TWI741629B - Machine spindle running-in pre-checking method and computer-readable medium - Google Patents

Abstract

The invention discloses a machine spindle running-in pre-checking method. First, collect measurement parameters by an agent module when a machine spindle running-in, and then uses a pre-processing technology to filter out invalid raw data of the measurement parameters by a data pre-processing analysis module. Secondly, use a machine learning technology to establish a spindle run-in pre-checking model, and uses the spindle run-in pre-checking model to analyze valid initial data in the measurement parameters, and then perform abnormal detection of the machine spindle during a machine spindle running-in test to provide abnormal spindle diagnosis results. Simultaneously, synchronous reference quality control score results when the machine spindle running-in and feedback the quality control score results to the spindle run-in pre-checking model to correct pre-checking parameters. The present invention also provides a computer-readable medium for a machine spindle running-in pre-checking method.

Description

Pre-inspection method for machine main shaft running-in and computer readable medium

The invention relates to a machine spindle running-in pre-inspection technology, in particular to a machine spindle running-in pre-inspection method and computer readable media.

The machine spindle is the spindle of precision machines (such as machine tools) and similar equipment, which guarantees the precision and performance of the precision machines. At the same time, after the machine spindle is assembled, it needs to undergo a running-in test of the machine spindle to detect the running status of the machine spindle, such as operating efficiency, speed, acceleration, temperature, running-in time, etc.

However, it is known that when the running-in test of the machine spindle encounters abnormal problems, it usually continues until the end of the test (usually 24-120 hours), and then relies on manual operation experience to determine and return to the manufacturing process for reorganization, inspection and retesting. As a result, production costs are increased and shipment schedules are affected.

Therefore, how to provide a novel or innovative machine spindle running-in pre-inspection technology to more effectively or quickly solve the abnormal problem of the machine spindle running-in test has become a major research topic for those skilled in the art.

The present invention provides a novel or innovative machine spindle running-in pre-inspection method and computer readable medium, which can quickly or effectively solve the abnormal problem of the machine spindle running-in test.

The machine spindle running-in pre-inspection method of the present invention includes: collecting the measurement parameters of the machine spindle running-in through an agent module, so that the data pre-processing analysis module uses the pre-processing technology to filter out the agent module The invalid original data in the collected measurement parameters; and the use of machine learning technology to establish a spindle running-in pre-inspection model to analyze the effective initial data in the measured parameters by using the spindle running-in pre-inspection model established by machine learning technology, According to the effective initial data analyzed, the abnormality detection of the machine spindle during the running-in test of the machine spindle is performed, and then the abnormal spindle diagnosis result is provided.

In one embodiment, when the abnormality detection is performed on the machine spindle to provide the abnormal spindle diagnosis result, the quality control scoring result during the running-in of the machine spindle is synchronously referred to, so as to feed back the quality control scoring result to the spindle running-in pre-check The model is used to calibrate the pre-inspection parameters of the spindle running-in pre-inspection model according to the results of the quality control score.

The present invention further provides a computer-readable medium used in a computing device or a computer, which stores instructions to execute the aforementioned method for pre-checking the machine spindle running-in.

In order to make the above-mentioned features and advantages of the present invention more comprehensible, embodiments are specifically described below in conjunction with the accompanying drawings. In the following description, the additional features and advantages of the present invention will be partially explained, and these features and advantages will be partly known from the description, or can be learned by practicing the present invention. It should be understood that both the foregoing general description and the following detailed description are only illustrative and explanatory, and are not intended to limit the scope of the present invention.

1: Machine spindle running-in device

10: Machine spindle

11: front axle

12: rear axle

21: Speed sensor

22: Accelerometer sensor

23: Eddy current sensor

24: Front axle temperature sensor

25: Rear axle temperature sensor

31: Pulley

32: belt

33: Motor

S10 to S20, S21 to S24, S41 to S45: steps

S51 to S55, S61 to S65, S71 to S76: steps

Figure 1 is a schematic diagram of the pre-inspection process of the spindle running-in in the method of the machine spindle running-in pre-inspection of the present invention;

Figure 2 is a schematic diagram of another process of pre-checking the spindle running-in in the method for pre-checking the machine spindle running-in according to the present invention;

Figure 3 is a schematic diagram of the machine spindle running-in device and measurement parameters in the machine spindle running-in pre-inspection method of the present invention;

Figure 4 is a flow diagram of the analysis of the correlation between spindle running-in sensor information and pre-check results in the machine spindle running-in pre-check method of the present invention;

Figure 5 is a schematic diagram of the pre-inspection process of the spindle running-in performance in the machine spindle running-in pre-inspection method of the present invention;

Figure 6 is a schematic flow diagram of the predictive spindle quality diagnosis in the machine spindle running-in pre-inspection method of the present invention; and

FIG. 7 is a schematic diagram of the flow chart of the prediction of the spindle running-in time in the machine spindle running-in pre-inspection method of the present invention.

The following describes the implementation of the present invention with specific specific embodiments. Those familiar with this technology can understand the other advantages and effects of the present invention from the contents disclosed in this specification, and can also implement other different specific equivalent embodiments. Or use.

The machine spindle running-in pre-inspection method of the present invention collects the measurement parameters of the machine spindle running-in through the agent module, and then the data pre-processing analysis module uses the pre-processing technology (such as data Quality indicator detection, principal component analysis, etc.) to filter out invalid raw data in the measurement parameters collected by the agent module. Next, use machine learning technology to establish a spindle running-in pre-inspection model, and use the spindle running-in pre-inspection model established by machine learning technology to analyze the effective initial data (effective measurement parameters) that have been analyzed. Valid initial data (valid measurement parameters) perform abnormality detection on the machine spindle during the test of the machine spindle running-in, and then provide the abnormal spindle diagnosis result. At the same time, when performing abnormality detection on the machine spindle to provide abnormal spindle diagnosis results, in order to strengthen the credibility of the spindle running-in pre-check model, you can synchronously refer to the quality control (QC) score during the machine spindle running-in. As a result, the quality control scoring results are fed back to the spindle running-in pre-inspection model, so as to correct the pre-inspection parameters of the spindle running-in pre-inspection model based on the quality control scoring results. Therefore, the present invention can solve the conventional problem that when the running-in test of the machine spindle encounters an abnormal problem, it usually continues until the end of the test (usually 24-120 hours), and then relies on manual operation experience to determine and return to the process for reorganization and reorganization. Inspection and re-testing, thus increasing production costs and affecting the delivery schedule. In addition, the method of the present invention for predicting the diagnosis result of spindle quality during the running-in period of the machine spindle can reduce the running-in test time and manual detection cost of abnormal spindles by more than 80%, thereby solving the machine spindle running-in more effectively or quickly. Abnormal problems during testing.

The machine spindle running-in pre-inspection method of the present invention may include the analysis of the relationship between the spindle running-in sensor information and the pre-inspection result, the spindle running-in pre-inspection parameter correction learning, the running efficiency prediction of the spindle running-in test, and the spindle running time Methods such as predictive and predictive spindle quality diagnosis, and the technical content of these methods are as follows.

The analysis method of the correlation between the information of the spindle running-in sensor and the pre-check results includes: (1) The running-in test of the abnormal spindle is used to collect the measurement parameters of the spindle. The abnormal spindle can include bearing damage, insufficient lubrication, assembly defects, and Poor balance, misalignment of the shaft, bent shaft, loose bearing screws Dynamic, gear damage, oil shock, rotor imbalance, etc., and the measurement parameters can include spindle speed, accelerometer (such as X, Y, Z direction acceleration in the time domain and dB value in the frequency domain), eddy current (such as The displacement of the front or rear axle in the X, Y, and Z directions), front axle temperature, rear axle temperature, ambient temperature, start time, end time, etc. (2) Establish correlation data between sensor information (such as speed, vibration spectrum, temperature...) and maintenance/fault prediction. (3) Establish an initial spindle running-in pre-check model through the supervised learning method to predict the abnormal state that may occur during the spindle running-in or the completion time of the normal spindle running-in, and this supervised learning method can be used It is an artificial intelligence (AI) supervised learning method. Therefore, the present invention can greatly reduce the running-in test time of the main shaft, so as to significantly solve the technical problem that the running-in test of the main shaft is time-consuming for a long time and needs to be detected by professional personnel based on experience.

The spindle running-in pre-inspection parameter correction learning method includes: (1) Based on the correlation analysis between the sensor information and the pre-inspection result during the spindle running-in, to establish the initial label (such as _Yo , the cause of the failure in the judgment result) or normal) sample (such as X _i, sensor information) repository. (2) Establish an initial pre-check model of spindle running-in through the supervised learning method. (3) Collect the sensor parameters (such as X _i , samples of pre-test test) when the machine spindle of the production line is running in. _{(4) The pre-inspection result (such as Yo} ) is determined according to the sensor parameters through the spindle running-in pre-inspection model. (5) After the pre-inspection result is determined by the quality control unit of the production line, a set of samples (such as X _{i ) with labels (such as Yo} ) can be added to the database. Therefore, the present invention can calibrate the pre-inspection parameters of the spindle running-in pre-inspection model by accumulating the number of labeled samples, so as to continuously strengthen the accuracy of the spindle running-in pre-inspection model, and shorten the quality control inspection time and speed up abnormal maintenance.

The running efficiency prediction method of the spindle running-in test includes: (1) Establishing the running efficiency level scoring mechanism of the spindle running-in result, such as the level of normal and possible abnormal causes. (2) establish an initial label (such as Y _o, performance scoring) of the sample (such as X _i, sensor information) repository with the spindle running-in preflight model. (3) Collect the sensor parameters (such as X _i , samples of the pre-test test) during the running-in of the machine spindle of the production line, and use the sensor parameters to score the operating efficiency level through the pre-test model of the spindle running-in. (4) Add a set of samples (such as X _i ) _{with tags (such as Yo} ) to the database based on the quality control scoring results of the machine spindle running in the production line. (5) Feed back the quality control scoring results to the spindle running-in pre-inspection model to correct the pre-inspection parameters of the spindle running-in pre-inspection model based on the quality control scoring results. Therefore, the present invention can achieve the function of scoring the operating efficiency of the spindle running-in test, so as to solve the problem of quality control over the operating efficiency level of the spindle test, and can be used as a reference for machine predictive maintenance and overhaul of after-sales service.

The spindle running-in time prediction method includes: (1) Continuous collection of sensor information (such as speed, acceleration gauge, front axle temperature, rear axle temperature, etc.) and running time of the spindle running-in test of the production line. Time prediction model sample (such as X _i, sensor information) (2) to establish a label (such as Y _o, running-time) of the spindle run-in database. (3) Continuously collect the spindle running-in test results of the production line to continuously increase the effective number of samples, and then feed back the effective sample number to the spindle running-in time prediction model to calibrate the spindle running-in pre-check model based on the effective sample number The preflight parameters. Therefore, the present invention can achieve the time prediction function of the spindle running-in test of the production line, and the abnormal spindle (for example, the test exceeds 120 hours) can be scheduled for maintenance in advance to solve the conventionally long time and unpredictable test time (24-120 hours). )The problem.

Fig. 1 is a schematic diagram of the flow of pre-checking the spindle running-in in the method for pre-checking the machine spindle running-in according to the present invention. As shown in the figure, by collecting the number of samples (such as measurement parameters) with labels (such as pre-check results), etc. (such as three) such as pre-checks for operating performance, pre-checks for quality diagnosis, and prediction of running-in time, in the database (see section Step S10 of Figure 1), where the same sample can correspond to the labels of multiple (such as three) spindle running-in pre-check models, and then use the supervised learning method to establish multiple (such as three) spindle running-in pre-check models, such as spindle running preflight model may be bonded spindle running-performance preflight operation model A (such as _{X i → Y o, 1,} → represents correspondence), the diagnostic quality of the running spindle preflight model B (such as _{X i → Y o, 2)} , The spindle running-in time prediction model C (such as X _i →Y _o,3 ).

Take the process of pre-checking the running-in performance of the spindle as an example. Initially select the pre-check model A of running-in running performance of the spindle (see step S11 in Figure 1), and collect the measurement parameters of the machine spindle running-in through the agent module. (See step S12 in Figure 1), where the measurement parameters (such as X) include the spindle speed (rpm), accelerometer (time domain or frequency domain), and eddy current recorded at intervals (such as 5 seconds) (Displacement), front axle temperature, rear axle temperature, ambient temperature, start time or end time of the main shaft running-in (see Figure 3 or Figure 4).

Then, the data pre-processing analysis module uses pre-processing techniques (such as data quality indicator detection, principal component analysis, etc.) on the collected measurement parameters to filter out invalid primitives in the measurement parameters collected by the agent module Data (see step S13 in Figure 1). Then, the spindle running-in operation performance pre-check model A is analyzed through the analyzed effective initial data (effective measurement parameters) collected in a fixed period of time (such as 30 minutes) to obtain the effective initial data (effective Find the corresponding spindle running-in pre-inspection result or running performance pre-inspection result (such as _Yo , label) in the measurement parameters) (see step S14 to step S15 in Figure 1).

If the running-in time of the machine spindle (such as T) is greater than or equal to the time threshold (such as T≧72 hours) (see step S16 in Figure 1), the running-in test is stopped and the spindle running-in pre-check result is output (such as Y _o ) and accuracy (see step S17 in Figure 1), and then based on the quality control scoring result or manual judgment result (see step S18 in Figure 1), a set of labeled samples can be added to the database.

In addition, if the running-in time of the machine spindle (such as T) is less than the time threshold (such as T<72 hours), the accuracy of the pre-check results of the spindle running-in will be scored independently (0-100%). And if the accuracy is greater than or equal to the accuracy threshold (such as accuracy ≧90%) (see step S19 in Figure 1), the spindle running-in pre-check result (such as _Yo ) and accuracy (see Figure 1 Step S20), and add a set of labeled samples to the database to increase the accuracy of the newly created spindle running-in pre-check model; conversely, if the accuracy is less than the accuracy threshold (such as accuracy <90 %), then return to the collection of measurement parameters (see step S11 in Figure 1), and continue to accumulate the measurement parameters to improve the accuracy of the pre-check.

Fig. 2 is a schematic diagram of another process of the pre-checking of the running-in of the spindle in the method of pre-checking the running-in of the machine spindle of the present invention. As shown in the figure, the machine spindle running-in is performed first (see step S21 in Figure 2), and then the measurement parameters of the machine spindle running-in are collected through the proxy module (see step S22 in Figure 2). Then, through a wired or wireless network (such as an intranet), the measurement parameters are sent to the data pre-processing analysis module in the core module, and the data pre-processing analysis module uses pre-processing technology (such as Data quality index detection, principal component analysis, etc.) to filter out the invalid original data in the measurement parameters collected by the agent module, and then use the spindle running-in pre-check model in the core module to analyze the effective initial data ( Effective measurement parameters) for analysis, in order to perform anomaly detection on the machine spindle during the test of the machine spindle running-in based on the analyzed effective initial data (effective measurement parameters), and then provide abnormal spindle diagnosis results (see section Step S23 in Figure 2). At the same time, in order to strengthen the credibility of the pre-inspection model of the spindle running-in, you can synchronously refer to the quality control scoring results (such as manual judgment results or manual quality inspection results) during the spindle running-in of the machine (see step S24 in Figure 2). Then the quality control score results are fed back to the spindle running-in pre-inspection model, so as to correct the pre-inspection parameters of the spindle running-in pre-inspection model based on the quality control scoring results.

Figure 3 is a schematic diagram of the machine spindle running-in device 1 and the measurement parameters in the machine spindle running-in pre-check method of the present invention. As shown in the figure, the machine spindle running-in pre-inspection method can perform the machine spindle running-in test on the machine spindle running-in device 1, and the machine spindle running-in device 1 may include the machine spindle 10 (main body) and the speed Sensor 21, accelerometer sensor 22, eddy current sensor 23, front axle temperature sensor 24, rear axle temperature sensor 25, ambient temperature sensor (not shown), Timer (not shown), pulley 31, belt 32, motor 33, etc.

The machine main shaft 10 can have a front shaft 11 and a rear shaft 12, the speed sensor 21 can be connected to the rear shaft 12 (pulley 31) of the machine main shaft 10 to sense the speed of the main shaft, and the accelerometer sensor 22 can be connected to the machine The front shaft 11 or the rear shaft 12 of the main shaft 10 can sense accelerometers, the eddy current sensor 23 can be connected to the rear shaft 12 of the machine main shaft 10 to sense eddy current, and the front shaft temperature sensor 24 can be connected to the machine main shaft The front axle 11 of 10 can sense the temperature of the front axle. The rear axle temperature sensor 25 can be connected to the rear axle 12 of the machine main shaft 10 to sense the temperature of the rear axle. The ambient temperature sensor can be installed in the machine main shaft running-in device. 1 is used to sense the ambient temperature, and the timer can be used to count the start time and end time of the machine spindle 10 running in. Therefore, the measurement parameters of the machine spindle 10 during running-in may include the spindle speed (rpm), accelerometer (time domain or frequency domain), eddy current (displacement), front axle temperature, rear axle temperature, ambient temperature, Start time, end time, etc.

Figure 4 is a flow diagram of the analysis of the correlation between the spindle running-in sensor information and the pre-check result in the machine spindle running-in pre-check method of the present invention, including the spindle running-in pre-check parameter correction learning. As shown in the figure, the analysis method of the correlation between the spindle running-in sensor information and the pre-check result is to collect the spindle measurement parameters through the running-in test of the abnormal spindle (see step S41 to step S42 in Figure 4), and then establish Correlation data of sensor information (such as speed, vibration spectrum, temperature...) and maintenance/fault prediction. The above-mentioned abnormal spindle may include bearing damage, insufficient lubrication, assembly defects, poor dynamic balance, shaft misalignment, shaft bending, bearing screw looseness, oil shock, gear damage, etc. The measurement parameters of the spindle may include speed, accelerometer (such as Acceleration in the X, Y, and Z directions in the time domain and dB value in the frequency domain), eddy current (such as the displacement of the front or rear axle in the X, Y, and Z directions), front axle temperature, rear axle temperature, and ambient temperature , Start time, end time, etc., bearing damage can include inner ring damage, outer ring damage, rolling body damage, and gear damage can include overall wear and partial contact tooth profile Error, pitch error eccentricity.

Furthermore, through supervised learning methods (such as AI supervised learning methods), establish a pre-check model for spindle running-in (see step S43 in Figure 4) to predict the abnormal state or normal spindle running-in that may occur during spindle running-in operation (See step S44 in Figure 4), and the result of this spindle running-in can be matched with the independent score or through manual judgment (corresponding to the correct f: X→Y) (see step S45 in Figure 4), Add a group label (e.g., Y _o) of the sample (e.g., X _i) number, and further continued to strengthen the running spindle preflight model performance and accuracy through the number of samples is accumulated label. Therefore, the present invention can greatly reduce the running-in test time of the main shaft, so as to significantly solve the technical problem that the running-in test of the main shaft is time-consuming for a long time and needs to be detected by professional personnel based on experience.

The above-mentioned method for continuously enhancing the performance and accuracy of the spindle running-in pre-check model by accumulating the number of samples with labels is the spindle running-in pre-check parameter correction learning method, and the spindle running-in pre-check parameter correction learning method may include: (1) based on the running spindle connected with the sensor information analysis results preflight off, to establish an initial label (e.g., Y _o, or the cause of the malfunction determination result of normal) samples (e.g., X _i, sensor information) in the database. (2) Establish an initial pre-check model of spindle running-in through the supervised learning method. _{(3) Collect sensor parameters (such as X i} , samples for pre-testing) when the machine spindle of the production line is running in. _{(4) The pre-inspection result (such as Yo} ) is determined according to the sensor parameters through the spindle running-in pre-inspection model. (5) After the pre-inspection result is judged by the quality control unit of the production line, a set of samples (such as X _{i ) with labels (such as Yo} ) can be added to the database. Therefore, the present invention can calibrate the pre-inspection parameters of the spindle running-in pre-inspection model by accumulating the number of samples with tags, so as to continuously strengthen the accuracy of the spindle running-in pre-inspection model, and shorten the quality control inspection time and speed up abnormal maintenance.

Figure 5 is a schematic diagram of the pre-checking process of the spindle running-in performance in the machine spindle running-in pre-checking method of the present invention, which can identify normality and its performance scores, various abnormal results and various abnormal results during the pre-checking of the spindle running-in running performance. Its abnormal level. As shown in the figure, the process of pre-checking the running-in performance of the spindle includes: (1) Establishing the running performance level scoring mechanism of the running-in result of the spindle, such as the level of normal and possible abnormal causes (see step S51 in Figure 5), For example, the pre-check result (such as _Yo ) is normal or abnormal (such as abnormal -1, abnormal -2,..., abnormal -N, N is an integer), and the normal performance score or the abnormal abnormality level can be 1- 3 points. (2) establishing an initial label (e.g., Y _o, scoring performance) of a sample (e.g., X _i, sensor information) in the database together with the spindle running preflight model (see step S51 of FIG. 5 and step S53) . (3) Collect the sensor parameters (such as X _i , the sample of the pre-test test) during the running-in of the machine spindle of the production line (see step S52 in Figure 5), and use the sensor parameters to pass the pre-test of the spindle running-in The model is rated for operating efficiency level (see the pre-check result _{Yo, t} shown in step S54 in Figure 5). (4) Add a set of samples (such as X _{i ) with tags (such as Yo} ) to the database according to the quality control scoring results of the machine spindle running-in by the production line (see step S55 in Figure 5). (5) Feed back the quality control scoring result to the spindle running-in pre-inspection model to correct the pre-inspection parameters of the spindle running-in pre-inspection model based on the quality control scoring result (see step S53 in Figure 5). Therefore, the present invention can achieve the function of scoring the running performance of the spindle running-in test, so as to solve the problem of controlling the running performance level of the spindle test by the quality control, and can be used as a reference for machine predictive maintenance and overhaul of after-sales service.

Figure 6 is a schematic diagram of the process of predictive spindle quality diagnosis in the pre-inspection method for machine spindle running-in according to the present invention, which can identify normal and various abnormal results during the pre-inspection of the spindle. As shown, the running operation of the spindle is not easy to establish the effectiveness of the preflight model sufficient label (e.g., Y _o) of the sample (e.g., X _i) number (see the step of FIG. 6 S61), thereby easily reducing spindle run Because of the accuracy of the pre-check model for combined running efficiency, the pre-check model for spindle running-in performance can be simplified. The fast pre-check method for the cause of abnormal spindle running-in can only predict normal and possible abnormal causes, which can shorten the process inspection and Time to check and repair the cause of the abnormality.

The predictive spindle quality diagnosis method is to collect the measurement parameters (such as X _{i, t} ) when the machine spindle is running in through the agent module (see step S62 in Figure 6), so that the data pre-processing analysis module uses the pre-processing Technology (such as data quality indicator detection, principal component analysis, etc.) to filter out invalid raw data in the measurement parameters collected by the agent module. In addition, machine learning technology is used to establish a pre-check model of the spindle running-in (see step S63 in Figure 6), so as to detect the abnormality of the machine spindle during the test of the machine spindle running-in, and then provide abnormal spindle diagnosis results (see step S63). _{The pre-check result Yo,t} shown in step S64 in Fig. 6). At the same time, in order to strengthen the credibility of the spindle running-in pre-inspection model, the quality control (QC) scoring results of the machine spindle running-in can be synchronously referred to, and the quality control scoring results can be fed back to the spindle running-in pre-inspection model. The quality control scoring result corrects the pre-inspection parameters of the spindle running-in pre-inspection model (see step S65 in Figure 6). Therefore, the present invention can solve the conventional problem that when the running-in test of the main shaft of the machine encounters an abnormal problem, it usually continues until the end of the test (24-120 hours), and then relies on manual operation experience to determine and return to process reorganization and inspection and retest. , Thus increasing production costs and affecting the delivery schedule. In addition, the present invention can reduce the running-in test time and manual detection cost of abnormal spindles by more than 80%, so as to solve long-standing technical problems.

Figure 7 is a schematic diagram of the flow of the prediction of the spindle running-in time in the machine spindle running-in pre-inspection method of the present invention, which includes training multiple (for example, more than 100) running-in test samples of the same model. As shown in the figure, the method for predicting the running-in time of the spindle includes: (1) Continuous collection of sensor information (such as speed, acceleration gauge, front axle temperature, rear axle temperature, etc.) and running-in time of the spindle running-in test of the production line ; for example, measuring parameter as shown in step S71 of the spindle of FIG. 7 (e.g., X _i) and the running time preflight (e.g., Y _o), a step S72 shown in FIG. 7 of the running test checks to preflight testing For the sample (such as X _i ), the measured parameters of the spindle (such as X _{i, t} ) and the actual running-in time (such as T _t ) shown in step S73 in Fig. 7. Samples (e.g., X _i, sensor information) (2) to establish a label (e.g., Y _o, running-time) database with the spindle running-time prediction model to predict the running time (e.g., T _p), and then output the actual time, predicted time and the prediction error value; e.g., step by step S71 shown in FIG. 7 of the tag (e.g., Y _o) of the sample (e.g., X _i), the step of FIG. 7 spindle running-time prediction model S74, the Step S75 in Fig. 7 shows the predicted running-in time (such as T _p ), and step S76 in Fig. 7 outputs the actual time, predicted time, and predicted error value. (3) Continuously collect the spindle running-in test results of the production line to continuously increase the effective number of samples, and then feed the effective sample number back to the spindle running-in time prediction model to correct the spindle running-in pre-inspection model based on the effective sample number Pre-check parameters (see step S74 in Figure 7). Therefore, the present invention can achieve the time prediction function of the spindle running-in test of the production line, and the abnormal spindle (for example, the test exceeds 120 hours) can be scheduled for maintenance in advance to solve the conventionally long time and unpredictable test time (24-120 hours). )The problem.

In addition, the present invention also provides a computer-readable medium for the pre-check method for machine spindle running-in, which is applied to a computing device or computer with a processor (for example, CPU, GPU, etc.) and/or memory, and Instructions are stored, and the computing device or computer can be used to execute the computer-readable medium through the processor and/or memory, so as to execute the above content when the computer-readable medium is executed.

In summary, the method for pre-checking machine spindle running-in and the computer-readable medium of the present invention can at least have the following characteristics, advantages, or technical effects.

1. The present invention provides correlation analysis between spindle running-in sensor information and pre-check results, spindle running-in pre-check parameter correction learning, spindle running-in test operation performance prediction, spindle running-in time prediction, predictive spindle quality diagnosis, etc. The method can quickly or effectively solve the abnormal problem of the machine spindle running-in test. That is, the present invention can solve the conventional problem that when the running-in test of the main shaft of the machine encounters an abnormal problem, it usually continues until the end of the test (24-120 hours), and then relies on manual operation experience to determine and return to the process for reorganization and inspection. Re-test, thus increasing production costs and affecting output Delivery schedule.

2. The present invention collects the measurement parameters of the machine spindle during running-in through the agent module, and uses pre-processing technologies such as data quality index detection and principal component analysis to facilitate screening out of the measurement parameters collected by the agent module Invalid source data for.

3. The present invention uses machine learning technology to establish a spindle running-in pre-inspection model, so as to provide abnormal spindle diagnosis results during the testing period of the machine spindle running-in.

4. In order to strengthen the credibility of the spindle running-in pre-inspection model, the present invention can synchronously refer to the quality control (QC) scoring results and feed back the quality control scoring results to the spindle running-in pre-inspection model to facilitate the basis of the quality control scoring results Correct the pre-inspection parameters of the pre-inspection model of the spindle running-in.

5. The method of the present invention for predicting the diagnosis result of spindle quality during the running-in period of the machine spindle can reduce the running-in test time and manual detection cost of abnormal spindles by more than 80%.

6. The present invention can greatly reduce the running-in test time of the spindle, so as to significantly solve the technical problem that the running-in test of the spindle is time-consuming for a long time and requires professional labor to detect abnormal causes based on experience.

7. The present invention can correct the pre-inspection parameters of the spindle running-in pre-inspection model by accumulating the number of labeled samples, so as to continuously strengthen the accuracy of the spindle running-in pre-inspection model, and shorten the quality control inspection time and speed up abnormal inspection.

8. The present invention can achieve the function of scoring the running efficiency of the spindle running-in test, so as to solve the problem of quality control over the running efficiency level of the spindle test, and can be used as a reference for machine predictive maintenance and overhaul of after-sales service.

9. The present invention can achieve the time prediction function of the spindle running-in test of the production line, and the abnormal spindle (for example, the test exceeds 120 hours) can be scheduled for maintenance in advance to solve the conventionally long time and unpredictable test time (24-120 hours) )The problem.

The above embodiments are only illustrative of the principles, features and effects of the present invention, and are not intended to limit the scope of implementation of the present invention. Anyone familiar with the art can comment on the above without departing from the spirit and scope of the present invention. Modifications and changes to the implementation form. Any equivalent changes and modifications made using the content disclosed in the present invention should still be covered by the scope of the patent application. Therefore, the protection scope of the present invention should be as listed in the scope of the patent application.

S10 to S20: steps

Claims (12)

A pre-inspection method for machine spindle running-in includes: Collect the measurement parameters when the machine spindle is running through the proxy module, so that the data pre-processing analysis module uses pre-processing technology to filter out the invalid raw data in the measurement parameters collected by the proxy module; and Use machine learning technology to establish a spindle running-in pre-inspection model, so as to use the spindle running-in pre-inspection model established by the machine learning technology to analyze the effective initial data in the measured parameters, so as to rely on the analyzed effective initial data During the running-in test of the machine spindle, the machine spindle is abnormally detected, and then the abnormal spindle diagnosis result is provided. For example, the method for pre-checking the machine spindle running-in as described in item 1 of the scope of patent application, wherein, when the machine spindle is abnormally detected to provide the abnormal spindle diagnosis result, the machine spindle running-in time is synchronously referred to The quality control scoring result is used to feed back the quality control scoring result to the spindle running-in pre-inspection model, so as to correct the pre-inspection parameters of the spindle running-in pre-inspection model according to the quality control scoring result. For example, the machine spindle running-in pre-inspection method described in item 1 of the scope of patent application is to test the machine spindle running-in device on the machine spindle running-in device, wherein the machine spindle running-in device includes the machine The main shaft and the speed sensor, accelerometer sensor, eddy current sensor, front axle temperature sensor and rear axle temperature sensor connected to the main shaft of the machine. For example, the method for pre-checking the machine spindle running-in as described in item 1 of the scope of the patent application further includes the following steps of analyzing the correlation between the spindle running-in sensor information and the pre-check results: Collect the measurement parameters through the running-in test of the abnormal spindle; Establish correlation data between sensor information and maintenance/fault prediction; and Through the supervised learning method, the initial spindle running-in pre-check model is established to predict the abnormal state that may occur during the spindle running-in operation or the completion time of the normal spindle running-in. As described in item 1 of the scope of patent application, the method for pre-inspection of machine spindle running-in further includes the steps of performing the following steps of the method for the correction and learning of the pre-inspection parameters of the spindle running-in: Based on the analysis of the correlation between the sensor information and the pre-inspection results when the spindle of the machine is running, establish a database of the initial labeled samples; Establish the initial pre-check model of spindle running-in through supervised learning method; Collect the sensor parameters when the main shaft of the machine is running in; Determine the pre-inspection result based on the sensor parameters through the spindle running-in pre-inspection model; and After determining the pre-inspection result, add a set of labeled samples to the database. For example, the method for pre-inspection of machine spindle running-in described in item 1 of the scope of patent application further includes the steps of performing the following method for predicting the running performance of the spindle running-in test: Establish a scoring mechanism for the running efficiency level of the spindle running-in results; Establish the database of the initial labeled samples and the pre-check model of the spindle running-in; Collect the sensor parameters of the main shaft of the machine during running-in, so as to use the sensor parameters to score the running performance level through the pre-check model of the main shaft running-in; Add a set of labeled samples to the database based on the quality control scoring results of the spindle of the machine during the running of the production line; and The quality control scoring result is fed back to the spindle running-in pre-inspection model, so as to correct the pre-inspection parameters of the spindle running-in pre-inspection model according to the quality control scoring result. For example, the method for pre-checking machine spindle running-in described in item 1 of the scope of patent application further includes the following steps regarding the method for predicting the running-in time of the spindle: Continuously collect the sensor information and running time of the spindle running-in test; Establish a database of labeled samples and a prediction model for the running-in time of the spindle to predict the running-in time; and Continuously collect the spindle running-in test results of the production line to continuously increase the effective number of samples, and then feed the effective sample number back to the spindle running-in time prediction model to correct the spindle running-in pre-check model based on the effective sample number The preflight parameters. For example, the machine spindle running-in pre-inspection method described in item 1 of the scope of patent application further includes the use of supervised learning methods to separately establish multiple spindle running-in pre-inspection models, where the multiple spindle running-in pre-inspection models include spindle running Pre-inspection model of combined operation efficiency, pre-inspection model of spindle running-in quality diagnosis and prediction model of spindle running-in time. For example, the method for pre-checking machine spindle running-in described in item 1 of the scope of patent application further includes the following steps regarding the pre-checking method for spindle running-in operation efficiency: Initially select the pre-check model for the running-in performance of the main shaft; and Collect the measurement parameters when the spindle of the machine is running in through the agent module, Among them, the measurement parameters include the rotation speed, accelerometer, eddy current, front axle temperature, rear axle temperature, ambient temperature, and the start time or end time of the spindle running-in recorded at every interval. For example, the machine spindle running-in pre-inspection method described in item 2 of the scope of patent application further includes that if the running-in time of the machine spindle is greater than or equal to the time threshold, the running-in test is stopped and the spindle running-in pre-inspection result is output And accuracy, and then add a set of labeled samples to the database based on the quality control scoring results, and if the running-in time of the machine spindle is less than the time threshold, the accuracy of the pre-check results for the spindle running-in Carry out self-scoring. For example, the machine spindle running-in pre-inspection method described in item 10 of the scope of patent application also includes outputting the spindle running-in pre-inspection result and accuracy if the accuracy is greater than or equal to the accuracy threshold, and a new set is added The labeled samples are in the database, and if the accuracy is less than the accuracy threshold, continue to collect to accumulate the measurement parameters. A computer-readable medium used in a computing device or computer, and storing instructions to execute the method for pre-checking machine spindle running-in as described in one of items 1 to 11 of the scope of patent application.

Images