How to monitor the condition of linear guide rails

Abstract

This provides a method for monitoring the condition of linear guide rails. [Solution] The method for monitoring the condition of a linear guide rail according to the present invention includes the steps of: generating a detection signal by detecting the linear guide rail during the monitoring stage; capturing numerical values ​​of multiple types of signal features from the detection signal as feature data based on the theoretical feature frequencies of the rolling element connectors of the linear guide rail; inputting the feature data into a standard model to project it into a feature space, determining whether the feature data is within a threshold range of the feature space, establishing the standard model based on unsupervised learning techniques and the detection results of the healthy linear guide rail; and defining that an abnormality has occurred in the rolling element connector if the feature data is not within the threshold range. This makes it possible to quickly and accurately detect abnormal parts of the linear guide rail and immediately perform the relevant processing.

Description

Technical Field

[0001] The present invention relates to a method for monitoring abnormalities, and more particularly to a method for monitoring the state of a linear guide rail.

Background Art

[0002] Many of the conventional detection mechanisms applied to the detection of the state of a linear guide rail can be easily identified from the appearance, have a high detectability, and are for the failure of the linear guide rail caused by damage to the track surface (for example, abnormal indentation, abnormal wear, load bias, fretting corrosion, exfoliation corrosion, etc., but the present invention is not limited thereto). In the conventional patent documents, for example, there are related descriptions such as the following Patent Documents 1 to Patent Documents 3.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, the failure of the linear guide rail due to damage that cannot be easily identified from the appearance, has a low detectability, and is not related to the track surface (for example, intrusion of foreign matter, generation of rust, failure of the rolling element connector, etc., but the present invention is not limited thereto) could not be effectively detected.

[0005] Therefore, the inventor of the present invention considered that the above-mentioned drawbacks could be improved, and as a result of intensive studies, the present invention was proposed to effectively improve the above-mentioned problems with a reasonable design.

[0006] The present invention has been made in view of these circumstances, and its purpose is to provide a method for monitoring the condition of a linear guide rail that enables high-speed and accurate diagnosis of abnormal parts of the linear guide rail, and allows workers to immediately perform the relevant processing on the linear guide rail. [Means for solving the problem]

[0007] To solve the above problems, the present invention employs the following means. A method for monitoring the state of a linear guide rail according to one aspect of the present invention comprises a linear guide rail, a slide rail, a slider covering the slide rail, a rolling element connector installed between the slide rail and the slider, and rolling elements installed in the rolling element connector. The process includes: (A) driving the linear guide rail to operate using a control unit; (B) detecting the linear guide rail using a sensor module and generating a detection signal during the monitoring phase; (C) capturing numerical values ​​of multiple signal features from the detection signal as feature data using a feature capture unit based on the theoretical feature frequencies of the rolling element connector; (D) determining whether the feature data is within a threshold range of the feature space by inputting the feature data into a standard model and projecting the feature data into a feature space using a cluster algorithm based on unsupervised learning technology, thereby establishing the standard model based on the unsupervised learning technology and the results of the sensor module detecting the linear guide rail in a healthy state; and (E) defining that an abnormality has occurred in the rolling element connector if the feature data is not within the threshold range.

[0008] According to one aspect of the present invention, prior to step (B), the method further includes the steps of: detecting with the sensor module that the linear guide rail has generated a detection signal; capturing a plurality of feature data from the detection signal with the feature capture unit based on the theoretical feature frequency of the rolling element connector, wherein each feature data is a numerical value of a plurality of signal features captured from the detection signal; and establishing the standard model with the standard establishment unit using the cluster algorithm to project at least a portion of these feature data into the feature space, estimating the corresponding difference value for each projected feature data, and estimating a threshold range to be provided to the feature space based on all the estimated difference values.

[0009] According to one aspect of the present invention, feature data capture is performed on the detection signal in the time domain or the frequency domain. When feature data capture is performed on the detection signal in the frequency domain, a signal preprocessing unit converts the detection signal from the time domain to the frequency domain, and then the feature capture unit captures the feature data from the detection signal in the frequency domain.

[0010] According to one aspect of the present invention, the step of establishing a standard model by using the cluster algorithm to project at least a portion of the feature data onto the feature space using the standard establishment unit, estimating the corresponding difference value for each projected feature data, and estimating the threshold range to be provided to the feature space based on all the estimated difference values, includes the steps of: dividing the feature data into a first feature dataset and a second feature dataset; establishing an initial model by projecting the first feature dataset onto the feature space using the cluster algorithm, estimating the corresponding difference value for each feature data in the first feature dataset, and estimating the threshold range based on all the estimated difference values; inputting the second feature dataset into the initial model, determining whether the second feature dataset is within the threshold range, and verifying the initial model; and, if the second feature dataset is within the threshold range, confirming that the initial model has passed, and completing the establishment of the standard model. The step of dividing the aforementioned feature data into the first feature dataset and the second feature dataset involves clustering the aforementioned feature data using a random sampling method, or clustering the aforementioned feature data based on the timestamp to which each feature data corresponds.

[0011] According to one aspect of the present invention, each feature data corresponding to a difference value is displayed based on a dimensionless distance or similarity. If the difference value is a dimensionless distance, the standard establishment unit obtains a confidence threshold by statistically stating all estimated difference values ​​and defines the threshold range in the feature space based on the confidence threshold so that all estimated difference values ​​are less than or equal to the confidence threshold. If the difference value is similarity, the standard establishment unit directly defines a similarity threshold based on all estimated difference values ​​and defines the threshold range in the feature space based on the similarity threshold so that all estimated difference values ​​are greater than or equal to the similarity threshold. Optionally, the method further includes the steps of: capturing a target signal portion from the detection signal using a signal preprocessing unit, the target signal portion referring to the signal component generated by the sensor module when the slider slides in one direction on the slide rail, and the capture of the target signal portion being based on a timestamp or direction marking; and filtering the target signal portion using the signal preprocessing unit based on the theoretical feature frequency of the rolling element connector. Each of the aforementioned feature data is captured from the filtered portion of the target signal in the time domain or frequency domain.

[0012] Another aspect of the present invention is a system for monitoring the condition of a linear guide rail, wherein the linear guide rail comprises a slide rail, a slider covering the slide rail, a rolling element connector installed between the slide rail and the slider, and rolling elements installed in the rolling element connector. The system is applied to perform a method for monitoring the condition of the linear guide rail.

[0013] The following information will become clear from the description in the specification and drawings described later. [Brief explanation of the drawing]

[0014] [Figure 1]It is a schematic diagram showing an example applied to the connection between the linear guide rail and the sensing module of the present invention. [Figure 2] It is a schematic exploded view showing the linear guide rail shown in FIG. 1. [Figure 3] It is a schematic cross-sectional view showing the example shown in FIG. 1. [Figure 4] It is an external perspective view showing an example of a rolling element connector and rolling elements of the linear guide rail shown in FIG. 2. [Figure 5] It is a schematic cross-sectional view showing an example of a rolling element connector and rolling elements shown in FIG. 4. [Figure 6] It is a functional block diagram showing an apparatus for detecting the state of a linear guide rail according to an embodiment of the present invention. [Figure 7] A method for detecting the state of a linear guide rail according to an embodiment of the present invention is a flowchart in the modeling stage. [Figure 8] A method for detecting the state of a linear guide rail according to an embodiment of the present invention is a flowchart in the monitoring stage. [Figure 9] It is a spectrum showing target signal sub-stages corresponding to a sound rolling element connector and an abnormal rolling element connector according to an embodiment of the present invention. [Figure 10] It is a distribution diagram showing characteristic frequencies in a plurality of captured characteristic data according to an embodiment of the present invention. [Figure 11] It is a distribution diagram showing the variance of characteristic frequencies in a plurality of captured characteristic data according to an embodiment of the present invention. [Figure 12] It is a distribution diagram showing the amplitude of characteristic frequencies in a plurality of captured characteristic data according to an embodiment of the present invention. [Figure 13] It is a distribution diagram showing the amplitude variation of characteristic frequencies in a plurality of captured characteristic data according to an embodiment of the present invention. [Figure 14] In an embodiment of the present invention, it is a distribution diagram of data points representing feature data mapped to a feature space and a reliability threshold. [Figure 15]In one embodiment of the present invention, it is a distribution diagram of a data point representing feature data mapped to a feature space and a similarity threshold value. [Figure 16] In one embodiment of the present invention, it is a distribution diagram of a data point representing feature data mapped to a feature space and a reliability threshold value.

Embodiments for Carrying Out the Invention

[0015] Hereinafter, embodiments of the present invention will be described in detail. However, the present invention is not limited thereto, and various modifications are possible within the described scope. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0016] First, an example of a specific embodiment of a system for monitoring the state of a linear guide rail of the present invention will be described while referring to FIGS. 1 to 16.

[0017] A system (hereinafter abbreviated as a monitoring system) for monitoring the state of a linear guide rail 10 according to an embodiment of the present invention diagnoses whether the linear guide rail 10 has failed by executing a method for monitoring the state of the linear guide rail 10 (hereinafter abbreviated as a monitoring method). Specifically, it diagnoses whether the rolling element connector 13 (or, referred to as a chain) has failed. By doing so, it enables an operator to immediately detect an abnormality in the rolling element connector 13 and to perform related measures.

[0018] The monitoring system further includes, in addition to the linear guide rail 10, a sensor module 20, a servo device 30 that is electrically connected to the linear guide rail 10 and can communicate with the sensor module 20, an external signal source 40 that is electrically connected to the servo device 30, and an application program 50.

[0019] The linear guide rail 10 comprises a slide rail 11, a slider 12 covering the slide rail 11, a rolling element connector 13 installed between the slide rail 11 and the slider 12, and rolling elements 14 installed in the rolling element connector 13.

[0020] The sensor module 20 comprises at least one sensor capable of detecting the motion state of the linear guide rail 10 during operation (for example, vibration or movement of the slider 12, collision of the rolling element 14 with surrounding objects, etc., but the present invention is not limited to these). The type of sensor is, for example, a sensor with directional measurement such as an accelerometer, speedometer, or displacement meter, or a sensor without directional measurement such as a microphone, but the present invention is not limited to these. The sensor can be placed on or near the linear guide rail 10, and the placement position of the sensor is determined by the type of sensor. For example, an accelerometer, speedometer, or displacement meter can be placed on the slider 12, and a microphone can be placed near the linear guide rail 10 and pick up sound directed towards the slider 12. In this embodiment, the sensor module 20 is implemented with one accelerometer and is placed on the outer side surface 121 of the slider 12.

[0021] The servo device 30 includes a storage 31, a processor module 32 electrically connected to the storage 31, an input / output interface module 33 which is connected to or communicates with the linear guide rail 10, the sensor module 20, and the external signal source 40, and is also electrically connected to the processor module 32, and a display module 34 which is electrically connected to the processor module 32.

[0022] The storage 31 may include one or more memory and / or hard disks for storing the application program 50. The application program 50 includes parameters, algorithms, program code, and other materials necessary to perform the detection method, and is executed by the processor module 32 to realize the detection method. In this way, one or more databases (not shown) are established in the storage 31, and a standard model 311 is established which is connected to at least one database and stored in the other databases.

[0023] The processor module 32 may comprise one processing unit or multiple processing units connected to each other, and when executing the application program 50, it defines a control unit 321, a signal preprocessing unit 322, a feature capture unit 323 connected to the signal preprocessing unit 322, a standard establishment unit 324 connected to the feature capture unit 323, an anomaly determination unit 325 connected to the feature capture unit 323, and a presentation generation unit 326 connected to the anomaly determination unit 325.

[0024] The input / output interface module 33 includes multiple wired or wireless transmission interfaces and provides a bridge for the processor module 32 to communicate with the linear guide rail 10, the sensor module 20, and the external signal source 40, respectively. In this way, the processor module 32 outputs control signals via the input / output interface module 33 to control the operation of the linear guide rail 10 and receives signals returned from the linear guide rail 10 (for example, motor signals, but the present invention is not limited thereto), outputs control signals via the input / output interface module 33 to control the operation of the sensor module 20 and receives detection signals returned from the sensor module 20, outputs a data call request to the external signal source 40 via the input / output interface module 33 and receives data returned from the external signal source 40.

[0025] The display module 34 extracts relevant material from the storage 31 based on an abnormality signal output by the processor module 32, and then shows the extracted material to the operator based on a predetermined display method. The type of display module 34 is determined by the display method. For example, in the case of audio presentation, the display module 34 may be an acoustic effect element, and this acoustic effect element may reproduce acoustic effect material. For example, in the case of visual presentation of an abnormality, the display module 34 may be a display module (for example, a display panel, but the present invention is not limited to these), and the display module displays a user interface (not shown) provided by the application program 50 using the processor module 32, and displays material such as characters or patterns on this user interface, but the present invention is not limited to these. Alternatively, the display module 34 may be an indicator lamp, and the indicator lamp operates based on a control command that serves as material.

[0026] The external signal source 40 is, for example, a programmable logic controller (PLC) system for controlling and recording the operation of each device in the facility to which the linear guide rail 10 is attached, but the present invention is not limited to this. Therefore, the processor module 32 requests the external signal source 40 to provide the necessary operation records via the input / output interface module 33 (for example, by providing the operation records of the linear guide rail 10 as an appendix to the operation status records of the PLC, but the present invention is not limited to this).

[0027] The monitoring method according to the present invention will be described exemplified below. This includes a flowchart for the model establishment stage and a flowchart for the monitoring stage, the monitoring stage being after the model establishment stage. The following is an illustrative flowchart of the model establishment stage.

[0028] <Step S101>: The theoretical characteristic frequency of the rolling element connector 13 is input to the user interface. In this way, the processor module 32 acquires the theoretical characteristic frequency via the user interface and saves it to the storage 31. The theoretical characteristic frequency of the rolling element connector 13 is obtained by calculation based on the specifications of the linear guide rail 10. Specifically, the speed at which the slider 12 moves on the slide rail 11 is V, the diameter of the rolling element 14 is Bd, the minimum thickness of the spacer 131 of the rolling element connector 13 is d, the theoretical characteristic frequency of the rolling element connector 13 is f, and f = V / (2 * (Bd + d)).

[0029] <Step S102>: The control unit 321 drives the linear guide rail 10 to operate, moving the slider 12 relative to the slide rail 11.

[0030] <Step S103>: The control unit 321 controls the sensor module 20 to detect the linear guide rail 10, generates a detection signal, and the sensor module 20 returns the detection result to the processor module 32. The signal to be detected is a time-domain signal.

[0031] <Step S104>: The signal preprocessing unit 322 captures the target signal portion from the detected signal. This target signal portion refers to the signal component generated by the sensor module 20 when the slider 12 slides in one direction on the slide rail 11.

[0032] Since the signal components for the outbound and return journeys are usually not exactly the same, if the outbound and return signal components are not used separately for model training, the established standard model 311 will be inaccurate, leading to errors in subsequent state diagnoses. In other words, if the outbound and return signal components are used separately for model training, or if the model is trained using only one-way signal components, the accuracy of the standard model 311 will increase, and the likelihood of misjudging the state will decrease. Training the model using only one-way signal components can also more effectively reduce the amount of computation required.

[0033] For example, since the detection signals generated by the slider 12 in the forward and return directions have corresponding timestamps, the signal preprocessing unit 322 can determine whether the signal component belongs to the forward or return direction based on the timestamp of the detection signal, and then capture the signal component of the forward (or return) direction as the target signal portion.

[0034] Alternatively, to give an example of a sensor with direction measurement, each signal component of the detection signal returned by the sensor module 20 has a direction marking (e.g., positive, negative marking) that indicates the direction of movement of the slider 12. This allows the signal preprocessing unit 322 to determine whether each signal component belongs to the forward or return path based on the direction marking of each signal component, and then capture the forward (or return) signal component as the target signal portion.

[0035] Alternatively, in an example where the motor signal of the linear guide rail 10 is transmitted to the processor module 32, the input and output signals when the motor (not shown) rotates in the forward and reverse directions have corresponding timestamps. This allows the signal preprocessing unit 322 to determine whether the signal component belongs to the forward or reverse path based on the timestamp of the motor signal, and then capture the signal component of the forward (or reverse) path as the target signal portion.

[0036] Alternatively, consider the example where a PLC signal provided by the equipment to which the linear guide rail 10 is attached (i.e., an external signal source 40) is transmitted to the processor module 32. Since the PLC system stores records related to the round trip of the linear guide rail 10, the signal preprocessing unit 322 can determine whether the signal component belongs to the forward or return path based on the timestamp of the PLC signal provided by the PLC system, and then capture the forward (or return) signal component as the target signal portion.

[0037] However, the target signal portion capture method according to the present invention is not limited to the above-described examples.

[0038] <Step S105>: Signal filtering is performed. That is, the signal preprocessing unit 322 digitally filters the detected signal based on the theoretical characteristic frequency of the rolling element connector 13 to remove noise and irrelevant signal components. For example, the signal preprocessing unit 322 performs a bandpass filter on the target signal portion, but the present invention is not limited to this. In this way, the accuracy of the model can be increased and the accuracy of the state diagnosis can be improved while reducing the amount of subsequent computation processing.

[0039] <Step S106>: Perform a signal conversion operation. That is, the signal preprocessing unit 322 converts the detected signal from the time domain to the frequency domain. The signal conversion operation is performed, for example, on the target signal portion, but the present invention is not limited thereto. The signal conversion operation is performed, for example, using the Fast Fourier Transform, but the present invention is not limited thereto. The target signal portion in the frequency domain is shown by a solid line in Figure 9, for example, but the present invention is not limited thereto.

[0040] <Step S107>: Perform the feature data capture operation. Specifically, the feature capture unit 323 captures multiple feature data from the detection signal based on the theoretical feature frequency of the rolling element connector 13, with each feature data being a numerical value of multiple types of signal features of the signal component. For example, the feature capture unit 323 captures these feature data from the target signal portion using spectral analysis techniques.

[0041] <Step S108>: Perform the grouping of feature data. That is, the feature capture unit 323 groups the multiple feature data captured in step S107. For example, it may be divided into a first feature dataset and a second feature dataset, but the present invention is not limited to this.

[0042] In one example, the method for grouping the captured feature data is to group these feature data using a random sampling method. For example, if 200 feature data points have been captured, a first feature dataset and a second feature dataset are formed by sampling these feature data points in an average and random manner.

[0043] In one example, the clustering method for captured feature data groups these captured feature data based on the timestamp at which each feature data was captured. For example, if 200 feature data have been captured, the first 100 feature data are defined as the first feature dataset, and the subsequent 100 feature data are defined as the second feature dataset.

[0044] <Step S109>: The initial model is established. That is, the standard establishment unit 325 uses a cluster algorithm based on unsupervised learning techniques (and may also selectively combine with a dimensionality reduction algorithm) to project the first feature dataset into the feature space, estimate the corresponding difference value for each projected feature data, and establish the initial model by estimating a threshold range to be provided to the feature space based on all the difference values. The cluster algorithm is, for example, the K-means algorithm, the SOM algorithm, the PCA algorithm, or other algorithms, but the present invention is not limited to these. The difference values ​​are displayed based on, for example, dimensionless distance or similarity, but the present invention is not limited to these. Specifically, the method of displaying the difference values ​​is determined by the cluster algorithm adopted. For example, if the K-means algorithm is adopted, the difference value is dimensionless distance or similarity. For example, if the SOM algorithm is adopted, the difference value is dimensionless distance. If the PCA algorithm is adopted, the difference value is similarity.

[0045] As an example where the clustering algorithm is the K-means algorithm and the difference value is a dimensionless distance, the standard probabilistic unit 325 uses the K-means algorithm (and may also selectively combine it with a dimensionality reduction algorithm) to project each captured feature data into the feature space, and each projected feature data has a corresponding data point in the feature space. Then, the standard probabilistic unit 325 calculates the dimensionless distance between each data point and the center of the cluster in which it exists as the difference value. Next, the standard probabilistic unit 325 uses a statistical algorithm to perform statistics on all the difference values ​​to obtain a confidence threshold (shown by the solid line representing the confidence threshold in Figure 14), and based on this confidence threshold, it defines a threshold range in the feature space (shown as the range below the solid line representing the confidence threshold in Figure 14). All estimated difference values ​​are below the confidence threshold, i.e., within the threshold range.

[0046] As an example where the clustering algorithm is the K-means algorithm and the difference value is similarity, the standard establishment unit 325 projects each captured feature data into the feature space using the K-means algorithm (and optionally a combination of dimensionality reduction algorithms). Then, the standard establishment unit 325 obtains a feature matrix representing the center of the cluster (i.e., one feature data) and a feature matrix representing the other data points of this cluster, and calculates the correlation coefficient (i.e., similarity) as the difference value using these two feature matrices. Next, the standard establishment unit 325 directly defines a similarity threshold (for example, 0.7, but the present invention is not limited to this; see Figure 15) based on all the difference values, and defines a threshold range in the feature space (shown as the range greater than or equal to the solid line representing the similarity threshold in Figure 15) based on this similarity threshold. All difference values ​​are greater than or equal to the similarity threshold, i.e., within the threshold range.

[0047] As an example where the clustering algorithm is the SOM algorithm and the difference value is a dimensionless distance, the standard establishment unit 325 uses the SOM algorithm to map (input) each captured feature data into an artificial neural network in the feature space (or phase space). The neurons repeatedly train on the input feature data and the competition mechanism to search for the winning neurons (or nodes) that represent the individual projected feature data within the artificial neural network, and repeat the training to obtain a model. Then, the standard establishment unit 325 inputs each feature data of the same feature data (i.e., all the feature data for generating the model) into this model, obtains the dimensionless distance between each feature data and its corresponding node as the difference value, and obtains a confidence threshold by statistically analyzing all the difference values ​​using a statistical algorithm. Based on this confidence threshold, a threshold range is defined in the feature space. All estimated difference values ​​are less than or equal to the confidence threshold, i.e., within the threshold range.

[0048] As an example where the clustering algorithm is the PCA algorithm and the difference value is similarity, the standard establishment unit 325 projects each captured feature data into the feature space using the PCA algorithm, and each projected feature data has a corresponding data point in the feature space. Then, the standard establishment unit 325 calculates the similarity corresponding to each data point as the difference value, for example, using Euclidean distance, cosine similarity, or Manhattan distance, but the present invention is not limited to these. Next, the standard establishment unit 325 directly defines a similarity threshold (for example, 0.7, but the present invention is not limited to this) based on all the difference values, and defines a threshold range in the feature space based on this similarity threshold. All difference values ​​are greater than or equal to the similarity threshold, i.e., within the threshold range.

[0049] <Step S110>: The standard establishment unit 325 performs the model validation. That is, the second feature dataset is fed into the initial model, and it is determined whether the second feature dataset is within the threshold range to verify whether the initial model has converged sufficiently (i.e., whether the state of the rolling element connector 13 has been diagnosed sufficiently accurately). If it has converged, it displays that the validation was successful and executes step S112. If it has not converged, it displays that the validation failed and executes step S111.

[0050] <Step S111>: The standard establishment unit 325 notifies the control unit 321 to control the display of the relevant page on the user interface, allowing the operator to change settings on the user interface. For example, the number of feature data to capture can be changed, but the present invention is not limited to this. After step S111 is completed, the process returns to step S104 and steps S104 through S111 are repeated until verification is successful.

[0051] <Step S112>: The standard establishment unit 325 defines the initial model that has been successfully verified as the standard model 311. The standard establishment unit 325 also saves the standard model 311 to the storage 31.

[0052] The present invention allows for the selection of the execution order of steps S104 to S106. For example, step S105 may be executed first, then step S104, and finally step S106; or step S106 may be executed first, then step S105, and finally step S104; or step S105 may be executed first, then step S106, and finally step S104.

[0053] In step S107, the present invention allows for the selection of capturing feature data from the time-domain detection signal (or target signal portion). In this case, step S106 can be omitted.

[0054] In other embodiments, depending on the situation (for example, a situation in which difference values ​​are displayed by similarity, but the present invention is not limited to these), steps S108, S110, S111, and S112 may be omitted. In this situation, the standard establishment unit 325 may directly use all of the feature data captured in step S107 in step S109 to establish an initial model, and may also directly use the initial model established in step S109 as the standard model 311 and store it in storage 31.

[0055] In this invention, the linear guide rail 10 used in the model establishment stage is the same normal and healthy linear guide rail. That is, all feature data for training the standard model 311 is normal and healthy feature data. Therefore, the training process for the standard model can be simplified, the amount of data computation required to establish the standard model can be reduced, the storage space for the standard model can be reduced, and practical operational feasibility can be met.

[0056] In this way, all linear guide rails 10 with the same specifications and model number can be monitored for their health status using the health model established through the flowchart in the model establishment stage.

[0057] The following is an illustrative flowchart of the monitoring phase.

[0058] <Step S201>: The control unit 321 drives the linear guide rail 10 to operate, moving the slider 12 relative to the slide rail 11.

[0059] <Step S202>: The control unit 321 controls the sensor module 20 to detect the linear guide rail 10, generates a detection signal, and the sensor module 20 returns the detection result to the processor module 32. This detection signal is a time-domain signal.

[0060] <Step S203>: Capture the target signal portion. The specific method for step S203 is described in step S104, and will not be repeated here.

[0061] <Step S204>: Perform signal filtering. The specific method for Step S204 is described in Step S105, and will not be repeated here.

[0062] <Step S205>: Perform the signal conversion operation. The specific method for Step S205 is described in Step S106, and will not be repeated here.

[0063] <Step S206>: Perform the feature data capture operation. For the specific method of Step S206, refer to Step S107, and the explanation will not be repeated here. For example, if the rolling element connector 13 is healthy, the feature data is captured from the target signal portion in the frequency domain shown by the solid line in Figure 9. If the rolling element connector 13 is abnormal, the feature data is captured from the target signal portion in the frequency domain shown by the dashed line in Figure 9.

[0064] <Step S207>: The anomaly detection unit 325 acquires the standard model 311 from the storage 31 and inputs one or more captured feature data into the standard model 311. The feature data is then projected into the feature space using a cluster algorithm based on unsupervised learning technology. The method for projecting the feature data into the feature space refers to the method for projecting the feature data into the feature space and forming corresponding data points as described in step S109, and that explanation will not be repeated here.

[0065] <Step S208>: The anomaly determination unit 325 determines whether the projected feature data is within the threshold range of the feature space. If it is within the threshold range, the anomaly determination unit 325 defines the rolling element connector 13 as healthy or normal and executes step S209. If it is not within the threshold range, the anomaly determination unit 325 defines the rolling element connector 13 as abnormal and executes step S210. For example, if the data points of the projected feature data are within the range enclosed by the confidence threshold represented by the thick solid line in the feature space shown in Figure 16, the anomaly determination unit 325 defines the rolling element connector 13 as healthy. Conversely, if the data points of the projected feature data are outside the range enclosed by the confidence threshold represented by the thick solid line in the feature space shown in Figure 16, the anomaly determination unit 325 defines the rolling element connector 13 as abnormal.

[0066] <Step S209>: The system indicates that it is normal. Specifically, the abnormality detection unit 325 generates a health signal and provides it to the presentation generation unit 326. Based on this health signal, the presentation generation unit 326 extracts relevant material from the storage 31 and forms the extracted material into presentation content based on the method defined by the application program 50, and provides it to the presentation display module 34. In this way, the presentation display module 34 shows this presentation content to the operator.

[0067] <Step S210>: An abnormality is indicated. Specifically, the abnormality detection unit 325 generates an abnormality signal and provides it to the indication generation unit 326. Based on this abnormality signal, the indication generation unit 326 extracts relevant material from the storage 31 and forms the extracted material as indication content based on the method defined by the application program 50, and provides it to the indication display module 34. In this way, the indication display module 34 shows this indication content to the operator. For example, an indicator lamp that serves as an example for the indication display module 34 may emit a warning signal light.

[0068] In the present invention, the linear guide rail 10 used in the monitoring stage may be the same linear guide rail 10 used in the model establishment stage. In this situation, if the system proceeds immediately to the monitoring stage without stopping immediately after the model establishment stage, step S201 can be omitted. However, if the system restarts the linear guide rail 10 under the same operating conditions after the model establishment stage and proceeds to the monitoring stage, step S201 is required. Alternatively, the linear guide rail 10 used in the monitoring stage may be another linear guide rail 10 having the same specifications and operating conditions as the linear guide rail 10 used in the model establishment stage. Therefore, in this situation, step S201 is required.

[0069] In the present invention, it is possible to adjust the execution order of steps S203 to S205. For example, step S204 may be executed first, then step S203, and finally step S205. Alternatively, step S205 may be executed first, then step S204, and finally step S203. Or, step S204 may be executed first, then step S205, and finally step S203.

[0070] In this invention, in step S206, it is possible to select to capture feature data from the time-domain detection signal (or target signal portion). In this case, step S205 can be omitted. However, both steps S107 and S206 must capture from either the time-domain detection signal (or target signal portion) or the frequency-domain detection signal (or target signal portion). In this way, the condition of the rolling element connector 13 can be successfully diagnosed.

[0071] In this way, the detection results obtained through visual analysis are provided quickly and accurately. Furthermore, by providing a status determination message, the operator can immediately know the time point or the start point of the process when the linear guide rail 10 has become invalid, and perform the appropriate processing.

[0072] The present invention allows for the selection of a response bandwidth of the sensor module 20 as BWr and satisfying the condition: BWr > 3f.

[0073] In this invention, the cluster algorithm and statistical method used in the abnormality determination unit 325 are the same as those used in the standard establishment unit 325. This ensures that the standards match, enabling successful diagnosis of the state of the rolling element connector 13.

[0074] In the present invention, the multiple signal features include at least two of the following: each feature frequency of the rolling element connector 13 (see Figure 10), the amplitude of each feature frequency (see Figure 12), the dispersion of each feature frequency (see Figure 11), the amplitude variation of each feature frequency (see Figure 13), and other signal features related to the rolling element connector 13. The cross marks in Figures 10 to 12 represent the results of detecting the linear guide rail 11 when the rolling element connector 13 is abnormal, and the circles represent the results of detecting the linear guide rail 10 when the rolling element connector 13 is healthy.

[0075] Although embodiments of the present invention have been described in detail above with reference to the drawings, the specific configuration is not limited to these embodiments, and design modifications and the like are also included within the scope of the gist of the present invention.

Claims

1. A method for monitoring the state of a linear guide rail, comprising a slide rail, a slider covering the slide rail, a rolling element connector installed between the slide rail and the slider, and rolling elements installed in the rolling element connector, Step (A) involves a control unit driving the linear guide rail to operate, In the monitoring phase, the sensor module detects the linear guide rail and generates a detection signal (B), The feature capture unit captures numerical values ​​of multiple signal features from the detection signal as feature data based on the theoretical feature frequencies of the rolling element connector (C), Step (D) involves inputting the feature data into a standard model using an anomaly detection unit, projecting the feature data into a feature space using a cluster algorithm based on unsupervised learning technology, and determining whether the feature data is within a threshold range in the feature space. The standard model is established based on the unsupervised learning technology and the results of the sensor module detecting the linear guide rail in a healthy state. A method for monitoring the condition of a linear guide rail, comprising the step (E) in which the abnormality determination unit defines that an abnormality has occurred in the rolling element connector if the characteristic data is not within the threshold range.

2. The aforementioned method, Before step (B) above, In the model establishment stage, the sensor module detects that the linear guide rail has generated a detection signal, The feature capture unit captures multiple feature data from the detection signal based on the theoretical feature frequency of the rolling element connector, and each of the feature data is a step which is a numerical value of multiple types of signal features captured from the detection signal. A method for monitoring the state of a linear guide rail according to claim 1, further comprising the steps of: using a standard establishment unit, projecting at least a portion of the feature data onto the feature space using the cluster algorithm; estimating the corresponding difference value for each projected feature data; and establishing the standard model by estimating a threshold range to be provided to the feature space based on all the estimated difference values.

3. A method for monitoring the state of a linear guide rail according to claim 1 or 2, characterized in that, when feature data capture is performed on the detection signal in the time domain or the frequency domain, and when feature data capture is performed on the detection signal in the frequency domain, the signal preprocessing unit converts the detection signal from the time domain to the frequency domain, and then the feature capture unit captures the feature data from the detection signal in the frequency domain.

4. The step of establishing the standard model by using the cluster algorithm to project at least a portion of the feature data onto the feature space using the standard establishment unit, estimating the corresponding difference value for each projected feature data, and estimating the threshold range to be provided to the feature space based on all the estimated difference values, is as follows: The steps include dividing the aforementioned feature data into a first feature dataset and a second feature dataset, The steps include: establishing an initial model by projecting the first feature dataset onto the feature space using the cluster algorithm, estimating the corresponding difference value for each feature data in the first feature dataset, and estimating the threshold range based on all the estimated difference values; The steps include: inputting the second feature dataset into the initial model to determine whether the second feature dataset falls within the threshold range and verifying the initial model; The steps include: confirming that the initial model passes if the second feature dataset is within the threshold range, and completing the establishment of the standard model; The step of dividing the aforementioned feature data into a first feature dataset and a second feature dataset is characterized in that the feature data is clustered by a random sampling method, or the feature data is clustered based on the timestamp to which each feature data corresponds, as described in claim 2.

5. A method for monitoring the state of a linear guide rail according to claim 2 or 4, characterized in that each of the feature data corresponding to a difference value is displayed based on dimensionless distance or similarity, and if the difference value is dimensionless distance, the standard establishment unit obtains a confidence threshold by statistically analyzing all estimated difference values, and defines the threshold range in the feature space based on the confidence threshold, so that all estimated difference values ​​are less than or equal to the confidence threshold, and if the difference value is similarity, the standard establishment unit directly defines a similarity threshold based on all estimated difference values, and defines the threshold range in the feature space based on the similarity threshold, so that all estimated difference values ​​are greater than or equal to the similarity threshold.

Images