Abnormality prediction device and program

Abstract

To automatically predict an abnormality of a plurality of parts in a press device.SOLUTION: An abnormality prediction device (200) includes: model storage means (213) for storing abnormality prediction models generated for each sensor group; acquisition means (202) for acquiring measurement data from sensors during operation of a press device; feature quantity calculation means (203) for calculating a feature quantity of each of the plurality of sensor groups based on the measurement data acquired by the acquisition means; detection means (206) for inputting the feature quantity calculated by the feature quantity calculation means to the corresponding abnormality prediction model for each sensor group and detecting whether or not the feature quantity is abnormal; combination information storage means (214) for storing combination information (TB) associating each target part with at least one sensor group; and determination means (207) for determining the presence or absence of an abnormality for each target part based on the detection result by the detection means and the combination information.SELECTED DRAWING: Figure 9

Description

Technical Field

[0001] The present invention relates to an abnormality prediction device and a program, and particularly to an abnormality prediction device and a program for predicting abnormalities in a plurality of parts in a press device.

Background Art

[0002] A forging press device presses a workpiece by rotationally driving a crankshaft to move a slide up and down. In such a forging press device, as described in Japanese Unexamined Patent Application Publication No. 2019-13947 (Patent Document 1), the load applied to the workpiece during processing is monitored, and detecting abnormalities in the product based on the peak value of the load has been conventionally performed. This is because it has been found that the peak value of the load has a high correlation with the quality of the product.

[0003] Further, as described in Japanese Unexamined Patent Application Publication No. 2019-13976 (Patent Document 2), a technique of using a plurality of sensors for detecting the operating state of a press machine for predicting failures of the press machine has been conventionally proposed.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] In Patent Document 1, defective products can be detected based on the peak value of the load, but the cause (abnormality cause) cannot be specified.

[0006] In Patent Document 2, it is difficult for an inexperienced operator to accurately predict a failure of a press machine because a monitoring PC checks for changes from the initial normal state of measurement results obtained from a plurality of sensors and predicts the failure of the press machine in light of past empirical rules.

[0007] The present invention has been made to solve the above-described problems, and an object thereof is to provide an abnormality prediction device and a program that can automatically predict abnormalities in a plurality of parts in a press device such as a forging press device.

Means for Solving the Problems

[0008] An abnormality prediction device according to an aspect of the present invention is an abnormality prediction device for predicting abnormalities in a plurality of parts in a press device, and includes a model storage unit, an acquisition unit, a feature amount calculation unit, a detection unit, a combination information storage unit, and a determination unit. The model storage unit stores an abnormality prediction model generated for each sensor group by machine learning the feature amounts of a plurality of sensor groups obtained based on measurement data of sensors mounted on the press device. The acquisition unit acquires measurement data from the sensors during operation of the press device. The feature amount calculation unit calculates the feature amounts of each of the plurality of sensor groups based on the measurement data acquired by the acquisition unit. The detection unit inputs the feature amounts calculated by the feature amount calculation unit into the corresponding abnormality prediction model for each sensor group and detects whether the feature amounts are abnormal. The combination information storage unit stores combination information associating each target part with at least one sensor group. The determination unit determines the presence or absence of an abnormality for each target part based on the detection result by the detection unit and the combination information stored in the combination information storage unit.

[0009] Preferably, in the model storage means, an abnormality prediction model for each sensor group is stored for each die pattern specified by at least the part number of the material. In this case, the acquisition means acquires die pattern data together with the measurement data, and it is desirable that the abnormality prediction device further includes a model discrimination means for discriminating an abnormality prediction model to be used from among a plurality of abnormality prediction models stored in the model storage means based on the die pattern data acquired by the acquisition means.

[0010] Preferably, the abnormality prediction device further includes an output means for outputting the degree of influence of the sensor group on the determination result and / or the reliability of the determination result together with the determination result by the determination means.

[0011] Preferably, the target parts include at least one of a material heater, a clutch, a brake, a main motor, a bearing, and a die, and a material that is a workpiece.

[0012] The target parts may include a plurality of dies. In this case, it is desirable that the plurality of sensor groups include a vibration feature amount on the upstream side of the frame and a vibration feature amount on the downstream side of the frame.

[0013] An abnormality prediction program according to another aspect of the present invention is an abnormality prediction program for predicting abnormalities of a plurality of parts in a press device, and includes a step of acquiring measurement data from sensors mounted on the press device during operation of the press device, a calculation step of calculating feature amounts of each of a plurality of sensor groups based on the acquired measurement data, a detection step of inputting the feature amounts calculated in the calculation step into corresponding abnormality prediction models for each sensor group to detect whether the feature amounts are abnormal, and a determination step of determining the presence or absence of an abnormality for each target part based on the detection result in the detection step and combination information associating each target part with at least one sensor group, and causing a computer to execute the steps.

Effects of the Invention

[0014] According to the present invention, it is possible to automatically predict abnormalities in a plurality of parts in a press device.

Brief Description of the Drawings

[0015]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Embodiments for Carrying Out the Invention

[0016] Embodiments of the present invention will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals and their description will not be repeated.

[0017] <Schematic Configuration of Forging Press Device> First, with reference to FIGS. 1 to 3, the schematic configuration of the forging press device 10 will be described. As shown in FIG. 1, the forging press device 10 includes a press machine 11 that presses a material as a workpiece, and a PLC (Programmable Logic Controller) 40 that controls the operation of the press machine 11.

[0018] As shown in FIG. 2, the press machine 11 includes a slide 1 and a bolster 2 provided so as to face each other vertically within a frame 3. The frame 3 includes a pair of columns 3a and 3b arranged on the front side of the press machine 11 and a pair of columns 3c and 3d arranged on the rear side of the press machine 11. The press machine 11 is configured such that the slide 1 connected to the connecting rod 5 moves up and down by rotationally driving the crankshaft 4.

[0019] A flywheel 8 that is rotationally driven by a main motor 6 via a belt 7 is connected to one end side of the crankshaft 4 via a clutch 9. A brake device 12 for stopping the rotation of the crankshaft 4 and stopping the slide 1 is attached to the other end side of the crankshaft 4. The press machine 11 rotationally drives the flywheel 8 at a constant speed by the main motor 6, engages the clutch 9, and transmits the rotation of the flywheel 8 to the crankshaft 4 to raise and lower the slide 1, thereby forging the workpiece placed on the bolster 2. The workpiece is processed into a size and shape corresponding to the pattern of a mold including a lower mold (not shown) installed on the bolster 2 and an upper mold (not shown) installed at the lower end of the slide 1, and becomes the final product. The "mold" means a mold used for forging. The mold is typically formed of special steel.

[0020] The forging press apparatus 10 in this embodiment is an apparatus for processing a workpiece by hot forging. The forging press apparatus 10 includes a plurality of sensors as described below in order to detect the operating status of the press machine 11. As will be described later, the press machine 11 may be a multi-stage press machine that processes a workpiece through a plurality of steps. That is, as shown in FIG. 4, a plurality (for example, three) of dies 51 to 53 may be mounted on the press machine 11. In the following description, when it is not necessary to distinguish between the dies 51 to 53, these are expressed as "die 50". · Load sensor 20: Detects the load applied to the workpiece. For example, it is composed of a strain gauge that detects the strain of the frame 3. · Angle sensor 21: By detecting the rotation angle (press angle) of the crankshaft 4, the position of the slide 1, and the speed (stroke) and acceleration of the slide 1 are detected. · Brake slack pressure sensor 22: Detects the slack pressure of the brake device 12. · Clutch pressure sensor 23: Detects the pressure of the clutch 9. · Clutch tank pressure sensor 24: Detects the pressure of the hydraulic oil that drives the clutch 9. · Brake tank pressure sensor 25: Detects the pressure of the cooling water of the brake device 12. · Cooling water flow sensor 26: Detects the flow rate of the cooling water of the brake device 12. · BKO position displacement sensor 27: Detects the position (BKO position) of the lower knockout device of the press machine 11. · SKO position displacement sensor 28: Detects the position (SKO position) of the upper knockout device. · Resistance temperature detector 29: Detects the temperature of each part of the press machine 11. · Lubricating fluid flow sensor 30: Detects the flow rate of the lubricating fluid that lubricates the die. · Air blow pressure sensor 31: Detects the pressure of the high-pressure air supplied to the press machine 11. · Material temperature sensor 32: Detects the temperature of the material supplied to the press machine 11. · Die height displacement sensor 33: Detects the die height (the distance between the bolster 2 and the bottom dead center position of the slide 1) of the press machine 11. · Type temperature sensor 34: Detects the temperature of the mold. · Vibration meter 35: For example, composed of an acceleration sensor, detects the vibration of the frame 3. · Motor speed sensor 36: Detects the speed (actual speed) of the main motor 6 by detecting the rotation angle of the rotating shaft of the main motor 6. · Motor vibration sensor 37: For example, composed of an acceleration sensor, detects the vibration of the main motor 6.

[0021] In this embodiment, the above-described load sensor 20 is attached to any one of the columns (for example, column 3c) of the frame 3. Also, vibration meters 35 are attached to the left and right columns 3c and 3d of the frame 3, one each. One of the columns 3c and 3d (column 3c) is located upstream of the mold in the transport direction, and the other of the columns 3c and 3d (column 3d) is located downstream of the mold in the transport direction. Each vibration meter 35 detects the vibration of the column 3c or 3d to which it is attached in a substantially horizontal direction (front-rear direction or left-right direction).

[0022] As shown in FIG. 3, the forging press apparatus 10 further includes a transport device 13 and a heater 14. The transport device 13 is composed of, for example, a belt conveyor, and transports the workpiece W from the previous process to the press 11 (the position of the bolster 2). The heater 14 is disposed upstream of the press 11 and is provided to prevent the workpiece W being transported by the transport device 13 from hardening.

[0023] The above-described material temperature sensor 32 detects the temperature of the workpiece W immediately after passing through the heater 14 (the workpiece W near the outlet of the heater 14). The mold temperature sensor 34 detects the temperature of the mold mounted on the press 11.

[0024] In addition to the above-described sensors, the forging press apparatus 10 may further include a timer (not shown) that detects the time required for the work W to reach the press position from the outlet of the heater 14 (heater outlet - press time). Further, a work presence / absence sensor (not shown) that detects the presence or absence of the work W may be provided near the outlet of the heater 14.

[0025] Referring to FIG. 1 again, the forging press apparatus 10 includes an abnormality prediction apparatus 200 for predicting abnormalities in a plurality of parts as shown in FIGS. 2 and 3. The parts to be subjected to abnormality prediction are hereinafter referred to as "target parts". It is desirable that the plurality of target parts include not only a plurality of components mounted on the forging press apparatus 10 but also the material as the work W. The abnormality prediction apparatus 200 predicts the presence or absence of an abnormality in the target part by using the abnormality prediction model generated by the learning apparatus 100. The learning apparatus 100 and the abnormality prediction apparatus 200 constitute an abnormality prediction system SYS. In this specification, "abnormality prediction" is a concept that includes "abnormality detection" that has actually occurred. Therefore, the abnormality prediction apparatus 200 can detect the presence or absence of an abnormality in the target part that has occurred suddenly.

[0026] The learning apparatus 100 and the abnormality prediction apparatus 200 acquire detection signals from a plurality of sensors via the PLC 40 or without passing through the PLC 40. The PLC 40 and these apparatuses 100, 200 are connected by wire or wirelessly. The learning apparatus 100 and the abnormality prediction apparatus 200 are realized by a computer including a processor such as a CPU (Central Processing Unit) and a memory. In this embodiment, an example in which the PLC 40 and the abnormality prediction apparatus 200 are provided separately is shown, but it is not limited, and the PLC 40 may have the function of the abnormality prediction apparatus 200 as described later.

[0027] <Learning apparatus> (Regarding the functional configuration) FIG. 5 is a block diagram showing the functional configuration of the learning device 100 according to the present embodiment. The learning device 100 mainly includes a data acquisition unit 101, a format conversion unit 102, a learning unit 104, and a model storage unit 114. In the model storage unit 114, a group of prediction models M1, M2,... for each die pattern are stored.

[0028] In the present embodiment, the "die pattern" is specified by at least the part number (model number) corresponding to the shape of the die. When the forging press device 10 continuously presses the workpiece W, the part numbers of these workpieces W are common. When the press 11 is a multi-stage type and, for example, three dies 51 to 53 (FIG. 4) are mounted, the "die pattern" is specified by the part number of the workpiece and the workpiece pattern indicating the arrangement pattern of the workpiece W on the dies 51 to 53. The arrangement patterns of the workpiece W on the dies 51 to 53 include a pattern in which the workpiece W is arranged on all of the dies 51 to 53, a pattern in which the workpiece W is arranged only on the first die 51, a pattern in which the workpiece W is arranged only on the last die 53, and the like. Although the arrangement mode of the workpiece W corresponding to the part number is defined one-to-one, since the workpiece W is sequentially conveyed, at the start and end of the press operation, etc., the workpiece W is arranged in a pattern different from the arrangement mode on only a part of the dies 51 to 53. Note that the arrangement modes of the workpiece W corresponding to the part number include a pattern in which the workpiece W is sequentially arranged on all of the dies 51 to 53, a pattern in which the workpiece W is arranged on the dies 51 to 53 skipping one by one, and the like. Information for specifying the die pattern may further include the projected area of the workpiece W.

[0029] The data acquisition unit 101 acquires measurement data as primary data from sensors included in the forging press device 10 during the operation of the forging press device 10. In the present embodiment, the data acquisition unit 101 acquires measurement data (hereinafter referred to as "cycle measurement data") and die pattern data for each press cycle.

[0030] The cycle measurement data acquired by the data acquisition unit 101 is stored in the primary data storage unit 111 in time series using a uniquely determined identification number (hereinafter referred to as "cycle No.") as a heading. Each cycle measurement data DT1 is stored in the primary data storage unit 111 together with the corresponding die pattern data. The die pattern data includes the part number of the target work W and the work pattern as described above. The part number of the work W is input by a user (operator or administrator) via an input unit (not shown), for example, before the start of the press operation by the forging press device 10. The work pattern is detected by work pattern detection means (not shown) provided in the forging press device 10, for example.

[0031] The format conversion unit 102 converts the format of the primary data into secondary data in a format suitable for machine learning. Specifically, the format conversion unit 102 functions as a feature quantity calculation means and calculates feature quantities for each detection item based on the cycle measurement data DT1. Specific examples of the detection items (types of feature quantities) are shown in Table 1 below. In Table 1, a column representing the type of detection item is provided for ease of understanding.

[0032]

Table 1

[0033] The detection items listed in Table 1 are briefly described below. The "material temperature" of item No. I01 is a feature quantity calculated based on the temperature data obtained from the material temperature sensor 32. The "C (clutch) / B (brake) signal feature quantity" of item No. I02 is a feature quantity calculated based on the clutch engagement signal (0 or 1) and the brake release signal (0 or 1). For example, the ON time of the clutch engagement signal, the FF time of the clutch engagement signal, and the OFF time of the brake release signal are represented as a three-dimensional vector with the time point when the brake release signal is ON as the reference time. This feature quantity can detect the timing of the signal ON / OFF.

[0034] The "clutch response feature amount" in item No. I03 is a feature amount (vector) calculated based on the clutch engagement signal and the pressure signal obtained from the clutch pressure sensor 23. The "brake response feature amount" in item No. I04 is a feature amount (vector) calculated based on the brake release signal and the pressure signal obtained from the brake release pressure sensor 22. These feature amounts can detect the pressure change response to the clutch / brake signal.

[0035] The "secondary side clutch pressure" in item No. I05 is a feature amount calculated based on the pressure signal obtained from the clutch pressure sensor 23, and includes at least one of, for example, the wave height vector and the frequency vector. The "brake release pressure" in item I06 is a feature amount calculated based on the pressure signal obtained from the brake release pressure sensor 22, and like the secondary side clutch pressure, includes at least one of, for example, the wave height vector and the frequency vector. Note that these pressure feature amounts only need to include at least one feature amount among the three types of time vector, wave height vector, and frequency vector.

[0036] The "actual main motor speed" in item No. I07 is a feature amount calculated based on the rotation angle signal obtained from the motor speed sensor 36, and is represented by, for example, the wave height vector. Items No. I08 and I09 are the speed and acceleration of slide 1, respectively, and are calculated based on the rotation angle signal from the angle sensor 21. The speed and acceleration of slide 1 are represented by, for example, the wave height vector.

[0037] The "load peak value" in item No. I10 and the "load increase slope" in item No. I11 are load feature amounts that can be calculated based on the load data obtained from the load sensor 20. The load increase slope is calculated as the slope of the linear equation passing through two points, the rising position and the peak position of the forming load curve.

[0038] The "die height value" of item No. I12 is a feature quantity calculated from the output value (die height data) of the die height displacement sensor 33. The "die contact time" of item No. I13 is a feature quantity that can be calculated based on the load data obtained from the load sensor 20, and the time between two points, namely the rising position and the peak position of the forming load curve, is calculated as the die contact time. The die contact time may be calculated as one feature quantity instead of for each of the dies 51 to 53.

[0039] The "die temperature" of items No. I14 to 16 is a feature quantity calculated based on the temperature data obtained from the mold temperature sensors 34 attached to the respective molds 50. The mold temperature sensors 34 are provided in the molds 51 to 53 (for example, the lower molds 51b, 52b, 53b) in each of the processes I to III, and the temperature of each mold 50 is detected. The feature quantity of the die temperature in each process is calculated, for example, as the average temperature of the specified area immediately before pressing (during workpiece transfer). In addition to / instead of the average temperature, the maximum temperature, the minimum temperature, etc. may be adopted as the feature quantity of the die temperature.

[0040] The "die lubricant amount" of items No. I17 to 19 is a feature quantity calculated based on the lubricant amount data obtained from the lubricant flow sensors 30 provided in the lubricant passages to the respective molds 50. The lubricant flow sensors 30 are provided for the molds 51 to 53 in each of the processes I to III, and the flow rate of the lubricant is detected for each mold 50. The feature quantity of the lubricant amount of each mold 50 is, for example, the lubricant amount (total amount) of each mold 50 immediately before pressing (during workpiece transfer).

[0041] The "vibration on the upstream side of the frame" for items No. I20 to I22 is a feature quantity calculated based on the signal from the vibration meter 35 provided on the left support column 3c of the frame 3. The "vibration on the downstream side of the frame" for items No. I23 to I25 is a feature quantity calculated based on the signal from the vibration meter 35 provided on the right support column 3d of the frame 3. The vibration feature quantity of the frame 3 is calculated in each of the three sections before the stroke, before mold contact, and after mold contact. The vibration feature quantity includes at least one of the wave height vector, time vector, and frequency vector. As shown in Table 1, although it is desirable for the vibration feature quantity of the frame 3 to include the wave height vector and frequency vector, it may also include the time vector and frequency vector. Details of the specific calculation method of the vibration feature quantity of the frame 3 will be described later.

[0042] The "vibration on the main motor body side" for items No. I26 to I28 is a feature quantity calculated based on the signal from the vibration meter 37 provided on the main motor 6 body. The "vibration on the motor load side" for items No. I29 to I31 is a feature quantity calculated based on the signal from the vibration meter 37 attached to the main motor 6 for detecting bearing vibration. The vibration feature quantity of the main motor 6 is calculated in each of the three analysis sections before the stroke, before mold contact, and after mold contact. The vibration feature quantity includes at least one of the wave height vector, time vector, and frequency vector. It is also desirable for the vibration feature quantity of the main motor 6 to include the wave height vector and frequency vector.

[0043] The cycle feature data DT2 including the feature quantities of a plurality of detection items calculated by the format conversion unit 102 is stored as secondary data in the secondary data storage unit 112. Each cycle feature data DT2 is stored in the secondary data storage unit 112 in association with, for example, the same cycle No. and mold pattern data as the original cycle measurement data DT1.

[0044] The learning unit 104 learns a large number of cycle feature data DT2 stored in the secondary data storage unit 112 for each classification of detection items (hereinafter referred to as "sensor groups"). Thereby, for each press cycle, an anomaly prediction model for predicting whether the feature amount of each sensor group is abnormal is generated. As the machine learning algorithm of the anomaly prediction model, a known algorithm, for example, isolation forest, is used. The isolation forest is a general machine learning algorithm used for anomaly detection and is a kind of algorithm that does not require teacher data.

[0045] The "Classification No." in Table 1 above is numbered in units of sensor groups. As shown in Table 1, except for some detection items, there is a one-to-one relationship between the detection items and the sensor groups. Regarding the exceptions, the clutch response feature amount and the brake response feature amount (item Nos. I03 to 04) are assigned to a common sensor group (group No. G03). Also, the load peak value, the load increase gradient, the die height value, and the die contact time (item Nos. I10 to I13) are assigned to a common sensor group (group No. G09) as load-related feature amounts. Note that the assignment method of the sensor groups shown in Table 1 is an example, and for example, the die height value and the die contact time may be individual sensor groups.

[0046] In this embodiment, the learning unit 104 learns the feature amounts of each sensor group for each "die pattern". That is, a group of learned models M1, M2,... are generated for the number of die patterns (that is, the number of combinations of product numbers and work patterns). Each of the learned model groups includes the same number of anomaly prediction models m1, m2,... as the sensor groups.

[0047] A plurality of learned model groups M1, M2, ··· are stored in the model storage unit 114. That is, in the model storage unit 114, for each mold pattern (associated with mold identification information for identifying the mold pattern), anomaly prediction models m1, m2, ··· for each sensor group are stored. Each anomaly prediction model is stored in association with identification information (classification No.) for identifying the corresponding sensor group.

[0048] The functions of the above-described data acquisition unit 101, format conversion unit 102, and learning unit 104 are realized by a processor executing software. The primary data storage unit 111, secondary data storage unit 112, and model storage unit 114 are typically constituted by non-volatile storage devices provided in a computer.

[0049] (Regarding the operation) FIG. 6 is a flowchart showing a method for generating an anomaly prediction model by the learning device 100 according to the present embodiment. The model generation process shown in FIG. 6 is realized by a processor of the learning device 100 executing a learning program stored in advance in a memory. Note that this process is started in response to an instruction to start learning being input by a user via an operation unit (not shown). In the present embodiment, it is assumed that a plurality of cycle measurement data DT1 are stored in time series in the primary data storage unit 111 in association with cycle No. and mold pattern data before this process is started.

[0050] Referring to FIG. 6, the learning device 100 specifies one mold pattern from the combinations of the part number of the workpiece W and the workpiece pattern (in a predetermined order) (step S2), and searches for the cycle measurement data DT1 associated with that mold pattern in the primary data storage unit 111 (step S4).

[0051] After that, the format conversion unit 102 reads all the retrieved (target) cycle measurement data DT1 from the primary data storage unit 111 (step S6), and converts each read cycle measurement data DT1 into cycle feature data DT2 (step S8). That is, the format conversion unit 102 calculates the feature amounts of the detection items shown in Table 1 above based on each data included in the cycle measurement data DT1, and records them in the secondary data storage unit 112 as the cycle feature data DT2. Here, as an example, an example of calculating the vibration feature amounts (item Nos. I20 to I25) of frame 3 will be described with reference to FIGS. 7 and 8.

[0052] FIG. 7(A) is a graph showing the stroke of slide 1 along the time axis. FIGS. 7(B) and (C) are graphs showing the transition of the magnitude of vibration (pulse) along the same time axis as FIG. 7(A), and the unit of the vertical axis is "m / s" as an example. 2 ". FIG. 8(B) is a graph showing the stroke of slide 1, similar to FIG. 7(A). FIG. 8(A) is a graph showing the brake release signal along the same time axis. FIG. 8(C) is a graph showing the transition of the magnitude of vibration (pulse), similar to FIGS. 7(B) and (C). FIG. 8(D) is a graph showing the transition of the power spectral density with the frequency on the horizontal axis (unit: Hz). Here, it is assumed that the vibration detected by the vibration meter 35 attached to the support column 3c on the upstream side (rear left side) of frame 3 is shown.

[0053] With reference to FIG. 7(B), the method for calculating the wave height vector will be described. The format conversion unit 102 extracts the pulses in the section TS that exceeds the baseline (predetermined value) BL, and divides the extracted section TS into d strip-shaped sections along the time axis. Then, the pulse data is processed into a d-dimensional vector (p = [h1, h2, ···, h j , ···, h d ) to calculate the wave height vector. However, "h j " is the average value within each strip.

[0054] Referring to FIG. 7(C), the method for calculating the time vector will be described. The format conversion unit 102 divides the pulses within the interval TS into d w pieces in the wave height direction, and acquires 2d w times that intersect this dividing line (horizontal line). Then, the time data is processed into a 2d w -dimensional vector (q = [w1, w2, ···, w j , ···, w 2d ) to calculate the time vector. However, for "w j ", two outer intersections with respect to the pulse peak position are selected. For example, focusing on the dividing line Lx, the times t1 and t4 of the outer intersections are selected, and the times t2 and t3 are not selected. The time t1 corresponds to the time when the upper molds of the molds 51 to 53 start to contact the workpiece W, and the time t4 corresponds to the time when the upper molds of the molds 51 to 53 start to separate from the workpiece W.

[0055] Referring to FIG. 8, the method for calculating the frequency vector will be described. The format conversion unit 102 cuts out three types of analysis intervals from the vibration waveform shown in FIG. 8(C), namely, the "pre-stroke interval", the "pre-mold contact interval", and the "mid-mold contact interval". The "pre-stroke interval" is the interval before the time point t11 when the brake release signal shown in FIG. 8(A) becomes ON. The "pre-mold contact interval" is the interval from the time point t11 to the time point t12 when the stroke position shown in FIG. 8(B) becomes the mold contact position. The "mid-mold contact interval" is the interval from the time point t12 to the time point t13 when the brake release signal shown in FIG. 8(A) becomes OFF.

[0056] The format conversion unit 102 performs Fourier transform on the vibration signal for each analysis interval and calculates the power spectral density PSD. FIG. 8(D) shows the waveform of the power spectral density from a frequency of 0 to fs / 2 Hz. "fs" is the sampling frequency, and in this example, fs / 2 = 50.

[0057] The format conversion unit 102, for example, has a lower frequency f min = 0 and an upper frequency f maxSet it as = fs / 2, divide this interval FS into d (for example, 20) strip-shaped segments, and process the power spectral density PSD within the interval FS into a d-dimensional vector (p = [h1, h2, ···, h j , ···, h d , where "h j " is the average value within each individual strip). In this way, a frequency vector for each analysis interval is obtained.

[0058] Note that the analysis interval used for calculating the frequency vector may also be used for calculating the wave height vector described above. That is, as schematically shown in Fig. 8(C), each analysis interval can be divided into d (for example, 20) strip-shaped segments, and for each analysis interval, the pulse data can be processed into a d-dimensional vector to calculate the wave height vector.

[0059] By the same calculation method as the method for calculating the vibration feature quantity of frame 3 as described above, it is possible to calculate, for example, the wave height vector and frequency vector of the analysis interval of the actual speed of the main motor.

[0060] The cycle feature data DT2 including the feature quantities for each detection item calculated by the above method is recorded in the secondary data storage unit 112 in association with the same cycle No. as the target cycle measurement data DT1. As shown in Fig. 5, the cycle feature data DT2 may also be stored in association with the target die pattern data.

[0061] The learning unit 104 performs machine learning on the cycle feature data DT2 obtained in step S8 in units of pre-set sensor groups (step S12). That is, the learning unit 104 generates an anomaly prediction model for each sensor group. When the learning unit 104 generates an anomaly prediction model, it stores it in the model storage unit 114 in association with the die identification information for identifying the die pattern (product number, work pattern) specified in step S2 (step S14).

[0062] Among the combination of the part number of the workpiece W and the workpiece pattern, if there is an untreated mold pattern (YES in step S16), the process returns to step S2 and the above process is repeated. When the processing for all the mold patterns is completed (NO in step S16), a series of model generation processes are terminated. As a result, in the model storage unit 114, a group of learned models M1, M2, ··· are stored in the number corresponding to the combination of the part number of the workpiece W and the workpiece pattern. In the example of FIG. 5, it is shown that the learned model group M1 includes an abnormality prediction model with a part number of "30" and a workpiece pattern No. of "15", and the learned model group M2 includes an abnormality prediction model with a part number of "80" and a workpiece pattern No. of "15".

[0063] <Abnormality prediction device> (Regarding the functional configuration) FIG. 9 is a block diagram showing the functional configuration of the abnormality prediction device 200 according to the present embodiment. The abnormality prediction device 200 mainly includes a model storage unit 213, a combination information storage unit 214, a data acquisition unit 202, a format conversion unit 203, a prediction unit 204, and an output unit 230.

[0064] The model storage unit 213 stores in advance a group of learned models M1, M2, ··· generated by the learning device 100 for each mold pattern. The learning device 100 and the abnormality prediction device 200 are connected via, for example, a network, and a communication unit (not shown) of the abnormality prediction device 200 receives the group of learned models M1, M2, ··· stored in the model storage unit 114 from a communication unit (not shown) of the learning device 100 and stores the received group of learned models M1, M2, ··· in the model storage unit 213. Alternatively, the group of learned models M1, M2, ··· stored in the model storage unit 114 of the learning device 100 may be stored in the model storage unit 213 using a removable recording medium.

[0065] The combination information storage unit 214 stores combination information associating each target part with at least one sensor group. In the present embodiment, the combination information is stored as a correspondence table TB. An example of the data structure of the correspondence table TB will be described later.

[0066] After the operation of the forging press device 10 starts, the data acquisition unit 202 acquires measurement data (cycle measurement data) DT11 of the same type as the cycle measurement data DT1 in the learning device 100, together with the die pattern data, for each press cycle.

[0067] The format conversion unit 203 functions as a feature quantity calculation means, similarly to the format conversion unit 102 of the learning device 100. That is, based on the cycle measurement data DT11, a plurality of types of feature quantities shown in Table 1 above are calculated, and cycle feature data DT12 including these feature quantities is generated.

[0068] The prediction unit 204 uses the abnormality prediction models m1, m2,... (Fig. 5) corresponding to the die pattern to detect the presence or absence of abnormality of each feature quantity, and determines the presence or absence of abnormality of the target part based on the detection result. Specifically, as shown in Fig. 9, the prediction unit 204 includes, as its functions, a model discrimination unit 205, an abnormality detection unit 206, and an abnormality determination unit 207.

[0069] The model discrimination unit 205 discriminates the abnormality prediction models m1, m2,... to be used by specifying the learned model group corresponding to the die pattern of the current cycle among the plurality of learned model groups M1, M2,... stored in the model storage unit 213 based on the acquired die pattern data.

[0070] The abnormality detection unit 206 inputs the feature quantity calculated by the format conversion unit 203 into the corresponding abnormality prediction model (among the plurality of abnormality prediction models m1, m2,... discriminated by the model discrimination unit 205) for each sensor group, and detects whether the feature quantity is abnormal.

[0071] The abnormality determination unit 207 determines the presence or absence of abnormality for each target part based on the detection result by the abnormality detection unit 206 and the correspondence table TB.

[0072] Here, an example of the data structure of the correspondence table TB is shown in FIG. 10. FIG. 10 illustrates a correspondence table in which sensor groups and additional logic are described in the left column, and multiple target part names are described in the upper column. In the column of the sensor group, the sensor group name (classification name of the detection item) corresponding to the classification No. in Table 1 is described one by one. In the additional logic column, when there is a determination item associated with each sensor group, the determination item name is described one by one. The additional determination item names are described in rows different from the row in which the sensor group name is described. In the correspondence table TB, at least one mark (〇 mark in FIG. 10) is attached to the column of each target part, and for each target part, the sensor group name (and additional determination item name) used for its abnormality determination is stored in association.

[0073] In the present embodiment, the target parts include the material heater 14, a C / B control device (not shown), the clutch 9, the pneumatic circuit of the clutch 9 (not shown), the brake device 12, the pneumatic circuit of the brake device 12 (not shown), the main motor 6, a bearing (not shown), the material which is the work W, and the mold 50 (51 to 53). Note that the target parts may not include all of these, or may include other components. Desirably, the target parts include at least one of a material heater, a clutch, a brake, a main motor, a bearing, and a mold, and the material which is the work.

[0074] Regarding an example of the additional logic, in the present embodiment, as additional determination items for each classification of "material temperature", "load-related feature quantity", "frame vibration", and each classification of "motor vibration", the presence or absence of continuous abnormality (whether it has been an abnormal value continuously for a predetermined number of times) is included. Also, "clutch reaction time" and "brake control angle" may be newly calculated, and whether each exceeds a threshold value may be used as an additional determination item.

[0075] In the example of FIG. 10, for the "material heater", marks are attached only to the rows of the material temperature (classification No. G01) and its additional logic (logic No. L01), and the material temperature and its additional logic are associated with and stored for the "material heater". In this case, the abnormality determination unit 207 determines the presence or absence of an abnormality in the "material heater" using the abnormality detection result of the material temperature and the abnormality detection result of its additional logic.

[0076] For the "main motor", marks are attached to the rows of the actual main motor speed, slide speed, motor body side vibration (before stroke, before die contact), and motor load side vibration (before stroke, before die contact). For the "bearing", similar to the "main motor", marks are attached not only to the rows of the actual in-motor speed, slide speed, motor body side vibration (before stroke, before die contact), and motor load side vibration (before stroke, before die contact), but also to the rows of the frame upstream side vibration (before stroke, before die contact), frame downstream side vibration (before stroke, before die contact), and the additional logic of the frame upstream side vibration or frame downstream side vibration.

[0077] For the "die" in each process, marks are attached to the row of the load-related feature quantity, the rows of the die temperature and lubricant amount of the corresponding process, and the row of the frame upstream side vibration and / or frame downstream side vibration (after die contact). Thereby, it is possible to individually determine the presence or absence of an abnormality in the dies 51 to 53. As shown in FIG. 10, an additional logic of the load-related feature quantity may be further used for the abnormality determination of the dies 51 to 53. Also, for example, when the lubricant amounts of the dies 51 to 53 can be measured for each of the upper die and the lower die, the presence or absence of an abnormality for each of the upper die and the lower die may be determined by using these feature quantities.

[0078] In this way, by changing the type and number of associated sensor groups for each target part, it becomes possible to automatically discriminate (predict) which part is abnormal. Note that the combination of a target part and at least one sensor group as shown in FIG. 10 is typically set in advance by an administrator.

[0079] The correspondence table TB is preferably updatable by an administrator. The update of the correspondence table TB may be executed via an input unit (operation unit) provided in the abnormality prediction device 20, or may be executed in a management device (not shown). Thereby, it is possible to increase or decrease the parts to be the target of abnormality prediction, or to add or change the type of sensor group associated with the target parts or additional logic.

[0080] Referring to FIG. 9 again, the output unit 230 outputs the prediction result by the prediction unit 204, that is, the determination result by the abnormality determination unit 207. The output unit 230 outputs the presence or absence of an abnormality of each target part for each press cycle. Thereby, the operator can grasp the presence or absence of an abnormality of the target part in real time. The output unit 230 may further output the degree of influence on the prediction result of the sensor group listed in Table 1 and / or the reliability of the abnormality prediction result for each target part.

[0081] The functions of the data acquisition unit 202, the format conversion unit 203, and the prediction unit 204 described above are realized by a processor executing software. The model storage unit 213 and the combination information storage unit 214 are typically configured by a non-volatile storage device included in a computer. The output unit 230 is typically configured by a display unit that displays a prediction result. Note that the output unit 230 may be configured by a communication unit that outputs the prediction result to the monitoring PC 42 shown in FIG. 1 via the PLC 40 or without passing through the PLC 40.

[0082] (Regarding the operation) FIG. 11 is a flowchart showing a method for predicting an abnormality of a target part by the abnormality prediction device 200 according to the present embodiment. The abnormality prediction process shown in FIG. 11 is realized by a processor of the abnormality prediction device 200 executing an abnormality prediction program stored in advance in a memory.

[0083] Referring to FIG. 11, when the operation of the forging press device 10 is started, the data acquisition unit 202 acquires cycle measurement data DT11 including output data from predetermined sensors provided in the forging press device 10, together with the die pattern data (step S26).

[0084] Subsequently, the format conversion unit 203 converts the cycle measurement data DT11 acquired by the data acquisition unit 202 in step S26 into cycle feature data DT12 (step S28). That is, the format conversion unit 203 calculates the feature amounts for each detection item shown in Table 1 above based on each data included in the cycle measurement data DT11.

[0085] Also, the model discrimination unit 205 of the prediction unit 204 specifies the learned model group corresponding to the current die pattern among the plurality of learned model groups M1, M2,... stored in the model storage unit 213 based on the die pattern data acquired by the data acquisition unit 202 in step S26, and discriminates a plurality of anomaly prediction models m1, m2,... to be used this time (step S29). The discrimination process of the anomaly prediction models may be performed in parallel with the data format conversion process.

[0086] The anomaly detection unit 206 of the prediction unit 204 inputs the plurality of types of feature amounts (for one cycle) included in the cycle feature data DT12 obtained in step S28 to the anomaly prediction models discriminated in step S29 for each sensor group (step S30). As shown in the image diagram of FIG. 12, each anomaly prediction model outputs either "anomaly present" or "no anomaly". Thereby, it is possible to detect (predict) whether the feature amounts of each sensor group are abnormal in terms of press cycle units. Note that each anomaly prediction model may output its reliability together with the detection result of the presence or absence of an anomaly. The reliability is calculated, for example, based on the difference between the input feature amount and the threshold value.

[0087] Subsequently, the anomaly determination unit 207 of the prediction unit 204 calculates the values of the determination items defined as additional logic in the correspondence table TB (step S32). For example, regarding "material temperature", the presence or absence of continuous anomalies is determined.

[0088] Thereafter, the abnormality determination unit 207 performs abnormality determination of the target parts based on the correspondence table TB (step S34). Specifically, based on the output of each abnormality prediction model (abnormal or not) and the calculation result of the value of the additional logic (additional determination result), and the correspondence table TB (combination information), the presence or absence of abnormality is determined for each target part. At this time, the abnormality determination unit 207 may perform weighting based on the reliability of the abnormality detection result for each sensor group in the abnormality detection unit 206 to perform abnormality determination of the target parts.

[0089] In step S34, it is desirable for the abnormality determination unit 207 to calculate the degree of influence of the sensor group on the abnormality determination result for each target part. Also, it is desirable to calculate the reliability of the abnormality determination result for each target part.

[0090] The output unit 230 outputs the determination result of the abnormality determination unit 207, that is, the presence or absence of abnormality of each target part (step S36). In the example of FIG. 9, it is displayed that the mold 51 in the first process is abnormal. Note that only the target parts determined to be abnormal may be displayed.

[0091] Also, the output unit 230 outputs, for example, by graphing the importance (degree of influence) of each sensor group on the abnormality determination result of the target part, and outputs its reliability as, for example, a numerical value (%).

[0092] As described above, in this embodiment, i) the abnormality prediction result for each target part, ii) the importance of each sensor group with respect to the abnormality prediction result, and iii) the reliability of the abnormality prediction result are output. Thereby, it is possible to easily take measures against the abnormality prediction result.

[0093] The processes of steps S26 to S38 are repeatedly executed until the press operation ends (NO in step S40). If there are parts determined to be abnormal in step S34, the operation of the forging press apparatus 10 may be automatically stopped. Thereby, it is possible to prevent or suppress the mass production of defective products by the press 11.

[0094] As described above, the abnormality prediction system SYS according to the present embodiment includes the learning device 100 that generates an abnormality prediction model for each sensor group. The abnormality prediction device 200 can determine the presence or absence of an abnormality in each target part for each press cycle by using these abnormality prediction models and the correspondence table TB. Therefore, according to the present embodiment, it is possible to automatically and real - time predict the abnormalities of a plurality of parts in the forging press apparatus 10 without relying on the experience of the operator.

[0095] Also, in the present embodiment, since the abnormality prediction model is learned for each die pattern and stored in the model storage unit 213 of the abnormality prediction device 200, it is possible to improve the accuracy of detecting abnormalities in feature amounts for each sensor group and predicting abnormalities for each target part.

[0096] Also, in the present embodiment, since the above - described additional logic is added to the abnormality prediction of the target part, the accuracy can be improved as compared with the form in which the abnormality of the target part is predicted only based on the output of the abnormality prediction model.

[0097] (Modification example) In the present embodiment, the abnormality prediction models m1, m2,... (learned model groups M1, M2,...) for each sensor group are stored in advance in the model storage unit 213 of the abnormality prediction device 200. However, the abnormality prediction models m1, m2,... in the model storage unit 213 may be updated according to the machine learning by the learning device 100. Alternatively, the abnormality prediction device 200 may have the functions of the learning device 100.

[0098] Further, although it has been assumed that the combination information storage unit 214 of the abnormality prediction device 200 stores a correspondence table TB in a table format as shown in FIG. 10 as combination information associating each target part with at least one sensor group, the present invention is not limited to such an example.

[0099] In the present embodiment, the forging press device 10 is provided with the abnormality prediction device 200. However, the present invention is not limited thereto, and the abnormality prediction device 200 may predict the presence or absence of an abnormality in a target part offline.

[0100] Note that the abnormality prediction method executed by the abnormality prediction device 200 or the learning method executed by the learning device 100 can also be provided as a program. Such a program can be recorded and provided on a computer-readable non-transitory recording medium such as an optical medium like a CD-ROM (Compact Disc-ROM) or a memory card. Further, the program can also be provided by downloading via a network.

[0101] The program according to the present invention may be a program that calls necessary modules in a predetermined array at a predetermined timing among program modules provided as part of a computer operating system (OS) to execute processing. In that case, the program itself does not include the above-mentioned modules, and the processing is executed in cooperation with the OS. A program that does not include such modules may also be included in the program according to the present invention.

[0102] Further, the program according to the present invention may be provided by being incorporated into a part of another program. Also in that case, the program itself does not include the modules included in the above-mentioned other program, and the processing is executed in cooperation with the other program. A program incorporated into such another program may also be included in the program according to the present invention.

[0103] The embodiments disclosed this time should be considered as illustrative in all respects and not restrictive. The scope of the present invention is shown not by the above description but by the claims, and it is intended that all modifications within the meaning and scope equivalent to the claims be included.

Explanation of Signs

[0104] 10 Forging press device, 11 Press machine, 100 Learning device, 200 Abnormality prediction device, 202 Data acquisition unit, 203 Format conversion unit, 204 Prediction unit, 205 Model discrimination unit, 206 Abnormality detection unit, 207 Abnormality determination unit, 213 Model storage unit, 214 Combination information storage unit, 230 Output unit, m1, m2, ··· Abnormality prediction models, SYS Abnormality prediction system, W Workpiece.

Claims

1. An abnormality prediction device for predicting abnormalities in a plurality of parts in a press device, By machine learning the feature amounts of a plurality of detection items obtained based on the measurement data of sensors mounted on the press device for each sensor group in which the plurality of detection items are classified by type, a model storage means for storing an abnormality prediction model generated for each sensor group; An acquisition means for acquiring measurement data from the sensors during operation of the press device; A feature amount calculation means for calculating the feature amount of each of the plurality of detection items based on the measurement data acquired by the acquisition means; For each sensor group, a detection means for inputting the feature amount calculated by the feature amount calculation means into the corresponding abnormality prediction model to detect whether the feature amount is abnormal; Combination information storage means for storing combination information associating each target part with at least one of the sensor groups; A determination means for determining the presence or absence of an abnormality for each target part based on the detection result by the detection means and the combination information stored in the combination information storage means, In the model storage means, the abnormality prediction models for each sensor group are stored at least by mold pattern specified by the part number of the material, The acquisition means acquires mold pattern data together with the measurement data, An abnormality prediction device further comprising a model discrimination means for discriminating an abnormality prediction model to be used from among the plurality of abnormality prediction models stored in the model storage means based on the mold pattern data acquired by the acquisition means.

2. The determination means determines the presence or absence of an abnormality for each target part by further using an additional determination result of the presence or absence of a continuous abnormality for at least one of the plurality of sensor groups. The abnormality prediction device according to claim 1.

3. The abnormality prediction device according to claim 1 or 2, further comprising an output means for outputting, together with the determination result by the determination means, the degree of influence of the sensor group on the determination result and / or the reliability of the determination result.

4. The target part includes at least one of a material heater, a clutch, a brake, a main motor, a bearing, and a mold, and a material that is a workpiece. The abnormality prediction device according to any one of claims 1 to 3.

5. The target part includes a plurality of molds, The abnormality prediction device according to any one of claims 1 to 4, wherein the plurality of sensor groups include a vibration feature amount on the upstream side of the frame and a vibration feature amount on the downstream side of the frame.

6. An abnormality prediction program for predicting an abnormality of a plurality of parts in a press device equipped with sensors, using an abnormality prediction model, The abnormality prediction model is generated by performing machine learning for each sensor group obtained by classifying a plurality of detection items into types, based on the feature amounts of the plurality of detection items obtained based on the measurement data of the sensors, for each die pattern specified by at least the part number of the material. A step of acquiring measurement data from the sensors mounted on the press device and die pattern data during operation of the press device; A calculation step of calculating the feature amount of each of the plurality of detection items based on the measurement data acquired in the acquisition step; A step of determining an abnormality prediction model to be used from among the plurality of abnormality prediction models based on the die pattern data acquired in the acquisition step; A detection step of inputting the feature amount calculated in the calculation step into the abnormality prediction model determined in the determination step for each sensor group, and detecting whether the feature amount is abnormal; An abnormality prediction program that causes a computer to execute a determination step of determining the presence or absence of an abnormality for each target part based on the detection result in the detection step and combination information associating each target part with at least one of the sensor groups.

Images