A kind of electric spindle fault diagnosis system and diagnosis method thereof

Abstract

The present invention relates to the technical field of fault diagnosis, and specifically discloses an electric spindle fault diagnosis system and a diagnosis method thereof, comprising: a data acquisition module, used for collecting original operation signal data of a historical electric spindle device through a signal acquisition sensor; a data processing module, used for performing data preprocessing and feature selection on the original operation signal data to obtain initial operation data features; a fault diagnosis module, used for inputting the initial operation data features of the historical operation into a preset fault prediction model for training to obtain a target feature sequence; and constructing a target attention mechanism through a Transformer architecture to obtain a target feature sequence, determine the fault type features corresponding to the current time period and each fault type feature parameter; and output the current model as a predicted fault feature; a data analysis module, used for obtaining a deviation coefficient between the real-time initial operation data features and the predicted fault features through a fault diagnosis strategy, and judging the fault diagnosis result according to the size of the deviation coefficient.

Description

Technical Field

[0001] The invention relates to the technical field of fault diagnosis, and in particular to an electric spindle fault diagnosis system and a diagnosis method thereof. Background Art

[0002] The electric spindle is one of the core components of modern CNC machine tools, and its performance directly affects the processing quality and production efficiency. However, due to long-term high-speed operation and complex working conditions, the electric spindle is prone to wear, looseness, fatigue and other faults. Therefore, it is of great significance to diagnose the fault of the electric spindle.

[0003] At present, the commonly used electric spindle fault diagnosis methods mainly include vibration analysis, temperature monitoring, oil analysis, etc., but these methods often require professional instruments and equipment and technicians, and have problems such as low diagnostic accuracy and poor real-time performance.

[0004] However, in the traditional electric spindle fault diagnosis process, it is difficult to link and combine the initialization feature data of the electric spindle device's operating status detection and the accuracy of the fault prediction of the intelligent network model. Further steps are needed to determine the accuracy of the electric spindle fault diagnosis results using the current diagnostic model, which in turn increases the time for the fault diagnosis prediction effect analysis equipment and effect data analysis, and increases the load on the electric spindle equipment's electrical energy operation. Summary of the invention

[0005] The purpose of the present invention is to provide an electric spindle fault diagnosis system and a diagnosis method thereof to solve the following technical problems:

[0006] How to achieve real-time analysis and judgment on the accuracy of electric spindle fault diagnosis prediction and reduce the error of fault diagnosis detection results.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] An electric spindle fault diagnosis system, comprising:

[0009] A data acquisition module, used for collecting original operation signal data of the historical electric spindle device through a signal acquisition sensor;

[0010] The original operation signal data includes: operation vibration signal data, operation temperature signal data and operation noise signal data;

[0011] A data processing module is used to perform data preprocessing and feature selection on the original operation signal data to obtain the initial operation data features;

[0012] The fault diagnosis module is used to input the initial data features of historical operation into the preset fault prediction model for training to obtain the target feature sequence; and to construct the target attention mechanism through the Transformer architecture to obtain the target feature sequence, determine the fault type features corresponding to the current time period and the feature parameters of each fault type; and use the current model output as the predicted fault feature;

[0013] The target attention mechanism of the fault diagnosis module obtains the real-time correlation between the target feature and the input feature within a specified time period by calculating:

[0014] By formula Calculate the correlation coefficient ;

[0015] in, The dot product scoring parameters for the attention mechanism; Run the initial data feature vector for input;

[0016] The target characteristic states include: fault state, suspected fault state and normal state;

[0017] The data analysis module is used to obtain the deviation coefficient between the real-time operation initial data characteristics and the predicted fault characteristics through the fault diagnosis strategy, and judge the fault diagnosis result according to the size of the deviation coefficient;

[0018] An alarm prompt module sends out an alarm signal according to the fault diagnosis result;

[0019] The data storage module is used to store the fault diagnosis data of the electric spindle.

[0020] Preferably, the input operation initial data feature vector The calculation method is:

[0021] ;

[0022] in, is the predicted feature vector state weight; is the real-time status of the encoder; is the real-time starting point, is the real-time end point; To predict the current state of the model features; is the current state weight of the prediction model feature; is the bias parameter; It represents the context of the current input features of the prediction model in the real-time state of the encoder.

[0023] Preferably, the fault diagnosis strategy of the data analysis module includes:

[0024] Obtain a feature vector of context input for running initial data features in real time;

[0025] Calculates a parameterized set of differences between adjacent eigenvectors within a specified time period and uses it as an influence function And output;

[0026] By formula Calculate the deviation coefficient of the current operating state of the electric spindle ;in, To run the vibration signal parameters, is the operating temperature signal parameter, is the operating noise signal parameter; is the influence function of the running vibration signal parameters, is the influence function of the operating temperature signal parameters, is the influence function of the running noise signal parameters; For standard operation vibration signal parameters, is the standard operating temperature signal parameter, is the standard operating noise signal parameter; It is the preset running vibration signal parameter deviation value; is the preset operating temperature signal parameter deviation value; It is the preset running noise signal parameter deviation value.

[0027] Preferably, it also includes: judging the deviation coefficient The threshold interval of the coefficient of standard deviation To compare:

[0028] like ∈ , the diagnosis result is judged to be qualified; otherwise, the diagnosis result is unqualified and an alarm signal is issued.

[0029] Preferably, the standard deviation coefficient threshold interval is obtained as follows:

[0030] By real-time correlation coefficient within a specified time period The model outputs the current weight of each fault feature; the minimum and maximum values ​​of each fault weight assignment output are used as the standard deviation coefficient threshold interval of the corresponding fault; and the standard deviation coefficient threshold intervals are divided into multiple intervals according to the fault type.

[0031] Preferably, the signal acquisition sensor comprises:

[0032] A vibration sensor is used to detect the vibration signal of the electric spindle and convert it into vibration signal data through a signal conditioner;

[0033] Temperature sensor, used to detect the operating temperature signal data of the electric spindle;

[0034] Piezoelectric sensor, used to detect the noise signal data of the electric spindle operation;

[0035] Preferably, the step of the data processing module acquiring the initial running characteristics comprises:

[0036] Perform data filtering and noise reduction preprocessing on the collected original operation signal data;

[0037] Feature selection is performed on the preprocessed original operation signal data to obtain corresponding features, and time series feature processing is performed on multiple features to obtain the initial operation data features.

[0038] A diagnostic method for an electric spindle fault diagnosis system, the method comprising the following steps:

[0039] Step 1: Collecting original operation signal data of the historical electric spindle device through the signal acquisition sensor of the data acquisition module; the original operation signal data includes: operation vibration signal data, operation temperature signal data and operation noise signal data;

[0040] Step 2: Use the data processing module to perform data preprocessing and feature selection on the original operation signal data to obtain the initial operation data features;

[0041] Step 3: Use the fault diagnosis module to input the initial data features of historical operation into the preset fault prediction model for training to obtain the target feature sequence; use the Transformer architecture to construct the target attention mechanism to obtain the target feature sequence, determine the fault type features corresponding to the current time period and the feature parameters of each fault type; and use the current model output as the predicted fault feature;

[0042] Step 4: Design a data analysis module to obtain the deviation coefficient between the real-time initial data characteristics and the predicted fault characteristics through the fault diagnosis strategy, and judge the rationality of the fault diagnosis result according to the size of the deviation coefficient, and send an alarm signal for the fault diagnosis result according to the alarm prompt module.

[0043] Beneficial effects of the present invention:

[0044] (1) The present invention sets a data analysis module to analyze the deviation between the real-time initial data characteristics and the predicted fault characteristics according to the fault diagnosis strategy, and further judges the accuracy of the fault diagnosis through the deviation coefficient, thereby ensuring the high accuracy and timeliness of the diagnosis results, and being able to make real-time corrections to erroneous judgment results and timely repair the equipment.

[0045] (2) The present invention processes the data after preprocessing and feature selection according to the fault diagnosis module, obtains the target feature sequence by inputting the initial data features of the historical operation into the preset fault prediction model for training; and obtains the target feature sequence by constructing the target attention mechanism through the Transformer architecture, and determines the fault type features corresponding to the current time period and the characteristic parameters of each fault type; the real-time data input model output results are used as the predicted fault features; it is achieved to ensure the acquisition of the correlation between the target features and the input features within the specified time period of the real-time input, and obtains the output of the corresponding fault type features according to the current real-time correlation, thereby improving the real-time detection effect of the electric spindle device operation fault.

[0046] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0048] Figure 1 This is a module diagram of an electric spindle fault diagnosis system of the present invention;

[0049] Figure 2 This is a step diagram of a method for diagnosing faults of an electric spindle according to the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Since it is difficult to link and combine the initialization feature data of the electric spindle device's operating status detection and the accuracy of the fault prediction of the intelligent network model in the traditional electric spindle fault diagnosis process, it is necessary to add further steps to determine the accuracy of the electric spindle fault diagnosis results using the current diagnostic model, which increases the time for fault diagnosis prediction effect analysis equipment and effect data analysis, and increases the load on the electric energy operation of the electric spindle equipment.

[0052] See also Figure 1 As shown, the present invention is an electric spindle fault diagnosis system, comprising:

[0053] A data acquisition module, used for collecting original operation signal data of the historical electric spindle device through a signal acquisition sensor;

[0054] The original operation signal data includes: operation vibration signal data, operation temperature signal data and operation noise signal data;

[0055] A data processing module is used to perform data preprocessing and feature selection on the original operation signal data to obtain the initial operation data features;

[0056] The fault diagnosis module is used to input the initial data features of historical operation into the preset fault prediction model for training to obtain the target feature sequence; and to construct the target attention mechanism through the Transformer architecture to obtain the target feature sequence, determine the fault type features corresponding to the current time period and the feature parameters of each fault type; and use the current model output as the predicted fault feature;

[0057] The target characteristic states include: fault state, suspected fault state and normal state;

[0058] The data analysis module is used to obtain the deviation coefficient between the real-time operation initial data characteristics and the predicted fault characteristics through the fault diagnosis strategy, and judge the fault diagnosis result according to the size of the deviation coefficient;

[0059] An alarm prompt module sends out an alarm signal according to the fault diagnosis result;

[0060] The data storage module is used to store the fault diagnosis data of the electric spindle.

[0061] In the above technical scheme, in order to solve the above technical problems, this design realizes the accuracy judgment of electric spindle fault diagnosis by setting six modules; realizes accurate judgment of the effect of electric spindle fault diagnosis prediction results and reduces the error of fault diagnosis detection.

[0062] Specifically, by setting up a data acquisition module, the data acquisition module is used to realize the collection and processing of the operating data of the electric spindle device; by setting up a signal acquisition sensor, the original operating signal data of the historical electric spindle device is collected; the signal acquisition sensor mainly depends on the signal acquisition system, such as using a non-contact vibration sensor to collect vibration signals, and this collection and measurement method has a high degree of accuracy; the collection of other signals in this embodiment also depends on the signal acquisition system and is determined by the specific setting of the sensor for collection according to the data type required for collection. Among them, this embodiment collects original operation signals based on the main parts of the electric spindle device: the operation failure mode and normal performance characteristics of the spindle and bearings. The original operation signal data includes: operation vibration signal data, operation temperature signal data and operation noise signal data; the collection of the operation vibration signal and temperature signal of the electric spindle device can reflect whether the spindle and bearings have failed; and according to the change in temperature, it can also reflect whether the current spindle and bearings are running smoothly, such as the frictional heat generated, the speed deviation caused by eccentricity, etc., which lead to wear and structural damage of the spindle and bearings. In addition, some structural problems are difficult to detect through direct performance characteristics. This design determines whether the operation of the electric spindle has problems such as insufficient lubrication, operating load, and spindle and bearing balance by detecting the changes in the operation noise state of the current electric spindle device; of course, based on such judgment, it can be determined whether it is necessary to conduct regular inspection and replacement of the equipment in a timely manner.

[0063] By setting up the data processing module, the data is further optimized and converted, and the input data is guaranteed to meet the operation conditions of the training model. Due to the different types of data obtained, data preprocessing and feature selection are two crucial steps when processing the original running signal data. First, data preprocessing aims to improve data quality, including removing noise, filling missing values, and processing outliers, to ensure data accuracy and consistency. Second, feature selection is to select the most influential features for model prediction from the original data to reduce data dimensions and improve the computational efficiency and prediction performance of machine learning and later models.

[0064] By setting up a fault diagnosis module, the data after preprocessing and feature selection is further trained and processed according to the current fault diagnosis module. Specifically, the initial data features of historical operation are input into the preset fault prediction model for training to obtain the target feature sequence; wherein, the preset fault prediction model is built based on a deep learning model through a multi-layer neural network.

[0065] Furthermore, a target attention mechanism is constructed through the Transformer architecture to obtain a target feature sequence; the fault type feature corresponding to the current time period and the characteristic parameters of each fault type are determined, and the current model output is used as the predicted fault feature; the target feature state includes: fault state, suspected fault state and normal state; the fault diagnosis module is used to ensure that the target feature and input feature relationship within the specified time period of the real-time input are correlated, and the output of the corresponding fault type feature is obtained according to the current real-time correlation, thereby improving the real-time detection effect of the electric spindle device operation fault.

[0066] By setting up a data analysis module, the deviation coefficient between the real-time operation initial data characteristics and the predicted fault characteristics is obtained according to the fault diagnosis strategy, and the fault diagnosis result is judged according to the size of the deviation coefficient; because most of the existing technologies default the fault diagnosis result to accurate reference information and default to the current diagnosis result; for example, in this embodiment, the corresponding fault type is directly judged based on the size of the correlation of the fault, while ignoring some data judgment errors in a series of data processing processes, and misjudgments caused by human operation deviations in equipment data acquisition. Especially for the judgment of the selected time period, it is very necessary to confirm whether the real-time collection of the original data of the electric spindle equipment is consistent with the fault diagnosis result.

[0067] Therefore, this design sets up a fault diagnosis strategy to further judge the accuracy of fault diagnosis by taking the deviation coefficient between the initial data characteristics of real-time operation and the predicted fault characteristics, ensuring the high accuracy and timeliness of the diagnosis results, and being able to make real-time corrections to erroneous judgment results and timely repair of equipment. In addition, by setting up an alarm prompt module, an alarm signal is sent to the operating end and related maintenance personnel according to the fault diagnosis results; a data storage module is also set up to store the fault diagnosis data of the electric spindle, which is convenient for improving data problems and improving the accuracy of fault problem prediction.

[0068] As an implementation mode of the present invention, the target attention mechanism of the fault diagnosis module obtains the real-time correlation between the target feature and the input feature within the specified time period by calculating:

[0069] By formula Calculate the correlation coefficient ;

[0070] in, The dot product scoring parameters for the attention mechanism; Run the initial data feature vector for input.

[0071] In the above technical solution, the fault diagnosis module in this embodiment also determines the correlation between the fault characteristics and the input electric spindle equipment operation characteristics by obtaining real-time correlation, so as to ensure that the diagnosis prediction is made in advance. Specifically, through the target attention mechanism based on the parallel processing capability and efficient long-distance dependency capture capability of Transformer, it is convenient to obtain the correlation between the target characteristics and the input characteristics within a specified time period.

[0072] As an embodiment of the present invention, the input operation initial data feature vector The calculation method is:

[0073] ;

[0074] in, is the predicted feature vector state weight; is the real-time status of the encoder; is the real-time starting point, is the real-time end point; To predict the current state of the model features; is the current state weight of the prediction model feature; is the bias parameter; It represents the context of the current input features of the prediction model in the real-time state of the encoder.

[0075] In the above technical solution, the running initial data feature vector in this embodiment is It depends on the correlation between the current prediction model characteristics and the predicted fault characteristics results, and then calculates the correlation coefficient Specifically, add dot product scores to the attention mechanism to construct a time series consisting of prediction results and The correlation between them is used to obtain the real-time correlation size; in the specific calculation process, a loss function is added to weight the time series of the prediction results and the current input features of the prediction model in the context of the real-time state in the encoder for subsequent calculations. It should be noted that the above calculation and analysis processes are all existing calculation methods of the target attention mechanism, and the specific calculation process will not be described in detail.

[0076] As an implementation mode of the present invention, the fault diagnosis strategy of the data analysis module includes:

[0077] Obtain a feature vector of context input for running initial data features in real time;

[0078] Calculates a parameterized set of differences between adjacent eigenvectors within a specified time period and uses it as an influence function And output;

[0079] By formula Calculate the deviation coefficient of the current operating state of the electric spindle ;in, To run the vibration signal parameters, is the operating temperature signal parameter, is the operating noise signal parameter; is the influence function of the running vibration signal parameters, is the influence function of the operating temperature signal parameters, is the influence function of the running noise signal parameters; For standard operation vibration signal parameters, is the standard operating temperature signal parameter, is the standard operating noise signal parameter; It is the preset running vibration signal parameter deviation value; is the preset operating temperature signal parameter deviation value; It is the preset running noise signal parameter deviation value.

[0080] In the above technical solution, the deviation coefficient is obtained by setting the fault diagnosis strategy, and the accuracy of the current fault diagnosis result is reflected according to the size of the deviation coefficient; specifically, the formula Calculate the deviation coefficient of the current operating state of the electric spindle ; and the influence function here is It is a parameter indicator for real-time acquisition of specified data according to a specified time period; when applied to this formula, it is calculated based on the experience summary of the deviation of the current operating state according to the change of the original operating signal data of the electric spindle device within the current time period, and an adjustment function is pre-set according to the historical data to ensure that the current deviation value is within a specific reasonable range.

[0081] It should be noted that the standard operating vibration signal parameters , Standard operating temperature signal parameters And standard operating noise signal parameters The standard data are obtained from the machine simulation analyzer under the normal operation state of the electric spindle according to the size of the operation value corresponding to the current electric spindle device; and the running vibration signal parameter deviation value is preset , preset operating temperature signal parameter deviation value And preset running noise signal parameter deviation value They are all obtained through real-time analysis of the network training based on conventional data, and are set in advance based on the historical database, which will not be described in detail here.

[0082] As an embodiment of the present invention, it also includes: judging the deviation coefficient The threshold interval of the coefficient of standard deviation To compare:

[0083] like ∈ , the diagnosis result is judged to be qualified; otherwise, the diagnosis result is unqualified and an alarm signal is issued.

[0084] In the above technical solution, the rationality of the diagnosis result of the current time period is judged by comparison and analysis in this embodiment, and the deviation coefficient is specifically The threshold interval of the coefficient of standard deviation To compare the size, When it falls within the set range, the diagnostic result is judged to be consistent with the actual operating status. If it is not in the current range, the diagnostic result is judged to be unqualified and further investigation is required, and an alarm signal is issued to ensure timely fault investigation.

[0085] As an implementation of the present invention, the standard deviation coefficient threshold interval is obtained as follows:

[0086] By real-time correlation coefficient within a specified time period The model outputs the current weight of each fault feature; the minimum and maximum values ​​of each fault weight assignment output are used as the standard deviation coefficient threshold interval of the corresponding fault; and the standard deviation coefficient threshold intervals are divided into multiple intervals according to the fault type.

[0087] In the above technical scheme, the specific method for obtaining the standard deviation coefficient threshold interval in this embodiment is: obtain the weights of various fault characteristics corresponding to the model output of the real-time correlation coefficient within the specified time period, and assign values ​​according to the weight of each fault characteristic and then output the minimum and maximum values ​​therein as the standard deviation coefficient threshold interval of the fault, and each fault type has its corresponding threshold interval, and the corresponding interval matching is performed according to the fault output by the model. For the situation that does not meet any interval, it is determined that the determination result of the current fault type does not meet the requirements, and an early warning is issued and corresponding inspections are performed according to the current fault type.

[0088] As an embodiment of the present invention, the signal acquisition sensor includes:

[0089] A vibration sensor is used to detect the vibration signal of the electric spindle and convert it into vibration signal data through a signal conditioner;

[0090] Temperature sensor, used to detect the operating temperature signal data of the electric spindle;

[0091] Piezoelectric sensor, used to detect the noise signal data of the electric spindle operation;

[0092] In the above technical scheme, the sensor equipment for signal acquisition sensors in this embodiment includes vibration sensors, temperature sensors and piezoelectric sensors. Of course, other sensors are also used to assist in acquiring the operation information of the electric spindle device according to the actual operation process. For example, by setting the electrical performance size and operating load conditions of the electric spindle, voltage sensors, current sensors, etc. are set. Of course, it is not limited to the above sensor types, and the selection of sensors is also specifically based on the electric spindle and the detection direction of the acquired signal.

[0093] As an embodiment of the present invention, the step of the data processing module acquiring the initial running characteristics includes:

[0094] Perform data filtering and noise reduction preprocessing on the collected original operation signal data;

[0095] Feature selection is performed on the preprocessed original operation signal data to obtain corresponding features, and time series feature processing is performed on multiple features to obtain the initial operation data features.

[0096] In the above technical scheme, this embodiment sets up a data processing module to preprocess the collected vibration, noise and temperature operation signals, and the preprocessing method adopts data filtering and noise reduction preprocessing; the original operation signal data after preprocessing selects the characteristics of each operation signal through the feature selection machine processing method to perform time series feature processing, outputs the initial operation data characteristics, and uses the data after the current data processing module as the basis for model analysis.

[0097] The present invention also designs a diagnostic method for an electric spindle fault diagnosis system, please refer to Figure 2 As shown, the method comprises the following steps:

[0098] Step 1: Collecting original operation signal data of the historical electric spindle device through the signal acquisition sensor of the data acquisition module; the original operation signal data includes: operation vibration signal data, operation temperature signal data and operation noise signal data;

[0099] Step 2: Use the data processing module to perform data preprocessing and feature selection on the original operation signal data to obtain the initial operation data features;

[0100] Step 3: Use the fault diagnosis module to input the initial data features of historical operation into the preset fault prediction model for training to obtain the target feature sequence; use the Transformer architecture to construct the target attention mechanism to obtain the target feature sequence, determine the fault type features corresponding to the current time period and the feature parameters of each fault type; and use the current model output as the predicted fault feature;

[0101] Step 4: Design a data analysis module to obtain the deviation coefficient between the real-time initial data characteristics and the predicted fault characteristics through the fault diagnosis strategy, and judge the rationality of the fault diagnosis result according to the size of the deviation coefficient, and send an alarm signal for the fault diagnosis result according to the alarm prompt module.

[0102] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined in this application, they shall all fall within the protection scope of the present invention.

Claims

1. An electric spindle fault diagnosis system, characterized in that: include: A data acquisition module, used for collecting original operation signal data of the historical electric spindle device through a signal acquisition sensor; The original operation signal data includes: operation vibration signal data, operation temperature signal data and operation noise signal data; A data processing module is used to perform data preprocessing and feature selection on the original operation signal data to obtain the initial operation data features; The fault diagnosis module is used to input the initial data features of historical operation into the preset fault prediction model for training to obtain the target feature sequence; and to construct the target attention mechanism through the Transformer architecture to obtain the target feature sequence, determine the fault type features corresponding to the current time period and the feature parameters of each fault type; and use the current model output as the predicted fault feature; The target attention mechanism of the fault diagnosis module obtains the real-time correlation between the target feature and the input feature within the specified time period by calculating: By formula Calculate the correlation coefficient ; in, The dot product scoring parameters for the attention mechanism; Run the initial data feature vector for input; The target characteristic state includes: fault state, suspected fault state and normal state; The data analysis module is used to obtain the deviation coefficient between the real-time operation initial data characteristics and the predicted fault characteristics through the fault diagnosis strategy, and judge the fault diagnosis result according to the size of the deviation coefficient; The fault diagnosis strategy of the data analysis module includes: Obtain a feature vector of context input for running initial data features in real time; Calculates a parameterized set of differences between adjacent eigenvectors within a specified time period and uses it as an influence function And output; By formula Calculate the deviation coefficient of the current operating state of the electric spindle ;in, To run the vibration signal parameters, is the operating temperature signal parameter, is the operating noise signal parameter; is the influence function of the running vibration signal parameters, is the influence function of the operating temperature signal parameters, is the influence function of the running noise signal parameters; For standard operation vibration signal parameters, is the standard operating temperature signal parameter, is the standard operating noise signal parameter; It is the preset running vibration signal parameter deviation value; is the preset operating temperature signal parameter deviation value; is the preset running noise signal parameter deviation value; An alarm prompt module sends out an alarm signal according to the fault diagnosis result; The data storage module is used to store the fault diagnosis data of the electric spindle.

2. The electric spindle fault diagnosis system according to claim 1, characterized in that: The input runs the initial data feature vector The calculation method is: in, is the predicted feature vector state weight; is the real-time status of the encoder; is the real-time starting point, is the real-time end point; To predict the current state of the model features; is the current state weight of the prediction model feature; is the bias parameter; It represents the context of the current input features of the prediction model in the real-time state of the encoder.

3. The electric spindle fault diagnosis system according to claim 2, characterized in that: Also includes: Judgment bias coefficient The threshold interval of the coefficient of standard deviation To compare: like ∈ , the diagnosis result is judged to be qualified; otherwise, the diagnosis result is unqualified and an alarm signal is issued.

4. The electric spindle fault diagnosis system according to claim 3, characterized in that: The standard deviation coefficient threshold interval is obtained as follows: By real-time correlation coefficient within a specified time period The model outputs the current weight of each fault feature; The minimum and maximum values ​​of each fault weight assignment output are used as the standard deviation coefficient threshold interval of the corresponding fault; and the standard deviation coefficient threshold intervals are divided into multiple intervals according to the fault type.

5. The electric spindle fault diagnosis system according to claim 1, characterized in that: The signal acquisition sensor comprises: A vibration sensor is used to detect the vibration signal of the electric spindle and convert it into vibration signal data through a signal conditioner; Temperature sensor, used to detect the operating temperature signal data of the electric spindle; Piezoelectric sensor is used to detect the noise signal data of the electric spindle operation.

6. The electric spindle fault diagnosis system according to claim 1, characterized in that: The step of obtaining the initial running characteristics by the data processing module comprises: Perform data filtering and noise reduction preprocessing on the collected original operation signal data; Feature selection is performed on the preprocessed original operation signal data to obtain corresponding features, and time series feature processing is performed on multiple features to obtain the initial operation data features.

7. A diagnostic method for an electric spindle fault diagnosis system according to any one of claims 1 to 6, characterized in that: The method comprises the following steps: Step 1: Collecting original operation signal data of the historical electric spindle device through the signal acquisition sensor of the data acquisition module; the original operation signal data includes: operation vibration signal data, operation temperature signal data and operation noise signal data; Step 2: Use the data processing module to perform data preprocessing and feature selection on the original operation signal data to obtain the initial operation data features; Step 3: Use the fault diagnosis module to input the initial data features of historical operation into the preset fault prediction model for training to obtain the target feature sequence; use the Transformer architecture to construct the target attention mechanism to obtain the target feature sequence, determine the fault type features corresponding to the current time period and the feature parameters of each fault type; and use the current model output as the predicted fault feature; Step 4: Design a data analysis module to obtain the deviation coefficient between the real-time initial data characteristics and the predicted fault characteristics through the fault diagnosis strategy, and judge the rationality of the fault diagnosis result according to the size of the deviation coefficient, and send an alarm signal for the fault diagnosis result according to the alarm prompt module.

Images