Sign detection and diagnostic device and method

Abstract

To specify a time point when wear or chipping of a tool in a machine tool occurs.SOLUTION: A sign detection and diagnostic device 1 comprises: a data acquisition unit 10 which acquires time-series data of a load current supplied to a motor of a machine tool 4; a reference data generation unit 12 which calculates, for each time, reference data for kernel density estimation on the basis of first batch data extracted from the time-series data during the reference data creation; and an inspection execution unit 13 which calculates, for each time, a kernel density estimation amount on the basis of second batch data extracted from the time-series data at the time of an inspection and the reference data, and calculates, for each time, a score indicating the degree of deviation of the second batch data from the first batch data on the basis of the estimation amount.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a prognostic detection and diagnosis device and method for detecting abnormalities in machine tools.

Background Art

[0002] When machining materials with a machine tool, defective products may occur due to tool wear or chipping. Conventionally, a technique is known in which the current value of a motor connected to the spindle of a machine tool or the like is measured, and the entire batch is evaluated based on this change to detect tool wear (see Patent Document 1). In addition, a technique is known in which tool abnormalities are detected for each type of tool by using the current value of the motor that controls tool change together with the current value of the motor connected to the spindle of the machine tool or the like (see Patent Document 2).

[0003] In the technique disclosed in Patent Document 1, although evaluation of the entire batch is possible, there is a problem that the location of tool wear or chipping cannot be specified in a batch consisting of a plurality of processes.

[0004] In addition, in the technique disclosed in Patent Document 2, there is a problem that a measuring device for the motor that controls tool change is separately required to cut out a plurality of processing sections in a batch. Furthermore, since the section cutting is performed for each tool change, there is a problem that continuous different processes using the same tool cannot be individually extracted.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0006] The present invention has been made to solve the above problems, and an object thereof is to provide a prognostic detection and diagnosis apparatus and method capable of specifying the time point when wear or chipping of a tool of a machine tool occurs.

Means for Solving the Problems

[0007] The prognostic detection and diagnosis apparatus of the present invention includes a data acquisition unit configured to acquire time-series data of a load current supplied to a motor of a machine tool, a data storage unit configured to store the time-series data, and a reference data creation unit configured to calculate reference data for performing kernel density estimation for each time based on first batch data cut out from the time-series data at the time of creating the reference data, and a second batch data cut out from the time-series data at the time of inspection and the reference data, and calculates a kernel density estimation amount for each time, and based on this estimation amount, a score indicating the degree of deviation of the second batch data from the first batch data is calculated for each time. It is characterized by comprising an inspection execution unit configured to do so.

[0008] Further, one configuration example of the prognostic detection and diagnosis apparatus of the present invention is characterized in that, among the batch data, an offset value is detected for a region where the current value is equal to or less than an offset threshold value and the state where the current value is equal to or less than the offset threshold value continues for a predetermined time or more, and a data correction execution unit configured to subtract or add the offset value from the batch data is further provided. Further, in one configuration example of the prognostic detection and diagnosis apparatus of the present invention, the data correction execution unit corrects the time positions of a plurality of the first batch data to match, and corrects the second batch data so as to match the time positions of these first batch data. It is characterized by being like this.

[0009] In addition, in one configuration example of the sign detection and diagnosis apparatus of the present invention, the reference data creation unit stores the feature amounts of the plurality of first batch data and the reference data in the data storage unit in association with the work type data input by the user. The inspection execution unit determines the work type at the time of inspection based on the feature amounts of the first batch data and the second batch data, and calculates the kernel density estimator based on the reference data corresponding to the work type at the time of inspection and the second batch data. In addition, in one configuration example of the sign detection and diagnosis apparatus of the present invention, the feature amounts of the first batch data are the average waveform and the average cycle length of these batch data. The inspection execution unit performs a first work type determination by comparing the average cycle length with the cycle length of the second batch data. When there are two or more work type candidates in the first work type determination, a second work type determination is performed based on the correlation coefficient or cross-correlation between the average waveform and the second batch data. When there are two or more work type candidates in the second work type determination, the kernel density estimator is calculated for each work type candidate remaining in the second work type determination, and the work type at the time of inspection is determined based on this kernel density estimator.

[0010] In addition, in one configuration example of the sign detection and diagnosis apparatus of the present invention, the reference data creation unit detects a processing section in the first batch data where work processing is estimated to be performed. The inspection execution unit calculates the kernel density estimator only for the section at the same time position as the processing section in the second batch data. Furthermore, in one configuration example of the symptom detection and diagnosis device of the present invention, the data acquisition unit takes in the time series data and acquires workpiece type data from the machine tool, the reference data creation unit associates the calculated reference data with the type data acquired by the data acquisition unit when the reference data was created and stores them in the data accumulation unit, and the inspection execution unit acquires the reference data corresponding to the type data acquired by the data acquisition unit during inspection from the data accumulation unit and calculates a kernel density estimator based on this reference data and the second batch data.

[0011] Furthermore, the sign detection and diagnosis method of the present invention is characterized by including a first step of acquiring time series data of a load current supplied to a motor of a machine tool, a second step of calculating reference data for each time period for performing Kernel density estimation based on a first batch of data extracted from the time series data at the time of creating reference data, a third step of calculating a Kernel density estimator for each time period based on the reference data and second batch of data extracted from the time series data at the time of inspection, and a fourth step of calculating a score for each time period indicating the degree of deviation of the second batch of data from the first batch of data based on the Kernel density estimator. Effect of the Invention

[0012] According to the present invention, reference data for performing Kernel density estimation is calculated for each time based on the first batch data extracted from the time series data when the reference data was created, a Kernel density estimate is calculated for each time based on the second batch data extracted from the time series data during inspection and the reference data, and a score indicating the degree of deviation of the second batch data from the first batch data is calculated for each time based on this estimate, thereby making it possible to identify the time point when wear or chipping of the tool of the machine tool is likely to have occurred. As a result, the present invention makes it possible to improve operations, such as reviewing the machining procedures within a batch. [Brief description of the drawings]

[0013]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

Figure 23

Figure 24

Figure 25

Figure 26

Figure 27

Figure 28

Figure 29

Mode for Carrying Out the Invention

[0014] [First Embodiment] Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a machine tool inspection system according to the first embodiment of the present invention. The machine tool inspection system includes an omen detection / diagnosis device 1, a CT (Current Transformer) 2 that is a current transformer for converting the load current supplied to the motor that drives the workpiece of the machine tool 4 into a current of a magnitude that can be handled by the omen detection / diagnosis device 1, and a display 3 for displaying the inspection results and the like by the omen detection / diagnosis device 1.

[0015] The machine tool 4 includes a tool (not shown) for machining a workpiece, a motor 40 for driving the workpiece, a control unit 41 for controlling the motor 40, and a CNC (Computer Numerical Control) 42 for controlling the entire machine tool.

[0016] The symptom detection and diagnosis device 1 includes a data acquisition unit 10 that acquires time series data of a load current waveform supplied to a motor 40 of a machine tool 4, a data storage unit 11 that stores the time series data, a reference data creation unit 12 that calculates reference data for performing Kernel density estimation for each time based on batch data for reference data (first batch data) extracted from the time series data at the time of creating the reference data, an inspection execution unit 13 that calculates a Kernel density estimate for each time based on batch data for inspection (second batch data) extracted from the time series data at the time of inspection and the reference data, and calculates a score for each time based on this estimate that indicates the degree of deviation of the batch data for inspection against the batch data for reference data, and a display data generation unit 14 that generates data to be displayed on a display 3.

[0017] First, a description will be given of the operation of the sign detection and diagnosis device 1 when creating reference data. Fig. 2 is a flowchart explaining the operation of the sign detection and diagnosis device 1 when creating reference data. A user of the symptom detection and diagnosis device 1 instructs the symptom detection and diagnosis device 1 to create reference data while the machine tool 4 is in a normal state.

[0018] When the data acquisition unit 10 of the symptom detection and diagnosis device 1 receives an instruction from a user, it acquires time series data of the load current waveform supplied to the motor 40 of the machine tool 4 via the CT 2 (step S100 in FIG. 2). The time series data for reference data acquired by the data acquisition unit 10 is stored in the data accumulation unit 11.

[0019] The reference data creation unit 12 of the omen detection and diagnosis device 1 extracts batch data for reference data from the time-series data for reference data stored in the data storage unit 11 according to preset conditions (step S101 in FIG. 2). The reference data creation unit 12 extracts, for example, the time-series data for one batch from the start point of processing to the end point of processing as the batch data for reference data. A timing signal indicating approximately one batch can be obtained from the CNC 42. Alternatively, the reference data creation unit 12 may determine a section where the current value is equal to or greater than a certain value and obtain the cycle length L. For example, when the current value becomes 0.1 A or more, it is determined as the start of one batch, and when the state where the current value is 0.1 A or less continues for 1 second or more, it is determined as the end of one batch, whereby the cycle length L can be obtained. The data acquisition unit 10 and the reference data creation unit 12 repeatedly execute the processes of steps S100 and S101 until they finish extracting batch data for a preset number of times.

[0020] Next, after the reference data creation unit 12 finishes extracting batch data for a preset number of times (YES in step S102 in FIG. 2), it calculates the bandwidth h and the standard deviation σ necessary for calculating the kernel density estimator executed during inspection based on the batch data for reference data (step S103 in FIG. 2). For the sample data x1, x2, ···, x n Regarding, the function f KDE (x) for calculating the kernel density estimator with the kernel function K(x) and the bandwidth h as the mediating variable is given by the following formula.

[0021]

Equation

[0022] The kernel function K(x) is given by Equation (2).

[0023]

Equation

[0024] The bandwidth h is given by Equation (3).

[0025]

Equation

[0026] Equations (1) to (3) mean that the function f KDE (x) is approximated by the sum of the same number of kernel functions K(x) as the sample data. The IQR (InterQuartile Range) in Equation (3) is the difference between the 75% percentile and the 25% percentile of the sample in terms of the interquartile range. min(a, b) is a function that takes the smaller value of a and b.

[0027] The reference data creation unit 12 calculates the standard deviation σ for each time of the sample data and the bandwidth h for each time based on the sample data at times before and after the time to be calculated. Here, the time refers to, for example, the elapsed time with the start time of processing as time 0. The sample data refers to the current value at a certain time in the batch data. The number of samples n is the value obtained by multiplying the width before and after the time for which the standard deviation σ and the bandwidth h are to be calculated (the number of sample data points at times before and after) by the number of batches of the batch data (the number of batches captured by the data acquisition unit 10).

[0028] Thus, the creation of reference data is completed. The calculation results of the reference data (standard deviation σ and bandwidth h) calculated by the reference data creation unit 12 are stored in the data storage unit 11. When reference data creation is performed using a data group as shown in Fig. 3(A), the reference data creation result as shown in Fig. 3(B) is obtained.

[0029] Next, the operation of the prognostic detection and diagnosis device 1 during inspection will be described. Fig. 4 is a flowchart for explaining the operation of the prognostic detection and diagnosis device 1 during inspection. The user of the prognostic detection and diagnosis device 1 instructs the prognostic detection and diagnosis device 1 to execute an inspection of the machine tool 4.

[0030] When the data acquisition unit 10 of the symptom detection and diagnosis device 1 receives an instruction from a user, it acquires time-series data of the load current waveform supplied to the motor 40 of the machine tool 4 via the CT 2 in the same manner as when the reference data was created (step S200 in FIG. 4). The time-series data for inspection acquired by the data acquisition unit 10 is stored in the data accumulation unit 11.

[0031] The inspection execution unit 13 of the symptom detection and diagnosis device 1 extracts inspection batch data according to preset conditions from the inspection time-series data stored in the data storage unit 11 in the same manner as when the reference data was created (step S201 in FIG. 4). Then, the inspection execution unit 13 calculates a kernel density estimator f based on the reference data (standard deviation σ and bandwidth h) and the inspection batch data extracted in step S201. KDE (x) is calculated for each time (step S202 in FIG. 4). As in the case of creating the reference data, the time means the elapsed time with the processing start time being time 0, for example.

[0032] The inspection execution unit 13 substitutes the sample data at the calculation target time in the inspection batch data extracted in step S201 and the reference data (standard deviation σ and bandwidth h) corresponding to the calculation target time into equations (1) and (2) to obtain a kernel density estimator f KDE All we need to do is calculate (x).

[0033] Next, the inspection execution unit 13 calculates the kernel density estimator f KDE Based on (x), a score SC indicating the degree of deviation of the inspection sample data from the reference sample data is calculated for each time (step S203 in FIG. 4). The score SC is obtained, for example, by the following formula.

[0034]

number

[0035] A higher score SC indicates a greater deviation of the inspection sample data from the reference sample data. When wear or the like occurs in the tool, the load current changes, so an abnormality can be detected based on the score SC. The time series data of the score SC is stored in the data accumulation unit 11.

[0036] The display data generating unit 14 of the sign detection and diagnosis device 1 associates, for example, the time series data of the score SC with the time series data for inspection taken in by the data acquiring unit 10, turns it into a graph, and displays it on the display device 3 (FIG. 4, step S204). The user can understand the presence or absence of an abnormality in the machine tool 4 by checking the graph displayed on the display device 3.

[0037] An example of the test results according to this embodiment is shown in Figures 5(A), 5(B) and 6. Here, Figure 5(A) is the time series data for reference data, and Figure 5(B) is the time series data for test. The peak value of the time series data for test is different from that of the time series data for reference data in the range around 15 seconds shown by the dashed line. The calculation result of the score SC according to this embodiment is shown in Figure 6, and it can be seen that the value of the score SC is high in the range around 15 seconds shown by the dashed line.

[0038] As described above, in this embodiment, it is possible to identify the time when tool wear or chipping is likely to occur. As a result, in this embodiment, it is possible to improve operations, such as reviewing the machining procedures within a batch.

[0039] [Second Example] Next, a second embodiment of the present invention will be described. In this embodiment, the configuration of the machine tool inspection system is similar to that in the first embodiment, so the same reference numerals as in FIG. In this embodiment, the data acquisition unit 10 of the symptom detection and diagnosis device 1 takes in time-series data of the load current waveform via the CT 2 and also acquires type data of the workpiece, which is the object to be machined, from the CNC 42 .

[0040] The reference data creation unit 12 of the omen detection and diagnosis device 1 stores the calculated reference data (standard deviation σ and bandwidth h) in association with the product type data acquired by the data acquisition unit 10 during the creation of the reference data in the data storage unit 11.

[0041] The inspection execution unit 13 of the omen detection and diagnosis device 1 acquires the reference data (standard deviation σ and bandwidth h) corresponding to the product type data acquired by the data acquisition unit 10 during the inspection from the data storage unit 11, and based on this reference data and the batch data cut out during the inspection, calculates the kernel density estimator f KDE (x).

[0042] Other operations are the same as those in the first embodiment. Thus, in this embodiment, the present invention can be applied to the machine tool 4 that processes a plurality of types of workpieces, and abnormalities of the machine tool 4 can be detected for each type of workpiece.

[0043] [Third Embodiment] Next, a third embodiment of the present invention will be described. Also in this embodiment, since the configuration of the machine tool inspection system is the same as that in the first and second embodiments, it will be described using the reference numerals in FIG. 1. In calculating the score value in the inspection process, if the score calculation is performed using only the values in the section where the workpiece is actually being processed, excluding the section without a signal, it is considered that the tendency such as wear can be clearly obtained without being affected by the noise in the section without a signal. Therefore, in the batch data, it is desirable to estimate the section where the workpiece is actually being processed and the section without a signal where no processing is being performed.

[0044] FIG. 7 is a flowchart for explaining the operation during the creation of the reference data of the omen detection and diagnosis device 1 in this embodiment. The processes in steps S100 to S102 in FIG. 7 are the same as those in the first and second embodiments. When the workpiece is being processed, the load current of the motor 40 of the machine tool 4 increases. The reference data creation unit 12 of the omen detection and diagnosis device 1 detects the point where the load current increases as the section where processing is estimated to be performed, and cuts out the processing section from the batch data for reference data (step S104 in FIG. 7).

[0045] FIG. 8 is a flowchart for explaining the machining section estimation process by the reference data creation unit 12. The reference data creation unit 12 calculates a threshold value TH from the batch data for reference data (step S300 in FIG. 8). The reference data creation unit 12 sets, for example, the current value that is half of the current peak value in the time-series data for one batch as the threshold value TH. Alternatively, the reference data creation unit 12 may classify the time-series data for one batch into two classes by, for example, Otsu's binarization method, and set the current value at the midpoint between the centroids of the two classes as the threshold value TH. Alternatively, the reference data creation unit 12 may create a histogram of the current values in the time-series data for one batch and set the median value as the threshold value TH.

[0046] Subsequently, as shown in FIG. 9(A), the reference data creation unit 12 detects, in the batch data for reference data, the section where the current value is equal to or greater than the threshold value TH as the machining section SP (step S301 in FIG. 8).

[0047] Furthermore, if there is a region in the batch data for reference data where the interval between the machining sections SP is vacant by a specified time width or more (YES in step S302 in FIG. 8), the reference data creation unit 12 returns to step S300. In this case, the reference data creation unit 12 calculates the threshold value TH for this region from the batch data in the region detected in step S302 (step S300).

[0048] Then, the reference data creation unit 12 detects, in the batch data in the detected region, the section where the current value is equal to or greater than the threshold value TH as the machining section SP (step S301). In the example of FIG. 9(B), machining sections SP are newly detected around 30 seconds and from 40 seconds to around 45 seconds.

[0049] The processes of steps S300 to S302 are repeatedly executed for a time width equal to or longer than a previously specified time width until there is no region where the intervals between the processing sections SP are empty. When the determination in step S302 is NO, a detection result as shown in FIG. 10 is obtained. The reference data creation unit 12 performs the above-described processing section estimation process for each batch data for reference data.

[0050] Next, the reference data creation unit 12 calculates the bandwidth h and the standard deviation σ for the batch data within the processing section SP for each processing section SP and for each time (step S103a in FIG. 7). Thus, the creation of the reference data is completed.

[0051] FIG. 11 is a flowchart for explaining the operation during the inspection of the prediction detection / diagnosis apparatus 1 of the present embodiment. The processes of steps S200 and S201 in FIG. 11 are the same as those in the first and second embodiments. The inspection execution unit 13 of the prediction detection / diagnosis apparatus 1 calculates the kernel density estimator f KDE (x) for each time based on the reference data (standard deviation σ and bandwidth h) and the batch data for inspection cut out in step S201 (step S202a in FIG. 11).

[0052] However, in the present embodiment, the reference data (standard deviation σ and bandwidth h) is calculated only for within the processing section SP by the reference data creation unit 12. Therefore, the inspection execution unit 13 calculates the kernel density estimator f only for the section where the elapsed time from the start of the batch data for inspection is the same as the processing section SP among the batch data for inspection. KDE (x).

[0053] Subsequently, the inspection execution unit 13 calculates a score SC indicating the degree of deviation of the sample data for inspection with respect to the sample data for reference data for each time based on the calculated kernel density estimator f KDE (x) (step S203a in FIG. 11). In the present embodiment, the inspection execution unit 13 calculates the score SC only for the section at the same time position as within the processing section SP. The process of step S204 in FIG. 11 is the same as those in the first and second embodiments.

[0054] As described above, in this embodiment, by estimating the machining section SP and calculating the score SC only within the machining section SP, it is possible to detect an abnormality in the machine tool 4 without being affected by noise in a signal-free section where machining is not performed.

[0055] [Fourth Embodiment] Next, a fourth embodiment of the present invention will be described. FIG. 12 is a block diagram showing the configuration of a machine tool inspection system according to the fourth embodiment of the present invention. The machine tool inspection system of this embodiment is composed of a sign detection and diagnosis device 1a, a CT 2, and a display 3.

[0056] The sign detection and diagnosis device 1a includes a data acquisition unit 10, a data storage unit 11, a reference data creation unit 12, an inspection execution unit 13, a display data generation unit 14, and a data correction execution unit 15.

[0057] When a DC offset component is superimposed on the batch data, it affects the results of the reference data creation process and the inspection process. In the reference data creation process, as shown in FIG. 13(A), when a data group with a small difference in offset amounts between a plurality of batch data is used, the reference data creation result as shown in FIG. 13(B) is obtained. On the other hand, as shown in FIG. 13(C), when a data group with a large difference in offset amounts between a plurality of batch data is used, the reference data creation result as shown in FIG. 13(D) is obtained. If the reference data creation results are different, the kernel density estimators calculated in the inspection process will also be different.

[0058] Even if a data group with a small difference in offset amounts is used in the reference data creation process, the calculation results of the kernel density estimators are different between the case of using data with a small difference in offset amounts from the batch data used in the reference data creation process and the case of using data with a large difference in offset amounts in the inspection process. For the above reasons, it is desirable to appropriately remove the offset component of the batch data.

[0059] 14 is a flow chart for explaining the operation of the sign detection and diagnosis device 1a of this embodiment when creating reference data. The processes of steps S100 to S102 and S104 in FIG. 14 are the same as those in the third embodiment.

[0060] The data correction execution unit 15 of the sign detection and diagnosis device 1a removes the offset components of the batch data for the reference data extracted in step S101 (step S105 in FIG. 14). FIG. 15 is a flowchart illustrating the offset removal process performed by the data correction execution unit 15.

[0061] The data correction execution unit 15 instructs the display data generation unit 14 to display the candidate area CA used to remove the offset component in the batch data for reference data on the display 3 (step S400 in FIG. 15). The data correction execution unit 15 detects, in the batch data for reference data, an area where the current value is equal to or less than a preset offset threshold OTH (OTH>0) and where the state where the current value is equal to or less than the offset threshold OTH continues for a predetermined time or more, as the candidate area CA. The offset threshold OTH is set to a value close to 0.

[0062] 16 shows an example of the candidate areas CA displayed on the screen 30 of the display device 3. The user of the sign detection and diagnosis device 1a selects one of the displayed candidate areas CA or specifies an arbitrary range.

[0063] The data correction execution unit 15 determines the offset value Ioff based on the batch data in the area specified by the user (step S401 in FIG. 15). The data correction execution unit 15 sets the current value at the point where the gradient of the current value is the smallest among the batch data in the area specified by the user as the offset value Ioff. Alternatively, the data correction execution unit 15 may set the most frequent value of the current value in the area specified by the user as the offset value Ioff.

[0064] The data correction execution unit 15 removes the offset component by subtracting the offset value Ioff from the batch data for the reference data (step S402 in FIG. 15). The data correction execution unit 15 performs the above offset removal process for each batch data for the reference data.

[0065] The reference data creation unit 12 of the present embodiment calculates the bandwidth h and the standard deviation σ for the batch data within the processing section SP after the offset removal process for each processing section SP and for each time (step S103b in FIG. 14). Thus, the creation of the reference data is completed.

[0066] FIG. 17 is a flowchart for explaining the operation during the inspection of the prognostic detection and diagnosis device 1a of the present embodiment. The processes in steps S200 and S201 in FIG. 17 are the same as those in the first to third embodiments. The data correction execution unit 15 of the prognostic detection and diagnosis device 1a removes the offset component of the batch data for inspection cut out in step S201 (step S205 in FIG. 17). The offset removal process at this time is the same as that at the time of creating the reference data. As the offset value Ioff, the value obtained at the time of creating the reference data may be used.

[0067] The inspection execution unit 13 of the prognostic detection and diagnosis device 1a calculates the kernel density estimator f KDE (x) for each time based on the reference data (standard deviation σ and bandwidth h) and the batch data after the offset removal process in step S205 (step S202b in FIG. 17). Similar to the third embodiment, the inspection execution unit 13 calculates the kernel density estimator f KDE (x) only for the section at the same time position as the processing section SP. The processes in steps S203a and S204 in FIG. 17 are the same as those in the third embodiment.

[0068] For example, even if a data group in which the difference in the offset amounts between a plurality of batch data is large as shown in FIG. 18(A) is used for creating the reference data, if the offset component is removed according to the present embodiment, batch data as shown in FIG. 18(B) can be obtained.

[0069] As described above, in this embodiment, since the offset component superimposed on the batch data can be removed, the influence of the offset component on the results of the reference data creation process and the inspection process can be reduced.

[0070] In the above description, an example in which the data correction execution unit 15 is applied to the third embodiment has been described, but it may also be applied to the first and second embodiments. That is, the process of step S104 in FIG. 14 may be omitted. In this case, the reference data creation unit 12 may calculate the bandwidth h and the standard deviation σ for the batch data for the reference data after the offset removal process at each time. Further, the inspection execution unit 13 calculates the kernel density estimator f KDE (x) at each time based on the reference data (standard deviation σ and bandwidth h) and the batch data for inspection cut out in step S201 and offset-removed in step S205.

[0071] [Fifth Embodiment] Next, a fifth embodiment of the present invention will be described. Also in this embodiment, the configuration of the machine tool inspection system is the same as that of the fourth embodiment, so it will be described using the reference numerals in FIG. 12. In the calculation of the reference data in the reference data creation process and the calculation of the score value in the inspection process, if jitter occurs in the batch data, it will affect the results of the reference data creation process and the inspection process. Therefore, it is desirable to appropriately correct the jitter.

[0072] FIG. 19 is a block diagram showing the configuration of the data correction execution unit 15 of this embodiment. The data correction execution unit 15 of this embodiment is composed of a cross-correlation calculation unit 150 and a correction unit 151.

[0073] FIG. 20 is a flowchart for explaining the operation at the time of creating reference data of the omen detection / diagnosis device 1a of this embodiment. The processes of steps S100 to S102 and S104 in FIG. 20 are the same as those in the fourth embodiment. The data correction execution unit 15 of this embodiment removes the offset component of the batch data for reference data cut out in step S101 in the same manner as in the fourth embodiment (step S105 in FIG. 20), and corrects the jitter of the batch data after the offset removal process (step S106 in FIG. 20).

[0074] FIG. 21 is a flowchart for explaining the jitter correction process of the data correction execution unit 15. The cross-correlation calculation unit 150 of the data correction execution unit 15 calculates the cross-correlation between the processing section SP of one batch data (hereinafter, representative data) among the batch data for creating reference data cut out by the reference data creation unit 12 a predetermined number of times, and a block section at the same time position as the processing section SP of the representative data among the other batch data other than the representative data (step S501 in FIG. 21). The cross-correlation calculation unit 150 repeatedly calculates the cross-correlation between the representative data and the block section of the other batch data within a predetermined time range (for example, TC - window to TC + window) centered on the timing TC of the start position of the processing section SP of the representative data (window is a predetermined maximum time width).

[0075] After finishing calculating all the cross-correlations within the predetermined time range, the correction unit 151 of the data correction execution unit 15 obtains the deviation amount with respect to the reference position at the time position where the cross-correlation is maximum, with the timing TC of the start position of the processing section SP as the reference position (step S502 in FIG. 21).

[0076] The cross-correlation calculation unit 150 and the correction unit 151 perform the processes of steps S501 and S502 for each processing section SP of the representative data. When the calculation process of the deviation amount is completed for all the processing sections SP, the correction unit 151 moves the time position of the block section of the other batch data for which the cross-correlation has been calculated by the amount of deviation calculated for the processing section SP corresponding to this block section, and corrects the jitter (step S503 in FIG. 21). The correction unit 151 performs the jitter correction process of step S503 for each processing section SP.

[0077] When the correction unit 151 moves the block section of other batch data, for example, in the direction of time delay, it will delete the data after the trailing end of this block section by the amount of the shift. Furthermore, since a blank portion without data will be created in front of the leading end of the block section, the correction unit 151 interpolates the blank portion of the amount of the shift using the data immediately before the block section.

[0078] Also, when the correction unit 151 moves the block section of other batch data in the direction of time advancement, it will delete the data before the leading end of the block section by the amount of the shift. Furthermore, since a blank portion without data will be created after the trailing end of the block section, the correction unit 151 interpolates the blank portion of the amount of the shift using the data immediately after the block section.

[0079] The cross-correlation calculation unit 150 and the correction unit 151 perform the processing of steps S500 to S503 for each batch data other than the representative data among the batch data for creating reference data of a specified number of times cut out by the reference data creation unit 12. When the processing of steps S500 to S503 is completed for all of the batch data other than the representative data (YES in step S504 of FIG. 21), the jitter correction process ends. In this way, the start times of each processing section SP of the batch data for creating reference data are corrected to be synchronized. The batch data after the jitter correction process is stored in the data storage unit 11.

[0080] The reference data creation unit 12 of this embodiment calculates the bandwidth h and the standard deviation σ for each processing section SP and for each time for the batch data within the processing section SP after the offset removal process and further the jitter correction process (step S103c in FIG. 20). Thus, the creation of reference data is completed.

[0081] FIG. 22 is a flowchart for explaining the operation during the inspection of the prognostic detection and diagnosis device 1a of this embodiment. The processing of steps S200 and S201 in FIG. 22 is the same as that of the fourth embodiment. The data correction execution unit 15 of this embodiment removes the offset component of the inspection batch data cut out in step S201 in the same manner as in the fourth embodiment (step S205 in FIG. 22), and corrects the jitter of the batch data after the offset removal process (step S206 in FIG. 22).

[0082] FIG. 23 is a flowchart for explaining the jitter correction process during the inspection of the data correction execution unit 15. The cross-correlation calculation unit 150 of the data correction execution unit 15 calculates the cross-correlation between the processing section SP of the representative data and the block section at the same time position as the processing section SP of the representative data among the inspection batch data after the offset removal process (step S601 in FIG. 23). The cross-correlation calculation unit 150 repeatedly calculates the cross-correlation between the representative data and the block section of the inspection batch data within a predetermined time range (for example, TC - window to TC + window) centered on the leading position timing TC of the processing section SP of the representative data.

[0083] After finishing calculating all the cross-correlations within the predetermined time range, the correction unit 151 of the data correction execution unit 15 uses the leading position timing TC of the processing section SP as the reference position, and obtains the deviation amount with respect to the reference position at the time position where the cross-correlation is maximized (step S602 in FIG. 23).

[0084] The cross-correlation calculation unit 150 and the correction unit 151 perform the processes of steps S601 and S602 for each processing section SP of the representative data. When the calculation process of the deviation amount is completed for all the processing sections SP, the correction unit 151 moves the time position of the block section of the inspection batch data by the amount of deviation calculated for the corresponding processing section SP to correct the jitter (step S603 in FIG. 23). The correction unit 151 performs the jitter correction process of step S603 for each processing section SP.

[0085] When the processing of step S603 is completed for all of the processing sections SP of the representative data, the jitter correction process ends. In this way, the correction is made so that the start times of the respective processing sections SP of the representative data and the batch data for inspection are synchronized. The batch data for inspection after the jitter correction process is stored in the data storage unit 11.

[0086] Based on the reference data (standard deviation σ and bandwidth h) and the batch data after the jitter correction process in step S206, the inspection execution unit 13 of the omen detection and diagnosis device 1a calculates the kernel density estimator f KDE (x) at each time (step S202c in FIG. 22). Similar to the third embodiment, the inspection execution unit 13 calculates the kernel density estimator f KDE (x) only for the section at the same time position as the processing section SP. The processes of steps S203a and S204 in FIG. 22 are the same as those in the third embodiment.

[0087] As described above, in this embodiment, since the jitter of the batch data can be corrected, the influence of the jitter on the result of the reference data creation process and the inspection process can be reduced.

[0088] Note that the following process may be performed as the jitter correction process. FIG. 24 is a flowchart for explaining another example of the jitter correction process at the time of creating the reference data. The processes of steps S501 and S502 in FIG. 24 are as described with reference to FIG. 21. When the correction unit 151 of the data correction execution unit 15 finishes the calculation process of the deviation amount for the first processing section SP of the representative data, the correction unit 151 moves all the time positions of the block sections of the other batch data corresponding to each processing section SP of the representative data by the amount of deviation calculated for the first processing section SP of the representative data to correct the jitter (step S505 in FIG. 24).

[0089] Next, when the correction unit 151 finishes calculating the shift amount for the second processing section SP of the representative data, it corrects the jitter by shifting all the time positions of the block sections from the second onward of the other batch data corresponding to each processing section SP from the second onward of the representative data by the amount of shift calculated for the second processing section SP of the representative data (step S505).

[0090] Subsequently, when the correction unit 151 finishes calculating the shift amount for the third processing section SP of the representative data, it corrects the jitter by shifting all the time positions of the block sections from the third onward of the other batch data corresponding to each processing section SP from the third onward of the representative data by the amount of shift calculated for the third processing section SP of the representative data (step S505).

[0091] The cross-correlation calculation unit 150 and the correction unit 151 perform the processes of steps S501, S502, and S505 as described above for each processing section SP of the representative data. Further, the cross-correlation calculation unit 150 and the correction unit 151 perform the processes of steps S500 to S502 and S505 for each batch data other than the representative data among the batch data for creating reference data, which is cut out by the reference data creation unit 12 a specified number of times in advance. When the processes of steps S500 to S503 are completed for all of the batch data other than the representative data (YES in step S506 of FIG. 24), the jitter correction process at the time of creating the reference data ends.

[0092] FIG. 25 is a flowchart for explaining another example of the jitter correction process during inspection. The processes of steps S601 and S602 in FIG. 25 are as described in FIG. 23. When the correction unit 151 of the data correction execution unit 15 finishes calculating the shift amount for the first processing section SP of the representative data, it corrects the jitter by shifting all the time positions of the block sections of the batch data for inspection corresponding to each processing section SP of the representative data by the amount of shift calculated for the first processing section SP of the representative data (step S605 in FIG. 25).

[0093] Next, when the correction unit 151 finishes the calculation process of the deviation amount for the second processing section SP of the representative data, it moves all the time positions of the block sections from the second onward of the inspection batch data corresponding to each processing section SP from the second onward of the representative data by the amount of deviation calculated for the second processing section SP of the representative data, and corrects the jitter (step S605).

[0094] Subsequently, when the correction unit 151 finishes the calculation process of the deviation amount for the third processing section SP of the representative data, it moves all the time positions of the block sections from the third onward of the inspection batch data corresponding to each processing section SP from the third onward of the representative data by the amount of deviation calculated for the third processing section SP of the representative data, and corrects the jitter (step S605).

[0095] The cross-correlation calculation unit 150 and the correction unit 151 perform the processes of steps S601, S602, and S605 as described above for each processing section SP of the representative data. When the processes of steps S601, S602, and S605 are completed for all of the processing sections SP of the representative data (YES in step S600 of FIG. 25), the jitter correction process at the time of inspection is completed. The processes described with reference to FIGS. 24 and 25 are effective when the jitter is large.

[0096] [Sixth Embodiment] Next, a sixth embodiment of the present invention will be described. Also in this embodiment, since the configuration of the machine tool inspection system is the same as that of the fourth and fifth embodiments, the description will be made using the reference numerals in FIG. 12. When machining workpieces of a plurality of varieties with one machine tool 4, it is necessary to hold reference data for the plurality of varieties. In the second embodiment, the variety data is acquired from the CNC 42, but obtaining the variety data often requires modification of the machine tool 4, which takes time and technology. Therefore, a function for discriminating the variety of the workpiece from the batch data is useful.

[0097] FIG. 26 is a flowchart for explaining the operation during creation of reference data of the prediction detection / diagnosis apparatus 1a of the present embodiment. The processes of steps S100 to S102 and S104 to S106 in FIG. 26 are the same as those in the fifth embodiment.

[0098] The reference data creation unit 12 of the present embodiment calculates the bandwidth h and the standard deviation σ for each processing section SP and for each time with respect to the batch data for reference data within the processing section SP after offset removal processing and further jitter correction processing (step S103d in FIG. 26). At this time, the reference data creation unit 12 uses the average waveform of a plurality of pieces of batch data for reference data used for calculating the bandwidth h and the standard deviation σ, and the average cycle length L ave of the plurality of pieces of batch data for reference data as characteristic amounts of the batch data for reference data.

[0099] Since the reference data creation unit 12 can obtain the cycle length of the time-series data for one batch by acquiring a timing signal indicating one batch from the CNC 42, the average cycle length L ave can be calculated by obtaining the cycle length for each of the plurality of pieces of batch data. Alternatively, the reference data creation unit 12 may determine a section where the current value is equal to or greater than a certain value and obtain the cycle length L. For example, when the current value becomes 0.1 A or more, it is determined as the start of one batch, and when the state where the current value is 0.1 A or less continues for 1 second or more, it is determined as the end of one batch, whereby the cycle length L can be obtained. Then, the reference data creation unit 12 stores reference data including the bandwidth h, the standard deviation σ, the average waveform of the batch data, the average cycle length L ave and the offset value Ioff determined by the data correction execution unit 15 in the data storage unit 11 in association with the work type data.

[0100] The reference data creation unit 12 performs the above-mentioned reference data creation process for each type of workpiece. The workpiece type data when creating the reference data is manually input for each type of workpiece by the user of the symptom detection and diagnosis device 1a. This enables the reference data creation unit 12 to perform the reference data creation process for each type of workpiece. Examples of average waveforms of batch data for product types A to E are shown in Figs. 27(A) to 27(E), respectively.

[0101] Fig. 28 is a flow chart for explaining the operation during inspection of the sign detection and diagnosis device 1a of this embodiment. The processes of steps S200 and S201 in Fig. 28 are the same as those in the fifth embodiment. The inspection execution unit 13 of this embodiment obtains the cycle length L of the batch data for inspection extracted in step S201 (step S207 in FIG. 28). The inspection execution unit 13 can obtain the cycle length L of one batch of time-series data by acquiring a timing signal indicating one batch from the CNC 42. Alternatively, the inspection execution unit 13 may determine the section in which the current value is equal to or greater than a certain value to obtain the cycle length L. For example, the cycle length L can be obtained by determining that the start of one batch is reached when the current value is 0.1 A or greater, and determining that the end of one batch is reached when the current value remains below 0.1 A for one second or more.

[0102] Next, the inspection execution unit 13 compares the cycle length L obtained in step S207 with the average cycle length L included in the reference data for each type of workpiece. ave Compared with the average cycle length L ave The inspection execution unit 13 judges how many types of workpieces satisfy a predetermined type judgment condition (step S208 in FIG. 28). ave If it is within ±5%, this average cycle length L ave The workpiece type corresponding to the reference data including the above is determined to satisfy the type determination condition and is set as a first type candidate.

[0103] When there is one first-round variety candidate, the inspection execution unit 13 determines that the first-round variety candidate is the variety of the workpiece during inspection (step S209 in FIG. 28). Further, when there is no first-round variety candidate, the inspection execution unit 13 determines that the variety cannot be determined (step S210 in FIG. 28). In the examples of FIGS. 27(A) to 27(E), variety A, variety B, variety D, and variety E are the first-round variety candidates.

[0104] When the inspection execution unit 13 determines that there are two or more first-round variety candidates, the data correction execution unit 15 of this embodiment corrects the jitter of the inspection batch data cut out in step S201 (step S211 in FIG. 28). At this time, the data correction execution unit 15 generates the batch data after jitter correction for each first-round variety candidate. For example, when variety A, variety B, variety D, and variety E are the first-round variety candidates, the data correction execution unit 15 uses the average waveform of the batch data of variety A to perform jitter correction processing on the inspection batch data, the average waveform of the batch data of variety B to perform jitter correction processing on the inspection batch data, the average waveform of the batch data of variety D to perform jitter correction processing on the inspection batch data, and the average waveform of the batch data of variety E to perform jitter correction processing on the inspection batch data, and generates the inspection batch data after jitter correction processing.

[0105] The inspection execution unit 13 calculates the correlation coefficient or cross-correlation between the jitter-corrected inspection batch data and the average waveform of the batch data included in the reference data of the first-round variety candidate for each first-round variety candidate (step S212 in FIG. 28), and determines how many varieties of workpieces whose correlation coefficient or cross-correlation satisfies the predetermined variety determination condition (step S213 in FIG. 28).

[0106] The inspection execution unit 13 sets the variety candidates whose correlation coefficient or cross-correlation is equal to or greater than the reference value as the second-round variety candidates that satisfy the variety determination condition. In the examples of FIGS. 27(A) to 27(E), variety A is below the reference value, and variety B, variety D, and variety E are above the reference value. As a result, variety B, variety D, and variety E become the second-round variety candidates.

[0107] When there is one candidate variety in the second round, the inspection execution unit 13 determines that the candidate variety in the second round is the variety of the workpiece during inspection (step S209). Further, when there is no candidate variety in the second round, the inspection execution unit 13 determines that the variety cannot be determined (step S210).

[0108] When the inspection execution unit 13 determines that there are two or more candidate varieties in the second round, the data correction execution unit 15 of the present embodiment generates batch data with the offset component removed by subtracting the offset value Ioff from the batch data for inspection for each candidate variety in the second round (step S214 in FIG. 28).

[0109] At this time, the data correction execution unit 15 subtracts the offset value Ioff included in the reference data of the same candidate variety from the inspection batch data that has been subjected to jitter correction processing using the average waveform of the batch data of the candidate variety in the second round. Therefore, the data correction execution unit 15 subtracts the offset value Ioff of variety B from the inspection batch data that has been subjected to jitter correction processing using the average waveform of the batch data of variety B, subtracts the offset value Ioff of variety D from the inspection batch data that has been subjected to jitter correction processing using the average waveform of the batch data of variety D, and subtracts the offset value Ioff of variety E from the inspection batch data that has been subjected to jitter correction processing using the average waveform of the batch data of variety E. If the position of the offset point is not in the candidate variety in the second round due to reasons such as insufficient cycle length, that variety is excluded from the candidate varieties in the second round.

[0110] Next, the inspection execution unit 13 calculates the kernel density estimator f KDE (x) for each candidate variety in the second round and for each time (step S215 in FIG. 28). Similar to the third embodiment, the inspection execution unit 13 calculates the kernel density estimator f KDE (x) only for the section at the same time position as the processing section SP. The inspection execution unit 13 substitutes the batch data that has been subjected to offset removal processing using the offset value Ioff of variety B and the reference data of the same variety B into equations (1) and (2) to calculate the kernel density estimator fKDE It is only necessary to calculate (x). The same applies to varieties D and E.

[0111] Subsequently, the inspection execution unit 13 calculates the score SC for each second variety candidate and for each time based on the calculated kernel density estimator f KDE (x), for example, according to Equation (4), and calculates the average value of the score SC for one batch for each second variety candidate (step S216 in FIG. 28). For example, the average value of the score SC for variety B was 2.11, the average value of the score SC for variety D was 5.83, and the average value of the score SC for variety E was 6.82.

[0112] The inspection execution unit 13 determines how many variety candidates have an average value of the score SC below a certain value (step S217 in FIG. 28). If there is no second variety candidate with an average value of the score SC below the certain value, the inspection execution unit 13 determines that the variety cannot be determined (step S210).

[0113] If there is one or more variety candidates with an average value of the score SC below the certain value, the inspection execution unit 13 determines the second variety candidate with the smallest average value of the score SC as the variety of the work at the time of inspection, and adopts the score SC calculated for this variety candidate as the inspection result (step S218 in FIG. 28). In the above example, variety B is determined as the variety of the work at the time of inspection.

[0114] Note that if there is only one first variety candidate or second variety candidate, the process may be performed based on the variety determination result. That is, the data correction execution unit 15 removes the offset component by subtracting the offset value Ioff included in the reference data of the same variety candidate from the batch data for inspection that has been subjected to jitter correction processing using the average waveform of the batch data of the variety candidate determined as the variety of the work at the time of inspection (step S219 in FIG. 28). However, when there is only one first variety candidate, the data correction execution unit 15 performs jitter correction processing on the batch data for inspection using the average waveform of the batch data of this variety candidate, and then performs offset removal processing.

[0115] Next, the inspection execution unit 13 substitutes the offset-removed batch data for inspection and the reference data of the candidate product types determined as the product types of the workpiece during inspection into Formula (1) and Formula (2), and calculates the kernel density estimator f KDE (x) at each time (step S220 in FIG. 28). Subsequently, the inspection execution unit 13 may calculate the score SC at each time based on the calculated kernel density estimator f KDE (x), for example, by using Formula (4) (step S221 in FIG. 28).

[0116] When the product type of the workpiece during inspection is determined by the process of step S218, since the calculation processes of the kernel density estimator f KDE (x) and the score SC have already been completed, the score SC calculated for the candidate product types determined as the product types of the workpiece during inspection as described above may be adopted as the inspection result. The process of step S204 in FIG. 28 is the same as that in the first to fifth embodiments.

[0117] In this embodiment, it is not necessary to acquire product type data from the machine tool 4, and it is only necessary to acquire the load current. Further, even when workpieces of the same product type are machined by the same machine tool 4, there are variations in the motor current profile due to the influence of the machining environment, accidental errors, machining defects, etc. In this embodiment, by using the kernel density estimator and correlation between the reference data and the target data, it is possible to determine the product type while tolerating these variations. In the above description, an example in which this embodiment is applied to the fifth embodiment has been described, but the process of step S104 in FIG. 26 may be omitted.

[0118] The prognostic detection and diagnosis devices 1 and 1a described in the first to sixth embodiments can be realized by a computer including a CPU (Central Processing Unit), a storage device, and an interface, and a program for controlling these hardware resources. A configuration example of this computer is shown in FIG. 29.

[0119] The computer includes a CPU 200, a storage device 201, and an interface device (I / F) 202. A display 3, a machine tool 4, and the like are connected to the I / F 102. In such a computer, a program for implementing the sign detection and diagnosis method of the present invention is stored in the storage device 201. The CPU 200 executes the processes described in the first to sixth embodiments in accordance with the program stored in the storage device 201. Furthermore, at least a part of the sign detection and diagnosis devices 1, 1a may be implemented by hardware. [Industrial Applicability]

[0120] The present invention can be applied to a technique for detecting an abnormality in a machine tool. [Explanation of symbols]

[0121] 1, 1a...predictive detection and diagnosis device, 2...CT, 3...display, 4...machine tool, 10...data acquisition unit, 11...data storage unit, 12...reference data creation unit, 13...inspection execution unit, 14...display data generation unit, 15...data correction execution unit, 40...motor, 41...control unit, 42...CNC, 150...cross-correlation calculation unit, 151...correction unit.

Claims

1. A data acquisition unit configured to acquire time-series data of a load current supplied to a motor of a machine tool; A data storage unit configured to store the time-series data; A reference data creation unit configured to calculate reference data for kernel density estimation for each time based on first batch data cut out from the time-series data at the time of creating the reference data; A test execution unit configured to calculate a kernel density estimation amount for each time based on second batch data cut out from the time-series data at the time of inspection and the reference data, and based on this estimation amount, calculate a score indicating the degree of deviation of the second batch data from the first batch data for each time, wherein the prognostic detection and diagnosis device is characterized by comprising the test execution unit.

2. In the prognostic detection and diagnosis device according to Claim 1, an offset value is detected for a region in the batch data where the current value is equal to or less than an offset threshold value and the state where the current value is equal to or less than the offset threshold value continues for a predetermined time or more, and the prognostic detection and diagnosis device further comprises a data correction execution unit configured to subtract or add the offset value from the batch data.

3. In the prognostic detection and diagnosis device according to Claim 2, the data correction execution unit corrects the time positions of a plurality of the first batch data to match, and corrects the second batch data so that the time positions match those of the first batch data, wherein the prognostic detection and diagnosis device is characterized by this configuration.

4. In the prognostic detection and diagnosis device according to Claim 3, the reference data creation unit stores the feature amounts of a plurality of the first batch data and the reference data in the data storage unit in association with the work type data input by the user; the test execution unit determines the work type at the time of inspection based on the feature amounts of the first batch data and the second batch data, and calculates the kernel density estimation amount based on the reference data corresponding to the work type at the time of inspection and the second batch data, wherein the prognostic detection and diagnosis device is characterized by this configuration.

5. In the prognostic detection and diagnosis device according to Claim 4, the feature amounts of the first batch data are the average waveform and the average cycle length of these batch data. The inspection execution unit performs a first variety determination by comparing the average cycle length with the cycle length of the second batch data. When there are two or more variety candidates in the first variety determination, a second variety determination is performed based on the correlation coefficient or cross-correlation between the average waveform and the second batch data. When there are two or more variety candidates in the second variety determination, the kernel density estimator is calculated for each variety candidate remaining in the second variety determination, and the variety of the workpiece at the time of inspection is determined based on this kernel density estimator. A prognostic detection and diagnosis device characterized by this.

6. In the prognostic detection and diagnosis device according to any one of Claims 1 to 5, The reference data creation unit detects a processing section in the first batch data where it is estimated that the processing of the workpiece is being performed. The inspection execution unit calculates the kernel density estimator only for the section at the same time position as the processing section in the second batch data. A prognostic detection and diagnosis device characterized by this.

7. In the prognostic detection and diagnosis device according to Claim 1, The data acquisition unit captures the time series data and acquires the variety data of the workpiece from the machine tool. The reference data creation unit associates the calculated reference data with the variety data acquired by the data acquisition unit at the time of creating the reference data and stores it in the data storage unit. The inspection execution unit acquires the reference data corresponding to the variety data acquired by the data acquisition unit at the time of inspection from the data storage unit, and calculates the kernel density estimator based on this reference data and the second batch data. A prognostic detection and diagnosis device characterized by this.

8. A first step of acquiring time series data of the load current supplied to the motor of the machine tool; A second step of calculating reference data for kernel density estimation for each time based on the first batch data cut out from the time series data at the time of creating the reference data; A third step of calculating the kernel density estimator for each time based on the second batch data cut out from the time series data at the time of inspection and the reference data; A prognostic detection and diagnosis method characterized by including a fourth step of calculating a score indicating the degree of deviation of the second batch data from the first batch data for each time based on the kernel density estimator.

Images