KR20240043640A - Method and apparatus for predictive maintenance of network transmission equipment - Google Patents

Abstract

A predictive maintenance method performed in a predictive maintenance device, wherein a network transmission device is configured and a plurality of data units and a plurality of switching units are interconnected to transmit and receive data frames through a plurality of buses, and the plurality of buses are connected in a bus direction. A step of periodically collecting bus status information for each bus direction from the network transmission device, dividing each collected bus status information into each of the plurality of data units and the plurality of switching units, and It includes monitoring abnormal signs on a data unit and switching unit basis through a predictive maintenance model that uses bus status information as input data.

Description

Predictive maintenance method and device for network transmission device {METHOD AND APPARATUS FOR PREDICTIVE MAINTENANCE OF NETWORK TRANSMISSION EQUIPMENT}

This disclosure relates to a predictive maintenance method and device for an MSPP (Multi-Service Provisioning Platform) device, which is a network transmission device.

The MSPP (Multi-Service Provisioning Platform) device, a network transmission device, is a very important device that connects customer servers that handle mission-critical tasks such as finance. Therefore, failure of the MSPP device causes high ripple effects, so high stability is required.

Previously, when an MSPP device malfunctioned, maintenance was carried out in a follow-up manner. However, MSPP devices have many connected devices and play an important role, so they require high safety. Therefore, the MSPP device needs to repair the equipment in advance by detecting abnormalities before a failure occurs.

Predictive Maintenance (PdM) refers to a method of equipment maintenance that analyzes data to identify abnormal signs in advance and take action. Predictive maintenance-type maintenance has recently been attracting attention in smart factories for the purpose of improving over-maintenance of equipment and increasing stability.

Most existing predictive maintenance studies deal with data in a form suitable for factory equipment using data sets related to bearings, batteries, engines, etc., and research on transmission equipment is lacking. In other words, previous studies applied predictive maintenance by focusing only on analog sensors coming from one target equipment.

However, network transmission devices such as MSPP devices have different characteristics from existing facilities, so it is necessary to consider their unique characteristics. For example, in a network transmission device, the values according to the internal bus state are related to equipment failure, and the input/output values of the values according to the internal bus state are affected by the failure of other connected network transmission devices.

Additionally, network transmission devices such as MSPP devices are not devices that break down frequently as safety is important. Because it is difficult to collect actual failure data, it is difficult to apply remaining life prediction methods that require a lot of failure data.

Additionally, there is a conventional method that utilizes NMS (Network Management System) data to determine when a device breaks down and repair the device after the breakage occurs.

However, in this method, the service safety of the MSPP device is damaged, and if the failure continues for a certain period of time, the telecommunication company that fails to provide customer service stably is responsible for compensation. In addition, there is a problem that trust in customer companies decreases and there is a possibility of defection from customer companies that must ensure high safety.

The present disclosure provides a method and device for applying predictive maintenance to an MSPP device, which is a network transmission device.

The present disclosure sets the switching unit and data unit as a fault repair unit among the elements constituting the MSPP device, and uses the bus status and connection status with other devices of each switching unit and data unit as learning data to develop an unsupervised learning-based Provides a predictive maintenance method and device for generating an anomaly detection model and detecting anomalies in units of MSPP devices using the anomaly detection model.

According to one feature, a predictive maintenance method performed in a predictive maintenance device, comprising a plurality of data units and a plurality of switching units interconnected to configure a network transmission device and transmit and receive data frames through a plurality of buses, the plurality of switching units Classifying buses by bus direction and periodically collecting bus status information for each bus direction from the network transmission device, and dividing each collected bus status information into the plurality of data units and the plurality of switching units, , It includes the step of monitoring abnormal signs in units of data units and switching units through a predictive maintenance model that uses the status information of each classified bus as input data.

The monitoring step includes using the status information of each bus as input data of a previously learned predictive maintenance model to obtain an abnormality score indicating an abnormality for the corresponding bus from the predictive maintenance model, for each bus Calculating predictive maintenance values in data unit and switching unit units through a weighted average calculation of the obtained anomaly scores, and comparing the predictive maintenance values in data unit and switching unit units with a preset threshold to output a visualization screen. It may include steps.

The preset threshold may be set differentially for each data unit and each switching unit.

Each of the bus status information may include a registry value that is a signaling level value for the bus, and a flag value that indicates an error state of a data frame passing through the bus.

Each of the bus state information may additionally include a connection topology value if it is bus state information between a data unit and a switching unit.

The connection topology value may be set as a flag value between a switching unit and a data unit constituting another connected network transmission device.

Each of the bus status information may include a plurality of first bus status information for a plurality of first buses heading from each data unit to a plurality of switching units, and a plurality of second bus status information for a plurality of second buses heading from each switching unit to the plurality of data units. A plurality of second bus state information, and a plurality of third bus state information for a plurality of third buses directed from the high-speed switching module included in the switching unit to the low-speed switching module, and a plurality of third bus state information directed from the low-speed switching module to the high-speed switching module. It may include a plurality of fourth bus status information for the fourth bus.

The step of obtaining the anomaly score includes obtaining an anomaly score for each bus from the predictive maintenance model using the plurality of first bus state information as input data, and calculating the predictive maintenance value includes data unit The predictive maintenance value for the data unit can be calculated by performing a weighted average calculation on the abnormality scores obtained for the plurality of first buses.

The step of obtaining the abnormality score includes an abnormality for each bus from the predictive maintenance model using the plurality of second bus state information, the plurality of third bus state information, and the plurality of fourth bus states as input data. The step of obtaining a score and calculating the predictive maintenance value includes calculating a weighted average of the abnormality scores obtained for the plurality of second buses, the plurality of third buses, and the plurality of fourth buses of the switching unit, and calculating the predictive maintenance value to the switching unit. Predictive conservation values can be calculated.

The predictive maintenance model may use different unsupervised learning-based models for the switching unit and the data unit.

In the collecting step, bus status information for each bus direction may be collected in window units, which are preset time units.

According to another feature, the predictive maintenance system includes a network transmission device including a plurality of data units and a plurality of switching units interconnected to transmit and receive data frames through a plurality of buses, and from the network transmission device to the plurality of buses. For each bus direction, a plurality of bus state information corresponding to each bus direction is periodically collected, the collected plurality of state information is divided into the plurality of data units and the plurality of switching units, and each divided bus state information is input data. It includes a predictive maintenance device that monitors abnormal signs on a data unit and switching unit basis through a predictive maintenance model used as a .

The predictive maintenance device calculates an abnormality score for each bus by inputting the status information of each bus into a predictive maintenance model, and sums the abnormality scores for each bus to calculate the predictive maintenance value in data unit and switching unit units. Calculate and output a screen visualizing the predictive maintenance value relative to the threshold value.

The plurality of data units include a plurality of bus registers configured for each bus unit that store status information of a plurality of buses heading from the plurality of data units to the plurality of switching units, and monitor status information for the plurality of buses. and a monitor process for storing the monitored state information in the bus registers, and the plurality of bus state information stored in the plurality of bus registers may be periodically transmitted to the predictive maintenance device.

The plurality of data units include a plurality of first bus registers each bus unit storing status information of a plurality of buses heading from the plurality of switching units to the plurality of data units, an internal high-speed switching module, and a low-speed switching module. A plurality of second bus registers composed of each bus unit that store status information of a plurality of buses for each bus direction between the buses, monitor status information for the plurality of buses, and transmit the monitored status information to the plurality of first bus registers, and and a monitor process for storing each of the plurality of second bus registers, wherein the plurality of bus status information respectively stored in the plurality of first bus registers and the plurality of second bus registers are periodically transmitted to the predictive maintenance device. It can be.

According to another feature, the predictive maintenance device includes a data collection unit that periodically collects a plurality of bus status information registered in a plurality of bus registers included in the network transmission device from the network transmission device, and the plurality of bus state information. A predictive maintenance performing unit that calculates an anomaly score indicating an abnormality using input data of a predictive maintenance model and generates a predictive maintenance value for the network transmission device through a weighted average calculation of the calculated anomaly score, and the predictive maintenance It includes a predictive maintenance result output unit that compares the value with the threshold value and outputs visualization data on the screen.

The plurality of bus status information may include a registry value that is a signaling level value for the bus and a flag value that indicates an error state of a data frame passing through the bus.

The network transmission device includes a plurality of data units that process data frames and a plurality of switching units that set a transmission path for the data frames, and the plurality of data units and the plurality of switching units connect a plurality of buses. They are connected to each other through, and the plurality of bus status information may be composed of a plurality of pieces for each bus direction.

The predictive maintenance performance unit calculates an anomaly score for each bus from the predictive maintenance model using a plurality of bus status information from the data unit to the switching unit as input data, and calculates the anomaly score for each bus by a weighted average. One value may be calculated as the predictive maintenance value of the data unit, and visualization data that compares the calculated predictive maintenance value with a first threshold set for the data unit may be generated and output to the predictive maintenance result output unit.

The predictive maintenance performance unit uses the plurality of bus state information from the switching unit to the data unit and the plurality of bus state information between the internal switching modules of the switching unit as input data to calculate an abnormality score for each bus from the predictive maintenance model. Visualization data is calculated by calculating the weighted average of the abnormality scores for each bus as the predictive maintenance value of the switching unit, and comparing the second threshold set for the switching unit with the calculated predictive maintenance value. Can be generated and output to the predictive maintenance result output unit.

According to an embodiment, the switching unit and data unit constituting the MSPP device are set as a fault repair unit, and abnormal signs of the MSPP device are detected using topology information, which is the bus status of the switching unit and data unit and the connection state with other devices. Detection enables predictive maintenance that takes advantage of the unique characteristics of MSPP devices and takes into account interactions with other equipment. Additionally, since the fault repair unit is set to a switching unit and a data unit, general-purpose design is possible and detailed monitoring is possible.

In addition, when an abnormality is detected through the bus status, which is the internal data of the MSPP device, a maintenance method that takes proactive measures is used, thereby ensuring high service stability of the MSPP device. This solves problems in the event of a service failure and provides the advantage of high stability compared to other telecommunication companies that do not use predictive maintenance for MSPP devices.

1 is a configuration diagram of a predictive maintenance system according to an embodiment.
Figure 2 is an internal configuration diagram of an MSPP device according to an embodiment.
Figure 3 is an internal configuration diagram of a predictive maintenance device according to an embodiment.
Figure 4 illustrates the correlation between the bus state value of the MSPP device and the occurrence of MSPP failure according to one embodiment.
Figure 5 illustrates the concept of calculating a PdM score for an individual unit according to one embodiment.
Figure 6 is an exemplary diagram of a predictive maintenance result screen according to an embodiment.
Figure 7 is an exemplary configuration diagram of MSPP data according to an embodiment.
Figure 8 is an exemplary configuration diagram of an individual register of a data unit according to an embodiment.
9 is an exemplary configuration diagram of an individual register of a switching unit according to an embodiment.
Figure 10 is a flowchart showing a predictive maintenance procedure according to an embodiment.
Figure 11 is a block diagram showing the hardware configuration of a predictive maintenance device according to an embodiment.

Below, with reference to the attached drawings, embodiments of the present disclosure will be described in detail so that those skilled in the art can easily practice them. However, the present disclosure may be implemented in many different forms and is not limited to the embodiments described herein. In order to clearly explain the present disclosure in the drawings, parts that are not related to the description are omitted, and similar parts are given similar reference numerals throughout the specification.

Throughout the specification, when a part is said to “include” a certain element, this means that it may further include other elements rather than excluding other elements, unless specifically stated to the contrary.

In addition, terms such as “…unit”, “…unit”, and “…module” used in the specification refer to a unit that processes at least one function or operation, which may be implemented through hardware or software or a combination of hardware and software. You can.

The devices described in the present invention are composed of hardware including at least one processor, a memory device, a communication device, etc., and a program that is executed in conjunction with the hardware is stored in a designated location. The hardware has a configuration and performance capable of executing the method of the present invention. The program includes instructions that implement the operating method of the present invention described with reference to the drawings, and executes the present invention by combining it with hardware such as a processor and memory device.

In this specification, “transmission or provision” may include not only direct transmission or provision, but also indirect transmission or provision through another device or by using a circuitous route.

In this specification, expressions described as singular may be interpreted as singular or plural, unless explicit expressions such as “one” or “single” are used.

In this specification, the same reference numbers refer to the same elements regardless of the drawings, and “and/or” includes each and all combinations of one or more of the mentioned elements.

In this specification, terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but the components are not limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, a first component may be referred to as a second component, and similarly, the second component may be referred to as a first component without departing from the scope of the present disclosure.

In the flowcharts described herein with reference to the drawings, the order of operations may be changed, several operations may be merged, certain operations may be divided, and certain operations may not be performed.

In this specification, an artificial intelligence model (AI model) is a machine learning model that learns at least one task, and may be implemented as a computer program executed by a processor. The task that an artificial intelligence model learns can refer to a task to be solved through machine learning or a task to be performed through machine learning. Artificial intelligence models can be implemented as computer programs running on computing devices, downloaded over a network, or sold in product form. Alternatively, the artificial intelligence model can link with various devices through the network.

In this specification, predictive maintenance (hereinafter collectively referred to as 'PdM') refers to a maintenance method that analyzes equipment data to identify abnormal signs in advance and take action.

In this specification, the MSPP (Multiservice Provisioning Platform) device is an optical transport device used in the dedicated line business of a communication service provider. The MSPP device encompasses network transmission devices that can accommodate various services, and is referred to as the MSPP device in embodiments of the present invention for unification of terminology.

Hereinafter, a configuration for applying predictive maintenance according to an embodiment of the present invention to an MSPP device will be described.

FIG. 1 is a configuration diagram of a predictive maintenance system according to an embodiment, FIG. 2 is an internal configuration diagram of an MSPP device according to an embodiment, FIG. 3 is an internal configuration diagram of a predictive maintenance device according to an embodiment, and FIG. 4 illustrates the correlation between the bus state value of the MSPP device and the occurrence of MSPP failure according to an embodiment, FIG. 5 illustrates the concept of calculating a PdM score for an individual unit according to an embodiment, and FIG. 6 shows a This is an example of a predictive maintenance result screen according to an embodiment.

Referring to FIG. 1, the predictive maintenance system may include a plurality of MSPP devices 100, a communication network management system 200, and a predictive maintenance device 300.

The plurality of MSPP devices 100 include a plurality of buses that serve as paths for data frames and a plurality of bus registers in which status information of each bus is stored.

The communication network management system 200 may periodically collect status information of each bus from the plurality of MSPP devices 100 and transmit it to the predictive maintenance device 300.

The communication network management system 200 may be an Element Management System (EMS) or a Network Management System (NMS).

The predictive maintenance device 300 may use the collected bus status information to perform predictive maintenance to predict abnormal signs, that is, the possibility of failure, of each MSPP device 100.

The predictive maintenance device 300 is a computing device that operates by at least one processor.

The predictive maintenance device 300 can calculate an anomaly score (AS) from input data using an artificial intelligence (AI) model. Anomaly Score (AS) represents a measure of the likelihood of failure or indication of anomaly.

The predictive maintenance device 300 includes a predictive maintenance model. A predictive conservation model is an artificial intelligence model trained to calculate anomaly scores (AS) from input data. Input data is a predictor that affects the calculation of anomaly score (AS) and includes bus status information.

At this time, the predictive maintenance device 300 is shown separately from the communication network management system 200, but depending on the embodiment, the communication network management system 200 may perform the operation of the predictive maintenance device 300.

A plurality of MSPP devices 100 may be interconnected. Any MSPP device 100 may be connected to at least one neighboring MSPP device 100.

At this time, the MSPP device that serves as the basis for explanation is described as MSPP device #1, and the neighboring devices of MSPP device #1 are described as MSPP device #2 and MSPP device #3, respectively, and their reference numerals are the same as '100'. It is given as

Each MSPP device 100 is composed of a plurality of slots, and each slot is equipped with one of a plurality of data units 110, a plurality of switching units 120, and a plurality of control units 130.

The data unit 110 serves to receive customer service and relay data frames. The data unit 110 receives data frames from the subscriber device or outputs data frames to the subscriber device. Here, the data frame may be a Synchronous Digital Hierarchy (SDH) frame, a Plesiochronous Digital Hierarchy (PDH) frame, an Ethernet frame, etc.

The switching unit 120 serves to transmit the data frame to a path matching the destination. That is, the switching unit 120 sets the transmission direction of the data frame so that the data frame can arrive at the destination and transmits the data frame in the set transmission direction. The switching unit 120 may transmit data frames and information to other external devices, that is, the neighboring MSPP device 100 and the communication network equipment management system 200.

The control unit 130 performs overall control operations of the MSPP device 100. At this time, since the control unit 130 is not subject to predictive maintenance, detailed description will be omitted.

Referring to FIG. 2, the MSPP device 100 may include a plurality of data units (U) 110, a plurality of switching units 120, and a system interlocking unit 130. At this time, the configuration in FIG. 2 briefly shows the configuration related to predictive maintenance.

The plurality of data units (U) 110 may be composed of n individual data units, that is, U ₁ , ..., U _N .

The plurality of switching (S) 120 may have a dual structure and may be composed of, for example, S _A and S _B.

Referring to FIGS. 1 and 2, the MSPP device 100 may be equipped with n data units (U ₁ , ..., U _N ) 110 in n slots, respectively. The MSPP device 100 may be equipped with two switching units (S _A , S _B ) 120 in two slots, respectively. The MSPP device 100 may be equipped with two control units 130 in each of two slots.

In the present invention, the units to which predictive maintenance is applied are n data units (U ₁ , ..., U _N ) (110) and two switching units (S _A , S _B ) (120).

A plurality of data units 110 and a plurality of switching units 120 are connected to a plurality of unidirectional buses. For a unidirectional bus, the direction of the bus is determined depending on the direction of data flow. Basically, a bus refers to the path of data frames.

The plurality of data units 110 may be divided into service units and subscriber units and may be configured in plural numbers. At this time, the plurality of buses may be divided into subscribers within the data unit 110 providing the same service, or may be divided into service units within the data unit 110 implemented on a subscriber basis.

The plurality of switching units 120 may have a redundant structure.

The number of unidirectional buses is determined depending on the unit type of the data unit 110.

The data unit 110 may have different names, different unit types, and different numbers of buses depending on capacity.

For example, the data unit 110 may have a data unit name of S16U, and in this case, the number of buses is four and can perform service acceptance and data transmission of STM (Synchronous Transport Module)-16 signals. At this time, there are two buses heading from the data unit 110 to the switching unit 120 and two buses heading from the switching unit 120 to the data unit 110.

Additionally, the data unit 110 may have a data unit name of ET3U, and in this case, the number of buses is two and can accommodate Ethernet services and perform data transmission. At this time, there is one bus heading from the data unit 110 to the switching unit 120 and one bus heading from the switching unit 120 to the data unit 110.

Additionally, the plurality of data units 110 may be composed of n data units (U ₁ , ..., U _N ) 110 having different data unit names, that is, different numbers of buses. For example, U ₁ may be S16U and U ₂ may be ET3U.

A plurality of data units 110 and a plurality of switching units 120 are configured onboard. Onboard, the CPU (central processing unit), memory, buffer, and I/O input/output modules are mounted on a single PCB (Printed Circuit Board) board.

The CPU may include a processor and a control unit. Of course, the CPU can be an MPU (Micro Processor Unit), MCU (Microcontroller Unit), etc.

The processor may include a plurality of bus registers and a monitor process. Bus registers store bus status information. The monitor process monitors the status of the bus, generates bus status information, and stores it in the corresponding bus register.

At this time, a plurality of bus registers are matched 1:1 with a plurality of buses. For example, if there are N buses, there will also be N bus registers.

U ₁ consists of a monitor process and a plurality of bus registers (r ₁₁ , ..., r _1M ), and the plurality of bus registers (r ₁₁ , ..., r _1M ) are buses heading from U ₁ to S _A. and U ₁ are created as many as the number of buses heading to S _B. U _N consists of a monitor process and a plurality of bus registers (r _N1 , ..., r _NM ), and a plurality of bus registers (r ₁₁ , ..., r _1M ) are buses heading from U _N to S _A. and are created as many as the number of buses heading from U _N to S _B.

At this time, to avoid complexity of the drawing, the bus is expressed only for U ₁ .

Unlike the data unit 110, the processor of the switching unit 120 additionally includes switching modules (HO, LO). The switching module consists of a high order (HO) switching module and a low order (LO) switching module. The HO switching module performs high-speed switching (320G). The LO switching module performs low-speed switching (10G).

S _A includes a monitor process, multiple registers, and switching modules (HO, LO). S _B includes a monitor process, multiple registers, and switching modules (HO, LO).

At this time, a plurality of registers in _SA are created as many as the number of buses heading from _SA to each data unit (U ₁ , ..., U _N ) (110) and buses between the HO switching module and the LO switching module. A plurality of registers in S _B are created as many as the number of buses heading from S _B to each data unit (U ₁ , ..., U _N ) (110) and buses between the HO switching module and the LO switching module. At this time, 16 buses are usually used between the HO switching module and the LO switching module, but this number is not limited.

The above-mentioned bus registers store status information of the matching bus, and the bus status information includes registry values and flag values.

The registry value is the signaling level value of the bus and is a 4-digit HEX value (eg, 0x0035). The flag value is a value indicating the error state of a data frame passing through the bus, and its value is defined in advance. For example, when the BIP (Bit Interleaved Parity) error in a data frame is greater than a certain threshold, the flag value is recorded as 'BIP8_ERROR'. If loss occurs in the data frame itself above a certain threshold, the flag value is recorded as 'OOF'.

The system interlocking unit 130 is connected to a plurality of data units (U ₁ , ..., U _N ) 110 and a plurality of switching units (S _A , S _B ) 120, and is connected to a communication network management system 200. is connected to

The system interlocking unit 130 receives each bus status information from the bus registers of each of the plurality of data units (U ₁ , ..., U _N ) 110 and the plurality of switching units (S _A , S _B ) 120. It can read periodically, process the read bus status information into MSPP data in a predetermined format, and transmit it to the communication network management system 200.

At this time, the read/transmit cycle of bus status information can be expressed as a collection cycle, and the time unit is called window (W ₁ , ... W _N ).

The window (W ₁ , ... W _N ) is set in advance by considering various factors such as the unit in which the failure can be handled and the point in time at which it can be captured. For example, it can be set in 30-minute increments, 1-hour increments, etc., and is set by the operator. The window is set by the operator because the time available to respond to a fault varies.

Referring again to FIG. 1, the communication network management system 200 stores in advance device connection configuration information between MSPP devices 100 and bus connection configuration information between internal units of individual MSPP devices 100. This device connection configuration information and bus connection configuration information is input by the operator. It may be directly input and stored in the communication network management system 200, or it can be input to each MSPP device 100 and retrieved from each MSPP device 100. It may also be transmitted to and stored in the communication network management system 200.

Additionally, the communication network management system 200 may transmit bus status information periodically collected for each MSPP device 100 to the predictive maintenance device 300.

A plurality of MSPP devices 100 are connected to the communication network management system 200 and periodically transmit their internal bus status information to the communication network management system 200.

The predictive maintenance device 300 uses bus status information as learning data to create a predictive maintenance model, and inputs periodically collected bus status information into the predictive maintenance model to calculate an anomaly score (AS) for each cycle. can do.

Referring to FIG. 3, the predictive maintenance device 300 is a computer device operated by at least one processor and may be a server device equipped with a monitor. The predictive maintenance device 300 may be a control server used by the operator.

The predictive maintenance device 300 may include a data collection unit 310, a data processing unit 320, a predictive maintenance performing unit 330, and a predictive maintenance result providing unit 340.

The data collection unit 310 is connected to the communication network management system 200 and can collect bus status information of each MSPP 100 transmitted from the communication network management system 200 through each system interlocking unit 130.

The data processing unit 320 performs data reconstruction by converting the bus state information collected by the data collection unit 310 into a predetermined format that can be used by the predictive maintenance performing unit 330.

Status information on buses designed for each MSPP 100 is transmitted as MSPP data in the form of a text file. The data processing unit 320 may convert such MSPP data into a designated format, for example, an Excel file. The Excel file may have items such as window, slot name, unit type, registry value, flag value, and connection topology value sorted by MSPP. These items are sorted by corresponding bus register. At this time, the registry value may be an analog data format, and the flag value/connection topology value may be a digital data format.

The predictive maintenance performance unit 330 calculates an anomaly score (AS) using the corresponding bus status information in individual bus register units, which can be defined as Equation 1.

[Equation 1]

That is, according to Equation 1, an abnormality score is calculated by using the registry value, flag value, and connection topology value as input data of the predictive maintenance model (Model).

In this way, by using bus status information such as registry values, flag values, and connection topology values as input data of the predictive maintenance model, it is possible to take into account the general characteristics of the MSPP device 100 in which characteristics of each equipment/chipset are shared.

Additionally, the flag value of another connected MSPP device is used as the connection topology value. Therefore, the connection topology value can detect abnormalities by considering factors caused by other MSPP devices connected to any MSPP device. This will be explained with reference to FIG. 8.

Since the MSPP device 100 is a transmission device, if a specific device malfunctions, other normal connected devices may also be recognized as malfunctioning. However, conventional methods deal with predictive maintenance for individual equipment and do not consider the relationships and characteristics between devices. However, in the present invention, the components of the MSPP device are divided into units and predictive maintenance is performed by considering the relationships and characteristics between them and the relationships with other MSPP devices. In addition, predictive maintenance is performed by considering the characteristics of the transmission equipment, namely the connection topology value.

Referring to FIG. 4, bus status values within Windows, that is, registry values, have time-series characteristics that change over time.

In the actual MSPP device 100, the bus state value of w ₄ , which is the section before w ₅ where the failure occurred, shows a different aspect from w ₁ to w ₃ . The abnormality score represents the bus state change value at w ₄ , and through this, abnormality symptoms can be detected.

At this time, when the predictive maintenance target is a data unit, the predictive maintenance model (Model _u ) is a registry read from the individual bus registers of the buses heading from the data unit (U) 110 to the switching unit (S) 120 in FIG. An anomaly score is calculated using the value, flag value, and connection topology value as input data.

In addition, when the predictive maintenance target is a data unit, the predictive maintenance model (Model _s ) is a registry read from the individual bus registers of the buses heading from the switching unit (S) 120 to the data unit (U) 110 in FIG. An abnormality score is calculated using the value, flag value, connection topology value, registry value read from the bus registers between the internal switching modules (HO↔LO) of the switching unit (S) 120, and flag value as input data. do.

At this time, since the first predictive maintenance model (Model _u ) and the second predictive maintenance model (Model _s ) have different input data, different types of models may be used.

The predictive conservation model calculates anomaly scores using an unsupervised learning method, and unsupervised learning methods can be used in a variety of ways.

According to one embodiment, STAT (statistics) may be used as an unsupervised learning method. According to this method, the 'change' in the value representing the bus state is defined as an abnormality score, and it can be determined whether the abnormality score exceeds the threshold. At this time, variance is used to define the variation within the window, and the average can be calculated using the history data collected in the first step and the variance can be calculated based on this.

According to another embodiment, machine learning (ML) may be used as an unsupervised learning method. For example, using Isolation forest, you can define an anomaly score based on the depth of the tree.

According to another embodiment, deep learning (DL) may be used as an unsupervised learning method. According to this method, reconstruction error can be obtained through an AE (AutoEncoder)-based model and defined as an abnormality score. At this time, AE can also be combined with CNN, and to model time, it can be combined with RNN (Recurrent Neural Network), LSTM (Long Short Term Memory), and GRU (Gated Recurrent Units). Alternatively, you can use a model such as DeepAnT, which detects anomalies through unsupervised learning by combining an anomaly detection module and a time series prediction module.

Referring to Figure 5, the abnormality score is calculated for each individual register. However, since the repair unit for the MSPP device 100 is a unit, the predictive maintenance value (PdM Score) must be calculated in unit units. To this end, the predictive maintenance performing unit 330 calculates a unit-level predictive maintenance value (PdM Score) using a weighted average value through Equation 2.

[Equation 2]

is a weight set by the operator in advance and is defined differently according to the characteristics of each equipment. At this time, if the importance of all individual bus registers is the same, is set to 1/n. In other words, if the weights are the same for each bus register, the predictive maintenance value (PdM Score) becomes the average value of the abnormal scores.

The predictive maintenance performance unit 330 uses Equation 2 based on abnormality scores derived through bus state information read from the bus registers of an arbitrary unit, for example, U ₁ (110) _. The predictive maintenance value (PdM score) can be calculated. The predictive maintenance result provider 340 may generate a predictive maintenance notification window that matches the predictive maintenance values in units and output the predictive maintenance notification window on the screen.

At this time, the predictive maintenance value can be displayed by comparing it with the threshold value set for each unit. For example, the unit-level predictive maintenance value can be visualized and displayed in the predictive maintenance notification window so that one can intuitively know whether the predictive maintenance value is larger or smaller based on the threshold.

The predictive maintenance value (PdM score) is used to determine whether the corresponding units 110 and 120 are abnormal according to a set threshold. The threshold can be arbitrarily set by the operator.

According to one embodiment, the threshold (Th) may be defined as Equation 3.

[Equation 3]

Here, IQR is an abbreviation for Inter Quartile Range and is equal to Q_3-Q_1. Q_3 refers to the tertile of predictive conservation value (PdM score). Q_1 refers to the 1st quartile of predictive conservation value (PdM score). At this time, the predictive conservation value (PdM score) for calculating the third and first quartiles can be calculated through the cumulative probability using the learning data used in the predictive conservation model learning step.

Referring to FIG. 6, the predictive maintenance result providing unit 340 can output a predictive maintenance notification window (P10) listing the connection structure between a plurality of MSPPs (MSPP ₁ , MSPP ₂ , MSPP ₃ , MSPP ₄ , and MSPP ₅ ). there is.

At this time, when any MSPP ₃ is clicked, the predictive maintenance result provider 340 provides the predictive maintenance results calculated for the plurality of units (U ₁ , ..., U _N , S _A , S _B ) constituting MSPP ₃ You can create and output a predictive maintenance notification window (P20) that visualizes the conservation value (PdM score) by comparing it with the threshold value of each unit. The predictive maintenance notification window (P20) can display the threshold value for each unit and the area where the predictive maintenance value (PdM score) of each unit is located. It can be visualized as a predictive maintenance value (PdM score) that is above a threshold, or a predictive maintenance value (PdM score) that is below a threshold.

The unit-level threshold can be defined in two types: data unit and switching unit (T _u , T _s ).

When using the threshold dynamically, the bus status values collected for each window can be stored in a separate history, and the threshold value can be dynamically changed and set based on the saved bus status values.

When a unit exceeding a set threshold value (T _u , T _s ) is discovered, the predictive maintenance performing unit 330 determines anomaly detection for the unit and visually/audibly generates an alarm indicating that maintenance is required. You can. In other words, after visualization considering the topology, an alarm can be generated to the equipment manager about the need for maintenance when the predictive maintenance value (PdM score) exceeds a predefined threshold.

The predictive maintenance performing unit 330, in conjunction with the predictive maintenance result providing unit 340, distinguishes the unit in which abnormalities are detected in the predictive maintenance alarm window (P20) from other units and performs visual functions such as changing the color or displaying a warning light. An alarm can be triggered. Additionally, the predictive maintenance result providing unit 340 may output a sound or voice notifying the detection of abnormalities. Through this, the abnormal situation of the entire MSPP device 100 and the units 110 and 120 constituting the device 100 can be viewed at a glance and the abnormal situation of the MSPP device can be easily identified through the process of checking the detailed score. there is.

Figure 7 is an exemplary configuration diagram of MSPP data according to an embodiment.

Referring to FIG. 7, it shows the configuration of MSPP data (P11, P12) generated in the data unit 110 of any MSPP #1 (100).

For example, the S16U type MSPP data (P11) mounted in the S01 slot has 4 unidirectional buses between the S16U type data unit (U ₁ ) mounted in the S01 slot and the switching unit (S _A ), so 4 bus lines This is created.

Information 1 displays status information of the bus heading from the data unit (U ₁ ) to the switching unit (S _A ).

Information 2 displays status information of the bus heading from the switching unit (S _A ) to the data unit (U ₁ ).

The ET3U type MSPP data (P12) mounted in the S02 slot creates two unidirectional buses because there are two unidirectional buses between the ET3U type data unit (U ₂ ) mounted in the S02 slot and the switching unit (S _A ). do.

Information 1 displays status information of the bus heading from the data unit (U ₂ ) to the switching unit (S _A ).

Information 2 displays status information of the bus heading from the switching unit (S _A ) to the data unit (U ₂ ).

At this time, the bus status information displayed in each MSPP data (P11, P12) is a registry value and a flag value.

Figure 8 is an exemplary configuration diagram of an individual register of a data unit according to an embodiment.

Referring to FIG. 8, the MSPP data (P21, P22) of MSPP #1 and MSPP #2 are the same as described in FIG. 7.

MSPP #1 and MSPP #2 are connected to each other. At this time, it is assumed that the switching units S _A and S _B are interconnected. For convenience of explanation, the symbols for switching units and data units of MSPP #1 and MSPP #2 are displayed differently.

U ₁ is a bus heading to S _A ( ) includes a register that stores status information. a register is the bus ( ) registry value, bus ( ) stores the flag values of the data frames passing through, and these values are written to the corresponding bus line (1 of information 1) of the MSPP #1 data (P21). And, the connection topology value is the flag value of the register included in S _B connected to S _A , that is, the d register. This is a bus ( ) is a data frame passing through the bus ( ), so the bus ( )'s status is the bus ( This is because it is related to the state of ).

S _A is the bus heading to U ₁ ( ) and a b register that stores status information. The b register is the bus ( ) registry value, bus ( ) stores the flag values of the data frames passing through, and these values are written to the corresponding bus line (1 of information 2) of the MSPP #1 data (P21). And, the connection topology value is the bus flowing into S _B connected to S _A ( ) The flag value of the c register, which is a register of ), is used. This is a bus ( ) is a data frame passing through the bus ( ), so the bus ( )'s status is the bus ( This is because it is related to the state of ).

U ₂ is a bus heading to S _B ( ) includes the c register, which stores status information. The c register is the bus ( ) registry value, bus ( ) stores the flag values of the data frames passing through, and these values are written to the corresponding bus line (1 of information 1) of the MSPP #2 data (P22). And, the connection topology value is the flag value of the register included in S _A connected to S _B , that is, the b register. This is a bus ( ) is a data frame passing through the bus ( ), so the bus ( )'s status is the bus ( This is because it is related to the state of ).

S _B is the bus heading to U ₂ ( ) and a d register that stores status information. The d register is the bus ( ) registry value, bus ( ) stores the flag values of the data frames passing through, and these values are written to the corresponding bus line (1 of information 2) of the MSPP #2 data (P22). And, the connection topology value is the bus flowing into S _A connected to S _B ( ) The flag value of the a register, which is a register of ), is used. This is a bus ( ) is a data frame passing through the bus ( ), so the bus ( )'s status is the bus ( This is because it is related to the state of ).

At this time, as shown in FIG. 8, information such as which MSPPs are connected to each other and through which buses within the MSPP the flows of different MSPPs are connected to each other are device connection configuration information and bus connection configuration information, which are previously It is stored in the communication network management system 200. This information can be retrieved by the monitor process (FIG. 2) from the communication network management system 200, stored in advance, used, or viewed whenever necessary. The monitor process can use this information to obtain registry values, flag values, and connection topology values from individual registers and configure MSPP data to be transmitted to the communication network management system 200.

In this way, the registry values, flag values, and connection topology values stored in each register are used as input data for the predictive maintenance model to calculate the abnormality score of each unit (U ₁ , U ₂ , S _A , S _B ) for each MSPP. .

Figure 9 is an exemplary configuration diagram of an individual register of a switching unit according to an embodiment.

Referring to FIG. 9, the MSPP data (P31) of MSPP #1 is the same as described in FIG. 7.

MSPP data (P31) includes bus status information between HO and LO, which are internal modules of a random switching unit (S _A ). At this time, for convenience of explanation, the bus status information between HO and LO is displayed, but the MSPP data (P31) is integrated with the MSPP data (P11, P12, P21, P22) described in Figures 7 and 8 and is divided into MSPP units and within MSPP. It is divided into bus connection information between the data unit 110 and the switching unit 120 and bus connection information between HO and LO within the switching unit 120, and bus status information can be displayed for each corresponding bus line.

The bus between HO and LO can be composed of 16 buses in the HO → LO direction and 16 buses in the LO → HO direction. Accordingly, in the MSPP data (P31), bus lines are formed for each bus direction. Information 1 displays status information of the bus in the HO → LO direction. Information 2 displays the status information of the bus in the LO → HO direction.

The switching unit (S _A ) has a register configured for each bus, but the bus register where the bus status between HO and LO is stored does not have a connection topology value, but a registry value and a flag value.

In addition, Figure 9 is for explaining the bus state between HO and LO, and the bus register stored by the switching unit (S _A ) is directed to the plurality of data units (U ₁ , ..., U _N ) described in Figure 8. It includes a plurality of bus registers that store the bus state, and a plurality of bus registers that store the bus state for each direction between HO and LO described in FIG. 9.

This configuration of the switching unit (S _A ) is similarly applied to all other switching units (S _B ).

Meanwhile, as described in FIG. 8, the data units (U ₁ , ..., U _N ) include a plurality of bus registers that store the status of each bus heading to a plurality of switching units (S _A, S _B ). .

FIG. 10 is a flowchart showing a predictive maintenance procedure according to an embodiment, and sequentially shows the operation of the predictive maintenance device 300 based on the configuration described in FIGS. 1 to 9.

Referring to FIG. 10, the predictive maintenance device 300 sets the device connection configuration information and bus connection configuration information of the MSPP devices 100 that are subject to predictive maintenance (S101).

The predictive maintenance device 300 receives bus status information (registry value, flag value) of each data unit 110 and each switching unit 120 constituting the MSPP devices 100 from the MSPP devices 100. , connection topology values) are collected periodically (i.e., on a per-window basis) (S102).

The predictive maintenance device 300 divides the collected bus state information into predictive maintenance units, that is, unit units.

The predictive maintenance device 300 divides the bus state information of the plurality of data units 110 into a plurality of individual register units for each data unit 110, and stores the classified bus state information in the first predictive maintenance model (Model _u ). Input to calculate the abnormality score (AS) of each individual register (S103).

The predictive maintenance device 300 calculates the PdM score in units of data units 110 through the weighted average of the calculated abnormality scores (S104).

The predictive maintenance device 300 divides the bus state information of the plurality of switching units 120 into a plurality of individual register units for each switching unit 120, and stores the classified bus state information in the second predictive maintenance model (Model _s ). Input to calculate the abnormality score (AS) of each individual register (S105).

The predictive maintenance device 300 calculates the PdM score for each switching unit 120 through a weighted average of the calculated abnormality scores (S106).

The predictive maintenance device 300 may visualize the result of comparing the PdM score calculated for each unit with each threshold value classified for each unit and output it on the screen (S107).

Meanwhile, Figure 11 is a block diagram showing the hardware configuration of a predictive maintenance device according to an embodiment.

Referring to FIG. 11 , the predictive maintenance device 300 described in FIGS. 1 to 10 may be a computing device 400 operated by at least one processor.

The computing device 400 includes at least one processor 410, a memory 420 for loading a program executed by the processor 410, a storage 430 for storing programs and various data, a communication interface 440, and It may include a bus 450 connecting them. In addition, the computing device 400 may further include various components. When loaded into the memory 420, the program may include instructions that cause the processor 410 to perform methods/operations according to various embodiments of the present disclosure. That is, the processor 410 can perform methods/operations according to various embodiments of the present disclosure by executing instructions. Instructions are a series of computer-readable instructions grouped by function and are a component of a computer program and are executed by a processor.

The processor 410 controls the overall operation of each component of the computing device 400. The processor 410 includes at least one of a Central Processing Unit (CPU), Micro Processor Unit (MPU), Micro Controller Unit (MCU), Graphic Processing Unit (GPU), or any type of processor well known in the art of the present disclosure. It can be configured to include. Additionally, the processor 410 may perform operations on at least one application or program to execute methods/operations according to various embodiments of the present disclosure.

Additionally, the processor 410 may support virtualization services based on cloud computing technology. That is, the processor 410 can create and execute a virtual machine.

The memory 420 stores various data, commands and/or information. Memory 420 may load one or more programs from storage 430 to execute methods/operations according to various embodiments of the present disclosure. The memory 420 may be implemented as a volatile memory such as random access memory (RAM), but the technical scope of the present disclosure is not limited thereto.

Storage 430 may store programs non-temporarily. The storage 430 is a non-volatile memory such as Read Only Memory (ROM), Erasable Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), flash memory, a hard disk, a removable disk, or a device well known in the art to which this disclosure pertains. It may be configured to include any known type of computer-readable recording medium.

The communication interface 440 supports wired and wireless communication of the computing device 400. To this end, the communication interface 440 may be configured to include a communication module well known in the technical field of the present disclosure.

Bus 450 provides communication functionality between components of computing device 400.

The embodiments of the present disclosure described above are not only implemented through devices and methods, but may also be implemented through programs that implement functions corresponding to the configurations of the embodiments of the present disclosure or recording media on which the programs are recorded.

Although the embodiments of the present disclosure have been described in detail above, the scope of the rights of the present disclosure is not limited thereto, and various modifications and improvements made by those skilled in the art using the basic concept of the present disclosure defined in the following claims are also possible. It falls within the scope of rights.

Claims (20)

As a predictive maintenance method performed in a predictive maintenance device,
For a plurality of data units and a plurality of switching units that constitute a network transmission device and are interconnected to transmit and receive data frames through a plurality of buses, the plurality of buses are divided by bus direction and bus status information for each bus direction is provided. periodically collecting from the network transmission device, and
Each collected bus state information is divided into the plurality of data units and the plurality of switching units, and anomalies are detected in each data unit and switching unit through a predictive maintenance model that uses the divided bus state information as input data. Steps to monitor
Method, including. In paragraph 1:
The monitoring step is,
Using the status information of each bus as input data of a previously learned predictive maintenance model, obtaining an anomaly score indicating an abnormality for the corresponding bus from the predictive maintenance model.
Calculating predictive maintenance values in data unit and switching unit units through a weighted average calculation of the abnormality scores obtained for each bus, and
Comparing the predictive maintenance values of the data unit and switching unit with a preset threshold and outputting a visualized screen.
Method, including. In paragraph 2,
The preset threshold is,
A method that is differentially set for each data unit and the switching unit. In paragraph 2,
Each of the bus status information above is,
A method comprising a registry value that is a signaling level value for the bus, and a flag value that indicates an error condition of a data frame passing through the bus. In paragraph 4,
Each of the bus status information above is,
In the case of bus state information between a data unit and a switching unit, the method further includes a connection topology value. In paragraph 5,
The connection topology value is,
A method set to a flag value between a switching unit and a data unit constituting another connected network transmission device. In paragraph 2,
Each of the bus status information above is,
A plurality of first bus status information for a plurality of first buses from each data unit to the plurality of switching units,
A plurality of second bus status information for a plurality of second buses from each switching unit to the plurality of data units, and
A plurality of third bus status information for a plurality of third buses heading from the high-speed switching module included in the switching unit to the low-speed switching module, and a plurality of third bus status information for a plurality of fourth buses heading from the low-speed switching module to the high-speed switching module. 4 Bus status information
Method, including. In paragraph 7:
The step of obtaining the above score is,
Obtaining an anomaly score for each bus from the predictive maintenance model using the plurality of first bus state information as input data,
The step of calculating the predictive maintenance value is,
A method for calculating a predictive maintenance value for the data unit by calculating a weighted average of the abnormality scores obtained for the plurality of first buses of the data unit. In paragraph 7:
The step of obtaining the above score is,
Obtaining an abnormality score for each bus from the predictive maintenance model using the plurality of second bus state information, the plurality of third bus state information, and the plurality of fourth bus states as input data,
The step of calculating the predictive maintenance value is,
A method for calculating a predictive maintenance value for the switching unit by calculating a weighted average of the abnormality scores obtained for the plurality of second buses, the plurality of third buses, and the plurality of fourth buses of the switching unit. In paragraph 2,
The predictive conservation model is,
A method in which different unsupervised learning-based models are used for the switching unit and the data unit. In paragraph 2,
The collecting step is,
A method of collecting bus status information for each bus direction in window units, which are preset time units. A network transmission device including a plurality of data units and a plurality of switching units interconnected to transmit and receive data frames through a plurality of buses, and
Periodically collect a plurality of bus state information corresponding to each bus direction for the plurality of buses from the network transmission device, and divide the collected state information into the plurality of data units and the plurality of switching units, A predictive maintenance device that monitors abnormalities in data units and switching units through a predictive maintenance model that uses the status information of each bus as input data.
Including a predictive maintenance system. In paragraph 12:
The predictive maintenance device is,
Input the status information of each bus into the predictive maintenance model to calculate an abnormality score for each bus, calculate the predictive maintenance value in data unit and switching unit units by adding the abnormality scores for each bus, and calculate the predictive maintenance value in units of data units and switching units. A predictive maintenance system that displays a screen visualizing values relative to threshold values. In paragraph 12:
The plurality of data units are:
A plurality of bus registers configured for each bus unit to store status information of a plurality of buses heading from the plurality of data units to the plurality of switching units, and
A monitor process that monitors status information for the plurality of buses and stores the monitored status information in the bus register,
The plurality of bus status information stored in the plurality of bus registers is:
A predictive maintenance system that is periodically transmitted to the predictive maintenance device. In paragraph 12:
The plurality of data units are:
A plurality of first bus registers configured for each bus unit to store status information of a plurality of buses heading from the plurality of switching units to the plurality of data units,
A plurality of second bus registers composed of each bus unit to store status information of a plurality of buses for each bus direction between the internal high-speed switching module and the low-speed switching module,
A monitor process that monitors status information for the plurality of buses and stores the monitored status information in the plurality of first bus registers and the plurality of second bus registers, respectively,
A plurality of bus status information stored in each of the first plurality of bus registers and the plurality of second bus registers is:
A predictive maintenance system that is periodically transmitted to the predictive maintenance device. A data collection unit that periodically collects a plurality of bus status information registered in a plurality of bus registers included in the network transmission device from the network transmission device,
Predictive maintenance that uses the plurality of bus status information as input data of a predictive maintenance model to calculate an abnormality score indicating an abnormality, and generates a predictive maintenance value for the network transmission device through a weighted average calculation of the calculated abnormality score. execution department, and
A predictive maintenance result output unit that compares the predictive maintenance value with the threshold value and outputs visualization data on the screen.
Including a predictive maintenance device. In paragraph 16:
The plurality of bus status information is,
A predictive maintenance device, including a registry value that is a signaling level value for the bus, and a flag value that indicates an error condition of a data frame passing through the bus. In paragraph 17:
The network transmission device,
It includes a plurality of data units that process data frames and a plurality of switching units that set a transmission path for the data frames,
The plurality of data units and the plurality of switching units are:
Connected to each other through multiple buses,
The plurality of bus status information is,
A predictive maintenance device consisting of a plurality of units for each bus direction. In paragraph 18:
The predictive conservation department,
An abnormality score for each bus is calculated from the predictive maintenance model using the status information of a plurality of buses heading from the data unit to the switching unit as input data, and the weighted average value of the abnormality score for each bus is calculated as the data unit. A predictive maintenance device that calculates a predictive maintenance value, generates visualization data that compares the calculated predictive maintenance value with a first threshold set for a data unit, and outputs it to the predictive maintenance result output unit. In paragraph 18:
The predictive conservation department,
Calculating an abnormality score for each bus from the predictive maintenance model using a plurality of bus state information from the switching unit to the data unit and a plurality of bus state information between internal switching modules of the switching unit as input data,
Calculate the weighted average of the abnormality scores for each bus as the predictive maintenance value of the switching unit, and generate visualization data that compares the calculated predictive maintenance value with the second threshold set for the switching unit. A predictive maintenance device that outputs the predictive maintenance result to the output unit.

Images