Abnormality detection device and program

Abstract

To provide an abnormality detection device with low calculation cost and high determination accuracy.SOLUTION: An abnormality determination device 1 includes: a data acquisition part 2 which acquires time-series data 6 indicating a contact degree of a tool and workpiece at a predetermined sampling interval; a list data generation part 3 which confirms the time-series data 6 in time series and generates list data of frequencies of continuous occurrences of a same contact degree; and a determination part 4 which calculates a representative value of the list data on each of the same contact degrees, and determines whether or not a shape of the workpiece is abnormal on the basis of the representative value.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an anomaly detection device that detects anomalies in time-series data. [Background technology]

[0002] Conventionally, in detecting anomalies in time-series data related to processing by a processing machine, a method has been widely adopted in which a normal pattern is first created, pattern matching is performed between the normal pattern and inspection data, and the degree of agreement or deviation between the patterns is calculated to determine whether the data is normal or abnormal. For example, Patent Document 1 discloses a method of learning a normal model by machine learning in order to detect anomalies in processing by a processing machine. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2023-026103 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, in conventional anomaly detection technologies including that of Patent Document 1, it is sometimes impossible or inaccurate to align the phase of a normal pattern and inspection data in pattern matching. Even if accurate phase alignment is possible, it requires huge calculation costs, and the judgment accuracy may be low due to individual differences in tools and workpieces, the influence of measurement conditions, inaccuracies in normal patterns, and the like.

[0005] The present invention has been made in consideration of the above-mentioned problems, and has an object to provide an anomaly detection device and the like that has low calculation costs and high determination accuracy. [Means for solving the problem]

[0006] A first invention for achieving the above-mentioned object is an abnormality detection device comprising a data acquisition unit that acquires time series data indicating the degree of contact between a tool and a workpiece at a predetermined sampling interval, a list data generation unit that checks the time series data in chronological order and generates list data of the number of times the same degree of contact appears in succession, and a judgment unit that calculates a representative value of the list data for each degree of contact and judges whether the shape of the workpiece is abnormal based on the representative value.

[0007] The judgment unit may judge whether or not the shape of the workpiece is abnormal based on the degree of deviation between each element of the list data and the representative value.

[0008] In addition, the time series data may be binary data indicating contact or non-contact between the tool and the workpiece, and the list data generation unit may generate the list data by alternately counting the number of times a value indicating contact appears consecutively and the number of times a value indicating non-contact appears consecutively.

[0009] The second invention is a program for causing a computer to function as an abnormality detection device comprising a data acquisition unit that acquires time series data indicating the degree of contact between a tool and a workpiece at a predetermined sampling interval, a list data generation unit that checks the time series data in chronological order and generates list data of the number of times the same degree of contact appears in succession, and a judgment unit that calculates a representative value of the list data for each degree of contact and judges whether the shape of the workpiece is abnormal based on the representative value. Effect of the Invention

[0010] The present invention can provide an anomaly detection device and the like that has low calculation costs and high determination accuracy. [Brief description of the drawings]

[0011] [Figure 1] FIG. 2 is a functional block diagram showing the configuration of an abnormality determination device according to the present invention; [Diagram 2]A flowchart showing an example of a process flow of a data acquisition unit in the abnormality determination device of FIG. 1. [Diagram 3] A flowchart showing an example of a process flow of a list data generating unit in the abnormality determination device of FIG. 1. [Figure 4] A flowchart showing an example of a process flow of a determination unit in the abnormality determination device of FIG. 1. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] The time-series data targeted by the anomaly detection device of the present invention is data indicating the degree of contact between a tool and a workpiece in machining by a numerically controlled machining machine, and is preferably periodic data, such as data relating to the chamfering of gears.

[0013] Gears have teeth of the same size and at equal intervals. In chamfering such gears, the tool and workpiece periodically come into contact and out of contact for chamfering (deburring), and the time series data has periodicity. The anomaly detection device in this embodiment focuses on this periodicity and does not perform pattern matching between normal patterns and inspection data as in conventional technology, but uses only its own inspection data to determine whether the shape of the workpiece is abnormal. This makes it possible to detect defects such as missing teeth in the gear workpiece.

[0014] The present embodiment will be described in detail below with reference to the drawings. Fig. 1 is a functional block diagram showing the configuration of the abnormality determination device of the present invention. As shown in Fig. 1, the abnormality determination device 1 includes a data acquisition unit 2 that acquires time-series data 6 indicating the degree of contact between a tool and a workpiece at a predetermined sampling interval, a list data generation unit 3 that checks the time-series data 6 in chronological order and generates list data of the number of times the same degree of contact appears consecutively, a determination unit 4 that calculates a representative value of the list data for each degree of contact and determines whether the shape of the workpiece is abnormal or not based on the representative value, and a display unit 5 that displays the determination result on a display device 7.

[0015] The abnormality determination device 1 is a computer having hardware such as a CPU (Central Processing Unit), memory, and an input / output interface. The CPU executes the processes of the data acquisition unit 2, list data generation unit 3, determination unit 4, and display unit 5 by reading and executing programs previously stored in the memory. The CPU also inputs and outputs data to and from external devices via the input / output interface. The connection for inputting and outputting data may be wireless or wired. The computer constituting the abnormality determination device 1 may be a PC (Personal Computer) or a PLC (Programmable The device may be a portable terminal such as a smartphone or a tablet.

[0016] The time series data 6 in this embodiment is binary data indicating contact or non-contact between the tool and the workpiece. The time series data 6 is, for example, data obtained by determining in advance at a predetermined sampling interval whether the tool and the workpiece are in contact or not, based on vibration data obtained from a vibration sensor installed in a numerically controlled processing machine. In the time series data 6, for example, 1 indicates contact (=ON) and 0 indicates non-contact (=OFF). Hereinafter, ON means contact and OFF means non-contact.

[0017] The data acquiring unit 2 may acquire CSV data, which has been previously determined to be in contact or not in contact, from its own storage device or another computer as the time-series data 6. In addition, the data acquiring unit 2 may itself or another computer determine whether the tool and the workpiece are in contact or not for data output in real time from a data logger such as a vibration sensor, and acquire the determination result as the time-series data 6.

[0018] In addition, if there is spike-like noise in the time series data 6, it may have an adverse effect on the processing described below. Therefore, when the data acquisition unit 2 itself determines whether the tool is in contact with the workpiece, it may perform a filter process to remove noise from the data output from the data logger. For example, when the data output from the data logger is generated based on vibration acceleration data, the data acquisition unit 2 applies a low-pass filter to the vibration frequency, and determines whether the tool is in contact with the workpiece based on the data after the low-pass filter process. The same applies when another computer performs these processes.

[0019] Fig. 2 is a flowchart showing an example of a process flow of a data acquisition unit in the abnormality determination device of Fig. 1. As shown in Fig. 2, the data acquisition unit 2 of the abnormality determination device 1 acquires time-series data 6 of N samples (step S1), and stores the time-series data 6 in an array D[j] (j=1, 2, . . . , N) in a memory (step S2).

[0020] Returning to the explanation of Fig. 1, the list data generating unit 3 checks the time series data 6 acquired by the data acquiring unit 2 and stored in the array D[j] in chronological order, and generates list data by alternately counting the number of times a value indicating contact (=ON) appears consecutively and the number of times a value indicating no contact (=OFF) appears consecutively. In this embodiment, the time series data 6 is binary data, and it is sufficient to count alternately, so that the calculation cost can be kept significantly low.

[0021] Fig. 3 is a flowchart showing an example of the flow of processing by the list data generating unit in the abnormality determination device of Fig. 1. As shown in Fig. 3, the list data generating unit 3 of the abnormality determination device 1 executes initialization of variables (step S11). Specifically, the list data generating unit 3 sets the previous value Prev, which is a variable that stores the previous value of the time-series data 6 to be processed, to 0 (=OFF), the consecutive OFF count Coff, which is a variable that stores the number of consecutive OFFs during counting, to 0, and the consecutive ON count Con, which is a variable that stores the number of consecutive ONs during counting, to 0.

[0022] Next, the list data generating unit 3 executes list initialization (step S12). Specifically, the list data generating unit 3 initializes the OFF list Loff, which is a list that stores the number of consecutive OFFs, and the ON list Lon, which is a list that stores the number of consecutive ONs, to an empty state without specifying the number of elements.

[0023] Next, the list data generation unit 3 extracts the time series data 6 stored in the array D[j] (j = 1, 2, ... N) one element at a time, and executes a repetitive process described later until the element number j of the array D[j] becomes N (step S13).

[0024] In the repetitive process, the list data generating unit 3 checks whether the value of the array D[j] extracted in step S13 is OFF (step S14). If the value of the array D[j] is OFF (Yes in step S14), the list data generating unit 3 checks whether the previous value Prev is OFF (step S15). If the previous value Prev is OFF (Yes in step S15), the list data generating unit 3 considers that OFF (non-contact) is continuous, and adds 1 to the number of consecutive OFFs Coff (step S16). If the previous value Prev is not OFF (No in step S15), the list data generating unit 3 considers that the continuous ON (contact) has ended and switched to OFF (non-contact), adds the value of the number of consecutive ONs Con to the end of the ON list Lon (step S17), and initializes the number of consecutive ONs Con to 0 (step S18).

[0025] On the other hand, when the value of the array D[j] is not OFF (No in step S14), that is, when the value of the array D[j] is ON, the list data generation unit 3 checks whether the previous value Prev is ON (step S19). When the previous value Prev is ON (Yes in step S19), the list data generation unit 3 regards that ON (contact) is continuous and adds 1 to the ON continuous count Con (step S20). When the previous value Prev is not ON (No in step S19), the list data generation unit 3 regards that the OFF (non-contact) continuity has ended and has changed to ON (contact), adds the value of the OFF continuous count Coff to the end of the OFF list Loff (step S21), and initializes the OFF continuous count Coff to 0 (step S22).

[0026] When one iteration of the repetition process is completed, the list data generation unit 3 compares the element number j of the array D[j] with N. If j < N, the value of the array D[j] is assigned to the previous value Prev, 1 is added to the element number j, and the process is repeated from step S13. If j = N, the process ends (step S23).

[0027] Return to the description of FIG. 1. The determination unit 4 uses the list data generated by the list data generation unit 3 to detect the presence or absence of heterogeneous continuous values of ON (contact) or OFF (non-contact), and determines whether the shape of the workpiece is abnormal. More specifically, the determination unit 4 determines whether the shape of the workpiece is abnormal based on the degree of deviation between each element of the ON list data and the representative value of the ON list data, and the degree of deviation between each element of the OFF list data and the representative value of the OFF list data. As a result, it is not necessary to match the phase with normal data as in the prior art, and the calculation cost can be significantly reduced. In addition, since the time series data 6 of other workpieces that are not the determination target is not used, it is not affected by individual differences in tools and workpieces, measurement conditions, accuracy of normal patterns, etc., and can be accurately determined.

[0028] In this embodiment, the representative value of the list data uses the average value, but it is not limited thereto. The representative value may be, for example, the median or the mode value, and known statistical quantities can be used as appropriate.

[0029] 4 is a flowchart showing an example of the flow of processing by the determination unit in the abnormality determination device of FIG. 1. As shown in FIG. 4, the determination unit 4 of the abnormality determination device 1 calculates the average value Mon of all elements of the ON list Lon (step S31). Next, the determination unit 4 checks whether or not there is an element in the ON list Lon that is equal to or less than 1 / X (X is a real number greater than 1) of the average value Mon of the ON list calculated in step S31 (step S32). If there is even one such element (Yes in step S32), the determination unit 4 determines that there is an abnormality during the continuous ON (=during contact between the workpiece and the tool) (step S33). On the other hand, if there is no such element (No in step S32), the determination unit 4 determines that all the continuous ON are normal (step S34).

[0030] Next, the judgment unit 4 of the abnormality judgment device 1 calculates the average value Moff of all elements of the OFF list Loff (step S35). Next, the judgment unit 4 checks whether or not there is an element in the OFF list Loff that is equal to or less than 1 / Y (Y is a real number greater than 1) of the average value Moff of the OFF list calculated in step S35 (step S36). If there is even one such element (Yes in step S36), the judgment unit 4 judges that there is an abnormality during the continuous OFF (= during the non-contact between the workpiece and the tool) (step S37). On the other hand, if there is no such element (No in step S36), the judgment unit 4 judges that all the continuous OFF is normal (step S38).

[0031] In steps S32 and S36, the judgment conditions are set to be equal to or less than 1 / X of the average value Mon of the ON list and equal to or less than 1 / Y of the average value Moff of the OFF list, but the judgment conditions are not limited to this. Any threshold value may be set as the judgment conditions as long as it is possible to judge normality or abnormality based on the degree of deviation between each element of the list data and the representative value.

[0032] In addition, if the judgment unit 4 judges that there is an abnormality in steps S32 and S36, it may store in a memory or storage device which element in array D[j] it has judged to be abnormal, in other words, the group of elements it has judged to be abnormal.

[0033] Returning to the explanation of Fig. 1, the display unit 5 displays the judgment results of the judgment unit 4 during continuous OFF and continuous ON on the display device 7. Furthermore, when a group of elements judged to be abnormal in the processing of the judgment unit 4 is stored, the display unit 5 may display the group of elements judged to be abnormal on the display device 7. If it is possible to identify in advance where on the workpiece the time-series data 6 starts, it is possible to identify the abnormal part of the workpiece from the group of elements judged to be abnormal.

[0034] In this embodiment, the means for communicating the judgment result to the user is to display it on the display device 7, but is not limited to this. The means for communicating the judgment result to the user may be to output it as a sound, to turn on a warning lamp if an abnormality occurs, to output it onto a paper medium by a printing device, to output it onto the screen of the robot's teaching pendant, to output it onto the touch panel of a PLC, or to send it to another computer (= the user's mobile terminal, etc.).

[0035] As described above, the abnormality determination device 1 in this embodiment does not perform pattern matching with a normal pattern, but determines whether the shape of the workpiece is abnormal using only its own time series data 6. Specifically, the abnormality determination device 1 checks the time series data 6 in chronological order, generates list data of the number of times the same contact degree appears consecutively, calculates a representative value of the list data for each of the same contact degrees, and determines whether the shape of the workpiece is abnormal based on the representative value. This eliminates the need for phase matching with normal data as in the conventional technology, and can significantly reduce calculation costs. In addition, since it does not use the time series data 6 of other workpieces that are not the object of determination, it can be determined with high accuracy without being affected by individual differences in tools and workpieces, measurement conditions, accuracy of the normal pattern, etc.

[0036] For example, in the chamfering of gears, the abnormality determination device 1 acquires time-series data 6 in real time and determines whether the workpiece is abnormal, thereby making it possible to detect a workpiece having defects such as missing teeth while the chamfering is being performed. In particular, the abnormality determination device 1 in this embodiment can significantly reduce the calculation cost for abnormality determination, so there is no delay in the actual machining work.

[0037] Although the preferred embodiments of the anomaly detection device and the like according to the present invention have been described above with reference to the attached drawings, the present invention is not limited to such examples. It is clear that a person skilled in the art can come up with various modified or revised examples within the scope of the technical ideas disclosed in this application, and it is understood that such examples also naturally fall within the technical scope of the present invention. [Explanation of symbols]

[0038] 1. Anomaly detection device 2. Data acquisition section 3. List data generation section 4……Judgment section 5……Display section 6. Time series data 7……Display device

Claims

1. a data acquisition unit that acquires time-series data indicating a degree of contact between a tool and a workpiece at a predetermined sampling interval; a list data generating unit that checks the time series data in chronological order and generates list data of the number of times the same contact degree appears consecutively; a determination unit that calculates a representative value of the list data for each same degree of contact and determines whether or not the shape of the workpiece is abnormal based on the representative value; An anomaly detection device comprising:

2. The determination unit determines whether or not the shape of the workpiece is abnormal based on the degree of deviation between each element of the list data and the representative value. The anomaly detection device according to claim 1 .

3. The time series data is binary data indicating contact or non-contact between the tool and the workpiece, The list data generating unit generates the list data by alternately counting the number of times a value indicating contact appears consecutively and the number of times a value indicating non-contact appears consecutively.

3. The abnormality detection device according to claim 1 or 2.

4. Computer, a data acquisition unit that acquires time-series data indicating a degree of contact between a tool and a workpiece at a predetermined sampling interval; a list data generating unit that checks the time series data in chronological order and generates list data of the number of times the same contact degree appears consecutively; a determination unit that calculates a representative value of the list data for each same degree of contact and determines whether or not the shape of the workpiece is abnormal based on the representative value; A program for causing the device to function as an anomaly detection device having the above-mentioned configuration.

Images