usually used as a circular standard for comparison and calibration, is generally between 10nm and 50nm; the roundness tolerance of any circular section of the curved mirror in the normal direction of the X-ray microscope is 5nm; the fields of aspheric lenses, free-form surfaces and microstructure optics also usually require the shape error of the workpiece to be at the nanometer level.

[0003] To ensure high-precision measurement, the error caused by the measuring instrument can only account for a small part of the allowable error of the workpiece, about (1 / 3 to 1 / 10). But in fact, many workpiece accuracy indicators now reach or exceed the accuracy level of the most accurate roundness measuring instrument currently used to measure roundness. If we want to further improve the measurement accuracy of the instrument, it is difficult to meet the growing requirements by relying solely on a large amount of investment in mechanical processing. Therefore, without further requiring the improvement of the spindle rotation accuracy, the precise measurement of the circular contour of the workpiece is a very valuable research issue.

[0004] Error separation technology is a measurement technology that separates useful signal components from error signals by transforming information sources or estimating model parameters. It can achieve a level of accuracy that is difficult to achieve with "hard technology" at a relatively low cost, so error separation technology has been quickly accepted by scholars and experts from various countries. However, the actual signal is not only the circular contour error signal and the spindle rotation error signal, but also the random errors such as environmental noise in the measurement process will be coupled with it, resulting in the sensor acquisition signal not being "pure" before error separation. Since the roundness error separation requires high accuracy, usually at the nanometer level, the random noise coupling and contamination of the sensor acquisition signal greatly affects the error separation effect, resulting in a reduction in error separation accuracy. In view of the above problems, a roundness error separation signal denoising method based on singular value decomposition of differential spectrum judgment is proposed to reduce the influence of random noise on the roundness error separation accuracy. Summary of the invention

[0005] In order to solve the problem that random noise is mixed into the sensor acquisition signal in the roundness error separation process, and is coupled with the circular contour error and the spindle rotation error to be separated, thus affecting the roundness error separation accuracy, the present invention further proposes a roundness error separation signal denoising method based on differential spectrum judgment singular value decomposition.

[0006] The technical solution adopted by the present invention to solve the above problems is: when the method described in the present invention adopts the singular value decomposition theory to perform signal filtering, if the first q singular value sequences represent useful signals in the signal, and the last pq smaller singular values ​​represent noise interference, then the last pq singular values ​​are set to 0.

[0007] Furthermore, in the singular value decomposition theory, any matrix A∈R m×n Written as a diagonal matrix A∈R m×n With two orthogonal matrices U∈R m×n V∈R m×n The product of A∈UΛV T , according to the diagonal matrix theory, Λ is expressed as:

[0008]

[0009] Λ=diag(λ1,λ2,…,λ q ), where λ1,≥λ2≥…≥λ q >0, q represents the rank of the matrix Λ, λ i Represents the singular values ​​of matrix A. In this case, matrix A is:

[0010]

[0011] According to formula (2), the matrix A is expressed as the product of each singular value and its basis matrix vector, and the singular value represents the energy distribution of the matrix; where i represents the counting sequence, and i = 1, 2, 3, ... q, u i Represented as an element in the orthogonal matrix U, v i Represents the elements in the orthogonal matrix V, and T represents the matrix symbol.

[0012] Furthermore, the matrix A is a Hankle matrix. The product of the m and n parameters of the Hankel matrix is ​​ensured to be the maximum. The number of rows n is determined to satisfy the following relationship. The number of columns m=N+1-n can be determined based on the number of rows n.

[0013]

[0014] Assume that the sensor collects the signal S = [y1, y2, ... y N ], according to Hankle theory, the matrix A can be obtained as:

[0015]

[0016] In formula (4), according to the difference definition: δ(i) = λ i -λ i+1,i=1,2,…,r-1, and the difference spectrum B=(δ(1),δ(2),…,δ(r-1)); Since there is a large difference between the main axis characteristic signal and the noise interference in the singular value sequence, the peak value of the difference spectrum is taken as the effective rank order of the singular value, and the singular value corresponding to the order is used as the threshold. The singular values ​​greater than the threshold are retained and the rest are set to zero. The processed singular value sequence can be inversely transformed to obtain the denoised sensor acquisition signal.

[0017] The beneficial effects of the present invention are as follows: in view of the situation that random noise is mixed into the sensor acquisition signal in the roundness error separation process, and is coupled with the circular profile error and the spindle rotation error to be separated, thus affecting the roundness error separation accuracy, the present invention proposes a roundness error separation signal denoising method based on singular value decomposition of differential spectrum judgment, thereby reducing the influence of random noise on the roundness error separation accuracy;

[0018] The present invention is a signal denoising method applied in the field of roundness measurement. The effect is as shown in the figure. After denoising, the given ideal signal can be kept consistent, verifying the effectiveness of denoising. In the following text, quantitative indicators of mean square error and signal-to-noise ratio before and after signal denoising are added, which can well reflect the quality improvement of the signal after denoising. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the original sensor collecting signals;

[0020] Figure 2 It is a schematic diagram of signal acquisition by a noisy sensor;

[0021] Figure 3 It is a schematic diagram comparing the signal after noise reduction and the signal containing noise;

[0022] Figure 4 It is a schematic diagram of signal denoising comparison. DETAILED DESCRIPTION

[0023] Specific implementation method 1: Figures 1 to 4 As shown in the figure, a roundness error separation signal denoising method based on singular value decomposition of differential spectrum judgment is proposed. When this method uses singular value decomposition theory to filter the signal, if the first q singular value sequences represent useful signals in the signal, and the last pq smaller singular values ​​represent noise interference, the last pq singular values ​​are set to 0.

[0024] Specific implementation method 2: Figures 1 to 4 As shown, in the singular value decomposition theory, any matrix A∈R m×n Written as a diagonal matrix A∈R m×n With two orthogonal matrices U∈R m×n Vv m×n The product of A∈UΛV T, according to the diagonal matrix theory, Λ is expressed as:

[0025]

[0026] Λ=diag(λ1,λ2,…,λ q ), where λ1,≥λ2≥…≥λ q >0, q represents the rank of the matrix Λ, λ i Represents the singular values ​​of matrix A. In this case, matrix A is:

[0027]

[0028] According to formula (2), the matrix A is expressed as the product of each singular value and its basis matrix vector, and the singular value represents the energy distribution of the matrix; where i represents the counting sequence, and i = 1, 2, 3, ... q, u i Represented as an element in the orthogonal matrix U, v i Represents the elements in the orthogonal matrix V, and T represents the matrix symbol.

[0029] Specific implementation method three: Figures 1 to 4 As shown, matrix A uses a Hankle matrix. The product of the m and n parameters of the Hankel matrix is ​​guaranteed to be the largest. The number of rows n is determined to satisfy the following relationship. According to the number of rows n, the number of columns m=N+1-n can be determined;

[0030]

[0031] Assume that the sensor collects the signal S = [y1, y2, ... y N ], according to Hankle theory, the matrix A can be obtained as:

[0032]

[0033] In formula (4), according to the difference definition: δ(i) = λ i -λ i +1,i=1,2,…,r-1, and the difference spectrum B=(δ(1),δ(2),…,δ(r-1)); Since there is a large difference between the main axis characteristic signal and the noise interference in the singular value sequence, the peak value of the difference spectrum is taken as the effective rank order of the singular value, and the singular value corresponding to the order is used as the threshold. The singular values ​​greater than the threshold are retained and the rest are set to zero. The processed singular value sequence can be inversely transformed to obtain the denoised sensor acquisition signal.

[0034] Effect verification of the present invention

[0035] The given sensor acquisition signal is: 2.8*cos(3.7*t+0.1*pi)+2.1*cos(10*t+0.3*pi)+1.3*cos(20*t+0.4*pi)+1.4*cos(24*t+0.17*pi). A noise signal is added to the given signal to simulate and verify the denoising ability of this method. The ideal signal and the noisy sensor acquisition signal are plotted, as shown in the figure. Figure 1 and Figure 2 As shown;

[0036] The signal obtained by using the roundness error separation signal denoising method based on differential spectrum determination singular value decomposition is compared with the noisy signal. Figure 3 As shown;

[0037] The mean square error (MSE) of the signal before and after denoising is reduced from 11.23 to 6.81, and the signal-to-noise ratio (SNR) is increased from 11.43 to 13.61, which shows that the signal performance has been greatly improved;

[0038] It can be seen that this method greatly reduces the pollution of random noise to the original signal, and to a certain extent reduces the coupling between random noise and circular contour error and spindle rotation error signals, providing a basis for subsequent accurate separation; for the case of applying the two-step method for error separation, this method is effective for the comparison of the signals before and after the noise reduction of the sensor acquisition signals before and after the indexing. Figure 4 shown.

[0039] The above is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent replacement and improvement made to the above embodiments without departing from the content of the technical solution of the present invention, based on the technical essence of the present invention, within the spirit and principles of the present invention, still fall within the protection scope of the technical solution of the present invention.

Claims

1. A roundness error separation signal denoising method based on differential spectrum determination singular value decomposition, characterized in that: When this method uses the singular value decomposition theory to filter signals, if the first q singular value sequences represent useful signals in the signal, and the last pq smaller singular values ​​represent noise interference, the last pq singular values ​​are set to 0.

2. The method for roundness error separation signal denoising based on differential spectrum determination singular value decomposition according to claim 1, characterized in that: In the singular value decomposition theory, any matrix A∈R m×n Written as a diagonal matrix A∈R m×n With two orthogonal matrices U∈R m×n V∈R m×n The product of A∈UΛV T , according to the diagonal matrix theory, Λ is expressed as: Λ=diag(λ1,λ2,...,λ q ), where λ1, ≥λ2 ≥…≥λ q > 0, q represents the rank of the matrix Λ, λ i Represents the singular values ​​of matrix A. In this case, matrix A is: According to formula (2), the matrix A is expressed as the product of each singular value and its basis matrix vector, and the singular value represents the energy distribution of the matrix; where i represents the counting number, and i = 1, 2, 3, .. q, u i Represented as an element in the orthogonal matrix U, v i Represents the elements in the orthogonal matrix V, and T represents the matrix symbol.

3. The method for roundness error separation signal denoising based on differential spectrum determination singular value decomposition according to claim 2, characterized in that: The matrix A is selected from Hankl e Matrix, the product of the m and n parameters of the Hankel matrix is ​​ensured to be maximum, and the number of rows n is determined to satisfy the following relationship. According to the number of rows n, the number of columns m = N + 1-n can be determined; Assume that the sensor collects the signal S = [y1, y2, ...y N ], according to Hankle theory, the matrix A can be obtained as: In formula (4), according to the difference definition: δ(i) = λ i -λ i +1, i=1, 2, ..., r-1, and the differential spectrum B=(δ(1), δ(2), ..., δ(r-1)); Since there is a large difference between the main axis characteristic signal and the noise interference in the singular value sequence, the peak value of the differential spectrum is taken as the effective rank order of the singular value, and the singular value corresponding to the order is used as the threshold. The singular values ​​greater than the threshold are retained and the rest are set to zero. The processed singular value sequence can be inversely transformed to obtain the denoised sensor acquisition signal.