JP7082461B2 - Failure prediction method, failure prediction device and failure prediction program - Google Patents

Description

The disclosed embodiments relate to failure prediction methods, failure prediction devices and failure prediction programs.

Conventionally, there has been known 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 / equipment (see, for example, Patent Document 1).

In addition, conventionally, for machinery and equipment, rule information including conditions for performing intervention operation is generated by executing machine learning based on past intervention operation results, and this is performed based on such rule information and the current status of machinery and equipment. Techniques for determining and directing intervention operations to be performed are known (see, for example, Patent Document 2). The above-mentioned rule information corresponds to a so-called predictive model.

Here, by applying the technique disclosed in Patent Document 2 to the technique disclosed in Patent Document 1, for example, by machine learning the sensor value indicated by the sensor when a failure occurs, it is considered that failure prediction by a prediction model becomes possible. ..

Japanese Unexamined Patent Publication No. 2011-230634 Japanese Unexamined Patent Publication No. 2017-049801

However, the above-mentioned prior art has room for further improvement in improving the failure prediction accuracy of mechanical equipment.

Specifically, it is practically difficult to collect and cover machine learning data showing the status of all failures in the first stage of generating a predictive model. In particular, when the mechanical equipment is a large mechatronics machine such as a large refrigerator or a plant, it can be said that it is more difficult because the number of sensors provided is enormous.

Even if data can be collected, machine learning is executed, and a prediction model can be generated, if an unexpected pattern still exists in the actual operation data, false detection will occur due to the unexpected pattern. There is a risk that it will end up.

One aspect of the embodiment is made in view of the above, and an object thereof is to provide a failure prediction method, a failure prediction device, and a failure prediction program capable of improving the failure prediction accuracy of mechanical equipment.

The failure prediction method according to one embodiment includes a collection step, an extraction step, a generation step, a derivation step, an evaluation step, a determination step, and an update step. The collection process collects sensor data from a plurality of sensors provided in mechanical equipment. In the extraction step, from the sensor data, a predetermined normal period in which the mechanical equipment is in a normal state and an evaluation portion at an arbitrary evaluation time are extracted. The generation step generates a correlation model that models the correlation of the normal state in the machine equipment by executing machine learning using the normal period. In the derivation step, sample data for the normal period is derived based on the output value of the correlation model obtained by inputting the normal period into the correlation model. In the evaluation step, the degree of deviation from the normal state of the mechanical equipment is evaluated based on the output value of the correlation model obtained by inputting the evaluation portion into the correlation model. The determination step determines a failure sign of the mechanical equipment based on the degree of deviation. In the update step, when the false detection of the deviation degree is estimated by the determination step, the additional learning portion including the sensor data corresponding to the false detection extracted by the extraction step and the normal period portion. The correlation model is updated by executing machine learning so that the sample data of the above is reflected together. In addition, the derivation step calculates the degree of deviation for the normal period based on the output value of the correlation model obtained by inputting the normal period into the correlation model, and out of the normal period. , The upper predetermined portion in descending order of the degree of deviation is selected as the sample data for the normal period that plays a bridging role between the correlation model before the update and the additional learning portion .

According to one aspect of the embodiment, it is possible to improve the failure prediction accuracy of mechanical equipment.

FIG. 1A is a schematic explanatory view (No. 1) of the failure prediction method according to the embodiment. FIG. 1B is a schematic explanatory view (No. 2) of the failure prediction method according to the embodiment. FIG. 1C is a schematic explanatory view (No. 3) of the failure prediction method according to the embodiment. FIG. 1D is a schematic explanatory view (No. 4) of the failure prediction method according to the embodiment. FIG. 1E is a schematic explanatory view (No. 5) 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 an explanatory diagram (No. 1) of the additional learning necessity determination process. FIG. 3B is an explanatory diagram (No. 2) of the additional learning necessity determination process. FIG. 4A is a flowchart (No. 1) showing a processing procedure executed by the failure prediction device. FIG. 4B is a flowchart (No. 2) showing a processing procedure executed by the failure prediction device. FIG. 5 is a hardware configuration diagram showing an example of a computer that realizes the function of the failure prediction device.

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

Further, in the following, the machine / equipment subject to the failure sign determination will be referred to as “target machine 100”. The target machine 100 is assumed to be a large-scale mechatronics machine such as a power plant.

First, an outline of the failure prediction method according to the present embodiment will be described with reference to FIGS. 1A to 1E. 1A to 1E are schematic explanatory views (No. 1) to (No. 5) of the failure prediction method according to the embodiment.

As shown in FIG. 1A, the target machine 100 includes a sensor group of sensors S-1 to Sn. Then, in the failure prediction method according to the present embodiment, the correlation between the sensors S-1 to Sn is grasped based on the sensor data from the sensor group, and the entire target machine 100 is based on the change in the correlation. It is intended to grasp the change in the behavior of.

Specifically, as shown in FIG. 1B, in the failure prediction method according to the present embodiment, machine learning is performed using the sensor data of each sensor S-1 to Sn showing the correlation of "normal period". This is executed (step S1), and the correlation model 12c is generated. In this embodiment, known methods such as random forest and deep learning can be used as the machine learning algorithm.

Here, the "normal period portion" refers to a predetermined period portion in which the target machine 100 was in a normal state at the initial stage of operation or the like. In this embodiment, the normal period is assumed to be "30 days".

The normal state of the target machine 100 can be modeled by the correlation model 12c generated based on the correlation between the sensors S-1 to Sn for the normal period. Then, in the failure prediction method according to the present embodiment, as shown in FIG. 1B, the sensor data of each sensor S-1 to Sn showing the correlation of the "evaluation portion" is input to the correlation model 12c. , The correlation error is calculated from the output value (regression value) of the correlation model 12c obtained as a result. Then, the degree of deviation from the normal state is evaluated based on the error of the correlation (step S2).

If the degree of deviation from the normal state is large, it is possible to predict the failure of the target machine 100 as a sign of failure. Here, the "evaluation portion" refers to a portion based on real-time data currently output from the target machine 100, for example.

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 correlation between the sensors S-1 to Sn, and the change in the behavior is modeled on the correlation in the normal state. It can be obtained from the output value of the converted correlation model 12c. Then, the failure of the target machine 100 is predicted by the magnitude of the degree of deviation from the normal state based on the output value. For example, if the degree of deviation from the normal state is equal to or higher than a predetermined determination threshold value indicating that there is a sign of failure, an alert is notified.

Therefore, according to the failure prediction method according to the present embodiment, all failure phenomena that are required if it is a failure prediction method using a prediction model generated by machine learning based on sensor data at the time of failure, for example. There is no need for complicated processes such as the implementation of modeling. That is, according to the failure prediction method according to the present embodiment, it is possible to easily grasp the failure sign of the target machine 100.

Not limited to the correlation between the sensors S-1 to Sn at one time point, in the failure prediction method according to the present embodiment, as shown in FIG. 1A, the sensors of the sensors S-1 to Sn are used. Machine learning is performed by including the time fluctuation of data, that is, the correlation of time series in the feature vector (hereinafter, simply referred to as "vector"), and it is possible to grasp the sign of failure appearing in the time fluctuation of the sensor data. As a result, the failure sign of the target machine 100 can be accurately grasped.

Further, in the failure prediction method according to the present embodiment, as shown in FIG. 1A, machine learning can be performed by including the time fluctuation of the ambient condition of the target machine 100, for example, the temperature data in the vector. For example, some of the sensor data of the sensors S-1 to Sn have a high correlation with the temperature, and therefore the sensor data may be affected by seasonal fluctuations.

Therefore, by performing machine learning that includes the time fluctuation of the temperature data in the vector, the influence of the seasonal fluctuation can be reduced. It should be noted that the reason why the time variation is extracted from the temperature data as in the case of the sensor data is that the influence of the temperature change acts on the sensor data with a time delay. This makes it possible to accurately grasp the failure sign of the target machine 100.

By the way, it is practically difficult to collect and cover all the data indicating the normal state (hereinafter referred to as "normal data") for the "normal period" from the beginning. Therefore, when the "evaluation" is evaluated using the unripe correlation model 12c at the initial stage of generation, the "evaluation" includes normal data that was not covered when the correlation model 12c was generated. If this is the case, there is a risk of erroneous detection as a sign of failure despite the normal state.

Specifically, it will be described with reference to FIG. 1C. FIG. 1C is obtained by inputting the operation data from the time d0 as the “evaluation portion” to the correlation model 12c generated using the normal data of the “normal period portion” shown in the figure and based on the output value. The degree of deviation of each sensor S-1 to Sn from the normal state is plotted along the time axis. For convenience of explanation, the threshold value for determining the degree of deviation is set to about 0.4, but the threshold value in actual operation is not limited.

Here, it is assumed that the "alert appropriate time point" that should actually be detected as a sign of failure is time d4. In the case of the example shown in FIG. 1C, in addition to the time d4, the time points exceeding the threshold value exist in various places such as the time d1 and the time d3, and therefore, for example, at the time d1, an "alert false alarm" is made as a sign of failure. It shall be.

In order to deal with such false positives, by re-learning the false positive operation data as indicating the normal state, the degree of deviation from the normal state when the false positive pattern operation data is input can be determined. It is possible to reduce it.

However, when the correlation model 12c is newly recreated, it may take too much time to generate the model because the machine learning is restarted from the beginning by adding the additional data for additional learning to the initial normal data. In order to deal with this point, it is conceivable to update the correlation model 12c by reflecting the additional data for additional learning in the existing correlation model 12c, but in this case, the vector space of the original model is destroyed. It may not be possible to express the original normal state.

Therefore, in the failure prediction method according to the present embodiment, the correlation model 12c is updated by reflecting not only the additional data for additional learning but also the normal sample data obtained by extracting a part of the original normal data. And said.

Specifically, in the failure prediction method according to the present embodiment, as shown in FIG. 1C, an additional period (for example, time) that first includes operation data of a falsely detected pattern and is regarded as a normal period to be added. Additional data for additional learning is extracted from d0 to time d2) (step S3).

Further, in the failure prediction method according to the present embodiment, normal sample data which is a part of the normal data for the “normal period” is extracted (step S4). Here, the normal sample data is, for example, data obtained by extracting about 10% of the normal data (that is, about 3 days). As the normal sample data, for example, among the normal data, the top 10% is selected from the data having a high degree of deviation from the normal state (close to the abnormal state).

The normal sample data selected in this way contains the vector component of the original model and shows the characteristics of the vector space of the original model, and also has the characteristics of the additional data because of the high degree of divergence. It also has. Therefore, if it is reflected in the original model together with the additional data, it plays a bridging role between the original model and the additional data, and makes it difficult to break the original model.

That is, in the failure prediction method according to the present embodiment, as shown in FIG. 1D, the additional data extracted in step S3 and the normal sample data extracted in step S4 are used for the pre-update model based on the normal state before additional learning. Is executed to generate an updated model by reflecting the above (step S5). As a result, the correlation model 12c is updated to the one in which the new tendency is learned while maintaining the original tendency.

Similar to FIG. 1C, the operation data from the time d0 is input as the "evaluation portion" for the correlation model 12c updated in this way, and the degree of deviation from the normal state based on the output value is plotted in the figure. Shown in 1E.

According to FIG. 1E, for the time d0 to time d2 specified as the additional period of machine learning based on the “alert false alarm” of FIG. 1C, the degree of deviation from the normal state is reduced by the updated correlation model 12c. It can be seen that at the time d1 when the alert is falsely reported, "no alert false alarm" is obtained.

Further, it can be seen that the degree of deviation from the normal state is generally lowered to less than the threshold value by the updated correlation model 12c for the other periods showing a pattern similar to the original time d0 to time d2. Then, it can be seen that the time d4, which should be actually detected as a sign of failure, indicates the degree of deviation above the threshold value and becomes "appropriately alertable".

As described above, in the failure prediction method according to the present embodiment, when additional learning of the correlation model 12c is required, not only the additional data for the additional learning but also the normal sample data obtained by extracting a part of the original normal data. It was decided to update the correlation model 12c by adding and reflecting the above.

Therefore, according to the failure prediction method according to the present embodiment, the correlation model 12c can be appropriately updated to the one in which the new tendency is learned while maintaining the original tendency, and false detection can be prevented to prevent the target machine. The failure prediction accuracy of 100 can be improved.

Further, according to the failure prediction method according to the present embodiment, since the correlation model 12c is not newly recreated, it is possible to prevent it from taking too much time to update the model. Therefore, it is possible to reduce the time loss in operation.

Whether or not additional learning of the correlation model 12c is necessary may be automatically determined by the system according to, for example, the degree of deviation from the normal state derived during operation, or the knowledge of a person such as an operator. May be judged by. This point will be described later with reference to FIGS. 3A and 3B.

Hereinafter, the configuration of the failure prediction system 1 to which the above-mentioned 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, the components necessary for explaining the features of the present embodiment are represented by functional blocks, and the description of general components is omitted.

In other words, each component shown in FIG. 2 is a functional concept and does not necessarily have to be physically configured as shown in the figure. 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 part of it is functionally or physically distributed in arbitrary units according to various loads and usage conditions. -It is possible to integrate and configure.

In the description using 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, a target machine 100, and an ambient situation providing unit 200. The failure prediction device 10 is connected to the target machine 100 and the surrounding condition providing unit 200 so as to be able to communicate with each other, and the failure predicting device 10 is provided with sensor data from the target machine 100 and surroundings from the surrounding condition providing unit 200. Situation data can be collected as appropriate.

The surrounding condition providing unit 200 is a facility or device that provides surrounding condition data. The ambient data is, for example, temperature data. In such a case, the surrounding condition providing unit 200 corresponds to, for example, a public or private weather information providing facility, a temperature sensor, or the like. Further, the surrounding situation providing unit 200 may provide the surrounding situation by manual input of the operator via the input device.

The failure prediction device 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, a normal sample derivation unit 11d, an evaluation unit 11e, a determination unit 11f, a notification unit 11g, and an update unit 11h.

The storage unit 12 is a storage device such as a hard disk drive, a non-volatile memory, or a register, and includes collected data 12a, training data set 12b, correlation model 12c, normal sample data 12d, evaluation data set 12e, and evaluation information. 12f and the additional learning data set 12g are stored. The evaluation information 12f includes a deviation degree 12fa and a contribution rate 12fb.

The control unit 11 controls the entire failure prediction device 10. The collecting unit 11a collects sensor data from the sensor group of the target machine 100 at a predetermined cycle and stores it in the collected data 12a. The predetermined cycle for collecting may be about 15 minutes to 1 hour in order to detect a gradual change in behavior that indicates a sign of failure due to aging or the like. Further, the collecting unit 11a also collects the surrounding condition data from the surrounding condition providing unit 200 at a predetermined cycle and stores it in the collected data 12a.

The extraction unit 11b extracts a data set for the normal period from the collected data 12a based on the normal period set at the first operation, and stores the data set in the learning data set 12b. Further, the extraction unit 11b extracts the data set for evaluation from the collected data 12a and stores it in the evaluation data set 12e.

Further, when the extraction unit 11b receives the instruction for extracting the additional data for the additional period from the determination unit 11f, the extraction unit 11b extracts the data set for the additional period from the collected data 12a and stores it in the additional learning data set 12g.

The generation unit 11c executes machine learning using the training data set 12b at the first operation, and generates the correlation model 12c.

The normal sample derivation unit 11d calculates the degree of deviation from the normal state based on the output value of the correlation model 12c with the learning data set 12b as an input. Further, the normal sample derivation unit 11d selects the top 10% of normal data from the learning data set 12b in order from the one with the highest degree of deviation from the calculated normal state, and stores it in the normal sample data 12d. Further, the normal sample derivation unit 11d derives a normal sample even at the time of additional learning. This point will be described later.

The evaluation unit 11e inputs the evaluation data set 12e extracted by the extraction unit 11b into the correlation model 12c, and receives the output result of the correlation model 12c. Then, the evaluation unit 11e calculates various evaluation values for determining a failure sign based on the received output result, and stores them in the evaluation information 12f. The evaluation values correspond to a deviation degree of 12fa and a contribution rate of 12fb.

Specifically, the evaluation unit 11e calculates the degree of deviation from the normal state based on the error between the input and the output when the evaluation data set 12e is input to the correlation model 12c, that is, the error of the correlation, and shifts to the degree of deviation 12fa. Store.

Further, the evaluation unit 11e acquires the contributions of the sensors S-1 to Sn from the correlation model 12c, and uses the formula “contribution rate i = contribution i / Σ contribution i” to obtain the contributions of the sensors S-1 to S. -Calculate the contribution rate for each n. Further, the evaluation unit 11e stores the calculated contribution rate in the contribution rate 12fb of the evaluation information 12f.

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

When the notification unit 11g receives a notification request instruction with a sign of failure from the determination unit 11f, the notification unit 11g notifies an external device such as a display unit of an alert. At this time, the notification unit 11g also refers to the deviation degree 12fa and the contribution rate 12fb, and notifies the deviation degree from the normal state and, for example, the names of the sensors S-1 to Sn having the higher contribution rate. can do.

The operator can know the degree of abnormality of the entire target machine 100 by checking, for example, the degree of deviation from the normal state. Further, for example, the cause of the abnormality can be estimated by checking the sensor with the higher contribution rate.

The determination unit 11f determines that additional learning is necessary when the degree of deviation from the normal state is equal to or higher than a predetermined threshold value, but this is presumed to be a false detection, and the extraction unit 11b determines that additional learning is required for the additional period. Instruct additional data extraction.

Here, the process of determining the necessity of additional learning will be described with reference to FIGS. 3A and 3B. 3A and 3B are explanatory views (No. 1) and (No. 2) of the additional learning necessity determination process. Note that FIG. 3A is an example of a case where the determination unit 11f executes the additional learning necessity determination process. Further, FIG. 3B is an example of a case where the additional learning necessity determination process is determined based on the knowledge of the operator.

First, FIG. 3A shows a case where the degree of deviation from the normal state is represented by a schematic waveform, and the waveform has a so-called impulse noise-like peak. In such a case, if the peak is equal to or higher than the failure prediction determination threshold TH1, the alert is normally notified as a failure sign, but the determination unit 11f considers this to be a false detection if the conditions described below are satisfied. Estimate and determine that additional learning is needed.

Specifically, as shown in FIG. 3A, the determination unit 11f is provided with an impulse-like determination threshold value TH2 smaller than, for example, a failure prediction determination threshold value TH1, and is less than the impulse-like determination threshold value TH2 over which predetermined sections i1 and i2 before and after the peak are applied. If so, the peak is in the form of impulse noise, and it is determined that additional learning is required for false detection (step S31). Then, the determination unit 11f has the section I including the peak extracted as additional data (step S32). That is, according to the example shown in FIG. 3A, the failure prediction accuracy of the target machine 100 can be improved by automatic additional learning.

It should be noted that the example shown in FIG. 3A is merely an example, and does not limit the processing procedure of the additional learning necessity determination process. For example, if the integrated value of the degree of deviation for the section I in FIG. 3A is less than a predetermined value, it may be presumed to be a false positive.

Further, as shown in FIG. 3B, it is determined that additional learning is necessary by estimating false detection based on the operator's knowledge from the notification result to the display unit or the like (step S31'), and the operator uses the input unit or the like (not shown). Additional data may be specified and additional learning may be instructed (step S32'). That is, according to the example shown in FIG. 3B, the failure prediction accuracy of the target machine 100 can be improved by additional learning of human knowledge.

Returning to the description of FIG. 2, the update unit 11h will be described subsequently. The update unit 11h updates the correlation model 12c by additional learning that reflects both the additional learning data set 12g and the normal sample data 12d.

When the correlation model 12c is updated by the update unit 11h, the normal sample derivation unit 11d calculates the degree of deviation from the normal state based on the output value of the correlation model 12c with the additional learning data set 12g as an input. Further, the normal sample derivation unit 11d selects additional data of about the top 10% from the additional learning data set 12g in order from the one having the highest degree of deviation from the calculated normal state, and adds and holds the additional data to the normal sample data 12d.

That is, in the normal sample data 12d, a normal sample based on the normal data at the time of initial generation of the correlation model 12c and a normal sample based on the additional data for each additional learning are accumulated.

Next, the processing procedure executed by the failure prediction device 10 will be described with reference to FIGS. 4A and 4B. 4A and 4B are flowcharts (No. 1) and (No. 2) showing a processing procedure executed by the failure prediction device 10. In the processing procedure shown here, it is assumed that the determination unit 11f executes the above-mentioned additional learning necessity determination process.

As shown in FIG. 4A, first, the control unit 11 determines whether or not it is the first operation (step S101). Here, in the case of the first operation (step S101, Yes), the control unit 11 subsequently sets the normal period (step S102).

In step S102, in addition to the normal period (for example, "30 days from XX days"), the ratio of data extracted as sample data (for example, "10%") is set on the system.

Subsequently, the generation unit 11c generates the correlation model 12c from the learning data set 12b extracted by the extraction unit 11b, that is, the normal data for the normal period by the machine learning algorithm (step S103).

Then, the normal sample derivation unit 11d calculates the degree of deviation from the normal state based on the output value of the correlation model 12c with the normal data as an input (step S104). Further, the normal sample derivation unit 11d extracts the higher-order normal data in descending order of the degree of deviation from the calculated normal state, and holds the normal sample data 12d (step S105).

If it is not the first operation in step S101 (steps S101 and No), control is transferred to step S106.

Subsequently, the evaluation unit 11e calculates the degree of deviation from the normal state based on the evaluation data set 12e extracted by the extraction unit 11b, that is, the output value of the correlation model 12c using the evaluation data as an input (step S106). Further, the evaluation unit 11e calculates the contribution ratio of each sensor S-1 to Sn (step S107).

Then, the determination unit 11f determines whether or not the degree of deviation calculated by the evaluation unit 11e is equal to or greater than a predetermined determination threshold value (step S108). Here, when the degree of deviation is equal to or higher than the determination threshold value (step S108, Yes), the notification unit 11g notifies the sensors S-1 to Sn having the degree of deviation from the normal state and the higher contribution rate (step S109). .. If the determination condition of step S108 is not satisfied (steps S108, No), control is transferred to step S110.

Next, as shown in FIG. 4B, the determination unit 11f determines the necessity of additional learning (step S110). Here, when it is determined that additional learning is required (step S110, Yes), the determination unit 11f sets an additional period (step S111), and the extraction unit 11b is made to extract the additional learning data set 12g. If it is determined that additional learning is unnecessary (steps S110, No), the process ends as it is.

Subsequently, the update unit 11h updates the correlation model 12c from the additional learning data set 12g (that is, additional data for the additional period) extracted by the extraction unit 11b and the normal sample data 12d held (step S112).

Then, the normal sample derivation unit 11d calculates the degree of deviation from the normal state based on the output value of the correlation model 12c with the additional data as an input (step S113).

Further, the normal sample derivation unit 11d extracts higher-order additional data in descending order of the degree of deviation from the calculated normal state, and adds and holds the additional data to the normal sample data 12d (step S114). Then, the process is terminated.

The failure prediction device 10 according to the embodiment is realized by, for example, a computer 60 having a configuration as shown in FIG. FIG. 5 is a hardware configuration diagram showing an example of a computer that realizes the function of the failure prediction device 10. The computer 60 includes a CPU (Central Processing Unit) 61, a RAM (Random Access Memory) 62, a ROM (Read Only Memory) 63, an HDD (Hard Disk Drive) 64, a communication interface (I / F) 65, and an input / output interface (I). It includes a / F) 66 and a media interface (I / F) 67.

The CPU 61 operates based on the program stored in the ROM 63 or the HDD 64, and controls each part. The ROM 63 stores a boot program executed by the CPU 61 when the computer 60 is started, 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 another device via the communication network and sends the data to the CPU 61, and the data generated by the CPU 61 is transmitted via the communication network. Send to other devices.

The CPU 61 controls an output device such as a display or a printer, and an input device such as a keyboard or 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 the program or data 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 MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. And so on.

When the computer 60 functions as the failure prediction device 10, the CPU 61 of the computer 60 executes the program loaded on the RAM 62 to execute the collection unit 11a, the extraction unit 11b, the generation unit 11c, the normal sample derivation unit 11d, and the evaluation unit. Each function of the unit 11e, the determination unit 11f, the notification unit 11g, and the update unit 11h is realized. Further, the HDD 64 realizes the function of the storage unit 12 and stores the collected data 12a and the like.

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

As described above, the failure prediction device 10 according to the embodiment includes a collection unit 11a, an extraction unit 11b, a generation unit 11c, a normal sample derivation unit 11d (corresponding to an example of the “deriving unit”), and an evaluation unit. 11e, a determination unit 11f, and an update unit 11h are included.

The collecting unit 11a collects sensor data of a plurality of sensors S-1 to Sn provided in the target machine 100 (corresponding to an example of "machinery and equipment"). The extraction unit 11b extracts the sensor data for a predetermined normal period in which the target machine 100 is in a normal state and the evaluation portion at an arbitrary evaluation time.

The generation unit 11c generates a correlation model 12c that models the correlation of the normal state in the target machine 100 by executing machine learning using the normal period. The normal sample derivation unit 11d derives the sample data for the normal period based on the output value of the correlation model 12c obtained by inputting the normal period portion into the correlation model 12c.

The evaluation unit 11e evaluates the degree of deviation of the target machine 100 from the normal state based on the output value of the correlation model 12c obtained by inputting the evaluation portion into the correlation model 12c. The determination unit 11f determines a failure sign of the target machine 100 based on the degree of deviation.

When the determination unit 11f estimates the erroneous detection of the degree of deviation, the update unit 11h includes additional learning including sensor data corresponding to the erroneous detection extracted by the extraction unit 11b, and a sample for the normal period. The correlation model 12c is updated by performing machine learning so that the data are reflected together.

Therefore, according to the failure prediction device 10 according to the present embodiment, the failure prediction accuracy of the target machine 100 can be improved.

In the above-described embodiment, random forest and deep learning are taken as examples of machine learning algorithms, but the algorithms are not limited. Therefore, machine learning may be executed by a regression analysis method such as support vector regression using a pattern classifier such as SVM (Support Vector Machine) to generate a correlation model 12c. Further, here, the pattern classifier is not limited to the SVM, and may be, for example, AdaBoost.

Further, in the above-described embodiment, the normal period is "30 days" and the ratio of the data extracted as sample data is "10%", but of course, this is just an example, for example, system operation. In addition, the set value adjusted to the optimum may be used.

Further effects and variations can be easily derived by those skilled in the art. For this reason, the broader aspects of the invention are not limited to the particular details and representative embodiments described and described above. Thus, various modifications can be made without departing from the spirit or scope of the overall concept of the invention as defined by the appended claims and their equivalents.

1 Failure prediction system 10 Failure prediction device 11 Control unit 11a Collection unit 11b Extraction unit 11c Generation unit 11d Normal sample derivation unit 11e Evaluation unit 11f Judgment unit 11g Notification unit 11h Update unit 12 Storage unit 12a Collected data 12b Learning data set 12c Correlation model 12d Normal sample data 12e Evaluation data set 12f Evaluation information 12fa Degree of deviation 12fb Contribution rate 12g Additional learning data set 100 Target machine 200 Surrounding condition provider S-1 to Sn sensors

Claims (5)

A collection process that collects sensor data from multiple sensors installed in machinery and equipment,
From the sensor data, an extraction step for extracting a predetermined normal period in which the machine and equipment were in a normal state and an evaluation portion at an arbitrary evaluation time, and
A generation step of generating a correlation model that models the correlation of the normal state in the machine equipment by executing machine learning using the normal period.
A derivation step of deriving sample data for the normal period based on the output value of the correlation model obtained by inputting the normal period into the correlation model.
An evaluation step of evaluating the degree of deviation from the normal state of the mechanical equipment based on the output value of the correlation model obtained by inputting the evaluation portion into the correlation model, and an evaluation step.
A determination process for determining a failure sign of the mechanical equipment based on the degree of deviation, and
When the false detection of the degree of deviation is estimated by the determination step, both the additional learning portion including the sensor data corresponding to the false positive extracted by the extraction step and the sample data for the normal period are both. It includes an update process that updates the correlation model by performing machine learning to be reflected.
The derivation step is
The degree of deviation for the normal period is calculated based on the output value of the correlation model obtained by inputting the normal period into the correlation model, and the degree of deviation is calculated in descending order of the normal period. A failure prediction method, characterized in that the upper predetermined portion of the above is selected as sample data for the normal period , which plays a bridging role between the correlation model before update and the additional learning portion . The derivation step is
When the correlation model is updated by the update step, sample data for the additional training is derived based on the output value of the correlation model obtained by inputting the additional training to the correlation model, and the above. The failure prediction method according to claim 1, wherein the data is added to the sample data for a normal period. The update process is
The correlation model is updated every time the additional learning portion is newly extracted by the extraction step.
The derivation step is
The failure prediction method according to claim 2, wherein sample data for the additional learning is derived and added to the sample data for the normal period each time the correlation model is updated by the update step. A collection unit that collects sensor data from multiple sensors installed in mechanical equipment,
From the sensor data, an extraction unit for extracting a predetermined normal period in which the machine and equipment were in a normal state and an evaluation portion at an arbitrary evaluation time, and
A generator that generates a correlation model that models the correlation of the normal state in the machine equipment by executing machine learning using the normal period.
A derivation unit that derives sample data for the normal period based on the output value of the correlation model obtained by inputting the normal period into the correlation model.
An evaluation unit that evaluates the degree of deviation from the normal state of the machinery and equipment based on the output value of the correlation model obtained by inputting the evaluation portion into the correlation model.
A determination unit that determines a failure sign of the mechanical equipment based on the degree of deviation, and a determination unit.
When the false detection of the degree of deviation is estimated by the determination unit, the additional learning portion including the sensor data corresponding to the false detection extracted by the extraction unit and the sample data are both reflected. It is equipped with an update unit that updates the correlation model by executing machine learning.
The derivation unit is
The degree of deviation for the normal period is calculated based on the output value of the correlation model obtained by inputting the normal period into the correlation model, and the degree of deviation is calculated in descending order of the normal period. A failure prediction device, characterized in that the upper predetermined portion of the above is selected as sample data for the normal period , which plays a bridging role between the correlation model before update and the additional learning portion . A collection procedure for collecting sensor data from multiple sensors installed in machinery and equipment,
From the sensor data, an extraction procedure for extracting a predetermined normal period in which the mechanical equipment was in a normal state and an evaluation portion at an arbitrary evaluation time, and
A generation procedure for generating a correlation model that models the correlation of the normal state in the machine equipment by executing machine learning using the normal period.
A derivation procedure for deriving sample data for the normal period based on the output value of the correlation model obtained by inputting the normal period into the correlation model, and a derivation procedure.
An evaluation procedure for evaluating the degree of deviation of the mechanical equipment from the normal state based on the output value of the correlation model obtained by inputting the evaluation portion into the correlation model, and an evaluation procedure.
A determination procedure for determining a failure sign of the mechanical equipment based on the degree of deviation, and a determination procedure.
When the false detection of the degree of deviation is estimated by the determination procedure, the additional learning including the sensor data corresponding to the false detection extracted by the extraction procedure and the sample data are both reflected. By executing machine learning, the computer is made to execute the update procedure for updating the correlation model.
The derivation procedure is
The degree of deviation for the normal period is calculated based on the output value of the correlation model obtained by inputting the normal period into the correlation model, and the degree of deviation is calculated in descending order of the normal period. A failure prediction program characterized in that the upper predetermined portion of the above is selected as sample data for the normal period that plays a bridging role between the correlation model before update and the additional learning portion .

Images