TWI749742B - Machine tool spindle diagnosis method - Google Patents

Abstract

A method for diagnosing the spindle of a machine tool is suitable for diagnosing whether a spindle of a machine tool to be tested is normal. A computer device uses a vibration sensor to detect the vibration signals generated by the spindle operating in a predetermined time interval and perform frequency spectrum Analyze and time domain analysis to obtain multiple time domain feature values and multiple time domain feature values, and perform principal component analysis on these frequency domain feature values and these time domain features to obtain multiple analysis data. Analyze the data, establish a Gaussian model based on the analysis data, and then perform Gaussian mixing on the established Gaussian model to obtain a Gaussian mixture model, and obtain a difference value according to the Gaussian mixture model and a preset model, and finally according to the Gaussian mixture model and a preset model. The difference value and a preset threshold value produce a diagnosis result indicating whether the machine tool under test is normal.

Description

Machine tool spindle diagnosis method

The invention relates to a diagnosis method, in particular to a diagnosis method for a machine tool spindle.

The development of machine tools is changing with each passing day. High-speed and high-precision have become the development trend of machine tools. Under this trend, if abnormal conditions of the machine tool spindle cannot be found in time during the machining process, it will often affect the production yield and the machine tool. The service life of the spindle.

However, the existing method of machine tool spindle diagnosis cannot be performed on the machine. The real-time diagnosis of the machine tool spindle is whether it is normal. Generally, the machine tool spindle is inspected, measured and calibrated before the machine tool is processed or after a period of processing. However, multiple shutdown corrections will increase processing time and production costs.

Therefore, the purpose of the present invention is to provide a method for diagnosing whether the machine tool spindle is normal during processing.

Therefore, the method for diagnosing the spindle of a machine tool of the present invention is suitable for diagnosing whether a spindle of a machine tool to be tested is normal or not, which is implemented by a computer device which is electromechanically connected with the tool to be tested, and the machine tool to be tested includes the spindle And a vibration sensor for sensing the spindle and electrically connected to the computer device. The computer device stores a preset model and a preset threshold. The method includes one step (A), one step (B), one Step (C), one step (D), one step (E), one step (F), one step (G), and one step (H).

In the step (A), the computer device receives a plurality of vibration signals generated by the vibration sensor sensing the spindle operation in a predetermined time interval.

In this step (B), the computer device performs frequency spectrum analysis on the vibration signals to obtain multiple frequency domain characteristic values.

In this step (C), the computer device performs time-domain analysis on the vibration signals to obtain multiple time-domain feature values.

In this step (D), the computer device performs principal component analysis on the frequency domain feature values and the time domain features to obtain multiple analysis data, and each analysis data includes multiple analysis feature values.

In the step (E), for each analysis data, the computer device establishes a Gaussian model based on the analysis characteristic value of the analysis data.

In this step (F), the computer device uses a Gaussian model mixture algorithm to perform Gaussian mixture on the Gaussian model corresponding to the analysis data to obtain a Gaussian mixture model.

In the step (G), the computer device obtains a difference value related to the difference between the Gaussian mixture model and the preset model according to the Gaussian mixture model and the preset model.

In the step (H), the computer device generates a diagnosis result indicating whether the machine tool under test is normal according to the difference value and the preset threshold value.

The effect of the present invention is that after the computer device performs spectrum analysis, time domain analysis, and principal component analysis based on the vibration signals sensed by the vibration sensor, the Gaussian mixture algorithm is established by the Gaussian model mixture algorithm. Model, and generate the diagnosis result according to the difference between the Gaussian mixture model and the preset model, so as to instantly diagnose whether the machine tool to be tested is normal.

Before the present invention is described in detail, it should be noted that in the following description, similar elements are denoted by the same numbers.

1, a computer device 11 used to implement an embodiment of the method for diagnosing a machine tool spindle of the present invention and a machine tool under test 12 electrically connected to the computer device 11 are illustrated. The computer device 11 stores a preset model and a preset threshold. The machine tool under test 12 includes a spindle 121 and a vibration sensor 122 disposed on the spindle 121 and used for sensing vibration of the spindle 121. It is worth noting that, in this embodiment, the main shaft 121 has, for example, two tools (not shown), and the vibration sensor 122 is, for example, an accelerometer, but it is not limited.

Referring to Figures 1 and 2, the steps included in this embodiment of the method for diagnosing a machine tool spindle of the present invention will be described below.

In step 21, the computer device 11 receives a plurality of vibration signals generated by the vibration sensor 122 to sense the spindle 121 during a predetermined time interval. It should be noted that, in this embodiment, the predetermined time interval is, for example, 100 seconds, but it is not limited thereto.

In step 22, the computer device 11 performs frequency spectrum analysis on the vibration signals to obtain multiple frequency domain feature values.

With reference to Figure 3, step 22 includes the following sub-steps:

In step 221, the computer device 11 converts the vibration signal into a frequency domain value. It is worth noting that in this embodiment, the computer device 11 uses Fourier transform to convert the vibration signals into frequency domain values, but it is not limited to this.

In step 222, the computer device 11 obtains a plurality of key frequency domain values from the frequency domain values. It is worth noting that in this embodiment, the computer device 11 first calculates a main frequency, and then uses a frequency domain value located at 1 to 30 times the main frequency as the key frequency domain value, where the main frequency is, for example, the The revolution per minute (RPM) of the spindle 121 is divided by 60 and then multiplied by the number of tools of the spindle 121. For example, if the revolution of the spindle 121 is 8000RPM, the main frequency is 8000/60*2=266.7, but Not limited to this.

In step 223, the computer device 11 performs filtering and outlier processing on the key frequency domain values to obtain the frequency domain characteristic values. It is worth noting that, in this embodiment, the computer device 11 uses Kalman filter to filter out noise values, and uses the Z-score method to filter out outliers, but not This is limited.

In step 23, the computer device 11 performs time-domain analysis on the vibration signals to obtain multiple time-domain feature values. It is worth noting that in this embodiment, the time-domain characteristic values are, for example, a Kurtosis value, a Root-Mean-Square (RMS) value, and a crest factor (Crest) of the vibration signals. At least two of a Factor value, a Skewness value, a Standard Deviation (SD), and a Variance value, but not limited to this.

， 其中， n為振動訊號的數量，

為第i個振動訊號，

為該等振動訊號的平均。 The kurtosis value K is expressed by the following formula:

, Where n is the number of vibration signals,

Is the i-th vibration signal,

It is the average of these vibration signals.

， 其中， n為振動訊號的數量，

為第i個振動訊號。 The root mean square value M is expressed by the following formula:

, Where n is the number of vibration signals,

It is the i-th vibration signal.

， 其中

為該等振動訊號中絕對值後最大的值，

為該等振動訊號的方均根。 The crest factor value C is expressed by the following formula:

, in

Is the largest value after the absolute value of these vibration signals,

Is the root mean square of these vibration signals.

， 其中， n為振動訊號的數量，

為第i個振動訊號，

為該等振動訊號的平均。 The skewness value S is expressed by the following formula:

, Where n is the number of vibration signals,

Is the i-th vibration signal,

It is the average of these vibration signals.

以下式表示：

其中， n為振動訊號的數量，

為第i個振動訊號，

為該等振動訊號的平均。 The standard deviation

The following formula represents:

Where n is the number of vibration signals,

Is the i-th vibration signal,

It is the average of these vibration signals.

The variance value is the square of the standard deviation value.

In step 24, the computer device 11 performs principal component analysis on the frequency domain feature values and the time domain features to obtain multiple pieces of analysis data, and each piece of analysis data includes multiple analysis feature values.

以下式表示：

其中，

為該分析資料的第i個分析特徵值，

為該分析資料中最小的分析特徵值，

為該分析資料中最大的分析特徵值。 In step 25, for each analysis data, the computer device 11 normalizes the analysis feature value of the analysis data to obtain a plurality of normalized values. Where these normalized values

The following formula represents:

in,

Is the i-th analysis characteristic value of the analysis data,

Is the smallest analysis characteristic value in the analysis data,

It is the largest analysis characteristic value in the analysis data.

In step 26, for each analysis data, the computer device 11 establishes a Gaussian model according to the normalized value corresponding to the analysis data. It is worth noting that, in this embodiment, for each analysis data, the computer device 11 obtains the average value and the variance of the normalized value corresponding to the analysis data according to the normalized value corresponding to the analysis data, so as to establish the Gaussian model.

，

，

，

， 其中， k為步驟26所獲得的對應該等分析資料的高斯模型的數量，

為混合加權值，

為第i個高斯模型，

為第i個高斯模型的中心點，即第i筆分析資料對應的正規化值之平均值，

為第i個高斯模型的變異數，即第i筆分析資料對應的正規化值之變異數。 In step 27, the computer device 11 uses a Gaussian model mixture algorithm to perform Gaussian mixture on the Gaussian model corresponding to the analysis data obtained in step 26 to obtain a Gaussian mixture model. The density function of the Gaussian mixture model is expressed by the following formula:

,

,

,

, Where k is the number of Gaussian models corresponding to the analysis data obtained in step 26,

Is the mixed weight value,

Is the i-th Gaussian model,

Is the center point of the i-th Gaussian model, that is, the average value of the normalized values corresponding to the i-th analysis data,

Is the variance of the i-th Gaussian model, that is, the variance of the normalized value corresponding to the i-th analysis data.

In step 28, the computer device 11 obtains a difference value related to the difference between the Gaussian mixture model and the preset model according to the Gaussian mixture model and the preset model. It is worth noting that in the embodiment, the difference value is 1 minus the overlap rate between the Gaussian mixture model and the preset model, because the feature of the present invention is not known to those skilled in the art Calculate the overlap ratio based on the Gaussian mixture model and the preset model. The detailed method is described in "Haojun Sun, Shengrui Wang. Measuring the component overlapping in the Gaussian mixture model . Computer Science Data Mining and Knowledge Discovery, 2011" For the sake of brevity, their details are omitted here. In other embodiments, the difference value may also be the distance between the Gaussian mixture model and the preset model, and is not limited to this.

In step 29, the computer device 11 generates a diagnosis result indicating whether the machine tool 12 under test is normal according to the difference value and the preset threshold value.

With reference to FIG. 4, step 29 includes sub-steps 291 to 293, and the sub-steps included in step 29 are described below.

In step 291, the computer device 11 determines whether the difference value is less than the preset threshold. When it is determined that the difference value is less than the preset threshold value, the process proceeds to step 292; and when the difference value is not less than the preset threshold value, the process proceeds to step 293.

In step 292, the computer device 11 generates the diagnosis result indicating that the machine tool 12 to be tested is normal.

In step 293, the computer device 11 generates the diagnosis result indicating that the machine tool 12 to be tested is abnormal.

It should be particularly noted that, in this embodiment, the preset model is established by performing steps 21 to 27 with a normal machine tool (not shown in the figure), but it is not limited to this.

To sum up, the method for diagnosing machine tool spindles of the present invention uses the computer device 11 to perform spectrum analysis, time domain analysis, and principal component analysis based on the vibration signals sensed by the vibration sensor 122, and then use the Gaussian The model mixing algorithm establishes the Gaussian mixture model, and generates the diagnosis result according to the difference between the Gaussian mixture model and the preset model, so as to instantly diagnose whether the machine tool 12 under test is normal, so it can indeed achieve the cost of the invention. the goal of.

However, the above are only examples of the present invention. When the scope of implementation of the present invention cannot be limited by this, all simple equivalent changes and modifications made in accordance with the scope of the patent application of the present invention and the content of the patent specification still belong to Within the scope covered by the patent of the present invention.

11: Computer device 12: machine tool to be tested 121: Spindle 122: Vibration Sensor 21~29: Steps 221~223: Steps 291~293: Steps

Other features and effects of the present invention will be clearly presented in the embodiments with reference to the drawings, in which: Figure 1 is a block diagram illustrating a computer device used to implement one embodiment of the machine tool spindle diagnosis method of the present invention; Figure 2 is a flowchart illustrating the embodiment of the method for diagnosing the machine tool spindle of the present invention; Fig. 3 is a flowchart to assist in explaining the sub-steps of step 22 in Fig. 2; and FIG. 4 is a flowchart to assist in explaining the sub-steps of step 29 in FIG. 2.

21~29: Steps

Claims (5)

A method for diagnosing the spindle of a machine tool to be tested is suitable for diagnosing whether a spindle of a machine tool to be tested is normal or not. It is implemented by a computer device which is electromechanically connected with the tool to be tested. The machine tool to be tested includes the spindle and a setting The computer device stores a preset model and a preset threshold in the vibration sensor for sensing the vibration of the spindle. The method includes the following steps: (A) Receiving multiple vibration signals generated by the vibration sensor sensing the spindle operation in a predetermined time interval; (B) Perform frequency spectrum analysis on these vibration signals to obtain multiple frequency domain characteristic values; (C) Perform time-domain analysis of these vibration signals to obtain multiple time-domain eigenvalues; (D) Perform principal component analysis on the frequency domain feature values and the time domain features to obtain multiple analysis data, each of which includes multiple analysis feature values; (E) For each analysis data, establish a Gaussian model based on the analysis characteristic value of the analysis data; (F) Using a Gaussian model mixture algorithm to perform Gaussian mixture on the Gaussian model corresponding to the analysis data to obtain a Gaussian mixture model; (G) According to the Gaussian mixture model and the preset model, obtain a difference value related to the difference between the Gaussian mixture model and the preset model; and (H) According to the difference value and the preset threshold value, a diagnosis result indicating whether the machine tool to be tested is normal is generated. The method for diagnosing a machine tool spindle according to claim 1, wherein step (B) includes the following sub-steps: (B-1) Convert these vibration signals into frequency domain values; (B-2) Obtain multiple key frequency domain values from these frequency domain values; and (B-3) Perform filtering and outlier processing on these key frequency domain values to obtain the frequency domain characteristic values. The method for diagnosing a machine tool spindle according to claim 1, wherein, in step (C), the time-domain characteristic values include a kurtosis value, a crest factor value, a skewness value, a root mean value, and a variation Value, and at least two of a standard deviation value. The method for diagnosing a machine tool spindle according to claim 1, wherein step (E) includes the following sub-steps: (E-1) For each analysis data, normalize the analysis feature value of the analysis data to obtain multiple normalized values; and (E-2) For each analysis data, establish the Gaussian model according to the normalized value corresponding to the analysis data. The method for diagnosing a machine tool spindle according to claim 1, wherein step (H) includes the following sub-steps: (H-1) Determine whether the difference value is less than the preset threshold; (H-2) When it is determined that the difference value is less than the preset threshold value, the diagnosis result indicating that the machine tool to be tested is normal is generated; and (H-3) When it is determined that the difference value is not less than the preset threshold value, a diagnosis result indicating the abnormality of the machine tool under test is generated.

Images