State diagnostic device and state diagnosis system
The condition diagnosis device uses a neural network to estimate bearing abnormalities from external sensor data, enhancing accuracy and reducing costs by eliminating the need for internal sensors.
Patent Information
- Application Number
- JP2024039272
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-03-13
- Publication Date
- 2025-09-29
AI Technical Summary
Existing condition diagnosis systems face challenges in accurately diagnosing bearing abnormalities due to interference from external vibrations and require expensive electrical components, leading to increased costs.
A condition diagnosis device that uses a neural network to estimate signal information from a virtual sensor installed near the bearing, eliminating the need for a physical sensor by learning the relationship between external and internal sensor signals through machine learning.
Accurately diagnoses bearing abnormalities while reducing costs by eliminating the need for expensive internal sensors and minimizing interference from external vibrations.
Smart Images
Figure 2025140097000001_ABST
Abstract
Description
[Technical Field]
[0001] The present invention relates to a condition diagnosis device and a condition diagnosis system for diagnosing the condition of an object such as a mechanical facility or an item for sale including a device or a vehicle. [Background technology]
[0002] Condition diagnosis devices and systems have been proposed for diagnosing the presence or absence of abnormalities in mechanical equipment, devices, and objects for sale, including vehicles. Patent Document 1 discloses an abnormality diagnosis device that uses a sensor to detect vibration signals generated from a rotating part, such as a motor in the mechanical equipment, performs envelope processing and frequency analysis on the detection results to extract frequency components, and compares the frequency components of a predetermined rotational frequency with a threshold value to diagnose the presence or absence of abnormalities. The threshold value is set for each frequency according to the level difference of the vibration response previously measured by impact testing for each part, so that the transmission distance and transmission path of the vibration signal between the bearing of the rotating part and the sensor are reflected. Furthermore, a method of installing a sensor near or inside the bearing has been proposed as a method for suppressing the influence of the transmission distance and transmission path of the vibration signal between the bearing and the sensor.
[0003] Patent Document 2 discloses a mounting structure for attaching a self-power-generating sensor externally to a bearing body, in which threaded portions are provided on the inner and outer rings of the bearing, and a sensor unit or cover equipped with a sensor, power generation unit, etc. is screwed onto the bearing. Patent Document 3 discloses a bearing with a wireless sensor, in which detection signals from a (built-in) sensor provided on the outer or inner ring of the bearing are processed by a wireless processing circuit, etc., provided in the seal. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] Patent No. 5146008 [Patent Document 2] Patent No. 6676901 [Patent Document 3] Patent No. 6743361 Summary of the Invention [Problem to be solved by the invention]
[0005] In Patent Document 1, as described above, in order to reduce the influence of the transmission distance and transmission path between the bearing and the sensor, a threshold value is set for each frequency according to the level difference of the vibration response measured in advance by impact testing for each part. However, the vibrations detected by the sensor include not only vibrations generated from the rotating parts mentioned above, but also vibrations from outside the machinery and equipment, and vibrations from other moving parts and drive sources such as motors within the machinery, making it difficult to accurately extract vibrations caused by bearing abnormalities.
[0006] Regarding the transmission distance and transmission path between the bearing and the sensor, the structure of Patent Document 2 reduces the influence of the transmission distance and transmission path by screwing the sensor unit into the bearing, but requires space around the bearing to install the sensor unit (externally attached). Furthermore, installation in a limited space requires electrical components such as small sensors and generators, but these components are expensive, so the cost increases depending on the number of sensor units used. Meanwhile, the structure of Patent Document 3 incorporates the sensors and other components into the bearing, so space for installing sensor units like in Patent Document 2 is not required around the bearing. However, as with Patent Document 2, it still requires the use of expensive electrical components such as small sensors and generators, so the cost increases depending on the number of sensors and other components used.
[0007] SUMMARY OF THE INVENTION In order to solve the above-mentioned problems of the prior art, an object of the present invention is to provide a condition diagnosis device and a condition diagnosis system that are capable of accurately diagnosing abnormalities and that are cost-effective. [Means for solving the problem]
[0008] Generally, the condition diagnosis device of the present invention inputs signal information from a first sensor externally attached to an object, such as a machine or equipment having a bearing, and applies a neural network to an estimation model that estimates and outputs signal information from a second sensor built into or installed near the bearing, thereby learning the relationship between the signals from the two sensors through machine learning. The training data is, for example, a pair of signal information from the first sensor and signal information from the second sensor. The condition diagnosis device of the present invention uses this trained estimation model to estimate signal information from the second sensor built into or installed near the bearing from the signal information from the first sensor installed in the machine or equipment, without using the second sensor (i.e., it estimates the signal information from an imaginary, non-existent second sensor located at a virtual position), and then diagnoses the condition of the equipment using this estimated sensor signal information. This not only enables accurate abnormality diagnosis, but also eliminates the need for the second sensor, thereby reducing costs.
[0009] The condition diagnosis device according to the present invention comprises: A condition diagnosis device that diagnoses the presence or absence of an abnormality in an object, an estimation unit including an estimation model and estimating signal information of a virtual sensor virtually existing at a virtual position in the object; a condition diagnosis unit that diagnoses whether or not there is an abnormality in the object based on the estimated signal information, the estimation unit estimates, using the estimation model, signal information of the measurement target at a target portion that is a moving part of the object or a target portion in its vicinity, from signal information including outputs of one or more external sensors that are attached to the object and measure at least one of vibration and temperature, the signal information being assumed to be measured at the virtual position by the virtual sensor; The estimation model is a model that includes the relationship between signal information of the external sensor and signal information of the measurement object obtained from an actual sensor installed at a position corresponding to the virtual position of the target portion of the object, based on signal information including the output of the external sensor attached to the object and signal information of the measurement object obtained from the actual sensor installed at a position corresponding to the virtual position of the target portion of the object. A condition diagnosis system according to the present invention includes the condition diagnosis device and the object. In the present invention, "diagnosis of the presence or absence of an abnormality" includes simply diagnosing the presence or absence of an abnormality, and although no specific example is given, it may also include diagnostic information that estimates the severity of the abnormality, the state of state changes including future prospects, the cause of the abnormality, and the location of the abnormality. It may also include recommended countermeasures (repairs, review of operating conditions).
[0010] The condition diagnosis device and system according to the present invention include an estimation unit that includes an estimation model and estimates signal information of a virtual sensor virtually existing at the virtual position, specifically, the estimation unit that uses the estimation model to estimate signal information of the measurement object of the target portion, which is assumed to have been measured at the virtual position by the virtual sensor, from signal information including the output of the external sensor measuring the measurement object, and also includes an estimation model that includes a relationship between signal information of both sensors, based on signal information including the output of the external sensor attached to the target and signal information of the measurement object acquired from an actual sensor installed at a position corresponding to the virtual position of the target portion of the target. With this configuration, the condition diagnosis device and system according to the present invention can use the estimation model to estimate signal information of the virtual sensor at the virtual position corresponding to, for example, a sensor (sensor unit) built into or installed near a bearing of the operating part, from signal information of the external sensor installed on the target object, such as mechanical equipment, and use the signal information for diagnosis. This eliminates the need for the expensive sensor unit, reducing costs, and also eliminates or reduces the difference in vibration response level caused by the influence of the transmission distance and transmission path between the bearings of the operating unit and the external sensor, allowing the external sensor to accurately diagnose abnormalities in mechanical equipment and other objects without the sensor unit. Furthermore, the influence of external disturbances such as vibrations applied from outside the device can be suppressed, making it possible to accurately diagnose the presence or absence of abnormalities. Furthermore, since there is no longer a need to set a threshold for each frequency depending on the influence of the transmission distance and transmission path, as in conventional technology, labor hours can be reduced, reducing costs.
[0011] The estimation model may be a neural network trained by machine learning using training data of signal information of the measurement target acquired from an actual sensor installed at a position corresponding to the virtual position of the target portion of the object, thereby enabling even more accurate abnormality diagnosis.
[0012] In the case of an estimation model of the above form, the estimation model may be applied to an abnormality diagnosis of another object corresponding to the one object for which the training data was obtained. As a result, the estimation model can be applied to the other object corresponding to the one object, such as an object that is a machine or equipment of the same type. Therefore, when a factory or the like has multiple machines or equipment of the same type and monitors their conditions, it is not necessary to install the above expensive sensor unit for each machine or equipment, thereby reducing costs.
[0013] A block that estimates signal information from a virtual sensor virtually existing at the virtual position of the object and a block that diagnoses the presence or absence of an abnormality in the object based on the estimated signal information may be separate. The above blocks are formed by functions, subroutines, modules, parts, mounting patterns, etc., configured of hardware, software, or a mixture thereof. As a result, typically, by separating the block of the estimation unit and the block of the condition diagnosis unit, it is possible to, for example, change threshold settings in the condition diagnosis unit or improve a diagnostic algorithm, etc., independently, thereby facilitating such improvements, thereby increasing the flexibility and maintainability of the condition diagnosis device and reducing costs.
[0014] The input to the estimation unit may include signal information indicating operating conditions of the object other than the signal information from the external sensor. The operating conditions are signal information other than the signal information from the external sensor and are information dependent on the object. By including the operating conditions and operating status of the object determined by the operator's operation or preset by the operator, such as the operating time, operation pattern, rotation speed, and motor current value indicating the load state of the operating unit, in the signal information, the load conditions and rotation speed acting on the operating unit can be used for estimation. Furthermore, it is possible to distinguish between vibrations and temperature changes caused by the operating conditions of the object and the magnitude, frequency, and temperature rise of vibrations caused by bearing damage, thereby improving estimation accuracy. This enables more accurate abnormality diagnosis.
[0015] The estimation unit may estimate signal information of the measurement target in the vicinity of a rotating part provided with a bearing that is the operating part of the object, or of the bearing, from signal information including the output of the external sensor attached to the object, thereby enabling abnormality diagnosis of the rotating part provided with a bearing that is the operating part of the object. [Effects of the Invention]
[0016] The condition diagnosis device and condition diagnosis system of the present invention are capable of accurately diagnosing abnormalities and reducing costs. [Brief explanation of the drawings]
[0017] [Figure 1] 1 is a block diagram showing a configuration of a condition diagnosis device according to an embodiment; [Figure 2] 2 is a partial perspective view showing a schematic configuration of a condition diagnosis system including the condition diagnosis device. FIG. [Figure 3] FIG. 2 is a schematic diagram illustrating a neural network of an estimation model in the condition diagnosis device. [Figure 4] FIG. 2 is a diagram illustrating a schematic configuration for acquiring training data used in the estimation model. [Figure 5]3A to 3C are waveform diagrams showing time-series data of acceleration acquired by an external sensor and an internal sensor in the condition diagnosis system. [Figure 6] FIG. 6 is a waveform diagram showing the results of envelope processing and frequency analysis performed on the acceleration data of FIG. 5. [Figure 7] FIG. 10 is a schematic diagram illustrating a procedure for applying an estimation model of one of the objects from which the training data is obtained to an abnormality diagnosis of another object. [Figure 8] 1 is a partial perspective view showing a schematic configuration of the condition diagnosis system in which a plurality of external sensors and the like are attached. FIG. DETAILED DESCRIPTION OF THE INVENTION
[0018] An exemplary embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a condition diagnosis device according to one embodiment for diagnosing the presence or absence of an abnormality in an object, and is a block diagram illustrating the flow of signal processing in the condition diagnosis device 100. The condition diagnosis device 100 includes an estimation unit 130 that includes an estimation model 133 and estimates signal information from a virtual sensor virtually located at a virtual position VP in the object 200, and a condition diagnosis unit 150 that diagnoses the presence or absence of an abnormality in the object 200 based on the estimated signal information. FIG. 2 is a partial perspective view showing the schematic configuration of a condition diagnosis system including the condition diagnosis device 100, in which sensors connected to the condition diagnosis device 100 are attached to a mechanical equipment 200, which is an example of an object to be diagnosed. As described above, the condition diagnosis system sys of this embodiment includes the condition diagnosis device 100 and the object (mechanical equipment) 200 to which one or more external sensors 300 are attached.
[0019] The signal information input values in Fig. 1 include the output values of the external sensor 300 attached to the mechanical equipment 200 in Fig. 2 as well as signal information of values that can be determined on the mechanical equipment side, such as detected values of the load and force on the mechanical equipment related to cutting conditions other than signals from such sensors, such as the operating conditions and operating status described below, and the output value (rotational speed) of a rotational speed sensor of the mechanical equipment 200. The external sensor 300 is, for example, a sensor that detects vibration, such as a vibration sensor or an acoustic sensor, or a temperature sensor, and is easily attached externally to the housing of the mechanical equipment 200, etc., and is fixed in a location suitable for detecting the vibration, temperature, etc. At least one of vibration and temperature is measured by the external sensor 300, and the measurement target may also include acoustic data, etc., as vibration data.
[0020] The estimation unit 130 in FIG. 1 uses the signal information input value to estimate the signal information of the measurement target at the target portion, which is the moving part of the object 200 or its vicinity, using an estimation model 133. The signal information is assumed to be measured by the virtual sensor at a virtual position VP. The estimation unit 130 in this embodiment estimates signal information acquired by a virtual sensor (not shown) installed at a position corresponding to the virtual position VP at the moving part of the object 200 or its vicinity, specifically, at or near the rotating shaft (rotating part) 235 on which the bearing 1, which is the moving part of the object 200, is mounted, or near or inside the bearing 1, and outputs the signal information as a signal information estimate. The virtual position VP is, for example, a location where a sensor for detecting vibrations, a temperature sensor, etc., would be installed in a conventional mechanical equipment, or a range including this location. The bearing 1 in this embodiment is, for example, a radial ball bearing.
[0021] The condition diagnosis unit 150 diagnoses an abnormality in the mechanical equipment 200, or in this embodiment, the condition of the target part, based on the input signal information estimated value. More specifically, the condition diagnosis unit 150 diagnoses the presence or absence of an abnormality in the bearing 1 in the mechanical equipment 200. If the signal information is vibration, for example, the condition diagnosis unit 150 performs envelope processing and frequency analysis on the vibration waveform data, and diagnoses the presence or absence of an abnormality by comparing the frequency component of a predetermined rotational frequency calculated from the rotational speed of the bearing 1 with a threshold, and outputs the diagnosis result. If the signal information is temperature, for example, the condition diagnosis unit 150 diagnoses the presence or absence of an abnormality by comparing the absolute value of the temperature, the temperature difference from a reference value, the rate of temperature rise per hour, etc. with a threshold, and outputs the diagnosis result.
[0022] The estimation model 133 is a model that includes the relationship between the signal information of both sensors (external sensor 300 and sensor sn) based on signal information including the output of external sensor 300 attached to the mechanical equipment 200 and signal information of the measurement target acquired from an actual sensor sn (FIG. 4, described below) installed at a position corresponding to the virtual position of the target portion of the mechanical equipment 200. The estimation model 133 in FIG. 1 uses, for example, the neural network shown in FIG. 3. FIG. 3 is a schematic diagram illustrating the structure of a neural network. A neural network has an input layer, an intermediate layer, and an output layer, and the intermediate layer consists of multiple node layers each containing one or more nodes. Note that the neural network shown in FIG. 3 is just an example, and the number of nodes in each layer and the number of intermediate layers may be adjusted. Furthermore, the output signal may be fed back as an input signal, or the layer and node structure may be adjusted to make it easier to handle time-series data.
[0023] The estimation model 133 of this embodiment is trained by machine learning using training data. Specifically, the estimation model 133 of this embodiment is a neural network that uses training data consisting of a pair of signal information including the output of the external sensor 300 attached to the mechanical equipment 200 and the signal information of the mechanical equipment 200 acquired from an actual sensor sn installed at a position corresponding to the virtual position VP of the target portion of the mechanical equipment 200, and learns, through machine learning, the relationship between the signal information including the output of the external sensor 300 attached to the mechanical equipment 200 and the signal information of the mechanical equipment 200 acquired from the actual sensor sn installed at a position corresponding to the virtual position VP of the mechanical equipment 200. Specifically, the training data uses data acquired by the sensor sn in a configuration shown in FIG. 4, which closely resembles the configuration of FIG. 2. The external sensor 300 in FIG. 4 is installed in the same manner as in FIG. 2. The virtual position VP is, for example, a position where a detection component such as the sensor sn would conventionally be installed, as described above. The position of the (built-in) sensor sn of the sensor-equipped bearing 11 in Figure 4 corresponds to the virtual position VP in Figure 2, but it may be the same or substantially the same, or may be near that position, for example, as long as it is thought that signal information assumed to be measured by the virtual sensor at the virtual position VP can be obtained. Here, the bearing 1 in Figure 4 is a radial ball bearing that constitutes the sensor-equipped bearing 11 and has a built-in sensor sn. Note that while the bearing 1 is shown as an example in which the sensor sn is built-in, the sensor sn unit may be attached externally and integrated into the bearing 1, or may be installed near the bearing, and data on the object to be measured may be acquired. Furthermore, in the sensor-equipped bearing 11, the sensor may be of either a wireless transmission type or a wired transmission type.
[0024] The estimation model 133 is trained to estimate and output signal information from a virtual sensor at a virtual position VP in FIG. 2 using signal information from an external sensor 300 installed on the mechanical equipment 200, such as information indicating the operating time, operation pattern, and rotation speed of the mechanical equipment, as training data, and signal information from a sensor sn built into the bearing as output. When vibration or acoustic signals are used as input and output, time-series vibration waveforms, frequency analysis data, data subjected to envelope processing and frequency analysis, or a combination of these can be appropriately selected and used. For example, the input may be a time-series vibration waveform, and the output may be data subjected to envelope processing and frequency analysis, or both the input and output may be data subjected to envelope processing and frequency analysis.
[0025] Furthermore, the input signal information (signal information input value) can include the operating conditions and operating status of the mechanical equipment 200, which are determined by the operator's operation or by the operator's pre-setting, such as the operating time, operation pattern, rotational speed, and motor current value indicating the load state of the operating part, such as the rotating part on which the bearing 1 is provided, of the mechanical equipment 200. This allows the load conditions and rotational speed acting on the bearing to be used for estimation, and also makes it possible to distinguish between vibrations and temperature changes caused by the operating conditions of the mechanical equipment and the magnitude and frequency of vibrations and the magnitude of temperature rise caused by damage to the bearing, thereby improving estimation accuracy. Note that the operating conditions of the object 200 are signal information other than the signal information from the external sensor 300, as described above, and are information dependent on the object.
[0026] Figure 5 shows time series data of acceleration G acquired by an external sensor 300 installed on a machine 200, which is an example of a test object, and by the built-in sensor sn of the sensor-equipped bearing 11, in an actual machine test. Here, the test conditions were as follows: a deep groove ball bearing was used as the bearing 1, a recessed portion simulating flaking was formed on the outer ring rolling surface, a load was applied in the radial direction, and the bearing 1 was rotated for 2000 min -1 The test was conducted by rotating the bearing 1 at a constant speed. The output of the built-in sensor sn is less affected by the transmission distance and transmission path between the bearing 1 and the built-in sensor sn, so the output amplitude is larger than that of the external sensor 300.
[0027] Figure 6 is a frequency characteristic diagram showing the results of envelope processing and frequency analysis of the vibration data (acceleration G) in Figure 5. This frequency characteristic diagram shows peaks at 103 Hz and higher frequencies due to the relative rotational speed of the rolling element with respect to the outer ring. As shown in Figure 6, the output of the built-in sensor sn has larger peak values than the output of the external sensor 300, demonstrating that it detects high-frequency vibrations with high sensitivity and that the transmission distance and transmission path between the bearing 1 and the built-in sensor sn have little effect. The built-in sensor sn directly measures vibrations of the rotating part on which the bearing 1 is mounted, or in its vicinity, of the bearing 1, which is the moving part of the testing machine (subject, mechanical equipment 200). Therefore, the effects of external vibrations and vibrations from other moving parts, motors, etc. within the mechanical equipment can be suppressed.
[0028] 5 and 6 show vibration data as an example, but similarly for temperature data, if estimation model 133 is trained using signal information from external sensor 300 and signal information from built-in sensor sn, it is possible to estimate the signal information of built-in sensor sn, i.e., the virtual sensor, from the signal information from external sensor 300. In this way, similarly for temperature data, the presence or absence of an abnormality can be diagnosed with high accuracy using signal information that reduces the effects of the transmission distance and transmission path between bearing 1 and external sensor 300.
[0029] FIG. 7 is a schematic diagram showing a configuration in which an estimation model 133 trained using signal information measured by a machine A, designated by reference numeral 200A, which has an internal sensor sn, is applied to other machines B, C, D, etc. (designated by reference numerals 200B, 200C, 200D, etc.) of the same type that do not have the internal sensor sn, thereby diagnosing the condition of the other machines with the same level of accuracy as that of the machine A. When the machines are of the same type, the transmission distance and influence of the transmission path between the bearing 1 including the internal sensor sn and the external sensor 300 (FIGS. 2 and 4) are similar. Therefore, by using a condition diagnosis device 100 that applies the estimation model 133 trained using signal information measured by the machine A in FIG. 7, it is possible to estimate the signal information of the internal sensor sn of the sensor-equipped bearing 11 and diagnose the condition of the machine. Here, the other objects corresponding to the object A for which training data was learned do not have to be of the same type, but may be similar or have similar or similar influences on the transmission distance and transmission path, or may be considered to be similar or similar. In this case, the same effect as in the case of the same type as described above can be achieved. Since built-in sensors such as sensor sn generally use small and expensive electrical components, in the operation described in Figure 7, for example, when a factory or the like has multiple machines of the same type and monitors their status, it is not necessary to install the above-mentioned expensive built-in sensors for each machine. Therefore, costs can be reduced.
[0030] 8 shows another example of a case where multiple external sensors 300 and sensor-equipped bearings 11 are installed on one piece of machinery and equipment 200 such as that shown in FIG. 4. In this way, multiple sensor-equipped bearings 11 may be installed to correspond to the bearings whose condition is to be diagnosed. Furthermore, the external sensors 300 are appropriately set to be installed in positions and in numbers (multiple sensors are possible) that are suitable for learning the estimation model 133, depending on the number and positions of the sensor-equipped bearings 11. In other words, for example, the number of external sensors 300 and sensor-equipped bearings 11 may or may not be the same.
[0031] In this embodiment, a block that estimates signal information from a virtual sensor virtually existing at a virtual position on an object and a block that diagnoses the presence or absence of an abnormality in the object based on the estimated signal information are separated. Specifically, as shown in the block diagram of Fig. 1, an estimation unit 130 and a condition diagnosis unit 150 are separated. Note that the block corresponding to the estimation unit 130 and the block corresponding to the condition diagnosis unit 150 are each formed of functions, subroutines, modules, components, implementation patterns, etc., which are configured using hardware, software, or a combination of both.
[0032] When using a neural network, it is typically possible to receive signal information input values, perform the learning and estimation, and directly output a diagnosis result. In other words, the learning and estimation may be implemented as an algorithm that is an integrated function, subroutine, module, implementation pattern, or the like, without being separated. This is because such an integrated configuration may be relatively simple in structure and development. However, in this embodiment, by separating the estimation unit 130 (block) and the condition diagnosis unit 150 (block), it is possible to change the threshold setting in the condition diagnosis unit 150 and improve the diagnostic algorithm, etc., independently. This makes it easy to make such improvements, and increases the flexibility and maintainability of the condition diagnosis device 100.
[0033] Although the embodiments of the present invention have been described above, the disclosed embodiments are illustrative in all respects and are not limiting. The scope of the present invention is defined by the claims, not by the above description, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]
[0034] 1. Bearings 11, 11A, 11B Sensor-equipped bearings 100 Condition diagnostic device 130 Estimation part 133 Estimation Model 150 Condition diagnosis section 200, 200A, 200B, 200C, 200D Machinery and equipment (objects) 235, 235A, 235B Rotating shaft (rotating part) 300, 300A, 300B external sensor sn (built-in) sensor VP Virtual Position
Claims
1. A condition diagnosis device that diagnoses the presence or absence of an abnormality in an object, an estimation unit including an estimation model and estimating signal information of a virtual sensor virtually existing at a virtual position in the object; a condition diagnosis unit that diagnoses whether or not there is an abnormality in the object based on the estimated signal information, the estimation unit estimates, using the estimation model, signal information of the measurement object at a target portion that is a moving part of the object or a target portion in its vicinity, from signal information including outputs of one or more external sensors that are attached to the object and measure at least one of vibration and temperature, the signal information being assumed to have been measured at the virtual position by the virtual sensor; the estimation model is a model including a relationship between signal information including an output of the external sensor attached to the object and signal information of the measurement target acquired from an actual sensor installed at a position corresponding to the virtual position of the target portion of the object, and signal information of the external sensor and the actual sensor. Condition diagnostic device.
2. The condition diagnosis device according to claim 1, the estimation model is a neural network trained by machine learning using training data as signal information of the measurement target acquired from an actual sensor installed at a position corresponding to the virtual position of the target portion of the object; Condition diagnostic device.
3. The condition diagnosis device according to claim 2, the estimation model is applied to an abnormality diagnosis of another object corresponding to the one object from which the training data was obtained. Condition diagnostic device.
4. The condition diagnosis device according to claim 1, a block for estimating signal information of a virtual sensor virtually existing at the virtual position of the object and a block for diagnosing the presence or absence of an abnormality in the object based on the estimated signal information are separated from each other; Condition diagnostic device.
5. The condition diagnosis device according to claim 1, The input of the estimation unit includes signal information indicating an operating condition of the object other than the signal information from the external sensor. Condition diagnostic device.
6. The condition diagnosis device according to claim 1, the estimation unit estimates signal information of the measurement target in the vicinity of a rotating part provided with a bearing, which is the operating part of the object, or of the bearing, from signal information including an output of the external sensor attached to the object; Condition diagnostic device.
7. A condition diagnosis device according to any one of claims 1 to 6, comprising the object. Condition diagnostic system.
Citation Information
Patent Citations
Shingodensohoshiki
JP1976046008A
Bearing-sensor mounting structure
JP6676901B2
Wireless sensor bearing
JP6743361B2