TWI719901B - Method for judging the decline of preload of ball screw - Google Patents

Abstract

A method for judging the precompression decay of a ball screw is implemented by a computer device connected to a first sensing unit with a signal. The first sensing unit periodically transmits a vibration signal related to the vibration of the ball in the circulating accessory to The computer device includes: (A) obtaining a vibration time domain data according to the received at least one vibration signal; (B) obtaining at least one vibrator frequency corresponding to the vibration time domain data according to the vibration time domain data Domain data; (C) For each vibrator frequency domain data, obtain a vibration eigenvector related to the vibrator frequency domain data; (D) According to the vibration eigenvectors, multiple vibration reference vectors and a preload detection Range, obtain a pre-compression judgment result; (E) According to the pre-compression judgment result, judge whether the pre-compression of the ball screw has declined.

Description

Method for judging the decline of ball screw preload

The present invention relates to a detection method, in particular to a method for detecting the preload of a ball screw.

At present, the ball screw has been widely used in machine tools that require precise positioning. The ball screw is made by screwing a screw cap through a rollable ball and the screw shaft, and through the nut and the circulation accessories (return accessories) ) The rolling of the ball to make the nut move linearly along the screw axis.

However, as the ball screw is used longer, the preload of the ball screw will gradually decrease. The lack of preload will make the nut easy to vibrate back and forth during the reciprocating movement, resulting in a decrease in accuracy. The pressure decays to a certain extent, and even causes backlash in the ball screw.

Refer to the Taiwan Certificate No. TW I653410B patent, its purpose is to detect whether the ball screw has produced backlash (no preload), but it is impossible to know whether the ball screw currently has only preload recession but no backlash.

Therefore, the object of the present invention is to provide a method that can determine whether the preload of the ball screw has only declined without backlash.

Therefore, the method for judging the precompression decay of the ball screw of the present invention is implemented by a computer device, the computer device signal is connected to a first sensing unit, the first sensing unit is installed on a nut of a ball screw and adjacent to A circulating part of the ball screw periodically transmits a vibration signal related to the vibration of the ball in the circulating part to the computer device. The method for judging the preload decline of the ball screw includes the following steps.

Step (A) is to obtain vibration time-domain data corresponding to the at least one vibration signal based on the received at least one vibration signal by the computer device.

Step (B) is to obtain at least one vibrator frequency domain data corresponding to the vibration time domain data based on the vibration time domain data by the computer device.

Step (C) is to obtain a vibration feature vector related to the frequency domain data of the vibrator by the computer device according to the frequency domain data of the vibrator for each vibrator frequency domain data.

Step (D) is to use the computer device to obtain a pre-compression determination result based on the vibration feature vectors, multiple vibration reference vectors, and a pre-compression detection range.

Step (E) is to use the computer device to determine whether the preload of the ball screw has decayed based on the result of the preload determination.

The effect of the present invention is: by obtaining the at least one vibration feature vector obtained according to the at least one vibration signal, the vibration reference vectors, and the pre-compression detection range, a method for determining whether the pre-compression of the ball screw has declined is obtained As a result of the preload determination, it can be detected that the ball screw has only decayed in preload without backlash.

Before the present invention is described in detail, it should be noted that in the following description, similar elements are denoted by the same numbers.

Referring to FIG. 1, an embodiment of the method for judging the decline of the ball screw preload of the present invention is implemented by a ball screw preloading decline judging system 100. The ball screw preloading decline judging system 100 includes a computer device 1, a signal connection The first sensing unit 2 of the computer device 1 and the second sensing unit 3 of the computer device 1 are connected by a signal.

Referring to Figures 1 and 2, the first sensing unit 2 is installed on a nut 41 of a ball screw 4 and adjacent to a circulating fitting 43 of the ball screw 4 and periodically transmits a piece of information related to the circulating fitting 43 The vibration signal of the vibration of the ball 44 is sent to the computer device 1. The second sensing unit 3 is installed on the nut 41 and periodically transmits an inertial force signal related to the inertial force of the nut 41 relative to the moving direction of the screw shaft 42 of the ball screw to the computer device 1 .

It is worth noting that the user can also install various types of ball screws 4 on the first sensing unit 2 and the second sensing unit 3 according to different types of ball screws (Figure 2 only shows the outer circulation type ball screws). Each corresponding position (indicating that the first sensing unit 2 and the second sensing unit 3 may be installed at the same position according to different types of ball screws 4) to obtain vibrations related to the balls 44 in the circulating fitting 43 The vibration signal and the inertial force signal related to the inertial force of the nut 41 relative to the screw axis of the ball screw 42.

The computer device 1 includes a communication module 11 that signals the first sensing unit 2 and the second sensing unit 3, a storage module 12, a display module 13, and an electrical connection to the communication module 11 The processing module 14 with the storage module 12 and the display module 13.

In this embodiment, the implementation aspect of the computer device 1 is, for example, a personal computer, a server, or a cloud host, but it is not limited to this.

In this embodiment, the implementation of the first sensing unit 2 and the second sensing unit 3 is an accelerometer, but it is not limited to this. In other embodiments, it can also be a displacement meter. Or a speedometer. Furthermore, the effective bandwidth of the first sensing unit 2 needs to cover 0.1~5 Hz, and the effective bandwidth of the second sensing unit 3 needs to cover 10 times the bandwidth of the screw shaft 42 rotation frequency (for example: 0.1 ~250Hz), and the digital resolution is 20bit.

The storage module 12 stores multiple first training vibration feature vectors, multiple second training vibration feature vectors, multiple first training inertial force feature vectors, and multiple second training inertial force feature vectors. Wherein, each training vibration feature vector includes a training kurtosis feature vector indicating the kurtosis of the frequency domain data corresponding to the training vibration feature vector, and a training kurtosis feature vector indicating the frequency domain data corresponding to the training vibration feature vector At least one of the training maximum peak frequency domain peak feature vector of the maximum peak value and a training total energy feature vector indicating the total energy of the frequency domain data corresponding to the training vibration feature vector, but not limited to the above. Wherein, each training inertial force feature vector includes a training peak-peak feature vector indicating the peak-peak value of the time domain data corresponding to the training inertial force feature vector, and a training peak-peak feature vector indicating the location of the training inertial force feature vector. The training maximum time domain peak eigenvector corresponding to the maximum peak of the time domain data, and an indication that the sum of the absolute value of the maximum peak value and the absolute value of the minimum peak value of the time domain data corresponding to the training inertial force eigenvector is averaged Train at least one of the average peak feature vectors, but not limited to the above.

Hereinafter, the computer device 1, the first sensing unit 2, and the second sensing unit 3 of the ball screw precompression decay judging system 100 will be described by the embodiment of the method for judging the precompression decay of the ball screw of the present invention. As for the operation details of the components, the method for judging the precompression decline of the ball screw of the present invention includes a precompression training program, a backlash training program, and a precompression judgment program.

Referring to Fig. 3, the pre-compression training program includes steps 50 to 53.

In step 50, the processing module 14 uses an unsupervised algorithm according to the first training vibration feature vectors to obtain a plurality of vibration reference vectors in the corresponding data space. Furthermore, when the unsupervised algorithm includes a clustering algorithm (e.g. K-mean), the vibration reference vectors include the vectors corresponding to the centers of the multiple vibration clusters obtained by the clustering algorithm; and when the non-supervised algorithm The supervised algorithm includes the self-organizing map (SOM, Self-Organizing Map). The vibration reference vectors include all neurons whose update times obtained by the self-organizing map algorithm are greater than a preset number of times. Vector, but not limited to the above example.

In step 51, for each second training vibration feature vector, the processing module 14 calculates a first vibration candidate distance between the second training vibration feature vector and each vibration reference vector. Furthermore, the Euclidean distance is used to calculate each first vibration candidate distance, but it is not limited to this.

In step 52, for each second training vibration feature vector for which the first vibration candidate distances have been obtained, the processing module 14 selects from the first vibration candidate distances corresponding to the second training vibration feature vector, Obtain a first vibration target distance corresponding to the shortest distance.

In step 53, the processing module 14 obtains a preload detection range according to the first vibration target distances. Furthermore, in this embodiment, the first vibration target distances are regarded as the normal distribution (Normal Distribution), and the 95% confidence interval (CI, Confidence interval) corresponding to the first vibration target distances is used as the preload The detection range, but not limited to this.

Referring to FIG. 4, the procedure for obtaining the backlash detection range includes steps 60 to 63.

In step 60, the processing module 14 uses another unsupervised algorithm to obtain a plurality of inertial force reference vectors in the corresponding data space according to the first training inertial force feature vectors. Furthermore, when the other unsupervised algorithm includes a clustering algorithm, the inertial force reference vectors include the vectors corresponding to the centers of the multiple inertial force clusters obtained by the clustering algorithm; and when the other Unsupervised algorithms include self-organizing mapping algorithms. These inertial force reference vectors include vectors corresponding to all neurons whose update times obtained by the self-organizing mapping algorithm are greater than another preset number. The above examples are limited.

In step 61, for each second training inertial force feature vector, the processing module 14 calculates a first inertial force candidate distance between the second training inertial force feature vector and each inertial force reference vector. Furthermore, the Euclidean distance is used to calculate each first inertial force candidate distance, but it is not limited to this.

In step 62, for each second training inertial force feature vector for which the first inertial force candidate distances have been obtained, the processing module 14 obtains the first inertial force feature vector corresponding to the second training inertial force feature vector Among the candidate distances, a first inertial force target distance corresponding to the shortest distance is obtained.

In step 63, the processing module 14 obtains a backlash detection range according to the first inertial force target distances. Furthermore, in this embodiment, the first inertial force target distances are regarded as the normal distribution, and the 95% confidence interval corresponding to the first inertial force target distances is used as the backlash detection range, but this is not limit.

Referring to Figures 5 and 6, the pre-compression determination procedure includes steps 70 to 85.

In step 70, the processing module 14 obtains vibration time-domain data corresponding to the at least one vibration signal according to the received at least one vibration signal.

In step 71, the processing module 14 obtains at least one vibrator frequency domain data corresponding to the vibration time domain data according to the vibration time domain data.

Referring to FIG. 7, it is worth noting that step 51 further includes step 710 to step 714.

In step 710, the processing module 14 performs envelope processing (Envelope) according to the vibration time domain data to obtain the processed vibration time domain data. Among them, envelope processing is a conventional technology and is not the focus of the present invention, so it will not be elaborated.

In step 711, the processing module 14 obtains a target vibration time domain data related to the constant velocity period when the ball screw 4 is moving according to the processed vibration time domain data. Furthermore, the constant speed period during the movement of the ball screw 4 can also be known from the preset motor speed for controlling the movement of the ball screw 4.

In step 712, the processing module 14 obtains at least one vibrator time domain data according to the target vibration time domain data. Furthermore, the processing module 14 cuts the target vibration time domain data according to a preset time interval of the same length to obtain the at least one vibrator time domain data, and each vibrator time domain data is Corresponds to time intervals of the same length. Conversely, in other embodiments, the processing module 14 may also combine the target vibration time domain data according to a plurality of pieces of target vibration time domain data and another preset time interval of the same length to obtain at least One other vibrator time-domain data, and each other vibrator time-domain data corresponds to another time interval of the same length.

In step 713, for each vibrator time-domain data, the processing module 14 performs a band pass filter (Band Pass Filter) according to the vibrator time-domain data to obtain filtered time-domain data of the vibrator. Furthermore, the frequency range reserved by the filtered time domain data of the vibrator is within 10 times the bandwidth of the rotational speed of the screw shaft 42. Therefore, the frequency range that can be sensed by the first sensing unit 2 is just When it covers a bandwidth of 10 times the frequency of the rotation speed of the screw shaft 42, the band-pass filtering described in step 713 is not required, and step 714 is directly performed.

In step 714, for each filtered time domain data of the vibrator, the processing module 14 performs Fourier Transform according to the filtered time domain data of the vibrator to obtain the corresponding filtered time domain data of the vibrator The frequency domain data of the vibrator of the domain data.

In step 72, for each vibrator frequency domain data, the processing module 14 obtains a vibration feature vector related to the vibrator frequency domain data according to the vibrator frequency domain data. Wherein, each vibration feature vector includes a kurtosis feature vector indicating the kurtosis of the vibrator frequency domain data corresponding to the vibration feature vector, and a kurtosis feature vector indicating the maximum peak value of the vibrator frequency domain data corresponding to the vibration feature vector At least one of the maximum frequency domain peak eigenvector of the vibration eigenvector, and a total energy eigenvector indicating the total energy of the frequency domain data of the vibrator corresponding to the vibration eigenvector, but not limited to the above.

In step 73, for each vibration feature vector, the processing module 14 calculates a second candidate vibration distance between the vibration feature vector and each vibration reference vector.

In step 74, for each vibration feature vector for which the second vibration candidate distances have been obtained, the processing module 14 obtains a corresponding shortest distance from the second vibration candidate distances corresponding to the vibration feature vector The second vibration target distance.

In step 75, the processing module 14 obtains the pre-compression detection range based on the second vibration target distances and the pre-compression detection range used to determine whether the vibration characteristic vectors indicate that the pre-compression of the ball screw 4 has declined. judgement result. For example, the processing module 14 averages the distances of the second vibration targets to obtain a vibration average, and determines whether the vibration average is within the preload detection range as the preload determination result Or, take the mode according to the second vibration target distances to obtain a vibration mode, and determine whether the vibration mode is within the preload detection range, as the preload determination result, but not The above algorithm is limited.

In step 76, the processing module 14 determines whether the preload of the ball screw 4 has declined according to the result of the preload determination. When it is determined that the pre-compression of the ball screw 4 has not decayed (e.g., the average number of vibrations is within the pre-compression detection range), proceed to step 77; when it is determined that the pre-compression of the ball screw 4 has decayed (e.g., the When the average number of vibrations is outside the preload detection range), proceed to step 78 of the process.

In step 77, the processing module 14 generates a pre-compression non-decaying message indicating that the pre-compression of the ball screw 4 has not decayed, and displays the pre-compression non-decaying message on the display module 13.

In step 78, the processing module 14 obtains at least one inertial force time domain data corresponding to the at least one inertial force signal according to the received at least one inertial force signal. Among them, each inertial force time domain data corresponds to a reciprocating cycle of the nut 41 of the ball screw 4 relative to the screw shaft 42, and each reciprocating cycle means that the nut 41 starts to move from an initial position of the screw shaft 42 To an end position, and then return to the initial position from the end position. Preferably, each inertial force time domain data can also only cover the acceleration (deceleration) period of the reciprocating cycle corresponding to the ball screw 4.

In step 79, for each inertial force time domain data, the processing module 14 obtains an inertial force feature vector related to the inertial force time domain data according to the inertial force time domain data. Wherein, each inertial force eigenvector includes a peak-to-peak eigenvector indicating the peak-to-peak value of the inertial force time-domain data corresponding to the inertial force eigenvector, and a peak-to-peak eigenvector indicating the inertial force time-domain data corresponding to the inertial force eigenvector The maximum peak value of the maximum time domain peak eigenvector, and an average peak eigenvector that indicates the sum of the absolute value of the maximum peak value and the absolute value of the minimum peak value of the inertial force time domain data corresponding to the inertial force eigenvector At least one, but not limited to the above.

Referring to FIG. 8, it is worth noting that step 79 further includes step 790 to step 792.

In step 790, for each inertial force time domain data, the processing module 14 performs envelope processing according to the inertial force time domain data to obtain the processed inertial force time domain data.

In step 791, for each processed time-domain inertial force data, the processing module 14 performs a low-pass filter based on the processed time-domain inertial force data to obtain a corresponding processed time-domain data. The filtered time-domain data of the inertial force of the time-domain data of the inertial force. Furthermore, the frequency range reserved by the filtered inertial force time-domain data is 0.1~5Hz bandwidth, therefore, when the frequency range that the second sensing unit 3 can sense only covers 0.1~5Hz bandwidth At this time, there is no need to perform the low-pass filtering described in step 791, and step 792 is directly performed.

In step 792, for each filtered time-domain inertial force data, the processing module 14 obtains the inertial force characteristic corresponding to the filtered time-domain inertial force data according to the filtered time-domain inertial force data vector.

In step 80, for each inertial force feature vector, the processing module 14 calculates a second inertial force candidate distance between the inertial force feature vector and each inertial force reference vector.

In step 81, for each inertial force feature vector that has obtained the second inertial force candidate distances, the processing module 14 obtains a corresponding inertial force candidate distance from the inertial force candidate distances corresponding to the inertial force feature vector. The shortest distance of the second inertial force target distance.

In step 82, the processing module 14 obtains the backlash detection range based on the second inertial force target distance and the backlash used to determine whether the inertial force feature vectors indicate that the ball screw 4 has generated backlash. The result of the gap determination. For example, the processing module 14 averages the distances of the second inertial force targets to obtain an average inertial force, and determines whether the average inertial force is within the detection range of the backlash as the backlash. Or, take the mode according to the second inertial force target distance to obtain an inertial force mode, and determine whether the inertial force mode is within the detection range of the back gap, as the back gap Judgment result, but not limited to the above algorithm.

In step 83, the processing module 14 determines whether the ball screw 4 has a backlash based on the backlash determination result. When it is determined that the ball screw 4 has no backlash (for example: the average number of inertial forces is within the backlash detection range), proceed to step 84; when it is determined that the ball screw 4 has generated backlash (for example: the inertial force) When the force average is within the backlash detection range), proceed to step 85 of the process.

In step 84, the processing module 14 generates a pre-compression-only decay message indicating that the pre-compression of the ball screw 4 has decayed but no backlash has occurred, and displays the pre-compression-only decay message on the display module 13 .

In step 85, the processing module 14 generates a backlash generated message indicating that the backlash of the ball screw 4 has been generated, and displays the backlash generated message on the display module 13.

In summary, the method for judging the precompression recession of the ball screw of the present invention uses the precompression judgment result for judging whether the precompression of the ball screw 4 has recessed, and the precompression judgment result for judging whether the ball screw 4 has generated back pressure. As a result of the backlash determination result, it can be detected that the preload of the ball screw 4 is not decayed, only decayed but backlash has not yet occurred, or backlash has occurred. Therefore, the objective of the invention can be achieved.

However, the above are only examples of the present invention. When the scope of implementation of the present invention cannot be limited by this, all simple equivalent changes and modifications made in accordance with the scope of the patent application of the present invention and the content of the patent specification still belong to Within the scope of the patent of the present invention.

100: Ball screw preload decay judging system 1: computer device 11: Communication module 12: Storage module 13: display module 14: Processing module 2: The first sensing unit 3: The second sensing unit 4: Ball screw 41: Nut 42: screw shaft 43: Loop accessories 44: Ball 50~53: Step 60~63: Step 70~85: Step 710~714: steps 790~792: steps

Other features and effects of the present invention will be clearly presented in the embodiments with reference to the drawings, in which: Figure 1 is a block diagram illustrating a ball screw preload decay judging system for implementing an embodiment of the ball screw preload decay judging method of the present invention; 2 is a schematic diagram illustrating a first sensing unit, a second sensing unit and a ball screw of the embodiment; Figure 3 is a flowchart illustrating a pre-compression training procedure of this embodiment; Figure 4 is a flowchart illustrating a backlash training procedure of this embodiment; Figure 5 is a flowchart illustrating steps 70 to 77 of a pre-compression determination program of this embodiment; Fig. 6 is a flowchart illustrating steps 78 to 85 of the pre-compression determination procedure of this embodiment; FIG. 7 is a flowchart illustrating how the preload determination procedure of this embodiment obtains the frequency domain data of the vibrator; and Fig. 8 is a flowchart illustrating how the pre-compression determination program of this embodiment obtains the inertial force feature vector.

70~77: Step

Claims (10)

A method for judging the pre-compression decay of a ball screw is implemented by a computer device whose signal is connected to a first sensing unit, the first sensing unit being installed on a nut of a ball screw and adjacent to the ball screw A circulating accessory periodically transmits a vibration signal related to the vibration of the ball in the circulating accessory to the computer device. The method for judging the preload decline of the ball screw includes the following steps: (A) Using the computer device, according to the received at least one vibration signal, obtain vibration time-domain data corresponding to at least one vibration signal; (B) Using the computer device to obtain at least one vibrator frequency domain data corresponding to the vibration time domain data based on the vibration time domain data; (C) For each vibrator frequency domain data, using the computer device, obtain a vibration eigenvector related to the vibrator frequency domain data according to the vibrator frequency domain data; (D) Using the computer device, obtain a preload determination result based on the vibration feature vectors, multiple vibration reference vectors, and a preload detection range; and (E) Using the computer device, determine whether the preload of the ball screw has declined according to the result of the preload determination. The method for judging the decline of the ball screw preload according to claim 1, wherein step (B) further includes the following steps: (B-1) Using the computer device to perform envelope processing based on the vibration time domain data to obtain the processed vibration time domain data; (B-2) According to the processed vibration time domain data by the computer device, obtain a target vibration time domain data related to the constant velocity period when the ball screw is moving; (B-3) Obtain the at least one vibrator frequency domain data according to the target vibration time domain data by the computer device. The method for judging the decline of the ball screw preload according to claim 2, wherein the step (B-3) further includes the following steps: (B-3-1) Obtain at least one vibrator time-domain data based on the target vibration time-domain data through the computer device; and (B-3-2) For each vibrator time domain data, the computer device performs Fourier transform according to the vibrator time domain data to obtain the vibrator frequency domain data corresponding to the vibrator time domain data. The method for judging the decline of the ball screw preload according to claim 1, wherein step (D) includes the following steps: (D-1) For each vibration feature vector, using the computer device, calculate a second candidate vibration distance between the vibration feature vector and each vibration reference vector; (D-2) For each vibration eigenvector for which the second vibration candidate distances have been obtained, by the computer device, from the second vibration candidate distances corresponding to the vibration eigenvector, obtain a corresponding one with the shortest The second vibration target distance of the distance; and (D-3) The computer device obtains the preload determination result according to the second vibration target distances and the preload detection range. The ball screw preload decay determination method according to claim 1, wherein in step (C), each vibration eigenvector includes a kurtosis eigenvector indicating the kurtosis of the corresponding vibrator frequency domain data, and a At least one of a maximum frequency domain peak eigenvector indicating the maximum peak of the corresponding vibrator frequency domain data, and a total energy eigenvector indicating the total energy of the corresponding vibrator frequency domain data. According to the method for judging the decline of preload of a ball screw according to claim 1, the computer device stores a plurality of first training vibration feature vectors and a plurality of second training vibration feature vectors, wherein, before step (D), it further includes The following steps: (F) Using the computer device, according to the first training vibration feature vectors, use an unsupervised algorithm to obtain the vibration reference vectors in the corresponding data space; and (G) By using the computer device, the preload detection range is obtained according to the vibration reference vectors and the second training vibration characteristic vectors. The ball screw preload decay determination method according to claim 5, wherein, in step (F), the unsupervised algorithm includes a grouping algorithm, and the vibration reference vectors include a plurality of vibrations obtained by the grouping algorithm The clusters respectively correspond to the vector of the center. The method for judging the decline of the ball screw preload according to claim 5, wherein, in step (F), the unsupervised algorithm includes a self-organizing mapping algorithm, and the vibration reference vectors include a self-organizing mapping algorithm The obtained update times are greater than a preset number of neurons corresponding to the vectors. The method for judging the decline of the ball screw preload according to claim 5, wherein step (G) further includes the following steps: (G-1) For each second training vibration feature vector, using the computer device, calculate a first vibration candidate distance between the second training vibration feature vector and each vibration reference vector; (G-2) For each second training vibration feature vector that has obtained the first vibration candidate distances, by the computer device, from the first vibration candidate distances corresponding to the second training vibration feature vector , Obtain the first vibration target distance corresponding to the shortest distance; and (G-3) Using the computer device, obtain the preload detection range according to the first vibration target distances. According to the method for judging the deterioration of the preload of the ball screw according to claim 1, the computer device is also signally connected to a second sensing unit, the second sensing unit is installed on the nut and periodically transmits a signal related to the nut The inertial force signal relative to the inertial force of the moving direction of a screw shaft of the ball screw is sent to the computer device, wherein, after step (E), the following steps are further included: (H) When it is determined that the preload of the ball screw has decayed, at least one pair of inertial force time-domain data corresponding to at least one inertial force signal is obtained by the computer device according to the received at least one inertial force signal; (I) For each inertial force time-domain data, use the computer device to obtain an inertial force eigenvector related to the inertial force time-domain data based on the inertial force time-domain data; (J) With the computer device, a backlash determination result is obtained based on the inertial force feature vectors, multiple inertial force reference vectors, and a backlash detection range; and (K) Using the computer device, determine whether the ball screw has backlash based on the backlash determination result.

Images