WO2020111317A1 - 기계학습 기법에 기반한 기계의 오류 데이터를 검출하기 위한 알고리즘 및 방법 - Google Patents
기계학습 기법에 기반한 기계의 오류 데이터를 검출하기 위한 알고리즘 및 방법 Download PDFInfo
- Publication number
- WO2020111317A1 WO2020111317A1 PCT/KR2018/014895 KR2018014895W WO2020111317A1 WO 2020111317 A1 WO2020111317 A1 WO 2020111317A1 KR 2018014895 W KR2018014895 W KR 2018014895W WO 2020111317 A1 WO2020111317 A1 WO 2020111317A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- data
- machine
- error
- time
- machine learning
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 66
- 238000010801 machine learning Methods 0.000 title claims abstract description 47
- 238000001514 detection method Methods 0.000 claims description 19
- 230000002950 deficient Effects 0.000 claims description 8
- 238000004519 manufacturing process Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims 1
- 230000007547 defect Effects 0.000 abstract description 8
- 238000004891 communication Methods 0.000 description 20
- 238000005259 measurement Methods 0.000 description 15
- 238000007405 data analysis Methods 0.000 description 11
- 238000005516 engineering process Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 230000007257 malfunction Effects 0.000 description 5
- 230000005856 abnormality Effects 0.000 description 4
- 238000007781 pre-processing Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000007796 conventional method Methods 0.000 description 2
- 238000002604 ultrasonography Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013524 data verification Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units
- B60W30/18—Propelling the vehicle
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
- B60W40/02—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to ambient conditions
- B60W40/06—Road conditions
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Definitions
- the present invention relates to an algorithm and method for detecting error data of a machine based on a machine learning technique.
- Smart factory refers to the overall technologies that make it easy to connect and easily manage machines and facilities in factories.
- One of the many technologies that make up this smart factory is managing machines. In other words, it is to monitor whether the machine is operating normally and to predict and manage the occurrence of errors in advance.
- a threshold value defined in the form of an absolute value independent of time is set, and when operation data out of the threshold is generated by collecting operation data of the machine from time to time, a specific error occurs I think it was done.
- the conventional threshold value is related to a set allowable value of a machine for production. In other words, it is a combination of an upper limit and a lower limit of the operation to prevent defects in the product being produced.
- the conventional threshold is a fixed value (absolute value fixed regardless of time within each section set by the operator) directly set by the operator of the factory.
- a threshold is set for a time period during which a product is manufactured during a period corresponding to one cycle in which the machine operates, and a separate threshold is not set for a time period not related to product manufacturing. Therefore, it is difficult to accurately determine whether or not the machine is in error, although it is possible to judge whether a product is defective.
- the present invention is to solve the above-mentioned problems of the prior art, by automatically extracting time-series threshold data from the motion data of the machine collected in real time based on the machine learning technique, grasping as the operator sets the threshold by hand
- the aim is to enable accurate prediction and detection of machine error data.
- an object of the present invention is to provide error data deviating from time-series threshold data so that it can be easily confirmed by an operator terminal.
- a method for detecting error data of a machine based on a machine learning technique includes: (a) collecting time-series operation data of at least one machine; (b) dividing the motion data at predetermined time intervals, and mapping the divided motion data to overlap on the same time domain; (c) generating time-series threshold data by deriving time-series standard data for the set of mapped motion data based on machine learning; And (d) when the motion data collected in real time deviates from the threshold data, determining as an error event and providing information on the error event to the operator terminal.
- a server for detecting error data of a machine based on a machine learning technique includes: a memory for storing a program for detecting error data of a machine based on a machine learning technique; And a processor for executing the program, wherein the processor, according to the execution of the program, collects time-series motion data of at least one machine, divides the motion data at predetermined time intervals, and The segmented motion data is mapped to overlap on the same time domain, and time-series threshold data is generated by deriving time-series standard data for the set of mapped motion data based on machine learning, and motion data collected in real time When is out of the threshold data, it is determined as an error event and information on the error event is provided to the worker terminal.
- the present invention automatically detects time-series threshold data from a server, and compares operation data in all time domains based on this, thereby accurately detecting defects or product defects of machines that could not be found when setting a conventional threshold (absolute value). Can be detected.
- the present invention can check the presence or absence of an abnormality in product and quality more precisely than in the conventional case, thereby enabling precise predictive maintenance.
- the threshold value is precisely generated in time series for all time domains, the accuracy of detecting a defect or malfunction of a machine or product is very high, and accordingly, the process capability index (Cp) of the machine can be greatly improved. Capability index detection is possible.
- FIG. 1 is a structural diagram of a system according to an embodiment of the present invention.
- FIG. 2 is a block diagram of the structure of a sensor assembly according to an embodiment of the present invention.
- 3 is a graph of three representative types of motion data measured from a machine.
- FIG. 4 is a block diagram of the structure of a server according to an embodiment of the present invention.
- FIG. 5 is a graph of pre-processed operation data according to an embodiment of the present invention.
- FIG. 6 is a graph in which motion data is divided in 60 second units according to an embodiment of the present invention and then mapped so that each divided region overlaps within a 60 second unit.
- FIG. 7 is a graph of standard data detected based on machine learning from motion data collected according to an embodiment of the present invention.
- FIG. 8 is a graph of time series threshold data according to an embodiment of the present invention.
- FIG. 9 is a graph of a case in which arbitrary operation data is input for threshold data verification according to an embodiment of the present invention.
- 10 is a graph when error data is generated according to an embodiment of the present invention.
- FIG 11 is an example of a user interface according to an embodiment of the present invention.
- 12 is a graph for comparing threshold data according to an embodiment of the present invention and the prior art.
- FIG. 13 is a flowchart illustrating a method for detecting error data of a machine based on machine learning according to an embodiment of the present invention.
- unit includes a unit realized by hardware, a unit realized by software, and a unit realized by using both. Further, one unit may be realized by using two or more hardware, and two or more units may be realized by one hardware.
- ' ⁇ unit' is not limited to software or hardware, and' ⁇ unit' may be configured to be in an addressable storage medium or may be configured to play one or more processors.
- ' ⁇ unit' refers to components such as software components, object-oriented software components, class components and task components, processes, functions, attributes, and procedures. , Subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, database, data structures, tables, arrays and variables.
- components and' ⁇ units' may be combined into a smaller number of components and' ⁇ units', or further separated into additional components and' ⁇ units'.
- the components and' ⁇ unit' may be implemented to play one or more CPUs in the device or secure multimedia card.
- the "terminal” mentioned below may be implemented as a computer or a portable terminal capable of accessing a server or another terminal through a network.
- the computer for example, a laptop equipped with a web browser (WEB Browser), desktop (desktop), laptop (laptop), VR HMD (eg, HTC VIVE, Oculus Rift, GearVR, DayDream, PSVR, etc.) It may include.
- VR HMD is for PC (e.g. HTC VIVE, Oculus Rift, FOVE, Deepon, etc.) and mobile (e.g. GearVR, DayDream, Storm Racing, Google Cardboard, etc.) and console (PSVR). Includes all Stand Alone models (eg Deepon, PICO, etc.) that are implemented independently.
- the portable terminal is, for example, a wireless communication device that guarantees portability and mobility, as well as a smart phone, tablet PC, and wearable device, as well as Bluetooth (BLE, Bluetooth Low Energy), NFC, RFID, and ultrasound (Ultrasonic) , Infrared, Wi-Fi (WiFi), LiFi (LiFi), and may include various devices equipped with a communication module.
- “network” refers to a connection structure capable of exchanging information between each node such as terminals and servers, a local area network (LAN), a wide area network (WAN), and the Internet. (WWW: World Wide Web), wired and wireless data communication networks, telephone networks, and wired and wireless television communication networks.
- wireless data communication networks examples include 3G, 4G, 5G, 3rd Generation Partnership Project (3GPP), Long Term Evolution (LTE), World Interoperability for Microwave Access (WIMAX), Wi-Fi, Bluetooth communication, infrared communication, ultrasound Communication, Visible Light Communication (VLC), LiFi, and the like are included, but are not limited thereto.
- 3GPP 3rd Generation Partnership Project
- LTE Long Term Evolution
- WIMAX World Interoperability for Microwave Access
- Wi-Fi Bluetooth communication
- infrared communication ultrasound Communication
- VLC Visible Light Communication
- LiFi and the like are included, but are not limited thereto.
- operation data' refers to data that can directly or indirectly indicate the operation of the machine, and may mean information such as temperature, humidity, pressure, and power of the machine.
- the term'object' means a part produced in a factory or a sub-component of the part, and means an object produced/produced by a single machine.
- 'error data means data related to at least one of a machine error and a defect of an object produced by the machine.
- the system of an embodiment of the present invention includes a sensor assembly 100 installed near a machine 10 in a factory, a worker terminal 150, a server 200, and a manager terminal 300.
- the worker terminal 150 means a terminal assigned to a worker in charge of any one of the line equipment in the factory
- the manager terminal 300 may mean a terminal assigned to a line manager or a factory manager.
- the worker terminal 150 and the manager terminal 300 may be collectively called a worker terminal.
- a system is a system capable of providing a smart factory service.
- This smart factory service provides the efficiency and convenience of management of the machine 10 of the manager by monitoring the operation status of the machine facility 10 in the factory in real time and notifying the manager immediately when there is a high possibility of malfunction or a malfunction. can do.
- IOT Internet of Things
- the factory manager can individually search for the machine 10 and remove the inconvenience of scrutinizing whether there is a problem, and the sensor assembly ( By performing machine learning on the value measured in 100), the threshold data can be automatically set on the server to determine whether there is a precise machine abnormality or a product abnormality.
- the sensor assembly 100 is an IOT integrated module terminal composed of at least one sensor.
- the sensor assembly 100 is installed in the vicinity of the machine 10 in the factory, and may be configured to be attached to any one surface of the machine 10.
- the sensor assembly 100 includes a sensor for measuring motion data of the machine 10 and a sensor for transmitting it to the server 200.
- the sensor assembly 100 includes a sensor controller 110, a communication module 120, a measurement sensor 130, and a connector 140.
- each of the sensor controllers and the measurement sensors 110-130 may be implemented in a physically independent form. That is, as shown in FIG. 1, each sensor may be formed to have similar or identical physical specifications in a hexahedral form, and it is very easy to be replaced for each required module. For example, if a failure occurs in any one of the sensor controller 110, the communication module 120, and the measurement sensor 130 or a change in use is required, the problem may be easily solved by replacing the corresponding sensor.
- the sensor assembly 100 may further include a cradle. The cradle functions to support the sensor controller 110 and the communication module 120 together.
- the cradle is formed with an area capable of covering the areas of the sensor controller 110 and the communication module 120, and the partition wall is formed in the border area to fix the sensor controller 110 and the communication module 120 so as not to escape outside. You can also perform functions.
- the sensor controller 110 and the communication module 120 may be arranged in a form of being stacked with each other on the cradle.
- the sensor controller 110 receives the electrical signal value (current or voltage value) measured by the measurement sensor 130 from the measurement sensor 130, and transmits the electrical signal value to the server 200 and collects the communication module 120 ). At this time, the sensor controller 110 is connected to at least one measurement sensor 130. If the existing measurement sensor 130 is replaced by another type of measurement sensor 130 through the replacement or addition of the measurement sensor 130, the sensor controller 110 is electrically connected to the currently connected measurement sensor 130. The signal value can be received and recognized.
- various sensors such as a temperature sensor, a pressure sensor, a humidity sensor, a current/voltage sensor, and a power sensor
- the sensor controller 110 may be connected.
- the sensor controller 110 simply stores the electrical signal value of the measurement sensor 130 (that is, A/D signal value: a signal converted from analog to digital) to the server ( 200), and when the firmware for all sensors is downloaded and installed in the sensor controller 110, even if each sensor is an object that is not compatible with different devices, the sensor controller 110 is applicable The signal from the sensor can be recognized.
- the sensor controller 110 converts the operation data received from the measurement sensor 130 into a standardized digital signal and transmits it to the communication module 120.
- signals transmitted from the temperature sensor and the pressure sensor to the sensor controller 110 may be electrical signals of different formats. If it is transmitted to the server 200 as it is, it may not accurately recognize what information the server 200 contains. To this end, the sensor controller 110 may also convert an analog signal or a digital signal into a standardized digital signal and convert it into a form recognizable by the server 200.
- the communication module 120 transfers information between the sensor controller 110 and the server 200 or the user terminal 300.
- the electrical signal value transmitted to the server 200 by the communication module 120 may be recognized as operation data by firmware stored in the server 200. That is, it is transmitted to the server 200 as a simple current value or voltage value, but since there is firmware in the server 200, it can be recognized as a value related to operation data such as temperature, pressure, and humidity.
- the measurement sensor 130 is a sensor that measures motion data of the machine 10.
- the measurement sensor 130 may be a sensor that measures any one of temperature, pressure, humidity, voltage, power, and vibration. This is only an example, and may include sensors for measuring various other motion data.
- the sensor controller 110, the communication module 120, and the measurement sensor 130 may be connected and fixed to each other through the connector 140.
- the connector 140 may be implemented in the form of wires or wires.
- the connector 140 may be implemented in the form of a plurality of pins formed in one region of each sensor.
- the sensors 140 formed in each sensor are disposed and connected to mesh with each other so that the sensors can be connected to each other.
- the server 200 may receive electrical signal values related to the operation data of the machine 10 from the sensor assembly 100 installed for each machine 10, and recognize which operation data is based on firmware. In addition, by automatically generating time-series threshold data of the machine based on the motion data collected through the analysis method based on big data analysis and machine learning, so that the operator does not have to manually input the threshold data corresponding to the absolute value. In addition, since it is threshold data based on machine learning, it is possible to accurately set the threshold, so it is possible to accurately determine whether the machine is in error or the product is defective.
- Information on the operation data of the server 200 and 100 may be provided to the worker terminal 150 and 150 and the manager terminal 300 and 300.
- the worker terminal 150, 150 is a terminal that is installed on the machine 10 or is disposed in the vicinity of the machine 10, and a worker in charge of a process directly monitors the status of the machine 10 in charge. Real-time operation data for the machine 10 can be displayed so that it can be checked.
- the manager terminals 300 and 300 may be installed with an application capable of providing a smart factory service.
- the application may receive information from the server 200 and 200 and process it in an easy-to-understand form to provide the user with information about the operation status of the machine 10 and the equipment 10, and the entire machine in the factory ( It can be displayed to inquire at a glance the operation data for 10).
- the representative type of the operation data of the machine 10 can be summarized to three degrees as shown in FIG. That is, as shown in 1, it has different amplitudes over time, but has certain periods, and is read-out operation data. As shown in 2, it has the same amplitude and period over time, but operation data whose frequency is different. After a period of time, it can be summarized as saturation motion data.
- a conventional machine 10 error or product error detection method is a method in which an operator inputs a threshold manually, and recognizes an error event when motion data exceeding the threshold is found.
- the threshold value is an absolute value defined regardless of time. Therefore, when determining the normal operation by setting the threshold for the saturation value as 3, the conventional method may be appropriate, but when the amplitude or frequency is different depending on the time, such as 1 and 2, use the existing method. The exact error event cannot be recognized. Accordingly, when applying the conventional method to 1, the absolute value threshold is applied only to a certain time period area directly connected to product production among all time domains. Therefore, it is impossible to detect an error event in another time period region to which the absolute value threshold is not applied.
- One embodiment of the present invention described later regardless of what type of motion data of the machine 10, generates time-series threshold data that matches the pattern of the motion data based on the machine learning technique, so all types of motion It can be referred to as a machine 10 error detection technique that can be applied to the machine 10 for data.
- a machine 10 error detection technique that can be applied to the machine 10 for data.
- the operation data of the machine 10 is of the type (1).
- the server 200 includes a memory in which a program (or application) performing a method for detecting error data of the machine 10 based on a machine learning technique and a processor executing the above program Can be.
- the processor may perform various functions according to the execution of a program stored in the memory, and the detailed components included in the processor according to each function may be a data preprocessing unit 210, a data analysis unit 220, and a threshold data detection unit 230 ), an error detection and determination unit 240 and an operation data providing unit 250.
- the data pre-processing unit 210 collects operation data of the machine 10 from the past to the present and performs pre-processing.
- the operation data of the machine 10 is of the type 1 in FIG. 3 described above, the operation data is formed in one pattern each time the machine 10 produces one object. That is, when the machine 10 manufactures one object for 60 seconds, the operation data has the same or similar amplitude (value for temperature, pressure, voltage, etc.) in units of 60 seconds. At this time, since the motion data collected while the machine 10 is operating to produce an object are meaningful data, the data during the time when the machine 10 is not operating is excluded and the data during the operation time are collected. , Compression can be performed on the collected data. Accordingly, the pre-processed operation data may be displayed in a graph as shown in FIG. 5.
- event data 410 having a pattern that deviates from the collected pattern of motion data may be detected.
- the event data 410 having a pattern deviating from the average pattern (for example, a graph shape) formed by the collected motion data may be detected.
- the amplitude value is different from the motion data in a specific time period. Compared to this, event data 410 that is much higher is shown.
- the event data 410 is data related to an error of the machine 10 or an object.
- the work data at the moment the event data 410 is detected (that is, information about whether the machine 10 is in error or an object is defective) is retrieved, and the work is performed.
- Information and event data 410 may be matched and stored in the server 200.
- the information stored matching with respect to the event data 410 includes information indicating a machine 10 error and object failure, information representing a machine 10 error and object failure, information representing a machine 10 normal and object failure, and a machine (10) It may be any one of information indicating normal and object normal.
- the data analysis unit 220 may perform a machine learning technique on the pre-processed motion data to derive standard data for the motion data.
- the data analysis unit 220 may divide the operation data at predetermined time intervals and map the divided operation data on a time domain having a length corresponding to the predetermined time interval.
- the predetermined time interval may be one cycle of operation data (eg, 60 seconds).
- One cycle may be the time it takes for the machine 10 to manufacture one object.
- the data analysis unit 220 divides one period into units for all collected time-series operation data, and maps values constituting the divided operation data on a graph having a length of one cycle, as shown in FIG. 6. Graphs can be derived. That is, all points constituting the divided motion data are mapped on the graph. According to the graph of FIG. 6, it can be seen that the operation data has a specific pattern and is repeated.
- the data analysis unit 220 extracts at least one time-series standard data based on an average value or a median value from the set of mapped motion data based on a machine learning technique and detects standard data having the highest K index. can do.
- the data analysis unit 220 extracts standard data based on the median value, it is possible to detect, as standard data, the most frequently divided motion data among all the divided motion data superimposed on the graph of FIG. 6.
- the K index may be a statistical index or mean an average value or a median value. It shows that the closer the K index of a specific graph is to 1, the closer the standard value of the motion data.
- the data analysis unit 220 repeatedly detects the standard data for the collected motion data, measures the K index, performs machine 10 learning, and detects the standard data having the highest K index. For example, a graph for standard data as shown in FIG. 7 can be derived.
- the threshold data detector 230 detects time-series threshold data based on standard data. Specifically, among motion data mapped on the graph of FIG. 6, upper limit threshold data is performed by performing a machine learning process performed by the data analysis unit 220 on motion data having a higher Y-axis value (amplitude value) than standard data (422) is detected. In addition, the threshold data detector 230 performs the same machine learning process on the motion data having a lower Y-axis value (amplitude value) than the standard data among motion data mapped on the graph of FIG. 6, thereby lower limit threshold data 423 Detects. At this time, the combination of the upper limit threshold data 422 and the lower limit threshold data 423 becomes threshold data.
- standard data 421 is mapped between the upper limit threshold data 422 and the lower limit threshold data 423.
- the threshold data is configured to have different values over time. That is, since the type of motion data was the type 1 in FIG. 3, the threshold data derived based on machine learning is also the type 1.
- the upper limit threshold data 422 and the lower limit threshold data 423 in FIG. 8 are described as having a difference of +10 or -10 from the standard data, this is only an example and may be configured to have different difference values.
- the error detection and determination unit 240 detects an error by comparing motion data collected in real time with threshold data and determines what type of error it is.
- the error detection and determination unit 240 may consider that the machine 10 and the object are in a normal state.
- the error detection and determination unit 240 may detect the error data 425.
- the error data 425 of FIG. 10 has a normal data pattern form until about 25 seconds, but has an abnormal pattern that deviates from the threshold data in a time period d between about 25 seconds and about 50 seconds.
- the error detection and determination unit 240 compares the pattern of the error data 425 with the pattern of the event data 410.
- the error detection and determination unit 240 provides information on whether the machine 10 of the event data 410 having a pattern corresponding to the error data 425 has an error and whether or not the object is defective according to the comparison result. 425). That is, the event data 410 includes the four types described above (machine 10 normal and object normal type, machine 10 normal and object bad type, machine 10 error and object normal type, machine 10 error and Object error type), and it is possible to determine which type of error data 425 corresponds to these four types.
- the operation data providing unit 250 may provide information on an error event to the operator terminal.
- the user interface provided to the worker terminal may have the form shown in FIG. 11.
- the user interface is identification information (for example, WF-11, WF-12, WF-21, WF-22, WF-31, WF of FIG. 11) included in the work site. -32) and the machine 10 displayed by matching the identification information for each machine 10 and the state information of the object produced by the machine 10. That is, status information may be provided by being divided into blocks for each machine 10.
- the current state information of the machine 10 can be expressed by differently displaying color or contrast for each state value (machine 10 malfunction, machine 10 normal, product defect, product normal).
- the user interface presents a picture of a structure in which a plurality of machines 10 are arranged, and presents identification information and status information for each machine 10 on the picture, in a map form. It is also possible to provide the operator with the status information for each machine 10 so that they can be easily checked at a glance.
- real-time collection information for a plurality of time-series operation data for the machine 10 is generated.
- a real-time graph of temperature, humidity, pressure, etc. of the machine 10 can be provided.
- the user interface may provide a worker with a real-time operation data graph and a graph for threshold data superimposedly displayed when the real-time graph is enlarged.
- the threshold data is generated once through the above-described machine learning and big data analysis process, since the motion data is continuously collected and accumulated, the same machine learning and big data analysis process including the subsequent motion data It can be updated by doing again.
- threshold data set by a worker according to the prior art is defined as absolute values such as USL (Upper Spec Limit) and LSL (Lower Spec Limit).
- USL Upper Spec Limit
- LSL Lower Spec Limit
- the threshold data is composed of the upper limit threshold data 422 and the lower limit threshold data 423 that change in time series. Accordingly, it is possible to detect error data in the E1 region that cannot be detected in the prior art.
- the server 200 first collects time-series operation data of the machine 10 and performs pre-processing (S110).
- the server 200 extracts a plurality of error patterns from pre-processed operation data (S120). When a pattern deviating from the average pattern indicated by the motion data is found, it is detected as an error pattern, and each error pattern is matched with the information in the work log to determine whether the error pattern indicates an abnormality of the machine 10 or the object. do.
- the server 200 detects time-series threshold data from all motion data collected based on machine learning (S130). Standard data having the highest K index is detected, and threshold data is detected based on the standard data.
- the server 200 detects error data outside the threshold data (S140).
- the server 200 detects an error state by comparing error data and a plurality of error patterns (S150). That is, the error state is detected by comparing the error data with the previously detected event data 410.
- the server 200 provides error state information according to a user interface preset as an operator terminal (S160).
- One embodiment of the invention may also be implemented in the form of a recording medium comprising instructions executable by a computer, such as program modules, being executed by a computer.
- Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media.
- the computer-readable medium may include any computer storage medium.
- Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data.
Landscapes
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Mechanical Engineering (AREA)
- Physics & Mathematics (AREA)
- Transportation (AREA)
- Aviation & Aerospace Engineering (AREA)
- Mathematical Physics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- General Physics & Mathematics (AREA)
- General Factory Administration (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
Claims (17)
- 서버에 의해 수행되는, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법에 있어서,(a) 적어도 하나의 기계의 시계열적인 동작데이터를 수집하는 단계;(b) 상기 동작데이터를 소정의 시간 간격으로 분할하고, 상기 분할된 동작데이터를 동일한 시간영역 상에 중첩되도록 매핑하는 단계;(c) 상기 매핑된 동작데이터의 집합에 대한 시계열적인 표준데이터를 머신러닝에 기반하여 도출함으로써 시계열적인 임계치 데이터를 생성하는 단계; 및(d) 실시간으로 수집되는 동작데이터가 상기 임계치 데이터를 벗어나는 경우, 오류 이벤트로 판단하여 상기 오류 이벤트에 대한 정보를 작업자 단말로 제공하는 단계를 포함하는, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 1 항에 있어서,상기 (a) 단계는,상기 기계가 동작하여 지정된 객체를 제작하는 과정을 수행할 때의 동작데이터를 수집하는 단계를 포함하는, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 1항에 있어서,상기 (a) 단계는,상기 수집된 동작데이터의 패턴으로부터 어긋나는 패턴을 갖는 이벤트 데이터를 검출하는 단계; 및상기 이벤트 데이터에 대하여 기계 오류 여부 및 상기 기계에 의해 제작되는 객체의 불량 여부에 대한 정보를 매칭하여 저장하는 단계;를 포함하는, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 3 항에 있어서,상기 이벤트 데이터에 대하여 매칭 저장되는 정보는 기계 오류 및 객체 불량을 나타내는 정보, 기계 오류 및 객체 정상을 나타내는 정보 및 기계 정상 및 객체 불량을 나타내는 정보 중 어느 하나인 것인, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 1 항에 있어서,상기 (b) 단계는,상기 동작데이터의 한 주기를 상기 소정의 시간 간격으로 지정하여, 상기 수집된 동작데이터를 분할하고, 상기 한 주기에 대응하는 길이의 상기 시간영역 상에 상기 분할된 동작데이터를 매핑하는 단계를 포함하는, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 5 항에 있어서,상기 동작데이터의 한 주기는, 상기 기계가 하나의 객체를 제조하는 데에 소요되는 시간인 것인, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 1 항에 있어서,상기 (c) 단계는,(c-1) 머신러닝 기법에 기반하여, 상기 매핑된 동작데이터의 집합으로부터 메디안(median) 값에 기반한 적어도 하나의 시계열적인 표준 데이터를 추출하고 통계 지수인 K 지수가 가장 높은 표준 데이터를 검출하는 단계를 포함하는, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 7 항에 있어서,상기 (c) 단계는,(c-2) 상기 (c-1) 단계 이후, 상기 매핑된 동작데이터 중 상기 표준 데이터보다 높은 값을 갖는 동작데이터에 대하여 상기 (c-1) 단계를 수행하여, 상한 임계치 데이터를 검출하는 단계; 및(c-3) 상기 매핑된 동작데이터 중 상기 표준 데이터보다 낮은 값을 갖는 동작데이터에 대하여 상기 (c-1) 단계를 수행하여, 하한 임계치 데이터를 검출하는 단계;를 포함하는, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 8 항에 있어서,상기 표준 데이터는 상기 상한 임계치 데이터와 상기 하한 임계치 데이터의 사이 값으로 구성되며,상기 시계열적인 임계치 데이터는 상기 상한 임계치 데이터와 상기 하한 임계치 데이터의 조합으로 구성되며, 시간에 따라 상이한 값을 갖도록 구성되는 것인, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 4 항에 있어서,상기 (d) 단계는,실시간으로 수집되는 동작데이터 중 어느 한 시점에서 상기 임계치 데이터를 벗어나는 오류데이터가 검출되는 경우, 상기 오류데이터의 패턴을 상기 이벤트 데이터의 패턴과 비교하는 단계; 및비교 결과에 따라, 상기 오류데이터와 대응하는 패턴을 갖는 이벤트 데이터의 기계 오류 여부 및 객체의 불량 여부에 대한 정보를 상기 오류데이터에 대한 정보로서 상기 작업자 단말로 제공하는 단계;를 포함하는, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 1 항에 있어서,상기 작업자 단말로 제공되는 사용자 인터페이스는,작업 현장에 포함된 복수의 기계에 관한 식별정보 및각 기계에 관한 식별정보 별로 매칭되어 표시되는 기계 및 상기 기계가 제작하는 객체의 상태 정보를 포함하며,기계 및 객체의 오류데이터에 대한 정보는 상기 오류 이벤트가 발생되는 경우, 상기 상태 정보로서 제공되는 것인, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 11 항에 있어서,상기 상태 정보는,기계 정상 및 객체 정상에 관한 상태, 기계 정상 및 객체 불량에 관한 상태, 기계 오류 및 객체 정상에 관한 상태 및 기계 오류 및 객체 오류에 관한 상태를 포함하는 것인, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 12 항에 있어서,상기 사용자 인터페이스에서 어느 하나의 기계에 대한 식별정보에 대한 작업자의 입력이 발생될 경우, 상기 기계에 대한 복수의 시계열적인 동작데이터에 대한 실시간 수집 정보가 제공되는 것인, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 1 항에 있어서,상기 시계열적인 임계치 데이터는,상기 동작데이터가 실시간으로 수집되고 누적됨에 따라 업데이트되는 것인, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 제 1 항에 있어서,상기 작업자 단말로 제공되는 사용자 인터페이스는,실시간으로 수집되는 시계열적인 동작데이터에 대한 그래프 및 상기 그래프 위에 중첩되어 표시된 상기 임계치 데이터를 포함하는 것인, 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법.
- 머신러닝 기법에 기반한 기계의 오류 데이터를 검출하는 서버에 있어서,머신러닝 기법에 기반한 기계의 오류 데이터를 검출하기 위한 프로그램을 저장하는 메모리; 및상기 프로그램을 실행하기 위한 프로세서;를 포함하며,상기 프로세서는, 상기 프로그램의 실행에 따라,적어도 하나의 기계의 시계열적인 동작데이터를 수집하고,상기 동작데이터를 소정의 시간 간격으로 분할하고, 상기 분할된 동작데이터를 동일한 시간영역 상에 중첩되도록 매핑하며,상기 매핑된 동작데이터의 집합에 대한 시계열적인 표준데이터를 머신러닝에 기반하여 도출함으로써 시계열적인 임계치 데이터를 생성하고,실시간으로 수집되는 동작데이터가 상기 임계치 데이터를 벗어나는 경우, 오류 이벤트로 판단하여 상기 오류 이벤트에 대한 정보를 작업자 단말로 제공하는 것인, 머신러닝 기법에 기반한 기계의 오류 데이터를 검출하는 서버.
- 제 1항에 따르는 머신러닝 기법에 기반한 기계의 오류 데이터 검출 방법을 수행하기 위한 컴퓨터 프로그램이 저장된 컴퓨터 판독가능 기록매체.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/KR2018/014895 WO2020111317A1 (ko) | 2018-11-29 | 2018-11-29 | 기계학습 기법에 기반한 기계의 오류 데이터를 검출하기 위한 알고리즘 및 방법 |
JP2021529123A JP7169706B2 (ja) | 2018-11-29 | 2018-11-29 | 機械学習手法に基づく機械のエラーデータを検出するためのアルゴリズム及び方法 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/KR2018/014895 WO2020111317A1 (ko) | 2018-11-29 | 2018-11-29 | 기계학습 기법에 기반한 기계의 오류 데이터를 검출하기 위한 알고리즘 및 방법 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020111317A1 true WO2020111317A1 (ko) | 2020-06-04 |
Family
ID=70853283
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/KR2018/014895 WO2020111317A1 (ko) | 2018-11-29 | 2018-11-29 | 기계학습 기법에 기반한 기계의 오류 데이터를 검출하기 위한 알고리즘 및 방법 |
Country Status (2)
Country | Link |
---|---|
JP (1) | JP7169706B2 (ko) |
WO (1) | WO2020111317A1 (ko) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2023287698A1 (en) * | 2021-07-14 | 2023-01-19 | Baker Hughes Holdings Llc | Configuration of asset monitoring systems |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20130061029A (ko) * | 2011-11-30 | 2013-06-10 | (주)한콘트롤스 | 열처리 설비장치의 지능형 관리 시스템 |
KR20160067510A (ko) * | 2014-12-04 | 2016-06-14 | 삼성전자주식회사 | 배터리의 상태를 추정하는 방법 및 장치 |
US20160202693A1 (en) * | 2015-01-09 | 2016-07-14 | Hitachi Power Solutions Co., Ltd. | Anomaly Diagnosis System and Anomaly Diagnosis Method |
KR101651104B1 (ko) * | 2015-12-09 | 2016-08-25 | 이정훈 | 진동 및 소음의 격차 패턴을 이용한 머신 정보 수집시스템 |
US20170139398A1 (en) * | 2015-11-16 | 2017-05-18 | Jtekt Corporation | Abnormality analysis system and analysis apparatus |
KR20190017121A (ko) * | 2017-08-10 | 2019-02-20 | 울랄라랩 주식회사 | 기계학습 기법에 기반한 기계의 오류 데이터를 검출하기 위한 알고리즘 및 방법 |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5480440B1 (ja) | 2013-12-03 | 2014-04-23 | 株式会社日立パワーソリューションズ | 異常予兆診断装置及び異常予兆診断方法 |
JP5827425B1 (ja) | 2015-01-09 | 2015-12-02 | 株式会社日立パワーソリューションズ | 予兆診断システム及び予兆診断方法 |
JP6828300B2 (ja) | 2015-11-16 | 2021-02-10 | 株式会社ジェイテクト | 異常解析システム及び解析装置 |
JP2020027342A (ja) | 2018-08-09 | 2020-02-20 | 富士電機株式会社 | 情報処理装置、監視装置、及び情報処理方法 |
-
2018
- 2018-11-29 JP JP2021529123A patent/JP7169706B2/ja active Active
- 2018-11-29 WO PCT/KR2018/014895 patent/WO2020111317A1/ko active Application Filing
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20130061029A (ko) * | 2011-11-30 | 2013-06-10 | (주)한콘트롤스 | 열처리 설비장치의 지능형 관리 시스템 |
KR20160067510A (ko) * | 2014-12-04 | 2016-06-14 | 삼성전자주식회사 | 배터리의 상태를 추정하는 방법 및 장치 |
US20160202693A1 (en) * | 2015-01-09 | 2016-07-14 | Hitachi Power Solutions Co., Ltd. | Anomaly Diagnosis System and Anomaly Diagnosis Method |
US20170139398A1 (en) * | 2015-11-16 | 2017-05-18 | Jtekt Corporation | Abnormality analysis system and analysis apparatus |
KR101651104B1 (ko) * | 2015-12-09 | 2016-08-25 | 이정훈 | 진동 및 소음의 격차 패턴을 이용한 머신 정보 수집시스템 |
KR20190017121A (ko) * | 2017-08-10 | 2019-02-20 | 울랄라랩 주식회사 | 기계학습 기법에 기반한 기계의 오류 데이터를 검출하기 위한 알고리즘 및 방법 |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2023287698A1 (en) * | 2021-07-14 | 2023-01-19 | Baker Hughes Holdings Llc | Configuration of asset monitoring systems |
Also Published As
Publication number | Publication date |
---|---|
JP7169706B2 (ja) | 2022-11-11 |
JP2022510858A (ja) | 2022-01-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR102017561B1 (ko) | 기계학습 기법에 기반한 기계의 오류 데이터를 검출하기 위한 알고리즘 및 방법 | |
WO2019132112A1 (ko) | 관리 대상 설비 이상 상태 모니터링 시스템 | |
US11641536B2 (en) | Capture and association of measurement data | |
WO2022071760A1 (ko) | 화재 방호 방법 및 화재 방호 시스템 | |
KR102319083B1 (ko) | 인공지능 기반의 화재예방 제공 장치 및 방법 | |
US20220308572A1 (en) | System and method for determining a health condition and an anomaly of an equipment using one or more sensors | |
WO2020111317A1 (ko) | 기계학습 기법에 기반한 기계의 오류 데이터를 검출하기 위한 알고리즘 및 방법 | |
CN113153262A (zh) | 一种基于电缆热特性的海上油田可接入容量评估方法 | |
WO2020130169A1 (ko) | 이종 품질 데이터의 효율적 관리를 위한 opc ua 기반의 플랫폼 구성 방법 | |
CN110708286A (zh) | 一种测试互联网的监管系统 | |
WO2023106504A1 (ko) | 서버 시스템 로그를 이용한 머신러닝 기반의 관측레벨 측정 및 이에 따른 위험도 산출 방법, 장치 및 컴퓨터-판독 가능 기록 매체 | |
WO2021261614A1 (ko) | 설비의 진단 방법, 프로그램 및 시스템 | |
CN114400769A (zh) | 智能换电站的3d可视化监控方法与装置 | |
CN110764025A (zh) | 变电设备外绝缘的故障检测装置和方法 | |
CN112153330A (zh) | 一种工业互联网实时监控系统 | |
CN110164118A (zh) | 计量装置的监控方法、终端、系统及服务器 | |
CN117806912B (zh) | 一种服务器异常监测方法及系统 | |
CN108880916A (zh) | 一种基于iic总线的故障定位方法及系统 | |
CN214954653U (zh) | 一种电信机房综合环境状态实时监测及报警系统 | |
CN102035253A (zh) | 电网视频及环境监控系统中可记忆学习的测试方法及装置 | |
CN116132845A (zh) | 一种用于电力计量数字实验室数据采集的方法及装置 | |
CN118137682A (zh) | 配电自动化主站硬件故障主动检测系统及方法 | |
WO2018230853A1 (ko) | 장비신뢰지수에 기초한 Lot 리스크 스코어 기반의 동적 Lot 계측 제어방법 및 시스템 | |
CN115371727A (zh) | 六氟化硫在线监控装置全镜像式快速远程运维方法 | |
CN117555321A (zh) | 设备故障诊断方法、装置、电子设备及可读存储介质 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 18941831 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2021529123 Country of ref document: JP Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 18941831 Country of ref document: EP Kind code of ref document: A1 |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 18941831 Country of ref document: EP Kind code of ref document: A1 |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 22.04.2022) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 18941831 Country of ref document: EP Kind code of ref document: A1 |