Module for detecting anomalies in rotating body using three-axis acceleration data

Abstract

A module for detecting anomalies in a rotating body using three-axis acceleration data, according to one aspect of the present invention, comprises: a collection unit for collecting real-time vibration data of a rotating body; a processing unit which detects an abnormal point of the data collected by the collection unit; and an output unit which visualizes the data processed by the processing unit and outputs a result.

Description

Rotating body anomaly detection module utilizing 3-axis acceleration data

[0001] The present invention relates to a rotating body anomaly detection module utilizing 3-axis acceleration data, and more specifically, to a rotating body anomaly detection module utilizing 3-axis acceleration data that is equipped in a rotating body to detect vibration and determine anomalies.

[0002] Generally, in industrial facilities, a rotating body refers to a machine part or equipment that rotates around an axis, and failures occur due to abnormal noise and vibration caused by damage to key components such as bearings and impellers.

[0003] Rotating machinery presents a problem in that damage to parts such as bearings and impellers causes downtime for repairs, leading to reduced productivity, and if the rotating machinery itself fails, it results in massive maintenance costs.

[0004] In addition, rotating bodies generate vibrations even in normal conditions, so regulations are needed to distinguish vibrations within the normal range, and an efficient algorithm is required to detect abnormalities by analyzing this in real time.

[0005] Conventional predictive maintenance systems detect problems before equipment failure, but they have the disadvantage of being difficult to clearly distinguish between unnecessary warnings, such as undetected or false positives, and the normal vibrations of rotating machinery.

[0006] The objective of the present invention is to provide a rotating body anomaly detection module utilizing 3-axis acceleration data that collects data on the vibration of a rotating body in real time and identifies the normal range of vibration using an algorithm.

[0007] Another objective of the present invention is to provide a rotating body anomaly detection module utilizing 3-axis acceleration data that can clearly define the normal vibration range of a rotating body through an efficient algorithm and accurately detect anomalies.

[0008] A rotating body anomaly detection module utilizing 3-axis acceleration data according to one aspect of the present invention for achieving the above-mentioned purpose may comprise a collection unit that collects real-time vibration data of a rotating body, a processing unit that detects anomalies in the data collected by the collection unit, and an output unit that visualizes the data processed by the processing unit and outputs the result.

[0009] The above-mentioned collection unit is equipped with an accelerometer that detects minute vibration changes of the rotating body, and the accelerometer can collect physical variables of displacement, velocity, and acceleration of vibration data of the rotating body over a wide frequency range.

[0010] The above collection unit may be equipped with a control board and a 6-axis sensor to measure the vibration and physical data of the rotating body.

[0011] The above collection unit may be equipped with a Bluetooth sensor and an IoT-based wireless network to collect data in real time through data measured and collected from the accelerometer, the control board, and the 6-axis sensor.

[0012] The processing unit and the output unit may be equipped with a 3D spatial model and a time series model, which is a cumulative sum algorithm, to track the average change amount and detect anomalies using data received in real time through the Bluetooth sensor and the IoT-based wireless network and to visualize them.

[0013] The above cumulative sum algorithm can improve the accuracy of data anomaly detection by optimizing the hyperparameters, namely the threshold and the drift.

[0014] The output unit can distinguish by color the section where a change in the vibration data of the rotating body begins, the section where an abnormal phenomenon occurs, and the section where it returns to normal after an alarm.

[0015] The above 3D spatial model outputs the direction and magnitude of the observed acceleration vector in space, and can diagnose the abnormality by distinguishing the color of the vector endpoint when it deviates from the set value and becomes an abnormal state.

[0016] According to the rotating body anomaly detection module utilizing 3-axis acceleration data according to the present invention, the processing unit and the output unit can track the average change amount and detect anomalies using data received in real time via a cumulative sum algorithm.

[0017] In addition, the cumulative sum algorithm can improve the accuracy of data anomaly detection by optimizing the hyperparameters, namely the threshold and the drift.

[0018] FIG. 1 is a drawing showing the appearance of a rotating body anomaly detection module utilizing 3-axis acceleration data according to one embodiment of the present invention.

[0019] FIGS. 2a to 2d are drawings showing a result table output from an output unit after applying a rotating body anomaly detection module utilizing 3-axis acceleration data according to another embodiment of the present invention to a small Sirocco fan.

[0020] FIG. 3 is a diagram illustrating a method for visualizing an algorithm of a rotating body anomaly detection module utilizing 3-axis acceleration data according to another embodiment of the present invention.

[0021] FIG. 4 is a diagram showing a graph output from an output unit after applying a rotating body anomaly detection module utilizing 3-axis acceleration data according to another embodiment of the present invention to a small Sirocco fan.

[0022] Hereinafter, an embodiment of the present invention will be described in detail with reference to the attached drawings. However, the present invention is not limited to this embodiment and can be modified in various forms.

[0023] In the drawings, parts unrelated to the description have been omitted to clearly and concisely explain the present invention, and the same reference numerals are used for identical or extremely similar parts throughout the specification. Additionally, in the drawings, thicknesses, widths, etc., are depicted enlarged or reduced to make the description clearer; however, the thicknesses, widths, etc., of the present invention are not limited to those depicted in the drawings.

[0024] And when any part of the specification is described as "including" another part, unless specifically stated otherwise, it does not exclude other parts and may include additional parts.

[0025]

[0026] FIG. 1 is a diagram showing the appearance of a rotating body anomaly detection module utilizing 3-axis acceleration data according to one embodiment of the present invention.

[0027] As illustrated in FIG. 1, a collection unit (100) for collecting real-time vibration data of a rotating body (10) according to one embodiment of the present invention, a processing unit (200) for detecting abnormalities in the data collected by the collection unit (100), and an output unit (300) for visualizing the data processed by the processing unit (200) and outputting the result.

[0028] (1) Collection section

[0029] The collection unit (100) may be equipped with an accelerometer (103) that detects minute vibration changes of the rotating body (10) and collects vibration data as physical variables such as displacement, velocity, and acceleration over a wide frequency range.

[0030] The collection unit (100) is equipped with a control board (101) and a 6-axis sensor (102) to measure vibration and physical data of a rotating body (10), and may be equipped with a Bluetooth sensor (201) and an IoT-based wireless network (202) to collect data in real time through data measured and collected from the accelerometer (103), the control board (101), and the 6-axis sensor (102).

[0031] The control board (101) may be equipped with an Arduino Uno board, the 6-axis sensor (102) may be equipped with an MPU6050, and the Bluetooth sensor (201) may be equipped with an HM-10.

[0032] (2) Processing unit and output unit

[0033] The processing unit (200) and output unit (300) may be equipped with a 3D spatial model and a time series model, which is a cumulative sum algorithm, to track the average change amount and detect anomalies using data received in real time through a Bluetooth sensor (201) and an IoT-based wireless network (202) and to visualize them.

[0034] The cumulative sum algorithm can improve the accuracy of data anomaly detection by optimizing the hyperparameters, namely the threshold and the drift.

[0035] The cumulative sum algorithm can be implemented as CUSUM(Cumulative Sum).

[0036] CUSUM (Cumulative Sum) is an algorithm used to detect the average variation of time-series data. By calculating the cumulative sum of the data, it can detect anomalies if a specific threshold is exceeded. Primarily, it can be used to detect abnormal vibrations in equipment in the manufacturing industry and to identify abnormal usage in network traffic.

[0037] In addition, normal and abnormal states can be distinguished through time series data, and anomaly detection can be performed when vibration levels in rotating equipment suddenly increase over time.

[0038] In addition, based on past data, the likelihood of failure can be predicted by analyzing the trend of temperature changes in the equipment through future values.

[0039] In addition, time series data can be represented as a graph to express temperature changes over time in a way that users can easily understand.

[0040] Hyperparameters refer to values ​​that must be set in advance in an algorithm, and threshold and drift are hyperparameters of the cumulative sum algorithm (CUSUM), and the outlier detection performance may vary depending on the values.

[0041] A threshold is a reference value set to determine whether a change is normal or abnormal; for example, if the amount of change in a vibration value exceeds the threshold, it can be determined to be abnormal.

[0042] The drift threshold is a criterion for distinguishing whether a change in data is an "average shift within a normal range" or an "abnormal change." If the drift threshold is too large, it may miss minute abnormalities, and if it is too small, it may cause an overreaction (false detection).

[0043] Therefore, to find the optimal value, it is essential to test various settings and select the value that demonstrates the highest diagnostic accuracy to enhance the reliability and accuracy of the algorithm.

[0044] Finally, the output unit (300) can provide information to the user on the monitor (301) by visualizing the data value processed by the processing unit (200) as a graph.

[0045] The output unit (300) can distinguish by color the section where a change in the vibration data of the rotating body (10) begins, the section where an abnormal phenomenon occurs, and the section where it returns to normal after an alarm.

[0046] However, if color is not displayed on the screen, the line shape can be distinguished by representing it as '…', '---', and '━━━'.

[0047] The 3D spatial model outputs the direction and magnitude of the observed acceleration vector in space, and if it deviates from the set value and becomes an abnormal state, it can diagnose the abnormality by distinguishing the color of the vector endpoint.

[0048] Here, the acceleration vector is an acceleration calculated based on the vibration data of a rotating body, which has magnitude (change in velocity) and direction (direction of vibration), and the 3D space model may refer to a system that visually represents this data in a three-dimensional coordinate system.

[0049] The vector endpoint refers to the end point (ending coordinate) of the acceleration vector, and by displaying a different color for when the endpoint is normal and when it is abnormal, abnormal vibrations can be intuitively identified through the color change.

[0050] However, if color representation is not possible, the line shape can be distinguished and expressed as ‘…’, ‘---’, and ‘━━━’.

[0051] Next, a rotating body anomaly detection module utilizing 3-axis acceleration data according to another embodiment of the present invention will be described. In the following description, only the parts that differ from the above-described embodiment will be described in detail, and detailed descriptions of identical or extremely similar parts will be omitted.

[0052]

[0053] FIGS. 2a to 2d are drawings showing a result table output from an output unit after applying a rotating body anomaly detection module utilizing 3-axis acceleration data according to another embodiment of the present invention to a small Sirocco fan, FIG. 3 is a drawing showing an algorithm visualization method of a rotating body anomaly detection module utilizing 3-axis acceleration data according to another embodiment of the present invention, and FIG. 4 is a drawing showing a graph output from an output unit after applying a rotating body anomaly detection module utilizing 3-axis acceleration data according to another embodiment of the present invention to a small Sirocco fan.

[0054] For small Sirocco fans, data is collected according to five attachment positions of the package module and 14 failure items are defined. After performing an in-depth analysis of four data points that showed a significant difference from the normal state using a cumulative sum algorithm (CUSUM), and selecting the optimal threshold and drift values ​​of the cumulative sum algorithm (CUSUM), it is possible to secure performance capable of detecting abnormal states with an accuracy of over 90% in data over 180 seconds during which normal and abnormal states alternate.

[0055] To explain in more detail,

[0056] 1) By attaching an anomaly detection module to five different locations on a small Sirocco fan, various data can be collected to analyze the impact of each location and increase reliability.

[0057] 2) The types of failures that may occur during fan operation (vibration, noise, imbalance, etc.) can be classified into 14 categories, and four key data points showing a significant difference from the normal state can be selected to conduct an in-depth analysis.

[0058] 3) At this time, an abnormal state can be detected by performing in-depth analysis with the cumulative sum algorithm (CUSUM), tracking the average change amount of the data to detect anomalies, and applying this algorithm to four major data points.

[0059] 4) Through algorithm optimization, the threshold and drift values ​​are reference values ​​set to detect abnormal conditions in the cumulative sum algorithm (CUSUM), and the accuracy of the analysis results can be improved by selecting optimal values.

[0060] 5) As a result of analyzing the data over 180 seconds during which the fan alternated from a normal state to an abnormal state, we achieved a performance capable of detecting abnormal states with an accuracy of over 90%, which demonstrates the high performance of the algorithm and system.

[0061] Referring to Figures 2a to 2d, the process of selecting optimal parameters for four types of data (impeller clogging failure, suction port clogging failure due to vinyl, failure due to slope formation, and failure due to foreign matter inside the impeller) can be seen in the result table.

[0062] Referring to Figures 3 and 4, a visualization method of the cumulative sum algorithm (CUSUM) and a graph of a real-time time series example can be seen.

[0063] The vibration data of the rotating body (10) can be classified into three sections: a section where the change begins (▶), a section where an abnormal phenomenon occurs and an alarm is provided (●), and a section where the normal state is maintained after the alarm (◀), and these sections can be distinguished by different colors to enable effective visualization.

[0064] However, if color is not displayed on the screen, effective visualization may be possible by using different types of lines such as ‘…’, ‘---’, and ‘━━━’.

[0065] The 3D spatial model outputs the direction and magnitude of the observed acceleration vector in space, and if it deviates from the set value and becomes an abnormal state, it can diagnose the abnormality by distinguishing the color of the vector endpoint.

[0066] Through this, the rotating body anomaly detection module utilizing 3-axis acceleration data according to the present invention can be utilized for resource management and integration with a Building Energy Management System (BEMS) by applying optimal hyperparameters to a real-time monitoring time-series anomaly diagnosis model to increase diagnosis accuracy. Therefore, it can improve energy efficiency management and increase equipment maintenance efficiency.

[0067] In addition, a real-time 3D spatial model capable of multidimensional analysis can complement the limitations of single-dimensional data analysis and local anomaly detection, precisely analyze abnormal states by calculating the magnitude and directionality of acceleration vectors, and enable efficient anomaly diagnosis by integrating with a time-series model. Therefore, more efficient anomaly diagnosis can be achieved by integrating this 3D spatial model with the time-series model.

[0068] In addition, about 55% of the causes of abnormal conditions in rotating bodies can be detected early, and up to 25% of total maintenance costs can be expected to be reduced.

[0069]

[0070] Although a rotating body anomaly detection module utilizing 3-axis acceleration data according to an embodiment of the present invention has been described above, the concept of the present invention is not limited to the embodiments presented in this specification. Furthermore, a person skilled in the art who understands the concept of the present invention may easily propose other embodiments within the scope of the same concept by adding, changing, deleting, or adding components, and such are also to be considered to fall within the scope of the concept of the present invention.

Claims

1. A collection unit that collects real-time vibration data of a rotating body, and A processing unit that detects anomalies in the data collected by the above collection unit, and A rotating body anomaly detection module utilizing 3-axis acceleration data, comprising an output unit that visualizes data processed by the processing unit and outputs the result.

2. In Paragraph 1, The above-mentioned collecting unit is equipped with an accelerometer that detects minute vibration changes of the rotating body, and The above accelerometer is a rotating body anomaly detection module utilizing 3-axis acceleration data, characterized by collecting physical variables of displacement, velocity, and acceleration from vibration data of the rotating body over a wide frequency range.

3. In Paragraph 2, A rotating body anomaly detection module utilizing 3-axis acceleration data, characterized in that the above-mentioned collection unit is equipped with a control board and a 6-axis sensor to measure the vibration and physical data of the rotating body.

4. In Paragraph 3, A rotating body anomaly detection module utilizing 3-axis acceleration data, characterized in that the above-described collection unit is equipped with a Bluetooth sensor and an IoT-based wireless network to collect data in real time through data measured and collected from the above-described accelerometer, the above-described control board, and the above-described 6-axis sensor.

5. In Paragraph 4, A rotating body anomaly detection module utilizing 3-axis acceleration data, characterized in that the processing unit and the output unit are equipped with a 3D spatial model and a time series model, a cumulative sum algorithm, to track average changes and detect anomalies using data received in real time through the Bluetooth sensor and the IoT-based wireless network.

6. In Paragraph 5, A rotating body anomaly detection module utilizing 3-axis acceleration data, characterized by the above-mentioned cumulative sum algorithm improving the accuracy of data anomaly diagnosis by optimizing the hyperparameters, namely the threshold and the drift.

7. In Paragraph 6, A rotating body anomaly detection module utilizing 3-axis acceleration data, characterized in that the output unit distinguishes by color the section where a change in the vibration data of the rotating body begins, the section where an anomaly occurs, and the section where it returns to normal after an alarm.

8. In Paragraph 7, A rotating body anomaly detection module utilizing 3-axis acceleration data, characterized by the above 3D spatial model outputting the direction and magnitude of the observed acceleration vector in space, and diagnosing the anomaly by distinguishing the color of the vector endpoint when the state deviates from the set value.

Images