Diagnosis system

Abstract

To provide a diagnosis system capable of relatively accurately grasping the start / stop state of a rotary machine without being affected by disturbances specific to ships, and capable of contributing to accurate abnormal diagnosis for the rotary machine.SOLUTION: A diagnosis system 1 for diagnosing a condition of a rotary machine M installed in a ship comprises: an acquisition unit 10 that acquires vibration information of the rotary machine M; a classification model that extracts a speed component at a rotational frequency of the rotary machine M and an acceleration component in a specific frequency band of the rotary machine M from the vibration information acquired by the acquisition unit 10, and determines a start / stop state of the rotary machine M; and a determination unit 20 that determines the start / stop state of the rotating machine M based on the extracted speed component and the acceleration component.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a diagnostic system.

Background Art

[0002] Currently, in order to diagnose abnormalities in a device having a rotating bearing (hereinafter referred to as a "rotating machine"), a technique is known in which vibration information in the operating state of the rotating machine is acquired and analyzed to determine the presence or absence of an abnormality and identify the location where the abnormality has occurred. In order to perform such abnormality diagnosis, it is necessary to grasp the start / stop state of the target rotating machine. However, in some cases, it is difficult to directly obtain a signal indicating the start / stop state from the rotating machine, and it is required to determine the start / stop state of the machine based on information obtained from various sensors externally attached to the rotating machine.

[0003] For example, in recent years, a start / stop determination method has been proposed in which the current value flowing through the power cable supplying power to the rotating machine is detected at a predetermined cycle using an external current sensor, and when the AC component included in the detected current value exceeds a predetermined threshold value, it is determined that the "rotating machine is in an operating state" (see Patent Document 1).

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, when adopting the start / stop determination method using a current sensor as described in Patent Document 1, it takes time and cost to install the current sensor, and there are problems with the stability of the behavior of the current sensor in an electromagnetic noise environment. For this reason, there has been a long-awaited start / stop determination method that avoids the disadvantages of adopting a current sensor and is not affected by disturbances peculiar to ships.

[0006] The present invention has been made in view of such circumstances, and an object thereof is to provide a diagnostic system that can relatively accurately grasp the start / stop state of a rotating machine without being affected by disturbances peculiar to ships and can contribute to accurate abnormality diagnosis of the rotating machine.

Means for Solving the Problems

[0007] The diagnostic system according to the present invention for achieving the above object is a system for diagnosing the state of a rotating machine installed in a ship, comprising: an acquisition unit that acquires vibration information of the rotating machine; and a classification model for determining the start / stop state of the rotating machine, and a determination unit that determines the start / stop state of the rotating machine based on the speed component at the rotation frequency of the rotating machine and the acceleration component in a specific frequency band of the rotating machine, which are extracted from the vibration information acquired by the acquisition unit, and the extracted speed component and acceleration component.

[0008] By adopting such a configuration, the speed component at the rotation frequency of the rotating machine and the acceleration component in a specific frequency band of the rotating machine can be extracted from the vibration information of the rotating machine, and the start / stop state of the rotating machine can be determined based on the classification model for determining the start / stop state of the rotating machine and the extracted speed component and acceleration component. Therefore, it is possible to relatively accurately grasp the start / stop state of the rotating machine without being affected by disturbances peculiar to ships (such as shaking caused by external environments such as waves and vibration of surrounding equipment accumulated in the ship), and to avoid the inconvenience when adopting a current sensor, and it becomes possible to contribute to accurate abnormality diagnosis of the rotating machine.

[0009] In the diagnostic system according to the present invention, when it is determined by the determination unit that the rotating machine is in the running state, a diagnostic unit for diagnosing the state of the rotating machine based on the vibration information acquired by the acquisition unit can be provided.

[0010] By adopting such a configuration, when it is determined that the rotating machine is in the running state, the state of the rotating machine can be diagnosed based on the vibration information acquired by the acquisition unit.

[0011] In the diagnostic system according to the present invention, a classification model created by performing logistic regression analysis or discriminant analysis on information indicating the start / stop state of a rotating machine can be adopted. Here, the information indicating the start / stop state of the rotating machine can be generated by performing cluster analysis on the speed component and the acceleration component in the vibration information of the rotating machine. Further, the cluster analysis can be performed on the speed component in the vibration information of the rotating machine acquired in the past and the first principal component score calculated by performing principal component analysis on the acceleration component in the vibration information of the rotating machine acquired in the past. And the determination unit can determine the start / stop state of the rotating machine based on the classification model, the speed component, and the information regarding the first principal component score.

[0012] In the diagnostic system according to the present invention, a specific frequency band in which the correlation ratio between the intensity of the vibration acceleration component of the rotating machine and the start / stop state exceeds a predetermined threshold can be defined as the specific frequency band. For example, when the rotating machine is provided with a rolling bearing, the specific frequency band can be set to 5 to 15 kHz.

Effects of the Invention

[0013] According to the present invention, it is possible to provide a diagnostic system that can relatively accurately grasp the start / stop state of a rotating machine without being affected by disturbances peculiar to a ship and can contribute to accurate abnormality diagnosis of the rotating machine.

Brief Description of the Drawings

[0014]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Mode for Carrying Out the Invention

[0015] Hereinafter, embodiments of the present invention will be described with reference to the drawings. It should be noted that the following embodiments are merely preferred application examples, and the scope of application of the present invention is not limited thereto.

[0016] First, the configuration of the diagnostic system 1 according to the present embodiment will be described with reference to FIG. 1. The diagnostic system 1 is a system for diagnosing the state of the rotating machine M installed in the ship's interior, and can relatively accurately grasp the start / stop state of the rotating machine M without being affected by disturbances peculiar to the ship, and can contribute to accurate abnormality diagnosis of the rotating machine M.

[0017] In this embodiment, the rotating machine M to be diagnosed does not rotate while maintaining perfect balance (without being biased), but rotates while generating minute rubbing rotations. That is, while the rotating machine M is operating, vibrations that can also be expressed as "unbalance vibrations" occur. The inventor of the present invention has found that the presence or absence and frequency of such "unbalance vibrations" are closely related to the start / stop state, and has completed the present invention. Examples of the rotating machine M include those having rolling bearings and those having sliding bearings.

[0018] As shown in FIG. 1, the diagnostic system 1 according to this embodiment includes an acquisition unit 10 that acquires vibration information of the rotating machine M, a determination unit 20 that determines the start / stop state of the rotating machine M based on the vibration information acquired by the acquisition unit 10, and a diagnosis unit 30 that diagnoses the state of the rotating machine M based on the vibration information acquired by the acquisition unit 10 when the determination unit 20 determines that the rotating machine M is in the starting state. The determination unit 20 and the diagnosis unit 30 are incorporated in the diagnostic device 2 installed at a location away from the rotating machine M.

[0019] The acquisition unit 10 has a vibration sensor 11 attached to the rotating machine M. The vibration sensor 11 acquires vibration information including a speed component at the rotational frequency of the rotating machine M and an acceleration component in a specific frequency band of the rotating machine M. The vibration information acquired by the vibration sensor 11 is sent to the diagnostic device 2 having the determination unit 20 and the diagnosis unit 30 via communication means and is used for start / stop determination and the like. The configuration of the communication means is not particularly limited. For example, as shown in FIG. 1, a slave unit 12 that performs wired communication with the vibration sensor 11, a relay unit 13 that performs wireless communication with the slave unit 12, and a master unit 14 that performs wireless communication with the relay unit 13 and wired communication with the diagnostic device 2 can be adopted.

[0020] The determination unit 20 functions to extract from the vibration information acquired by the acquisition unit 10 the velocity component at the rotational frequency of the rotating machine M and the acceleration component in a specific frequency band of the rotating machine M. Since the "rotational frequency" is information unique to the rotating machine M, the determination unit 20 can store in advance the rotational frequency corresponding to the type of the rotating machine M to be determined. The determination unit 20 functions to identify the time history of the vibration velocity from the vibration information acquired by the acquisition unit 10, convert this into the frequency domain, and then extract the intensity of the component at the pre-stored rotational frequency as the "velocity component at the rotational frequency of the rotating machine M". When there are a plurality of rotational frequencies of the rotating machine M, the determination unit 20 can extract the sum or average value of the intensities of the components at the plurality of rotational frequencies as the "velocity component at the rotational frequency of the rotating machine M".

[0021] Also, the "specific frequency band" is a specific frequency band in which the correlation ratio between the intensity of the vibration acceleration component of the rotating machine M and the start / stop state exceeds a predetermined threshold value. For example, as shown in FIG. 2, the time history of the vibration acceleration when the rotating machine M is in the "ON (started)" state and the "OFF (stopped)" state is acquired in advance, converted into the frequency domain, and then the correlation ratio between the intensity of the vibration acceleration and the ON / OFF state is calculated for each frequency. Then, the determination unit 20 can extract and store a specific frequency band (specific frequency band) in which this correlation ratio exceeds a predetermined threshold value (for example, 0.6). When the rotating machine M has a rolling bearing, the specific frequency band is 5 to 15 kHz.

[0022] The determination unit 20 can function to identify the time history of the vibration acceleration from the vibration information acquired by the acquisition unit 10, convert this into the frequency domain, and then extract the intensity of the component in a specific frequency band stored in advance as the "acceleration component in the specific frequency band of the rotating machine M". At this time, when creating the classification model described later, the determination unit 20 can calculate the first principal component score from the intensity of the component in the specific frequency band using the same weight coefficient as that used in the principal component analysis performed on the acceleration components in the vibration information of the rotating machine M acquired in the past, and extract it as the "acceleration component in the specific frequency band of the rotating machine M".

[0023] Further, the determination unit 20 functions to determine the start / stop state of the rotating machine M based on the classification model for determining the start / stop state of the rotating machine M and the extracted speed component and acceleration component. Here, the "classification model" is created by performing logistic regression analysis on the information indicating the start / stop state of the rotating machine M, and is the core of the present invention.

[0024] To explain the concept of the "classification model", first, the "information indicating the start / stop state of the rotating machine M" will be explained. The "information indicating the start / stop state of the rotating machine M" in the present embodiment is generated by performing cluster analysis on the velocity component and the acceleration component in the vibration information of the rotating machine M acquired in the past. For example, the cluster analysis can be performed on the velocity component in the vibration information of the rotating machine M acquired in the past and the first principal component score calculated by performing principal component analysis on the acceleration component in the vibration information of the rotating machine M acquired in the past. Further, the cluster analysis can be performed with the number of clusters set to 2, and the larger of the central values of the velocity components of the two generated clusters can be specified as the "starting state" of the rotating machine M, and the smaller of the central values of the velocity components can be specified as the "stopping state" of the rotating machine M. Here, the "principal component analysis" is a machine learning method that summarizes the shape of data using variance, and summarizes each data as a principal component score using a plurality of weight coefficients. In the "principal component analysis", as shown in FIG. 3, the axis with the largest variation in the scatter plot of the acquired data is specified, and where each data is located on this axis is calculated as the "first principal component score". The vibration information of the rotating machine M is plotted as shown in FIG. 4 with the first principal component score (acceleration component) calculated by such "principal component analysis" on the vertical axis and the velocity component on the horizontal axis, and cluster analysis is performed on the plotted data to classify it into two types of clusters as shown in FIG. 5. The data obtained in this way is the "information indicating the start / stop state of the rotating machine M" in the present embodiment.

[0025] As an evaluation method for the clustering result, for example, a method using the silhouette coefficient can be mentioned. According to the evaluation method using the silhouette coefficient, it is preferable because both the cohesion within the same cluster and the separation between different clusters can be evaluated. Cohesion a (i) is the average value of the distances from a certain point within a certain cluster to other points within the same cluster, and is calculated by the following formula (1).

Equation

Number

Number

[0026] If the silhouette coefficient is less than 0.55 (or there is a negative value in the silhouette coefficient), the clustering may be inappropriate. Therefore, it is preferable to perform parameter adjustment such as normalization and standardization and then perform the logistic regression analysis described later. On the other hand, if the silhouette coefficient is 0.55 or more, it is determined that the clustering is appropriate, and the logistic regression analysis described later is performed.

[0027] The "classification model" is created by performing a logistic regression analysis on this "information indicating the starting and stopping states of the rotating machine M". The logistic regression analysis can be performed using the following formula (4) of the sigmoid function.

Number

[0028] Based on the "classification model" created in this way and the extracted velocity component and acceleration component (the acceleration component among these is the "first principal component score" already described), the determination unit 20 functions to determine the starting and stopping state of the rotating machine M. For example, if the data of the extracted velocity component and acceleration component falls within the "ON region" of FIG. 6, the determination unit 20 determines that the rotating machine M is in the starting state.

[0029] When the determination unit 20 determines that the rotating machine M is in the starting state, the diagnosis unit 30 functions to diagnose the state of the rotating machine M based on the vibration information acquired by the acquisition unit 10. As the diagnosis unit 30, one that notifies that there may be a failure in the rotating machine M when the O / A value of the vibration velocity or vibration acceleration acquired from the rotating machine M is greater than a predetermined reference value can be adopted. Also, one that specifies the cause of the failure (for example, shaft bending, bolt loosening, bearing abnormality, etc.) according to the intensity of the vibration velocity or vibration acceleration at a specific frequency of the rotating machine M can also be adopted as the diagnosis unit 30.

[0030] In the diagnostic system 1 according to the embodiment described above, from the vibration information of the rotating machine M, the velocity component at the rotation frequency of the rotating machine M and the acceleration component in a specific frequency band of the rotating machine M are extracted, and a classification model for determining the start / stop state of the rotating machine M and the extracted velocity component and acceleration component are used to determine the start / stop state of the rotating machine M in the determination unit 20. Therefore, without being affected by disturbances specific to ships (such as shaking caused by external environments such as waves, vibrations of surrounding equipment accumulated inside the ship, etc.), and while avoiding the inconveniences when adopting current sensors, the start / stop state of the rotating machine M can be grasped relatively accurately, which contributes to accurate abnormality diagnosis of the rotating machine M. And when it is determined that the rotating machine M is in the starting state, the state of the rotating machine M can be diagnosed by the diagnosis unit 30 based on the vibration information acquired by the acquisition unit 10.

[0031] Next, examples of the present invention will be described.

[0032] <First Embodiment> First, by applying the diagnostic system 1 according to the present invention to a first ship, the start / stop determination of a rotating machine M (auxiliary blower) mounted on the first ship was performed. A vibration sensor 11 (acquisition unit 10) was attached to the rotating machine M, and the vibration information of the rotating machine M was acquired by this vibration sensor 11. As the vibration information, the sum of the intensity of the velocity components at the rotation frequencies of 58.75 kHz and 60.00 kHz of the rotating machine M and the acceleration component in the rotation frequency range of 5 to 15 kHz (specific frequency band) were adopted. Then, the data was plotted as shown in FIG. 7(A) with the first principal component score calculated by performing principal component analysis on this acceleration component as the vertical axis and the velocity component as the horizontal axis, and information indicating the start / stop state of the rotating machine M was generated by performing cluster analysis on this plotted data. By performing logistic regression analysis on this information, a "classification model" as shown in FIG. 7(B) was generated. Then, based on this "classification model" and the velocity component and acceleration component extracted from the vibration information acquired by the vibration sensor 11, the start / stop state of the rotating machine M was determined. As a result, the accuracy rate of the start / stop determination was "0.89".

[0033] <Second Embodiment> Next, by applying the diagnostic system 1 according to the present invention to a second ship, for the rotating machine M (auxiliary blower) mounted on the second ship, a start / stop determination was made in the same manner as in the first embodiment. As a result, the matching rate of the start / stop determination was "0.96". FIG. 8(A) is a diagram plotting the data obtained in this embodiment, and FIG. 8(B) is the "classification model" generated in this embodiment.

[0034] <Third Embodiment> Subsequently, by applying the diagnostic system 1 according to the present invention to a third ship, for the rotating machine M (auxiliary blower) mounted on the third ship, a start / stop determination was made in the same manner as in the first embodiment. As a result, the matching rate of the start / stop determination was "0.97". FIG. 9(A) is a diagram plotting the data obtained in this embodiment, and FIG. 9(B) is the "classification model" generated in this embodiment.

[0035] <First Comparative Example> On the other hand, an example of making a start / stop determination of the rotating machine M (auxiliary blower) mounted on the first ship by a method using a conventional O / A (overall) value without adopting the diagnostic system 1 according to the present invention will be described. A vibration sensor 11 (acquisition unit 10) was attached to the rotating machine M, and vibration information of the rotating machine M was acquired by this vibration sensor 11. As the vibration information, the vibration speed of the rotating machine M was adopted. Then, the O / A value was calculated from this vibration speed, and the start / stop state of the rotating machine M was determined based on whether or not this O / A value exceeded a predetermined threshold value. As a result, the matching rate of the start / stop determination was "0.60".

[0036] <Second Comparative Example> Next, for the rotating machine M (auxiliary blower) mounted on the second ship, a start / stop determination was made in the same manner as in the first comparative example by a method using a conventional O / A value. As a result, the matching rate of the start / stop determination was "0.54".

[0037] <Third Comparative Example> Subsequently, for the rotating machine M (auxiliary blower) mounted on the third ship, start / stop determination was performed in the same manner as in the first comparative example using the conventional O / A value method. As a result, the accuracy rate of the start / stop determination was "0.54".

[0038] According to the results of the above embodiments and comparative examples, it became clear that the accuracy rate of the start / stop determination when the diagnostic system 1 according to the present invention is adopted is significantly higher than the accuracy rate of the start / stop determination when the method using the conventional O / A value is adopted.

[0039] In the above embodiment, an example was shown in which "logistic regression analysis" was adopted for the information indicating the start / stop state of the rotating machine M when generating the classification model. However, a classification model can also be generated by adopting "discriminant analysis" for the information indicating the start / stop state of the rotating machine M.

[0040] Also, in the above embodiment, by performing cluster analysis on the speed component and the acceleration component in a specific frequency band (the first principal component score obtained as a result of principal component analysis) in the vibration information of the rotating machine M, information indicating the start / stop state of the rotating machine M was generated, and a classification model was generated by performing logistic regression analysis on this information. An example (that is, an example in which a "classification model" was generated by adopting "principal component analysis", "cluster analysis", and "logistic regression analysis" when there is no "correct data" regarding the start / stop state of the rotating machine M) was described. However, the present invention can also be applied when "correct data" regarding the start / stop state of the rotating machine M exists.

[0041] That is, the speed component and the acceleration component in a specific frequency band when the rotating machine M is in the operating state are acquired, and the speed component and the acceleration component in a specific frequency band when the rotating machine M is in the stopped state are acquired, and these are plotted as "correct data" (information indicating the start / stop state of the rotating machine M), and the "classification model" generated based on this "correct data" can be used for the start / stop determination of the rotating machine M.

[0042] Also, in the above embodiments, an example was shown in which cluster analysis was performed using the first principal component score calculated by performing "principal component analysis" for the purpose of dimensionality reduction on the acceleration component in the vibration information of the rotating machine M acquired in the past. However, cluster analysis can be performed without necessarily performing "principal component analysis" on the acceleration component (that is, even when there are multiple parameters for the acceleration component).

[0043] The present invention is not limited to the above embodiments, and those in which a person skilled in the art appropriately makes design changes to such embodiments are also included in the scope of the present invention as long as they have the features of the present invention. Also, each element included in the above embodiments can be combined as long as it is technically possible, and those obtained by combining them are also included in the scope of the present invention as long as they include the features of the present invention.

Industrial Applicability

[0044] The present invention is useful when providing a diagnostic system that can relatively accurately grasp the start / stop state of a rotating machine without being affected by disturbances peculiar to a ship and can contribute to accurate abnormality diagnosis of the rotating machine.

Explanation of Signs

[0045] 1…Diagnostic system 10…Acquisition unit 20…Determination unit 30…Diagnostic unit M…Rotating machine

Claims

1. A diagnostic system for diagnosing the state of a rotating machine installed inside a ship, comprising: an acquisition unit that acquires vibration information of the rotating machine; an extraction unit that extracts, from the vibration information acquired by the acquisition unit, a velocity component at the rotational frequency of the rotating machine and an acceleration component in a specific frequency band of the rotating machine, and a classification model for determining the starting and stopping state of the rotating machine; and a determination unit that determines the starting and stopping state of the rotating machine based on the extracted velocity component and acceleration component and the classification model; A diagnostic system comprising the above.

2. The diagnostic system according to claim 1, further comprising a diagnosis unit that diagnoses the state of the rotating machine based on the vibration information acquired by the acquisition unit when the determination unit determines that the rotating machine is in a starting state.

3. The classification model in the diagnostic system according to claim 1 is created by performing logistic regression analysis or discriminant analysis on information indicating the starting and stopping state of the rotating machine.

4. The information indicating the starting and stopping state of the rotating machine in the diagnostic system according to claim 3 is generated by performing cluster analysis on the velocity component and the acceleration component in the vibration information of the rotating machine.

5. The cluster analysis in the diagnostic system according to claim 4 is performed on the velocity component in the vibration information of the rotating machine acquired in the past, and the first principal component score calculated by performing principal component analysis on the acceleration component in the vibration information of the rotating machine acquired in the past.

6. The determination unit in the diagnostic system according to claim 5 determines the starting and stopping state of the rotating machine based on the classification model, the velocity component, and information regarding the first principal component score.

7. The specific frequency band in the diagnostic system according to claim 1 is a specific frequency band in which the correlation ratio between the intensity of the vibration acceleration component of the rotating machine and the starting and stopping state exceeds a predetermined threshold value.

8. The rotating machine in the diagnostic system according to claim 7 is provided with a rolling bearing.

9. The specific frequency band in the diagnostic system according to claim 8 is 5 to 15 kHz.

Images