Bearing pressing machine control method and system

Abstract

The invention relates to the technical field of mechanical automation control, in particular to a bearing pressing machine control method and system. The method comprises the following steps: acquiring multi-dimensional monitoring data in a press fitting process; based on the difference between the monitoring data at the current moment and the data at each moment in the preset window, constructing a deviation degree sequence of each dimension at the current moment; adjusting the length of a preset window based on the correlation of the deviation degree sequence to obtain a target window at the current moment; determining a truncation distance of clustering analysis according to the data point distribution sparseness of the monitoring data in the target window; taking the truncation distance as a parameter, and performing clustering analysis on the data points in the target window through a density clustering algorithm to identify abnormal data points; and when the monitoring data at the current moment is abnormal, the bearing press-fitting machine is controlled to execute a preset response action. According to the method, the analysis window and the clustering truncation distance are adaptively adjusted, so that the accuracy of abnormal state recognition under the multi-stage working condition is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mechanical automation control, and in particular to a bearing press fitting machine control method and system. BACKGROUND

[0002] Bearing press fitting is a key link in assembly process, and its core is to press the bearing into the shaft or bearing seat smoothly and accurately through a hydraulic or servo system. During the press fitting process, the changes of parameters such as pressure and displacement directly reflect the assembly quality. In a healthy press fitting process, these parameters have different physical correlations in each stage, but the trajectories formed in the state space of the same stage have high consistency.

[0003] However, if the press fitting process has problems such as jamming and partial load, the physical correlation and performance between parameters will be destroyed, resulting in abnormal fluctuations of the press fitting force, and further causing damage to parts and even equipment. Therefore, it is necessary to analyze the parameter change characteristics of the press fitting process to evaluate the press fitting state, identify possible press fitting abnormal conditions and immediately control the equipment to stop for maintenance, so as to avoid serious damage to the assembled parts.

[0004] Considering that once there is an abnormal fluctuation, the data performance will deviate from the change characteristics of the normal stage, resulting in a local density of abnormal data points being significantly lower than that of normal stage data points, therefore, the density peak clustering algorithm can be used to perform density clustering on the data points to identify possible abnormal points. However, the actual press fitting process has multiple stages such as idle stroke, contact press-in, and pressure holding. For example, in the idle stroke stage, the collected multi-dimensional monitoring data changes dramatically and often presents a low-density distribution. The existing density peak clustering algorithm classifies based on a globally fixed cut-off distance threshold. This strategy is easy to mistakenly mark the low-density data under normal working conditions as abnormal points. This will cause the normal running state to be misjudged as abnormal, resulting in a large number of false alarms, thereby affecting the production efficiency. SUMMARY

[0005] To solve the technical problem that the density peak clustering algorithm is difficult to adapt to the multi-stage variable density working conditions of the bearing press fitting process due to the use of a globally fixed cut-off distance, resulting in misjudgment of normal low-density data as abnormal, the present application provides solutions in the following aspects.

[0006] In a first aspect, the present application provides a bearing press fitting machine control method, which comprises the steps of: Acquire multi-dimensional monitoring data of the bearing press at multiple moments during the pressing process; for the monitoring data at the current moment, determine the deviation of each moment in the preset window relative to the current moment based on the difference between the monitoring data at the current moment and the monitoring data at each moment in the preset window, so as to construct a deviation degree sequence of each dimension at the current moment; based on the correlation between the deviation degree sequences of each dimension at the current moment in the preset window, adjust the window length of the preset window to obtain a target window that is in the same pressing stage as the current moment; determine the cutoff distance for clustering analysis based on the sparsity of the data point distribution of the monitoring data in the target window; use the cutoff distance as a parameter to perform cluster analysis on the data points in the target window through a density clustering algorithm to identify abnormal data points; when the monitoring data at the current moment is identified as the abnormal data point, control the bearing press to execute a preset abnormal response action.

[0007] The present invention proposes a dual adaptive control method. Existing technologies are unable to adapt to changes in data density at different stages of the press-fitting process, often misjudging normal sparse data as anomalies, resulting in a high false alarm rate. The present invention first dynamically adjusts the target window based on data correlation to ensure that the analyzed data originates from the same press-fitting stage, and then adaptively determines the cutoff distance for cluster analysis based on the sparsity of the data within the window. This method can intelligently adapt to the dynamic changes of the press-fitting process, significantly improving the accuracy and reliability of anomaly detection under multi-stage, variable-density conditions, effectively avoiding false alarms, and improving production efficiency and the intelligence level of the control system.

[0008] Preferably, the deviation between each moment in the preset window and the current moment satisfies the relationship: ;in, It is In the preset window of time, At the moment Deviation of the dimension; It is At the moment Dimension monitoring data values; It is In the preset window of time, At the moment Dimension monitoring data values; It is In the preset window of time, The maximum value of the dimension monitoring data value; is the maximum function; It is At the moment and within its preset window, The time difference between the moments; is the total number of data points of the preset window; is the absolute value symbol; is the preset first infinitesimal value; is the preset second infinitesimal value.

[0009] The present application introduces time difference as a weight factor. This makes the data points closer to the current time have a greater impact on the deviation degree, while the impact of the distant data points is weakened. It can more accurately reflect the dynamic evolution process of the press-fitting state, so that the deviation degree calculation not only focuses on the numerical deviation, but also considers the proximity in time sequence, thereby more accurately quantifying the difference between the current state and the recent historical state.

[0010] Preferably, the window length of the preset window is adjusted based on the correlation between the deviation degree sequences of each time point in the preset window to obtain a target window in the same press-fitting stage as the current time point, comprising: calculating the correlation coefficient of the deviation degree sequence of each dimension in the preset window; obtaining the standard deviation of the correlation coefficient of the target window corresponding to each historical time point in the historical press-fitting process of the press-fitting workpiece at the current time point, to form a standard deviation set; determining a differentiation threshold value in the same press-fitting stage based on the standard deviation set; and adjusting the preset window size to make the standard deviation of the correlation coefficient in the preset window satisfy the differentiation threshold value in the same press-fitting stage, to obtain the target window.

[0011] The present application determines whether the data in the window belongs to the same press-fitting stage by calculating the standard deviation of the correlation coefficient. This method provides a data-driven and automated means to identify the stage boundary of the press-fitting process, ensuring the consistency of the data for subsequent analysis, avoiding analysis errors caused by window fragmentation or mixing of data from different stages, and improving the accuracy of stage division.

[0012] Preferably, the adjustment of the preset window size to make the standard deviation of the correlation coefficient in the preset window satisfy the differentiation threshold value in the same press-fitting stage comprises: when the standard deviation of the preset correlation coefficient at the current time point is less than or equal to the differentiation threshold value, increasing the preset window length by a preset step; and when the standard deviation of the preset correlation coefficient at the current time point is greater than the differentiation threshold value, decreasing the preset window length by a preset step.

[0013] The present application forms a closed-loop feedback regulation mechanism by setting to increase or decrease the window length when the standard deviation of the correlation coefficient is less than or greater than the threshold value. Compared with the fuzzy adjustment strategy, this explicit control logic makes the adaptive adjustment process of the target window more efficient and stable, and can quickly converge to the optimal window size that matches the current press-fitting stage.

[0014] Preferably, determining the distinguishing threshold in the same pressing stage based on the standard deviation set includes: arranging the data in the standard deviation set in ascending order; calculating the preset quantile of the sorted data using linear interpolation, and using the preset quantile as the distinguishing threshold.

[0015] Preferably, the cutoff distance satisfies the relationship: ;in, It is The cutoff distance corresponding to the moment; is the preset global cutoff distance; It is In the target window at the moment, Moment and The Euclidean distance between the data points at the moment; It is The total number of data points in the target window at time instant; is the standard normalization function, It is the preset value adjustment coefficient.

[0016] The cutoff distance calculated by this invention is directly related to the sparsity of the data point distribution within the target window. This is the key step that this invention takes to address the core pain points of existing technologies. It enables the sensitivity of the clustering algorithm to automatically adjust based on whether the data is sparse or dense, fundamentally resolving the misjudgment problem caused by traditional fixed cutoff distances and significantly improving the anomaly identification accuracy of density clustering algorithms under variable density conditions.

[0017] Preferably, the cutoff distance is used as a parameter and a density clustering algorithm is used to perform cluster analysis on the data points in the target window to identify abnormal data points, including: calculating the local density of each data point based on the cutoff distance; determining the cluster center based on the local density and the relative distance between the data points; and identifying the data points that do not belong to any cluster after the cluster center is determined as abnormal data points.

[0018] Preferably, the multi-dimensional monitoring data at least includes: pressure data applied by the press and displacement data generated by the press head.

[0019] Preferably, the abnormal response action includes at least one of the following: triggering an audible and visual alarm device; controlling a switch valve to cut off the pressing process.

[0020] In a second aspect, the present invention provides a bearing press control system, which includes a memory and a processor. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a bearing press control method according to the first aspect of the present invention is implemented.

[0021] By adopting the technical scheme, the bearing press fitting machine control method of the first aspect of the present application is generated into a computer program and stored in a memory to be loaded and executed by a processor, so that a terminal device is manufactured according to the memory and the processor, and use is facilitated.

[0022] The present application has the following advantages: The present application can dynamically adjust the target window length by analyzing the correlation stability of the local bias degree sequence, ensure that the data in the segment is in the same press fitting stage, effectively avoid cross-stage data interference, and thus improve the state quantization precision. The present application adaptively adjusts the clustering cutoff distance by quantizing the sparsity of the current segment data point distribution. The neighborhood is expanded in the idle stroke sparse area and reduced in the pressure maintaining dense area, which significantly improves the abnormal state recognition accuracy of the density peak clustering in the variable density working condition. The present application feeds back the abnormal recognition result to the control mechanism to realize adaptive control of the device. When an abnormal state is recognized, control processing can be immediately taken to ensure the quality of the device and the product. The present application avoids the misjudgment of normal low-density data as an abnormal point by the traditional method, reduces false alarms, and improves production efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 A flowchart of a bearing press fitting machine control method provided for an embodiment of the present application is shown in the figure. Figure 2 A structural block diagram of a bearing press fitting machine control system provided for an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0024] The first aspect of the embodiment of the present application provides a bearing press fitting machine control method, as shown in the figure, which comprises steps S100-S500. Figure 1 Step S100, acquiring multi-dimensional monitoring data of the bearing press fitting machine at multiple moments in the press fitting process.

[0025] It should be noted that during the working process of the bearing press fitting machine, the changes of parameters such as pressure and displacement directly reflect whether the press fitting is smooth, whether there are problems such as jamming or unbalanced load, etc. Therefore, in order to realize fine control, high-frequency data acquisition needs to be performed on the press fitting process so as to capture transient changes.

[0026] Specifically, the state values at the current moment in multiple monitoring dimensions are collected in real time through sensors integrated on the press fitting head and the hydraulic cylinder. The monitoring dimensions specifically include: the main axial pressure applied by the press fitting machine can be acquired by a pressure sensor, and the displacement of the press head relative to the initial position can be acquired by a displacement sensor or an encoder. In order to ensure that transient changes in the press fitting process, such as pressure or displacement mutations caused by jamming, unbalanced load, etc., are captured, data acquisition is performed at a preset acquisition frequency. ​

[0027] As a preferred embodiment, the preset acquisition frequency can be set to acquire data once every 0.1 to 0.3 seconds. When the acquisition frequency is lower than once every 0.1 second, it may cause transient abnormal data to be smoothed or missed, reducing the timeliness of abnormal identification; when the acquisition frequency is higher than once every 0.3 second, it will significantly increase the data processing burden and storage requirements, causing redundant computing power. Therefore, controlling the acquisition frequency to once every 0.2 second can ensure data precision while taking into account system response speed and computing efficiency.

[0028] At this point, the multi-dimensional monitoring data of the bearing press-fitting machine at multiple moments in the press-fitting process is obtained.

[0029] Step S200, for the monitoring data of the current moment, based on the difference between the monitoring data of the current moment and the monitoring data of each moment in the preset window, determine the deviation degree of each moment in the preset window relative to the current moment, to construct the deviation degree sequence of each dimension of the current moment.

[0030] It should be noted that the normal bearing press-fitting process has obvious stage characteristics, for example, the idle stroke stage, the pressurization stage and the pressure maintaining stage. The data distribution characteristics of these stages are significantly different. In order to accurately quantify the current state, the change degree of the current state data relative to the recent history data needs to be obtained. By constructing the deviation degree sequence, the change of the current state can be more accurately quantified, thereby providing a basis for subsequent stage division and abnormal identification.

[0031] Specifically, the current moment is taken as the starting point and a preset number of moments are sampled forward as the preset window of the current moment. For example, taking the current moment as the starting point, sampling 20 moments as the preset window of the current moment. The data points can be preset to 20 moments. If the number of data points in the window is insufficient due to insufficient data, mean value or known data can be used for supplementary processing.

[0032] Further, for the preset window of the moment, the deviation degree of the moment in the dimension is calculated , and the deviation degree satisfies the relationship: ; Wherein, is the deviation degree of the moment in the dimension in the preset window of the moment; is the monitoring data value of the moment in the dimension; is the preset window of the time point is the maximum value of the monitoring data value in the preset window of the time point is the maximum value of the monitoring data value in the preset window of the time point dimension monitoring data value; is the preset window of the time point is the maximum value of the monitoring data value in the preset window of the time point is the maximum value of the monitoring data value in the preset window of the time point is the maximum value function; is the time difference between the time point is the time difference between the time point is the time difference between the time point is the total number of data points in the preset window, is the absolute value symbol; is the preset first infinitesimal value, used to prevent is 0, which can be set to 0.01, or can be set according to requirements; is the preset second infinitesimal value, used to prevent is 0, which can be set to 0.01, or can be set according to requirements.

[0033] In the formula, reflects the difference in value between the current time and other times in the preset window. The larger the value, the more obvious the state deviation. The second part reflects the time sequence proximity. Through weighted processing, the closer to the current time point, the higher the weight, thereby reducing the interference of data in different stages on the deviation degree calculation, and more accurately quantifying the deviation degree of the current time relative to each time in the window.

[0034] Repeat the above calculation to obtain the deviation degree of the current time and each time in the preset window in each dimension, and construct a deviation degree sequence of each dimension in the preset window of the current time. The data in the deviation degree sequence is arranged in ascending order according to the collection time.

[0035] Thus, the deviation degree sequence of each dimension of the current time is obtained.

[0036] Step S300, based on the correlation between the deviation degree sequences of each dimension of the current time in the preset window, adjust the window length of the preset window to obtain a target window in the same pressing stage as the current time.

[0037] It should be noted that the phase characteristics of the pressing process result in different data change patterns in each stage. If a fixed length window is used for analysis, the data in the same stage may be split or mixed with information in different stages, thereby affecting the accuracy of the analysis. By introducing correlation analysis of the deviation degree sequences of each dimension in the window, the characteristics of different stages can be distinguished, and the window length can be adaptively adjusted according to the consistency of the correlation to ensure that the data in the window belongs to the same pressing stage.

[0038] Specifically, first, the correlation coefficient of the deviation degree sequence of each dimension at the current time in the preset window is calculated, for example, the correlation coefficient of the deviation degree sequence of the two dimensions of pressure and displacement. For parameter changes in the same pressing stage, the correlation coefficient in the deviation degree sequence should be relatively concentrated. This step uses the Pearson correlation coefficient to obtain the correlation coefficient of the deviation degree sequence of the two dimensions in the segment to which the current time belongs. The Pearson correlation coefficient is a prior art and will not be described here.

[0039] Secondly, the standard deviation of the correlation coefficient of the target window corresponding to each historical time in the entire pressing history process at the current time is obtained to form a standard deviation set.

[0040] Then, the data in the standard deviation set is arranged in ascending order, and the preset quantile is calculated by using the linear interpolation method, and the quantile is taken as the threshold for distinguishing the same pressing stage. The preset quantile can be set to 10% quantile, or it can be set according to requirements.

[0041] Finally, the length of the preset window is adjusted so that the standard deviation of the correlation coefficient in the target window meets the threshold for distinguishing.

[0042] The specific adjustment process is: when the standard deviation of the correlation coefficient of the current window is less than or equal to the threshold for distinguishing, the length of the target window is increased by a preset step, for example, 1; when it is greater than the threshold for distinguishing, the length of the target window is decreased by a preset step. Repeat this process until the condition is met, and the data points in the finally obtained target window are considered to be in the same pressing stage.

[0043] At this point, the target window in the same pressing stage as the current time is obtained, and the possibility of misjudgment caused by the difference between the data across stages is eliminated.

[0044] Step S400, according to the sparsity of the data point distribution of the monitoring data in the target window, determine the cut-off distance for clustering analysis.

[0045] It should be noted that although the data has been divided into the same pressing stage, there are still differences in the density characteristics of the data in different stages, for example, the data in the empty stroke stage is sparse, and the data in the pressure maintaining stage is dense. The density peak clustering algorithm is an efficient clustering method, which can quickly and accurately cluster and divide the data by calculating the local density and relative distance of the data points. Its advantage is that it can effectively find the natural clustering structure in the data without pre-setting the number of clusters, which provides strong support for identifying abnormal points in the data.

[0046] However, the traditional density peak clustering algorithm adopts a globally fixed cut-off distance, which makes it difficult to adapt to the bearing press-fitting scenario with variable density conditions. Therefore, it is necessary to adaptively adjust the cut-off distance according to the sparsity of each stage after stage division to improve the recognition ability of the clustering algorithm to the real state.

[0047] Specifically, first, normalize all dimensional data at all times in the target window. Then, use the normalized data to construct data points, and calculate the mean of the Euclidean distance between any two data points in the target window to obtain the sparsity of the data points in the target window. According to this sparsity, the preset global cut-off distance is adaptively adjusted to obtain the cut-off distance corresponding to the current time, which satisfies the relationship: ; Wherein, is the cut-off distance corresponding to the time; is the preset global cut-off distance; is the target window at the time, the time and the time data points between the Euclidean distance; is the total number of data points of the target window at the time; is a standard normalization function, is a preset value adjustment coefficient.

[0048] In this formula, represents the mean of the Euclidean distance between any two data points in the target window. The larger this value is, the sparser the data distribution is, which may belong to the initial idle stroke stage, at which time the cut-off distance needs to be increased to ensure that there are enough data points in the neighborhood; otherwise, the smaller the value is, the denser the data distribution is, which may belong to the pressure maintaining stage, at which time the cut-off distance needs to be reduced to improve the accuracy of anomaly recognition.

[0049] It should be noted that, is a standard normalization function, which is used to quantize to the interval , which can be implemented by, for example, minimum-maximum normalization, Z-score standardization, etc. All of them are prior art and will not be described here.

[0050] It should also be noted that, is a value adjustment coefficient for , which sets to 0.5 to adjust the value of the value to the interval , so that the calculated On the basis of the value, the bidirectional optimization of increase and decrease is realized. is a preset global truncation distance, which can be set to 0.5, or can be set according to requirements.

[0051] At this point, the truncation distance used for clustering analysis at each time is obtained, which is referred to as an adaptive truncation distance hereinafter.

[0052] In step S500, the truncation distance is taken as a parameter, and a density clustering algorithm is used to perform clustering analysis on the data points in the target window to identify abnormal data points; when the monitoring data at the current time is identified as the abnormal data points, a preset abnormal response action of the bearing press-fit machine is controlled to be performed.

[0053] It should be noted that this step is a core link for identifying the abnormal state of the press-fit process. The adaptive truncation distance obtained in step S400 is used to replace the fixed truncation distance in the traditional density peak clustering algorithm in this step, so as to improve the accuracy of the abnormal state identification of the algorithm under the variable density condition. The traditional density peak clustering algorithm uses a fixed truncation distance, which is easy to misjudge the data points in the normal sparse area as abnormal under the variable density condition. By using the adaptive truncation distance, this problem can be effectively solved, thereby improving the accuracy of the clustering algorithm for identifying the abnormal state of the press-fit process.

[0054] Specifically, the data points in the target window corresponding to the current time and its historical times are taken as clustering samples. The adaptive truncation distance at each time is used to replace the fixed truncation distance in the traditional density peak clustering algorithm to perform clustering.

[0055] Firstly, the local density of each data point is calculated based on the adaptive truncation distance; then, the clustering center is determined based on the local density and the relative distance between the data points; finally, the data points not belonging to any cluster are identified as abnormal data points. The clustering process of the traditional density peak clustering algorithm is prior art, and the present application only modifies the truncation distance. Therefore, the detailed clustering process will not be described here.

[0056] If the monitoring data at the current time is identified as abnormal, it indicates that the bearing press-fit machine is in an abnormal running state. At this time, the corresponding audible and visual alarm device is immediately controlled to issue an alarm, and the corresponding on-off valve is immediately controlled to cut off the press-fit process, so as to prevent damage to the workpiece or equipment.

[0057] The second aspect of the embodiment provides a bearing press-fit machine control system, as shown in Figure 2 The bearing press-fit machine control system includes a memory and a processor, and the memory stores computer program instructions, which realize the bearing press-fit machine control method of the first aspect of the present application when executed by the processor.

[0058] The bearing press-fit machine control system also includes other components well known to those skilled in the art such as a communications bus and communications interface, the arrangement and function of which are known in the art and thus will not be described herein.

[0059] In this application, the aforementioned memory can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory, a dynamic random access memory, a static random access memory, an enhanced dynamic random access memory, a high bandwidth memory, a hybrid memory cube, or the like, or any other medium that can be used to store the desired information and that can be accessed by an application, module, or both. Any such computer storage media can be part of a device or accessible or connectable thereto.

[0060] The above are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, therefore: any equivalent changes made in the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A bearing press control method, characterized in that: Including steps: Obtain multi-dimensional monitoring data of the bearing press at multiple moments during the press-fitting process; For the monitoring data at the current moment, based on the difference between the monitoring data at the current moment and the monitoring data at each moment in the preset window, determine the deviation degree of each moment in the preset window relative to the current moment, so as to construct a deviation degree sequence for each dimension at the current moment; Based on the correlation between the deviation degree sequences of each dimension at the current moment within the preset window, adjusting the window length of the preset window to obtain a target window at the same pressing stage as the current moment; Determining a cutoff distance for cluster analysis based on the sparsity of data point distribution of the monitoring data within the target window; Using the cutoff distance as a parameter, cluster analysis is performed on the data points within the target window using a density clustering algorithm to identify abnormal data points; When the monitoring data at the current moment is identified as the abnormal data point, the bearing press is controlled to execute a preset abnormal response action.

2. The bearing press control method according to claim 1, characterized in that: The deviation of each moment in the preset window relative to the current moment satisfies the relationship: ； in, It is In the preset window of time, At the moment Deviation of the dimension; It is At the moment Dimension monitoring data values; It is In the preset window of time, At the moment Dimension monitoring data values; It is In the preset window of time, The maximum value of the dimension monitoring data value; is the maximum function; It is At the moment and within its preset window, The time difference between the moments; is the total number of data points in the preset window; is the absolute value symbol; is the preset first minimum value; It is the preset second smallest value.

3. The bearing press control method according to claim 1, characterized in that: The adjusting the window length of the preset window based on the correlation between the deviation degree sequences at each moment in the preset window to obtain a target window in the same pressing stage as the current moment includes: Calculating the correlation coefficient of the deviation degree sequence of each dimension in the preset window; Obtaining the standard deviation of the correlation coefficient of the target window corresponding to each historical moment in the historical pressing process of the pressed workpiece at the current moment, and forming a standard deviation set; Determining a distinction threshold at the same pressing stage based on the standard deviation set; The target window is obtained by adjusting the preset window size so that the standard deviation of the correlation coefficient within the preset window meets the distinction threshold of the same pressing stage.

4. The bearing press control method according to claim 3, characterized in that: The adjusting the preset window size so that the standard deviation of the correlation coefficient within the preset window meets the discrimination threshold of the same pressing stage includes: When the standard deviation of the preset correlation coefficient at the current moment is less than or equal to the discrimination threshold, the preset window length is increased by a preset step size; When the standard deviation of the preset correlation coefficient at the current moment is greater than the distinction threshold, the preset window length is reduced by a preset step size.

5. The bearing press control method according to claim 3, characterized in that: The determining of a distinction threshold at the same pressing stage based on the standard deviation set includes: Arrange the data in the standard deviation set in ascending order; The preset quantile of the sorted data is calculated using linear interpolation, and the preset quantile is used as the discrimination threshold.

6. The bearing press control method according to claim 1, characterized in that: The cutoff distance satisfies the relationship: ; in, It is The cutoff distance corresponding to the moment; is the preset global cutoff distance; It is In the target window at the moment, Moment and The Euclidean distance between the data points at the moment; It is The total number of data points in the target window at time instant; is the standard normalization function, It is the preset value adjustment coefficient.

7. The bearing press control method according to claim 1, characterized in that: The method of using the cutoff distance as a parameter and performing cluster analysis on the data points in the target window by a density clustering algorithm to identify abnormal data points includes: Calculating the local density of each data point based on the cutoff distance; Determining cluster centers based on the local density and relative distances between data points; After the cluster center is determined, data points that do not belong to any cluster are identified as abnormal data points.

8. The bearing press control method according to claim 1, characterized in that: The multi-dimensional monitoring data includes at least: pressure data applied by the press and displacement data generated by the press head.

9. The bearing press control method according to claim 1, characterized in that: The controlling the bearing press to execute a preset abnormal response action includes: Triggering the sound and light alarm device; Control the on-off valve to cut off the pressing process.

10. A bearing press control system, characterized in that: The bearing press control system includes a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, a bearing press control method according to any one of claims 1 to 9 is implemented.

Images