Inspection object information acquisition system and inspection object information acquisition method

Abstract

To generate correspondence information between inspection object information and plant operation state information.SOLUTION: Abnormal sound detection system 100 of one aspect of the present invention includes a learning data creation unit 22 that creates learning data obtained by learning the feature amount of the recorded data of the operation sound of an inspection object 200 acquired by a microphone unit 5, an analysis unit 11 that calculates the degree of abnormality of the inspection object 200 based on the learning data and the feature amount of the recorded data acquired by the microphone unit 5, a data disclosure unit 413 that generates abnormality degree operation state information correspondence information in which information on the degree of abnormality is associated with operation state information indicating the transition of the operation state of the plant and causes a display unit 42 to display it.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an inspection target information acquisition system and an inspection target information acquisition method. [Background technology]

[0002] At production sites such as power plants, chemical plants, and steel plants, workers listen to the operating sounds of motors, pumps, and other equipment and judge whether the equipment is operating normally based on the information from the sounds they hear. However, determining whether a sound is normal or abnormal requires skill, and this skill is cultivated through experience. Therefore, it can be difficult for inexperienced workers to distinguish between normal and abnormal sounds.

[0003] When a skilled worker with such skills passes on his skills to a less experienced worker, the less experienced worker may be able to distinguish between normal and abnormal sounds. However, due to factors such as the aging of skilled workers, it is becoming more difficult to pass on skills. Therefore, in recent years, a system has been proposed that collects the operating sounds of equipment with a microphone and analyzes the collected sound information to determine whether the equipment is normal or abnormal.

[0004] Patent document 1 describes an anomaly detection device that uses a learning model (outlier detector) stored in a model memory unit to determine whether acoustic data input from a sensor attached to a machine tool is normal or abnormal. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent Publication No. 2021-22311 Summary of the Invention [Problem to be solved by the invention]

[0006] Incidentally, the operating sounds of motors, pumps, and the like in the equipment also change when the operating state of the plant changes. The operating state of a plant changes, for example, when the products produced in the plant change. Furthermore, when an abnormality occurs in the plant equipment, the pressure value in the equipment, as indicated by the analog meter, also changes. In other words, there is a close relationship between the state of the objects to be inspected, such as motors, pumps, and analog meters, and the operating state of the plant.

[0007] Therefore, by knowing the correspondence between information on inspection objects such as analog meters and information on the operating status of the plant, the plant manager or the like can more accurately grasp information on the plant equipment. However, Patent Document 1 does not disclose a technology for providing such correspondence information.

[0008] The present invention has been made in consideration of the above-mentioned circumstances, and an object of the present invention is to make it possible to generate correspondence information between information on inspection objects and information on the operating status of a plant. [Means for solving the problem]

[0009] An inspection object information acquisition system according to one embodiment of the present invention includes a learning data creation unit that creates learning data by learning features of the status information of the inspection object acquired by the information acquisition unit, an analysis unit that calculates the degree of abnormality of the inspection object based on the learning data and the features of the status information of the inspection object acquired by the information acquisition unit, and a data disclosure unit that generates abnormality degree operating status information correspondence data that matches the abnormality information with operating status information that indicates the transition of the operating status of a plant having the inspection object, and displays the data on a display unit. The learning data creation unit includes a first feature extraction unit that extracts features of the state information of the inspection object, clusters the features, and generates cluster information indicating the correspondence between the generated clusters and the features, an operation status linking unit that links the clusters generated by the first feature extraction unit with the operation status information and outputs only cluster information of the clusters that have been linked with the operation status information, and a learning data creation processing unit that creates learning data using the features indicated in the cluster information input from the operation status linking unit.The analysis unit also includes a learning data storage unit that stores the learning data created by the learning data creation unit, and a learning data processing unit that extracts features of the state information of the inspection object and outputs the extracted cluster information. The feature amount output do The apparatus includes a second feature extraction unit, and an anomaly degree / cluster information calculation unit which performs clustering of the features output from the second feature extraction unit, generates cluster information indicating the correspondence between the generated clusters and the features, and calculates and outputs an anomaly degree based on the learning data stored in the learning data storage unit. The anomaly degree is information indicating the amount of deviation of the features of the state information of the inspection object acquired by the information acquisition unit from the learning data, and the data disclosure unit associates the anomaly degree information with the operation state information based on information on the acquisition date and time of the state information of the inspection object, which is included in the cluster information input from the anomaly degree / cluster information calculation unit, and information on the date and time of each operation state, which is included in the operation state information.

[0010] Further, the inspection object information acquisition method according to one aspect of the present invention includes: acquiring learning data obtained by learning feature quantities of state information of the inspection object acquired by the information acquisition unit; The learning data creation departmentThe abnormality degree of the inspection object is determined based on the procedure for creating the abnormality degree, the learning data, and the feature amount of the state information of the inspection object acquired by the information acquisition unit. The analysis part and generating abnormality degree operation status information correspondence data in which the abnormality degree information corresponds to operation status information showing the transition of the operation status of the plant having the inspection object. Data Disclosure Department and a step of displaying the information on the display unit. The learning data creation unit includes a first feature extraction unit that extracts features of the state information of the inspection object, clusters the features, and generates cluster information indicating the correspondence between the generated clusters and the features, an operation status linking unit that links the clusters generated by the first feature extraction unit with the operation status information and outputs only cluster information of the clusters that have been linked with the operation status information, and a learning data creation processing unit that creates learning data using the features indicated in the cluster information input from the operation status linking unit.The analysis unit also includes a learning data storage unit that stores the learning data created by the learning data creation unit, and a learning data processing unit that extracts features of the state information of the inspection object and outputs the extracted cluster information. The feature amount output do The apparatus includes a second feature extraction unit, and an anomaly degree / cluster information calculation unit which performs clustering of the features output from the second feature extraction unit, generates cluster information indicating the correspondence between the generated clusters and the features, and calculates and outputs an anomaly degree based on the learning data stored in the learning data storage unit. The anomaly degree is information indicating the amount of deviation of the features of the state information of the inspection object acquired by the information acquisition unit from the learning data, and the data disclosure unit associates the anomaly degree information with the operation state information based on information on the acquisition date and time of the state information of the inspection object, which is included in the cluster information input from the anomaly degree / cluster information calculation unit, and information on the date and time of each operation state, which is included in the operation state information. Effect of the Invention

[0011] According to at least one aspect of the present invention, it becomes possible to generate correspondence information between information on an inspection object and information on the operating state of a plant. Problems, configurations and effects other than those described above will become apparent from the following description of the embodiments. [Brief description of the drawings]

[0012] [Figure 1] 1 is a block diagram showing an example of the overall configuration of an abnormal sound detection system according to an embodiment of the present invention. [Diagram 2] 1 is a diagram illustrating an example of how a wireless sensor according to an embodiment of the present invention is attached to an inspection target. [Diagram 3] FIG. 2 is a block diagram showing an example of the hardware configuration of each device constituting the abnormal sound detection system according to the embodiment of the present invention. [Figure 4] 5A to 5C are diagrams illustrating examples of power spectra of recorded data in each operating state according to an embodiment of the present invention. [Diagram 5] 11 is a graph showing an example of correspondence between clusters and operational states when the clusters and operational states are associated with each other by an operational state associating unit of a learning terminal according to one embodiment of the present invention. [Figure 6] FIG. 13 is a diagram illustrating an example of a display of abnormality level operation status correspondence information according to an embodiment of the present invention. [Figure 7] 5 is a flowchart illustrating an example of the steps of a learning data generation method performed by the abnormal sound detection system according to one embodiment of the present invention. [Figure 8] 10 is a flowchart showing an example of a procedure for generating abnormality level operating status correspondence information by the abnormal sound detection system according to one embodiment of the present invention. [Figure 9] 1 is a block diagram showing a configuration example of an automatic inspection system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, examples of modes for carrying out the present invention (hereinafter referred to as "embodiments") will be described with reference to the accompanying drawings. The present invention is not limited to the embodiments, and various numerical values ​​in the embodiments are merely examples. In addition, in this specification and drawings, the same components or components having substantially the same functions are given the same reference numerals, and duplicated explanations will be omitted.

[0014] <Configuration of abnormal sound detection system> First, the configuration of an abnormal sound detection system (an example of an inspection target information acquisition system) according to this embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing an example of the overall configuration of an abnormal sound detection system 100.

[0015] The abnormal sound detection system 100 calculates the degree of abnormality of an inspection object based on information on sounds (operating sounds) emitted from the inspection object such as a motor, pump, compressor, turbine, boiler, etc. operating in a building such as a plant. Buildings in which inspection objects are installed include, for example, power plants, chemical plants, steel plants, and other plants, substations, etc.

[0016] The abnormal sound detection system 100 generates abnormality level operating state correspondence information D15 (see FIG. 6) in which information on the abnormality level of the inspection object and information on the operating state of the plant are associated and visualized, and displays the information on the display unit 42 of the monitoring terminal 4. In this embodiment, the abnormal sound detection system 100 generates learning data based on information on the sound of the inspection object collected during the trial run of the plant, and calculates the abnormality level of the inspection object using the learning data during the actual operation of the plant. Note that the timing of generating the learning data is not limited to the trial run of the plant, and may be any timing, such as the timing when the operating state is changed by implementing a PoC (Proof of Concept) or the like.

[0017] As shown in FIG. 1, the abnormal sound detection system 100 includes a wireless sensor 1, a learning terminal 2, a wireless receiver 3, and a monitoring terminal 4.

[0018] [Wireless sensor configuration] The wireless sensor 1 includes a microphone unit 5, a power supply unit 10, an analysis unit 11, and a wireless communication unit 12. Each unit included in the wireless sensor 1 is stored in a waterproof and dustproof housing (not shown).

[0019] The microphone unit 5 (an example of an information acquisition unit) is composed of, for example, a microphone and an A / D (Analog / Digital) converter (both not shown). The A / D conversion unit samples and quantizes the amplitude of an analog signal corresponding to the sound of the inspection object 200 picked up by the microphone at a predetermined period, converts the analog signal into a digital value, and outputs it as recording data. The recording data (an example of status information of the inspection object) is output to the feature extraction unit 111 during the trial operation period, and is output to the learning data storage unit 112 of the learning terminal 2 after the start of regular operation.

[0020] The power supply unit 10 is formed of, for example, a secondary battery, and supplies power to each component of the wireless sensor 1 .

[0021] The analysis unit 11 includes a feature extraction unit 111 , a learning data storage unit 112 , and anomaly level / cluster information calculation unit 113 .

[0022] The feature extraction unit 111 (an example of a second feature extraction unit) extracts features of the recorded data generated by the microphone unit 5. For example, the feature extraction unit 111 extracts a power spectrum of the recorded data as the feature, and outputs information on the extracted feature (power spectrum) to the anomaly level / cluster information calculation unit 113.

[0023] The learning data storage unit 112 stores the learning data generated in the learning terminal 2 during the test run period.

[0024] The anomaly level / cluster information calculation unit 113 generates cluster information by clustering the features input from the feature extraction unit 111 using a Gaussian mixture model, Principal Component Analysis (PCA), or other such methods. The cluster information is information indicating which cluster the feature (power spectrum) belongs to. Note that the cluster information is also associated with information on the recording date and time of the recording data from which the cluster is generated.

[0025] Furthermore, the abnormality degree / cluster information calculation unit 113 calculates information on the abnormality degree of the inspection object 200, based on the learning data stored in the learning data storage unit 112 and information on the feature amount extracted by the feature amount extraction unit 111. The abnormality degree is a value indicating the amount of deviation (degree of deviation) of the feature amount extracted by the feature amount extraction unit 111 from the learning data of the sound recording data collected during normal operation of the plant.

[0026] Here, an example of a method for calculating the degree of anomaly by the anomaly level / cluster information calculation unit 113 will be described. First, the anomaly level / cluster information calculation unit 113 randomly samples a plurality of feature amounts of the sound recording data indicated by the learning data and feature amounts of the sound recording data picked up by the microphone unit 5. Next, the anomaly level / cluster information calculation unit 113 inputs each pair of sampled feature amounts individually into a Gaussian mixture model or the like, and calculates a score for each pair.

[0027] Then, the anomaly level / cluster information calculation unit 113 obtains the degree of deviation of the feature extracted by the feature extraction unit 111 from the learning data of the sound recording data during normal operation as the anomaly level by calculating the difference between the calculated scores. Note that the information on the anomaly level is also associated with information on the recording date and time of the sound recording data from which the feature was extracted.

[0028] The wireless communication unit 12 transmits the cluster information and anomaly level information generated by the anomaly level / cluster information calculation unit 113 to the monitoring terminal 4 via the wireless receiver 3. In this embodiment, the recorded data itself is not transmitted from the wireless communication unit 12 of the wireless sensor 1 to the wireless receiver 3, so that the amount of communication data can be reduced, and therefore the communication bandwidth can be saved. Therefore, according to this embodiment, it is no longer necessary to set up a separate line for data communication or to expand the bandwidth of an existing line.

[0029] Furthermore, in this embodiment, the power consumption of the wireless sensor 1 can be reduced by not transmitting the recorded data itself from the wireless communication unit 12 of the wireless sensor 1 to the wireless receiver 3. This makes it possible for the system to operate using only the internal power supply unit 10 without requiring an external power supply. Therefore, users of the abnormal sound detection system 100 can easily introduce the system.

[0030] [Configuration of learning device] The learning terminal 2 includes a data storage unit 21 and a learning data creation unit 22. The data storage unit 21 stores operating state information D1 and recording data D2.

[0031] The operating status information D1 is information that indicates the transition of the operating status during the trial run period of the plant. Specifically, it is information that associates the operating statuses such as "operating status A" to "operating status C" shown in Fig. 6 with information on the date and time when the plant was operating in that operating status. This information is obtained in advance from the plant manager (user of the abnormal sound detection system 100). The audio recording data D2 is audio recording data output from the microphone unit 5.

[0032] The learning data creation unit 22 includes a feature extraction unit 221 , an operating state linking unit 222 , a learning data creation processing unit 223 , and a learning data writing unit 224 .

[0033] The feature extraction unit 221 (an example of a first feature extraction unit), like the feature extraction unit 111 of the wireless sensor 1, extracts, for example, a power spectrum of the sound recording data as a feature of the sound recording data D2. The feature extraction unit 221 also clusters the extracted features and generates cluster information. The feature extraction unit 221 outputs the generated cluster information to the operating status linking unit 222.

[0034] The operating state linking unit 222 links the cluster input from the feature extraction unit 221 with the operating state information read from the data storage unit 21. The operating state linking unit 222 has in advance information on the correspondence between each operating state (such as "operating state A" to "operating state C" shown in FIG. 6) indicated in the operating state information and a cluster, and links the cluster with the operating state information by classifying the cluster input from the feature extraction unit 221 into one of the clusters associated with the operating state.

[0035] If linking is possible, the operating status linking unit 222 determines that the sound recording data corresponding to the cluster for which linking is possible is normal sound recording data, selects the feature of the sound recording data, and outputs it to the learning data creation processing unit 223. On the other hand, if linking is not possible, the operating status linking unit 222 determines that the cluster is a cluster when an abnormality occurs, and does not output the feature of the sound recording data corresponding to the cluster to the learning data creation processing unit 223.

[0036] The learning data creation processing unit 223 creates learning data by learning the feature amount input from the operating state linking unit 222. The learning data creation processing unit 223 outputs the created learning data to the learning data writing unit 224.

[0037] The learning data writing unit 224 writes the learning data input from the learning data creation processing unit 223 into the learning data storage unit 112 of the wireless sensor 1 .

[0038] The wireless receiver 3 includes a wireless communication unit 31, and the wireless communication unit 31 transmits the anomaly level information and cluster information transmitted from the wireless communication unit 12 of the wireless sensor 1 to the monitoring terminal 4. Note that when the physical distance between the wireless sensor 1 and the monitoring terminal 4 is long, a wireless repeater or the like may be provided between the wireless sensor 1 and the wireless receiver 3 as necessary.

[0039] [Monitoring terminal configuration] The monitoring terminal 4 uses the abnormality degree information and cluster information transmitted from the wireless sensor 1 to generate abnormality degree operating state correspondence information D15 that associates the abnormality degree with the operating state of the plant, and displays the abnormality degree operating state correspondence information on the display unit 42.

[0040] The monitoring terminal 4 monitors the state of the inspection target object 200 based on the information transmitted from the wireless sensor 1. The monitoring terminal 4 also generates abnormality level operating state correspondence information D15 using the information transmitted from the wireless sensor 1, and displays it on the display unit 42. The abnormality level operating state correspondence information D15 will be described in detail with reference to FIG. 6 described later.

[0041] The monitoring terminal 4 includes a data management unit 41 and a display unit 42. The data management unit 41 includes a wireless communication unit 411, a data storage unit 412, and a data publishing unit 413.

[0042] The wireless communication unit 411 receives the data (information on the degree of abnormality and cluster information) transmitted from the wireless sensor 1 via the wireless receiver 3, and stores the received data in the data storage unit 412.

[0043] Cluster information D11, operating status information D12, abnormality level information D13, and other information D14 are stored in the data storage unit 412. The cluster information D11 is cluster information transmitted from the wireless sensor 1.

[0044] The operation status information D12 is information showing the transition of the operation information during the actual operation period of the plant. The operation status information D1 and the operation status information D12 stored in the data storage unit 21 of the learning terminal 2 have the same information, except that the corresponding periods (trial operation period or actual operation period) are different.

[0045] The abnormality degree information D13 is the degree of abnormality transmitted from the wireless sensor 1, i.e., information on the degree of abnormality calculated by the abnormality degree / cluster information calculation unit 113. As described above, the information on the degree of abnormality is associated with information on the recording date and time of the sound recording data from which the feature amount from which the degree of abnormality was calculated was extracted.

[0046] The other information D14 is composed of the identification number or accompanying information such as a drawing of the inspection object 200, map information within the plant, etc. The information written in the other information D14 is displayed on the display unit 42 in response to an operation by the user, etc.

[0047] The data disclosure unit 413 generates information on the trend of the abnormality degree using the information on the abnormality degree read from the abnormality degree information D13. The trend of the abnormality degree is information on the transition of the abnormality degree. The data disclosure unit 413 also generates the abnormality degree operation state correspondence information D15 by associating each abnormality degree in the trend of the abnormality degree with the operation state information and the date and time information contained in the information. The data disclosure unit 413 then displays the generated abnormality degree operation state correspondence information D15 on the display unit 42.

[0048] Furthermore, when the degree of abnormality read from the abnormality degree information D13 is not associated with any operating state, i.e., is associated with an "abnormal state", the data disclosure unit 413 notifies the display unit 42 of an alert. This is because in such a case, there is a high possibility that an abnormality has occurred in the inspection target object 200 which is the source of the recorded data for which the degree of abnormality was calculated. The alert on the display unit 42 can be issued, for example, by displaying a message such as "An abnormal state has been detected" on a menu bar (not shown) of the monitoring terminal 4, or by displaying a pop-up message on the display unit 42 when the worker logs in, etc.

[0049] [Wireless sensor installation example] Next, an example of attaching the wireless sensor 1 will be described with reference to Fig. 2. Fig. 2 is a diagram illustrating an example of attaching the wireless sensor 1 to an inspection target 200. In the example shown in Fig. 2, the microphone unit 5 of the wireless sensor 1 is attached so as to be in close contact with a portion of the inspection target 200 that produces operating sounds (near the rotation mechanism). The wireless sensor 1 can be fixed to the inspection target 200 using, for example, a belt or clay. It is preferable to select a position for attaching the microphone unit 5 where the surrounding environmental sounds are minimized.

[0050] 2 shows an example in which the microphone unit 5 is attached closely to the inspection target 200, but the present invention is not limited to this. The microphone unit 5 may be provided in the vicinity of the inspection target 200. Also, different wireless sensors 1 may be attached to different parts of one inspection target 200.

[0051] [Computer hardware configuration] Next, the hardware configuration for realizing the functions of the control system of each device (wireless sensor 1, learning terminal 2, monitoring terminal 4) that constitutes abnormal sound detection system 100 shown in Fig. 1 will be described with reference to Fig. 3. Fig. 3 is a block diagram showing an example of the hardware configuration of each device that constitutes abnormal sound detection system 100.

[0052] 3, the computer 50 includes a CPU (Central Processing Unit) 501, a ROM (Read Only Memory) 502, a RAM (Random Access Memory) 503, a non-volatile storage 504, and a network interface 505. The components constituting the computer 50 are connected to each other via a bus B so as to be able to communicate with each other.

[0053] The CPU 501 reads out from the ROM 502 software programs that realize the functions of the devices that make up the abnormal sound detection system 100, and loads the programs into the RAM 503 and executes them. That is, the functions of the parts that make up the abnormal sound detection system 100 are realized by the CPU 501 executing the programs. Note that instead of the CPU, other processing devices such as an MPU (Micro Processing Unit) may be used.

[0054] The ROM 502 is used as an example of a computer-readable non-transitory recording medium that stores a program executed by the computer 50. The ROM 502 stores an operating system (OS), various parameters, a program for causing the computer 50 to function, and the like.

[0055] Variables, parameters, and the like generated during the arithmetic processing of the CPU 501 are temporarily written to the RAM 503. The variables, parameters, and the like written to the RAM 503 are read by the CPU 501 as appropriate.

[0056] The non-volatile storage 504 is composed of a hard disk drive (HDD), a solid state drive (SSD), a flexible disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a non-volatile memory, etc., and stores various data. The functions of the data storage unit 21 of the learning terminal 2 and the data storage unit 412 of the monitoring terminal 4 are realized by the non-volatile storage 504.

[0057] The program may be stored in the non-volatile storage 504. When the program is stored in the non-volatile storage 504, the non-volatile storage 504 is used as an example of a computer-readable non-transitory recording medium that stores a program executed by a computer.

[0058] The network interface 505 is configured, for example, by a network interface card (NIC). The network interface 505 controls the transmission and reception of various communication data between the network interface 505 and other components constituting the abnormal sound detection system 100 via a line connected to the NIC. The functions of the wireless communication unit 12 of the wireless sensor 1 and the wireless communication unit 411 of the monitoring terminal 4 are realized by the network interface 505.

[0059] The computer 50 that realizes the functions of the monitoring terminal 4 is provided with a display device such as an LCD (Liquid Crystal Display) or an organic EL (Electro Luminescence) display, and an operation input device such as a mouse and a keyboard (both not shown). The display device and the operation input device may be a touch panel in which a panel as the display device and a touch sensor as the operation input device are integrally formed. The function of the display unit 42 of the monitoring terminal 4 is realized by the display device.

[0060] [Examples of power spectra for each operating state] Next, an example of the power spectrum of the recorded data recorded in each operating state of the plant will be described with reference to Fig. 4. Fig. 4 is a diagram showing an example of the power spectrum of the recorded data in each operating state.

[0061] FIG. 4A shows the power spectrum of the recorded data in "operating state A", and FIG. 4B shows the power spectrum of the recorded data in "operating state B". FIG. 4C shows the power spectrum of the recorded data in "operating state C", and FIG. 4D shows the power spectrum of the recorded data in an "abnormal state". In each of FIG. 4A to FIG. 4D, the vertical axis indicates the loudness (dB) of the recorded data, and the horizontal axis indicates the frequency (Hz). As shown in FIG. 4A to FIG. 4D, the loudness and frequency waveform of the recorded data differ for each operating state.

[0062] [Association process between cluster and operating status by the operating status association unit] Next, the process of linking a cluster with an operating state by the operating state linking unit 222 of the learning terminal 2 will be described with reference to Fig. 5. Fig. 5 is a graph showing an example of the correspondence between a cluster and an operating state when the cluster and the operating state are linked by the operating state linking unit 222 of the learning terminal 2.

[0063] The vertical and horizontal axes of the graph shown in Fig. 5 indicate the coordinates of a two-dimensional space in which each feature amount is plotted. In the example shown in Fig. 5, "cluster C1" plotted near the upper left of the graph is associated with "operational state A," and "cluster C2" plotted in the area to the right of it is associated with "operational state B." In addition, "cluster C3" plotted below clusters C1 and C2 is associated with "operational state C," and "cluster C4" plotted in the lower right area of ​​the graph is associated with "abnormal state."

[0064] Cluster C4 is a cluster in which features that could not be associated with any of the operating states A to C are plotted. If an abnormal state occurs during the trial run period (while the learning data is being generated by the learning data creation unit 22 (see FIG. 1)) and cluster information of the features of the recording data in the abnormal state can be obtained, a category of "abnormal state" can be set in advance as one of the operating states to which the cluster is linked.

[0065] [Example of abnormality level operating status response information display] Next, a display example of the abnormality level operational state corresponding information D15 on the display unit 42 will be described with reference to Fig. 6. Fig. 6 is a diagram showing a display example of the abnormality level operational state corresponding information D15. The vertical axis of Fig. 6 indicates the magnitude (level) of the abnormality level, and the horizontal axis indicates time.

[0066] In the graph shown in Fig. 6, the transition of the abnormality degree is shown by a line graph, and each operating state is shown by a color (pattern in the figure) against the background of the line graph. In the example shown in Fig. 6, the operating state transitions in the order of "A" → "B" → "C" → "B" → "C" → "A" → "C" → "abnormal state", and when the operating state is "abnormal state", it is shown that the abnormality degree is high.

[0067] By displaying the abnormality level operation status correspondence information D15 in such a graph, the user can confirm the information on the transition of the abnormality level within a predetermined period. For example, if the user confirms that the abnormality level has been increasing, the user can make a plan to replace the inspection target with an inventory item or to purchase a new one.

[0068] In addition, because the transition of the operation status can be confirmed along with the information on the transition of the abnormality level, the user can understand even if the operation status transition occurs in a different order from the originally scheduled order. Therefore, the user can investigate the cause of the change in the order of the operation status and take measures.

[0069] [Method of acquiring information on inspection objects using an abnormal sound detection system] Next, a method for acquiring information about an inspection target by the abnormal sound detection system 100 according to this embodiment will be described with reference to Fig. 7 and Fig. 8. Fig. 7 is a flowchart showing an example of the steps of a learning data generation method by the abnormal sound detection system 100, and Fig. 8 is a flowchart showing an example of the steps of a process for generating abnormality level operating status correspondence information D15 by the abnormal sound detection system 100.

[0070] [Learning data generation method] First, a method for generating learning data by the abnormal sound detection system 100 will be described with reference to FIG. First, the microphone unit 5 (see FIG. 1) of the wireless sensor 1 picks up the operating sound of the inspection target 200 and generates recorded data (step S1).

[0071] Next, the feature extraction unit 221 of the learning terminal 2 extracts features from the recording data generated in step S1 (step S2). Next, the feature extraction unit 221 clusters the features extracted in step S2 to generate cluster information (step S3). Next, the operation status linking unit 222 of the learning terminal 2 determines whether or not the cluster generated in step S3 can be linked to the operation status information D1 read from the data storage unit 21 (step S4).

[0072] If it is determined in step S4 that linking is possible (if step S4 is determined as YES), the learning data creation processing unit 223 selects the feature amount of the sound recording data from which the cluster was generated (step S5). Next, the learning data creation processing unit 223 generates learning data using the feature amount of the sound recording data selected in step S5 (step S6). Next, the learning data writing unit 224 writes the learning data generated in step S6 to the learning data storage unit 112 of the wireless sensor 1 (step S7). After the process of step S7, the learning data generation process by the abnormal sound detection system 100 ends.

[0073] On the other hand, if it is determined in step S4 that the cluster and the operating state information D1 are not linked (if step S4 is judged as NO), the operating state linking unit 222 determines whether the cluster is an unlearned cluster, that is, a cluster that has not been learned (step S8). If it is determined in step S8 that the cluster is an unlearned cluster (if step S8 is judged as YES), it is determined whether new operating state information corresponding to the unlearned cluster exists (step S9). Whether a corresponding new operating state exists is determined, for example, by whether the operating state corresponding to the recording time of the sound recording data at which the cluster was generated in the operating state information D1 is an operating state that has not been learned (not associated with any cluster) so far.

[0074] If it is determined in step S9 that corresponding new operating status information exists (if the determination in step S9 is YES), the learning data creation processing unit 223 performs the processes in steps S5 and after. That is, the learning data creation processing unit 223 selects the feature amount of the corresponding sound recording data, creates learning data using the feature amount of the selected sound recording data, and writes the learning data to the wireless sensor 1.

[0075] On the other hand, if it is determined in step S9 that there is no corresponding operating state (if step S9 is a NO judgment), or if it is determined in step S8 that the cluster is not an unlearned cluster (if step S8 is a NO judgment), the learning data creation processing unit 223 determines that an abnormality has occurred (step S10). After the process of step S10, the learning data creation processing unit 223 does not generate learning data, and the learning data generation process by the abnormal sound detection system 100 ends.

[0076] That is, when it is an unlearned cluster and there is no corresponding new operating state, or when it is a learned cluster but cannot be associated with an operating state, the learning data creation processing unit 223 determines that an abnormality has occurred in the inspection target object 200. Then, the learning data creation processing unit 223 does not use the sound recording data at the time of the abnormality occurrence in creating the learning data.

[0077] [Generation process of abnormality level operating status correspondence information] Next, a process for generating the abnormality level-operating state correspondence information D15 by the abnormal sound detection system 100 will be described with reference to Fig. 8. First, the microphone unit 5 (see Fig. 1) of the wireless sensor 1 picks up the operating sound of the inspection target 200 and generates recorded data (step S11).

[0078] Next, the feature extraction unit 111 of the wireless sensor 1 extracts features from the sound recording data generated in step S11 (step S12). Next, the anomaly level / cluster information calculation unit 113 of the wireless sensor 1 clusters the features extracted in step S12 to generate cluster information (step S13).

[0079] Next, the anomaly degree / cluster information calculation unit 113 calculates the degree of anomaly of the feature extracted in step S12 by using the learning data stored in the learning data storage unit 112 (step S14). That is, the anomaly degree / cluster information calculation unit 113 calculates the amount of deviation of the feature extracted in step S12 from the learning data under normal conditions as the degree of anomaly.

[0080] Next, the wireless communication unit 12 of the wireless sensor 1 transmits the cluster information and the abnormality level information generated in step S14 to the monitoring terminal 4 via the wireless receiver 3 (step S15). Next, the wireless communication unit 411 of the monitoring terminal 4 stores the cluster information and the abnormality level information received from the wireless receiver 3 in the data storage unit 412 (step S16).

[0081] Next, the data publishing unit 413 of the monitoring terminal 4 associates the abnormality degree and the operation state read from the data storage unit 412, and generates abnormality degree-operation state correspondence information D15 (see FIG. 6) (step S17). Next, the data publishing unit 413 judges whether or not there is an abnormality degree associated with the abnormal state (step S18). If it is judged in step S18 that there is an abnormality degree associated with the abnormal state (if the judgment in step S18 is YES), the data publishing unit 413 notifies the operator of an alert (step S19).

[0082] After the process of step S19, or if the result of step S18 is NO, the data disclosure unit 413 displays the abnormality level operating state correspondence information D15 on the display unit 42 (see FIG. 1) (step S20). After the process of step S20, the abnormal sound detection system 100 ends the process of generating the abnormality level operating state correspondence information D15.

[0083] In the above-described embodiment, the abnormal sound detection system 100 as an example of an inspection object information acquisition unit includes a learning data creation unit 22 that creates learning data by learning features of the sound recording data (status information of the inspection object) picked up by the microphone unit 5, an analysis unit 11 that calculates the degree of abnormality of the inspection object 200 based on the learning data and the features of the sound recording data picked up by the microphone unit 5, and a data disclosure unit 413 that generates abnormality degree-operation status correspondence information D15 that matches the abnormality degree information with operation status information that shows the transition of the operation status of the plant having the inspection object 200, and displays it on the display unit 42.

[0084] Therefore, according to this embodiment, it becomes possible to generate correspondence information (abnormality level-operating state correspondence information D15) between information on the inspection object 200 (information on the degree of abnormality) and information on the operating state of the plant. This enables the user of the abnormal sound detection system 100 to more accurately grasp information on the equipment of the plant by checking the abnormality level-operating state correspondence information D15.

[0085] Moreover, the degree of abnormality indicated in the abnormality degree-operating condition correspondence information D15 is information indicating the amount of deviation of the feature amount of the sound recording data picked up by the microphone unit 5 from the learning data. Therefore, the user of the abnormal sound detection system 100 can grasp the degree of abnormality occurring in the inspection target object 200.

[0086] Furthermore, in the above-described embodiment, information on the degree of abnormality of the inspection target 200 is automatically generated by the analysis unit 11 of the wireless sensor 1. Therefore, the inspection target 200 can be inspected without an operator having to go to the site, so that human intervention can be minimized and the inspection target 200 can be appropriately inspected even when there is a shortage of operators.

[0087] Furthermore, in the above-described embodiment, the analysis unit 11 of the wireless sensor 1 automatically generates information on the degree of abnormality of the inspection target 200, so that even if an abnormality occurs in the inspection target 200 and an abnormal sound is generated during a time period when inspection by an operator is not being performed, the information can be reliably acquired and information on the degree of abnormality corresponding to the abnormal sound is generated. Therefore, the monitoring terminal 4 of the abnormal sound detection system 100 can reliably acquire information on the abnormality that has occurred in the inspection target 200 without missing any information.

[0088] In the above-described embodiment, the operating status linking unit 222 of the learning terminal 2 determines that the sound recording data corresponding to the cluster that has been linked to the operating status information is normal sound recording data and outputs it. Then, the learning data creation processing unit 223 generates learning data using the feature amount of the sound recording data input from the operating status linking unit 222. Therefore, according to this embodiment, the learning data that the anomaly level / cluster information calculation unit 113 of the wireless sensor 1 refers to when generating the anomaly level is made using only sound recording data in normal times, so that the anomaly level / cluster information calculation unit 113 can accurately calculate the anomaly level.

[0089] Furthermore, in the above-described embodiment, the learning data can be always kept up to date by creating learning data by the learning terminal 2 not only during the trial operation period but also when the type of product produced by the plant is changed, etc. Therefore, according to this embodiment, the user of the abnormal sound detection system 100 can easily respond to changes in the equipment and products (outputs) of the plant without manually performing additional learning or changing or modifying the program for additional learning.

[0090] In the above-described embodiment, the information on the degree of abnormality calculated by the anomaly degree / cluster information calculation unit 113 of the wireless sensor 1 may be recorded in the form of an anomaly information database, etc. By performing such processing, it becomes possible to perform analysis of the type of abnormal state of the inspection target 200, the occurrence scene, etc., and therefore it becomes possible to perform a more sophisticated inspection of the inspection target 200.

[0091] [Automatic inspection system configuration] The abnormal sound detection system 100 described above can also be incorporated into an automatic inspection system. The automatic inspection system is a system that inspects the inspection object 200 based on information such as the measured temperature of the inspection object 200, the current value of the drive current, and photographed images of the inspection object 200. Fig. 9 is a block diagram showing an example of the configuration of an automatic inspection system 300.

[0092] 2 includes wireless sensors 1A, 1', and 1B. The wireless sensors 1A and 1' are wirelessly connected to a wireless repeater 10A, and the wireless sensor 1B is wirelessly connected to a wireless repeater 10B. The wireless repeaters 10A and 10B are connected to a wireless receiver 3', and the wireless receiver 3' is connected to a monitoring terminal 4' via a network N.

[0093] Wireless sensor 1A includes a microphone unit 5A, which picks up the operating sound of inspection object 200A, which is constituted by a motor, a pump, or the like, to generate recorded data. Wireless sensor 1B includes a microphone unit 5B, which picks up the operating sound of inspection object 200B, which is constituted by a motor, a pump, or the like, to generate recorded data. The functions of wireless sensors 1A and 1B are the same as those of wireless sensor 1 shown in FIG. 1, so a description thereof will be omitted here.

[0094] The wireless sensor 1' includes a camera unit 6, which captures an image of a portion of the inspection target 200C that displays values, such as an analog meter or counter, and outputs the captured image to the wireless sensor 1'. The wireless sensor 1' analyzes the captured image input from the camera unit 6, reads the value of the analog meter, converts it into a digital value, and transmits the digital read value to the wireless repeater 10A.

[0095] The monitoring terminal 4' performs the same process as the monitoring terminal 4 shown in Fig. 1. The monitoring terminal 4' may calculate the degree of abnormality using information on the read value of the inspection object 200C in addition to information on the recorded data acquired by the microphone unit 5A or 5B. By performing such a process, it becomes possible to detect information on an abnormality in the equipment more accurately. The monitoring terminal 4' may also display information on the read value of the inspection object 200C in the abnormality degree operating state correspondence information D15 (Fig. 6).

[0096] In addition to the above-mentioned processing, the monitoring terminal 4' also performs abnormality detection processing of the read values ​​of the inspection object 200C and alarm notification processing when an abnormality is detected, generates and displays information on trends of the read values, and monitors the inspection object 200C based on the read values.

[0097] In the example shown in Fig. 9, an information acquisition unit such as a microphone unit or a camera unit is provided for one inspection object, but the present invention is not limited to this. For example, multiple information acquisition units may be provided for one inspection object, or one information acquisition unit may be provided for multiple inspection objects. For example, the microphone unit 5A in Fig. 9 may acquire information on both inspection objects 200A and 200B.

[0098] Furthermore, the types of the information acquisition unit are not limited to the two, that is, the microphone unit and the camera unit, shown in Fig. 9. For example, a pressure sensor, a color sensor, a flow rate sensor, a weight sensor, a voltage sensor, a frequency sensor, an illuminance sensor, or the like may be provided as the information acquisition unit.

[0099] In the automatic inspection system 300 shown in Fig. 9, a plurality of (two) wireless repeaters 10A and 10B are connected to one wireless receiver 3', and a plurality of wireless sensors 1A, 1', and 1B are connected to the wireless repeaters 10A and 10B, respectively. Then, information acquired from each of the inspection objects 200A to 200C is transmitted to the monitoring terminal 4' periodically or irregularly. In other words, a plurality of packets are sent from the plurality of wireless sensors 1A, 1', and 1B to the monitoring terminal 4' at any time.

[0100] Generally, wireless communication networks for monitoring installed in plants and the like have stricter limitations on communication capacity and communication speed than general mobile communication services. However, according to the automatic inspection system 300 shown in Fig. 9, even in such a wireless communication environment, each piece of information obtained from the inspection objects 200A to 200C can be stably transmitted to the monitoring terminal 4'. Therefore, according to this embodiment, the monitoring terminal 4' can efficiently and reliably monitor the inspection objects 200A to 200C from a remote location.

[0101] It should be noted that the present invention is not limited to the above-described embodiment, and various other applications and modifications are possible without departing from the gist of the present invention as set forth in the claims.

[0102] For example, the above-described embodiments provide detailed and specific descriptions of the configurations of the devices (wireless sensor, learning terminal, wireless receiver, monitoring terminal) and systems (abnormal sound detection system) in order to clearly explain the present invention, and are not necessarily limited to those having all of the configurations described.

[0103] In addition, in Fig. 1, the control lines or information lines shown by solid lines or solid arrows are those considered necessary for explanation, and do not necessarily show all control lines or information lines in the product. In reality, it can be considered that almost all components are connected to each other.

[0104] In addition, in this specification, the processing steps describing chronological processing include not only processing that is performed chronologically in the order described, but also processing that is not necessarily performed chronologically but is performed in parallel or individually (for example, parallel processing or processing by objects). [Explanation of symbols]

[0105] 1...wireless sensor, 2...learning terminal, 3...wireless receiver, 4...monitoring terminal, 5...microphone unit, 10...power supply unit, 11...analysis unit, 21...data storage unit, 22...learning data creation unit, 41...data management unit, 42...display unit, 100...abnormal sound detection system, 111...feature extraction unit, 112...learning data storage unit, 113...anomaly level / cluster information calculation unit, 222...operation status linking unit, 223...learning data creation processing unit, 224...learning data writing unit, 300...automatic inspection system, 412...data storage unit, 413...data publication unit

Claims

1. a learning data creation unit that creates learning data by learning feature quantities of the state information of the inspection object acquired by the information acquisition unit; an analysis unit that calculates an abnormality degree of the inspection object based on the learning data and a feature amount of the state information of the inspection object acquired by the information acquisition unit; a data disclosure unit that generates abnormality level operation status information correspondence information in which the information on the abnormality level corresponds to operation status information indicating a transition of an operation status of a plant having the inspection object, and displays the abnormality level operation status information on a display unit; The learning data creation unit: a first feature extraction unit that extracts features of state information of the inspection object, clusters the features, and generates cluster information indicating a correspondence between the generated clusters and the features; an operation status linking unit that links the clusters generated by the first feature amount extraction unit with the operation status information, and outputs only cluster information of the clusters that have been linked with the operation status information; a learning data creation processing unit that creates the learning data by using the feature amount indicated in the cluster information input from the operation state linking unit, The analysis unit is a learning data storage unit in which the learning data created by the learning data creation unit is stored; a second feature extraction unit that extracts a feature of state information of the inspection object and outputs the extracted feature; an anomaly degree / cluster information calculation unit that performs clustering on the feature amounts output from the second feature amount extraction unit, generates cluster information indicating a correspondence between the generated clusters and the feature amounts, and calculates and outputs the anomaly degree based on the learning data stored in the learning data storage unit, the degree of abnormality is information indicating a deviation amount of a feature amount of state information of the inspection object acquired by the information acquisition unit from the learning data, The data disclosure unit associates the information on the degree of anomaly with the operation status information based on information on the acquisition date and time of the state information of the inspection object, which is included in the cluster information input from the degree of anomaly / cluster information calculation unit, and information on the date and time of each operation status, which is included in the operation status information. Inspection object information acquisition system.

2. The operation status linking unit determines that the status information of the inspection object corresponding to the cluster that has been linked with the operation status information is normal status information. The inspection target information acquisition system according to claim 1 .

3. The operation state linking unit has correspondence information between each operation state indicated in the operation state information and the cluster, and links the cluster with the operation state information by classifying the new cluster generated by the first feature amount extraction unit into one of the clusters associated with the operation state. The inspection target information acquisition system according to claim 2 .

4. A learning data creation unit creates learning data that has learned feature amounts of the state information of the inspection object acquired by the information acquisition unit; A step of an analysis unit calculating an abnormality degree of the inspection object based on the learning data and feature amounts of state information of the inspection object acquired by the information acquisition unit; a step of generating abnormality degree operation status information correspondence information in which the information on the abnormality degree corresponds to operation status information showing a transition of an operation status of a plant having the inspection object, and displaying the information on the abnormality degree on a display unit by a data disclosure unit, The learning data creation unit: a first feature extraction unit that extracts features of state information of the inspection object, clusters the features, and generates cluster information indicating a correspondence between the generated clusters and the features; an operation status linking unit that links the clusters generated by the first feature amount extraction unit with the operation status information, and outputs only cluster information of the clusters that have been linked with the operation status information; a learning data creation processing unit that creates the learning data by using the feature amount indicated in the cluster information input from the operation state linking unit, The analysis unit is a learning data storage unit in which the learning data created by the learning data creation unit is stored; a second feature extraction unit that extracts a feature of state information of the inspection object and outputs the extracted feature; an anomaly degree / cluster information calculation unit that performs clustering on the feature amounts output from the second feature amount extraction unit, generates cluster information indicating a correspondence between the generated clusters and the feature amounts, and calculates and outputs the anomaly degree based on the learning data stored in the learning data storage unit, the degree of abnormality is information indicating a deviation amount of a feature amount of state information of the inspection object acquired by the information acquisition unit from the learning data, The data disclosure unit associates the information on the degree of anomaly with the operation status information based on information on the acquisition date and time of the state information of the inspection object, which is included in the cluster information input from the degree of anomaly / cluster information calculation unit, and information on the date and time of each operation status, which is included in the operation status information. Method for obtaining information on inspection objects.

Images