Three-point error separation method capable of resisting rotating speed fluctuation

Abstract

A three-point error separation method resistant to rotating speed fluctuation comprises the steps that firstly, three displacement sensors and a key phase sensor are arranged on the same section of a rotor, phase alignment is conducted on signals of the three displacement sensors through synchronously-collected signals of the key phase sensor, n circles of signals are intercepted respectively, and then spectrum data are spread in a frequency domain; the method comprises the following steps: aligning the lengths of displacement signals of different circles, obtaining time domain displacement signals with consistent lengths through inverse Fourier transform, carrying out weighted fusion on multiple circles of signals, and searching a weight optimal solution by using a simulated annealing algorithm, so that the total mean square error of the fused signals and the multiple circles of signals before fusion is minimum, and the fusion accuracy is improved. The influence of random noise errors on a separation result is reduced to the maximum extent, and finally three-point method error separation is carried out on the fused displacement signals of the three different angles to obtain a roundness error signal and a rotation error signal; according to the invention, while the random error is eliminated, the feature information during multi-circle signal fusion is reserved, and the error separation precision of the three-point method is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of precision measurement of rotating devices, and in particular relates to a three-point error separation method that is resistant to rotational speed fluctuations. Background Art

[0002] Rotating devices such as electric spindles and motors are the core power parts of many equipment. In order to accurately measure the rotation error of the rotating device, it is necessary to separate the roundness error of the rotor from the signal measured by the displacement sensor. The three-point error separation technology has the advantage of online error separation and is the most widely used error separation method in industrial fields. This separation method is to arrange three displacement sensors distributed at a certain angle on a cross section of the rotating part of the rotating device. The data collected synchronously by these three displacement sensors are used to separate the roundness error of the rotating device in this cross section through the error separation equation, thereby obtaining a relatively pure rotation error signal and improving the accuracy of subsequent data processing.

[0003] Traditional three-point error separation technology, such as the patent application entitled "Ultra-precision spindle measurement method based on hybrid three-point error separation technology" (publication number CN117705446 A), is based on the spatial arrangement of multiple sensors and separates the roundness error of the measured section and the sensor installation error through geometric relationships. Since the separation equation utilizes the periodic law of the roundness error and rotation error signals, only single-turn signals are intercepted for error separation during actual processing, that is, the signal collected during one rotation of the rotating device; in order to accurately obtain single-turn signals, there are generally two methods: 1) Single-turn signal marking method, such as the patent application entitled "A spindle rotation error measurement method and device with detachable installation eccentricity" (publication CN103983227B), using a key phase sensor or encoder to mark the single-turn displacement signal, synchronously collect the signals of the displacement sensor and the key phase sensor or encoder, and intercept the single-turn displacement signal through the marked position. 2) Multi-turn signal fusion method: Based on the single-turn signal marking method, in order to minimize the impact of random noise errors on the separation results, the marked multi-turn signals are averaged and the averaged single-turn signals are used for error separation.

[0004] However, both of the above methods assume that the speed is constant and do not consider the tiny speed fluctuations in actual working conditions, resulting in the following problems during the separation process: the single-loop signal marking method ignores the dynamic correlation between multi-loop signals, which will cause spectral leakage and feature distortion in the separation results when the speed fluctuates; and the multi-loop signal fusion method will cause signal inaccuracy. Speed ​​fluctuations lead to inconsistent phases and period lengths of signals in different loops. The existing method generally selects the shortest signal length as the benchmark, directly discards the excess signal lengths of other loops, and finally performs average fusion. This method will destroy the phase consistency of the signal, thereby introducing false geometric coupling errors. Summary of the Invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a three-point error separation method that is resistant to speed fluctuations. While eliminating random errors, it further retains the characteristic information when multi-turn signals are fused, thereby improving the accuracy of three-point error separation.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A three-point error separation method for resisting speed fluctuations includes the following steps:

[0008] Step 1) Acquiring signals and aligning phases;

[0009] Step 2) aligning the lengths of displacement sensor signals of different circles in the frequency domain;

[0010] Step 3) Perform weighted fusion on the multi-circle signals and use the simulated annealing algorithm to search for the optimal weight solution;

[0011] Step 4) Repeat steps 2) and 3) to intercept n-turn signals collected by the three displacement sensors. By aligning the signal lengths and finding the optimal weights, three sets of fused displacement signals are obtained. Three-point error separation is performed on them to obtain roundness error signals and rotation error signals.

[0012] The step 1) is specifically as follows:

[0013] 1.1) Arrange three displacement sensors at a certain angle on the same cross-section of the rotating part of the rotating device. At the same time, affix a marking sticker to the rotating part and place a key phase sensor at the corresponding position of the marking sticker. Determine and calibrate each sensor.

[0014] 2.2) Stabilize the rotating device to a certain speed and use a high sampling frequency to synchronously collect the signals of the three displacement sensors and the key phase sensor;

[0015] 2.3) According to the key phase sensor signal, the three displacement sensor signals are phase-aligned and n-turn signals are intercepted respectively.

[0016] The step 2) is specifically as follows:

[0017] 2.1) Let the n-turn signal of one displacement sensor be X1, X2, X3…X n , the length of each signal is l1,l2,l3...l n , the maximum value is recorded as l max ;

[0018] 2.2) For the k-th circle signal X k , if l k <l max, after Fourier transform, we get the frequency domain signal F k ={f1,f2,…f k …f lk}, continue with subsequent processing; if l k ≥l max , no operation is performed;

[0019] 2.3) Using the symmetry of the frequency domain components of the Fourier transformed signal, a portion of zero signal is inserted into the high frequency part of the signal. The specific operation is: define an all-zero array The length is l max , the frequency domain signal F k The first half of the signal is assigned to the array one by one The second half of the signal from The high frequency part in the middle of the frequency domain signal is still 0, that is,

[0020]

[0021] 2.4) The interpolated frequency domain signal Inverse Fourier transform to obtain its time domain signal Repeat steps 2.2) and 2.3) above to obtain the interpolated n-circle signal, which is recorded as The signal length is l max .

[0022] The step 3) is specifically as follows:

[0023] 3.1) Assume that the signal weights of each circle are W1, W2, W3…W n , and W1+W2+W3…+W n =1, the fused signal is Then there is The total mean square error between the fused signal and the multi-circle signal before fusion is

[0024] 3.2) Taking the total mean square error as the optimization target, the optimal weight is calculated by the simulated annealing algorithm to make the best signal after fusion The total mean square error between the multi-turn signal and the original signal is the smallest.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] 1. This invention uses spectrum expansion and inverse transformation technology to unify signals of different periods into the same length, providing a data basis for subsequent multi-turn signal fusion, avoiding errors introduced by speed fluctuations, and solving the failure problem of traditional time domain alignment methods under variable speed conditions.

[0027] 2. The present invention aims to minimize the total mean square error between the fusion signal and each noisy signal, automatically balances the weights of each circle signal, and is particularly suitable for non-stationary noise environments.

[0028] 3. Compared with optimization methods such as gradient descent, the simulated annealing algorithm adopted in the present invention can effectively avoid falling into local optimality, converge stably in nonlinear, multi-peak weight search space, and significantly reduce the signal noise level after weighted fusion.

[0029] 4. The present invention reduces the impact of speed fluctuations and random errors on algorithm accuracy through data length alignment and weighted fusion, and effectively improves the accuracy of the rotating part roundness error and rotation error signal separated by the three-point method. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Flowchart of an embodiment of the present invention.

[0031] Figure 2 This is a schematic diagram of sensor installation according to an embodiment of the present invention.

[0032] Figure 3 Schematic diagram of signal alignment between the displacement sensor and the key phase sensor according to an embodiment of the present invention.

[0033] Figure 4 Schematic diagram comparing the roundness errors of signals separated before and after fusion according to an embodiment of the present invention.

[0034] Figure 5 Schematic diagram comparing the rotation errors obtained by separating the signals before and after fusion according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The present invention is described in detail below with reference to the embodiments and accompanying drawings.

[0036] Reference Figure 1 , a three-point error separation method for resisting speed fluctuation, comprising the following steps:

[0037] Step 1) Acquire the signal and align the phase. The specific measures are as follows:

[0038] 1.1) If Figure 2 As shown, three displacement sensors S1, S2, and S3 are arranged at a certain angle on the same cross section of the rotating part of the rotating device. At the same time, a marking sticker is affixed to the rotating part, and a key phase sensor P is arranged at the corresponding position of the marking sticker. Each sensor is calibrated and calibrated.

[0039] 2.2) Stabilize the rotating device to a certain speed and use a higher sampling frequency to synchronously collect the signals of the three displacement sensors and the key phase sensor;

[0040] 2.3) According to the key phase sensor signal, such as Figure 3 As shown, the three displacement sensor signals are phase-aligned and n-turn signals are intercepted respectively;

[0041] Step 2) Align the lengths of displacement signals of different circles in the frequency domain. The core idea is to utilize the symmetry and periodicity of the discrete Fourier transform, expand the spectrum data in the frequency domain, and then generate time-domain displacement signals of consistent length through the inverse Fourier transform. The specific measures are as follows:

[0042] 2.1) Let the n-turn signal of one displacement sensor be X1, X2, X3…X n , the length of each signal is l1,l2,l3...l n , the maximum value is recorded as l max ;

[0043] 2.2) For the k-th circle signal X k , if l k <l max , after Fourier transform, we get the frequency domain signal F k ={f1,f2,…f k …f lk}, continue with subsequent processing; if l k ≥l max , no operation is performed;

[0044] 2.3) Using the symmetry of the frequency domain components of the Fourier transformed signal, a portion of zero signal is inserted into the high frequency part of the signal, so that the signal length can be extended without changing the original signal spectrum. The specific operation is: define an all-zero array The length is l max , the frequency domain signal F k The first half of the signal is assigned to the array one by one The second half of the signal from Assign a value to the tail of the signal, so that the high frequency part in the middle of the frequency domain signal is still 0, that is,

[0045] 2.4) The interpolated frequency domain signal Inverse Fourier transform to obtain its time domain signal Repeat the above steps 2.2) and 2.3) to get the interpolated n-circle signal, which is recorded as The signal length is l max ;

[0046] Step 3) Considering the correlation between multi-loop signals and minimizing the impact of random noise errors on the separation results, multi-loop signals are weighted and fused, and the simulated annealing algorithm is used to search for the optimal weight solution. The specific measures are as follows:

[0047] 3.1) Assume that the signal weights of each circle are W1, W2, W3…W n , and W1+W2+W3…+W n =1, the fused signal is Then there is The total mean square error between the fused signal and the multi-circle signal before fusion is

[0048] 3.2) Taking the total mean square error as the optimization target, the optimal weight is calculated by the simulated annealing algorithm to make the best signal after fusion The total mean square error between the multi-turn signal is the smallest;

[0049] Step 4) Repeat steps 2) and 3) to intercept n-turn signals collected by the three displacement sensors. By aligning the signal lengths and finding the optimal weights, three sets of fused displacement signals are obtained. Three-point error separation is performed on them to obtain roundness error signals and rotation error signals.

[0050] Reference Figure 4 、 Figure 5 , Figure 4 and Figure 5 The figures are respectively a comparison diagram of the roundness error and the rotation error separated from the signals before and after fusion. The dotted part in the figure is the roundness error and rotation error obtained by intercepting multiple circles of signals before fusion and performing three-point error separation on each circle of signals. The solid line part is the separation result after length alignment and signal fusion. It can be seen that the separation result of the signal before fusion contains noise, and the error between the separation results is more obvious. The separation result of the signal after fusion is better.

Claims

1. A three-point error separation method for resisting speed fluctuation, characterized in that: The following steps are involved: Step 1) Acquiring signals and aligning phases; Step 2) aligning the lengths of displacement sensor signals of different circles in the frequency domain; Step 3) Perform weighted fusion on the multi-circle signals and use the simulated annealing algorithm to search for the optimal weight solution; Step 4) Repeat steps 2) and 3) to intercept n-turn signals collected by the three displacement sensors. By aligning the signal lengths and finding the optimal weights, three sets of fused displacement signals are obtained. Three-point error separation is performed on them to obtain roundness error signals and rotation error signals.

2. The method according to claim 1, characterized in that The step 1) is specifically as follows: 1.1) Arrange three displacement sensors at a certain angle on the same cross-section of the rotating part of the rotating device. At the same time, affix a marking sticker to the rotating part and place a key phase sensor at the corresponding position of the marking sticker. Determine and calibrate each sensor. 2.2) Stabilize the rotating device to a certain speed and use a high sampling frequency to synchronously collect the signals of the three displacement sensors and the key phase sensor; 2.3) According to the key phase sensor signal, the three displacement sensor signals are phase-aligned and n-turn signals are intercepted respectively.

3. The method according to claim 1, characterized in that The step 2) is specifically as follows: 2.1) Let the n-turn signal of one displacement sensor be X1, X2, X3…X n , the length of each signal is l1,l2,l3...l n , the maximum value is recorded as l max ; 2.2) For the k-th circle signal X k , if l k <l max , after Fourier transform, we get the frequency domain signal F k ={f1,f2,…f k …f lk }, continue with subsequent processing; if l k ≥l max , no operation is performed; 2.3) Using the symmetry of the frequency domain components of the Fourier transformed signal, a portion of zero signal is inserted into the high frequency part of the signal. The specific operation is: define an all-zero array The length is l max , the frequency domain signal F k The first half of the signal is assigned to the array one by one The second half of the signal from The high frequency part in the middle of the frequency domain signal is still 0, that is, 2.4) The interpolated frequency domain signal Inverse Fourier transform to obtain its time domain signal Repeat steps 2.2) and 2.3) above to obtain the interpolated n-circle signal, which is recorded as The signal length is l max .

4. The method according to claim 3, characterized in that The step 3) is specifically as follows: 3.1) Assume that the signal weights of each circle are W1, W2, W3…W n , and W1+W2+W3…+W n =1, the fused signal is Then there is The total mean square error between the fused signal and the multi-circle signal before fusion is 3.2) Taking the total mean square error as the optimization target, the optimal weight is calculated by the simulated annealing algorithm to make the best signal after fusion The total mean square error between the multi-turn signal and the original signal is the smallest.

Images