TW202001703A - System of machine tool processing behavior abnormal analysis and predictive maintenance and method thereof - Google Patents

Abstract

A system of machine tool processing behavior abnormal analysis and predictive maintenance and method thereof are disclosed. The method comprises: capturing a sensed value of a spindle or a key component of a machine by a machine sensor capturing module to calculate a time-domain vibration value and a frequency-domain vibration value, capturing an operation information of a spindle rotating speed, operating status or processing program of the machine by a machine controller capturing module; providing a warning for an abnormal time-domain vibration value or an abnormal frequency-domain vibration value of the spindle or key component and analyzing abnormal machining behavior or faulty component cause of the machine according the time-domain vibration value, frequency-domain vibration value, spindle rotating speed, operating status or machining program; and integrating information from the machine sensor capturing module and the machine controller capturing module to generate integrated information, and providing maintenance method and maintenance timing of the machine before abnormal machining behavior or component failure of the machine.

Description

System and method for analyzing and predicting abnormality of machine processing behavior

The present invention relates to a technique for analyzing machining behavior of a machine, and particularly to a system and method for analyzing and predicting and maintaining abnormal machining behavior of a machine.

In response to the trend of Industry 4.0 and the development of the smart machinery industry, how to reduce downtime and improve factory production capacity through fault prediction is the main focus of smart machinery. The practice of SCADA (Supervisory Control And Data Acquisition) system in the general industry mostly focuses on machine monitoring, such as machine information visualization, remote monitoring, alarm notification and other functions, but it lacks predictive maintenance of the machine technology.

Furthermore, the prior art proposes a method and system for monitoring the abnormal state of the machine. The monitoring end of the system connects a monitoring unit to the machine through an image unit or electrical, and the monitoring end performs monitoring on the machine when the machine is found. When an abnormality occurs in the station, an abnormal signal is sent to the central control terminal to determine that the machine is in an abnormal state. However, this prior art only discloses that the abnormality of the machine is determined through the image unit or the electrical connection, but does not further analyze the abnormality and cause of the machine.

Therefore, how to solve the shortcomings of the above-mentioned conventional technology has become a major issue for those skilled in the art.

The invention provides a system and method for abnormal analysis and prediction and maintenance of machine tool processing behavior, which can analyze the abnormal processing behavior of the machine tool or the cause of the faulty component, and provide the maintenance mode or maintenance timing of the machine tool.

The system for analyzing and predicting the abnormality of machine processing behavior in the present invention includes: a machine sensor acquisition module, which captures the sensor's sense value of the machine spindle or key components to calculate the spindle or key Time-domain vibration value and frequency-domain vibration value of the component; the machine controller acquisition module is used to acquire the machine spindle speed, operating status or operation information of the processing program through the controller; the processing behavior abnormal analysis module is According to the time-domain vibration value and frequency-domain vibration value of the spindle or key components calculated by the machine sensor acquisition module, and the machine spindle speed, operation status or machine tool acquisition module acquisition The operation information of the processing program provides an alarm for the abnormal time-domain vibration value or abnormal frequency-domain vibration value of the spindle or key components, and analyzes the abnormal processing behavior of the machine or the cause of the faulty component; and the predictive maintenance analysis module is integrated from The machine sensor captures the time-domain vibration value and frequency-domain vibration value of the spindle or key components of the module and the machine spindle speed, operating status or operation information of the processing program from the machine controller acquisition module Generate integrated information. When the integrated information conforms to the established characteristic mathematical model, the predictive maintenance analysis module provides the maintenance method or maintenance timing of the machine before the abnormal processing behavior of the machine or component failure.

The method for analyzing and predicting the maintenance abnormality of the machine tool of the present invention includes: the sensor sensor acquisition module captures the sensor's sensing value of the machine tool's main shaft or key element to calculate the main shaft or key element Time-domain vibration value and frequency-domain vibration value, and the machine controller acquisition module captures the machine spindle speed, operating status or operation information of the processing program through the controller; the processing behavior abnormal analysis module is based on the machine The time-domain vibration value and frequency-domain vibration value of the spindle or key component calculated by the sensor acquisition module, and the spindle speed, operating state or processing program of the machine acquired by the machine controller acquisition module Operation information, provide alarms for abnormal time-domain vibration values or abnormal frequency-domain vibration values of spindles or key components, and analyze the abnormal processing behavior of the machine or the cause of the faulty component; and integrate the machine sensing from the machine by the predictive maintenance analysis module The machine captures the time-domain vibration value and frequency-domain vibration value of the spindle or key components of the module and the operation information of the spindle speed, operating state or processing program of the machine from the machine controller to retrieve the module to generate integrated information, When the integrated information conforms to the established characteristic mathematical model, the predictive maintenance analysis module provides the maintenance method and timing of the machine before the abnormal processing behavior of the machine or component failure.

In order to make the above-mentioned features and advantages of the present invention more obvious and understandable, the embodiments are specifically described below in conjunction with the accompanying drawings for detailed description. Additional features and advantages of the present invention will be partially explained in the following description, and these features and advantages will be partially apparent from the description, or may be learned by practicing the present invention. The features and advantages of the present invention are recognized and achieved by means of the elements and combinations particularly pointed out in the scope of the patent application. It should be understood that both the foregoing general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the claimed scope of the invention.

1‧‧‧ Abnormal analysis and predictive maintenance system of machine processing

10‧‧‧machine

11‧‧‧spindle

12‧‧‧Key components

13‧‧‧Sensor

14‧‧‧Controller

20‧‧‧ machine sensor acquisition module

30‧‧‧ machine controller capture module

40‧‧‧Analysis module for abnormal processing behavior

41‧‧‧Processing program analysis module

42‧‧‧Amplitude analysis module in time domain

421‧‧‧ Time domain processing model analysis unit

422‧‧‧Processing abnormality analysis unit

423‧‧‧ Frequency domain processing model analysis unit

43‧‧‧Processing analysis module

431‧‧‧Process abnormality analysis unit

432‧‧‧Process abnormality analysis unit

44‧‧‧ Frequency domain amplitude analysis module

441‧‧‧Frequency domain spindle life and key component model analysis unit

45‧‧‧Spindle life analysis module

451‧‧‧Analysis unit for abnormal type of the first component

452‧‧‧Analysis unit for abnormality of the first component

46‧‧‧Key component analysis module

461‧‧‧Analysis unit for abnormal type of the second component

462‧‧‧Analysis unit for abnormality of the second component

50‧‧‧Processing Behavior History Module

51‧‧‧ Abnormal processing behavior characteristic analysis unit

52‧‧‧Analysis unit of spindle life and abnormal machining behavior of key components

60‧‧‧Predictive Maintenance Analysis Module

61‧‧‧ Feature Model Learning Unit

62‧‧‧ Expert maintenance diagnosis unit

63‧‧‧Prediction result unit

70‧‧‧Mathematical model update module

80‧‧‧ Abnormal model setting module for time-domain amplitude processing behavior

90‧‧‧ Frequency domain amplitude processing abnormality model setting module

Steps S1 to S5‧‧‧

Figure 1 is a schematic diagram of an architecture of an abnormal analysis and predictive maintenance system of machine processing behavior of the present invention; Figure 2 is a schematic diagram of another architecture of an abnormal analysis and predictive maintenance system of machine processing behavior of the present invention; The schematic flow chart of the method for abnormal analysis and predictive maintenance of machine processing behavior of the invention; and FIGS. 4A to 20D are schematic views of embodiments of the method for abnormal analysis and predictive maintenance of machine processing behavior of the present invention.

The following describes the embodiments of the present invention by specific specific embodiments. Those familiar with this technology can easily understand other advantages and effects of the present invention from the contents disclosed in this specification, and can also be implemented by other different specific embodiments. Or application.

FIG. 1 is a schematic structural diagram of a system 1 for analyzing and predicting maintenance abnormality of machine processing behavior of the present invention. As shown in the figure, the machine tool processing behavior abnormality analysis and prediction maintenance system 1 includes a sensor 13, a controller 14, a machine sensor acquisition module 20, a machine controller acquisition module 30, and a machining behavior abnormality The analysis module 40, the processing history module 50, the predictive maintenance analysis module 60, and the mathematical model update module 70 may also include a time domain amplitude processing behavior abnormality model setting module 80 and a frequency domain amplitude processing behavior abnormality model setting module Group 90.

The sensor 13 may be, for example, an accelerometer, etc., and is disposed on the main shaft 11 or the key element 12 of the machine 10, or outside of the machine 10. The present invention is not limited thereto. The machine sensor capturing module 20 can capture the (real-time) sensing value of the spindle 13 or the key component 12 of the machine 10 by the sensor 13 to calculate the time-domain vibration value of the spindle 11 or the key component 12 Vibration value with frequency domain.

The controller 14 may be installed on the machine 10. The machine controller acquisition module 30 can be connected (communication) with the controller 14, and the operation information such as the spindle speed, the operating state, or the processing program of the machine 10 can be acquired through the controller 14.

The processing behavior abnormality analysis module 40 can be based on the time domain vibration value and the frequency domain vibration value of the spindle 11 or the key component 12 calculated by the machine sensor acquisition module 20, and the machine controller acquisition module 30. Retrieved operating information such as spindle speed, operating status or machining program of the machine 10, providing four abnormal analysis methods such as time-domain amplitude, cutting processing, spindle life and key components, so that the abnormal processing behavior analysis module 40 targets the spindle 11 or the abnormal time domain vibration value or abnormal frequency domain vibration value of the key component 12 provides a preliminary alarm, and automatically analyzes or judges the abnormal processing behavior of the machine 10 or the cause of the faulty component.

The processing behavior history module 50 can collect or analyze abnormal processing behavior and operation information of the faulty component analyzed or determined by the processing behavior abnormality analysis module 40 to establish characteristic data of the abnormal processing behavior. This characteristic data can include abnormal processing characteristics and frequency-domain vibration values of the processing process, and can also include (1) abnormal spindle life and key component characteristics, and (2) frequency-domain vibration values of the processing process.

The predictive maintenance analysis module 60 can learn the predictive maintenance mathematical model and characteristic parameters unique to each machine 10, and integrate the information from the machine sensor acquisition module 20 (such as the time-domain vibration value of the spindle 11 or the key component 12) And frequency domain vibration value) and the information from the machine controller acquisition module 30 (such as the spindle speed of the machine 10, the operating status or the operation information of the machining program) to generate integrated information, so that the integrated information conforms to the established learning For the characteristic mathematical model, the predictive maintenance analysis module 60 (expert maintenance diagnosis unit 62 in FIG. 2) provides the machine maintenance method and machine maintenance timing before abnormal processing behavior or component failure occurs according to the pre-set.

The mathematical model update module 70 can feed back the learning results of the abnormal characteristics of the machine 10 predicted by the maintenance analysis module 60 to the frequency domain machining model analysis unit 423 of the machining behavior abnormality analysis module 40 or the frequency domain spindle life and key components The model analysis unit 441 (see FIG. 2) enables the frequency domain machining model analysis unit 423 of the machining behavior abnormality analysis module 40 or the frequency domain spindle life and the abnormal component diagnosis method of the key component model analysis unit 441 to quickly converge.

The time-domain amplitude processing behavior abnormality model setting module 80 allows the user to set the time-domain amplitude processing behavior abnormality model of the machine 10 as shown in FIG. 5A, for example, the state of the machine 10, the processing program, the time-domain amplitude waveform, the duration Model conditions such as time, number of occurrences...

The frequency domain amplitude machining behavior abnormality model setting module 90 allows the user to set the frequency domain amplitude machining behavior abnormality model of the machine 10 as shown in FIG. 5B, such as setting processing, spindle life, and key components. This abnormal model of frequency domain amplitude machining behavior can include the main frequency f of the machine 10, the harmonic frequencies 2f, 3f, 4f, etc. and the contribution of each frequency band, or the bearing retainer damage frequency (such as FTF=RPM×40~60% ).

FIG. 2 is another schematic diagram of the system for analyzing and predicting the maintenance abnormality of the machine processing behavior 1 of the present invention. As shown in the figure, the machine tool processing behavior abnormality analysis and predictive maintenance system 1 may include the machine sensor acquisition module 20, the machine controller acquisition module 30, and the machining behavior abnormality analysis as shown in FIG. Module 40, machining behavior history module 50, predictive maintenance analysis module 60 and mathematical model update module 70, time domain amplitude machining behavior abnormal model setting module 80 and frequency domain amplitude machining behavior abnormal model setting module 90, also The sensor 13 and the controller 14 of FIG. 1 may be further included, but not limited thereto.

The processing behavior abnormality analysis module 40 may include a processing program analysis module 41, a time domain amplitude analysis module 42, a processing analysis module 43, a frequency domain amplitude analysis module 44, a spindle life analysis module 45, and a key component analysis module 46. The time domain amplitude analysis module 42 may have a time domain processing model analysis unit 421, a processing abnormality analysis unit 422, and a frequency domain processing model analysis unit 423. The processing analysis module 43 may have a processing abnormality cause analysis unit 431 and a processing abnormality degree The analysis unit 432, the frequency domain amplitude analysis module 44 may have a frequency domain spindle life and key component model analysis unit 441, and the spindle life analysis module 45 may have a first component abnormality analysis unit 451 and a first component abnormality analysis unit 452 The critical component analysis module 46 may have a second component abnormality analysis unit 461 and a second component abnormality analysis unit 462. The processing behavior history module 50 may include an abnormal processing behavior characteristic analysis unit 51 and a spindle life and key component abnormal processing behavior characteristic analysis unit 52, and the predictive maintenance analysis module 60 may include a characteristic model learning unit 61, an expert maintenance diagnosis unit 62 and prediction Results unit 63.

The processing program analysis module 41 of the processing behavior abnormality analysis module 40 can read the machine processing code of the machine controller retrieval module 30 from FIG. 1 to interpret the machine from the machine processing code Processing behavior. At the same time, the processing program analysis module 41 can identify the processing actions required by the processing analysis module 43 (such as grinding, screw processing, drilling and cutting, etc.), and can also identify the spindle life analysis module 45 or key component analysis The analysis actions required by the module 46 (such as warm start of the machine, spindle running-in, tool compensation adjustment... etc.). If the machining program analysis module 41 recognizes the machining action required by the machining analysis module 43, the time domain amplitude analysis module 42 is entered; and if the machining program analysis module 41 recognizes the spindle life analysis module 45 or key component analysis The analysis action required by the module 46 enters the frequency domain amplitude analysis module 44.

In the time-domain amplitude analysis module 42, when the processing program analysis module 41 recognizes the processing actions (such as grinding, screw processing, drilling, etc.) required by the processing analysis module 43, the time domain The time domain processing model analysis unit 421 of the amplitude analysis module 42 analyzes the time domain processing model of the machine 10, and the processing abnormality analysis unit 422 of the time domain amplitude analysis module 42 analyzes the processing abnormality of the machine 10, and the time domain The frequency domain processing model analysis unit 423 of the amplitude analysis module 42 analyzes the frequency domain processing model of the machine 10.

For example, the time-domain processing model analysis unit 421 can read the real-time measurement value of the time-domain amplitude from the machine sensor acquisition module 20 shown in FIG. 1 and based on the time domain shown in FIG. 5A Data collection and formula calculation of abnormal amplitude processing behavior models, such as: grinding processing, screw processing, drilling cutting, etc., and various time-domain abnormal models of various processing operations (such as normal time-domain amplitude waveform, operation Status, abnormal duration, number of abnormal occurrences...), where the abnormal model of time-domain amplitude processing behavior can be set through the platform or tool.

In the time-domain processing model analysis unit 421, the analysis formula of the time-domain processing model may be, for example: TM_Cutting(Return TM_C_Alarm)_i=F1(C_Process,Vmg(t) _ref ,Vmg(t),Tcontinued,Count,Status) . The aforementioned TM_Cutting(Return TM_C_Alarm)_i is the analysis of the i-th time domain processing model, where TM_C_Alarm=True/False. C_Process is a processing action. Vmg(t) _ref is the normal time-domain amplitude waveform, where t is time. Vmg(t) is the real-time measurement time-domain amplitude waveform, where t is time. Tcontinued is the abnormal duration. Count is the number of abnormal occurrences. Status is the running state of the machine.

The processing abnormality analysis unit 422 may perform processing abnormality judgment according to the analysis or calculation result of the time-domain processing model analysis unit 421. When the time-domain processing model is established (such as TM_C_Alarm=True), the processing abnormality analysis unit 422 can immediately generate a processing (such as C_Process) abnormality alarm, and then enter the frequency-domain processing model analysis unit 423 to analyze the cause and degree of abnormality.

The frequency domain processing model analysis unit 423 can read the FFT (Fast Fourier Transform; Fast Fourier Transform) fast frequency domain amplitude value of the machine sensor acquisition module 20 shown in FIG. 1, and the processing behavior is abnormal according to the frequency domain amplitude Model setting module 90's abnormal model of frequency domain amplitude machining behavior performs data collection and formula calculation, such as grinding processing, screw processing, drilling cutting, etc. various processing actions and their corresponding abnormal causes (such as high speed, tool Wear, mechanical loosening, etc.), as well as frequency-domain anomaly models for various causes of anomalies (such as main frequency f, harmonic frequencies 2f, 3f, 4f..., etc., and the contribution of each frequency band), where the frequency domain The abnormal model of amplitude processing behavior can be set by platform or tool.

In the frequency domain processing model analysis unit 423, the analysis formula of the frequency domain processing model may be, for example: FM_Cutting(Return FM_C_Alarm)_i=F2(C_Process, C_Error, C_Value, FFT(f, n, kn), Status). The aforementioned FM_Cutting(Return FM_C_Alarm)_i is the analysis of the i-th frequency domain machining model, where FM_C_Alarm=True/False. C_Process is a processing action. C_Error is the cause of the exception. C_Value is the amplitude of the abnormal frequency domain. FFT(f,n,kn) is the abnormal frequency domain amplitude formula, where f is the rotation frequency of the machine spindle, n is the number of harmonic frequencies, and kn is the contribution of each frequency band. Status is the running state of the machine.

The processing analysis module 43 can determine the cause of the abnormality for the (cutting) processing action (C_Process) of the machine 10 according to the analysis or calculation result of the frequency domain processing model of the frequency domain processing model analysis unit 423. When the frequency domain processing model is established (such as FM_C_Alarm=True), the processing analysis module 43 immediately informs the system that the cause of the abnormal occurrence of action processing (C_Process) is C_Error, and stores the analysis result in the processing abnormality cause analysis unit 431. At the same time, the processing analysis module 43 performs processing abnormality analysis based on the real-time measurement FFT (f, n, kn) size, and stores the analysis result in the processing abnormality analysis unit 432.

In the frequency-domain amplitude analysis module 44, the machining program analysis module 41 recognizes the spindle life and key component analysis (such as warm-up of the machine, spindle running-in, tool compensation adjustment, etc.). In a specific behavior, the frequency domain spindle life and key component model analysis unit 441 can perform frequency domain spindle life and key component model analysis.

For example, the frequency domain spindle life and key component model analysis unit 441 of the frequency domain amplitude analysis module 44 can read the frequency domain amplitude value (such as FFT) from the machine sensor acquisition module 20 shown in FIG. 1 Frequency domain amplitude value), and data collection and formula calculation based on the frequency domain amplitude processing behavior abnormal model setting module 90 frequency domain amplitude processing behavior abnormal model, such as: warm start of the machine, spindle running-in, tool compensation adjustment ... and other specific behaviors of various machines, corresponding abnormal or faulty components (such as rolling bearing outer race, ball, fund. Train, inner race), and frequency domain abnormal models of various abnormal or faulty components (such as bearing retainer damage) Frequency FTF=RPM×40~60%), in which the abnormal model of frequency domain amplitude processing behavior can be set by platform or tool.

In the frequency domain spindle life and key component model analysis unit 441, the analysis formula of the frequency domain spindle life and key component model can be, for example: FM_Machine(Return FM_M_Alarm)_i=F3(M_Process,M_Error,M_Value,FFT(f,n ,kn),Status). The aforementioned FM_Machine(Return FM_M_Alarm)_i is the i-th frequency domain spindle life and key component model analysis, where FM_M_Alarm=True/False. M_Process is the specific behavior of the machine. M_Error is an abnormal or faulty component. M_Value is the amplitude of the abnormal frequency domain. FFT(f,n,kn) is the abnormal frequency domain amplitude formula, where f is the rotation frequency of the machine spindle, n is the number of harmonic frequencies, and kn is the contribution of each frequency band. Status is the running state of the machine.

The spindle life analysis module 45 can perform spindle life analysis on the specific behavior (M_Process) of the machine 10 according to the analysis or calculation results of the frequency domain spindle life and the key component model analysis unit 441. When the frequency domain spindle life and key component models are established (such as FM_M_Alarm=True), the spindle life analysis module 45 can immediately notify the system about the abnormal causes or faulty components of the spindle life (such as M_Error), and store the analysis results in the first In the component abnormality analysis unit 451. At the same time, the spindle life analysis module 45 can also perform component abnormality analysis according to the size of FFT (f, n, kn), and store the analysis result in the first component abnormality analysis unit 452.

The analysis method of the key component analysis module 46 is similar to the analysis method of the spindle life analysis module 45 described above, and an abnormal cause analysis or damage degree evaluation can be performed for other key components 12 of the machine 10. That is, the key component analysis module 46 can analyze the key component 12 for the specific behavior (M_Process) of the machine 10 according to the analysis or calculation result of the frequency domain spindle life and the key component model analysis unit 441. When the frequency domain spindle life and the key component model are established (such as FM_M_Alarm=True), the key component analysis module 46 can immediately notify the system about abnormal or faulty components of the key component (such as M_Error), and store the analysis results in the second component Abnormal type analysis unit 461. At the same time, the key component analysis module 46 can perform component abnormality analysis according to the size of FFT (f, n, kn), and store the analysis result in the second component abnormality analysis unit 462.

The processing behavior history module 50 can analyze or judge the abnormal processing behavior for the processing analysis module 43, the spindle life analysis module 45, and the key component analysis module 46, and automatically collect processing abnormalities from the processing abnormality analysis unit 431, from Abnormality of processing abnormality analysis unit 432, abnormality of component from first abnormality analysis unit 451 (second abnormality analysis unit 461), abnormality of first abnormality analysis unit 452 (analysis of abnormality of second component) Unit 462) the abnormal degree of the component, and filter or store the complete FFT frequency domain amplitude value of the time of abnormal processing behavior.

The abnormal processing behavior characteristic analysis unit 51 of the processing behavior history module 50 can analyze the abnormal processing behavior characteristics, and the spindle life and critical component abnormal processing behavior characteristic analysis unit 52 can analyze the spindle life and critical component abnormal processing behavior characteristics. According to this, the characteristic data table of abnormal processing behavior is established.

It is stated that the abnormal processing behavior characteristic analysis unit 51 can create a characteristic data table of processing abnormal characteristics (abnormal causes and abnormal degrees) and abnormal process FFT frequency domain amplitude values. The data format of this characteristic data table can be, for example: DB_Cutting_Table(C_Process , C_Error, C_Level, C_FFT(start, end))ij. The aforementioned DB_Cutting_Table is a characteristic data table of processing abnormal processing behavior. C_Process is a processing action. C_Error is the cause of the exception. C_Level is the abnormal level. C_FFT(Tstart, Tend) is the amplitude value of FFT frequency domain in abnormal process, where Tstart is the start time of abnormal processing behavior and Tend is the end time of abnormal processing behavior. DB_Cutting_Table(C_Process, C_Error, C_Level, C_FFT(start, end)) ij is the jth abnormal process FFT frequency domain amplitude value data table for the i-th abnormal processing feature (abnormal cause and abnormal level).

The characteristic analysis unit 52 for spindle life and abnormal processing behavior of key components can establish a characteristic data table of spindle life and abnormal characteristics of abnormal components (abnormal problems and abnormal degrees) and FFT frequency domain amplitude values of abnormal processes. The data format of this characteristic data table can be, for example It is: DB_Machine_Table(M-Process, M_Error, M_Level, M_FFT(start, end))ij. The aforementioned DB_Machine_Table is a characteristic data table of spindle life and abnormal machining behavior of key components. M_Process is the specific behavior of the machine. M_Error is an abnormal component problem. M_Level is the abnormal level of the component. M_FFT(Tstart, Tend) is the amplitude value of the FFT frequency domain of the component abnormal process, where Tstart is the start time of component abnormal behavior and Tend is the end time of component abnormal behavior. DB_Machine_Table(M_Process, M_Error, M_Level, M_FFT(start, end)) ij is the jth abnormal process FFT frequency domain amplitude of the i-th spindle life and abnormal characteristics (abnormal problems and abnormal degrees) of key components.

The predictive maintenance analysis module 60 can collect the characteristic data tables of abnormal processing behaviors required for predictive maintenance analysis through the processing behavior history module 50, and the number of characteristic data tables of abnormal processing behaviors can be set through a platform or a tool.

The characteristic model learning unit 61 of the predictive maintenance analysis module 60 can analyze and establish a mathematical model of abnormal processing behavior, and the learning method of the mathematical model of abnormal processing behavior can be achieved using techniques such as regression analysis, neural network-like... and so on. At the same time, the mathematical model update module 70 can feedback the learning results of the feature model learning unit 61 to the frequency domain machining model analysis unit 423 and the frequency domain spindle life and key component model analysis unit 441, so that the frequency domain machining model analysis unit 423 and The frequency domain spindle life and the anomaly analysis and diagnosis method of the key component model analysis unit 441 converge quickly.

The expert maintenance diagnosis unit 62 of the predictive maintenance analysis module 60 may be provided with a machine maintenance method for abnormal processing behavior, and the machine maintenance method may be set through a platform or tool. When the integrated information of the machine sensor acquisition module 20 and the machine controller acquisition module 30 shown in FIG. 1 conforms to the characteristic mathematical model that has been learned and established, the prediction result unit 63 may provide abnormal processing behavior or components Failure maintenance methods and machine maintenance timing.

In a word, the characteristic model learning unit 61 can analyze the characteristic data table of the spindle life and the abnormal processing behavior of key components, and use the characteristic model learning method (such as regression analysis method, neural network...) to establish a characteristic mathematical model , Where the feature model learning method can be selected or set through the platform or tool. At the same time, the feature model learning unit 61 can output the correlation coefficient index to indicate the credibility of the feature mathematical model. When the correlation coefficient index is greater than the set target, the feature model learning unit 61 can store the feature mathematical model and notify the mathematical model update module 70 Conduct feedback and update of learning results.

In the feature model learning unit 61, the formula for feature model learning may be, for example: Model_Learning(Return Finish, Index, FFT_Model(f, n, kn))_i=F4(DB_Table, Error, n, Method). The aforementioned Model_Learning(Return Finish, Index, FFT_Model(f, n, kn))_i is the i-th feature model learning formula, where Finish=True (complete modeling)/False (in modeling). Index is the index of the correlation coefficient of the characteristic model. FFT_Model(f,n,kn) is a characteristic mathematical model, where f is the rotation frequency of the machine spindle, n is the number of harmonic frequencies, and kn is the contribution of each frequency band. DB_Table is DB_Cutting_Table or DB_Machine_Table, Error is the cause of abnormality (C_Error) or component abnormality problem (M_Error), n is the number of data tables, and Method is the feature model learning method.

In addition, in the feature model learning unit 61, when the correlation coefficient index is greater than the set target, the mathematical model update module 70 may, according to the correlation coefficient index of the feature model of the feature model learning unit 61, convert the FFT_Model(f, n, kn) Feedback and update of feature mathematical model. For example, when Error is an abnormal cause (such as C_Error), the mathematical model update module 70 may update the frequency domain processing model of the frequency domain processing model analysis unit 423; and when Error is an abnormal component problem (such as M_Error), the mathematical model is updated The module 70 can update the frequency domain spindle life and key component model of the frequency domain spindle life and key component model analysis unit 441.

The expert maintenance diagnosis unit 62 can provide a dedicated machine maintenance method for the abnormal reasons of cutting processing, spindle life and key components, and analyze and recommend suitable maintenance opportunities based on the degree of abnormality. The data format of the expert maintenance diagnosis data table can be: DB_Diagnosis_Table_i (Process, Error, Level, Maintain_Method, Maintain_Time). The aforementioned DB_Diagnosis_Table_i is the i-th expert maintenance diagnosis data table, Process is C_Process or M_Process, Error is C_Error or M_Error, Level is C_Level or M_Level, Maintain_Method is the machine maintenance method, and Maintain_Time is the machine maintenance timing.

When the integrated information of the machine sensor acquisition module 20 and the machine controller acquisition module 30 shown in FIG. 1 matches the learned mathematical model FFT_Model(f,n,kn), the prediction result unit 63 can compare or query the expert maintenance diagnosis unit 62 to provide a description of the predicted results, such as: abnormal processing behavior (Process), abnormal cause or component abnormal problem (Error), machine maintenance method (Maintain_Method) and machine maintenance timing (Maintain_Time ).

Figure 3 is a schematic flow chart of the abnormal processing analysis and predictive maintenance method of the machine tool of the present invention. The main technical contents can be described as follows and refer to Figures 1 to 2; for the remaining technical contents, please refer to Figure 1 Detailed descriptions to FIGS. 2 and 4A to 20D.

In step S1 of FIG. 3, the sensor sensor acquisition module 20 captures the sensor 13 to the spindle 10 or the key element 12 of the machine 10 to calculate the spindle 11 or the key element 12 The vibration value in the time domain and the vibration value in the frequency domain are acquired by the machine controller acquisition module 30 through the controller 14 to acquire the operating information of the spindle speed, the operating state, or the processing program of the machine 10. At the same time, the time domain amplitude processing behavior abnormal model setting module 80 sets the time domain amplitude processing behavior abnormal model of the machine 10, and the frequency domain amplitude processing behavior abnormal model setting module 90 sets the frequency domain amplitude processing behavior of the machine 10 Abnormal model.

In step S2 in FIG. 3, the processing behavior abnormality analysis module 40 calculates the time domain vibration value and the frequency domain vibration value of the spindle 11 or the key component 12 calculated by the machine sensor acquisition module 20, and the machine The operation information of the spindle speed, running state or processing program of the machine 10 captured by the table controller acquisition module 30 provides an alarm for the abnormal time-domain vibration value or abnormal frequency-domain vibration value of the spindle 11 or the key component 12, And analyze the abnormal processing behavior of the machine 10 or the cause of the faulty component.

To proclaim, the processing program analysis module 41 of the processing behavior abnormality analysis module 40 can read the machine processing code from the machine controller acquisition module 30 to interpret the machine from the machine processing code Processing behavior. At the same time, the time-domain amplitude analysis module 42 of the processing behavior abnormality analysis module 40 can analyze the time-domain processing model, processing abnormality, or frequency-domain processing model of the machine 10. Furthermore, the processing analysis module 43 of the processing behavior abnormality analysis module 40 can determine the abnormality of the processing action of the machine 10 according to the analysis result of the frequency domain processing model.

In addition, the frequency domain amplitude analysis module 44 of the processing behavior abnormality analysis module 40 can read the frequency domain amplitude value from the machine sensor acquisition module 20 and based on the frequency domain amplitude processing behavior abnormal model of the machine 10 Conduct data collection. At the same time, the spindle life analysis module 45 of the machining behavior abnormality analysis module 40 can analyze the abnormal cause or malfunction component of the spindle life of the machine 10, and the machine can be analyzed by the key component analysis module 46 of the machining behavior abnormality analysis module 40 10 Abnormal causes or damage degree of key components.

In step S3 of FIG. 3, the processing behavior history module 50 collects the abnormal analysis results from the processing behavior abnormal analysis module 40 for the abnormal processing behavior of the machine 10, so as to establish the characteristic data of the abnormal processing behavior based on the abnormal analysis results table.

In step S4 of FIG. 3, the predictive maintenance analysis module 60 integrates the information from the machine sensor acquisition module 20 (such as the time domain vibration value and the frequency domain vibration value of the spindle 11 or the key component 12) and Retrieve information from the module 30 of the machine controller (such as the spindle speed, operating status, or operation information of the machining program of the machine 10) to generate integrated information. When the integrated information conforms to the established characteristic mathematical model, it is predicted The maintenance analysis module 60 provides the maintenance method and timing of the maintenance of the machine 10 before the abnormal processing behavior of the machine 10 occurs or the component fails. In addition, the predictive maintenance analysis module 60 can also learn the exclusive predictive maintenance mathematical model and characteristic parameters of the machine, and establish the cutting processing and spindle life of the machine 10 according to the characteristic data table of abnormal machining behavior from the machining behavior history module 50 Mathematical model of abnormal features with key components.

In step S5 of FIG. 3, the mathematical model update module 70 feeds back the learning results of the abnormal characteristics of the predictive maintenance analysis module 60 to the machine 10 to the machining behavior abnormality analysis module 40, so that the machining behavior abnormality analysis module The abnormal analysis and diagnosis methods of group 40 converge.

FIGS. 4A to 20D are schematic diagrams of an embodiment of a system and method for analyzing and predicting the maintenance abnormality of the machine tool of the present invention, and the working procedures thereof can be described as follows with reference to FIGS. 1 to 2. It should be noted that FIGS. 4A to 20D are only illustrated in a schematic manner of the embodiment, but the present invention is not limited thereto.

Procedure (1): As shown in Figures 4A to 4B, the factory is equipped with the machine 10 of Figure 1, such as a machine tool or a processing machine (such as a multi-axis lathe integrated processing machine). Before the start of daily operation, first perform the warm-up operation mode of the machine 10, and when the warm-up of the machine 10 is completed, the machine 10 starts screw processing, drilling, cutting, and cutting according to the work order requirements... Wait for processing. At the same time, the time-domain amplitude processing behavior model of the machine 10 (see FIG. 4A) and the frequency-domain amplitude processing behavior model (see FIG. 4B) can be measured by the sensor 13 and the machine sensor acquisition module 20 Obtain or provide data from the original machine manufacturer.

Program (2): As shown in Figures 5A to 5B, for processing analysis (such as screw processing, drilling, cutting, etc...), the module can be set through the abnormal model of the time-domain amplitude processing behavior 80 Set the abnormal model of time-domain amplitude processing behavior in Figure 5A (such as amplitude waveform, abnormal duration, number of abnormal occurrences, etc.), in order to judge the processing abnormality and provide processing abnormality alarm, and through the frequency domain amplitude processing abnormality The model setting module 90 sets the frequency domain amplitude processing abnormality model of Figure 5B (such as the main frequency f, harmonic frequency 2f, 3f, 4f, etc., and the contribution of each frequency band) to facilitate automatic judgment of processing abnormalities the reason.

In the time-domain processing model analysis unit 421, the analysis formula of the time-domain processing model may be, for example: TM_Cutting(Return TM_C_Alarm)_i=F1(C_Process,Vmg(t) _ref ,Vmg(t),Tcontinued,Count,Status) , I=1~3. The aforementioned C_Process is a machining operation (including screw machining, drilling, cutting, cutting), Vmg(t) _ref is the time-domain amplitude waveform of the normal machining operation (obtained by program (1)), and Vmg(t) is the real-time measurement time domain Amplitude waveform, Tcontinued is the abnormal duration (3 seconds), Count is the number of abnormal occurrences (1 time), and Status is the machine running status (processing status).

In the frequency domain processing model analysis unit 423, the analysis formula of the frequency domain processing model may be, for example: FM_Cutting(Return FM_C_Alarm)_i=F2(C_Process,C_Error,C_Value,FFT(f,n,kn),Status), i =1~3. The aforementioned C_Process is a machining action (including screw processing, drilling, cutting, cutting), C_Error is an abnormal cause (including excessive speed, tool wear, and mechanical loosening), and C_Value is an abnormal frequency domain amplitude size (obtained by program (1), FFT(f,n,kn) is the abnormal frequency domain amplitude formula (obtained by program (1)), and Status is the machine running state (processing state).

Program (3): As shown in Figure 6, in the analysis of spindle life and key components (such as warm start), the frequency domain amplitude machining behavior abnormal model setting module 90 can set the frequency domain amplitude machining behavior abnormal model to automatically determine the fault Components, this example is based on the analysis of the abnormal model of ball bearings.

In the frequency domain spindle life and key component model analysis unit 441, the analysis formula of the frequency domain spindle life and key component model can be, for example: FM_Machine(Return FM_M_Alarm)_i=F3(M_Process,M_Error,M_Value,FFT(f,n ,kn),Status),i=1~4. The aforementioned M_Process is the specific behavior of the machine (warm start of the machine), M_Error is an abnormal or faulty component (such as rolling bearing outer race, ball, fund.Train, inner race), and M_Value is the abnormal frequency domain amplitude size (by the program (1 ) Obtained), FFT (f, n, kn) is the abnormal frequency domain amplitude formula (obtained by the program (1)), Status is the machine running state (processing state).

Program (4): After completing the settings of the above program (2) to program (3), the machine 10 will first perform the warm-up operation mode before the machine 10 is started daily. At the same time, the processing program analysis module 41 will read the machine processing code of the machine controller acquisition module 30, and after automatically analyzing or judging the processing action as a warm start of the machine, the frequency domain amplitude analysis module 44 Analyze the abnormal frequency domain amplitude of each abnormal or faulty component (such as M_Error). For example, as shown in FIG. 7, the analysis procedure of the frequency-domain amplitude analysis module 44 is: [a] analyzing the operating state of the machine 10 (see the top of FIG. 7), [b] analyzing the machine sensor Retrieve the frequency domain amplitude value of the module 20 and the spindle rotation frequency of the machine 10 (see Figure 7), [c] set the frequency domain amplitude processing behavior abnormal model of the module 90 according to the frequency domain amplitude processing behavior abnormal model Data collection and formula calculation (see bottom of Figure 7).

Program (5): analyze the analysis or calculation result of the above program (4) through the spindle life analysis module 45, and store the analysis result of the abnormality in the first component abnormality type analysis unit 451 and the first component abnormality level Analysis unit 452. For example, as shown in Figure 8, the analysis procedure of the spindle life analysis module 45 is: [a] Perform spindle life analysis on the behavior of warm boot (see the top of Figure 8), [b]Ball M_Value(40) >28 (see bottom of Figure 8), [c] The frequency domain spindle life model is abnormally established, FM_M_Alarm=True (see bottom of Figure 8).

Procedure (6): After the warm-up of the machine 10 is completed, the machine 10 starts processing such as screw processing, drilling, cutting, etc. according to work order requirements. The processing program analysis module 41 will read the machine processing code of the machine controller acquisition module 30 to automatically analyze or judge the processing action as screw processing, drilling cutting, cutting...etc. The time-domain amplitude analysis module 42 performs alarm analysis of abnormal processing. For example, as shown in FIG. 9, the analysis procedure of the time-domain amplitude analysis module 42 is: [a] analyzing the operating state of the machine 10 (see the top of FIG. 9), [b] analyzing the machine sensor Retrieve the time-domain amplitude value of the module 20 (see the middle of Figure 9), [c] set up the frequency-domain amplitude processing behavior abnormal model of the module 90 according to the frequency-domain amplitude processing behavior abnormal model for data collection and formula calculation (see page 9 below).

Procedure (7): When the machining (such as drilling and cutting) abnormality of the machine 10 occurs, the time-domain amplitude analysis module 42 can analyze the abnormal frequency-domain amplitude magnitude of each anomaly cause (such as C_Error). For example, as shown in FIG. 10, the analysis procedure of the time-domain amplitude analysis module 42 is: [a] Analysis of the FFT frequency-domain amplitude value of the machine sensor acquisition module 20 (see top of FIG. 10) , [B] analyze the spindle rotation frequency amplitude of the machine 10 (see the middle of Fig. 10), [c] set the data collection and formula calculation according to the frequency domain amplitude processing abnormality model setting module 90 frequency domain amplitude processing abnormality model (See Figure 10 below).

Program (8): The analysis or calculation result of the above program (7) is analyzed through the processing analysis module 43, and the analysis result of the abnormality is stored in the processing abnormality analysis unit 431 and the processing abnormality by the processing analysis module 43, respectively Degree analysis unit 432. For example, as shown in FIG. 11, the analysis program of the machining analysis module 43 is: [a] cutting machining analysis for drilling cutting behavior (see top of FIG. 11), [b] tool wear C_Value(33) >18 (see the top of Figure 11), [c] The abnormality of the frequency-domain machining model is established, FM_C_Alarm=True (see the bottom of the Figure 11), [d] notify the system about the abnormal cause of drilling and cutting of the machining analysis due to tool wear And store the analysis results (see the bottom of Figure 11).

Program (9): The processing behavior history module 50 automatically collects the analysis results of the above programs (5) and (8), and filters and stores the complete FFT frequency domain amplitude value of the occurrence time of the abnormal processing behavior, and then passes the abnormal processing behavior The characteristic analysis unit 51 and the spindle life and abnormal processing behavior of the key components The characteristic analysis unit 52 completes the establishment of the characteristic data table of the abnormal processing behavior. For example, as shown in FIG. 12A, the data format or characteristic data of the characteristic data table created by the abnormal processing behavior characteristic analysis unit 51 includes the data table (j), C_Error, C_Level, and C_FFT(start, end). As shown in FIG. 12B, the data format or characteristic data of the characteristic data table created by the characteristic analysis unit 52 for spindle life and abnormal processing behavior of key components includes the data table (j), M_Error, M_Level, and M_FFT(start, end).

Program (10): As shown in Figures 13A to 13B, after the establishment of the characteristic data table of abnormal machining behavior, the predictive maintenance analysis module 60 is directed to C_Process (drilling cutting) and C_Error (tool wear), And M_Process (warm boot) and M_Error (Ball) set the number of data tables required for predictive maintenance analysis to 3, so as to automatically filter the contents of the data table (j) with a higher abnormal level (C_Level or M_Level).

Program (11): As shown in FIGS. 14A to 14B, the feature model learning unit 61 is used to analyze and establish the mathematical model of abnormal processing behavior, in which the number of data tables (n) is 3, and the feature model learning method ( Method) Use regression analysis.

Procedure (12): As shown in Figures 15A to 15B, use the regression analysis method to obtain the key spectrum parameters (correlation coefficient (R ² )>0.8), and complete the characteristic mathematical model FFT_Model(f ,n,kn), where Figure 15A uses C_Process (drilling cutting), C_Error (tool wear), FFT_Model(f,n,kn), and Figure 15B uses M_Process (warm boot), M_Error(Ball), FFT_Model (f,n,kn).

Procedure (13): As shown in FIGS. 16A to 16B, after completing the feature mathematical model FFT_Model(f,n,kn), determine whether FFT_Model(f,n,kn) meets the update through the mathematical model update module 70 If the conditions are met, the mathematical model update module 70 updates the frequency domain machining model of the frequency domain machining model analysis unit 423 and the frequency domain spindle life and key component model analysis unit 441 of the frequency domain spindle life model and key component model.

In this embodiment, the target value of the correlation coefficient is set to 0.8, because the correlation coefficients (0.9919 and 0.9911) of FFT_Model (f, n, kn) are greater than 0.8, which meets the update conditions, so the mathematical model update module 70 will correct and update The related formulas of the frequency domain machining model analysis unit 423, and the related formulas of the frequency domain spindle life and key component model analysis unit 441, to overcome the model errors caused by machine differences and component aging, let the frequency domain machining model analysis unit 423 and frequency domain The spindle life and critical component model analysis unit 441 anomaly analysis and diagnosis methods quickly converge.

Procedure (14): The predictive maintenance analysis module 60 will store the characteristic mathematical model of the above procedure (12), such as FFT_Model(f,n,kn). When the integrated information of the machine sensor acquisition module 20 and the machine controller acquisition module 30 at the front end conforms to FFT_Model(f,n,kn) and Index>0.8, the abnormal processing behavior will be provided by the prediction result unit 63 Or the maintenance method of component failure and the timing of machine maintenance. The maintenance mode of the expert maintenance diagnosis unit 62 can be set through a platform or tool. The explanation is as follows:

[a] As shown in FIGS. 17A to 17D, when M_Process is a warm boot, read the integrated information of the machine sensor acquisition module 20 and the machine controller acquisition module 30 four times, where Figures 17C to 17D are consistent with M_Process for warm boot, M_Error for Ball, FFT_Model(f,n,kn), Index>0.8.

[b] As shown in FIGS. 18A to 18D, under C_Process for drilling, read the integrated information of the machine sensor acquisition module 20 and the machine controller acquisition module 30 four times. Among them, figures 18C to 18D conform to C_Process for drilling and cutting, C_Error for tool wear, FFT_Model(f,n,kn), and Index>0.8.

[c] As shown in FIG. 19, according to the original manufacturer's maintenance manual of the machine 10, the maintenance method of the expert maintenance diagnosis unit 62 is set through the platform or tool.

[d] As shown in FIGS. 20A to 20D, for the analysis results of [a] and [b] in the above procedure (14), the prediction result unit 63 provides maintenance methods and machines for abnormal machining behavior or component failure Maintenance timing to achieve the predicted maintenance purpose of the machine.

In summary, the system and method for analyzing and predicting maintenance abnormalities of the machine tool of the present invention may have, for example, the following features, advantages, or technical effects:

(1) The present invention can combine the controller and the machine controller acquisition module through the sensor and machine sensor acquisition module to target the abnormal time-domain vibration value or abnormal frequency of the machine spindle or key components Alarms in the field vibration value, and automatically analyze or judge the abnormal processing behavior of the machine (such as grinding processing, screw processing, drilling and cutting, etc.) or faulty components (such as rolling bearing outer race, ball, fund.Train, inner race) cause.

(2) The processing behavior abnormality analysis module (processing analysis module) of the present invention can perform processing feature analysis for specific processing behaviors (such as grinding processing, screw processing, drilling cutting, etc.) through the processing program, and according to the time Domain amplitude processing behavior models (such as amplitude size, duration, number of occurrences... etc.) automatically raise alarms for processing abnormalities, and based on frequency domain amplitude processing behavior models (such as main frequency f, harmonic frequency 2f, 3f, 4f... . Etc. and the contribution of each frequency band) Automatically determine the cause of abnormal processing (such as high speed, tool wear, mechanical loosening...) or abnormality index (such as FFT (mm/s ² ) size level of mechanical loosening) .

(3) The frequency spindle life and key component model analysis unit of the present invention can use the processing program to perform the spindle life and the specific behavior of the machine (such as warm start of the machine, spindle run-in, tool compensation adjustment... etc.) Characteristic analysis of key components, and through the frequency domain abnormal model (such as bearing retainer damage frequency FTF=RPM×40~60%) to automatically determine the faulty components (such as rolling bearing outer race, ball, fund.Train, inner race) and abnormality index (Such as FFT (mm/s ² ) size level of FTF frequency).

(4) The processing behavior history module of the present invention can automatically collect abnormal analysis from the processing behavior abnormal analysis module (processing analysis module, spindle life analysis module and key component model analysis module) for the abnormal processing behavior of the machine The result is to establish a characteristic data table of abnormal machining behaviors, such as cutting machining characteristics (abnormal causes and abnormal degrees) and abnormal process FFT frequency domain amplitude values, as well as spindle life and key component characteristics (abnormal problems and abnormal degrees) and abnormal process FFT frequency Domain amplitude value and other two association tables. Accordingly, the present invention can solve the problem that the SCADA (Supervisory Control and Data Acquisition) system in the general industry must store complete raw data, which is easy to cause the problem of huge system data and poor query efficiency.

(5) The processing behavior history module of the present invention can automatically collect abnormal processing behavior and the operation information of the faulty component, and the predictive maintenance analysis module can learn the predictive maintenance mathematical model and characteristic parameters unique to each machine, and analyze the abnormal processing behavior Provide machine maintenance methods and maintenance timing before the occurrence or component failure, so as to achieve the purpose of reducing failure downtime and increasing factory production capacity.

(6) The predictive maintenance analysis module of the present invention can integrate the information from the machine sensor acquisition module and the machine controller acquisition module information, when the integrated information conforms to the established characteristic mathematical model, The predictive maintenance analysis module can provide the machine maintenance method and machine maintenance timing before the abnormal processing behavior of the machine or component failure.

(7) The predictive maintenance analysis module of the present invention can use the characteristic data table of abnormal processing behavior collected by the processing behavior history module to further establish the cutting processing, spindle life and key component abnormalities of each machine using the characteristic model learning unit Characteristic mathematical model, and the learning method can be achieved by using regression analysis method or neural network, etc., and provide the exclusive machine maintenance method and maintenance timing setting method in the expert maintenance diagnosis unit.

(8) The mathematical model update module of the present invention can provide a mathematical model feedback mechanism and feed back the learning results of abnormal features to the processing behavior abnormal analysis module (frequency domain processing model analysis unit and spindle life and key component model analysis Unit), allowing the abnormal analysis and diagnosis methods of the abnormal processing analysis module to quickly converge to overcome the model error caused by machine differences and component aging.

The above-mentioned embodiments only exemplarily illustrate the principles, characteristics and effects of the present invention, and are not intended to limit the scope of the invention. Anyone who is familiar with this skill can do the above without departing from the spirit and scope of the present invention. The embodiment is modified and changed. Any equivalent changes and modifications made using the disclosure of the present invention should still be covered by the scope of the patent application. Therefore, the scope of protection of the rights of the present invention should be as listed in the scope of patent application.

1‧‧‧ Abnormal analysis and predictive maintenance system of machine processing

10‧‧‧machine

11‧‧‧spindle

12‧‧‧Key components

13‧‧‧Sensor

14‧‧‧Controller

20‧‧‧ machine sensor acquisition module

30‧‧‧ machine controller capture module

40‧‧‧Analysis module for abnormal processing behavior

50‧‧‧Processing Behavior History Module

60‧‧‧Predictive Maintenance Analysis Module

70‧‧‧Mathematical model update module

80‧‧‧ Abnormal model setting module for time-domain amplitude processing behavior

90‧‧‧ Frequency domain amplitude processing abnormality model setting module

Claims (20)

A machine tool processing behavior abnormality analysis and predictive maintenance system includes: a machine sensor acquisition module, which captures the sensor's sensing value of the machine tool's main shaft or key component to calculate the main shaft or key Time-domain vibration value and frequency-domain vibration value of the component; the machine controller acquisition module is to acquire the spindle spindle speed, operating status or operation information of the processing program of the machine through the controller; the processing behavior abnormal analysis module, It is based on the time domain vibration value and frequency domain vibration value of the spindle or key component calculated by the machine sensor acquisition module, and the machine spindle captured by the machine controller acquisition module Rotational speed, operating status or operation information of the machining program provides an alarm for the abnormal time-domain vibration value or abnormal frequency-domain vibration value of the spindle or key components, and analyzes the abnormal processing behavior of the machine or the cause of the faulty component; and predicts maintenance The analysis module integrates the time-domain vibration value and frequency-domain vibration value of the main shaft or key components from the machine sensor acquisition module and the main shaft of the machine from the machine controller acquisition module Rotational speed, operating status or operation information of the machining program to generate integrated information. When the integrated information conforms to the established characteristic mathematical model, the predictive maintenance analysis module is used before the abnormal processing behavior of the machine or component failure occurs. Provide the maintenance method or timing of the machine. The system as described in item 1 of the patent application scope, wherein the processing behavior abnormality analysis module has a processing program analysis module for reading the machine processing program code from the machine controller acquisition module, In order to interpret the machine processing behavior from the machine processing code. The system as described in item 1 of the patent application scope, wherein the processing behavior abnormality analysis module has a time-domain amplitude analysis module for analyzing the machine's time-domain processing model, processing abnormality or frequency-domain processing model. The system as described in item 3 of the patent application scope, wherein the processing behavior abnormality analysis module further has a processing analysis module, which is based on the analysis result of the frequency domain processing model to determine the abnormal cause of the processing action of the machine . The system as described in item 1 of the patent application scope, wherein the processing behavior abnormality analysis module has a frequency domain amplitude analysis module for reading the frequency domain amplitude value from the machine sensor acquisition module And collect data based on the frequency domain amplitude processing abnormal model of the machine. The system as described in item 1 of the patent application scope, wherein the processing behavior abnormality analysis module has a spindle life analysis module and a key component analysis module, and the spindle life analysis module analyzes the abnormality of the spindle life of the machine Cause or faulty component, and the critical component analysis module analyzes the abnormal cause or damage degree of the critical component of the machine. The system as described in item 1 of the patent application scope also includes a processing behavior history module, which collects abnormal analysis results from the abnormal processing analysis module of the processing behavior for the abnormal processing behavior of the machine, and establishes based on the abnormal analysis results The characteristic data table of the abnormal processing behavior. The system as described in item 7 of the patent application scope, wherein the predictive maintenance analysis module further learns the predictive maintenance mathematical model and characteristic parameters specific to the machine, and based on the characteristics of the abnormal processing behavior from the processing behavior history module The data table establishes the mathematical model of the machine's cutting process, spindle life and abnormal characteristics of key components. The system described in item 1 of the patent application scope further includes a mathematical model update module, which feeds back the learning results of the predictive maintenance analysis module to the abnormal characteristics of the machine to the processing behavior abnormality analysis module, to The abnormal analysis and diagnosis method of the abnormal analysis module for processing behavior is converged. The system as described in item 1 of the patent application scope further includes a time domain amplitude processing behavior abnormality model setting module and a frequency domain amplitude processing behavior abnormality model setting module. The time domain amplitude processing behavior abnormality model setting module is used for setting The abnormal model of the time-domain amplitude processing behavior of the machine, and the frequency-domain amplitude-processing abnormal model model setting module is used to set the abnormal model of the frequency-domain amplitude processing behavior of the machine. A method for analyzing and predicting the maintenance abnormality of a machine tool includes: the sensor sensor acquisition module captures the sensor's sensing value of the machine tool spindle or key element to calculate the spindle or key element The vibration value in the time domain and the vibration value in the frequency domain are captured by the machine controller acquisition module through the controller to acquire the spindle speed, operating status or operation information of the machining program of the machine; The time-domain vibration value and frequency-domain vibration value of the spindle or key component calculated by the machine sensor acquisition module, and the spindle speed and operation of the machine acquired by the machine controller acquisition module The operation information of the status or processing program provides an alarm for the abnormal time-domain vibration value or abnormal frequency-domain vibration value of the spindle or key component, and analyzes the abnormal processing behavior of the machine or the cause of the faulty component; and the predictive maintenance analysis model Integrate the time-domain vibration value and frequency-domain vibration value of the spindle or key components from the machine sensor acquisition module and the machine spindle speed and operating status from the machine controller acquisition module Or the operation information of the machining program to generate integrated information. When the integrated information conforms to the established characteristic mathematical model, the predictive maintenance analysis module provides the machine before the abnormal machining behavior of the machine or component failure Maintenance methods and timing of maintenance. The method as described in item 11 of the patent application scope further includes the processing program analysis module reading the machine processing code from the machine controller retrieval module to interpret it from the machine processing code Machine processing behavior. The method as described in item 11 of the patent application scope further includes the time-domain amplitude analysis module analyzing the time-domain processing model, the processing abnormality or the frequency-domain processing model of the machine. The method as described in item 13 of the patent application scope further includes the processing analysis module based on the analysis result of the frequency domain processing model to determine the abnormal cause of the processing action of the machine. The method as described in item 11 of the patent application scope further includes reading the frequency domain amplitude value from the machine sensor acquisition module by the frequency domain amplitude analysis module and processing according to the frequency domain amplitude of the machine Abnormal behavior model for data collection. The method described in item 11 of the patent application scope further includes the spindle life analysis module to analyze the abnormal cause or faulty component of the spindle life of the machine, and the key component analysis module to analyze the abnormality of the key components of the machine Cause or degree of damage. The method as described in item 11 of the patent application scope further includes collecting abnormal analysis results from the abnormal analysis module of the processing behavior for the abnormal processing behavior of the machine by the processing behavior history module to establish the abnormal analysis results based on the abnormal analysis results Characteristic data sheet for abnormal processing behavior. The method described in item 17 of the patent application scope further includes the predictive maintenance analysis module learning the predictive maintenance mathematical model and characteristic parameters of the machine, and based on the characteristics of the abnormal processing behavior from the processing behavior history module The data table establishes the mathematical model of the machine's cutting process, spindle life and abnormal characteristics of key components. The method as described in item 11 of the patent application scope further includes that the mathematical model update module feeds back the learning result of the predictive maintenance analysis module to the abnormal characteristics of the machine to the processing behavior abnormal analysis module, so that The abnormal analysis and diagnosis method of the abnormal analysis module for processing behavior is converged. The method as described in item 11 of the patent application scope further includes setting the model of the time-domain amplitude machining behavior abnormality model through the time-domain amplitude processing behavior abnormality model setting module, and setting the module through the frequency-domain amplitude machining behavior abnormality model setting module Set the frequency domain amplitude processing abnormal model of this machine.

Images