JP7600679B2 - Machine learning device and failure analysis device - Google Patents

Description

The present invention relates to a machine learning device and a fault analysis device.

If a fault occurs in the communication equipment or line equipment used in the power plant's network, the location of the fault is identified based on an alarm displayed in the NMS (Network Management System), an alarm output to the line management system and related business offices, and a communication check.

For example, Patent Document 1 discloses a technology in which a security utility server (SUS) allows a network management element to set thresholds and/or other criteria, and when the thresholds or criteria are exceeded, outputs an alert that results in a specific action, such as sending a notification to an NMS, while nodes in the network detect anomalies using an anomaly detection mechanism.

JP 2020-017300 A

However, when identifying the location of a fault based on an alarm display in the NMS included in the power plant network, the time required to identify the fault varies depending on the knowledge of the person in charge about the communication equipment and line equipment. Furthermore, alarms in the NMS are not only output when a fault occurs in the communication equipment or line equipment, but are also output when communication equipment or line equipment is shut down, for example, in connection with power outage or communication work. For this reason, if a fault occurs in the communication equipment or line equipment during work that generates many alarms, the alarm display associated with the fault may be mixed in with the alarm displays associated with the power outage or communication work, and the person in charge may not notice the alarm display associated with the fault until they receive a communication fault report from the operation site.

The present invention aims to provide a machine learning device and a fault analysis device that can automatically identify faults based on alarm displays, regardless of the amount of knowledge about communication equipment and line equipment.

To achieve the above objective, the present invention has the following configuration.

(1) A machine learning device that constructs a learning model for identifying faults in communication facilities or line facilities from facility information acquired through a network, the machine learning device comprising: a first input data acquisition means that acquires information related to an alarm occurring in a monitoring device of the network as the facility information as first input data; a first label acquisition means that acquires a fault determination result corresponding to the alarm as a first label; and a first learning model construction means that performs supervised learning using a pair of the first input data and the first label as teacher data to construct a first learning model for generating the determination result corresponding to the alarm occurring in the monitoring device.

According to (1), it is possible to provide a machine learning device that can automatically identify faults based on alarm displays, regardless of the amount of knowledge about communication equipment and line equipment.

(2) In the machine learning device described in (1), it is preferable that the device further comprises a second input data acquisition means for acquiring information related to an output waveform from a microwave radio as the communication equipment as second input data, a second label acquisition means for acquiring a fault judgment result corresponding to the output waveform as a second label, and a second learning model construction means for constructing a second learning model for generating the judgment result corresponding to the output waveform by performing supervised learning using a pair of the second input data and the second label as teacher data.

(2) It is possible to provide a machine learning device that can automatically identify faults based on the output waveform from a microwave radio, regardless of the amount of knowledge about communication equipment and line equipment.

(3) A fault analysis device using the first learning model constructed by the machine learning device described in (1) or (2), comprising: an alarm acquisition means for acquiring a new alarm; an alarm sorting means for sorting the alarms acquired by the alarm acquisition means into alarms that have been registered in advance and alarms that have not been registered in advance; and a first fault determination means for determining the occurrence of the fault using the alarms that have not been registered in advance and the first learning model.

(3) makes it possible to provide a fault analysis device that automatically identifies faults based on alarm displays, regardless of the level of knowledge about communication equipment and line equipment.

(4) In the fault analysis device described in (3), it is preferable that the first fault determination means further includes a confirmation item output means for determining, in addition to the fault, the probability of the fault occurring, and outputting a confirmation item for increasing the probability of the fault occurring if the probability of the fault occurring is not 100%.

(4) makes it possible to increase the probability of occurrence of a fault as a result of the determination made by the first learning model.

(5) In the fault analysis device described in (3) or (4), the fault analysis device uses the second learning model constructed by the machine learning device described in (2), and includes a second fault determination means that determines the occurrence of the fault using the output waveform and the second learning model when the log of the non-preregistered alarm indicates that the output from the microwave radio exceeds a threshold value, and it is preferable that the first fault determination means determines the occurrence of the fault using the non-preregistered alarm and the first learning model when the log of the non-preregistered alarm does not indicate that the output exceeds a threshold value.

(5) makes it possible to provide a fault analysis device that can automatically identify faults based on the output waveform from a microwave radio, regardless of the level of knowledge about communication equipment and line equipment.

According to the present invention, it is possible to identify faults based on alarm displays, regardless of the level of knowledge about communication equipment and line equipment.

1 is a diagram showing an overall configuration of a fault analysis system according to an embodiment of the present invention; FIG. 2 is a functional block diagram of a first learning unit included in the machine learning device according to the embodiment of the present invention. 1 is a functional block diagram of a failure analysis device according to an embodiment of the present invention. 4 is a flowchart showing an operation of the machine learning device according to the embodiment of the present invention. 4 is a flowchart showing an operation of the machine learning device according to the embodiment of the present invention. 4 is a flowchart showing an operation of the failure analysis device according to the embodiment of the present invention.

The following describes an embodiment of the present invention with reference to Figures 1 to 6.

[Configuration of 1 embodiment]
1.1 Overall Configuration
First, the configuration of a fault analysis system 100 according to this embodiment will be described. Fig. 1 is a diagram showing the overall configuration of the fault analysis system 100. As shown in Fig. 1, the fault analysis system 100 includes a machine learning device 10, a fault analysis device 20, a monitoring device 30, a terminal device 40, and a network 50. Note that the machine learning device 10, the fault analysis device 20, the monitoring device 30, and the terminal device 40 may be one or more.

The machine learning device 10, the fault analysis device 20, the monitoring device 30, and the terminal device 40 are each connected to a network 50 and can communicate with each other via the network 50. The network 50 is, for example, a local area network (LAN), the Internet, a public telephone network, or a combination of these. There are no particular limitations on the specific communication method in the network 50, or whether the connection is wired or wireless. The machine learning device 10 and the fault analysis device 20 may be directly connected via a connection unit, rather than communicating via the network 50, and the fault analysis device 20 and the monitoring device 30 may be directly connected via a connection unit.

The machine learning device 10 is a device that constructs a learning model to be used by the fault analysis device 20 through supervised learning. Therefore, as shown in FIG. 1, the machine learning device 10 includes a first learning unit 11 and a second learning unit 12.

The fault analysis device 20 is a device that analyzes faults occurring in the NMS by applying alarms detected by the monitoring device 30 (described below) to the learning model constructed by the machine learning device 10.

The monitoring device 30 is a device that monitors the status of the NMS by detecting failures occurring in, for example, communication equipment or line equipment provided in the NMS based on alarms obtained from the communication equipment or line equipment.

The terminal device 40 is a device that acquires and displays the fault analysis results output from the fault analysis device 20 and the monitoring results output from the monitoring device 30, and also outputs input from the user of the terminal device 40 to the fault analysis device 20 and the monitoring device 30. The terminal device 40 is preferably realized by, for example, a mobile phone such as a smartphone, or a mobile terminal such as a tablet, but is not limited to this. Also, in FIG. 1, multiple terminal devices 40a to 40q are shown as the terminal device 40, but this is not limited to this, and there may be only one, for example.

[1.2 Configuration of machine learning device]
As described above, the machine learning device 10 includes the first learning unit 11 and the second learning unit 12. The first learning unit 11 constructs a first learning model for determining a fault occurring in the NMS from information related to an alarm occurring in the monitoring device 30 of the NMS. The second learning unit constructs a second learning model for determining a fault occurring in the NMS from information related to an output waveform from a microwave radio provided in the NMS.

Figure 2 is a functional block diagram of the first learning unit 11. As shown in Figure 2, the first learning unit 11 includes a first input data acquisition unit 111, a first label acquisition unit 112, a first learning model construction unit 113, and a first learning model storage unit 114.

The first input data acquisition unit 111 acquires information related to alarms occurring in the monitoring device of the NMS as the first input data. Note that examples of this alarm-related information include an alarm called "N150MOPT High-Speed Section REC", an alarm called "MJ", an alarm called "4WVA1 Low-Speed IF Section EQP1", an alarm called "Bulk Fault", and an alarm called "Equipment Abnormality" in a specific power plant.

Here, "N150MOPT" refers to the 150M (150 Mega) optical transmission section of the normal line. In particular, "N" means that it is a "normal" line, and when the normal line stops, it switches to the backup line indicated by "E" (Emergency). "OPT" indicates light emission/transmission.
In addition, the "high speed" in "high speed section" refers to a high speed line with a high multiplexing speed, which is defined by the international standard ISO and comes in many types, such as transmission speeds of STM-1 (155.52 Mbps) and STM-16 (2488.32 Mbps), and a "section" refers to the area or section in an optical multiplexing device through which a high speed line passes.
"REC" is an alarm output indication that occurs when optical input is cut off or frame synchronization is lost.
Also, "4WVA1" is the name of a board that accommodates 4W lines (mainly voice signals, telephone lines, etc.) in an optical multiplexing device.
Moreover, the "low-speed IF unit" refers to an interface used for a low-speed line.
Additionally, "EQP1" is an abbreviation for "Equipment 1."
In addition, a "lump fault" is an alarm that is issued when an alarm occurs within a device in order to make it easier to identify the device in question, and detailed alarm verification and confirmation are required for the device identified by the alarm.
In addition, "equipment abnormality" refers to an alarm that notifies of a warning of a unit that makes up the equipment, rather than of the entire equipment, and does not involve equipment or line shutdown, but indicates a state in which reliability is reduced (for example, in the case of a redundant system, one of the systems has stopped).

The first label acquisition unit 112 acquires the fault determination result corresponding to the alarm as the first label. Examples of such determination results include an IC failure in the "N150MOPT" board, a faulty optical core, a failure of the "4WVA1" board, or an abnormality in the device power supply.

The first learning model construction unit 113 performs supervised learning using a pair of the first input data and the first label as teaching data to construct a first learning model for determining a fault occurring in the NMS from information related to an alarm occurring in the NMS monitoring device 30, and transmits the constructed first learning model to the fault analysis device 20.

The first learning model construction unit 113 can be realized, for example, by using a support vector machine (hereinafter also referred to as SVM). For example, by using a logistic regression equation and an accuracy rate table in addition to the support vector machine, the first learning model constructed by this embodiment is preferably not only able to determine the fault corresponding to the alarm, but also able to output the probability that each fault has occurred (for example, Kazuko Takahashi, "Proposal of Ensemble Learning in Support Vector Machines," 23rd National Conference of the Japanese Society for Artificial Intelligence).

This first learning model is a learning model for determining the occurrence and the likelihood of a fault, for example, based on a combination of whether each of a number of alarms is ON or OFF.

The first learning model memory unit 114 stores the learning model constructed by the first learning model construction unit 113.

The second learning unit 12 includes a second input data acquisition unit 121, a second label acquisition unit 122, a second learning model construction unit 123, and a second learning model storage unit 124. Note that the configuration of the second learning unit 12 is similar to the configuration of the first learning unit 11 shown in FIG. 2, and therefore is not shown in the figure.

The second input data acquisition unit 121 acquires, as second input data, information related to the output waveform from a microwave radio serving as communication equipment provided in the NMS.

The second label acquisition unit 122 acquires the fault determination result corresponding to the output waveform as the second label. Note that this determination result may be, for example, an event such as rainfall or a fallen tree.

The second learning model construction unit 123 constructs a second learning model for determining a fault occurring in the NMS from information related to the output waveform from the microwave radio provided in the NMS by performing supervised learning using a pair of the second input data and the second label as teaching data, and transmits the constructed second learning model to the fault analysis device 20. This second learning model is a learning model for determining whether the fault occurring is, for example, rainfall or a fallen tree, depending on the degree of disturbance when the output waveform from the microwave radio is disturbed by rainfall or a fallen tree.

[1.3 Configuration of the fault analysis device]
3 is a functional block diagram of the failure analysis device 20. The failure analysis device 20 includes a storage unit 21, a control unit 22, a communication unit 23, and a display unit 24.

The memory unit 21 stores the first learning model and the second learning model obtained from the machine learning device 10.

The memory unit 21 further stores a pre-registration database 211, an alarm fault response database 212, and a fault characteristics database 213.

The pre-registration database 211 is a database that pre-registers work alarms that occur when communication equipment or line equipment is shut down for power outage work or communication work. For example, a user of the fault analysis device 20 can input information related to the work alarm from the terminal device 40, and the work alarm that will occur in the future is registered in the pre-registration database 211.

The alarm fault correspondence database 212 is a database that defines the correspondence between an alarm and a fault corresponding to the alarm. For example, the alarm fault correspondence database 212 defines the correspondence between a combination of multiple alarms, each of which is ON or OFF, and a fault that has occurred. As described below, a combination of multiple alarms that have actually been output is applied to the first learning model to determine a fault that may have occurred and its accuracy. At this time, by referring to the alarm fault correspondence database 212, if the fault has actually occurred, the original ON-OFF combination of the alarms can be determined. By comparing this with the ON-OFF of the multiple alarms that have actually been output, it can be determined that an alarm that should have been ON is actually OFF, or that an alarm that should have been OFF is actually ON.

The fault characteristics database 213 is a database that defines the correspondence between an occurring fault, the location of the fault, the priority of dealing with the fault, the presence and storage location of fault recovery and spare parts used to recover from the fault, the required travel time to the location of the fault, and the expected time to recover from the fault.

The control unit 22 is a part that controls the entire fault analysis device 20, and realizes various functions in this embodiment by appropriately reading and executing various programs from a storage area such as a ROM, RAM, flash memory, or hard disk (HDD). The control unit 22 may be a CPU. The control unit 22 includes an alarm acquisition unit 221, an alarm distribution unit 222, a first fault determination unit 223, a second fault determination unit 224, a fault candidate output unit 225, a confirmation item output unit 226, a fault identification unit 227, a priority output unit 228, a fault recovery product output unit 229, a required travel time output unit 230, and an expected recovery time output unit 231.

The alarm acquisition unit 221 acquires new alarms from the monitoring device 30 via the communication unit 23 described below, separately from the alarms acquired when constructing the first learning model and the second learning model.

The alarm distribution unit 222 refers to the pre-registration database 211 to determine whether the alarm acquired by the alarm acquisition unit 221 is a pre-registered work alarm or a fault alarm that has not been pre-registered and is output as a result of an actual fault occurring.

The first fault determination unit 223 applies a fault alarm that has not been registered in advance to the first learning model stored in the memory unit 21 to infer and determine the fault that may have occurred and its probability of occurrence.

When a log of a fault alarm not registered in advance indicates that the output from the microwave radio exceeds a threshold, the second fault determination unit 224 infers and determines a fault that may have occurred by applying the output waveform from the microwave radio to the second learning model stored in the memory unit 21. Furthermore, the second fault determination unit 224 outputs the fault that may have occurred to the terminal device 40 via the communication unit 23. Alternatively, the second fault determination unit 224 may output the fault that may have occurred to the display unit 24 described below.

The fault candidate output unit 225 sorts the fault candidates determined by the first fault determination unit 223 in descending order of probability if there are multiple candidates, and outputs the candidates together with the probability to the terminal device 40 via the communication unit 23. Alternatively, the fault candidate output unit 225 may output the fault candidates together with the probability to the display unit 24 described below.

The confirmation item output unit 226 outputs confirmation items for the user of the fault analysis device 20 by referring to the alarm fault response database 212. Specifically, when a combination of multiple alarms not registered in advance that is assigned by the alarm assignment unit 222 and that corresponds to a fault candidate output by the fault candidate output unit 225 is partially different from the alarm combination described in the alarm fault response database 212, the event corresponding to the different alarm is output to the terminal device 40 via the communication unit 23 as a confirmation item for the user. Alternatively, the confirmation item output unit 226 may output the confirmation items for the user to the display unit 24 described below.

For example, suppose that the combination of multiple failure alarms not registered in advance that are distributed by the alarm distribution unit 222 is (ALM1, ALM2, ALM3, ALM4) = (ON, OFF, ON, OFF), and that as a result of applying this to the first learning model, it is determined that failure D has occurred with a 90% probability. In this case, if the original failure alarm combination corresponding to failure D in the alarm failure response database 212 is (ALM1, ALM2, ALM3, ALM4) = (ON, OFF, OFF, OFF), the confirmation item output unit 226 outputs the confirmation item corresponding to ALM3 based on the fact that ALM3, which should be OFF originally, is ON.

The fault identification unit 227 identifies the fault that has actually occurred. For example, the fault identification unit 227 may identify the fault that has actually occurred based on a selection result selected by the user via the terminal device 40 from among the multiple fault candidates output by the fault candidate output unit 225. Alternatively, the fault identification unit 227 may identify the fault that has actually occurred based on a confirmation action by the user for the fault output by the second fault determination unit 224.

The priority output unit 228 refers to the fault characteristics database 213 and outputs the priority of the response to the fault identified by the fault identification unit 227 to the terminal device 40 via the communication unit 23. Alternatively, the priority output unit 228 may output the priority of the response to the fault to the display unit 24 described below.

The failure recovery product output unit 229 refers to the failure characteristics database 213, and outputs to the terminal device 40 via the communication unit 23 the presence or absence and storage location of failure recovery products and spare products used to recover from the failure identified by the failure identification unit 227. Alternatively, the failure recovery product output unit 229 may output the presence or absence and storage location of failure recovery products and spare products used to recover from the failure to the display unit 24 described below.

The required travel time output unit 230 refers to the fault characteristics database 213 and outputs the required travel time to the location of the fault identified by the fault identification unit 227 to the terminal device 40 via the communication unit 23. Alternatively, the required travel time output unit 230 may output the required travel time to the display unit 24 described below. At this time, for example, GPS information acquired by a GPS information acquisition unit (not shown) provided in the terminal device 40 may be used to calculate the required travel time.

The expected recovery time output unit 231 refers to the fault characteristics database 213 and outputs the expected recovery time from the fault identified by the fault identification unit 227 to the terminal device 40 via the communication unit 23. Alternatively, the expected recovery time output unit 231 may output the expected recovery time to the display unit 24 described below.

The communication unit 23 is a communication interface used by the fault analysis device 20 to communicate with the machine learning device 10, the monitoring device 30, and the terminal device 40.

The display unit 24 is a monitor that displays, for example, fault candidates output by the fault candidate output unit 225, confirmation items output by the confirmation item output unit 226, priorities output by the priority output unit 228, the presence and storage location of fault recovery items and spare items output by the fault recovery item output unit 229, the required travel time output by the required travel time output unit 230, and the expected recovery time output by the expected recovery time output unit 231.

2. Operation of the embodiment
Next, the operation of the fault analysis system 100 according to the present embodiment will be described. First, the operation of the machine learning device 10 will be described with reference to FIGS.

2.1 Operation of the Machine Learning Device 10
FIG. 4 is a flowchart showing the operation of the first learning unit 11.
In step S11, the first input data acquisition unit 111 acquires information related to an alarm occurring in the monitoring device 30 of the NMS as first input data.

In step S12, the first label acquisition unit 112 acquires the fault determination result corresponding to the alarm as the first label.

In step S13, the first learning model construction unit 113 sets the pair of the first input data and the first label as teacher data.

In step S14, the first learning model construction unit 113 performs supervised learning using the teacher data from step S13.

If machine learning has ended in step S15 (S15: YES), the process proceeds to step S16. If machine learning has not yet ended (S15: NO), the process proceeds to step S11.

In step S16, the first learning model construction unit 113 transmits the constructed first learning model to the fault analysis device 20. Then, all processing is terminated.

FIG. 5 is a flowchart showing the operation of the second learning unit 12.
In step S21, the second input data acquisition unit 121 acquires, as second input data, information related to an output waveform from a microwave radio provided in the NMS.

In step S22, the second label acquisition unit 122 acquires the fault determination result corresponding to the output waveform as a second label.

In step S23, the second learning model construction unit 123 sets the pair of the second input data and the second label as teacher data.

In step S24, the second learning model construction unit 123 performs supervised learning using the teacher data from step S23.

If machine learning has ended in step S25 (S25: YES), the process proceeds to step S26. If machine learning has not yet ended (S25: NO), the process proceeds to step S21.

In step S26, the second learning model construction unit 123 transmits the constructed second learning model to the fault analysis device 20. Then, all processing is terminated.

2.2 Operation of the Fault Analysis Device 20
Next, the operation of the failure analysis device 20 will be described with reference to FIG.

In step S31, the alarm acquisition unit 221 acquires a new alarm from the monitoring device 30 via the communication unit 23 described below, separately from the alarms acquired when constructing the first learning model and the second learning model.

In step S32, the alarm distribution unit 222 refers to the pre-registration database 211 to determine whether the alarm acquired by the alarm acquisition unit 221 is a pre-registered work alarm or a fault alarm that has not been pre-registered and is being output as a result of an actual fault occurring.

In step S33, if the fault alarm log does not contain data indicating that the output from the microwave radio has exceeded the threshold (S33: NO), processing proceeds to step S34. If the fault alarm log contains data indicating that the output from the microwave radio has exceeded the threshold (S33: YES), processing proceeds to step S35.

In step S34, the first fault determination unit 223 applies the fault alarm that has not been registered in advance to the first learning model stored in the memory unit 21 to infer and determine the fault that may have occurred and its probability of occurrence.

In step S35, the second fault determination unit 224 infers and determines a fault that may have occurred by applying the output waveform from the microwave radio to the second learning model stored in the memory unit 21. Furthermore, the second fault determination unit 224 outputs the fault that may have occurred.

In step S36, the fault candidate output unit 225 sorts the fault candidates determined by the first fault determination unit 223 in descending order of probability if there are multiple candidates, and outputs them together with the probability.

In step S37, if the accuracy of the fault candidate output by the fault candidate output unit 225 is 100% (S37: YES), the process proceeds to step S39. Otherwise (S37: NO), the process proceeds to step S38.

In step S38, the confirmation item output unit 226 outputs confirmation items for the user of the fault analysis device 20 by referring to the alarm fault response database 212.

In step S39, the fault identification unit 227 identifies the fault that has actually occurred.

In step S40, the priority output unit 228 outputs the priority of responding to the fault identified by the fault identification unit 227 by referring to the fault characteristics database 213.

In step S41, the failure recovery product output unit 229 references the failure characteristics database 213 to output the presence or absence and storage location of failure recovery products and spare products to be used to recover from the failure identified by the failure identification unit 227.

In step S42, the required travel time output unit 230 outputs the required travel time to the location of the fault identified by the fault identification unit 227 by referring to the fault characteristics database 213.

In step S43, the expected recovery time output unit 231 outputs the expected recovery time from the failure identified by the failure identification unit 227 by referring to the failure characteristics database 213. Then, all processing is terminated.

[3 Modifications]
Although the embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment. Furthermore, the effects described in the present embodiment are merely a list of the most preferable effects resulting from the present invention, and the effects of the present invention are not limited to those described in the present embodiment.

3.1 Modification 1
In the above embodiment, the machine learning device 10 and the fault analysis device 20 are shown as separate entities, but the present invention is not limited to this. For example, the machine learning device 10 may be incorporated into the housing of the fault analysis device 20, and thus the two may be realized as an integrated unit.

3.2 Modification 2
In the above embodiment, the first fault determination unit 223 infers and determines a fault that may have occurred and the probability of the fault, but this is not limited to the above. For example, in addition to the first fault determination unit 223, the second fault determination unit 224 may also infer and determine the probability of the fault that may have occurred. Furthermore, the second fault determination unit 224 may output a plurality of fault candidates after sorting them by the probability.

3.3 Modification 3
In the above embodiment, the confirmation item output unit 226 outputs confirmation items for the user, but is not limited to this. For example, in addition to simply outputting confirmation items, the confirmation results from the user may be input to the fault analysis device 20, and the probability corresponding to the fault candidate may be recalculated based on the input result.

The fault analysis method by the fault analysis system 100 is realized by software. When realized by software, the programs that make up this software are installed on a computer. These programs may also be recorded on removable media and distributed to users, or may be distributed by being downloaded to the user's computer via a network.

LIST OF SYMBOLS 10 Machine learning device 11 First learning unit 12 Second learning unit 20 Fault analysis device 21 Memory unit 22 Control unit 23 Communication unit 24 Display unit 30 Monitoring device 40 Terminal device 50 Network 100 Fault analysis system 111 First input data acquisition unit 112 First label acquisition unit 113 First learning model construction unit 114 First learning model storage unit 121 Second input data acquisition unit 122 Second label acquisition unit 123 Second learning model construction unit 124 Second learning model storage unit 211 Pre-registration database 212 Alarm fault response database 213 Fault characteristic database 221 Alarm acquisition unit 222 Alarm distribution unit 223 First fault determination unit 224 Second fault determination unit 225 Fault candidate output unit 226 Confirmation item output unit 227 Fault identification unit 228 Priority output unit 229 Fault recovery product output section 230 Required travel time output section

Claims (4)

A machine learning device that constructs a learning model for identifying faults in communication facilities or line facilities from facility information acquired through a network,
a first input data acquisition means for acquiring information related to an alarm occurring in a monitoring device of the network as the facility information as first input data;
a first label acquisition means for acquiring a result of a fault determination corresponding to the alarm as a first label;
a first learning model construction means for constructing a first learning model for generating the determination result corresponding to an alarm occurring in the monitoring device by performing supervised learning using a pair of the first input data and the first label as teacher data;
Equipped with
second input data acquiring means for acquiring information related to an output waveform from a microwave radio device serving as the communication equipment as second input data;
a second label acquiring means for acquiring a fault determination result corresponding to the output waveform as a second label;
a second learning model constructing means for constructing a second learning model for generating the determination result corresponding to the output waveform by performing supervised learning using a pair of the second input data and the second label as teacher data;
The machine learning device further comprises : A fault analysis device using the first learning model constructed by the machine learning device according to claim 1 ,
an alarm acquisition means for acquiring the new alarm;
an alarm sorting means for sorting the alarms acquired by the alarm acquisition means into alarms that have been registered in advance and alarms that have not been registered in advance;
a first fault determination means for determining the occurrence of the fault by using the alarm not registered in advance and the first learning model;
A failure analysis device comprising: The first fault determination means determines the probability of occurrence of the fault in addition to the fault,
3. The fault analysis device according to claim 2 , further comprising a check item output unit that outputs a check item for increasing the probability of the fault occurring when the probability of the fault occurring is not 100%. A fault analysis device using the second learning model constructed by the machine learning device according to claim 1 ,
a second fault determination means for determining the occurrence of a fault using the output waveform and the second learning model when the log of the alarm not registered in advance indicates that the output from the microwave radio device exceeds a threshold value;
Equipped with
The fault analysis device according to claim 2 or claim 3, wherein the first fault determination means determines the occurrence of the fault using the alarm that is not pre-registered and the first learning model when the log of the alarm that is not pre-registered does not indicate that the output exceeds a threshold value.

Images