JP2018147443A - Malfunction prediction method, malfunction prediction device and malfunction prediction program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately predict a malfunction of machinery having a drive unit.SOLUTION: A malfunction predicting method according to an embodiment comprises a collection step, a creation step, an evaluation step, and a determination step. In the collection step, sensor data of a plurality of sensors provided in machinery having a drive unit are collected. In the creation step, on the basis of the sensor data, machine learning is carried out such that a time change of driving speed of the drive unit in a normal operation of the machinery and a time change of driving force of the drive unit corresponding to such a driving speed are included in a feature vector, thereby creating a correlation model between the sensors. In the evaluation step, a degree of deviation from a normal state of the machinery is evaluated on the basis of an output value of the correlation model obtained by inputting sensor data into the correlation model, the sensor data corresponding to the driving speed and the driving force for evaluation in a given evaluation period later than the normal operation. In the determination step, a prediction about a malfunction of the machinery is determined on the basis of the degree of deviation.SELECTED DRAWING: Figure 2

Description

  Embodiments disclosed herein relate to a failure prediction method, a failure prediction device, and a failure prediction program.

  Conventionally, a technique for predicting the occurrence of a failure by detecting a failure sign by monitoring the sensor value of a sensor provided in the machine facility is known (for example, see Patent Document 1).

  The technique disclosed in Patent Document 1 compares a sensor value of a sensor mounted on a vehicle with a normal threshold value, and determines whether or not an abnormality has occurred in the sensor based on the comparison result. Then, when it is determined that an abnormality has occurred in the sensor, an evaluation index for evaluating a sensor failure sign using the integrated value of the difference between the sensor value and the normal threshold, the abnormality occurrence duration, etc. And a failure sign of the sensor is detected using the evaluation index.

JP 2011-230634 A

  However, the above-described conventional technology has room for further improvement in accurately capturing a failure sign of a mechanical facility having a drive unit.

  Specifically, although the above-described prior art is suitable for detecting a clear abnormal state exceeding a threshold value, it appears in the behavior of the entire system due to, for example, aging of each component of the drive unit. It is unsuitable for detecting an unclear abnormal state that does not show a sensor value until the threshold value is exceeded. Therefore, it is insufficient to accurately capture a failure sign and prevent, for example, a sudden failure or abnormal stop of the drive unit.

  One aspect of the embodiments has been made in view of the above, and provides a failure prediction method, a failure prediction apparatus, and a failure prediction program capable of accurately capturing a failure sign of a mechanical facility having a drive unit. Objective.

  The failure prediction method according to an aspect of the embodiment includes a collection process, a generation process, an evaluation process, and a determination process. The collecting step collects sensor data of a plurality of sensors provided in a mechanical facility having a drive unit. Based on the sensor data, the generation step includes a time variation of the driving speed of the driving unit during normal operation of the mechanical equipment and a time variation of the driving force of the driving unit corresponding to the driving speed in a feature vector. By executing machine learning as described above, a correlation model between the sensors is generated. The evaluation step outputs the correlation model obtained by inputting the sensor data corresponding to the driving speed and the driving force for the evaluation at an arbitrary evaluation time after the normal operation to the correlation model. Based on the value, the degree of deviation from the normal state of the mechanical equipment is evaluated. In the determination step, a failure sign of the mechanical equipment is determined based on the degree of deviation.

  According to one aspect of the embodiment, it is possible to accurately capture a failure sign of mechanical equipment having a drive unit.

FIG. 1A is a schematic explanatory diagram (part 1) of the failure prediction method according to the embodiment. FIG. 1B is a schematic explanatory diagram (part 2) of the failure prediction method according to the embodiment. FIG. 1C is a schematic explanatory diagram (part 3) of the failure prediction method according to the embodiment. FIG. 2 is a block diagram of the failure prediction system according to the embodiment. FIG. 3A is a diagram illustrating a precondition of the first verification example. FIG. 3B is a diagram (part 1) illustrating a verification result of verification example 1. FIG. 3C is a diagram (part 2) illustrating a verification result of verification example 1. FIG. 3D is a diagram (part 3) illustrating a verification result of the verification example 1. FIG. 4A is a diagram illustrating a precondition of the second verification example. FIG. 4B is a diagram (part 1) illustrating a verification result of verification example 2. FIG. 4C is a diagram (part 2) illustrating a verification result of verification example 2. FIG. 4D is a diagram (part 3) illustrating a verification result of the verification example 2. FIG. 4E is a diagram (part 4) illustrating a verification result of the verification example 2. FIG. 4F is a diagram (part 5) illustrating a verification result of the verification example 2. FIG. 5 is a flowchart showing a processing procedure executed by the failure prediction apparatus. FIG. 6 is a hardware configuration diagram illustrating an example of a computer that realizes the function of the failure prediction apparatus.

  Hereinafter, embodiments of a failure prediction method, a failure prediction device, and a failure prediction program disclosed in the present application will be described in detail with reference to the accompanying drawings. In addition, this invention is not limited by embodiment shown below.

  In the following, the machine equipment that is the target of failure sign determination is referred to as “target machine 100”. The target machine 100 is assumed to be a machine having a drive unit that is driven by a motor.

  First, an outline of the failure prediction method according to the present embodiment will be described with reference to FIGS. 1A to 1C. 1A to 1C are schematic explanatory views (No. 1) to (No. 3) of a failure prediction method according to an embodiment.

  As illustrated in FIG. 1A, the target machine 100 includes a sensor group of sensors S-1 to Sn. The failure prediction method according to the present embodiment grasps the correlation between the sensors S-1 to Sn (hereinafter sometimes referred to as “sensor correlation”) based on the sensor data from the sensor group. Then, the change in the behavior of the entire target machine 100 is grasped based on the change in the sensor correlation.

  Specifically, as shown in FIG. 1B, in the failure prediction method according to the present embodiment, machine learning is performed using sensor data of the sensors S-1 to Sn indicating sensor correlation of “normal state”. (Step S1) to generate the sensor correlation model 12c. In the present embodiment, deep learning is used as an algorithm for machine learning. Since deep learning is publicly known, detailed description is omitted.

  Here, the “normal state portion” refers to a portion based on each sensor data during normal operation in the initial operation stage of the target machine 100 or the like. With the sensor correlation model 12c generated based on the sensor correlation for the normal state, the normal state of the target machine 100 can be modeled.

  In the failure prediction method according to the present embodiment, as shown in FIG. 1B, the sensor data of the sensors S-1 to S-n indicating the sensor correlation of “evaluation” is obtained for the sensor correlation model 12c. The correlation shift amount is calculated from the output value (regression value) of the sensor correlation model 12c obtained as a result of the input. Then, the degree of deviation from the normal state is evaluated based on the amount of deviation of the correlation (step S2).

  If the degree of divergence is large, a failure of the target machine 100 can be predicted by indicating a failure sign. Here, the “evaluation portion” refers to, for example, a portion based on real-time data output in a progressive form from the target machine 100.

  As described above, in the failure prediction method according to the present embodiment, the behavior of the entire target machine 100 is grasped by the sensor correlation, and the change in the behavior is output from the sensor correlation model 12c obtained by modeling the sensor correlation in the normal state. It can be obtained by value. A failure of the target machine 100 is predicted based on the degree of deviation from the normal state based on the output value.

  Therefore, according to the failure prediction method according to the present embodiment, for example, all failure phenomena that are necessary if the failure prediction method uses a failure model generated by machine learning based on sensor data at the time of failure occurrence. A complicated process such as implementation of modeling is not necessary. Therefore, a failure sign of the target machine 100 can be easily and accurately captured.

  In the failure prediction method according to the present embodiment, as shown in FIG. 1A, the time variation of the sensor data of each sensor S-1 to Sn, that is, the time-series correlation is represented by a feature vector (hereinafter simply referred to as “vector”). Machine learning included in the above), it is possible to grasp the sign of failure that appears in the time fluctuation of sensor data. Thereby, a failure sign of the target machine 100 can be accurately captured.

  Further, in the present embodiment, in particular, based on the failure prediction method described above, the failure sign for the drive unit of the target machine 100 is accurately captured. Specifically, as illustrated in FIG. 1C, in the failure prediction method according to the present embodiment, a sensor based on sensor data from sensors S-1 to Sn indicating the behavior of the drive unit 101 included in the target machine 100. A correlation model 12c is generated.

  The drive unit 101 includes at least a motor M. Here, the sensors S-1 to Sn include at least a motor speed sensor, a motor torque sensor, and a vibration sensor. Further, it does not matter whether the sensors S-1 to Sn are mounted on the drive unit 101 or other than the drive unit 101.

  More specifically, as shown in FIG. 1C, in the failure prediction method according to the present embodiment, the “motor speed”, “motor torque” of the motor M and the target machine are related to the machine learning in step S1 (see FIG. 1B). Machine learning including 100 “vibrations” in the vector is performed.

  First, in the failure prediction method according to the present embodiment, time series data at the time of maximum acceleration / deceleration for the motor speed of the motor M is used as reference data, and motor torque data and vibration data corresponding to the reference data are extracted. Predetermined preprocessing is performed on the motor torque data and vibration data.

  In the pre-processing (step S11) relating to the motor torque data, for example, the operating frequency including the noise component is cut by filter processing using a secondary high-pass filter or the like, and then the root mean square so as to indicate the magnitude of the ripple component Convert to Root Mean Square (RMS) value.

  In the preprocessing (step S12) relating to the vibration data, the vibration value is converted into an RMS value so as to indicate the magnitude of the vibration component. Note that a plurality of vibration sensors may be provided so as to be arranged at various locations on the target machine 100.

  In the failure prediction method according to the present embodiment, the sensor correlation model 12c is generated by deep learning using the motor speed data, motor torque data, and vibration data as a learning data set (step S13).

  As illustrated in the neural network of FIG. 1C, in the sensor correlation model 12c, for example, an “auto encoder” can be used as an analysis algorithm for learning only normality.

  After the generation of the sensor correlation model 12c, the real time data of the motor speed data, the motor torque data and the vibration data are input to the sensor correlation model 12c, and the error (reconstruction error) between the input value and the output value is calculated. A parameter indicating the degree of deviation from the state can be obtained. Then, by determining the degree of divergence based on a predetermined determination threshold value, a failure sign of the drive unit 101 can be determined.

  When a plurality of vibration sensors are provided, for example, the contribution rank of each vibration sensor is calculated based on the sensor data of each vibration sensor. This order of contribution corresponds to the location of each vibration sensor, which can contribute to the identification of the failure location.

  As described above, the sensor correlation model 12c is generated based on each sensor data related to the drive unit 101, and the real-time data of the drive unit 101 is evaluated by the sensor correlation model 12c. Predictive failure can be accurately captured.

  As a result, it is possible to grasp in advance the failure signs of the drive unit 101 and perform maintenance such as repair or replacement of parts, and the drive unit 101 may suddenly fail or stop abnormally. Can be prevented.

  Hereinafter, the configuration of the failure prediction system 1 to which the above-described failure prediction method is applied will be described more specifically.

  FIG. 2 is a block diagram of the failure prediction system 1 according to the present embodiment. In FIG. 2, constituent elements necessary for describing the features of the present embodiment are represented by functional blocks, and descriptions of general constituent elements are omitted.

  In other words, each component illustrated in FIG. 2 is functionally conceptual and does not necessarily need to be physically configured as illustrated. For example, the specific form of distribution / integration of each functional block is not limited to the one shown in the figure, and all or a part thereof is functionally or physically distributed in an arbitrary unit according to various loads or usage conditions.・ It can be integrated and configured.

  In the description with reference to FIG. 2, the description of the components already described so far may be simplified or omitted.

  As shown in FIG. 2, the failure prediction system 1 includes a failure prediction device 10 and a target machine 100. The failure prediction device 10 and the target machine 100 are provided so as to be connected to each other via a network, and the failure prediction device 10 is provided so that sensor data from the target machine 100 can be collected as appropriate. The sensor data includes motor speed data, motor torque data, and vibration data related to the drive unit 101.

  The failure prediction apparatus 10 includes a control unit 11 and a storage unit 12. The control unit 11 includes a collection unit 11a, an extraction unit 11b, a generation unit 11c, an evaluation unit 11d, a determination unit 11e, and a notification unit 11f.

  The storage unit 12 is a storage device such as a hard disk drive, a nonvolatile memory, or a register, and stores collected data 12a, a learning data set 12b, a sensor correlation model 12c, an evaluation data set 12d, and evaluation information 12e. . The evaluation information 12e includes a divergence degree 12ea and a contribution rank 12eb.

  The control unit 11 performs overall control of the failure prediction apparatus 10. The collection unit 11a collects sensor data from the sensor group of the target machine 100 at a predetermined period, and stores it in the collection data 12a. The predetermined period is, for example, 1 to 20 milliseconds.

  The extraction unit 11b extracts from the collected data 12a a data set for a normal state using the motor speed data, motor torque data, and vibration data during normal operation of the target machine 100 as a vector, and stores the data set in the learning data set 12b.

  Further, the extraction unit 11b extracts a data set using the motor speed data, motor torque data, and vibration data at the time of evaluation from the collected data 12a as a vector, and stores the data set in the evaluation data set 12d.

  In addition, the extraction part 11b performs the pre-processing of step S11 (refer FIG. 1C) mentioned above with respect to motor torque data. Moreover, the extraction part 11b performs the pre-processing of step S12 (refer FIG. 1C) mentioned above with respect to vibration data.

  The generation unit 11c performs machine learning by deep learning using the learning data set 12b as an input, and generates a sensor correlation model 12c.

  For each sensor S-1 to Sn, the evaluation unit 11d inputs the evaluation data set 12d extracted by the extraction unit 11b to the sensor correlation model 12c, and receives an output result from the sensor correlation model 12c.

  Then, the evaluation unit 11d calculates various evaluation values for determining a failure sign based on the received output result, and stores the calculated evaluation values in the evaluation information 12e. The evaluation value corresponds to the divergence degree 12ea and the contribution rank 12eb.

  Specifically, the evaluation unit 11d inputs the evaluation data set 12d to the sensor correlation model 12c, and calculates the deviation from the normal state from the output value of the sensor correlation model 12c. The evaluation unit 11d stores the calculated divergence degree in the divergence degree 12ea of the evaluation information 12e.

  Further, the evaluation unit 11d calculates the contribution rank of each vibration sensor based on the sensor data of each vibration sensor, and stores it in the contribution rank 12eb of the evaluation information 12e.

  The determination unit 11e refers to the divergence degree 12ea, determines that there is a failure sign when the divergence degree from the normal state is equal to or greater than a predetermined determination threshold value, and issues a notification request instruction to the notification unit 11f.

  When the notification unit 11f receives a notification request instruction with a sign of failure from the determination unit 11e, the notification unit 11f notifies the external device of an alert notification. Further, at this time, the notification unit 11f refers to the divergence degree 12ea and the contribution rank 12eb, and indicates the divergence degree from the normal state and, for example, the names and arrangement positions of the sensors S-1 to Sn that are higher in the contribution rank. It can also be notified.

  For example, the user can know the degree of abnormality of the drive unit 101 by confirming the “degree of deviation from the normal state”. Further, for example, by confirming “a sensor with a higher contribution ranking”, it is possible to estimate the location and cause of the abnormality.

  Next, verification example 1 of the present embodiment is shown. FIG. 3A is a diagram illustrating a precondition of the first verification example. 3B to 3D are diagrams (No. 1) to (No. 3) showing the verification results of the verification example 1. FIG.

  As shown in FIG. 3A, in the verification example 1, the driving unit 101 using the servo motor 101a as the motor M is configured, and the driving mechanism 101 is driven by the driving unit 101 to obtain the verification result.

  As shown in FIG. 3A, the servo motor 101a includes an output shaft 101b, and the distal end side of the output shaft 101b is rotatably supported by a bearing 101c. The output shaft 101b has a drive gear 101d, and the drive gear 101d meshes with the driven gear 102a of the driven mechanism 102.

  The driven mechanism 102 includes a driven shaft 102b, and the driven shaft 102b is rotatably supported at both ends by bearings 102c and 102d. In order to reproduce the abnormal time, it was possible to apply vibration to the driven shaft 102b by the vibration applying mechanism 150 that mechanically generates vibration. And the vibration sensor S was arrange | positioned so that the vibration of the bearing 101c could be measured.

  Verification results (No. 1) to (No. 3) verified in such a configuration example are shown in FIGS. 3B to 3D. 3B to 3D, the case where the servo motor 101a is rotated without applying vibration by the vibration applying mechanism 150 is “normal”.

  On the other hand, the case where the servo motor 101a is rotated while applying the vibration by the vibration applying mechanism 150 is defined as “abnormal time”. It is assumed that the sensor correlation model 12c has already been generated based on the time series data of the motor speed at the maximum acceleration / deceleration of the servo motor 101a during normal operation and the corresponding motor torque data and vibration data.

  First, as (No. 1) of the verification example 1, as shown in FIG. 3B, when the servo motor 101a is rotated so that the amplitude decreases stepwise from the maximum acceleration / deceleration speed (-3000 rpm to 3000 rpm). Verified.

  In such a case, the evaluation result of “normal” shows that the degree of deviation from the normal state does not always exceed the determination threshold Th and is not determined as having a failure sign.

  On the other hand, in the “abnormal” evaluation result, even when the deviation from the normal state appropriately exceeds the determination threshold Th and the amplitude of the motor speed is changed stepwise, the motor speed data, the motor Based on the torque data and vibration data, the verification result that failure prediction is possible was obtained. That is, according to the present embodiment, a failure sign of the target machine 100 having the drive unit 101 can be accurately captured.

  Further, as (No. 2) of Verification Example 1, as shown in FIG. 3C, the servo motor 101a is rotated so that the motor speed periodically repeats the amplitude at the maximum acceleration / deceleration (-3000 rpm to 3000 rpm). Verified.

  Even in such a case, the evaluation result of “normal” shows that the degree of deviation from the normal state does not always exceed the determination threshold Th and is not determined as having a failure sign.

  On the other hand, in the evaluation result of “abnormal”, even when the deviation from the normal state exceeds the repeated determination threshold Th and the motor speed is periodically changed at the maximum acceleration / deceleration, the motor speed data Based on the motor torque data and vibration data, a verification result that failure prediction is possible was obtained. That is, according to the present embodiment, a failure sign of the target machine 100 having the drive unit 101 can be accurately captured.

  Further, as (No. 3) of the verification example 1, as shown in FIG. 3D, the case where the servo motor 101a was rotated so as to periodically repeat the amplitude of −500 rpm to 500 rpm where the motor speed is relatively low was verified. .

  Even in such a case, the evaluation result of “normal” shows that the degree of deviation from the normal state does not always exceed the determination threshold Th and is not determined as having a failure sign.

  On the other hand, in the evaluation result of “abnormal”, the degree of deviation from the normal state repeatedly exceeds the determination threshold Th, and even when the motor speed is periodically changed with a relatively low amplitude, the motor Based on the speed data, motor torque data, and vibration data, a verification result that failure prediction is possible was obtained. That is, according to the present embodiment, a failure sign of the target machine 100 having the drive unit 101 can be accurately captured.

  Next, verification example 2 of the present embodiment will be described. FIG. 4A is a diagram illustrating a precondition of the second verification example. 4B to 4F are diagrams (No. 1) to (No. 5) showing the verification results of the verification example 2. FIG.

  As shown in FIG. 4A, in the second verification example, the motor M configures the driving unit 101 that is driven and controlled by the inverter 101e, and the driving unit 101 is driven while mechanically changing the conditions of other components of the driving unit 101. The verification result was obtained.

  As shown in FIG. 4A, the motor M includes an output shaft 101f, and the proximal end side and the distal end side of the output shaft 101f are rotatably supported by bearings 101g and 101h, respectively. In addition, the output shaft 101f can be mechanically changed between “no eccentricity” and “with eccentricity”. In addition, the bearing 101h can be mechanically changed to “normal” or “abnormal” of the bearing. And the vibration sensor S was arrange | positioned so that the vibration of the bearing 101h could be measured.

  The verification results (No. 1) to (No. 5) verified in such a configuration example are shown in FIGS. 4B to 4F. In FIG. 4B to FIG. 4F, the condition (1) is defined as the case where there is no abnormality in any of the above-described bearings and eccentricity. In addition, conditions (2) to (4) were defined when there was an abnormality in at least one of the bearing and the eccentricity. As in the first verification example, the sensor correlation model 12c has already been generated based on the time-series data of the motor speed at the maximum acceleration / deceleration of the motor M during normal operation and the corresponding motor torque data and vibration data. Suppose that

  First, as (part 1) of the verification example 2, as shown in FIG. 4B, the case where the motor M was driven and controlled by the motor speed command of “speed 0 → 100%” from the inverter 101e was verified.

  In such a case, in the evaluation result of the condition (1), the verification result that the deviation from the normal state does not always exceed the determination threshold Th and is not determined as having a failure sign is obtained.

  On the other hand, in the evaluation results of the conditions (2) to (4), the deviation from the normal state changes so as to exceed the determination threshold Th as the motor M rotates at high speed, and at least the motor M is at high speed. Based on the motor speed data, motor torque data, and vibration data, the verification result that failure prediction is possible is obtained. That is, according to the present embodiment, a failure sign of the target machine 100 having the drive unit 101 can be accurately captured.

  Further, as (part 2) of the verification example 2, as shown in FIG. 4C, the case where the motor M was driven and controlled by the motor speed command of “speed 100 → 0%” from the inverter 101e was verified.

  Even in such a case, in the evaluation result of the condition (1), the verification result that the degree of deviation from the normal state does not always exceed the determination threshold Th and is not determined as having a failure sign is obtained.

  On the other hand, in the evaluation results of the conditions (2) to (4), although the degree of deviation from the normal state decreases as the motor M rotates at a low speed, at least if the motor M rotates at a relatively high speed, Based on the speed data, motor torque data, and vibration data, a verification result that failure prediction is possible was obtained. That is, according to the present embodiment, a failure sign of the target machine 100 having the drive unit 101 can be accurately captured.

  Further, as verification example 2 (part 3), as shown in FIG. 4D, a case where the motor M was driven and controlled by a motor speed command of “speed 0 → 80%” from the inverter 101e was verified.

  Even in such a case, in the evaluation result of the condition (1), the verification result that the degree of deviation from the normal state does not always exceed the determination threshold Th and is not determined as having a failure sign is obtained.

  On the other hand, in the evaluation result of the condition (2) or (4), as the motor M rotates at high speed, the deviation from the normal state changes so as to exceed the determination threshold Th, and at least the motor M is at high speed. Based on the motor speed data, motor torque data, and vibration data, the verification result that failure prediction is possible is obtained. That is, according to the present embodiment, a failure sign of the target machine 100 having the drive unit 101 can be accurately captured.

  In the evaluation result of condition (3), the degree of deviation from the normal state does not exceed the determination threshold Th, but at least only the bearing 101h is abnormal in the configuration example of the drive unit 101 of this verification example 2. It is understood that failure prediction may be difficult unless the motor M rotates at a high speed exceeding 80%.

  Further, as (No. 4) of Verification Example 2, as shown in FIG. 4E, the case where the motor M was driven and controlled by the motor speed command of “speed 80% → 30%” from the inverter 101e was verified.

  Even in such a case, in the evaluation result of the condition (1), the verification result that the degree of deviation from the normal state does not always exceed the determination threshold Th and is not determined as having a failure sign is obtained.

  On the other hand, in the evaluation result of the condition (2) or (4), although the degree of deviation from the normal state decreases as the motor M rotates at a low speed, at least if the motor M rotates at a relatively high speed, the motor Based on the speed data, motor torque data, and vibration data, a verification result that failure prediction is possible was obtained. That is, according to the present embodiment, a failure sign of the target machine 100 having the drive unit 101 can be accurately captured.

  In the evaluation result of condition (3), it can be seen that failure prediction may be difficult unless the motor M exceeds 80% and rotates at a high speed, as in the case of verification example 2 (part 3).

  Further, as verification example 2 (part 5), as shown in FIG. 4F, the case where the motor M was driven and controlled by the motor speed command of “speed 100%” from the inverter 101e was verified.

  Even in such a case, in the evaluation result of the condition (1), the verification result that the degree of deviation from the normal state does not always exceed the determination threshold Th and is not determined as having a failure sign is obtained.

  On the other hand, in the evaluation results of the conditions (2) to (4), the divergence in the normal state generally exceeds the determination threshold Th, although it is relatively lower in the case of the condition (3). Based on the data and vibration data, the verification result that failure prediction is possible was obtained. That is, according to the present embodiment, a failure sign of the target machine 100 having the drive unit 101 can be accurately captured.

  Next, a processing procedure executed by the failure prediction apparatus 10 will be described with reference to FIG. FIG. 5 is a flowchart showing a processing procedure executed by the failure prediction apparatus 10.

  As shown in FIG. 5, first, the control unit 11 determines whether or not it is a learning target period (step S101). The learning target period is a period during which the target machine 100 is operating normally, such as at the initial operation stage.

  Here, when it is a learning target period (step S101, Yes), the extraction part 11b extracts the time series data at the time of the maximum acceleration / deceleration for the motor speed (step S102). Further, the extraction unit 11b extracts the motor torque data corresponding to the motor speed data while performing preprocessing (step S103).

  The extraction unit 11b also extracts vibration data corresponding to the motor speed data while pre-processing (step S104).

  And the production | generation part 11c produces | generates the sensor correlation model 12c by the deep learning which used each data extracted by the extraction part 11b as a learning data set (step S105). In addition, when it is not a learning object period (step S101, No), control is transferred to step S106.

  Subsequently, the extraction unit 11b extracts data at the time of evaluation of motor speed, motor torque, and vibration, that is, real-time data (step S106).

  Then, the evaluation unit 11d derives the degree of divergence from the normal state from the output value of the sensor correlation model 12c that receives each data extracted by the extraction unit 11b (step S107). Further, the evaluation unit 11d derives a contribution rank based on the location of the vibration sensor S (step S108).

  Subsequently, the determination unit 11e determines whether or not the deviation degree derived by the evaluation unit 11d is greater than or equal to a predetermined determination threshold value Th (step S109). Here, when the divergence degree is equal to or greater than the determination threshold Th (step S109, Yes), the notification unit 11f notifies the divergence degree from the normal state and the contribution rank (step S110), and the process is terminated.

  If the determination condition in step S109 is not satisfied (No in step S109), the process is terminated. In this case, the notification unit 11f may notify that there is no failure sign.

  Note that the failure prediction apparatus 10 according to the embodiment is realized by a computer 60 having a configuration as shown in FIG. 6, for example. FIG. 6 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the failure prediction apparatus 10. The computer 60 includes a central processing unit (CPU) 61, a random access memory (RAM) 62, a read only memory (ROM) 63, a hard disk drive (HDD) 64, a communication interface (I / F) 65, an input / output interface (I). / F) 66 and a media interface (I / F) 67.

  The CPU 61 operates based on a program stored in the ROM 63 or the HDD 64 and controls each unit. The ROM 63 stores a boot program executed by the CPU 61 when the computer 60 is activated, a program depending on the hardware of the computer 60, and the like.

  The HDD 64 stores a program executed by the CPU 61, data used by the program, and the like. The communication interface 65 corresponds to a communication unit (not shown) with the target machine 100, receives data from other devices via the communication network, sends the data to the CPU 61, and transmits the data generated by the CPU 61 via the communication network. Send to other devices.

  The CPU 61 controls an output device such as a display and a printer and an input device such as a keyboard and a mouse via the input / output interface 66. The CPU 61 acquires data from the input device via the input / output interface 66. Further, the CPU 61 outputs the generated data to the output device via the input / output interface 66.

  The media interface 67 reads a program or data stored in the recording medium 68 and provides it to the CPU 61 via the RAM 62. The CPU 61 loads the program from the recording medium 68 onto the RAM 62 via the media interface 67 and executes the loaded program. The recording medium 68 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. Etc.

  When the computer 60 functions as the failure prediction device 10, the CPU 61 of the computer 60 executes a program loaded on the RAM 62, thereby collecting the collection unit 11a, the extraction unit 11b, the generation unit 11c, the evaluation unit 11d, and the determination unit 11e. And each function of the alerting | reporting part 11f is implement | achieved. The HDD 64 realizes the function of the storage unit 12 and stores collected data 12a and the like.

  The CPU 61 of the computer 60 reads these programs from the recording medium 68 and executes them, but as another example, these programs may be acquired from other devices via a communication network.

  As described above, the failure prediction apparatus 10 according to the embodiment includes the collection unit 11a, the generation unit 11c, the evaluation unit 11d, and the determination unit 11e.

  The collection unit 11a collects sensor data of the plurality of sensors S-1 to Sn provided in the target machine 100 having the drive unit 101 (corresponding to an example of “mechanical equipment”).

  Based on the sensor data, the generation unit 11c changes the motor speed of the drive unit 101 (corresponding to an example of “drive speed”) during normal operation of the target machine 100 and the motor torque of the drive unit 101 corresponding to the motor speed. By executing machine learning so that the temporal variation of (corresponding to an example of “driving force”) is included in the feature vector, a sensor correlation model 12c (corresponding to an example of “correlation model between sensors”) is generated.

  The evaluation unit 11d outputs an output value of the sensor correlation model 12c obtained by inputting sensor data corresponding to the motor speed and motor torque corresponding to the evaluation at an arbitrary evaluation time after the normal operation to the sensor correlation model 12c. Based on this, the degree of deviation from the normal state of the target machine 100 is evaluated. The determination unit 11e determines a failure sign of the target machine 100 based on the degree of deviation.

  Therefore, according to the failure prediction apparatus 10 according to the present embodiment, a failure sign of the target machine 100 can be accurately captured.

(Other embodiments)
By the way, in the embodiment described above, machine learning including motor speed, motor torque, and vibration in a vector is executed, and the deviation degree from the normal state of the behavior of the drive unit 101 is determined by the sensor correlation model 12c generated as a result. Although the example to evaluate was given, secondary handling may be sufficient about a vibration. In other words, machine learning using at least the motor speed and the motor torque as vectors can contribute to accurately grasping a failure sign of the drive unit 101.

  In addition to the above-described embodiment, the machine learning vector may further include other parameters that are considered to contribute to the evaluation of the degree of deviation from the normal state, such as the temperature and the position of the target machine 100. . Thereby, it can contribute more to grasping the failure sign of the target machine 100 having the drive unit 101 with high accuracy.

  In the above-described embodiment, deep learning is used as a machine learning algorithm, but the algorithm to be used is not limited. Accordingly, the sensor correlation model 12c may be generated by performing machine learning by a regression analysis method such as support vector regression using a pattern classifier such as SVM (Support Vector Machine). Here, the pattern discriminator is not limited to SVM, and may be, for example, AdaBoost. A random forest or the like may be used.

  In the above-described embodiment, the drive source of the drive unit 101 is basically a rotary motor, but may be a linear motor. In such a case, the driving force of the driving unit 101 corresponds to thrust rather than torque.

  Further effects and modifications can be easily derived by those skilled in the art. Thus, the broader aspects of the present invention are not limited to the specific details and representative embodiments shown and described above. Accordingly, various modifications can be made without departing from the spirit or scope of the general inventive concept as defined by the appended claims and their equivalents.

DESCRIPTION OF SYMBOLS 1 Failure prediction system 10 Failure prediction apparatus 11a Collection part 11b Extraction part 11c Generation part 11d Evaluation part 11e Determination part 11f Notification part 12a Collected data 12b Learning data set 12c Sensor correlation model 12d Evaluation data set 12e Evaluation information 12ea Deviation degree 12eb Contribution rank DESCRIPTION OF SYMBOLS 100 Target machine 101 Drive part 101a Servo motor 101e Inverter M Motor S-1 to Sn Sensor

Claims (9)

A collecting step of collecting sensor data of a plurality of sensors provided in a mechanical facility having a drive unit;
Based on the sensor data, machine learning is performed so that the time variation of the driving speed of the driving unit during normal operation of the mechanical equipment and the time variation of the driving force of the driving unit corresponding to the driving speed are included in the feature vector. A generating step of generating a correlation model between the sensors by executing;
Based on the output value of the correlation model obtained by inputting the sensor data corresponding to the driving speed and the driving force of the evaluation at an arbitrary evaluation time after the normal operation to the correlation model An evaluation process for evaluating the degree of deviation from the normal state of the mechanical equipment;
And a determination step of determining a failure sign of the mechanical equipment based on the degree of divergence. The generating step includes
The failure prediction method according to claim 1, wherein the machine learning is executed based on the sensor data so that a time variation of vibration of the mechanical equipment is included in a feature vector. The generating step includes
The machine learning is performed based on the sensor data corresponding to the driving force, which is filtered to remove the operating frequency and is converted to a root mean square value so as to indicate the magnitude of the ripple component. The failure prediction method according to claim 2, wherein: The generating step includes
The machine learning is performed based on the sensor data corresponding to the vibration converted to a root mean square value so as to indicate a magnitude of a ripple component. Failure prediction method. The generating step generates the correlation model using an auto encoder,
The evaluation step includes
An error between the input value and the output value of the correlation model is calculated as the degree of deviation from the normal state of the mechanical equipment,
The determination step includes
The failure prediction method according to claim 2, 3 or 4, wherein, when the degree of deviation is equal to or greater than a predetermined determination threshold value, it is determined that the mechanical equipment has a failure sign. The evaluation step includes
The contribution ranking of each of the sensors is calculated based on the sensor data of each of the sensors when a plurality of the sensors for measuring the vibration are provided. The failure prediction method described in 1. A notification step of performing an alert notification to an external device when it is determined by the determination step that there is a failure sign,
The determination step includes
The information on the sensor corresponding to the contribution rank calculated by the evaluation step is included in the alert notification of the notification step when it is determined that the failure sign is present. Failure prediction method. A collection unit that collects sensor data of a plurality of sensors provided in a mechanical facility having a drive unit;
Based on the sensor data, machine learning is performed so that the time variation of the driving speed of the driving unit during normal operation of the mechanical equipment and the time variation of the driving force of the driving unit corresponding to the driving speed are included in the feature vector. A generating unit that generates a correlation model between the sensors by executing,
Based on the output value of the correlation model obtained by inputting the sensor data corresponding to the driving speed and the driving force of the evaluation at an arbitrary evaluation time after the normal operation to the correlation model An evaluation unit that evaluates the degree of deviation from the normal state of the mechanical equipment;
A failure prediction apparatus comprising: a determination unit that determines a failure sign of the mechanical equipment based on the degree of deviation. A collection procedure for collecting sensor data of a plurality of sensors provided in mechanical equipment having a drive unit;
Based on the sensor data, machine learning is performed so that the time variation of the driving speed of the driving unit during normal operation of the mechanical equipment and the time variation of the driving force of the driving unit corresponding to the driving speed are included in the feature vector. A generation procedure for generating a correlation model between the sensors by executing;
Based on the output value of the correlation model obtained by inputting the sensor data corresponding to the driving speed and the driving force of the evaluation at an arbitrary evaluation time after the normal operation to the correlation model An evaluation procedure for evaluating the degree of deviation from the normal state of the mechanical equipment;
A failure prediction program that causes a computer to execute a determination procedure for determining a failure sign of the mechanical equipment based on the degree of deviation.