KR102261899B1 - Apparatus and method for diagnosing fault of drone - Google Patents

Abstract

A drone abnormality diagnosis apparatus of the present invention includes: a vibration sensor unit for collecting raw vibration data generated when the drone is driven; a filter unit that passes only an effective frequency band for diagnosing an abnormality of the drone among the raw vibration data collected by the vibration sensor; a conversion unit that converts the data in the time domain that has passed through the filter unit into the frequency domain and outputs it as inspection data; and an abnormality detection unit that stores vibration data when the drone is in a normal state in advance as reference data, and compares the inspection data output from the converter with the reference data to detect whether the drone is abnormal, FFT transformation By processing the data by applying an algorithm, the processing speed and data capacity can be dramatically improved, and by using the Euclidean distance calculation formula, it is possible to implement a low-capacity algorithm that can be mounted on a small microprocessor by simplifying the calculation process. There are advantages.

Description

Drone abnormality diagnosis device and method {APPARATUS AND METHOD FOR DIAGNOSING FAULT OF DRONE}

The present invention relates to an apparatus and method for diagnosing an abnormality of a drone, and more particularly, to an apparatus and method for diagnosing an abnormality of a drone for diagnosing an abnormality in a power unit of a drone using a vibration sensor.

Recently, although the application of drones in commercial and industrial fields is rapidly increasing and the size is accelerating, with this, fall accidents and secondary damage are also increasing, which is an obstacle to the development of the drone industry.

Accordingly, there is a need for a technology capable of preventing a fall accident due to a failure of a power unit (eg, a motor, a propeller, a frame, etc.) of an industrial drone or a poor maintenance due to human causes.

On the other hand, looking at previous studies for automatically diagnosing abnormalities in systems with rotating bodies, linear discriminant analysis, artificial neural networks, principal component analysis, Mahalanobis distance classifier ( Various analysis methods such as Mahalanobis distance classifier and logistic regression were used to determine the classification method. Among them, the 'Mahalanobis distance'-based classification method showed superior performance compared to other analysis methods. It has been used in many studies.

The basic approach to the detection of anomalies suggested in previous studies is to determine as an abnormal state when there is a large deviation in the collected data obtained through monitoring based on the vibration signal collected in the operating state of the normal system.

However, these conventional studies could not overcome the limited environmental characteristics of the drone system (eg, signal processing system performance, capacity, physical size, etc.), and due to this, there was a problem that the conventional studies could not be applied to the drone system. .

1. Lee, SH and Lim, G., “Performance Comparison of Mahalanobis- Taguchi System and Logistic Regression-A Case Study,” Journal of the Korean Institute of Industrial Engineers, Vol.39, No.5, pp.393-402 , 2013. 2. Bakonyi Mihaly, 1995, “The Euclidian Distance Matrix Completion Problem,” pp.646-654. 3. De Maesschalck, Roy, 2000, “The mahalanobis distance,” Chemometrics and Intelligent Laboratory Systems, pp.1-18. 4. Park, S. G., Park, W. S., Jung, J. E., Lee, Y. Y., and Oh, J. E., “A Fault Diagnosis on the Rotating Machinery Using Mahalanobis Distance,” Transactions of the Korean Society of Mechanical Engineers, Vol. 32, No. 7, pp. 556-560, 2008. 5. Yu Jeong-won, Jang Jae-yeol, Yoo Jae-young, Kim Seong-shin (2016). Failure detection method of steam boiler tube using Mahalanobis distance. Journal of the Korean Institute of Intelligent Systems, 26(3), 246-252.

Accordingly, an object of the present invention is to provide an apparatus and method for diagnosing abnormalities of a drone capable of signal processing in an environment of an ultra-small vibration sensor module that can be mounted on a small industrial drone.

Another object of the present invention is to provide an apparatus for diagnosing an abnormality of a drone and a method thereof, which minimizes data processing capacity according to the limited environmental characteristics of a small industrial drone.

In addition, the present invention provides a drone abnormality diagnosis apparatus and method that can dramatically improve processing speed and data capacity by processing the data by applying an FFT conversion algorithm after collecting vibration data generated when the drone is driven want to

In addition, the present invention provides a drone abnormality diagnosis apparatus capable of implementing a low-capacity algorithm that can be mounted on a small microprocessor by simplifying the calculation process by using the Euclidean distance calculation formula when comparing the collected data with the reference data, and We want to provide that method.

In order to achieve the above object, the apparatus for diagnosing an abnormality of a drone provided by the present invention includes a vibration sensor unit for collecting raw vibration data generated when the drone is driven; a filter unit that passes only an effective frequency band for diagnosing an abnormality of the drone among the raw vibration data collected by the vibration sensor; a conversion unit that converts the data in the time domain that has passed through the filter unit into the frequency domain and outputs it as inspection data; and an abnormality detection unit that stores vibration data when the drone is in a normal state in advance as reference data, and compares the inspection data output from the converter with the reference data to detect whether the drone is abnormal. do.

Preferably, the apparatus may further include an effective frequency determination unit that detects a frequency pattern that appears by creating a random failure situation with respect to the drone, and determines the effective frequency band based on the result.

Preferably, the effective frequency determiner may process data by special classification for a region in which harmonics are generated according to a rotation speed of a propeller among frequency bands existing within the effective frequency band.

Preferably, the transform unit may perform a Fast Fourier Transform (FFT) for performing a Fourier transform of only some of the signals of the frequency band that have passed through the filter unit.

Preferably, the abnormality detection unit generates reference data by calculating an average vector of the plurality of vibration data after collecting a plurality of vibration data based on a predetermined condition, and a Euclidean distance value between the reference data and the inspection data By calculating (D) and comparing the Euclidean distance value (D) with a preset threshold value (D _th ), it is possible to determine whether an abnormality has occurred in the drone.

Preferably, the abnormality detection unit collects vibration data in a steady state at regular frequency intervals with respect to a frequency of a first frequency band, which is an effective frequency band for diagnosing abnormalities of the drone, and calculates an average vector of the collected data and, with respect to a frequency of a second frequency band that is a frequency band in which a harmonic is generated according to the rotation speed of the drone, the peak value of the harmonic may be calculated as an average vector.

Preferably, when the rotation speed of the propeller of the drone is greater than or equal to a specific speed, the abnormality detection unit may generate the reference data whenever a preset reference speed is reached.

In addition, in order to achieve the above object, the method for diagnosing abnormalities of a drone provided by the present invention includes: a reference data generation step of learning vibration data when the drone is in a normal state and generating it as reference data; A data collection step of collecting raw data of vibrations generated when the drone is driven; a filtering step of selecting only a valid frequency band for abnormal diagnosis of the drone from among the collected vibration raw data; a conversion step of converting the data of the time domain selected in the filtering step into the frequency domain and outputting the data as inspection data; and comparing the inspection data with the reference data to detect an abnormality of the drone.

Preferably, the method may further include a frequency band determining step of detecting a frequency pattern that appears by creating a random failure situation for the drone, and determining the effective frequency band based on the result.

Preferably, the step of generating the reference data may include an average vector calculating step of collecting a plurality of vibration data based on a predetermined condition and then calculating an average vector of the plurality of vibration data to generate the reference data.

Preferably, in the calculating of the average vector, for a frequency of a first frequency band that is an effective frequency band for diagnosing abnormalities of the drone, vibration data in a steady state is collected at regular frequency intervals, and an average vector of the collected data is collected. a first average vector calculation step of calculating and a second average vector calculating step of calculating the peak value of the harmonic as an average vector with respect to a frequency of a second frequency band that is a frequency band in which harmonics are generated according to the rotation speed of the drone.

Preferably, in the step of generating the reference data, when the rotation speed of the propeller of the drone is greater than or equal to a specific speed, the reference data may be generated whenever a preset reference speed is reached.

Preferably, in the transforming step, a Fast Fourier Transform (FFT) of performing a Fourier transform of only a portion of the signals of the frequency band selected in the filtering step may be performed.

Preferably, the abnormal detection step comprises: a distance calculation step of calculating the Euclidean distance value (D) of the reference data and the inspection data; and comparing the Euclidean distance value D with a preset threshold value D _th to determine whether an abnormality occurs in the drone.

The present invention collects vibration data generated when a drone is driven by a micro-vibration sensor, compares the collected data with preset reference data in a steady state, and applies an FFT conversion algorithm to process the collected data to speed up processing and data capacity can be dramatically improved. In addition, when comparing the collected data with the reference data, there is an advantage in that it is possible to implement a low-capacity algorithm that can be mounted on a small microprocessor by simplifying the calculation process by using the Euclidean distance calculation formula. Accordingly, the present invention has an advantage in that it is possible to provide an apparatus and method for diagnosing abnormalities of a drone suitable for the limited environmental characteristics of a small industrial drone.

1 is a schematic block diagram of a drone abnormality diagnosis apparatus according to an embodiment of the present invention.
2 to 4 are diagrams illustrating changes in frequency patterns in the first to third intended failure situations.
5 is a schematic flowchart of a method for diagnosing abnormalities in a drone according to an embodiment of the present invention.
6 is a schematic flowchart of an anomaly detection process according to an embodiment of the present invention.
7 to 18 are diagrams for explaining an embodiment of the present invention.

Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings, but it will be described in detail so that those of ordinary skill in the art to which the present invention pertains can easily practice the present invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. On the other hand, in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification. In addition, even if the detailed description is omitted, descriptions of parts that can be easily understood by those skilled in the art are omitted.

Throughout the specification and claims, when a part includes a certain element, it means that other elements may be further included, rather than excluding other elements, unless otherwise stated.

1 is a schematic block diagram of a drone abnormality diagnosis apparatus according to an embodiment of the present invention. Referring to FIG. 1 , the drone abnormality diagnosis apparatus 100 according to an embodiment of the present invention includes a vibration sensor 110 , a low pass filter (LPF) 120 , and a fast Fourier transform (FFT) converter 130 . ), an anomaly detection unit 140 and an effective frequency determination unit 150 .

The vibration sensor 110 is attached to an arm to collect vibration raw data. At this time, the vibration sensor 110 preferably has a resolution of 10 bits or more, a sensor axis is 3 axes or more, a scan speed is 4Ksps or more, and a communication speed is 400 kbps or more, and a miniature 3-axis acceleration sensor can be used.

The low-pass filter (LPF) 120 passes only a part of the raw vibration data collected by the vibration sensor 110 . That is, the low frequency filter (LPF) 120 passes only a valid frequency band (eg, 1 KHz band or less) for diagnosing an abnormality of the drone among the vibration raw data collected by the vibration sensor 110 . To this end, the low-frequency filter (LPF) 120 may apply a digital filter (T-filter). Meanwhile, the effective frequency band may be determined by the effective frequency determiner 150 .

The effective frequency determining unit 150 determines an effective frequency band for diagnosing an abnormality of the drone according to a frequency pattern that appears in a drone failure situation. To this end, the effective frequency determination unit 150 creates an arbitrary failure situation (eg, unbalance Prop. or loosely coupled bolts) for the drone, and the frequency indicated in the failure situation. detect the pattern. In addition, the effective frequency band may be determined based on the frequency pattern. For example, the effective frequency determination unit 150 may determine the effective frequency band by creating intended failure situations and detecting a frequency pattern for each. As described above, examples in which the effective frequency determination unit 150 detects a frequency pattern for each failure situation are exemplified in FIGS. 2 to 4 .

2 to 4 are diagrams illustrating changes in frequency patterns in the first to third intended failure situations.

2 is a diagram illustrating a change in a frequency pattern in a first failure situation (ie, a drone prop balance abnormal), and FIG. 2 (a) is a frequency when a drone prop in a normal state rotates at a speed of 210 RPM. The pattern is exemplified, and (b) of FIG. 2 illustrates the frequency pattern when the prop of the drone with an abnormal prop balance rotates at a speed of 210 RPM, and (c) of FIG. 2 is the prop of the drone in a normal state. Illustrating the frequency pattern when rotating at a speed of 620 RPM, (d) of FIG. 2 illustrates a frequency pattern when a prop of a drone having an abnormality in the prop balance rotates at a speed of 620 RPM, (e) of FIG. ) illustrates a frequency pattern when a prop of a drone in a normal state rotates at a speed of 1000 RPM, and (f) of FIG. 2 is a frequency pattern when a prop of a drone with an abnormal prop balance rotates at a speed of 1000 RPM exemplifies Referring to (b), (d), and (f) of FIG. 2, looking at the change in the frequency pattern of the drop where the prop balance is abnormal, the frequency pattern change in the 1KHz band and 100Hz band according to the rotation speed (RPM) of the prop found to be large. On the other hand, by comparing (a) and (b), (c) and (d), (e) and (f) of FIG. 2 , respectively, looking at the pattern change according to the abnormal state at the same RPM, low speed (210 RPM) No significant change was found in the 1KHz and 100Hz bands, but it can be seen that the pattern change in the 100Hz band appears significantly after 500 RPM. In addition, even in the 1KHz band, there was a significant change in some sections.

3 is a view illustrating a change in a frequency pattern in a second failure situation (that is, more than the first prop bolt of the drone is loosened), and (a) of FIG. 3 shows that the prop of the drone in a normal state rotates at a speed of 620 RPM. exemplifies the frequency pattern at the time, and (b) of FIG. 3 illustrates the frequency pattern when the prop of the drone, which is the second failure situation, rotates at a speed of 620 RPM. Meanwhile, FIG. 4 is a view illustrating a change in a frequency pattern in a third failure situation (ie, the second prop bolt of the drone is loosened), and FIG. 4(a) is a view showing the prop of the drone in a normal state at 620 RPM speed. The frequency pattern when rotating is exemplified, and Fig. 4 (b) illustrates the frequency pattern when the prop of the drone, which is the third failure situation, rotates at a speed of 620 RPM. Referring to FIGS. 3 and 4 , it can be seen that the pattern change occurs simultaneously in the 1KHz band and the 100Hz band.

Accordingly, the effective frequency determination unit 150, as illustrated in FIGS. 2 to 4 , may determine a frequency band of 0 to 1 KHz, which exhibits a significant pattern change in an abnormal state, as an effective frequency band.

Meanwhile, referring to FIGS. 2 to 4 , the 0 to 100 Hz band is a region in which harmonics according to the rotation speed of the propeller are generated, and in particular, the first harmonic within the 100 Hz band always coincides with the rotation frequency of the propeller can be known Therefore, the effective frequency determination unit 150 can process the data by special classifying the 0 to 100 Hz band as an important band for monitoring the abnormality of the propeller.

Therefore, in this case, the effective frequency determination unit 150 determines the effective frequency band to be less than or equal to 1 KHz band for diagnosing the abnormality of the drone, and transmits this information to the low frequency filter (LPF) 120 and the abnormality detection unit 140 . can transmit

The FFT (Fast Fourier Transform) converter 130 converts the data in the time domain into the frequency domain and outputs it as inspection data. That is, the FFT (Fast Fourier Transform) transformer 130 converts the time domain data output through the low frequency filter (LPF) 120 into the frequency domain, but by Fourier transforming only a part of the continuously sampled signal. After sampling, it is output as inspection data. Due to this, the FFT converter 130 can remarkably improve its processing speed and data capacity.

The abnormality detection unit 140 stores vibration data when the drone is in a normal state in advance as reference data, and compares the inspection data output from the FFT converter 130 with the reference data to detect whether the drone is abnormal. To this end, the abnormality detection unit 140 must first generate and store reference data. Accordingly, the abnormality detection unit 140 collects a plurality of vibration data based on a predetermined condition, and then calculates an average vector of the plurality of vibration data to generate reference data. That is, the abnormality detection unit 140 may extract a statistical index from the vibration signal data detected at a specific RPM, and store a vector composed of the statistical index in advance as the reference data. For example, the abnormality detection unit 140 may detect a normal state at regular frequency intervals with respect to the frequency of the first frequency band (eg, 1 KHz) determined as an effective frequency for the abnormal diagnosis of the drone by the effective frequency determining unit 150 . After collecting vibration data, an average vector of the collected data is calculated to generate reference data. In addition, the abnormality detection unit 140 calculates the peak value of the harmonic as an average vector with respect to a frequency of a second frequency band (eg, 100 Hz) in which a harmonic is generated according to the rotation speed of the drone to obtain reference data. create Meanwhile, when the rotation speed of the drone propeller is greater than or equal to a specific speed, the abnormality detection unit 140 may generate the reference data whenever a preset reference speed is reached. For example, the abnormality detection unit 140 designates 2-3 RPM (eg, 800, 1000, 1200, etc.) in an area where the rotation speed of the drone propeller exceeds 500 RPM and generates reference data at that point. can do. This is because the pattern change in the 100 Hz band appears greatly when the rotation speed of the drone propeller exceeds 500 RPM.

And after calculating the Euclidean distance value (D) of the reference data and the inspection data, the abnormality detection unit 140 compares the Euclidean distance value (D) with a preset threshold value (D _th ) to cause an abnormality in the drone determine whether it has occurred. That is, the abnormality detection unit 140 calculates the Euclidean distance value (D) of the reference data corresponding to the inspection data generated from the vibration data collected in real time at a specific RPM, and the Euclidean distance value (D) is a preset threshold. If the value (D _th ) or more, it is determined that an abnormality has occurred in the drone.

5 is a schematic flowchart of a method for diagnosing abnormalities in a drone according to an embodiment of the present invention. 1 and 5 , in step S110, reference data is generated. That is, in step S110, the abnormality detection unit 140 learns the vibration data when the drone is in a normal state and generates it as reference data. To this end, in step S110, the abnormality detection unit 140 collects a plurality of vibration data based on a predetermined condition, and then calculates an average vector of the plurality of vibration data to generate reference data. That is, in step S110, the abnormality detection unit 140 may extract a statistical index from the vibration signal data detected at a specific RPM, and store a vector composed of the statistical index in advance as the reference data. For example, in step S110 , the abnormality detection unit 140 may generate reference data by dividing it into two cases. First, the first effective frequency determined by the effective frequency determining unit 150 as an effective frequency for diagnosing an abnormality of the drone With respect to a frequency of a frequency band (eg, 1 KHz), the abnormality detection unit 140 may generate reference data by collecting vibration data in a steady state at regular frequency intervals and calculating an average vector of the collected data. And, with respect to a frequency of a second frequency band (eg, 100 Hz) in which a harmonic is generated according to the rotation speed of the drone, the abnormality detection unit 140 calculates the peak value of the harmonic as an average vector to obtain reference data. can create

On the other hand, in step S110, when the rotation speed of the drone propeller is greater than or equal to a specific speed, the abnormality detection unit 140 additionally generates the reference data whenever the rotation speed of the drone propeller reaches a preset reference speed. It may further include the step of For example, the abnormality detection unit 140 designates 2-3 RPM (eg, 800, 1000, 1200, etc.) in an area where the rotation speed of the drone propeller exceeds 500 RPM and generates reference data at that point. can do. This is because the pattern change in the 100 Hz band appears greatly when the rotation speed of the drone propeller exceeds 500 RPM.

If the reference data is generated in this way, in step S120, data is collected. That is, in step S120, the vibration sensor 110 collects raw vibration data generated when the drone is driven.

In step S130, filtering is performed. That is, in step S130 , the low frequency filter (LPF) 120 selects only a frequency band effective for diagnosing abnormalities of the drone from among the raw vibration data collected in step S120 . To this end, step S130 may further include a frequency band determining step (not shown) of determining the effective frequency band.

In the frequency band determining step (not shown), the effective frequency determining unit 150 determines the effective frequency band according to a frequency pattern appearing in a drone failure situation. To this end, the effective frequency determination unit 150 creates a random failure situation (eg, prop balance abnormal (unbalance Prop.) or prop bolt loosening (Loosely coupled bolts), etc.) for the drone. In addition, the effective frequency determination unit 150 may determine the effective frequency band by detecting the frequency pattern that appears in the failure situation. For example, the effective frequency determining unit 150 may determine the effective frequency band by creating a plurality of intended failure situations and detecting a frequency pattern for each. As such, an example of detecting a frequency pattern for each failure situation in the effective frequency determining unit 150 is illustrated in FIGS. 2 to 4 , and referring to FIGS. 2 to 4 , the effective frequency determining unit 150 is in an abnormal state. A frequency band of 0 to 1 KHz showing a significant pattern change may be determined as an effective frequency band.

In this way, when the frequency band of 0 to 1 KHz is determined as the effective frequency band in the step of determining the frequency band, in step S130, only the frequency band of 0 to 1 KHz is passed.

In step S140 , the FFT converter 130 converts the time domain data filtered in step S130 into the frequency domain and then outputs it as inspection data. In this case, in step S140, a Fast Fourier Transform (FFT) of performing a Fourier transform of only some of the signals of the frequency band selected in step S130 is performed.

In step S150, the abnormality detection unit 140 detects whether the drone is abnormal by comparing the inspection data with the reference data. An example of a more specific processing for detecting whether the drone is abnormal in step S150 is illustrated in FIG. 6 .

6 is a schematic flowchart of an anomaly detection process according to an embodiment of the present invention. With reference to FIGS. 1, 5 and 6 , a detailed processing process for the abnormality detection process ( S150 ) will be described as follows.

First, in step S151, the abnormality detection unit 140 calculates the Euclidean distance value D between the reference data and the inspection data. That is, the reference data generated in step S110 and the Euclidean distance value D of the inspection data generated from the vibration data collected in real time are calculated.

In steps S153 and S155, the abnormality detection unit 140 compares the Euclidean distance value D with a preset threshold value D _th to determine whether an abnormality occurs in the drone. That is, as a result of the comparison in step S153, if the Euclidean distance value D _{is equal to or greater than a preset threshold value D th} , it is determined in step S155 that an abnormality has occurred in the drone.

7 to 18 are diagrams for explaining an embodiment of the present invention, and are diagrams for explaining an experimental procedure for verifying the algorithm of the present invention.

7 to 12 are diagrams for explaining processes of generating reference data from a normally operating drone as examples. The abnormality detection unit 140 illustrated in FIG. 1 illustrates processes of generating reference data as an example. 13 to 16 are diagrams for explaining processes of producing a random failure situation and detecting a frequency pattern for it as an example, and the processing processes of the effective frequency determining unit 150 illustrated in FIG. 1 as an example 17 and 18 are diagrams illustrating average values of Euclidean distances calculated for each frequency band in a drone failure situation.

First, FIGS. 7 to 9 show that when a frequency band determined as an effective frequency for diagnosing an abnormality of a drone is 1 KHz, after collecting vibration data in a steady state at a predetermined frequency interval (ie, 100 Hz interval) with respect to the frequency of the 1 KHz band, As diagrams illustrating a process of generating reference data by calculating an average vector of the collected data as an example, FIG. 7 is an example of vibration data obtained when a propeller in a steady state is set to a rotation speed of 1,000 RPM and then operated. 8 is a diagram showing the vibration data as illustrated in FIG. 7 obtained five times and sampled, and then the 1KHz band is divided into 10 sections (F1 to F10) in units of 100Hz to obtain the average level for each section It is a diagram showing the measurement result in a table, and FIG. 9 is a diagram showing the result of calculating the average vector U for each of the 10 sections based on the data illustrated in FIG. 8 in a table.

Meanwhile, in FIGS. 10 to 12, when the frequency band in which harmonics are generated according to the rotation speed of the drone is 100 Hz, with respect to the frequency of the 100 Hz band, the peak value of the harmonic is calculated as an average vector to generate reference data. As drawings for explaining the process as an example, FIG. 10 is a view showing an example of a harmonic peak value obtained in a 100 Hz band when the propeller in a steady state is set to 1,000 RPM and then operated, and FIG. 11 is as illustrated in FIG. It is a table showing the results of measuring the same harmonic peak value 5 times in order up to 6 (P1 to P6), and FIG. 12 is an average vector (UL) for the 100Hz band harmonic peak value based on the data exemplified in FIG. It is a diagram showing the results of the calculation in a table.

13 to 16 are diagrams for explaining a situation in which an abnormality has occurred in the propeller balance of the drone as an example and processes of detecting a frequency pattern therefor, and FIG. 13 is a propeller in an abnormal state (ie, a propeller balance abnormal state). is a view showing an example of the vibration data measured after setting the rotation speed to 1,000 RPM, and FIG. 14 is a vibration data for each 10 sections (F1 to F10) within the 1KHz band of the propeller balance abnormal state as illustrated in FIG. 13. It is a diagram showing the average measured value in a table, and FIG. 15 is a diagram showing an example of the harmonic peak value obtained in the 100 Hz band when the propeller in an abnormal state (ie, propeller balance abnormal state) was set to 1,000 RPM and then operated, FIG. 16 is a diagram shown in a table by sequentially detecting up to six harmonic peak values (P1 to P6) as illustrated in FIG. 15 .

17 and 18 are views exemplifying the average value of the Euclidean distance calculated for each frequency band in the propeller balance abnormality of the drone, and FIG. 17 is the average value of the Euclidean distance (D) for the 1KHz section in the propeller balance abnormality of the drone 18 is a diagram illustrating an average value of the Euclidean distance (DL) for a 100 Hz section in an abnormal state of the propeller balance of the drone.

A calculation formula for calculating the Euclidean distance D using the values illustrated in FIG. 17 is exemplified in Equation 1.

In this case, d represents the number of 100 Hz unit sample intervals, and in the above example, the value is 10.

Meanwhile, a calculation formula for calculating the Euclidean distance DL using the values illustrated in FIG. 18 is exemplified in Equation (2).

In this case, d represents the number of harmonic peaks, and the value is 6 in the above example.

Referring to Equations 1 and 2, in the propeller balance abnormal state, the Euclidean distance D in the 1KHz band is 20.33, and the Euclidean distance DL in the 100Hz band is 37.10.

_{In general, when the Euclidean distance (D normal} ) is obtained based on experimental data sampled in a steady state, most of them are expressed in Equation (3).

Therefore, it is sufficiently meaningful to determine whether the drone operates abnormally by the values calculated in Equations 1 and 2 above.

That is, the present invention has the advantage of being able to detect in advance whether the drone is operating abnormally by the method as described above, and thereby prevent an accident in advance.

In the exemplary system described above, the methods are described on the basis of a flowchart as a series of steps or blocks, but the present invention is not limited to the order of steps, and some steps may occur in a different order or concurrently with other steps as described above. can

In addition, those skilled in the art will understand that the steps shown in the flowchart are not exhaustive and that other steps may be included or that one or more steps in the flowchart may be deleted without affecting the scope of the present invention.

Claims (14)

A vibration sensor unit that is implemented as a miniature sensor and collects raw vibration data generated when the drone is driven;
a filter unit that passes only an effective frequency band for diagnosing an abnormality of the drone among the raw vibration data collected by the vibration sensor;
a conversion unit that converts the data in the time domain that has passed through the filter unit into the frequency domain and outputs it as inspection data;
an abnormality detection unit that stores vibration data when the drone is in a normal state in advance as reference data, and compares the inspection data output from the converter with the reference data to detect whether the drone is abnormal; and
and an effective frequency determiner configured to detect a frequency pattern that appears by creating a random failure situation with respect to the drone, and determine the effective frequency band based on the result. delete The method of claim 1, wherein the effective frequency determination unit
An apparatus for diagnosing an abnormality of a drone, characterized in that the data is processed by special classification for a region in which harmonics are generated according to the rotation speed of the propeller among the frequency bands existing within the effective frequency band. According to claim 1, wherein the conversion unit
An apparatus for diagnosing an abnormality of a drone, characterized in that a Fast Fourier Transform (FFT) of only a part of the signals in the frequency band that has passed through the filter unit is Fourier-transformed. The method of claim 1, wherein the abnormality detection unit
After collecting a plurality of vibration data based on a predetermined condition, an average vector of the plurality of vibration data is calculated to generate reference data,
Calculate the Euclidean distance value (D) of the reference data and the inspection data,
An apparatus for diagnosing abnormalities of a drone, characterized in that it is determined whether an abnormality has occurred in the drone by comparing the Euclidean distance value (D) with a preset threshold value (D _{th ).} The method of claim 5, wherein the abnormality detection unit
For a frequency of a first frequency band, which is an effective frequency band for diagnosing abnormalities of the drone, steady state vibration data is collected at regular frequency intervals, and an average vector of the collected data is calculated,
With respect to a frequency of a second frequency band, which is a frequency band in which a harmonic is generated according to the rotation speed of the drone, the peak value of the harmonic is calculated as an average vector. The method of claim 6, wherein the abnormality detection unit
When the rotation speed of the propeller of the drone is greater than or equal to a specific speed, the apparatus for diagnosing an abnormality of a drone, characterized in that the reference data is generated whenever a preset reference speed is reached. A reference data generation step of learning the vibration data when the drone is in a normal state and generating it as reference data;
A data collection step of collecting raw vibration data generated when the drone is driven by using a micro sensor;
a filtering step of selecting only a valid frequency band for abnormal diagnosis of the drone from among the collected vibration raw data;
a conversion step of converting the data of the time domain selected in the filtering step into the frequency domain and outputting the data as inspection data;
an abnormality detection step of detecting whether the drone is abnormal by comparing the inspection data with the reference data; and
and a frequency band determining step of detecting a frequency pattern that appears by creating a random failure situation with respect to the drone, and determining the effective frequency band based on the result. delete The method of claim 8, wherein the generating of the reference data comprises:
and an average vector calculation step of generating reference data by calculating an average vector of the plurality of vibration data after collecting a plurality of vibration data based on a predetermined condition. 11. The method of claim 10, wherein the calculating of the average vector comprises:
a first average vector calculation step of collecting steady state vibration data at regular frequency intervals with respect to a frequency of a first frequency band, which is an effective frequency band for diagnosing abnormalities of the drone, and calculating an average vector of the collected data; and
and a second average vector calculation step of calculating the peak value of the harmonic as an average vector with respect to a frequency of a second frequency band, which is a frequency band in which harmonics are generated according to the rotation speed of the drone. method of diagnosing abnormalities. The method of claim 8, wherein the generating of the reference data comprises:
When the rotation speed of the propeller of the drone is greater than or equal to a specific speed, the reference data is generated whenever a preset reference speed is reached. The method of claim 8, wherein the converting step
A method for diagnosing an abnormality of a drone, characterized in that a Fast Fourier Transform (FFT) of performing a Fourier transform of only some of the signals of the frequency band selected in the filtering step is performed. The method of claim 8, wherein the detecting anomaly comprises:
a distance calculation step of calculating a Euclidean distance value (D) between the reference data and the inspection data; and
and comparing the Euclidean distance value (D) with a preset threshold value (D _th ) to determine whether an abnormality occurs in the drone.

Images