Machine status detection device

The machine state detection device addresses the limitations of existing technologies by using feature vectors and classification IDs to identify and link damage sections, enabling accurate detection of machine states and reducing unexpected damage through timely maintenance.

JP7591129B2Active Publication Date: 2024-11-27HITACHI CONSTRUCTION MACHINERY CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
JP2023511237
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Priority Date
2021-04-02
Filing Date
2022-03-28
Publication Date
2024-11-27
Estimated Expiration
2042-03-28

AI Technical Summary

Technical Problem

Existing machine state detection technologies fail to accurately identify unknown operations that may not cause immediate damage and do not account for damage detection before abnormalities occur, relying on preset recognition methods that cannot handle unforeseen states or motions.

Method used

A machine state detection device that collects and processes state information over time, using feature vectors and classification IDs to identify machine states, detect damage, and link damage sections with classification IDs, allowing for the detection of unknown operations and damage states.

Benefits of technology

Enables easy detection of machine states, including unknown operations, by linking damage sections with classification IDs, facilitating timely maintenance and reducing the frequency of unexpected damage conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007591129000001
    Figure 0007591129000001
  • Figure 0007591129000002
    Figure 0007591129000002
  • Figure 0007591129000003
    Figure 0007591129000003
Patent Text Reader

Abstract

Provided is a machine state detecting device capable of easily detecting a state of a machine, irrespective of whether an unknown operation that is not assumed to cause damage has been performed. A state detecting device 100 is provided with: a collecting device 1a for collecting state information 1f indicating a state of the machine that varies as a time series, at a plurality of time points; and an arithmetic processing device 1b for detecting the state of the machine by processing the plurality of items of state information 1f collected at the plurality of time points. The arithmetic processing device 1b includes: a classification processing portion 1i which classifies a feature quantity vector, having a type of the state information 1f as a feature quantity, into any of a plurality of clusters, and assigns a classification ID; a detection processing portion 1j which detects damage suffered by the machine on the basis of state information 1g indicating a state of the machine that varies as a time series; and an association processing portion 1l which associates a sector in which damage was detected by the detection processing portion 1j with the classification ID assigned by the classification processing portion 1i in said sector.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a device for detecting a state of a machine. [Background technology]

[0002] The following Patent Documents 1, 2, and 3 are related technologies in this technical field. Patent Document 1 describes a device that learns the operating tendency of a machine when an abnormality occurs in the machine. Patent Document 2 describes a detection device that detects a section on a time axis where stress applied to a machine occurs, or a section at a position where the machine was located. Patent Document 3 describes a damage estimation device that estimates the amount of damage a machine has received from the recognition results of the operation performed by the machine. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2018-97616 A [Patent Document 2] Patent No. 6215446 [Patent Document 3] Patent No. 6458055 Summary of the Invention [Problem to be solved by the invention]

[0004] The techniques described in the above Patent Documents 1 to 3 have the following problems. (a) Patent Document 1 does not describe a means for determining whether the machine's operation when an abnormality occurs is a known operation or an unknown operation, and does not describe an output method. In addition, since the machine's operation tendency is learned using an alarm indicating an abnormality in the machine, the patent document does not describe detecting damage (cause of the abnormality) that the machine suffered before the alarm indicating the abnormality is generated. (b) In Patent Document 2, the section where damage (repeated stress) occurred before causing an abnormality (breakage of the machine) is detected, and the cause of the damage is identified using a position state recognition means. However, the position state recognition means used here recognizes a preset state, and there is no mention of what kind of processing should be performed when an unpredictable state occurs. (c) In Patent Document 3, a table is created in advance that links the action recognition result with the amount of damage the machine will receive when that action is performed. Then, during operation, the action performed by the machine is recognized by the action recognition method, and the amount of damage is estimated based on the recognition result and the amount of damage corresponding to that action. However, the action recognition method used here recognizes actions that are preset, as in (b) above, and does not state what kind of processing should be performed when an action that is not preset is performed.

[0005] The present invention has been made in consideration of the above, and aims to provide a machine condition detection device that can easily detect the machine condition regardless of whether or not an unknown operation that is not expected to cause damage has been performed. [Means for solving the problem]

[0006] In order to solve the above problem, the machine state detection device of the present invention is a machine state detection device comprising: a collection device that collects status information indicating a time-series changing state of the machine at a plurality of times; and an arithmetic processing device that processes the plurality of status information collected at the plurality of times to detect the state of the machine, wherein the arithmetic processing device comprises: a classification processing unit that time-divides the plurality of status information collected by the collection device to identify a plurality of feature vectors having a type of the status information as a feature, classifies the identified feature vectors into one of a plurality of predetermined sets, and assigns a classification ID to the classified feature vector that identifies the set to which the feature vector belongs; a detection processing unit that detects damage suffered by the machine based on the status information indicating the time-series changing state of the machine collected by the collection device; and a linking processing unit that links a section in which the damage was detected by the detection processing unit to the classification ID assigned by the classification processing unit in that section. Effect of the Invention

[0007] According to the present invention, the state of the machine can be easily detected regardless of whether or not an unknown action that is not expected to cause damage has been performed. Problems, configurations and effects other than those described above will become apparent from the following description of the embodiments. [Brief description of the drawings]

[0008] [Figure 1] FIG. 2 is a processing block diagram of the state detection device according to the first embodiment. [Diagram 2] FIG. 4 is a process flow diagram of the state detection device. [Diagram 3] FIG. 4 is an explanatory diagram of repeated stress amplitude and wavelength. [Figure 4] An example of information recorded in the classification model database. [Diagram 5] Example of sensor and control information for state classification. [Figure 6] An example of the results of the classification process. [Figure 7] An explanatory diagram of classification ID patterns. [Figure 8] Example of information recorded in the Damage Status Classification ID Database (part 1). [Figure 9] Another example of cutting out damaged sections. [Figure 10] Example of information recorded in the Damage Status Classification ID Database (part 2). [Figure 11] Example of information recorded in the Damage Status Classification ID Database (part 3). [Figure 12] FIG. 11 is a processing block diagram of a state detection device according to a second embodiment. [Figure 13] An illustration of the cluster distribution of classification IDs. [Figure 14] FIG. 11 is a processing block diagram of a state detection device according to a third embodiment. [Figure 15] Example of information recorded in the Damage Status Classification ID Database (part 4). [Figure 16] An illustration of keyframe animation. [Figure 17] Example of information recorded in the Damage Status Classification ID Database (part 5). [Figure 18] FIG. 13 is a processing block diagram of a state detection device according to a fourth embodiment. [Figure 19] FIG. 13 is a processing block diagram of a state detection device according to a fifth embodiment. [Figure 20] An example of cumulative damage. [Figure 21] An example of the change in cumulative damage over time. [Figure 22] 13 is an example of visualization of the integrated value of individual damage amounts performed in the state detection device of embodiment 6. [Figure 23] An example of visualizing the number of occurrences of a damage state. [Figure 24] 13 shows an example (part 1) of a display screen in the state detection device of embodiment 7. [Diagram 25] Example of display screen (part 2). [Figure 26] Example of display screen (part 3). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, embodiments of the present invention will be described with reference to the drawings. Note that components with the same reference numerals in each embodiment have the same functions in each embodiment unless otherwise specified, and the description thereof will be omitted.

[0010] [Embodiment 1] A state detection device 100 according to the first embodiment will be described with reference to FIGS. 1 to 11. FIG. 1 is a processing block diagram of the state detection device 100 of the first embodiment. FIG. 2 is a processing flow diagram of the state detection device 100. FIG. 3 is an explanatory diagram of repeated stress amplitude and wavelength. FIG. 4 is a diagram showing an example of information recorded in a classification model database 1h. FIG. 5 is a diagram showing an example of sensor control information 1f for state classification. FIG. 6 is a diagram showing an example of a processing result of the classification process 1i. FIG. 7 is an explanatory diagram of a classification ID pattern. FIG. 8 is a diagram showing an example (part 1) of information recorded in a damage state classification ID database 1o.

[0011] The state detection device 100 is a device that detects the state of various machines, including work machines such as construction machines or industrial robots. In particular, the state detection device 100 is a device that detects the state of a machine that changes over time (for example, the operating state of a machine). 1c, 1d, and 1e in FIG. 1 show examples of machines to which the state detection device 100 is applicable. 1c is a transport vehicle such as a dump truck. 1d is a vehicle-based construction machine such as a shovel. 1e is an industrial robot used in manufacturing plants and the like. The state detection device 100 is not limited to these, and can be applied to various machines.

[0012] The condition detection device 100 can be roughly divided into a collection device 1a that collects information for detecting the machine condition, and a calculation processing device 1b that performs processing for detecting the machine condition using the collected information. That is, the condition detection device 100 includes the collection device 1a that collects condition information indicating the machine condition that changes over time at multiple times, and the calculation processing device 1b that processes the multiple pieces of condition information collected at multiple times to detect the machine condition.

[0013] The collecting device 1a is, for example, a device that stores various sensor information provided in a machine, and is configured by a controller mounted on the machine. Each of the collecting device 1a and the arithmetic processing device 1b may be configured by a general-purpose or industrial computer, or may be configured by dedicated hardware using a microcomputer. Each of the collecting device 1a and the arithmetic processing device 1b is configured to include a CPM, a ROM, and a RAM, and the CPU executes a program stored in the ROM to realize each function. The arithmetic processing device 1b may suck up the information collected by the collecting device 1a to an external cloud computer using a communication line, and process the information using the cloud computer. In FIG. 1, the collecting device 1a and the arithmetic processing device 1b are illustrated separately, but the state detection device 100 may be configured as a single device by accommodating the collecting device 1a and the arithmetic processing device 1b in a single housing.

[0014] The collecting device 1a includes a sensor and control information collecting unit for state classification 1f that collects sensor and control information for state classification 1f as state information indicating the state of the machine that changes over time, and a sensor and control information collecting unit for damage detection 1g that collects sensor and control information for damage detection 1g. The sensor and control information for state classification 1f is sensor measurement information used to classify the state of the machine and information for controlling each part of the machine. The sensor and control information for damage detection 1g is sensor measurement information used to detect damage received by the machine and information for controlling each part of the machine. The sensor and control information for state classification 1f and the sensor and control information for damage detection 1g may be the same type of information or different types of information. In this embodiment, the sensor and control information for state classification 1f is also referred to as "state information 1f". In this embodiment, the sensor and control information for damage detection 1g is also referred to as "state information 1g".

[0015] The collecting device 1a can collect the sensor measurement information included in the status information 1f and the status information 1g by directly acquiring the sensor measurement information from a strain gauge or an angle sensor attached to the machine, or by acquiring information flowing through a network such as a controller area network (CAN) mounted on the machine. The collecting device 1a can collect the control information included in the status information 1f and the status information 1g by directly acquiring the control information from a control device of the machine, or by acquiring information flowing through a network mounted on the machine.

[0016] The arithmetic processing device 1b includes a detection processing unit 1j that executes a detection process 1j of damage received by the machine, a determination processing unit 1k that executes a determination process 1k of the amount of damage received by the machine, a classification processing unit 1i that executes a classification process 1i that classifies the state of the machine at each time based on the state information 1f collected by the collection device 1a, a linking processing unit 1l that executes a linking process 1l between the processing result of the detection process 1j (and the determination process 1k) and the processing result of the classification process 1i, a determination processing unit 1m that executes a determination process 1m of whether the classification ID linked by the linking process 1l is an unknown classification ID, a display processing unit 1n that executes a display process 1n that displays that the classification ID linked by the linking process 1l is an unknown classification ID, and a registration processing unit 1p that executes a registration process 1p that registers the classification ID determined to be an unknown classification ID by the determination process 1m in the classification ID database 1o. Furthermore, the arithmetic processing device 1b includes a damage state classification ID database 1o and a classification model database 1h as a recording device. For ease of explanation, the configuration of each part in the collection device 1a and the arithmetic processing device 1b and the processing in each part are denoted by the same reference numerals.

[0017] The arithmetic processing device 1b has a function (detection processing unit 1j) for executing a detection process 1j of damage suffered by the machine as described above. The detection processing unit 1j performs a process for detecting damage suffered by the machine based on the status information 1g collected by the collection device 1a. The detection process 1j detects a section (period) on the time axis in which the damage suffered by the machine is large. The damage targeted by the detection process 1j is, for example, damage caused by stress fatigue damage suffered by each part of the machine, such as a hydraulic cylinder or a hydraulic motor. The damage targeted by the detection process 1j also includes, in addition to physical damage such as stress fatigue damage, situations that pose a safety problem, such as a situation where the machine is out of balance. A situation where the machine is out of balance can be detected by attaching an inclinometer to the machine.

[0018] As a damage detection process 1j, there is a method described in Patent Document 2. This method is a method for detecting the start and end times of repeated stress using stress values ​​measured by a strain gauge attached to a dump truck. FIG. 3 shows an example of detection process 1j using this method. The horizontal axis of FIG. 3 indicates time. The vertical axis of FIG. 3 indicates stress (3a), stress amplitude (3b), and fatigue damage rate (3c) along with their respective time axes. 3h in FIG. 3 is a stress waveform, which is a graph of the measured value of a strain gauge attached to a predetermined position of the machine. Fatigue damage depends on the magnitude of the amplitude of repeated stress, not the absolute value of stress. 3h in FIG. 3 shows the absolute value of stress, and it is not possible to know where the large fatigue damage occurred as it is. Therefore, in detection process 1j, the amplitude and wavelength of repeated stress are detected from this stress waveform 3h, and the section where the large fatigue damage occurred is detected. This method is a method for detecting the wavelength of stress by improving the rainflow method, which is a type of stress cycle counting method. In Figure 3, 3i and 3j show the amplitude and wavelength of the repeated stress detected from the stress waveform in 3h using this method. The rectangles in 3i and 3j show the detected amplitude and wavelength of the stress. The length along the vertical axis of the rectangles in 3i and 3j shows the magnitude of the amplitude of the repeated stress. The length along the horizontal axis of the rectangles in 3i and 3j shows the wavelength of the repeated stress. The time corresponding to the apex on the left side of the base of the rectangles in 3i and 3j shows the start time of the repeated stress. The time corresponding to the apex on the right side of the base of the rectangles in 3i and 3j shows the end time of the repeated stress.

[0019] The loading of soil into a dump truck is performed in the following sequence: the dump truck moves in an empty state (3d), soil is loaded into the stopped dump truck in several batches (3e), the dump truck moves in a loaded state (3f), and the loaded soil is discharged (3g). 3j in FIG. 3 shows the amplitude and wavelength of the repeated stress that the dump truck receives when soil is loaded into the dump truck (3e). 3i in FIG. 3 shows the amplitude and wavelength of the repeated stress that the dump truck receives when the dump truck discharges soil (3g). With this method, the detection process 1j can detect the magnitude of damage that the machine has received (the magnitude of the amplitude of the repeated stress) and the section in which the machine has received damage (the period from the start time to the end time of the amplitude of the repeated stress). 3c in FIG. 3 shows the fatigue damage that the machine receives per unit time (fatigue damage rate). This fatigue damage rate 3c is calculated by calculating various fatigue damages per unit time from various repeated stresses based on the amplitude and wavelength of the repeated stress shown in 3b, and integrating these fatigue damages per unit time. This allows the detection process 1j to detect the magnitude of damage at each time.

[0020] The above method described in Patent Document 2 is an example of a method for detecting the magnitude of damage to the machine and the section in which the machine is damaged. Other methods may be used for the detection process 1j as long as they can detect the magnitude of damage to the machine and the section in which the machine is damaged.

[0021] In this embodiment, the magnitude of damage suffered by the machine (the magnitude of the amplitude of the repeated stress) is also referred to as the "damage amount." In this embodiment, the section on the time axis in which the machine is damaged (the period from the start time to the end time of the damage) is also referred to as the "damage section." In this embodiment, the state of the machine when it is damaged (damage section) is also referred to as the "damage state." In this embodiment, the amount of damage suffered by the machine in the damage section is also referred to as the "individual damage amount."

[0022] Furthermore, the arithmetic processing device 1b has a function (determination processing unit 1k) for executing a determination process 1k of the amount of damage received by the machine as described above. The determination processing unit 1k performs a process for determining whether or not the amount of damage detected by the detection process 1j is an amount of damage that will affect the lifespan of each part of the machine. The damage amount determination process 1k determines that the amount of damage detected by the detection process 1j is an amount of damage that will affect the lifespan if it is equal to or greater than a predetermined threshold, and determines that the amount of damage will not affect the lifespan if it is less than the threshold. If the amount of damage detected by the detection process 1j is an amount of damage that will not affect the lifespan, a process for acquiring the status information 1g and the status information 1f is performed (corresponding to "NO" in 2a of FIG. 2). On the other hand, if the amount of damage detected by the detection process 1j is an amount of damage that will affect the lifespan, a linking process 1l to be described later is performed (corresponding to "YES" in 2a of FIG. 2).

[0023] The arithmetic processing device 1b may not set a threshold in the judgment process 1k, and may perform the linking process 1l for all the damage amounts detected by the detection process 1j. For example, if the damage amount is small but the number of occurrences of the damage state is high, it may be difficult to set a threshold for determining the damage amount that affects the lifespan. In such a case, the arithmetic processing device 1b may not perform the judgment process 1k, and may perform the linking process 1l for all the damage amounts detected by the detection process 1j.

[0024] Furthermore, the arithmetic processing device 1b has a function (classification processing unit 1i) for executing a classification process 1i for classifying the machine status at each time based on the status information 1f collected by the collection device 1a as described above. The classification processing unit 1i performs a process of time-dividing the multiple pieces of status information 1f collected by the collection device 1a at multiple times and identifying multiple feature vectors having the types of the status information 1f as features. The classification processing unit 1i then performs a process of classifying the identified feature vector into one of multiple predetermined sets (clusters) and assigning a classification ID to the classified feature vector, which identifies the set (cluster) to which the feature vector belongs.

[0025] Specifically, the classification process 1i is performed based on information previously recorded in the classification model database 1h. In the classification model database 1h, the items (types) of the state information 1f are recorded as items of feature vectors (vector components defined as features) (if the state classification sensor / control information has n items, the feature vectors are n-dimensional feature vectors), and the feature vectors indicating the states and the classification IDs that identify the states are recorded in advance in association with each other for each of the various states of the machine.

[0026] FIG. 4 shows an example of information recorded in the classification model database 1h. 4a in FIG. 4 shows the items of feature vectors from the sensor A to the operation amount Z. 4b in FIG. 4 shows the classification ID. The classification model database 1h can be constructed, for example, by using a clustering method. The classification process 1i chronologically divides a plurality of pieces of learning state information 1f, which is collected by the collection device 1a at a plurality of times and in which various states of the machine are recorded, to identify a plurality of feature vectors. Then, the classification process 1i classifies the identified plurality of feature vectors into one of a plurality of clusters using a clustering method, and assigns a classification ID. For example, a method such as k-means may be used as the clustering method. The k-means is a method of grouping similar feature vectors into a number of clusters according to a preset division number, and returning a classification ID that identifies each cluster. As a result, the feature vectors of the state information 1f are classified according to the set division number, and the classification ID of the cluster to which the classified feature vector belongs is assigned. The feature vector divides the representative state of the machine by a set division number m, and shows the characteristics of each state. In the example of Figure 4, the feature vector with the classification ID "3" is (M 31 , M 32 , ,M 3n ) There are m types of classification IDs, which is the number of divisions. In this embodiment, the feature vectors recorded in the classification model database 1h shown in FIG. 4 are also referred to as "feature vectors of classification models."

[0027] The classification ID indicates one state (or scene) of the machine. For example, if the machine is a dump truck, the classification ID indicates a state in which the dump truck's vessel is raised and earth is being discharged, or if the machine is a shovel, the classification ID indicates a state in which the front arm is raised and excavation is being performed. The state of the machine is indicated by a feature vector. The classification model database 1h records the feature vector indicating the state of the machine in association with the classification ID that identifies the feature vector. The classification model database 1h of the first embodiment does not record natural language indicating the state of the machine, such as "a state in which the dump truck's vessel is raised and earth is being discharged" explained as an example, in direct association with the classification ID.

[0028] The classification process 1i performs a process of searching for the classification ID of a feature vector that is most similar to the feature vector of the status information 1f of the classification target among the feature vectors recorded in the classification model database 1h. Similar means that the distance (e.g., Euclidean distance) between the feature vectors is closer than others. The classification process 1i searches for the feature vector that is closest to the feature vector of the classification target among the feature vectors recorded in the classification model database 1h, and sets the classification ID of the searched feature vector as the classification ID that indicates the state of the machine at the time when the status information 1f on which the feature vector of the classification target is based was collected.

[0029] FIG. 5 shows an example of state information 1f on which the feature vector to be classified is based. The state information 1f at "time T3" in FIG. 5 is (F 13 , F 23 , , F m3 ) This (F 13 , F 23 , , F m3 5 constitutes a feature vector having each item (type) of the status information 1f at "time T3" as a feature. In this embodiment, the feature vector of the status information 1f shown in FIG. 5 is also referred to as the "feature vector of collected data."

[0030] The classification process 1i searches for a feature vector of a classification model that is most similar to the feature vector of the collected data at each time shown in FIG. 5, and obtains a classification ID of the feature vector of the searched classification model. In the process of searching for a similar feature vector, the distance (e.g., Euclidean distance) between the feature vector of the collected data and the feature vector of the classification model may be calculated as the similarity between the feature vector of the collected data and the feature vector of the classification model. Then, the classification process 1i may obtain a classification ID of the feature vector of the classification model with the closest distance. In calculating the distance between the feature vector of the collected data and the feature vector of the classification model, in addition to the Euclidean distance, the Mahalanobis distance may be calculated, and various distance calculation methods can be used. Through such processing, a sequence of classification IDs arranged in chronological order as shown in FIG. 6 is obtained as the processing result of the classification process 1i.

[0031] Furthermore, the arithmetic processing device 1b has a function (correlation processing unit 1l) for executing a linking process 1l between the processing result of the detection process 1j (and the determination process 1k) and the processing result of the classification process 1i, as described above. The linking processing unit 1l performs a process for linking a damage section, which is a section in which damage is detected by the damage detection process 1j, with a classification ID assigned to the damage section by the classification process 1i. This enables the arithmetic processing device 1b to identify the damage state, which is the state of the machine in the damage section, by the classification ID assigned by the classification process 1i.

[0032] FIG. 7 shows a diagram in which the processing results of detection process 1j (and judgment process 1k) and the processing results of classification process 1i are arranged in chronological order. 7a in FIG. 7 shows the processing results of detection process 1j (and judgment process 1k), and shows the amplitude and wavelength of the repeated stress, similar to 3j in FIG. 3. 7d in FIG. 7 shows the start time of the repeated stress, and 7e in FIG. 7 shows the end time of the repeated stress. 7b and 7c shown in the middle and bottom of FIG. 7 show the processing results of classification process 1i. Numbers in parentheses such as "(3)" shown in the middle and bottom of FIG. 7 indicate the value of the classification ID.

[0033] The processing result of the classification process 1i is obtained as a sequence of classification IDs arranged in time series. There are two types of sequence of classification IDs arranged in time series. One type is a type in which the sequence of classification IDs in the damage section between the start time 7d and the end time 7e is composed of the same classification ID, as shown in 7b of FIG. 7. In the example of FIG. 7, the sequence of classification IDs in the section between the start time 7d and the end time 7e is composed of all classification IDs (5). The other type is a type in which the sequence of classification IDs in the damage section between the start time 7d and the end time 7e is composed of a characteristic pattern, as shown in 7c of FIG. 7. In the example of FIG. 7, the sequence of classification IDs in the damage section is composed of the pattern of classification IDs (6), (6), (4), (7), (7), and (7). The type in which the sequence of classification IDs is composed of the same classification ID is observed when there are few machine operations. The type in which the sequence of classification IDs is composed of a pattern is observed when the machine performs an operation corresponding to the pattern. Since a change in the state of the machine appears as a change in the classification ID, the sequence of classification IDs has a pattern corresponding to the state of the machine. Therefore, the sequence of classification IDs can identify the damage state, which is the state of the machine in the damaged section.

[0034] Furthermore, the arithmetic processing device 1b has a function (determination processing unit 1m) for executing a determination process 1m of whether or not the classification ID linked by the linking process 1l is an unknown classification ID, as described above. The determination processing unit 1m performs a process of determining whether or not the classification ID linked by the linking process 1l is an unknown classification ID, based on the classification ID recorded in the damage state classification ID database 1o.

[0035] The determination process 1m performs different processing depending on the type of the above-mentioned classification ID column obtained as a processing result of the classification process 1i. If the above-mentioned classification ID column is of a type consisting of the same classification ID, the determination process 1m refers to a record in the classification ID database 1o in which only one classification ID is recorded. In the example of 7b in FIG. 7, if the classification ID (5) linked to the damaged section is recorded in the classification ID database 1o, the determination process 1m determines that this linked classification ID is a known classification ID (not an unknown classification ID). On the other hand, if the classification ID (5) linked to the damaged section is not recorded in the classification ID database 1o, the determination process 1m determines that this linked classification ID is an unknown classification ID.

[0036] If the above-mentioned classification ID string is of a type that is configured with a pattern, the determination process 1m refers to a record of the classification ID database 1o in which items such as those shown in FIG. 8 are recorded. 8a in FIG. 8 shows an item of the classification ID pattern. 8b in FIG. 8 shows an item of the ID that identifies the classification ID pattern of 8a. The determination process 1m performs a similarity determination between the classification ID pattern linked to the damaged section such as that shown in 7c in FIG. 7 and the classification ID pattern recorded in the classification ID database 1o such as that shown in 8a in FIG. For this similarity determination, a pattern recognition process using DP matching (Dynamic Programming matching) may be used.

[0037] DP matching is a pattern matching method using dynamic programming, and is a method that takes elastic matching of patterns into consideration. Elastic matching is a method in which, for example, in 7c of FIG. 7, the occurrence of classification IDs (6) and (7) multiple times in succession is considered to be "stretched", and the effect of "stretching" is eliminated, and matching processing is performed assuming that classification IDs (6) and (7) each appear once. Therefore, the classification ID pattern shown in 7c of FIG. 7 and the classification ID pattern shown in #1 of 8a of FIG. 8 are recognized as the same pattern. Note that DP matching can also calculate the similarity. If all classification IDs included in the pattern match, the similarity is 100%, and if only some classification IDs included in the pattern match, the similarity is lower than 100%. The determination process 1m may determine that the classification ID pattern linked to the damaged section and the classification ID pattern recorded in the classification ID database 1o are the same pattern if the similarity is not 100% but is somewhat high. If the pattern of the classification ID linked to the damaged section is identical to the pattern of the classification ID recorded in the classification ID database 1o, the determination process 1m determines that the linked classification ID is a known classification ID (not an unknown classification ID). On the other hand, if the pattern of the classification ID linked to the damaged section is not identical to the pattern of the classification ID recorded in the classification ID database 1o, the determination process 1m determines that the linked classification ID is an unknown classification ID.

[0038] If it is determined that the linked category ID by the linking process 1l is a known category ID, a process for acquiring status information 1g and status information 1f is performed (corresponding to "NO" in 2b of FIG. 2). On the other hand, if it is determined that the linked category ID by the linking process 1l is an unknown category ID, a display process 1n and a registration process 1p, which will be described later, are performed (corresponding to "YES" in 2b of FIG. 2).

[0039] Furthermore, the arithmetic processing device 1b has a function (display processing unit 1n) of executing a display process 1n that displays that the classification ID linked by the linking process 1l is an unknown classification ID, as described above. The fact that the classification ID linked by the linking process 1l is an unknown classification ID means that an unknown damage state has occurred. The display processing unit 1n performs a process of displaying information indicating that an unknown damage state has occurred on a display device such as a display, or issuing a report by an audio output device such as a speaker. The arithmetic processing device 1b may omit executing the display process 1n.

[0040] Furthermore, the arithmetic processing device 1b has a function (registration processing unit 1p) for executing a registration process 1p for registering a classification ID determined to be an unknown classification ID by the determination process 1m in the classification ID database 1o as described above. In the example of 7b in FIG. 7, if the classification ID (5) is an unknown classification ID, the registration processing unit 1p performs a process for registering the classification ID (5) in the classification ID database 1o. In the example of FIG. 8, the registration processing unit 1p performs a process for registering an unknown classification ID pattern in the section 8a written as "undefined" in #3 in FIG. 8, and registering an ID that identifies the pattern in the section 8b. When registering an unknown classification ID pattern in the section 8a, a classification ID that has appeared multiple times in succession is registered as having appeared once.

[0041] As described above, the state detection device 100 of the first embodiment includes a collection device 1a that collects state information 1f indicating the state of the machine that changes over time at multiple times, and a calculation processing device 1b that processes the multiple state information 1f collected at multiple times to detect the state of the machine. The calculation processing device 1b includes a classification processing unit 1i that time-divides the multiple state information 1f collected by the collection device 1a, classifies a feature vector having the type of the state information 1f as a feature into one of multiple clusters (sets), and assigns a classification ID. Furthermore, the calculation processing device 1b includes a detection processing unit 1j that detects damage suffered by the machine based on the state information 1g indicating the state of the machine that changes over time collected by the collection device 1a. Furthermore, the calculation processing device 1b includes a linking processing unit 1l that links a section in which damage is detected by the detection processing unit 1j with a classification ID assigned to the section by the classification processing unit 1i.

[0042] With such a configuration, the state detection device 100 of the first embodiment can specify the damage state, which is the state of the machine in the section where the machine is damaged, by the classification ID assigned by the classification processing unit 1i. As a result, the state detection device 100 can easily detect that a damage state has occurred in the damage section linked to the classification ID simply by referring to the classification ID assigned by the classification processing unit 1i. Furthermore, if the state detection device 100 records the classification ID that specifies the damage state, it can easily determine whether or not the detected damage state is an unknown state that is not expected to cause damage, based on the recorded classification ID. Therefore, the state detection device 100 of the first embodiment can easily detect the state of the machine, regardless of whether or not an unknown operation that is not expected to cause damage has been performed.

[0043] Furthermore, in the condition detection device 100 of the first embodiment, the arithmetic processing device 1b includes a classification ID database 1o in which classification IDs that identify a damage state are pre-recorded. Furthermore, the arithmetic processing device 1b includes a determination processing unit 1m that determines whether or not a classification ID linked by the linking processing unit 1l based on a classification ID recorded in the classification ID database 1o is an unknown classification ID. Furthermore, the arithmetic processing device 1b includes a registration processing unit 1p that registers a classification ID determined to be an unknown classification ID by the determination processing unit 1m in the classification ID database 1o.

[0044] With such a configuration, the state detection device 100 of the first embodiment can easily expand the classification ID database 1o in which classification IDs that specify the damage state are recorded, and can more reliably and easily determine whether the detected damage state is an unknown state. Furthermore, the state detection device 100 can reduce the frequency with which an unknown damage state occurs, making it easier to accurately detect the current state of the machine. Therefore, the state detection device 100 of the first embodiment can easily detect the state of the machine regardless of whether an unknown operation that is not expected to cause damage has been performed, and can easily maintain the soundness of the machine, such as by making it easier to perform maintenance of the machine in a timely manner.

[0045] FIG. 9 is a diagram showing another example of cutting out a damaged section.

[0046] In the above description, an example in which the detection process 1j detects damage based on the amplitude and wavelength of the repeated stress has been described, but the detection process 1j may detect damage based only on the strength of the waveform related to the stress. 9a in FIG. 9 shows an example in which a part of the waveform of the fatigue damage rate shown in 3c in FIG. 3 is cut out. The detection process 1j sets a threshold value 9b that determines the strength of the waveform 9a, and can detect the time 9c at which the waveform 9a changes to equal to or greater than the threshold value 9b and the time 9d at which the waveform 9a changes to less than the threshold value 9b as the start time 9c and end time 9d of the damage section. The detection process 1j may detect the maximum value of the waveform 9a in the damage section as the amount of damage suffered by the machine in the damage section. Alternatively, the detection process 1j may detect the average value of the waveform 9a in the damage section as the amount of damage suffered by the machine in the damage section.

[0047] Fig. 10 is a diagram showing an example (part 2) of information recorded in the damage state classification ID database 1o, and Fig. 11 is a diagram showing an example (part 3) of information recorded in the damage state classification ID database 1o.

[0048] The linking process 1l may not only link the damage section detected by the detection process 1j with the classification ID assigned to the damage section by the classification process 1i, but may also link the amount of damage (individual damage amount) suffered by the machine in the damage section. The classification ID database 1o may record the individual damage amount linked to the classification ID assigned to the damage section by the classification process 1i, as shown in Figs. 10 and 11.

[0049] Fig. 10 shows a case where the above-mentioned classification ID column obtained as a processing result of the classification process 1i is a type that is composed of the same classification ID. 10a in Fig. 10 shows the classification ID item. 10b in Fig. 10 shows the individual damage amount item. As shown in Fig. 10, the individual damage amount item 10b may be recorded separately for each part of the machine.

[0050] FIG. 11 shows a case where the above-mentioned classification ID string obtained as a processing result of the classification process 1i is a type that is configured with a pattern. 11a in FIG. 11 shows an item of the classification ID pattern. 11b in FIG. 11 shows an item of the ID that identifies the classification ID pattern. 11c in FIG. 11 shows an item of the individual damage amount. The item 11b of the individual damage amount may be recorded separately for each part of the machine, as in FIG. 10. In the case of FIG. 11, since the individual damage amount varies within the damage section (see FIG. 9), the average value, maximum value, or cumulative value of the individual damage amount in the damage section may be used as a representative value of the individual damage amount to be linked to the pattern of the classification ID. Naturally, a statistical value other than this average value, etc. may be used as a representative value of the individual damage amount in the damage section.

[0051] With this configuration, the state detection device 100 of the first embodiment can specify not only the damage state by the classification ID given by the classification processing unit 1i, but also the individual damage amount. As a result, the state detection device 100 can easily detect the occurrence of a damage state by simply referring to the classification ID given by the classification processing unit 1i, and can easily detect the amount of damage the machine has received in the damage section. This allows the state detection device 100 to more accurately detect the current fatigue damage level of the machine. Therefore, the state detection device 100 of the first embodiment can easily and accurately detect the state of the machine, and can further maintain the soundness of the machine, such as by performing maintenance of the machine more timely.

[0052] The function of executing the detection process 1j provided in the arithmetic processing device 1b corresponds to an example of the "detection processing unit" described in the claims. The classification model database 1h provided in the arithmetic processing device 1b corresponds to an example of the "second recording unit" described in the claims. The function of executing the classification process 1i provided in the arithmetic processing device 1b corresponds to an example of the "classification processing unit" described in the claims. The function of executing the judgment process 1m provided in the arithmetic processing device 1b corresponds to an example of the "first judgment processing unit" described in the claims. The damage state classification ID database 1o provided in the arithmetic processing device 1b corresponds to an example of the "first recording unit" described in the claims. The function of executing the registration process 1p provided in the arithmetic processing device 1b corresponds to an example of the "first registration processing unit" described in the claims.

[0053] [Embodiment 2] The state detection device 100 of the second embodiment will be described with reference to Figures 12 and 13. In the state detection device 100 of the second embodiment, the description of the same configuration and operation as those of the previous embodiments will be omitted. Fig. 12 is a process block diagram of the state detection device 100 according to the second embodiment. Fig. 13 is an explanatory diagram of the cluster distribution of classification IDs.

[0054] The state detection device 100 of the first embodiment searches for a feature vector that is closest to the feature vector of the classification target identified by the classification process 1i among the feature vectors pre-recorded in the classification model database 1h. Then, the state detection device 100 of the first embodiment assigns the classification ID of the searched feature vector to the feature vector of the classification target as a classification ID indicating the state of the machine at the time when the state information 1f on which the feature vector of the classification target is based was collected.

[0055] In this case, a problem as shown in FIG. 13 may occur. FIG. 13 shows a two-dimensional feature space in which one feature A among the feature vectors is described on the horizontal axis and another feature B among the feature vectors is described on the vertical axis. Although the feature vectors actually used in the present invention have a very large number of dimensions, the number of dimensions is reduced to simplify the explanation and the explanation is given in a two-dimensional feature space. 13a, 13b, 13c, and 13d in FIG. 13 respectively show the position coordinates of the feature vectors of the classification IDs (5), (6), (1), and (3) previously recorded in the classification model database 1h. 13e surrounded by 13g in FIG. 13 shows the position coordinates of the feature vector group of the classification target identified in the newly detected damage section. The position coordinates 13g of the feature vector group are far from the position coordinates of the feature vectors previously recorded in the classification model database 1h, and appear not to belong to any cluster of the classification ID. This is thought to have occurred because the learning section used to construct the classification model database 1h did not include any machine conditions similar to the newly detected damaged section.

[0056] Therefore, the state detection device 100 of the second embodiment has a processing function as shown in Fig. 12. The calculation processing device 1b of the second embodiment has a function (determination processing unit 12a) that executes a determination process 12a for determining whether or not a classification ID of a feature vector similar to a feature vector of a classification target identified by a classification process 1i is recorded in a classification model database 1h. That is, the determination processing unit 12a performs a process of determining whether or not a classification ID to be assigned to the feature vector of a classification target identified by the classification process 1i is recorded in the classification model database 1h, based on the feature vector of the classification target identified by the classification process 1i and the feature vectors recorded in advance in the classification model database 1h.

[0057] As described in the first embodiment, the determination process 12a calculates the distance between the feature vector to be classified and the feature vector recorded in the classification model database 1h. If the calculated distance is equal to or less than a preset threshold, the determination process 12a determines that the classification ID of the feature vector similar to the feature vector to be classified is recorded in the classification model database 1h. That is, in this case, the determination process 12a determines that the classification ID to be assigned to the feature vector to be classified is recorded in the classification model database 1h. Thereafter, the linking process 11 is performed as in the first embodiment. On the other hand, if the calculated distance is greater (farther) than a preset threshold, the determination process 12a determines that the classification ID of the feature vector similar to the feature vector to be classified is not recorded in the classification model database 1h. That is, in this case, the determination process 12a determines that the classification ID to be assigned to the feature vector to be classified is not recorded in the classification model database 1h. Thereafter, in the second embodiment, the generation process 12b and the registration process 12c described later are performed.

[0058] The arithmetic processing device 1b of the second embodiment has a function (generation processing unit 12b) of executing a generation process 12b for generating a new classification model when a classification ID of a feature vector similar to a feature vector to be classified is not recorded in the classification model database 1h. That is, the generation processing unit 12b performs a process of generating a new classification ID when a classification ID to be assigned to a feature vector to be classified is not recorded in the classification model database 1h.

[0059] Furthermore, the arithmetic processing device 1b of the second embodiment has a function of executing a registration process 12c (a registration processing unit 12c) for registering the new classification model generated by the generation process 12b in the classification model database 1h. That is, the registration processing unit 12c performs a process of registering the new classification ID generated by the generation processing unit 12b in the classification model database 1h in association with the feature vector to which the classification ID should be assigned.

[0060] The generation process 12b and the registration process 12c may be performed by either of the following two methods (method 1 or method 2). In method 1, the generation process 12b adds a group of feature vectors to be classified, which are identified in the newly detected damage section, to a learning section for constructing the classification model database 1h, and performs a process of reconstructing the classification model database 1h using a classification method such as clustering. As a result, as shown in 13f of FIG. 13, the generation process 12b generates a representative coordinate value (value of each feature of the representative feature vector) of a representative feature vector representing the group of feature vectors to be classified, which are identified in the newly detected damage section. Then, the generation process 12b generates a new classification ID to be assigned to the representative feature vector having the generated representative coordinate value. The registration process 12c associates the representative feature vector having the generated representative coordinate value with the generated new classification ID and registers them in the classification model database 1h. The registration process 12c registers the generated new classification ID in the classification ID database 1o as a classification ID for identifying a damage state.

[0061] In method 2, the generation process 12b calculates a representative coordinate value of a feature vector group from a feature vector group to be classified that is identified in a newly detected damage section. The representative coordinate value of the feature vector group may be calculated, for example, by calculating a center of gravity coordinate value or an average coordinate value of the feature vector group. The generation process 12b generates a new classification ID to be assigned to the representative feature vector having the calculated representative coordinate value. The registration process 12c associates the representative feature vector having the calculated representative coordinate value with the generated new classification ID and registers them in the classification model database 1h. The registration process 12c registers the generated new classification ID in the classification ID database 1o as a classification ID that identifies a damage state.

[0062] By using the above-described method 1 or method 2, the state detection device 100 of the second embodiment can solve the above problem as described with reference to Fig. 13. Normally, re-learning the classification model database 1h as in method 1 places a large processing load on the arithmetic processing device 1b and takes a long processing time, so method 2 is more effective. When it is necessary to generate a new classification ID, the generation process 12b may perform a hybrid process of methods 1 and 2, in which method 2 is performed the first few times (a number of times equal to or less than a preset threshold value) and, once newly generated classification IDs have been accumulated, method 1 is used to re-learn the classification model database 1h.

[0063] In this way, when a classification ID to be assigned to a feature vector identified by the classification process 1i is not recorded, the state detection device 100 of the second embodiment generates a new classification ID, associates it with the feature vector, and registers it in the classification model database 1h.

[0064] As a result, the state detection device 100 of the second embodiment can easily expand the classification model database 1h in which the feature vectors that specify the state of the machine are recorded, and can classify the state of the machine more reliably and easily. The state detection device 100 can reduce the frequency of recording the classification ID to be assigned to the feature vectors specified by the classification process 1i, making it easier to accurately detect the current state of the machine. Therefore, the state detection device 100 of the second embodiment can easily detect the state of the machine regardless of whether an unknown operation that is not expected to cause damage has been performed, and can easily maintain the soundness of the machine, such as by making it easier to perform maintenance of the machine in a timely manner.

[0065] The function of executing the judgment process 12a provided in the arithmetic processing device 1b corresponds to an example of a "second judgment processing unit" as set forth in the claims. The function of executing the generation process 12b provided in the arithmetic processing device 1b corresponds to an example of a "generation processing unit" as set forth in the claims. The function of executing the registration process 12c provided in the arithmetic processing device 1b corresponds to an example of a "second registration processing unit" as set forth in the claims.

[0066] [Embodiment 3] The state detection device 100 of the third embodiment will be described with reference to Figures 14 to 17. In the state detection device 100 of the third embodiment, the description of the same configuration and operation as those of the previous embodiments will be omitted. Fig. 14 is a processing block diagram of the state detection device 100 of the third embodiment. Fig. 15 is a diagram showing an example (part 4) of information recorded in the damage state classification ID database 1o. Fig. 16 is an explanatory diagram of key frame animation. Fig. 17 is a diagram showing an example (part 5) of information recorded in the damage state classification ID database 1o.

[0067] In the third embodiment, a method for visualizing the state of a machine will be described. The arithmetic processing device 1b of the third embodiment has a function (extraction processing unit 14a) for executing an extraction process 14a for extracting a classification ID given by the classification process 1i in a predetermined section including a section linked to a classification ID by the linking process 1l. That is, the extraction processing unit 14a performs a process for extracting a classification ID given by the classification process 1i in the predetermined section. The classification ID extracted by the extraction process 14a is used when generating an animation in a reproduction process 14b described later. The predetermined section extracted by the extraction process 14a may be the damaged section itself, which is a section linked to a classification ID by the linking process 1l, or may be a section including the time before and after the damaged section.

[0068] For example, as in 7b of FIG. 7, in the case where the above-mentioned classification ID string obtained as a processing result of the classification process 1i is of a type that is composed of the same classification ID, the specified section extracted by the extraction process 14a may be a section that includes the time before and after the damaged section (in the case of 7b of FIG. 7, a section that includes classification IDs (3) and (2) before and after the damaged section). Even in this type, if it is sufficient to reproduce the state of the machine in the damaged section itself, the specified section extracted by the extraction process 14a may be the damaged section itself. In this case, an input device that can be input by a user's operation may be provided to allow the user to select the specified section to be extracted by the extraction process 14a.

[0069] On the other hand, for example, as shown in 7c of FIG. 7, when the string of the above-mentioned classification IDs obtained as a result of the processing of the classification process 1i is a type that is configured with a pattern, the predetermined section cut out by the cutting out process 14a may be the damaged section itself. The cutting out process 14a may integrate the classification IDs that appear multiple times in succession after cutting out the classification IDs in the predetermined section, or may leave the classification IDs that appear multiple times in succession in the cut-out state without integrating them. However, when generating an animation in the reproduction process 14b, the playback speed of the animation is required. When the classification IDs are integrated, it is better to attach time information such as the start time of integration to the classification ID. FIG. 15 shows an example in which an item related to information on the classification ID for animation display is added to the classification ID database 1o. 15a of FIG. 15 shows an item related to information on the classification ID for animation display. FIG. 15 shows an example in which the classification IDs that appear multiple times in succession are integrated.

[0070] The arithmetic processing device 1b of the third embodiment has a function (reproduction processing unit 14b) of executing reproduction processing 14b that generates an animation (moving image) that reproduces the state of the machine in a predetermined section in which the classification ID is extracted by the extraction processing 14a. The reproduction processing unit 14b performs a process of generating an animation (moving image) that reproduces the state of the machine in the predetermined section based on the feature vector corresponding to the classification ID extracted by the extraction processing 14a.

[0071] The reproduction process 14b uses key frame animation used in computer graphics to generate an animation that reproduces the state of a machine in a predetermined section. FIG. 16 shows an example of a generated animation that reproduces the state of a hydraulic excavator, which is one of the machines. The feature vector corresponding to the classification ID extracted by the extraction process 14a includes information on the bending angle of each joint of the front working machine as the state of the machine (excavator). For example, this is information on the bending angle of the joints of the boom, arm, and bucket. The key frame animation is a method of generating this bending angle between classification IDs using an interpolation method. For example, the classification IDs for animation display are classification ID (11) shown in 16f, classification ID (12) shown in 16g, and classification ID (13) shown in 16h in FIG. 16. In an animation that does not use key frame animation, frames 16a, 16b, and 16c that reproduce the states of the excavator corresponding to classification IDs (11), (12), and (13), respectively, are generated and displayed by switching between them on the time axis. For this reason, animations that do not use keyframe animation will be jerky and unable to smoothly reproduce the movements of the shovel. Note that the frames mentioned here are not still images used in cel animation, but still images generated by projecting a 3D shape model generated using 3D computer graphics onto a 2D projection screen.

[0072] Therefore, the reproduction process 14b performs key frame interpolation by generating frames that reproduce the state of the shovel between frames 16a, 16b, and 16c using key frame animation so that the movement of the shovel can be reproduced smoothly. In the example of FIG. 16, the reproduction process 14b performs key frame interpolation for each bending angle of each joint of the shovel. 16i in FIG. 16 indicates the bending angle of the boom, 16j indicates the bending angle of the arm, and 16k indicates the bending angle of the bucket. Each curve shown in 16i, 16j, and 16k in FIG. 16 indicates the result of interpolating the bending angle of each joint between three points 16f, 16g, and 16h in FIG. 16 as the control points of the interpolation function. Various functions can be used as the interpolation function. For example, a spline function. By using these interpolation functions, the reproduction process 14b can smoothly interpolate the bending angle of each joint at the time between the control points of the interpolation function, and generate an animation that can smoothly reproduce the movement of the shovel. 16d and 16e in FIG. 16 show interpolated frames, which are interpolated between frames 16a and 16b, and between frames 16b and 16c. Of course, it is also possible to generate a frame that is interpolated between frames 16a and 16d. In order to reproduce the actual operating speed of the machine, the occurrence time of the classification ID used as the key frame is required. The reproduction process 14b also interpolates the time information of the interpolated frame based on the time information of the key frame. In this way, the reproduction process 14b can reproduce the operating speed of the machine. The reproduction process 14b may also generate a moving image other than an animation.

[0073] The arithmetic processing device 1b of the third embodiment has a function (display processing unit) for executing a display process for displaying the animation (moving image) generated by the reproduction process 14b on a display device. Although the function for executing this display process is not shown in Fig. 14, it may be included in the function for executing the display process 1n, or may be included in the function for executing the reproduction process 14b, for example.

[0074] In this way, the state detection device 100 of the third embodiment extracts the classification ID assigned by the classification process 1i in a predetermined section including a section linked to the classification ID by the linking process 1l. Then, the state detection device 100 generates an animation that smoothly reproduces the state of the machine based on the feature vector corresponding to the extracted classification ID, and displays the animation on a display device.

[0075] As a result, the state detection device 100 of the third embodiment can visually express in an easily understandable manner what state the machine was in in the damaged section, allowing the user to intuitively grasp the damaged state. Also, in the state detection device 100 of the third embodiment, an animation is generated using a feature vector corresponding to a classification ID, so that an animation can be generated using less information than the actually collected state information 1f.

[0076] The state detection device 100 of the third embodiment may generate animation based on the state information 1g, rather than generating animation based on the feature vector corresponding to the extracted classification ID, i.e., the state information 1f constituting the feature vector. The state detection device 100 of the third embodiment may convert the content of the generated animation into text, and register it in the classification ID database 1o as a character string for text display, as shown in 17a of Fig. 17. At this time, the character string for text display 17a is recorded in the classification ID database 1o in association with the information 15a of the classification ID for animation display.

[0077] Moreover, the function of executing the cut-out process 14a provided in the arithmetic processing device 1b corresponds to an example of the "cut-out processing unit" described in the claims. The function of executing the reproduction process 14b provided in the arithmetic processing device 1b corresponds to an example of the "reproduction processing unit" described in the claims. The function of executing the reproduction process 14b or the animation display process included in the function of executing the display process 1n provided in the arithmetic processing device 1b corresponds to an example of the "display processing unit" described in the claims.

[0078] [Embodiment 4] The state detection device 100 of the fourth embodiment will be described with reference to Fig. 18. In the state detection device 100 of the fourth embodiment, the description of the same configuration and operation as those of the previous embodiments will be omitted. FIG. 18 is a processing block diagram of the state detection device 100 according to the fourth embodiment.

[0079] In the fourth embodiment, a method of recognizing a damage state using a damage state classification ID database 1o will be described. The state detection device 100 of the fourth embodiment recognizes a damage state without using the state information 1g. In order to acquire the state information 1g, a sensor that is retrofitted to the machine and a device for acquiring control information are required, but it is possible to recognize a damaged section and a damaged state without using this information.

[0080] The arithmetic processing device 1b of the fourth embodiment has a function (a recognition processing unit 18a) for executing a recognition process 18a for recognizing a damage state specified by a classification ID given by the classification process 1i. The recognition processing unit 18a performs a process for recognizing a damage state specified by the classification ID given by the classification process 1i, based on the classification ID given by the classification process 1i and the classification ID recorded in the classification ID database 1o.

[0081] The recognition process 18a performs a process of searching whether or not the above-mentioned string of classification IDs obtained as a processing result of the classification process 1i is recorded in the classification ID database 1o as a pattern of classification IDs as shown in, for example, 8a in FIG. 17. If the string of classification IDs is a type that is composed of the same classification ID (#3 in FIG. 17), the recognition process 18a may determine whether or not the classification ID is recorded in the classification ID database 1o and perform a search. If the string of classification IDs is a type that is composed of a pattern (#1 or #2 in FIG. 17), the recognition process 18a may perform a search using a pattern matching method such as the above-mentioned DP matching. This allows the recognition process 18a to detect a damage state without using the state information 1g. The recognition process 18a may recognize a damage section by assuming that the time when the first classification ID of the above-mentioned string of classification IDs obtained as a processing result of the classification process 1i is assigned is the start time of the damage state, and the time when the last classification ID is assigned is the end time of the damage state.

[0082] Furthermore, the arithmetic processing device 1b of the fourth embodiment has a function (display processing unit 18b) for executing a display process 18b that displays the recognition result of the damaged state by the recognition process 18a on a display device. If the recognition process 18a can find that the above-mentioned string of classification IDs obtained as a processing result of the classification process 1i is recorded in the classification ID database 1o, the display processing unit 18b can acquire, for example, the classification ID 15a for animation display and the character string 17a for text display in Fig. 17. The display processing unit 18b performs a process of displaying an animation reproducing the damaged state on a display device using the acquired classification ID 15a for animation display and the character string 17a for text display.

[0083] In this way, the state detection device 100 of the fourth embodiment can recognize the damaged state without using the state information 1g. As a result, even if a sensor or the like that acquires the state information 1g breaks down, the state detection device 100 of the fourth embodiment can easily detect the damaged state by the classification ID assigned by the classification processing unit 1i. The state detection device 100 of the fourth embodiment can easily determine whether or not the detected damaged state is an unknown state that is not expected to cause damage, based on the classification ID recorded in the classification ID database 1o. Therefore, the state detection device 100 of the fourth embodiment can easily detect the state of the machine, regardless of whether or not an unknown operation that is not expected to cause damage has been performed.

[0084] The function of executing the recognition process 18a provided in the arithmetic processing device 1b corresponds to an example of the "recognition processing unit" described in the claims. The function of executing the display process 18b provided in the arithmetic processing device 1b corresponds to an example of the "display processing unit" described in the claims.

[0085] [Embodiment 5] The state detection device 100 of the fifth embodiment will be described with reference to Figures 19 to 21. In the state detection device 100 of the fifth embodiment, the description of the same configuration and operation as those of the previous embodiments will be omitted. Fig. 19 is a processing block diagram of the state detection device 100 of the fifth embodiment. Fig. 20 is a diagram showing an example of the accumulated damage degree. Fig. 21 is a diagram showing an example of the time-series change in the accumulated damage degree.

[0086] In the state detection device 100 of the fifth embodiment, the arithmetic processing device 1b has a function (estimation processing unit 19a) of performing a damage amount estimation process 19a based on the processing result of the recognition process 18a of the fourth embodiment. In the classification ID database 1o, the individual damage amounts received by each part of the machine in the damaged section may be recorded in advance in association with the classification ID (or a pattern of classification IDs) as shown in Figs. 10 and 11. When the estimation processing unit 19a recognizes a damage state based on the recognition result of the recognition process 18a, it performs a process of identifying the individual damage amount from the classification ID (e.g., classification ID (1) of #1 in Fig. 10) that identifies the damage state. That is, the estimation processing unit 19a performs a process of identifying the individual damage amount corresponding to the damage state recognized by the recognition process 18a based on the classification ID that identifies the damage state recognized by the recognition process 18a and the classification ID recorded in the classification ID database 1o. The individual damage amount shown in Figs. 10 and 11 is the amount of damage received by the machine due to one event.

[0087] When the damage amount is a measured value of stress or a value that can estimate stress, the estimation process 19a can convert the individual damage amount into a damage degree using an SN diagram. The SN diagram shows the relationship between the repeated stress and the number of times the stress is repeated until the material breaks. The damage degree is calculated from one repeated stress (stress amplitude) using the SN diagram. The cumulative damage degree, which is the cumulative value of the damage degree, indicates that the material will break when it reaches 1. In other words, the damage degree can also be said to be the consumption of the machine's life. Therefore, the estimation process 19a can convert the individual damage amounts shown in Figures 10 and 11 into damage degrees using the SN diagram. The estimation process 19a can obtain the SN diagram in advance by using the FAT class of the International Institute of Welding (IIW) or by actually conducting a life test of the machine. Note that when it is difficult to obtain an SN diagram, the estimation process 19a obtains the tendency of the damage degree from the slope of a general SN diagram. In this case, the estimation process 19a cannot determine the absolute value of the damage degree, but can determine the tendency of the damage degree (whether it is large or small). By calculating the degree of damage, the estimation process 19a can estimate the impact that a single damage state will have on the machine's lifespan, calculate the accumulated damage degree for each part of the machine to estimate the lifespan consumed or remaining lifespan, and predict the lifespan of each part of the machine from the time-series changes in the accumulated damage degree.

[0088] Furthermore, the arithmetic processing device 1b of the fifth embodiment has a function of executing a display process 19b (display processing unit 19b) for displaying the estimation result of the estimation process 19a on a display device. The display processing unit 18b performs processes such as visualizing the influence of one damage state estimated by the estimation process 19a on the lifespan of the machine, visualizing the life consumption or remaining lifespan of each part of the machine, and visualizing the predicted result of the lifespan of each part of the machine.

[0089] FIG. 20 and FIG. 21 are examples of visualization. FIG. 20 shows the cumulative damage degree of each part of the machine. 20f in FIG. 20 shows the line where the cumulative damage degree is 1.0. When the cumulative damage degree reaches 1.0, each part of the machine reaches the end of its life. 20a to 20e in FIG. 20 show the cumulative damage degree of each part of the machine. It can be seen that the cumulative damage degree 20a of part A in FIG. 20 is closest to the line 20f and has the shortest remaining life. The display process 19b may change the display mode of the graph of the cumulative damage degree shown in FIG. 20 according to the remaining life. For example, the display process 19b may change the color of the graph shown in FIG. 20 from green to yellow to red, etc., as the remaining life decreases.

[0090] FIG. 21 shows the time series change in the cumulative damage degree. 21e in FIG. 21 shows the line where the cumulative damage degree is 1.0. 21a to 21c in FIG. 21 show the time series change in the cumulative damage degree of each part of the machine. 21d in FIG. 21 is a line showing the current time. 21f to 21h in FIG. 21 are straight lines showing the prediction results of the time series change in the cumulative damage degree 21a to 21c. The time shown by the intersection of the straight lines 21f to 21h showing the prediction results and the line 21e where the cumulative damage degree is 1.0 shows the time when the life is reached. For example, the part 21a in FIG. 21 is predicted to reach the end of its life at the time 21i shown by the intersection 21k of the straight line 21f showing the prediction results and the line 21e. The display process 19b can display the remaining life time (the time from time 21d to time 21i) which is the time until the life is reached, display information encouraging the ordering of parts for repairs, and change and display the color of the graph shown in Fig. 21 in the same manner as in Fig. 20. Furthermore, the display process 19b can display a drawing of the part when the graph shown in Fig. 20 or 21 is operated by the user (for example, clicked with a mouse, etc.).

[0091] In this way, the state detection device 100 of the fifth embodiment can estimate the amount of individual damage from the processing result of the recognition process 18a. As a result, the state detection device 100 of the fifth embodiment can calculate the damage degree and the cumulative damage degree from the estimated individual damage amount. The state detection device 100 of the fifth embodiment can estimate the effect of one damage state on the life of the machine, estimate the life consumption or remaining life of each part of the machine, and predict the life of the machine from the time-series change in the cumulative damage degree. Therefore, the state detection device 100 of the fifth embodiment can detect not only what state the machine was in in the damaged section, but also how the state of the machine will change in the future from various viewpoints.

[0092] The function of executing the estimation process 19a provided in the arithmetic processing device 1b of the fifth embodiment may also be provided in the arithmetic processing device 1b of the first to third embodiments. For example, the arithmetic processing device 1b of the first to third embodiments may be provided with a function of executing the estimation process 19a to calculate the damage degree and the cumulative damage degree from the individual damage amount, estimate the life consumption or remaining life for each part of the machine, or predict the life of each part of the machine from the time-series change in the cumulative damage degree. In the above-mentioned Patent Document 3, it was possible to calculate the damage degree only for a predetermined damage state, so it was not possible to estimate the damage degree in an unexpected damage state. However, by providing the arithmetic processing device 1b of the first to third embodiments with the execution function of the estimation process 19a, it becomes possible to estimate the damage degree in an unexpected damage state.

[0093] The function of executing the estimation process 19a provided in the arithmetic processing device 1b corresponds to an example of the "estimation processing unit" described in the claims. The function of executing the display process 19b provided in the arithmetic processing device 1b corresponds to an example of the "display processing unit" described in the claims.

[0094] [Embodiment 6] 22 and 23, the state detection device 100 of the sixth embodiment will be described. In the state detection device 100 of the sixth embodiment, the description of the same configuration and operation as those of the previous embodiments will be omitted. Fig. 22 is a diagram showing an example of visualization of the integrated value of the individual damage amount performed in the state detection device 100 of embodiment 6. Fig. 23 is a diagram showing an example of visualization of the number of occurrences of a damage state.

[0095] In the sixth embodiment, a method of identifying and visualizing the state of the machine that is the main cause of the amount of damage received by each part of the machine will be described. The arithmetic processing device 1b of the fifth embodiment has the execution functions of the estimation process 19a and the display process 19b, and can calculate and visualize the damage degree and the accumulated damage degree for each part of the machine from the individual damage amount recorded in the classification ID database 1o as shown in Figs. 10 and 11. The arithmetic processing device 1b of the sixth embodiment also has the execution functions of the estimation process 19a and the display process 19b. However, the arithmetic processing device 1b of the sixth embodiment can calculate and visualize an accumulated value by accumulating the individual damage amount of a certain part recorded in the classification ID database 1o for each classification ID. Then, the arithmetic processing device 1b of the sixth embodiment can estimate the accumulated damage amount by accumulating the multiple accumulated values ​​for each calculated classification ID.

[0096] That is, the estimation process 19a of the sixth embodiment identifies an individual damage amount corresponding to the damage state recognized by the recognition process 18a based on the classification ID that identifies the damage state recognized by the recognition process 18a and the classification ID recorded in the classification ID database 1o. Then, the estimation process 19a of the sixth embodiment calculates an integrated value of the individual damage amount for each classification ID based on the identified individual damage amount and the number of occurrences of the damage state corresponding to the individual damage amount (i.e., the number of times the individual damage amount has been received). Then, the estimation process 19a of the sixth embodiment accumulates the multiple integrated values ​​calculated for each classification ID to estimate the accumulated damage amount received by the machine. Then, the display process 19b of the sixth embodiment displays the processing result of the estimation process 19a on a display device.

[0097] Specifically, the estimation process 19a of the sixth embodiment calculates an integrated value of the individual damage amount for each classification ID by multiplying an individual damage amount associated with a certain classification ID by the number of occurrences of the damage state corresponding to the individual damage amount. The estimation process 19a estimates the cumulative damage amount by calculating the sum of each of the multiple integrated values ​​calculated for each classification ID. Then, the estimation process 19a of the sixth embodiment identifies the damage state that was the main cause of the machine receiving the cumulative damage amount based on the proportion of each of the multiple integrated values ​​estimated for each classification ID to the cumulative damage amount.

[0098] For example, in FIG. 10, it is assumed that the damage state of category ID (1) occurred once, and the damage state of category ID (3) occurred 1000 times. In this case, the cumulative value of the individual damage amount of the part A in category ID (1) is calculated as 1050×1=1050. The cumulative value of the individual damage amount of the part A in category ID (3) is calculated as 5×1000=5000. The cumulative damage amount received by the part A is estimated as 1050+5000=6050. The proportion of the cumulative value of the individual damage amount of 5000 in category ID (3) out of the cumulative damage amount of 6050 received by the part A is greater than the proportion of the cumulative value of the individual damage amount of 1050 in category ID (1). Therefore, it is identified that the damage state that was the main cause of the part A receiving the cumulative damage amount of 6050 is the damage state of category ID (3). This situation is visualized in FIG. 22 and FIG. 23.

[0099] FIG. 22 shows the cumulative amount of individual damage received by part A for each category ID. FIG. 23 shows the number of occurrences of damage states corresponding to the individual amount of damage received by part A for each category ID. FIG. 22 shows that the damage state that was the main cause of the cumulative damage received by part A was the damage state with category ID (3). FIG. 23 shows that the damage state with category ID (3) is the damage state that occurs most frequently, and that the cumulative damage received by part A can be reduced by reducing the number of occurrences of the damage state with category ID (3).

[0100] The estimation process 19a of the sixth embodiment can convert the accumulated damage amount into the accumulated damage degree, as in the fifth embodiment. The estimation process 19a of the sixth embodiment can estimate the life consumption or remaining life for each part of the machine, and predict the life of each part of the machine from the time-series change in the accumulated damage degree, as in the fifth embodiment. In the examples of Figs. 22 and 23, when the life of part A comes before the next maintenance time, it is preferable not to cause the damage state of classification ID (3) to occur too much. In this case, the arithmetic processing device 1b of the sixth embodiment may combine the straight lines 21f to 21h showing the prediction results of the time-series changes 21a to 21c of the accumulated damage degree as shown in Fig. 21 with the method of this embodiment to generate a control command that does not cause the damage state of classification ID (3) to occur too much, and transmit the control command to the control device that controls the operation of the machine. In the examples of Figs. 22 and 23, the classification ID (3) is a damage state classified by a clustering method, and it may be difficult for the user to understand what kind of state it is. In this case, the calculation processing device 1b of embodiment 6 has the function of executing the extraction process 14a and reproduction process 14b of embodiment 3, and can generate an animation that reproduces the damage state of classification ID (3) and display it on a display device.

[0101] In the above description, the estimation process 19a has been described as calculating the integrated value of the individual damage amount for each classification ID, but when the individual damage amount is linked to an ID that identifies a pattern of the classification ID as shown in Fig. 11, the estimation process 19a calculates the integrated value for each ID that identifies a pattern of the classification ID. Also, in Fig. 20, Fig. 22, and Fig. 23, the calculation processing device 1b visualizes using a bar graph, but visualization using other formats such as a pie chart may also be used.

[0102] In this way, the state detection device 100 of the sixth embodiment calculates an integrated value of the individual damage amount for each classification ID, and accumulates the multiple integrated values ​​calculated for each classification ID to estimate the cumulative damage amount received by the machine. As a result, the state detection device 100 of the sixth embodiment can identify the damage state that was the main cause of the cumulative damage amount received by the machine from the proportion of the multiple integrated values ​​estimated for each classification ID to the cumulative damage amount. Therefore, the state detection device 100 of the sixth embodiment can appropriately take measures such as controlling the machine so as to prevent the damage state that was the main cause from occurring, thereby further maintaining the soundness of the machine.

[0103] The function of executing the estimation process 19a and the display process 19b included in the arithmetic processing device 1b of the sixth embodiment may also be included in the arithmetic processing device 1b of the first to third embodiments. For example, the arithmetic processing device 1b of the first to third embodiments may be provided with a function of executing the estimation process 19a that calculates an integrated value of the individual damage amount for each classification ID based on the individual damage amount detected by the detection process 1j and the number of times the individual damage amount is detected (i.e., the number of times the damage state occurs), and accumulates the multiple integrated values ​​calculated for each classification ID to estimate the accumulated damage amount received by the machine. As a result, the arithmetic processing device 1b of the first to third embodiments can appropriately estimate the accumulated damage amount even if an unexpected damage state is the main cause of the accumulated damage amount.

[0104] [Embodiment 7] The state detection device 100 of the seventh embodiment will be described with reference to Figures 24 to 26. In the state detection device 100 of the seventh embodiment, the description of the same configuration and operation as those of the previous embodiments will be omitted. Fig. 24 is a diagram showing an example (part 1) of a display screen in the state detection device 100 of the seventh embodiment. Fig. 25 is a diagram showing an example (part 2) of a display screen. Fig. 26 is a diagram showing an example (part 3) of a display screen.

[0105] In the seventh embodiment, the visualization of the various processing results described in the first to sixth embodiments will be described. The arithmetic processing device 1b of the seventh embodiment has a function of executing the extraction process 14a and the reproduction process 14b, and also has a function of executing a display process for displaying an animation (moving image) generated by the reproduction process 14b on a display device. The display process of the seventh embodiment displays at least one of the time-series change of the status information 1f collected in the predetermined section in which the classification ID is extracted by the extraction process 14a, the time-series change of the individual damage amount in the predetermined section, the accumulated damage amount, the time-series change of the accumulated damage amount, and the prediction result of the machine's lifespan together with the animation generated by the reproduction process 14b on the display device.

[0106] FIG. 24 shows an example of a display screen when a machine is damaged in a way that affects its lifespan. 24a in FIG. 24 shows the entire display screen. 24b in FIG. 24 shows the pressure waveform of the damaged bucket cylinder, i.e., the time-series change in the status information 1f. 24i in FIG. 24 shows the waveform of the individual damage amount detected based on the pressure waveform 24b of the bucket cylinder, i.e., the time-series change in the individual damage amount. 24f in FIG. 24 shows an animation that reproduces the operation (hereinafter also referred to as "damaged operation") that the damaged machine was performing in a specified section.

[0107] When the machine is a construction machine such as a hydraulic excavator, the display device that displays the display screen 24a may be a display device installed in the driver's seat of the construction machine, or may be a display device installed in the monitoring room of the machine. The display screens 25a and 26a described later are also displayed on a similar display device. In the display process of the seventh embodiment, when the machine is damaged, the display screen 24a may be displayed immediately in real time, or at a predetermined timing such as after the machine finishes working. Alternatively, in the display process of the seventh embodiment, the display screen 24a may be displayed when the user operates a display button or the like. The display screens 25a and 26a are also displayed in real time or at a predetermined timing.

[0108] In the display process of the seventh embodiment, when the machine is damaged, the display screen 24a can be displayed on the display device. In this case, the display process of the seventh embodiment can notify the user by displaying information indicating the occurrence of a damage state, such as "A damage state has occurred," and information on the time when the machine was damaged, as shown in 24c and 24e of FIG. 24b. In the display process of the seventh embodiment, the damaged machine part can be displayed in text as shown in 24d of FIG. 24, or a highlighting display that changes the color or shape of the damaged machine part can be superimposed on the animation 24f as shown in 24h of FIG. 24.

[0109] In this case, the highlighting may be different depending on the cumulative damage level of the damage. For example, if the cumulative damage level of the damaged part of the machine is high, the color may be "red," if the cumulative damage level is not as high as "red" but requires attention, the color may be "yellow," and if neither of these conditions apply, the color may be "blue." Of course, the colors to be displayed are not limited to the three types of red, yellow, and blue, and these three colors are merely examples.

[0110] Furthermore, the display process of the seventh embodiment is not limited to highlighting the damaged machine parts, but may display the animation 24f showing the entire machine by dividing the main parts of the machine (such as the boom, arm, bucket, or running body if the machine is a shovel) by color. As a result, the display process of the seventh embodiment allows the user to easily grasp the cumulative damage level of the entire machine each time damage occurs, such as which parts have a high cumulative damage level and which parts require attention, even for parts other than the highlighted parts.

[0111] The display process of embodiment 7, as shown in 24b and 24i of FIG. 24, can display the time series changes in the status information 1f and the time series changes in the individual damage amounts, thereby informing the user of the type and amount of damage the machine has received.

[0112] Fig. 25 shows an example of a display screen when the remaining life of a certain part of a machine is reduced. 25a in Fig. 25 shows the whole display screen. 25b in Fig. 25 shows the cumulative damage degree of each part of the machine, that is, the cumulative amount of damage each part of the machine has received. 25f in Fig. 25 shows an animation similar to 24f in Fig. 24.

[0113] In the display process of the seventh embodiment, when the remaining life of a certain part of the machine is reduced, the display screen 25a can be displayed on the display device. In this case, the display process of the seventh embodiment can display information indicating that the cumulative damage degree has increased and the remaining life has decreased in text as shown in 25d and 25e of FIG. 25, or can display it as a bar graph or the like as shown in 25b of FIG. 25. In addition, the display process of the seventh embodiment can display the highlighting of the part whose remaining life has decreased superimposed on the animation 25f. Although the classification ID is associated with the damage state, there are cases where a character string for displaying text as shown in 17a of FIG. 17 is not associated with the damage state. Even in this case, the display process of the seventh embodiment reproduces the damage state in which the remaining life has decreased as the animation 25f, so that the user can be informed of the state of the machine in an easy-to-understand manner.

[0114] Fig. 26 shows an example of a display screen displaying the prediction results of the machine's lifespan. 26a in Fig. 26 shows the entire display screen. 26b in Fig. 26 shows the time-series change in the cumulative damage level in a certain part of the machine and the prediction results of its lifespan. That is, 26b in Fig. 26 shows the time-series change in the cumulative damage level in a certain part of the machine and the prediction results of its lifespan. 26f in Fig. 26 shows an animation similar to 25f in Fig. 25.

[0115] In the display process of the seventh embodiment, when the remaining life of a certain part of the machine decreases, the life can be predicted and the display screen 26a can be displayed on the display device. In this case, the display process of the seventh embodiment can notify the user by displaying information indicating an increase in the cumulative damage degree and information indicating the remaining life in text, as shown in 26c and 26d of FIG. 26. In addition, the display process of the seventh embodiment can notify the user by displaying an action to be taken by the user in text, as shown in 26e of FIG. 26. If the action to be taken by the user is to contact the agency, the user may tap (or point and click with the pointer) the display 26e of "Contact the agency" to display the agency's contact information such as a phone number or a simple procedure for taking measures.

[0116] 24 to 26 are examples of display screens, and the presence or absence of text display for informing the user can be set arbitrarily. For example, the display process of the seventh embodiment may first display only the animations 24f to 26f. Then, when the user taps the highlighting superimposed on the animations 24f to 26f, more detailed information may be displayed. As a result, in the display process of the seventh embodiment, a wide display area for the animations 24f to 26f is secured, allowing the user to accurately grasp which part of the entire machine is damaged (whose cumulative damage level is high), and only the information that the user considers necessary can be notified.

[0117] In this way, the condition detection device 100 of the seventh embodiment displays at least one of the time series change in the sensor control information 1f collected in a predetermined section including the damaged section, the time series change in the individual damage amount, the cumulative damage amount, the time series change in the cumulative damage amount, and the prediction result of the machine's lifespan, together with an animation that reproduces the damaged state. As a result, the condition detection device 100 of the seventh embodiment can visualize the damaged state and can visualize future changes in the machine's state from various perspectives, allowing the user to intuitively grasp them.

[0118] [others] The present invention is not limited to the above-described embodiments, and includes various modified examples. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those including all of the configurations described. In addition, it is possible to replace a part of the configuration of a certain embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of a certain embodiment. In addition, it is possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

[0119] In addition, the above-mentioned configurations, functions, processing units, processing means, etc. may be realized in part or in whole by hardware, for example by designing them in an integrated circuit. In addition, the above-mentioned configurations, functions, etc. may be realized by software, in which a processor interprets and executes a program that realizes each function. Information such as the program, tape, file, etc. that realizes each function can be stored in a memory, a recording device such as a hard disk or SSD (solid state drive), or a recording medium such as an IC card, SD card, DVD, etc.

[0120] In addition, the control lines and information lines shown are those that are considered necessary for the explanation, and not all control lines and information lines in the product are necessarily shown. In reality, it can be considered that almost all components are connected to each other. [Explanation of symbols]

[0121] Reference Signs List 1a...collection device, 1b...arithmetic processing device, 1i...classification processing (classification processing unit), 1j...detection processing (detection processing unit), 1l...linking processing (linking processing unit), 1m...determination processing (first determination processing unit), 1n, 18b, 19b...display processing (display processing unit), 1p...registration processing (first registration processing unit), 1o...classification ID database (first recording unit), 1h...classification model database (second recording unit), 12a...determination processing (second determination processing unit), 12b...generation processing (generation processing unit), 12c...registration processing (second registration processing unit), 14a...extraction processing (extraction processing unit), 14b...reproduction processing (reproduction processing unit), 18a...recognition processing (recognition processing unit), 19a...estimation processing (estimation processing unit), 100...state detection device.

Claims

1. A machine status detection device comprising: a collection device that collects status information indicating a time-series changing status of a machine at a plurality of times; and a calculation processing device that processes the plurality of status information collected at the plurality of times to detect the status of the machine, The arithmetic processing device includes: a classification processing unit that time-divides the plurality of pieces of status information collected by the collecting device to identify a plurality of feature vectors having a type of the status information as a feature, classifies the identified feature vectors into one of a plurality of predetermined sets, and assigns a classification ID to the classified feature vector, the classification ID identifying the set to which the feature vector belongs; a detection processing unit that detects damage sustained by the machine based on status information indicating a time-series changing status of the machine collected by the collecting device; a linking processing unit that links a section in which the damage is detected by the detection processing unit with the classification ID assigned to the section by the classification processing unit. A machine status detection device characterized by:

2. The arithmetic processing device includes: a first recording section in which the classification ID for identifying a damage state, which is a state of the machine in the section in which the machine was damaged, is recorded in advance; The apparatus further includes a first determination processing unit that determines whether the classification ID linked by the linking processing unit is an unknown classification ID based on the classification ID recorded in the first recording unit, and a first registration processing unit that registers the classification ID determined to be the unknown classification ID by the first determination processing unit in the first recording unit.

2. The machine condition detection device according to claim 1.

3. The arithmetic processing device includes: a second recording unit in which the feature vector and the classification ID are recorded in advance in association with each other; a second determination processing unit that determines whether or not the classification ID to be assigned to the feature vector identified by the classification processing unit is recorded in the second recording unit, based on the feature vector identified by the classification processing unit and the feature vector recorded in the second recording unit; a generation processing unit that generates a new classification ID when the classification ID to be assigned to the feature vector identified by the classification processing unit is not recorded in the second recording unit, and a second registration processing unit that associates the classification ID generated by the generation processing unit with the feature vector to which the classification ID should be assigned and registers the association result in the second recording unit.

2. The machine condition detection device according to claim 1.

4. The arithmetic processing device includes: an extraction processing unit that extracts the classification ID assigned by the classification processing unit in a predetermined section including the section linked to the classification ID by the linking processing unit; a reproduction processing unit that generates a video reproducing the state of the machine in the predetermined section based on the classification ID extracted by the extraction processing unit; a display processing unit that displays the moving image generated by the reproduction processing unit on a display device.

2. The machine condition detection device according to claim 1.

5. The arithmetic processing device further includes an estimation processing unit that estimates an amount of damage received by the machine, and the detection processing unit detects an individual damage amount, which is the amount of damage received by the machine in the section, based on the status information collected in the section, the linking processing unit links the individual damage amount detected by the detection processing unit to the classification ID assigned by the classification processing unit in the section in which the individual damage amount was detected, The estimation processing unit: calculating an integrated value of the individual damage amount for each of the classification IDs based on the individual damage amount detected by the detection processing unit and the number of times the individual damage amount has been detected; The integrated values ​​calculated for each classification ID are accumulated to estimate the amount of accumulated damage received by the machine.

2. The machine condition detection device according to claim 1.

6. The arithmetic processing device includes: an extraction processing unit that extracts the classification ID assigned by the classification processing unit in a predetermined section including the section linked to the classification ID by the linking processing unit; a reproduction processing unit that generates a video reproducing the state of the machine in the predetermined section based on the classification ID extracted by the extraction processing unit; a display processing unit that displays the moving image generated by the reproduction processing unit on a display device, The display processing unit causes the display device to display, together with the video, at least one of the time-series change in the status information collected in the predetermined section, the time-series change in the individual damage amount in the predetermined section, the cumulative damage amount, the time-series change in the cumulative damage amount, and a prediction result of the life of the machine.

6. The machine condition detection device according to claim 5.

7. A machine status detection device comprising: a collection device that collects status information indicating a time-series changing status of a machine at a plurality of times; and a calculation processing device that processes the plurality of status information collected at the plurality of times to detect the status of the machine, The arithmetic processing device includes: a classification processing unit that time-divides the plurality of pieces of status information collected by the collecting device to identify a plurality of feature vectors having a type of the status information as a feature, classifies the identified feature vectors into one of a plurality of predetermined sets, and assigns a classification ID to the classified feature vector, the classification ID identifying the set to which the feature vector belongs; a first recording section in which the classification ID for identifying a damage state, which is a state of the machine in a section in which the machine is damaged, is recorded in advance; a recognition processing unit that recognizes the damage state specified by the classification ID assigned by the classification processing unit, based on the classification ID assigned by the classification processing unit and the classification ID recorded in the first recording unit. A machine status detection device characterized by:

8. The arithmetic processing device further includes an estimation processing unit that estimates an amount of damage that the machine has received, and the first recording unit records an individual amount of damage that is an amount of damage that the machine has received in the section, the individual amount of damage being linked to the classification ID in advance, The estimation processing unit: Identifying the individual damage amount corresponding to the damage state recognized by the recognition processing unit based on the classification ID that identifies the damage state recognized by the recognition processing unit and the classification ID recorded in the first recording unit; calculating an integrated value of the individual damage amount for each of the classification IDs based on the identified individual damage amount and the occurrence count of the damage state corresponding to the individual damage amount; The integrated values ​​calculated for each classification ID are accumulated to estimate the amount of accumulated damage received by the machine.

8. The machine condition detection device according to claim 7.

9. The arithmetic processing device includes: a cut-out processing unit that cuts out the classification ID assigned by the classification processing unit in a predetermined section including the section corresponding to the damage state recognized by the recognition processing unit; a reproduction processing unit that generates a video reproducing the state of the machine in the predetermined section based on the classification ID extracted by the extraction processing unit; a display processing unit that displays the moving image generated by the reproduction processing unit on a display device, The display processing unit causes the display device to display, together with the video, at least one of the time-series change in the status information collected in the predetermined section, the time-series change in the individual damage amount in the predetermined section, the cumulative damage amount, the time-series change in the cumulative damage amount, and a prediction result of the life of the machine.

9. The machine condition detection device according to claim 8.

Citation Information

Patent Citations

  • Thermal mass measuring device

    JP1987015446A

  • Editing device

    JP1989058055A

  • Numerical control device and machine learning device

    JP2018097616A

  • Device and method for diagnosing railway vehicle apparatus

    JP2018185256A

  • Action grasping system

    JP2022083497A