Abnormality detection device, abnormality detection method, and program

Abstract

To appropriately detect an abnormality without removing a non-processing section in time-series data related to a motor acquired in a machine tool.SOLUTION: An abnormality detection device 2 comprises: a reception unit 21 which receives time-series data related to a motor in a machine tool; an acquisition unit 23 which acquires reference time-series data being the reference corresponding to a single cycle of repetitive processing in the machine tool and detection target time-series data being the detection target of an abnormality corresponding to the single cycle from the time-series data; a processing unit 24 which aligns the reference time-series data with the detection target time-series data acquired from the time-series data such that the total sum of the distances between the data of respective times is minimized in a state where a time shift equal to or less than a predetermined threshold is allowed; and an abnormality detection unit 25 which detects the abnormality in the detection target time-series data by comparing the aligned reference time-series data with detection target time-series data.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an abnormality detection device or the like that detects abnormalities based on time-series data acquired in a machine tool. [Background technology]

[0002] Conventionally, in machine tools that rotate a spindle to which either a workpiece or a tool is attached, and perform cutting processing on the workpiece using the tool, the current of the motor that drives the spindle, etc. is measured and used to detect abnormalities such as wear and breakage of the tool (see, for example, Patent Document 1).

[0003] For example, as a tool becomes worn, it becomes harder to cut compared to a normal tool that is not worn, so the machine tool controls the tool to cut with a greater force than a normal tool, and as a result, the current value of the spindle motor increases more than normal. Also, if the tool breaks, the tool will no longer contact the workpiece, so the current value of the spindle motor decreases more than normal. Therefore, by comparing the current value of the spindle motor with the normal current value, it is possible to detect abnormalities such as tool wear and tool breakage.

[0004] In mass-production machine tools, the same machining operations are repeated repeatedly, so tool abnormalities can be detected by obtaining and comparing the data for one cycle, from the start to the end of the repeated machining operations, from the time-series data of the current value of the motor that drives the spindle. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent Publication No. 2022-093204 Summary of the Invention [Problem to be solved by the invention]

[0006] FIG. 3 is a waveform diagram showing time-series data of the current value of the motor that rotates the spindle, acquired by a current sensor of the machine tool. The periods PE1 to PE4 shown in FIG. 3 are machining sections in which the machine tool cuts the workpiece to be machined. As shown in FIG. 3, the time-series data includes machining sections and non-machining sections in which the rotation speed of the spindle is increased or decreased before and after machining. As can be seen from FIG. 3, the waveform in the non-machining section is steep, so if there is even a slight variation in the waveform in the time direction, the difference between the two waveforms when comparing two time-series data for abnormality detection becomes large, and there is a possibility that an abnormality will be erroneously detected as occurring depending on the difference in the waveform in the non-machining section that is not related to the detection of an abnormality.

[0007] In order to avoid such false detection, it is conceivable to delete the waveform of the non-machining section, but with a wide variety of machine tools and machining methods, it is difficult to completely distinguish between the waveform of the machining section and the waveform of the non-machining section, and it is difficult to delete the waveform of the non-machining section. Also, in order to avoid false detection of an abnormality, it is conceivable to set a larger value for the threshold value for detecting an abnormality, but there is also a problem in that if the threshold value for detecting an abnormality is made larger, there is a possibility that an abnormality may be overlooked.

[0008] The present invention has been made to solve the above-mentioned problems, and has an object to provide an abnormality detection device, etc. that is capable of properly detecting abnormalities without removing non-processing sections. [Means for solving the problem]

[0009] In order to achieve the above object, an abnormality detection device according to one embodiment of the present invention includes a receiving unit that receives first time series data related to a first motor in a machine tool, an acquisition unit that acquires, from the first time series data, reference time series data that serves as a reference corresponding to one cycle of machining repeated in the machine tool and detection target time series data that is a target for detection of abnormalities corresponding to one cycle, a processing unit that aligns the reference time series data and the detection target time series data acquired from the first time series data so that the sum of the distances between the data at each time is minimized while allowing a time shift of less than a predetermined threshold, and an abnormality detection unit that detects abnormalities in the detection target time series data by comparing the reference time series data and the detection target time series data after the alignment.

[0010] With this configuration, by combining the reference time series data and the time series data to be detected, it is possible to eliminate time deviations below a threshold value for steep waveforms in the non-processed section, while not eliminating time deviations in the processed section that are greater than the threshold value. This reduces the possibility of erroneously detecting that an abnormality has occurred due to differences in the waveforms of the non-processed section, and as a result, it becomes possible to properly detect abnormalities.

[0011] In addition, in the anomaly detection device according to one aspect of the present invention, the processing unit may perform upsampling on the reference time series data and the detection target time series data acquired by the acquisition unit, and then combine both time series data.

[0012] With this configuration, it becomes possible to match two sets of time series data with higher accuracy.

[0013] In addition, in an abnormality detection device according to one aspect of the present invention, the receiving unit may also receive second time series data related to a second motor in the machine tool, the acquiring unit may also acquire reference time series data and detection target time series data for the second time series data, the processing unit may also match the reference time series data acquired from the second time series data with the detection target time series data, and the abnormality detection unit may also detect an abnormality in the detection target time series data by comparing the reference time series data acquired from the second time series data with the matched reference time series data and the detection target time series data.

[0014] With this configuration, it is possible to increase the amount of time-series data used for anomaly detection, thereby improving the accuracy of anomaly detection.

[0015] In addition, in an abnormality detection device according to one aspect of the present invention, the first time series data is data on a current of a spindle motor that rotates a spindle in a machine tool, and the second time series data is data on a current of a servo motor, the range of the first time series data is larger than the range of the second time series data, and the resolution of the first time series data may be the same as the resolution of the second time series data.

[0016] With this configuration, it is possible to use ranges appropriate for the first and second time-series data, respectively, and therefore it is possible to perform appropriate abnormality detection and also to prevent an increase in the amount of data.

[0017] In the abnormality detection device according to an aspect of the present invention, the threshold value may be shorter than the time of the shortest machining section in one cycle.

[0018] With this configuration, when an abnormality occurs, it is possible to avoid matching the abnormal waveform.

[0019] In the anomaly detection device according to an aspect of the present invention, the reference time series data may be time series data received by the receiving unit before fitting is performed.

[0020] With this configuration, it is possible to prevent the length of the reference time series data from changing.

[0021] In addition, an abnormality detection method according to one aspect of the present invention is an abnormality detection method processed using a receiving unit, an acquiring unit, a processing unit, and an abnormality detection unit, and includes the steps of: the receiving unit accepting first time series data related to the rotation of a spindle in a machine tool; the acquiring unit acquiring, from the first time series data, reference time series data that serves as a standard corresponding to one cycle of machining repeated in the machine tool, and detection target time series data that is to be detected for an abnormality corresponding to one cycle; the processing unit aligning the reference time series data and the detection target time series data acquired from the first time series data so that the sum of the distances between the data at each time is minimized while allowing a time shift of less than a predetermined threshold; and the abnormality detection unit detecting an abnormality in the detection target time series data by comparing the reference time series data and the detection target time series data after the alignment. Effect of the Invention

[0022] According to an abnormality detection device, etc., of one aspect of the present invention, it is possible to reduce the possibility of erroneously detecting that an abnormality has occurred due to differences in the waveform of the non-processed section, and as a result, it becomes possible to properly detect abnormalities. [Brief description of the drawings]

[0023] [Figure 1] FIG. 1 is a block diagram showing a configuration of an information processing system according to an embodiment of the present invention. [Diagram 2] A flowchart showing the operation of the abnormality detection device according to the embodiment. [Diagram 3] FIG. 2 is a diagram showing an example of time-series data in the embodiment; [Figure 4] FIG. 2 is a diagram showing an example of reference time series data and detection target time series data in the embodiment; [Diagram 5] FIG. 13 is a diagram for explaining matching of reference time series data and detection target time series data in the embodiment; [Figure 6] FIG. 13 is a diagram showing an example of a waveform in a non-processing section in the embodiment; [Figure 7] FIG. 13 is a diagram showing an example of a waveform in a processing section in the embodiment; [Figure 8] FIG. 2 shows an example of the configuration of a computer system according to the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0024] The anomaly detection device and anomaly detection method according to the present invention will be described below using embodiments. In the following embodiments, components and steps with the same reference numerals are the same or equivalent, and repeated description may be omitted. The anomaly detection device according to this embodiment appropriately detects anomalies by comparing two pieces of time series data to be compared after matching them for time differences of less than a threshold value.

[0025] 1 is a block diagram showing a configuration of an information processing system 100 according to the present embodiment. The information processing system 100 according to the present embodiment includes a machine tool 1 and an abnormality detection device 2. The machine tool 1 and the abnormality detection device 2 may be connected to each other by wire or wirelessly.

[0026] The machine tool 1 repeatedly performs machining on a workpiece. The machine tool 1 may be, for example, an NC lathe or a machining center. In the NC lathe, a spindle on which a workpiece is attached is rotated, a tool is brought into contact with the rotating workpiece, and the workpiece and the tool are moved relatively to perform machining such as cutting. In the NC lathe, a tool is switched by, for example, rotating a turret. On the other hand, in the machine tool 1, which is a machining center, a spindle on which a tool is attached is rotated, a rotating tool is brought into contact with the workpiece, and the workpiece and the tool are moved relatively to perform machining such as cutting. In the machining center, a tool is switched by, for example, an auto tool changer (ATC).

[0027] As an example, as shown in FIG. 1, the machine tool 1 may have a first sensor 11 for acquiring first time series data related to a first motor of the machine tool 1 and a second sensor 12 for acquiring second time series data related to a second motor different from the first motor. The time series data related to the motor may be, for example, time series data related to the rotation of the motor. The first and second motors of the machine tool 1 may be, for example, spindle motors that rotate a main shaft, or servo motors that move a workpiece or a tool. The spindle motor may, for example, rotate a workpiece or may rotate a tool used in milling or the like. The servo motor may be, for example, a servo motor for an X-axis, a Y-axis, a Z-axis, an A-axis, a B-axis, a C-axis, or the like. In this embodiment, the case where the first time series data is time series data of a current of a spindle motor for rotating a spindle, acquired by a first sensor 11 which is an ammeter, and the second time series data is time series data of a current of a servo motor for moving a workpiece or a tool, acquired by a second sensor 12 which is an ammeter, will be mainly described. However, as described later, the first and second time series data may be other time series data. Also, when an abnormality is detected only for the first time series data, the second time series data may not be acquired. In this case, the machine tool 1 may not have the second sensor 12. Also, when an abnormality is detected for three or more pieces of time series data, the machine tool 1 may have three or more sensors for acquiring the three or more pieces of time series data, respectively. It is preferable that the two or more pieces of time series data acquired in the machine tool 1 are synchronized. In this embodiment, the case where an abnormality is detected only for the first time series data will be mainly described, and the case where an abnormality is detected for the first and second time series data will be described later.

[0028] The abnormality detection device 2 includes a reception unit 21, a storage unit 22, an acquisition unit 23, a processing unit 24, an abnormality detection unit 25, and an output unit 26. The abnormality detection device 2 may be, for example, a general-purpose device such as a personal computer, or may be a dedicated device for detecting an abnormality using time-series data.

[0029] The reception unit 21 receives first time series data related to a first motor in the machine tool 1. As an example, the reception unit 21 may also receive second time series data related to a second motor in the machine tool 1. The received first and second time series data may be stored in the storage unit 22. The reception unit 21 may receive the first and second time series data, for example, in real time, or at regular or irregular time intervals. When the first and second time series data are received in real time, for example, the first and second time series data may be configured in the storage unit 22 by receiving a signal at each time point from the machine tool 1 and storing the signal in the storage unit 22. Even in this case, the reception unit 21 repeatedly receives each signal constituting the first and second time series data, and as a result, the reception unit 21 receives the first and second time series data.

[0030] Receiving unit 21 may receive information transmitted via, for example, a wired or wireless communication line. In this embodiment, a case will be mainly described in which the first and second time-series data transmitted from machine tool 1 are received by receiving unit 21. Note that receiving unit 21 may or may not include a device for receiving the data (for example, a modem, a network card, etc.). Also, receiving unit 21 may be realized by hardware, or may be realized by software such as a driver that drives a predetermined device.

[0031] The first and second time series data are stored in the storage unit 22. Other information may be stored in the storage unit 22. For example, reference time series data acquired from the first and second time series data, detection target time series data, and the like may be stored in the storage unit 22. The storage unit 22 is preferably realized by a non-volatile recording medium, but may also be realized by a volatile recording medium. The recording medium may be, for example, a semiconductor memory, a magnetic disk, or the like.

[0032] FIG. 3 is a waveform diagram showing the first time series data which is the current of the spindle motor. In FIG. 3, the horizontal axis is time, and the vertical axis is current. The waveform shown in FIG. 3 includes peaks corresponding to acceleration or deceleration of the spindle. FIG. 3 shows a case where the first time series data is time series data for one cycle, but it goes without saying that the first time series data may be time series data for a longer period. As described above, the periods PE1 to PE4 each correspond to a machining section. The machining section may be, for example, a period during which cutting is continuously performed on a workpiece using one or more tools. In addition, in the non-machining section between the machining sections, the machine tool 1 switches tools.

[0033] The acquiring unit 23 acquires, from the first time series data, reference time series data serving as a reference corresponding to one cycle of machining repeated in the machine tool 1, and detection target time series data serving as a detection target for anomaly corresponding to one cycle of machining. The acquiring unit 23 is capable of identifying the range of each one cycle included in the first time series data. Here, one cycle is a period from the start to the end of machining having a plurality of steps on a machining target in the machine tool 1. The acquiring unit 23 may identify the range of one cycle, for example, by identifying the start time of one cycle in the first time series data and identifying the end time of one cycle that is a predetermined time after the start time. The identification of the start time of one cycle may be performed, for example, by identifying the time when the spindle motor starts to rotate. The acquiring unit 23 may then acquire the reference time series data and the detection target time series data by extracting time series data corresponding to the identified range of one cycle from the first time series data.

[0034] The acquisition unit 23 may acquire time series data for each cycle from the beginning of the first time series data, for example. When an abnormality in wear of the tool is detected, the time series data corresponding to the first cycle of the first time series data may be the reference time series data, and the time series data corresponding to the latest cycle from the second cycle onward may be the detection target time series data. When an abnormality in breakage of the tool is detected, the time series data corresponding to the Nth cycle of the first time series data may be the reference time series data, and the time series data corresponding to the N+1th cycle may be the detection target time series data. N is an integer of 1 or more. In this way, it becomes possible to detect breakage of the tool without being affected by wear of the tool. The reference time series data and the detection target time series data may be appropriately stored in the storage unit 22. For example, when an abnormality in breakage is detected, the detection target time series data corresponding to the N+1th cycle of the first time series data is used as the reference time series data when the next abnormality in breakage is detected. Therefore, after acquiring the detection target time series data corresponding to the N+1th cycle of the first time series data, the acquiring unit 23 may accumulate the time series data corresponding to the N+1th cycle in the storage unit 22, and when detecting the next damage, read out the time series data corresponding to the N+1th cycle as reference time series data from the storage unit 22. In this way, the acquisition of the reference time series data may be, for example, reading out the reference time series data from the storage unit 22.

[0035] The processing unit 24 adjusts the reference time series data and the detection target time series data acquired from the first time series data so that the sum of the distances between the data at each time is minimized while allowing a time shift of a predetermined threshold or less. For example, this threshold is preferably set to a value that allows appropriate adjustment of the steep waveform of the non-machined section and does not allow adjustment of the difference in the waveform of the processed section when an abnormality occurs. As an example, the threshold may be shorter than the time of the shortest processed section in one cycle. By using such a threshold, for example, even when an abnormality occurs, it is possible to avoid adjustment to an abnormal waveform. The threshold is not particularly limited, but may be, for example, 5 seconds or less, 3 seconds or less, or 1 second or less. In addition, this threshold may be, for example, 0.1 seconds or more, or 0.5 seconds or more. Hereinafter, this adjustment process will be described.

[0036] If the reference time series data is P and the detection target time series data is Q, P and Q are as follows. i is the data at time i in the reference time series data P, and q j is data at time j in the detection target time series data Q. The indexes i and j indicating time take integer values ​​from 1 to m and from 1 to n, such as i=1,2,...,m and j=1,2,...,n, in the reference time series data P and the detection target time series data Q, respectively. m and n are integers equal to or greater than 1. Note that the larger the values ​​of i and j, the later the time. Since both the reference time series data P and the detection target time series data Q are time series data corresponding to one cycle, ideally the number of data included in both is the same (i.e., m=n), but in reality there may be a difference of several pieces (i.e., m≠n). P={p 1 ,p 2 ,…,p m} Q={q 1 ,q 2 ,…,q n}

[0037] Let D(P,Q) be the sum of the distances between the data for each time period related to the reference time series data P and the detection target time series data Q, in a combination that provides the smallest distance between the data for each time period. However, this sum is calculated while allowing for a time shift of a predetermined threshold or less. In other words, the reference time series data P and the detection target time series data Q are adjusted to accommodate time shifts of less than the threshold. Also, in a combination that provides the smallest distance between the data for each time period related to the reference time series data P and the detection target time series data Q, the times of the reference time series data P and the detection target time series data Q are in ascending or descending order. In other words, when a certain combination is (p i ,q j ), other combinations (p i+1 ,q α ), (p β ,q j+1 ), where α≧j and β≧i, where α and β are integers. D(P,Q) is given by the following equation. D(P,Q)=f(m,n)

[0038] Here, f(i,j) is expressed by the following formula: However, as described above, only time lags equal to or smaller than a predetermined threshold are allowed, so |ij|≦M, where M is an integer equal to or greater than 1. The larger M is, the larger the time lag is allowed. f(i,j)=d(p i ,q j )+min{f(i,j-1),f(i-1,j),f(i-1,j-1)}

[0039] Here, d(p i ,q j ) is the data p i and data q at time j j For example, d(p i ,q j )=│p i -q j │, it can be the absolute value of the difference between the two data, d(p i ,q j )=(pi -q j ) 2 As shown above, it may be the square of the difference between the two data, or it may be the value of another increasing function that uses the absolute value of the difference between the two data as an argument.

[0040] The sum D(P,Q) calculated as above is expressed as the distance d(p i ,q j ) to d(p 1 ,q 1 ) to d(p m ,q n ) is the result of adding up the data sets that correspond to each other through matching. D(P,Q)=d(p 1 ,q 1 )+…+d(p m ,q n )

[0041] The distance d(p i ,q j ) for which the distance is calculated i and data q j The pairs of i, j are data pairs that have become corresponding to each other through the fitting. Therefore, the fitting of the reference time series data P and the detection target time series data Q by the processing unit 24 may be to obtain a time pair (i, j) of data that has become corresponding to each other through the fitting.

[0042] The above-mentioned matching process will be described with reference to specific waveform diagrams. FIG. 4 is a waveform diagram showing the reference time series data P and the detection target time series data Q. FIG. 5 is a waveform diagram showing each data p included in the reference time series data P. 1 ,p 2 ,…,p 12 , and each data q included in the detection target time series data Q 1 ,q 2 ,…,q 12 4 and 5, m=n=12 is used for the convenience of explanation, but in reality, m and n are larger values. Also, as described above, m and n may be different.

[0043] For example, if M=2, then only time lags of │ij│≦2 are allowed, so data p i and data q j In other words, only correspondences corresponding to the white squares that are not hatched in FIG. 5 are allowed. And for all of the white squares, the distance between the data d(p i ,q j ) is calculated, and in the white squares in Figure 5, the lower left (p 1 ,q 1 ) square, and then the (p 12 ,q 12 ) squares, the distance d(p i ,q j The minimum sum of all the distances d(p i ,q j ) corresponding data p i and data q j The combinations of P and Q are shown in the squares as black circles, as shown in Figure 5. Also, the distances d(p i ,q j ) corresponding to the data set (p i ,q j ) is shown in Figure 4 by connecting both sets of data with a thin line segment.

[0044] In the examples of Figures 4 and 5, for k=1 to 5, data p k is data q k Supports data p 6 is data q 6 , q 7 For k = 7 to 10, the data p k is data q k+1 Supports data p 11 , p 12 is data q 12 As an example, this data p i and data q jThe information indicating the correspondence relationship may be a result of matching the reference time series data P with the detection target time series data Q. Also, the processing unit 23 may acquire a result of changing the detection target time series data so as to match it with the reference time series data.

[0045] The time series data to be detected is changed to match the reference time series data by changing each data q j The data p i For example, data q j And data p i If and correspond to each other, data q j may be used as the data for time i. In addition, multiple data q j ,q j+1 is the data p i If it corresponds to multiple data q j ,q j+1 The representative value of may be used as the data at time i. The representative value may be, for example, the average value, the maximum value, or the minimum value. j There are multiple data p i ,p i+1 If it corresponds to data q j may be data at time i and time i+1. For example, in the examples of Figs. 4 and 5, the processing unit 23 may change the detection target time series data Q as follows. Here, the detection target time series data after the change is Q c ={q c 1 ,q c 2 ,…,q c n}. q c k =q k (k=1~5) q c 6 =q 6 and q 7 Representative value of q c k =q k+1(k=7~11) q c 12 =q 12

[0046] Since the matching between the reference time series data and the detection target time series data is relative, the processing unit 23 may match the detection target time series data to the reference time series data, or may match the reference time series data to the detection target time series data. That is, for the matching, the detection target time series data may be shifted in the time direction, or the reference time series data may be shifted in the time direction. Also, for example, when the detection target time series data is matched to the reference time series data, the entire length of the detection target time series data may be expanded or contracted. Therefore, if the detection target time series data after the expansion or contraction is used as the next reference time series data, further expansion or contraction may occur, and the length of the time series data may change with each cycle. Therefore, it is preferable to use the original time series data, that is, the time series data before the matching that is accepted by the accepting unit 21, as the reference time series data. Therefore, for example, even when the detection target time series data is aligned with the reference time series data, the detection target time series data before the alignment may be stored in the memory unit 22, and in the next cycle, the detection target time series data before the alignment may be read out from the memory unit 22 and used as the reference time series data.

[0047] The anomaly detection unit 25 detects an anomaly in the detection target time series data by comparing the reference time series data after fitting with the detection target time series data. Comparing the reference time series data after fitting with the detection target time series data may mean, for example, comparing both the reference time series data and the detection target time series data after changing one of them so as to fit with the other, or may mean comparing the data according to the correspondence indicated by the fitting result when comparing both the time series data.

[0048] The above-mentioned reference time series data P and the detection target time series data Q after fitting cIn the example shown in FIG. 1, the anomaly detection unit 25 detects, for example, the reference time series data P={p 1 ,p 2 ,…,p n} and the detection target time series data Q after fitting c ={q c 1 ,q c 2 ,…,q c n} may be used to detect an abnormality. The abnormality detection unit 25 may detect, for example, an abnormality related to wear of a tool, an abnormality related to damage of a tool, or both of these abnormalities. The abnormality detection unit 25 may compare, for example, the reference time-series data after fitting with the detection target time-series data, and detect an abnormality if the difference between the two is greater than a threshold value, and may not detect an abnormality if not.

[0049] The comparison of the two time series data after fitting may be performed, for example, for each machining step. That is, a comparison for detecting an abnormality in the time series data may be performed for each section obtained by dividing one cycle into a plurality of sections for each tool switching timing. The tool switching timing may be detected, for example, using the current value of the tool switching motor in the machine tool 1. For example, the time when the tool switching motor is detected to have operated based on the current value of the tool switching motor may be set as the tool switching time. The time series data indicating the operating status of the tool switching motor may be acquired, for example, in the machine tool 1 and passed to the abnormality detection device 2. Then, the abnormality detection device 2 may detect the tool switching time using the time series data. In addition, the comparison of the two time series data after fitting may be performed, for example, by acquiring a feature value from each time series data and comparing the acquired feature values. The feature value is not particularly limited, and may be, for example, a representative value such as an average value, a median value, a minimum value, or a maximum value, a statistical value such as a variance or a standard deviation, information indicating the characteristics of the waveform shape of the time series data, or other feature values ​​indicating the characteristics of the time series data.

[0050] When an abnormality is detected by the abnormality detection unit 25, the output unit 26 may perform an output related to the detection of the abnormality. The output related to the detection of the abnormality may be, for example, an output of information indicating that an abnormality has been detected, or may be an output of an instruction corresponding to control performed in response to the detection of the abnormality, such as an instruction to stop the machine tool 1. When the abnormality detection unit 25 detects both an abnormality in tool wear and an abnormality in tool breakage, for example, different outputs may be performed depending on the content of the detected abnormality. As an example, when an abnormality in tool wear is detected, information indicating that may be output, and when an abnormality in tool breakage is detected, a command to stop the machine tool 1 may be output to the machine tool 1.

[0051] Here, this output may be, for example, a display on a display device (e.g., a liquid crystal display or an organic EL display), a transmission to a predetermined device via a communication line, a printing by a printer, an audio output by a speaker, storage on a recording medium, or a transfer to another component. Note that output unit 26 may or may not include a device that performs output (e.g., a display device or a printer). Also, output unit 26 may be realized by hardware, or may be realized by software such as a driver that drives those devices.

[0052] Next, the operation of the abnormality detection device 2 will be described with reference to the flowchart of FIG. (Step S101) The accepting unit 21 judges whether or not the first time series data has been accepted. If the first time series data has been accepted, the process proceeds to step S102, and if not, the process proceeds to step S103.

[0053] (Step S102) The receiving unit 21 accumulates the received first time-series data in the storage unit 22. Then, the process returns to step S101.

[0054] (Step S103) Acquiring unit 23 determines whether or not to detect an abnormality. If an abnormality is to be detected, proceed to step S104; if not, return to step S101. Acquiring unit 23 may, for example, determine whether to detect an abnormality at every predetermined period. The predetermined period may be, for example, the cycle of machining repeated in machine tool 1. Also, in this embodiment, the case where acquiring unit 23 determines the timing to detect an abnormality will be mainly described, but the timing to detect an abnormality may also be determined by abnormality detection unit 25, for example.

[0055] (Step S104) The acquiring unit 23 acquires reference time series data. When both wear and breakage abnormalities of the tool are detected, reference time series data corresponding to each abnormality detection may be acquired.

[0056] (Step S105) The acquisition unit 23 acquires detection target time-series data.

[0057] (Step S106) The processing unit 24 performs matching between the reference time series data and the detection target time series data. When two reference time series data are acquired, each of the reference time series data may be matched with the detection target time series data.

[0058] (Step S107) The abnormality detection unit 25 compares the reference time series data after fitting with the detection target time series data to determine whether or not there is an abnormality. If there is an abnormality, the process proceeds to step S108, and if not, the process returns to step S101. The detection of this abnormality may be, for example, detection of wear of the tool, detection of damage to the tool, or both. Furthermore, the abnormality detection unit 25 may detect an abnormality when there is a difference between the reference time series data after fitting and the detection target time series data that exceeds a threshold value.

[0059] (Step S108) The output unit 26 performs output regarding the detection of the abnormality. Then, the process returns to step S101. When an abnormality in the wear of the tool is detected, the output unit 26 may output, for example, a message that the tool is worn. When breakage of the tool is detected, the output unit 26 may output, to the machine tool 1, an instruction to stop the machine tool 1.

[0060] When the first time-series data acquired by the machine tool 1 is received in real time by the abnormality detection device 2, the processes of steps S101 and S102 may be performed in parallel with other processes. The order of the processes in the flowchart of Fig. 2 is an example, and the order of the steps may be changed as long as the same results can be obtained. In the flowchart of Fig. 2, the process ends when the power is turned off or an interrupt is issued to end the process.

[0061] Next, the operation of the abnormality detection device 2 according to this embodiment will be described using a specific example. When machining is started in the machine tool 1, first time-series data is acquired by the first sensor 11 accordingly, and the first time-series data is transmitted in real time to the abnormality detection device 2. The reception unit 21 of the abnormality detection device 2 receives the first time-series data and accumulates it in the storage unit 22 (steps S101, S102).

[0062] The acquisition unit 23 determines to detect an anomaly when the first time series data stored in the storage unit 22 includes two cycles, acquires reference time series data corresponding to the first cycle and detection target time series data corresponding to the second cycle from the first time series data, and passes them to the processing unit 24 (steps S103 to S105). Upon receiving the reference time series data and the detection target time series data, the processing unit 24 matches them together and passes the result to the anomaly detection unit 25 (step S106). For example, as shown in FIG. 6, this matching causes the detection target time series data Q in the non-processing section to be shifted in the time direction so as to overlap with the reference time series data P, and the difference between the two time series data becomes smaller. 7, for example, by this fitting, the detection target time series data Q in the processing section is shifted in the time direction to overlap with the reference time series data P as shown by the solid arrow at the portion shifted in the time direction by less than the threshold, but is not shifted in the time direction to overlap with the reference time series data P at the portion shifted in the time direction beyond the threshold as shown by the dashed arrow. Therefore, for the portion of the detection target time series data Q where an abnormality occurs and which is significantly different from the reference time series data P, the difference between the two remains even after fitting, so that the abnormality can be appropriately detected.

[0063] When the abnormality detection unit 25 receives the reference time series data and the detection target time series data after fitting from the processing unit 24, the abnormality detection unit 25 detects an abnormality. For example, the detection of the abnormality may be detection of wear of the tool, detection of damage to the tool, or both. When at least one of abnormality of wear and damage to the tool is detected, the abnormality detection unit 25 passes a notice of the presence of an abnormality of wear and / or damage to the output unit 26 (step S107). When the notice of the presence of an abnormality is received, the output unit 26 performs an output corresponding to the abnormality (step S108). For example, when an abnormality of wear is detected, the output unit 26 may output the notice to the operator of the machine tool 1. The operator who knows about the wear of the tool may replace the tool of the machine tool 1, for example. Furthermore, when an abnormality of damage is detected, the output unit 26 may output an instruction to the machine tool 1 to stop the machine tool 1 and output a notice of the detection of an abnormality of damage to the operator of the machine tool 1. By stopping machine tool 1 in response to the breakage of the tool, for example, it is possible to prevent an unexpected situation from occurring in response to the breakage of the tool. Also, an operator who knows about the breakage of the tool may, for example, replace the tool of machine tool 1 and resume machining.

[0064] As described above, according to the anomaly detection device 2 of this embodiment, by matching the reference time series data and the detection target time series data so that the sum of the distances between the data at each time is minimized while allowing a time shift of a predetermined threshold or less, it is possible to eliminate the time shift of the steep waveform of the non-processing section that is less than the threshold, and to prevent the time shift of the processing section that is greater than the threshold from being eliminated. As a result, it is possible to reduce the possibility of erroneously detecting that an anomaly has occurred due to a difference in the waveform of the non-processing section, and to appropriately detect the anomaly.

[0065] In this embodiment, the processing unit 24 may perform upsampling on the reference time series data and the detection target time series data acquired by the acquisition unit 23, and then combine both time series data. By performing upsampling, the sampling period is made shorter than before upsampling, and the number n of data included in the reference time series data P and the detection target time series data Q can be increased, and more accurate combination can be achieved. In particular, for a portion with large changes such as a non-processing section, the difference between the data included in the reference time series data P and the data included in the detection target time series data Q can be made smaller by interpolating data through upsampling, and the combination of both time series data can correct the deviation with higher accuracy.

[0066] Furthermore, the abnormality detection device 2 may detect an abnormality not only in the first time series data but also in the second time series data. In this case, the abnormality detection device 2 may perform the matching process and the abnormality detection process independently for each of the first and second time series data. In this case, for example, the reception unit 21 may also receive second time series data related to a second motor in the machine tool 1. Furthermore, the acquisition unit 23 may also acquire reference time series data and detection target time series data for the second time series data. This process may also be performed by acquiring time series data for each cycle in the second time series data, similar to the case of acquiring the reference time series data and the detection target time series data from the first time series data. Furthermore, the processing unit 24 may also perform matching between the reference time series data acquired from the second time series data and the detection target time series data. This matching process may also be performed similar to the matching between the reference time series data acquired from the first time series data and the detection target time series data. The anomaly detection unit 25 may also detect an anomaly in the detection target time series data by comparing the reference time series data after fitting, acquired from the second time series data, with the detection target time series data. This anomaly detection may be performed in the same manner as the anomaly detection by comparing the reference time series data after fitting, acquired from the first time series data, with the detection time series data. The second time series data may be, for example, time series data of the current of a servo motor. In this case, for example, the anomaly can be detected by using the current of the servo motor.

[0067] In addition, when the first motor is a spindle motor that rotates a spindle, the first time series data may be at least any one of data on the current of the first motor, data on the torque of the first motor, data on the load applied to a non-rotating object in contact with a rotating object rotated by the spindle measured by a cutting dynamometer when a workpiece is machined by the machine tool 1, data on acceleration measured by an acceleration sensor attached to a non-rotating part of the first motor, measurement data obtained by an AE sensor attached to the non-rotating object, and acoustic data including the machining sound of the workpiece.

[0068] When the first time series data includes data on the torque of the spindle motor, the first sensor 11 may include, for example, a torque sensor provided on the rotating shaft of the spindle motor. Since there is a correlation between the current value and the torque of the motor, the rotation status of the spindle can be grasped using the torque as well as the current value. When the first time series data includes data on the load applied to a non-rotating object, the first sensor 11 may include a cutting dynamometer to which a non-rotating object is fixed and which abuts against a rotating object rotated by the spindle. When the machine tool 1 is an NC lathe, for example, the rotating object may be a workpiece and the non-rotating object may be a tool. When the machine tool 1 is a machining center, for example, the rotating object may be a tool and the non-rotating object may be a workpiece. The rotation status of the spindle during machining can be grasped by the data on the load applied to the non-rotating object acquired by the cutting dynamometer. When the first time series data includes data on acceleration, the first sensor 11 may include an acceleration sensor attached to a non-rotating part of the spindle motor. Since the vibration generated according to the rotation of the spindle motor can be detected by the acceleration data, the rotation status of the spindle can be grasped even by using the acceleration. Furthermore, when the first time series data includes measurement data acquired by an AE sensor, the first sensor 11 may include an AE sensor attached to a non-rotating object. The rotation status of the spindle during processing can be grasped by the measurement data acquired by the AE sensor attached to the non-rotating object. Furthermore, when the first time series data includes acoustic data, the first sensor 11 may include a microphone arranged at a position where the processing sound of the workpiece can be acquired. The rotation status of the spindle during processing can be grasped by the acoustic data of the processing sound of the workpiece. Since the rotation status of the spindle can be grasped by using such first time series data related to the rotation of the spindle, an abnormality can be detected by using such first time series data.

[0069] In addition, when the first motor is a servo motor for moving a workpiece or a tool, the first time series data may be at least any one of current data of the first motor, torque data of the first motor, acceleration data measured by an acceleration sensor attached to a non-rotating part of the first motor, and acoustic data including sound generated in response to the rotation of the first motor.

[0070] When the first time series data includes data on the torque of the servo motor, the first sensor 11 may include, for example, a torque sensor provided on the rotating shaft of the servo motor. When the first time series data includes data on acceleration, the first sensor 11 may include an acceleration sensor attached to a non-rotating part of the servo motor. Since vibrations occurring in response to the rotation of the servo motor can be detected by the acceleration data, the rotation status of the servo motor can be grasped even by using the acceleration. When the first time series data includes acoustic data, the first sensor 11 may include a microphone disposed at a position capable of acquiring sounds occurring in response to the rotation of the first motor. The rotation status of the servo motor can be grasped by acoustic data including sounds generated by the rotation of the servo motor. Since the rotation status of the servo motor can be grasped by using the first time series data, the first time series data can be used to detect an abnormality. For example, when a tool is worn, a larger force is required to move the workpiece or the tool during processing, and when the tool is broken, the workpiece and the tool are no longer in contact with each other, so that the workpiece or the tool can be moved with a smaller force, and therefore, an abnormality can also be detected by using the first time series data related to the rotation of the servo motor. Also, the first time series data may be other time series data related to the first motor as long as it can be used to detect an abnormality.

[0071] In addition, when the first time series data is data of a current of a spindle motor that rotates a spindle in the machine tool 1, and the second time series data is data of a current of a servo motor of the machine tool 1, the range of the first sensor 11 that acquires the first time series data may be larger than the range of the second sensor 12 that acquires the second time series data, and the resolution of the first and second sensors 11, 12 may be the same. That is, the range of the first time series data may be larger than the range of the second time series data, and the resolution of the first time series data may be the same as the resolution of the second time series data. The range of the sensor or time series data is a width from a minimum value to a maximum value. Usually, the range of the current of the spindle motor is larger than the range of the current of the servo motor. As an example, the range of the current of the spindle motor may be 0 to 120 (A), and the range of the current of the servo motor may be 0 to 10 (A). In this example, the range of the first time series data may be 12 times the range of the second time series data. The resolution of the first and second sensors 11, 12, that is, the resolution of the first and second time series data, may both be R bits. R is not particularly limited as long as it is an arbitrary positive integer, but it is usually preferable that R is 8 or more. In this way, it is possible to set a range suitable for the first and second time series data, and it is possible to perform appropriate abnormality detection and prevent the amount of data from increasing. In addition, if there is no problem even if the amount of data increases, as an example, the ranges of the first and second time series data may be combined and the resolution of the second time series data may be higher than the resolution of the first time series data. In this case, as an example, the range of the second time series data may also be set to 0 to 120 (A). On the other hand, as an example, the resolution of the second time series data may be set to 12 times the resolution of the first time series data.

[0072] Furthermore, when the second motor is a servo motor for moving a workpiece or a tool, the second time series data may be at least any of data on the current of the second motor, data on the torque of the second motor, data on acceleration measured by an acceleration sensor attached to a non-rotating part of the second motor, and sound data including sound generated in response to the rotation of the second motor. The second time series data may be acquired by, for example, the second sensor 12. Furthermore, the second time series data may be other time series data related to the second motor as long as it can be used to detect an abnormality.

[0073] Furthermore, the abnormality detection device 2 may receive, for example, three or more pieces of time series data and perform matching and abnormality detection for each piece of time series data. The three or more pieces of time series data may include, for example, time series data related to the spindle motor in the machine tool 1 and multiple pieces of time series data respectively related to multiple servo motors in the machine tool 1.

[0074] In addition, in the above embodiments, each process or function may be realized by centralized processing in a single device or a single system, or may be realized by distributed processing in multiple devices or multiple systems.

[0075] In addition, in the above embodiments, the transfer of information between each component may be performed, for example, by one component outputting information and the other component receiving information if the two components transferring the information are physically different, or, if the two components transferring the information are physically the same, by shifting from a processing phase corresponding to one component to a processing phase corresponding to the other component.

[0076] In the above embodiment, information related to the processing performed by each component, such as information accepted, acquired, selected, generated, transmitted, or received by each component, and information such as thresholds, formulas, and addresses used by each component in processing, may be temporarily or long-term stored in a recording medium (not shown) even if not specified in the above description. Furthermore, each component or a storage unit (not shown) may store information in the recording medium (not shown). Furthermore, each component or a reading unit (not shown) may read information from the recording medium (not shown).

[0077] In the above embodiment, if the information used by each component, such as the thresholds, addresses, and various setting values ​​used by each component in processing, may be changed by the user, the user may be able to change the information as appropriate, even if not specified in the above description. If the information is changeable by the user, the change may be realized, for example, by a reception unit (not shown) that receives a change instruction from the user, and a change unit (not shown) that changes the information in response to the change instruction. The reception unit (not shown) may receive the change instruction from an input device, may receive information transmitted via a communication line, or may receive information read from a predetermined recording medium.

[0078] In addition, in the above embodiment, when two or more components included in the anomaly detection device 2 have a communication device, an input device, etc., the two or more components may have a physically single device, or may have separate devices.

[0079] In the above embodiment, each component may be configured by dedicated hardware, or a component that can be realized by software may be realized by executing a program. For example, each component may be realized by a program execution unit such as a CPU reading and executing a software program recorded on a recording medium such as a hard disk or semiconductor memory. During execution, the program execution unit may execute the program while accessing the storage unit or recording medium. The software that realizes the abnormality detection device 2 in the above embodiment is a program such as the following. In other words, this program causes a computer to execute the steps of accepting first time series data related to the rotation of a spindle in a machine tool; acquiring from the first time series data reference time series data that serves as a standard corresponding to one cycle of machining repeated in the machine tool, and detection target time series data that is to be subject to detection of anomalies corresponding to one cycle; aligning the reference time series data and the detection target time series data acquired from the first time series data so that the sum of the distances between the data at each time is minimized while allowing a time shift of less than a predetermined threshold; and detecting an abnormality in the detection target time series data by comparing the reference time series data and the detection target time series data after alignment.

[0080] In addition, in the program, the steps of receiving information and outputting information do not include processes that are performed only by hardware, such as processes performed by a modem or interface card in the step of receiving information.

[0081] This program may be executed by being downloaded from a server or the like, or may be executed by reading out a program recorded on a predetermined recording medium (for example, an optical disk such as a CD-ROM, a magnetic disk, a semiconductor memory, etc.). This program may be used as a program constituting a program product.

[0082] Furthermore, the program may be executed by a single computer or a plurality of computers. That is, the program may be executed by centralized processing or distributed processing.

[0083] Fig. 8 is a diagram showing the configuration of a computer system 900 that executes the above program to realize the anomaly detection device 2 according to the above embodiment. In Fig. 8, the computer system 900 includes a computer 901, a keyboard 902, a mouse 903, and a monitor 904. The computer 901 includes a CD-ROM drive 905, an MPU (Micro Processing Unit) 911, a ROM 912 for storing programs such as a boot-up program, a RAM 913 connected to the MPU 911 and temporarily storing instructions of application programs and providing a temporary storage space, a hard disk 914 for storing application programs, system programs, and data, and a bus 915 for connecting the MPU 911, the ROM 912, and the like to each other. The computer 901 may include a network card (not shown) that provides a connection to a LAN, a WAN, or the like.

[0084] A program for causing computer system 900 to execute the functions of anomaly detection device 2 according to the above embodiment may be stored on CD-ROM 921, inserted into CD-ROM drive 905, and transferred to hard disk 914. Alternatively, the program may be transmitted to computer 901 via a network (not shown) and stored on hard disk 914. The program is loaded into RAM 913 at the time of execution. The program may be loaded directly from CD-ROM 921 or the network. The program may also be read into computer system 900 via another recording medium (e.g., a DVD) instead of CD-ROM 921.

[0085] The program does not necessarily include an operating system (OS) or a third party program that causes the computer 901 to execute the functions of the anomaly detection device 2 according to the above embodiment. The program may include only an instruction portion that calls appropriate functions or modules in a controlled manner to achieve a desired result. How the computer system 900 operates is well known, and a detailed description will be omitted.

[0086] The above-mentioned embodiments are merely examples for specifically implementing the present invention, and are not intended to limit the technical scope of the present invention. The technical scope of the present invention is indicated by the claims, not by the description of the embodiments, and is intended to include modifications within the literal scope of the claims and the scope of equivalent meanings. [Explanation of symbols]

[0087] 1 Machine tools 2. Anomaly detection device 21 Reception 22 Memory section 23 Acquisition Department 24 Processing section 25 Anomaly detection section 26 Output section

Claims

1. a reception unit that receives first time series data related to a first motor in the machine tool; an acquisition unit that acquires, from the first time series data, reference time series data that serves as a reference corresponding to one cycle of machining repeated by the machine tool, and detection target time series data that is a detection target for an abnormality corresponding to one cycle; a processing unit that aligns the reference time series data and the detection target time series data acquired from the first time series data so that a sum of distances between data at each time is minimized while allowing a time shift of a predetermined threshold or less; and an anomaly detection unit that detects an anomaly in the detection target time series data by comparing the adjusted reference time series data with the detection target time series data.

2. The anomaly detection device according to claim 1 , wherein the processing unit combines the reference time series data and the detection target time series data acquired by the acquisition unit after upsampling the two time series data.

3. The receiving unit also receives second time-series data related to a second motor of the machine tool, The acquisition unit acquires reference time series data and detection target time series data for the second time series data, The processing unit also performs matching between the reference time series data acquired from the second time series data and the detection target time series data, 2. The anomaly detection device according to claim 1, wherein the anomaly detection unit also detects an anomaly in the detection target time series data by comparing the detection target time series data with the reference time series data after fitting, which is obtained from the second time series data.

4. the first time-series data is data of a current of a spindle motor that rotates a spindle in the machine tool, the second time series data is data of a current of a servo motor, a range of the first time series data is greater than a range of the second time series data; 4. The anomaly detection device according to claim 3, wherein the resolution of the first time series data is the same as the resolution of the second time series data.

5. 5. The abnormality detection device according to claim 1, wherein the threshold value is shorter than the time of the shortest machining section in one cycle.

6. 5. The anomaly detection device according to claim 1, wherein the reference time series data is time series data received by the receiving unit before fitting is performed.

7. An anomaly detection method that is processed using a reception unit, an acquisition unit, a processing unit, and an anomaly detection unit, The receiving unit receives first time series data related to rotation of a spindle in a machine tool; The acquiring unit acquires, from the first time series data, reference time series data serving as a reference corresponding to one cycle of machining repeated by the machine tool, and detection target time series data that is a detection target for an abnormality corresponding to one cycle; a step in which the processing unit aligns the reference time series data and the detection target time series data acquired from the first time series data so that a sum of distances between data at each time becomes minimum while allowing a time shift of a predetermined threshold or less; and detecting an anomaly in the detection target time series data by the anomaly detection unit comparing the adjusted reference time series data with the detection target time series data.

8. On the computer, receiving first time series data relating to rotation of a spindle of a machine tool; acquiring, from the first time series data, reference time series data serving as a reference corresponding to one cycle of machining repeated by the machine tool, and detection target time series data serving as a detection target for anomaly corresponding to one cycle; A step of aligning the reference time series data and the detection target time series data acquired from the first time series data so that a sum of distances between data at each time is minimized while allowing a time shift of a predetermined threshold or less; a step of detecting an anomaly in the detection target time series data by comparing the adjusted reference time series data with the detection target time series data;

Images