JP2022142459A - Abnormality prediction device and program - Google Patents

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

The present invention relates to an abnormality prediction device and program, and more particularly to an abnormality prediction device and program for predicting an abnormality of a plurality of parts in a press machine.

A forging press device presses a workpiece by rotationally driving a crankshaft to move a slide up and down. In such a forging press apparatus, as described in Japanese Patent Application Laid-Open No. 2019-13947 (Patent Document 1), the load applied to the workpiece during processing is monitored, and abnormalities in the product are detected based on the peak value of the load. Detecting is conventional. This is because it is known that the peak value of the load has a high correlation with the quality of the product.

Further, as described in Japanese Patent Laying-Open No. 2019-13976 (Patent Document 2), there has been conventionally proposed a technique of using a plurality of sensors for detecting the operating conditions of the press machine for failure prediction of the press machine. there is

JP 2019-13947 A JP 2019-13976 A

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

In Patent Document 2, changes from the initial normal state of the measurement results obtained from a plurality of sensors are confirmed by a monitoring PC, and past empirical rules are also compared to predict failure of the press machine. However, it is difficult to accurately predict failures.

The present invention has been made to solve the above problems, and its object is to provide an abnormality prediction device capable of automatically predicting abnormality of a plurality of parts in a press machine such as a forging press machine, and to provide the program.

An abnormality prediction device according to one aspect of the present invention is an abnormality prediction device for predicting abnormality of a plurality of parts in a press machine, comprising model storage means, acquisition means, feature quantity calculation means, detection means, A combination information storage means and a determination means are provided. The model storage means stores an abnormality prediction model generated for each sensor group by machine-learning feature amounts of a plurality of sensor groups obtained based on measurement data of sensors mounted on the press. The acquisition means acquires measurement data from the sensors during operation of the press device. The feature amount calculation means calculates feature amounts of each of the plurality of sensor groups based on the measurement data acquired by the acquisition means. The detection means inputs the feature quantity calculated by the feature quantity calculation means to the corresponding abnormality prediction model for each sensor group, and detects whether or not the feature quantity is abnormal. The combination information storage means stores combination information that associates each target part with at least one sensor group. The judging means judges whether or not there is an abnormality for each target part based on the result of detection by the detecting means and the combination information stored in the combination information storage means.

Preferably, the model storage means stores an abnormality prediction model for each sensor group at least for each mold pattern specified by the product number of the material. In this case, the acquisition means acquires the mold pattern data together with the measurement data, and the abnormality prediction device stores the plurality of abnormality prediction models stored in the model storage means based on the mold pattern data acquired by the acquisition means. Among others, it is desirable to further include model discrimination means for discriminating the abnormality prediction model to be used.

Preferably, the abnormality prediction apparatus further includes output means for outputting 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.

Preferably, the target parts include at least one of material heaters, clutches, brakes, main motors, bearings, and dies, and materials that are workpieces.

A target part may include multiple molds. In this case, it is desirable that the plurality of sensor groups include the frame upstream side vibration feature amount and the frame downstream side vibration feature amount.

An abnormality prediction program according to another aspect of the present invention is an abnormality prediction program for predicting an abnormality in a plurality of parts in a press machine, wherein measurement data is collected from sensors mounted on the press machine during operation of the press machine. A step of obtaining, a calculation step of calculating feature values for each of a plurality of sensor groups based on the obtained measurement data, and inputting the feature values calculated in the calculation step to the corresponding abnormality prediction model for each sensor group. Then, based on a detection step for detecting whether or not the feature amount is abnormal, the detection result in the detection step, and combination information that associates each target part with at least one sensor group, an abnormality is determined for each target part. and a judgment step of judging the presence/absence.

ADVANTAGE OF THE INVENTION According to this invention, the abnormality of several parts in a press apparatus can be automatically predicted.

BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows schematic structure of the forging press apparatus which concerns on embodiment of this invention. BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows typically the press and sensors with which the forging press apparatus which concerns on embodiment of this invention is provided. FIG. 3 is a diagram schematically showing the peripheral structure of the press and the operating state information detected by sensors in the embodiment of the present invention; It is a figure which shows typically the example of arrangement|positioning of the metal mold|die with which the press in embodiment of this invention is equipped. 1 is a block diagram showing the functional configuration of a learning device according to an embodiment of the present invention; FIG. 4 is a flow chart showing a prediction model generation method by the learning device according to the embodiment of the present invention. (A) to (C) are graphs for explaining a method of calculating a feature amount of frame vibration. (A) to (D) are graphs for explaining a method of calculating a feature amount of frame vibration. 1 is a block diagram showing the functional configuration of an abnormality prediction device according to an embodiment of the present invention; FIG. It is a figure which shows the data structure example of the correspondence table in embodiment of this invention. 4 is a flow chart showing an abnormality prediction method for a target part by the abnormality prediction device according to the embodiment of the present invention; FIG. 12 is an image diagram of prediction processing in step S30 of FIG. 11;

An embodiment of the present invention will be described in detail with reference to the drawings. The same or corresponding parts in the drawings are denoted by the same reference numerals, and the description thereof will not be repeated.

<Schematic configuration of forging press>
First, referring to FIGS. 1 to 3, a schematic configuration of a forging press device 10 will be described. As shown in FIG. 1, the forging press apparatus 10 includes a press machine 11 that presses a workpiece material, and a PLC (Programmable Logic Controller) 40 that controls the operation of the press machine 11 .

As shown in FIG. 2, the press machine 11 has a slide 1 and a bolster 2 which are provided in a frame 3 so as to face each other vertically. The frame 3 includes a pair of struts 3 a and 3 b arranged on the front side of the press machine 11 and a pair of struts 3 c and 3 d 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 .

A flywheel 8 rotatably driven by a belt 7 is connected via a clutch 9 to one end of the crankshaft 4 . A brake device 12 is attached to the other end of the crankshaft 4 to stop the slide 1 by stopping the rotation of the crankshaft 4 . The press machine 11 rotates the flywheel 8 at a constant speed with the main motor 6, engages the clutch 9, and transmits the rotation of the flywheel 8 to the crankshaft 4, thereby elevating the slide 1 and moving it on the bolster 2. Forging the work placed on. The workpiece is processed into a size and shape according to the pattern of the mold including the lower mold (not shown) installed on the bolster 2 and the upper mold (not shown) installed at the lower end of the slide 1. , the final product. "Die" means a die used for forging. The mold is typically made of special steel.

A forging press device 10 in the present embodiment is a device for processing a work by hot forging. The forging press machine 10 is provided with a plurality of sensors as shown below in order to detect the operating conditions of the press machine 11. As shown in FIG. As will be described later, the press machine 11 may be a multi-stage press machine that processes a workpiece through a plurality of processes. That is, as shown in FIG. 4, the press machine 11 may have a plurality (for example, three) of molds 51 to 53 mounted thereon. In the following description, when the molds 51 to 53 do not need to be distinguished, they are referred to as "mold 50".
Load sensor 20: Detects the load applied to the work. For example, it is composed of a strain gauge that detects 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 rate sensor 26 : Detects the flow rate of cooling water for the brake device 12 .
BKO position displacement sensor 27: detects the position of the lower knockout device of the press machine 11 (BKO position).
SKO position displacement sensor 28: detects the position of the upper knockout device (SKO position).
· Temperature measuring resistor 29 : Detects the temperature of each part of the press machine 11 .
- Lubricant flow rate sensor 30: detects the flow rate of the lubricating liquid that lubricates the mold.
· 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 of the press machine 11 (the distance between the bolster 2 and the bottom dead center position of the slide 1).
- Mold temperature sensor 34: detects the temperature of the mold.
Vibration meter 35 : configured by, for example, an acceleration sensor, and detects vibrations 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 rotation shaft of the main motor 6;
· Motor vibration sensor 37 : It is composed of, for example, an acceleration sensor, and detects vibration of the main motor 6 .

In the present embodiment, the load sensor 20 described above is attached to one of the columns of the frame 3 (for example, the column 3c). Vibration meters 35 are attached to the left and right struts 3c and 3d of the frame 3, respectively. One of the columns 3c and 3d (column 3c) is located on the upstream side in the conveying direction of the mold, and the other of the columns 3c and 3d (column 3d) is located on the downstream side of the mold in the conveying direction. Each vibrometer 35 detects vibration in the substantially horizontal direction (back and forth direction or left and right direction) of the struts 3c and 3d to which it is attached.

As shown in FIG. 3 , the forging press device 10 further includes a carrier device 13 and a heater 14 . The conveying device 13 is composed of, for example, a belt conveyor, and conveys the work W from the previous process to the press machine 11 (position of the bolster 2). The heater 14 is arranged on the upstream side of the press machine 11 and is provided to prevent hardening of the work W being transported by the transport device 13 .

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

In addition to the above-described sensors, the forging press machine 10 includes a timer (not shown) for detecting the time required for the workpiece W to reach the press position from the exit of the heater 14 (heater exit - press time). may further include Further, a work presence/absence sensor (not shown) for detecting the presence/absence of the work W may be further provided near the outlet of the heater 14 .

Referring to FIG. 1 again, the forging press machine 10 includes an abnormality prediction device 200 for predicting abnormality of a plurality of parts as shown in FIGS. 2 and 3. FIG. A part targeted for abnormality prediction is hereinafter referred to as a “target part”. The plurality of target parts desirably include not only the plurality of component parts mounted on the forging press device 10 but also the material as the work W. The abnormality prediction device 200 uses the abnormality prediction model generated by the learning device 100 to predict the presence or absence of abnormality in the target part. The learning device 100 and the abnormality prediction device 200 constitute an abnormality prediction system SYS. In this specification, "abnormality prediction" is a concept including "abnormality detection" that actually occurred. Therefore, the abnormality prediction device 200 can detect whether or not there is a sudden abnormality in the target part.

Learning device 100 and abnormality prediction device 200 acquire detection signals from a plurality of sensors via PLC 40 or without PLC 40 . The PLC 40 and these devices 100 and 200 are connected by wire or wirelessly. Learning device 100 and anomaly prediction device 200 are implemented by a computer including a processor such as a CPU (Central Processing Unit) and a memory. In the present embodiment, an example in which the PLC 40 and the abnormality prediction device 200 are provided separately is shown, but this is not limiting, and the PLC 40 has the function of the abnormality prediction device 200 as described later. may

<Learning device>
(About functional configuration)
FIG. 5 is a block diagram showing the functional configuration of 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 . The model storage unit 114 stores prediction model groups M1, M2, . . . for each mold pattern.

In the present embodiment, the "mold pattern" is specified by at least a product number (model number) corresponding to the shape of the mold. When the works W are continuously pressed by the forging press machine 10, the part numbers of these works W are common. When the press machine 11 is of a multistage type and, for example, three molds 51 to 53 (FIG. 4) are mounted, the "mold pattern" is the part number of the workpiece and the workpiece to the molds 51 to 53. It is specified by a work pattern that indicates the W arrangement pattern. The arrangement patterns of the workpieces W on the molds 51 to 53 include a pattern in which the workpieces W are arranged in all the molds 51 to 53, a pattern in which the workpieces W are arranged only in the first mold 51, and a pattern in which the workpieces W are arranged only in the first mold 51. A pattern in which the workpiece W is arranged only at 53, and the like are included. Although the arrangement mode of the work W corresponding to the product number is determined on a one-to-one basis, since the work W is conveyed sequentially, the pattern different from the arrangement mode of the work W is used at the start and end of the press operation. , the workpiece W is placed only on a part of the molds 51 to 53 . The arrangement mode of the workpiece W corresponding to the part number includes a pattern in which the workpieces W are arranged sequentially on all of the molds 51 to 53, a pattern in which the workpieces W are arranged in the molds 51 to 53 by skipping one, and the like. be The information for specifying the mold pattern may further include the projected area of the workpiece W.

The data acquisition unit 101 acquires measurement data as primary data from sensors included in the forging press device 10 when the forging press device 10 is in operation. In this embodiment, the data acquisition unit 101 acquires measurement data (hereinafter referred to as “cycle measurement data”) and die pattern data for each press cycle.

The cycle measurement data acquired by the data acquisition unit 101 is stored in the primary data storage unit 111 in time series with a unique identification number (hereinafter referred to as “cycle No.”) as an index. Each cycle measurement data DT1 is stored in the primary data storage unit 111 together with the corresponding mold pattern data. The mold pattern data includes the product number and work pattern of the target work W, as described above. The product number of the workpiece W is input by a user (operator or administrator) via an input unit (not shown) before the forging press device 10 starts the press operation, for example. The work pattern is detected, for example, by work pattern detection means (not shown) provided in the forging press device 10 .

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

The detection items listed in Table 1 are briefly described below. Item no. "Material temperature" of I01 is a feature amount calculated based on the temperature data obtained from the material temperature sensor 32. FIG. Item no. The "C (clutch)/B (brake) signal feature amount" of I02 is a feature amount calculated based on the clutch engagement signal (0 or 1) and the brake loosening signal (0 or 1). The ON time of the signal, the FF time of the clutch engagement signal, and the OFF time of the brake slackness signal are represented by a three-dimensional vector with the point of time when the brake slackness signal is turned ON as a reference time. The timing of signal ON/OFF can be detected from this feature amount.

Item no. The “clutch response feature quantity” of I03 is a feature quantity (vector) calculated based on the clutch engagement signal and the pressure signal obtained from the clutch pressure sensor 23 . Item no. The “brake response feature quantity” of I04 is a feature quantity (vector) calculated based on the brake slackness signal and the pressure signal obtained from the brake slackness pressure sensor 22 . These feature quantities can be used to detect the pressure change response to the clutch/brake signal.

Item no. The "clutch secondary pressure" of I05 is a characteristic quantity calculated based on the pressure signal obtained from the clutch pressure sensor 23, and includes at least one of wave height vector and frequency vector, for example. The item I06 "brake slack pressure" is a characteristic quantity calculated based on the pressure signal obtained from the brake slack pressure sensor 22, and is similar to the clutch secondary side pressure, for example, at least one of a wave height vector and a frequency vector. include. It should be noted that these pressure feature amounts should include at least one feature amount out of three types of time vector, wave height vector, and frequency vector.

Item no. The "actual main motor speed" of 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, a wave height vector. Item no. I08 and I09 are the velocity and acceleration of the slide 1, respectively, and are calculated based on the rotation angle signal from the angle sensor 21. FIG. Velocity and acceleration of the slide 1 are represented, for example, by wave height vectors.

Item no. I10 "load peak value" and item No. The “load increase slope” of I11 is a load feature quantity 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 a linear expression passing through two points, the rising position and the peak position of the forming load curve.

Item no. The “die height value” of I12 is a feature amount calculated from the output value (die height data) from the die height displacement sensor 33 . Item no. The "mold contact time" of I13 is a feature quantity that can be calculated based on the load data obtained from the load sensor 20, and the time between the rising position and the peak position of the forming load curve is the mold contact time. calculated as The mold contact time may be calculated not for each of the molds 51 to 53 but as one feature quantity.

Item no. The “mold temperature” of I14 to I16 is a feature quantity calculated based on the temperature data obtained from the mold temperature sensor 34 attached to each mold 50. FIG. The mold temperature sensor 34 is provided on each of the molds 51 to 53 (for example, lower molds 51b, 52b, 53b) in steps I to III, and detects the temperature of each mold 50. FIG. The feature quantity of the mold temperature in each process is calculated as, for example, the average temperature of the specified area immediately before pressing (while the work is being conveyed). In addition to/instead of the average temperature, the maximum temperature, the minimum temperature, and the like may be adopted as the feature amount of the mold temperature.

Item no. The “mold lubricating fluid amount” of I17 to I19 is a characteristic amount calculated based on the lubricating fluid amount data obtained from the lubricating fluid flow rate sensor 30 provided in the lubricating fluid passage to each mold 50 . The lubricating fluid flow rate sensor 30 is provided for each of the molds 51 to 53 in processes I to III, and the flow rate of the lubricating fluid is detected for each mold 50 . The feature amount of the lubricating fluid amount of each die 50 is, for example, the lubricating fluid amount (total amount) for each die 50 immediately before pressing (while the work is being conveyed).

Item no. “Frame upstream vibration” of I20 to I22 is a feature quantity calculated based on the signal from the vibrometer 35 provided on the left column 3c of the frame 3. FIG. Item no. “Frame downstream side vibration” of I23 to I25 is a feature amount calculated based on the signal from the vibration meter 35 provided on the support 3d on the right side of the frame 3. FIG. The vibration feature amount of the frame 3 is calculated in each of the three sections before stroke, before contact with the mold, and after contact with the mold. A vibration feature quantity includes at least one of a wave height vector, a time vector, and a frequency vector. As shown in Table 1, the vibration feature amount of frame 3 preferably includes a wave height vector and a frequency vector, but may include a time vector and a frequency vector. Details of a specific method for calculating the vibration feature amount of the frame 3 will be described later.

Item no. “Motor main body side vibration” of I26 to I28 is a feature amount calculated based on the signal from the vibration meter 37 provided in the main motor 6 main body. Item no. "Motor load side vibration" of I29 to I31 is a feature amount calculated based on a signal from a vibrometer 37 attached to the main motor 6 and detecting bearing vibration. The vibration feature amount of the main motor 6 is calculated in each of the three analysis intervals before stroke, before mold contact, and after mold contact. A vibration feature quantity includes at least one of a wave height vector, a time vector, and a frequency vector. It is desirable that the vibration feature quantity of the main motor 6 also include a wave height vector and a frequency vector.

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

The learning unit 104 learns a large number of cycle feature data DT2 stored in the secondary data storage unit 112 for each detection item classification (hereinafter referred to as "sensor group"). As a result, an abnormality prediction model for predicting whether or not the feature amount of each sensor group is abnormal is generated for each press cycle. As a machine learning algorithm for the anomaly prediction model, a known algorithm such as isolation forest is used. Isolation forest is a general-purpose machine learning algorithm used for anomaly detection, and is a type of algorithm that does not require teacher data.

"Classification No." in Table 1 above is numbered for each sensor group. As shown in Table 1, except for some detection items, there is a one-to-one relationship between detection items and sensor groups. As an exception, a common sensor group (group No. G03) is assigned to the clutch response feature amount and the brake response feature amount (item Nos. I03 to I04). A common sensor group (group No. G09) is assigned to load peak value, load increase slope, die height value, and mold contact time (item Nos. I10 to I13) as load-related feature quantities. Note that the method of assigning the sensor groups shown in Table 1 is an example, and for example, the die height value and the mold contact time may be assigned to individual sensor groups.

In this embodiment, the learning unit 104 learns the feature amount of each sensor group for each “mold pattern”. That is, the learned model groups M1, M2, . . . Each learned model group includes the same number of abnormality prediction models m1, m2, . . . as sensor groups.

A plurality of learned model groups M1, M2, . . . That is, the model storage unit 114 stores the abnormality prediction models m1, m2, . Each abnormality prediction model is stored in association with identification information (classification No.) that identifies the corresponding sensor group.

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

(About operation)
FIG. 6 is a flowchart showing a method of generating an anomaly prediction model by learning device 100 according to the present embodiment. The model generation process shown in FIG. 6 is implemented by the processor of learning device 100 executing a learning program stored in advance in the memory. Note that this process is started in response to a user inputting an instruction to start learning via an operation unit (not shown). In the present embodiment, a plurality of cycle measurement data DT1 are stored in the primary data storage unit 111 before this process is started. and mold pattern data are stored in chronological order.

Referring to FIG. 6, learning device 100 identifies one mold pattern (in a predetermined order) from a combination of the product number of workpiece W and the workpiece pattern (step S2), and in primary data storage unit 111, Cycle measurement data DT1 associated with the mold pattern is retrieved (step S4).

After that, the format conversion unit 102 reads all of the searched (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 characteristic amounts of the detection items shown in Table 1 based on each data included in the cycle measurement data DT1, and records them in the secondary data storage unit 112 as the cycle characteristic data DT2. . Here, as an example, a calculation example of the vibration feature amount (item Nos. I20 to I25) of frame 3 will be described with reference to FIGS. 7 and 8. FIG.

FIG. 7A is a graph showing strokes of slide 1 along the time axis. 7B and 7C are graphs showing changes in the magnitude of vibration (pulse) along the same time axis as in FIG. 7A. ² ”. FIG. 8(B) is a graph showing the stroke of slide 1, like FIG. 7(A). FIG. 8A is a graph showing the brake slack signal along the same time axis. FIG. 8(C) is a graph showing changes in the magnitude of vibration (pulse), like FIGS. 7(B) and (C). FIG. 8(D) is a graph showing the transition of the power spectral density with the frequency as the horizontal axis (unit: Hz). Here, it is assumed that the vibration detected by the vibrometer 35 attached to the support 3c on the upstream side (rear left side) of the frame 3, for example, is shown.

A method of calculating the wave height vector will be described with reference to FIG. 7(B). The format conversion unit 102 extracts a pulse in a section TS exceeding the baseline (predetermined value) BL, and divides the extracted section TS into d strips along the time axis. Then, the pulse data is processed into a _{d -} dimensional vector (p ₌ [h1,h2,..., _hj ,...,hd]) to calculate the wave height vector. However, “h _j ” is the average value within each strip.

A method of calculating the time vector will be described with reference to FIG. The format conversion unit 102 divides the pulse in the section TS into _dw pieces in the wave height direction, and obtains _2dw pieces of time that intersect the division line (horizontal line). Then, the time data is processed into a 2d _w -dimensional vector (q=[w ₁ , w ₂ , . . . , w _j , . . . , w _2d ]) to calculate the time vector. However, for 'w _j ', select the two outer intersections with respect to the pulse apex position. For example, looking at the dividing line Lx, the outer intersection times t1 and t4 are selected and the times t2 and t3 are not selected. Time t1 corresponds to the time when the upper molds of the molds 51 to 53 start to come into contact with the work W, and time t4 corresponds to the time when the upper molds of the molds 51 to 53 start to separate from the work W.

A method of calculating the frequency vector will be described with reference to FIG. The format conversion unit 102 cuts out three types of analysis intervals, that is, a “pre-stroke interval,” a “pre-mold contact interval,” and a “during-mold contact interval,” from the vibration waveform shown in FIG. 8(C). The "pre-stroke section" is a section before time t11 when the brake loosening signal shown in FIG. 8A is turned ON. The “pre-mold contact section” is a section from time t11 to time t12 when the stroke position shown in FIG. 8B becomes the mold contact position. The "during mold contact section" is a section from time t12 to time t13 when the brake loosening signal shown in FIG. 8A is turned OFF.

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

Format conversion section 102 sets, for example, lower limit frequency f _min =0 and upper limit frequency f _max =fs/2, divides this section FS into d pieces (for example, 20 pieces) in a strip shape, and converts the power spectrum in section FS to Let the density PSD be a _d -dimensional vector (p ₌ [h ₁ , h ₂ , . . . , h _j , . value). Thereby, a frequency vector for each analysis interval is obtained.

Note that the analysis interval used for calculating the frequency vector may be used for calculating the above-described wave height vector. That is, as schematically shown in FIG. 8C, each analysis section is divided into d strips (for example, 20 pieces), and the pulse data is processed into a d-dimensional vector for each analysis section, and the wave height A vector may be calculated.

For example, the wave height vector and the frequency vector of the analysis section of the main motor actual speed can be calculated by the same calculation method as the calculation method of the vibration feature amount of the frame 3 as described above.

The cycle feature data DT2 including the feature amount for each detection item calculated by the above method has the same cycle number as the target cycle measurement data DT1. is recorded in the secondary data storage unit 112 in association with . As shown in FIG. 5, the cycle feature data DT2 may also be stored in association with the target mold pattern data.

The learning unit 104 machine-learns the cycle feature data DT2 obtained in step S8 in units of preset sensor groups (step S12). That is, the learning unit 104 generates an abnormality prediction model for each sensor group. After generating the abnormality prediction model, the learning unit 104 associates the mold pattern (product number, work pattern) identified in step S2 with the mold identification information and stores it in the model storage unit 114 (step S14).

If there is an unprocessed die pattern among the combinations of the product number of the work W and the work pattern (YES in step S16), the process returns to step S2 and repeats the above process. When the processing for all mold patterns is completed (NO in step S16), the series of model generation processing ends. As a result, the model storage unit 114 stores learned model groups M1, M2, . In the example of FIG. 5, the learned model group M1 includes an abnormality prediction model with product number "30" and work pattern No. "15", and the learned model group M2 includes product number "80" and work pattern No. "15." ” anomaly prediction model.

<Abnormality prediction device>
(About functional configuration)
FIG. 9 is a block diagram showing the functional configuration of 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 .

The model storage unit 213 preliminarily stores learned model groups M1, M2, . . . generated by the learning device 100 for each mold pattern. Learning device 100 and anomaly prediction device 200 are connected, for example, via a network. , and store the received learned model groups M1, M2, . . . in the model storage unit 213 . Alternatively, the group of learned models M1, M2, .

The combination information storage unit 214 stores combination information that associates each target part with at least one sensor group. In this embodiment, combination information is stored as a correspondence table TB. A data structure example of the correspondence table TB will be described later.

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

The format conversion unit 203 functions as feature amount calculation means, like 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 are calculated, and cycle feature data DT12 including these feature quantities are generated.

The prediction unit 204 detects the presence or absence of abnormality in each feature quantity using abnormality prediction models m1, m2, . Determine presence/absence. Specifically, as shown in FIG. 9, the prediction unit 204 includes a model determination unit 205, an abnormality detection unit 206, and an abnormality determination unit 207 as its functions.

Based on the acquired mold pattern data, the model determination unit 205 selects the mold pattern corresponding to the current cycle among the plurality of trained model groups M1, M2, . . . stored in the model storage unit 213. By specifying the trained model group, the abnormality prediction models m1, m2, . . . to be used are determined.

For each sensor group, the abnormality detection unit 206 converts the feature amount calculated by the format conversion unit 203 into a corresponding abnormality (among the plurality of abnormality prediction models m1, m2, . . . Input to the prediction model and detect whether the feature value is abnormal or not.

The abnormality determination unit 207 determines whether or not there is an abnormality for each target part based on the result of detection by the abnormality detection unit 206 and the correspondence table TB.

FIG. 10 shows an example data structure of the correspondence table TB. FIG. 10 exemplifies 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 sensor group column, the classification No. of Table 1 is displayed. The corresponding sensor group names (classification names of detection items) are described one by one. In the additional logic column, if there is a determination item associated with each sensor group, the name of the determination item is described one by one. Additional judgment item names are described on lines different from the lines describing the sensor group names. In the correspondence table TB, at least one mark (in FIG. 10, ◯ mark) 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 abnormality determination is given. are stored in association with each other.

In the present embodiment, the target parts are the material heater 14, the C/B control device (not shown), the clutch 9, the air pressure circuit of the clutch 9 (not shown), the brake device 12, and the air pressure of the brake device 12. It includes a circuit (not shown), the main motor 6, bearings (not shown), the material that is the workpiece W, and the mold 50 (51-53). Note that the target parts may not include all of these, and may include other components. Desirably, the target parts include at least one of material heaters, clutches, brakes, main motors, bearings, and dies, and materials that are workpieces.

To explain an example of the additional logic, in the present embodiment, as an additional determination item for each classification of "material temperature", "load-related feature amount", "frame vibration", and each classification of "motor vibration", continuous The presence or absence of any abnormalities (whether abnormal values occurred a predetermined number of times in a row) is included. Also, the "clutch reaction time" and the "brake control angle" may be newly calculated, and whether or not each exceeds the threshold value may be used as an additional determination item.

In the example of FIG. 10, for "material heater", (only) the row of the material temperature (classification No. G01) and its additional logic (logic No. L01) is marked, and the material temperature and its additional logic are stored in association with the "material heater". In this case, the abnormality determination unit 207 determines whether or not there is an abnormality in the "material heater" using the abnormality detection result of the material temperature and the abnormality detection result of the additional logic.

For "main motor", marks are added to the lines for main motor actual speed, slide speed, motor body side vibration (before stroke, before contact with mold), and motor load side vibration (before stroke, before contact with mold). there is As for "Bearing", same as "Main motor", the row of in-motor actual speed, slide speed, motor body side vibration (before stroke, before mold contact), motor load side vibration (before stroke, before mold contact) , as well as upstream frame vibration (pre-stroke, pre-mold contact), downstream frame vibration (pre-stroke, pre-mold contact), and upstream frame vibration or downstream frame vibration lines of additional logic are marked.

For the "mold" of each process, the row of the load-related feature quantity, the row of the mold temperature and lubricant amount of the corresponding process, and the frame upstream vibration and/or frame downstream vibration (after mold contact) lines and are marked. Thereby, it is possible to individually determine whether or not there is an abnormality in the molds 51 to 53 . Note that, as shown in FIG. 10, additional logic of the load-related feature amount may be used for the abnormality determination of the dies 51 to 53 . Further, for example, if the amount of lubricating fluid in the molds 51 to 53 can be measured for each of the upper mold and the lower mold, by using these characteristic amounts, it is possible to determine whether there is an abnormality in each of the upper mold and the lower mold. You may do so.

In this manner, by changing the type and number of sensor groups to be associated with each target part, it is possible to automatically determine (predict) which part is abnormal. Note that the combination of the target part and at least one sensor group as shown in FIG. 10 is typically set in advance by the administrator.

It is desirable that the correspondence table TB can be updated by an administrator. The correspondence table TB may be updated via an input unit (operation unit) included in the abnormality prediction device 20, or may be updated by a management device (not shown). As a result, it is possible to increase or decrease the number of parts targeted for abnormality prediction, and to add or change the types of sensor groups or additional logic associated with the target parts.

Referring to FIG. 9 again, output unit 230 outputs the result of prediction by prediction unit 204 , that is, the result of determination by abnormality determination unit 207 . The output unit 230 outputs whether or not there is an abnormality in each target part for each press cycle. As a result, the operator can grasp the presence or absence of an abnormality in the target part in real time. The output unit 230 may also output the degree of influence of the sensor groups listed in Table 1 on the prediction result and/or the reliability of the abnormality prediction result for each target part.

The functions of the data acquisition unit 202, the format conversion unit 203, and the prediction unit 204 described above are realized by the processor executing software. The model storage unit 213 and combination information storage unit 214 are typically configured by a non-volatile storage device included in the computer. The output unit 230 is typically configured by a display unit that displays prediction results. In addition, the output part 230 may be comprised by the communication part etc. which output a prediction result to monitoring PC42 shown in FIG.

(About operation)
FIG. 11 is a flow chart showing an abnormality prediction method for 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 the processor of the abnormality prediction device 200 executing an abnormality prediction program stored in advance in the memory.

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

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 amount for each detection item shown in Table 1 above based on each data included in the cycle measurement data DT11.

Also, the model determination unit 205 of the prediction unit 204 determines the plurality of learned model groups M1, M2, . . . Among them, a group of learned models corresponding to the mold pattern of this time is specified, and a plurality of abnormality prediction models m1, m2, . . . The abnormality prediction model discrimination process may be performed in parallel with the data format conversion process.

The anomaly detection unit 206 of the prediction unit 204 extracts, for each sensor group, multiple types of feature amounts (for one cycle) included in the cycle feature data DT12 obtained in step S28 into the anomaly prediction model determined in step S29. (step S30). As shown in the image diagram of FIG. 12, each abnormality prediction model outputs either "abnormal" or "abnormal". As a result, it is possible to detect (predict) whether or not the feature amount of each sensor group is abnormal for each press cycle. Each anomaly prediction model may output its reliability along with the detection result of the presence or absence of an anomaly. The reliability is calculated by, for example, the difference between the input feature amount and the threshold.

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

After that, the abnormality determination unit 207 performs abnormality determination of the target part based on the correspondence table TB (step S34). Specifically, based on the output of each abnormality prediction model (with or without abnormality), the calculation result of the additional logic value (additional judgment result), and the corresponding table TB (combination information), for each target part Determine the presence or absence of anomalies. At this time, the abnormality determination unit 207 may weight the abnormality detection result for each sensor group in the abnormality detection unit 206 according to the reliability, and perform abnormality determination of the target part.

In step S34, the abnormality determination unit 207 preferably calculates 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.

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

Further, the output unit 230 outputs the degree of importance (degree of influence) of each sensor group on the abnormality determination result of the target part in a graph form, for example, and outputs the degree of reliability as a numerical value (%), for example (step S38).

Thus, in the present 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. This makes it possible to easily take measures for the result of prediction of abnormality.

The processes of steps S26 to S38 are repeatedly executed until the press operation ends (NO in step S40). Note that if there is a part determined to be abnormal in step S34, the operation of the forging press machine 10 may be automatically stopped. As a result, mass production of defective products by the press machine 11 can be prevented or suppressed.

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. By using TB, it is possible to determine the presence or absence of abnormality in each target part for each press cycle. Therefore, according to the present embodiment, abnormalities of a plurality of parts in the forging press machine 10 can be predicted automatically and in real time without relying on the experience of the operator.

Further, in the present embodiment, the abnormality prediction model is learned for each mold pattern and stored in the model storage unit 213 of the abnormality prediction device 200, so that the abnormality detection of the feature amount for each sensor group and each target part can improve the accuracy of anomaly prediction.

In addition, in the present embodiment, since the above-described additional logic is added to the abnormality prediction of the target part, the accuracy is improved compared to the form in which the abnormality prediction of the target part is performed only by the output of the abnormality prediction model. be able to.

(Modification)
In the present embodiment, the model storage unit 213 of the abnormality prediction device 200 stores in advance the abnormality prediction models m1, m2, ... (learned model groups M1, M2, ...) for each sensor group. However, the abnormality prediction models m1, m2, . . . Alternatively, the abnormality prediction device 200 may have the function of the learning device 100 .

The combination information storage unit 214 of the abnormality prediction device 200 also stores a table-format correspondence table TB as shown in FIG. 10 as combination information that associates each target part with at least one sensor group. However, it is not limited to such an example.

Further, in the present embodiment, the forging press apparatus 10 is provided with the abnormality prediction device 200, but the abnormality prediction device 200 may predict the presence or absence of an abnormality in the target part offline.

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 provided by being recorded on an optical medium such as a CD-ROM (Compact Disc-ROM) or a computer-readable non-transitory recording medium such as a memory card. The program can also be provided by downloading via a network.

The program according to the present invention may call necessary modules out of program modules provided as part of the operating system (OS) of the computer in a predetermined sequence at a predetermined timing to execute processing. . In that case, the program itself does not include the above module, and the process is executed in cooperation with the OS. Programs that do not include such modules may also be included in the programs according to the present invention.

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

It should be considered that the embodiments disclosed this time are illustrative in all respects and not restrictive. The scope of the present invention is indicated by the scope of the claims rather than the above description, and is intended to include all modifications within the meaning and range of equivalents of the scope of the claims.

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 determination 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 model, SYS Abnormality prediction system, W Work.

Claims (7)

An abnormality prediction device for predicting abnormality of a plurality of parts in a press machine,
model storage means for storing an abnormality prediction model generated for each sensor group by performing machine learning on feature amounts of a plurality of sensor groups obtained based on measurement data of sensors mounted on the press device;
Acquisition means for acquiring measurement data from the sensors during operation of the press device;
feature quantity calculation means 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 for inputting the feature amount calculated by the feature amount calculation means to the corresponding abnormality prediction model for each sensor group and detecting whether or not the feature amount is abnormal;
combination information storage means for storing combination information that associates each target part with at least one sensor group;
An abnormality prediction apparatus, comprising: 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. The model storage means stores at least the abnormality prediction model for each sensor group for each mold pattern specified by the product number of the material,
The acquisition means acquires mold pattern data together with the measurement data,
3. A model discriminating means for discriminating a failure prediction model to be used from among the plurality of failure prediction models stored in the model storage means based on the mold pattern data acquired by the acquisition means. 2. The abnormality prediction device according to 1. 3. The method according to claim 1 or 2, wherein the determining means determines the presence or absence of an abnormality for each target part by further using continuous additional determination results of the presence or absence of an abnormality for at least one of the plurality of sensor groups. An anomaly prediction device as described. 4. The method according to any one of claims 1 to 3, further comprising output means for outputting a determination result by said determination means, as well as the degree of influence of said sensor group on said determination result and/or the reliability of said determination result. Anomaly prediction device. The abnormality prediction device according to any one of claims 1 to 4, wherein the target parts include at least one of material heaters, clutches, brakes, main motors, bearings, and dies, and materials that are workpieces. The target part includes a plurality of molds,
6. The abnormality prediction device according to claim 1, wherein said plurality of sensor groups include a frame upstream side vibration feature amount and a frame downstream side vibration feature amount. An abnormality prediction program for predicting abnormality of a plurality of parts in a press machine,
model storage means for storing an abnormality prediction model generated for each sensor group by performing machine learning on feature quantities of a plurality of sensor groups obtained based on measurement data of sensors;
a step of acquiring measurement data from sensors mounted on the press device during operation of the press device;
a calculation step of calculating the feature amount of each of the plurality of sensor groups based on the acquired measurement data;
A detection step of inputting the feature amount calculated in the calculation step into a corresponding abnormality prediction model for each sensor group and detecting whether or not the feature amount is abnormal;
Abnormality prediction, causing 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 that associates each target part with at least one of the sensor groups. program.

Images