Comparative analysis between data-driven and visual-based approach in drone propeller imbalance detection

Abstract

Early monitoring of the drone’s critical components, such as the motor, propeller, and electronic speed controller (ESC), can prevent unwanted accidents from happening to the drone. A failure of one of the drone’s components may lead the drone to crash. The fault detection approach can be made by analyzing certain parameters such as vibration, acoustic, current, and speed. In this paper, a comparative analysis between data-driven and visual-based approaches in detecting an imbalanced propeller in the drone is presented. The analysis will focus on vibration and acoustic parameters, with the use of CZN-15 and ADXL335 MEMS sensors to capture vibration and acoustic data, respectively. The Silky EvCam vibration camera will be used for the visual-based approach. Experimental results demonstrated that both data-driven and visual-based approaches could detect the imbalanced UAV propeller. There is a shift of vibration frequency when imbalances occur in the UAV propeller. However, in the ground and thrust modes, the acoustic-based method has an accuracy of 54% and 93%, respectively, which is lower compared to the vibration-based (100% accuracy in both modes) and visual-based techniques.

Keywords

Acoustic camera drone fault detection vibration

1. Introduction

Drones, also known as unmanned aerial vehicles (UAVs), have revolutionized the way we interact with the world. These versatile flying machines have gained significant popularity in recent years. Equipped with cameras, sensors, and advanced technologies such as GPS and artificial intelligence, drones have been utilized in fields such as aerial photography Siean et al. (2021), mapping Gupta and Shukla (2018), search and rescue operations Martinez-Alpiste et al. (2021), precision agriculture Ghazali et al. (2022), and package delivery Kellermann et al. (2020).

While drones, particularly multirotors, offer exciting opportunities, it is crucial to address the safety issue surrounding their operation. Crashes and collisions commonly occur when operating a drone due to various reasons such as pilot error, environmental factors, collisions with animals, and failure of its components Balestrieri et al. (2021). Among these factors, the failure of drone components can be prevented early. It is essential to monitor critical components like the motor and propeller of a drone before initiating a flight. This preflight maintenance or checking procedure can effectively mitigate the risk of drone crashes.

In this paper, a fault detection approach in drones, focusing on the propeller imbalance case is proposed. Two different MEMS sensors will be utilized to capture the vibration and acoustic data. The main contribution of this paper is the comparative analysis between vibration (data-driven and visual-based) and acoustic parameters in detecting an imbalanced propeller before flying the drone. The aim is to identify the advantages and limitations of each approach and determine the best one to be applied. The rest of the paper is organized as follows. Section 2 discusses the literature review related to fault detection techniques in drones, and Section 3 presents the methodology for this study. Sections 4 and 5 explain the data-driven and visual-based propeller imbalanced detection, respectively, including the experimental results. The comparison between each approach is discussed in Section 6, and the study is concluded in Section 7.

2. Literature review

There have been many reported studies involving fault detection and diagnosis of drone components over the years, as shown in Table 1. Analyzing the vibration data is one of the best ways to determine the condition of the drone, and a growing body of literature exists on the subject. Ghalamchi and Mueller (2018) employed the onboard accelerometer to assess the vibration signal of the drone as a means of determining the location of the damaged propeller. Subsequently, their research was extended to enable real-time implementation Ghalamchi et al. (2019). Bondyra et al. (2017) presented an approach for identifying physical damage on UAV rotor blades. They employed the support vector machine (SVM) technique to classify signals and evaluate the presence and extent of a fault on the rotor blades. The analysis relied on features extracted from the vibration signal, obtained through methods such as fast Fourier transform (FFT) and wavelet packet decomposition (WPD).

Table 1.

Reported studies related to fault detection and diagnosis of drone components.

Study	Author and year	Component involved	Parameter	Drone’s type
1	Pechan and Sescu (2015)	Propeller	Acoustic	Components only
2	Bondyra et al. (2017)	Propeller	Vibration	Quadcopter
3	Saied et al. (2017)	Propeller and motor	Speed and current	Octocopter
5	Ghalamchi and Mueller (2018)	Propeller	Vibration	Quadcopter
6	Pourpanah et al. (2018)	Propeller and motor	Current and vibration	Quadcopter
7	Medeiros et al. (2018)	Propeller	Current	Quadcopter
8	de Jesus Rangel-Magdaleno et al. (2018)	Propeller	Acoustic	Quadcopter
9	Ghalamchi et al. (2019)	Propeller and motor	Vibration	Quadcopter and hexacopter
10	Iannace et al. (2019)	Propeller	Acoustic	Quadcopter
11	Banerjee et al. (2020)	Motor	Vibration	Octocopter
12	Lee et al. (2021)	Propeller and motor	Speed and current	Hexacopter
13	Altinors et al. (2021)	Propeller and motor	Acoustic	Components only
14	Zhang et al. (2021)	Propeller	Vibration	Quadcopter
15	Yang et al. (2021)	Propeller	Acoustic	Quadcopter

Pourpanah et al. (2018) examined the vibration characteristics of quadcopter propellers to distinguish between those in good condition and those with various degrees of damage (normal, 5%, 10%, or 15%). Based on the vibration data acquired from four accelerometers installed, the results revealed that the vibration signal generated by a healthy propeller exhibited a smoother pattern compared to that of a faulty propeller. Banerjee et al. (2020) presented a method for classifying unbalanced faults in UAVs by analyzing frequency-domain vibration signals. They utilized an ADXL335 accelerometer to capture vibration data from a quadcopter in a controlled indoor laboratory setting. The results demonstrated a maximum frequency shifting percentage of 4.2% when transitioning from a healthy state.

Iannace et al. (2019) utilized the noise generated by the drone to develop a classification model based on neural networks (NN) in order to identify broken propeller blades. The model they proposed demonstrated a 97% accuracy in detecting faulty propeller blades on drones. In the study conducted by Altinors et al. (2021), they proposed a method for detecting faults in drone motors using acoustic data. They employed machine learning techniques and achieved an accuracy of over 99% in detecting motor failures. Pechan and Sescu (2015) conducted experiments to investigate the noise generated by multiple propellers with various surface defects. The results revealed that certain surface flaws led to a slight increase in total noise at low rotational speeds. However, at higher speeds, there was a slight decrease in the overall noise level.

The detection and diagnosis of faults in drones primarily involve the analysis of parameters such as vibration, acoustic, and current. Previous studies have shown that vibration and acoustic data are commonly used to identify mechanical faults, such as broken and unbalanced propellers. On the other hand, current signals are better suited for detecting electrical faults, such as motor and ESC failures. Existing literature does not include any comparative analysis between different parameters for detecting similar fault cases. Furthermore, previous research has not addressed visual-based fault detection in drones, focusing solely on data-driven approaches. Choosing the appropriate method is crucial to ensure that the fault detection process achieves consistent success across various fault scenarios and operating conditions. Therefore, it is imperative to conduct further investigations into this matter.

3. Methodology

Figure 1 shows the flowchart of the proposed data-driven and visual-based propeller imbalanced detection approaches. Both approaches start with determining the fault detection mode, which can be divided into ground and thrust modes. For the data-driven method, the vibration (time-domain) and acoustic data for the healthy and imbalanced propeller are acquired using sensors. The time-domain vibration data are further analyzed in terms of maximum amplitude and root mean square (RMS). The vibration data are also converted to frequency-domain vibration data using the Fourier transform technique. The acoustic data are analyzed in terms of maximum and average amplitude. For the visual-based method, the vibration data of healthy and imbalanced propellers are analyzed in terms of frequency, which is directly obtained from the vibration camera.

Figure 1.

The flowcharts of the data-driven and visual-based approaches.

This research aims to compare data-driven techniques with the visual-based method for detecting imbalances in UAV propellers. This process is similar to airline pilots’ preflight checklist before takeoff, which serves as a preliminary step before deploying the UAV for a flight. The procedure involves conducting tests in indoor environments where the vibration and noise generated by the UAV are measured. Analyzing these measurements makes it possible to identify an imbalance in the propeller. To achieve this, vibration and acoustic data, as well as visual images, are collected from the UAV under four different operating conditions: a healthy propeller in ground mode, a healthy propeller in thrust mode, an unbalanced propeller in ground mode, and an unbalanced propeller in thrust mode.

The experimental setups for the data-driven and visual-based approaches are depicted in Figure 2. Based on Figure 2, an ADXL335 accelerometer was installed close to the vibration source (motor) to capture the vibration data. A CZN-15 sound sensor is installed in a way that the microphone faces the sound source (motor), as shown in Figure 2. The Silky EvCam camera from CenturyArks Co., Ltd is adopted for the visual-based approach. The camera incorporates event-based vision technology, an imaging approach that records scene changes instead of capturing individual image frames. Figure 2 shows how the camera is set to determine the vibration characteristic.

Figure 2.

The experimental setups for the data-driven and visual-based approaches.

ADXL335 accelerometer is a low-power sensor equipped with a built-in circuit for signal conditioning. It can measure static and dynamic acceleration within the ±3 g range. Its operational principle resembles the ADXL345 accelerometer, whereby movement of the mass causes a deflection resulting in an output signal from the sensor that is directly proportional to the acceleration Ghazali and Rahiman (2022). The omnidirectional microphone on the CZN-15E sound sensor (refer to Figure 2) converts the sensor’s sound waves into resistance. This resistance changes as the sound is louder or softer. The sensor’s sensitivity can also be adjusted by turning the potentiometer. The CZN-15E sound sensor has no physical unit. The technical specifications of both sensors are listed in Tables 2 and 3.

Table 2.

The technical specification of the ADXL335 accelerometer.

No	Specification	Information
1	Dimension	2.3 cm × 2 cm
2	Weight	1 g
3	Voltage used	2.5 V to 6 V
4	Output	Analog
5	Operating current	300 μA
6	Linear range	±3g
7	Sensitivity	300 mV/g

Table 3.

The technical specification of the CZN-15E sound sensor.

No	Specification	Information
1	Dimension	5 cm × 1 cm
2	Weight	3 g
3	Voltage used	3.3 V to 5 V
4	Output	Digital
5	Frequency range	20 to 16000 Hz
6	Sensitivity	$\approx$ −52 dB

A strip of paper tape was applied to the upper surface of the propeller to simulate the unbalanced condition, as depicted in Figure 3. The application of paper tape significantly alters the aerodynamics of the blade, thus characterizing the vibration and noise generated by the rotating blade Iannace et al. (2019). The ground mode in this study refers to the state where the UAV is on the ground with motors spinning at low speed, typically between 1200 and 1500 rpm. The UAV can also be viewed as “warming up” at this stage. The thrust mode is when the UAV’s motors are spinning at a faster rate than in the ground mode while still on the ground, but the generated propeller force is still insufficient to lift the UAV. The typical rotational speeds of the motor in this state are approximately 3000 to 4500 rpm.

Figure 3.

The unbalanced propeller configuration used in the experiment.

The experimental works are also conducted in varying environments to determine the performance of the data-driven and visual-based methods under disturbances. The propeller imbalanced detection process is carried out in the presence of wind (3.3 m/s wind speed measured using a digital anemometer) and outdoor environment. The wind is simulated by a standing fan positioned 20 cm from the sound sensor, whereas the outdoor temperature is approximately 33°C.

4. Data-driven propeller imbalanced detection

The data-driven propeller imbalanced detection can be divided into vibration and acoustic-based methods.

4.1. Vibration-based propeller imbalanced detection

The vibration results acquired from the ADXL335 accelerometer are divided into ground and thrust modes. The measurements from the ADXL335 accelerometer were first analyzed with MATLAB by transforming the time-domain data into frequency-domain data via the Fourier transform technique. The experiments are repeated five times at each mode to obtain more reliable data and results. The datasets are named Dataset 1, 2, 3, 4, and 5, representing the vibration data collected at experimental trials 1, 2, 3, 4, and 5, respectively.

4.1.1. Ground mode

Figure 4 shows the vibration frequencies obtained from the ADXL335 accelerometer for the healthy and imbalanced propeller conditions. As listed in Table 4, the frequencies obtained through the Fourier transform have an average of 56.732 Hz and 51.6665 Hz over five datasets for the healthy and imbalance propeller conditions, respectively. There are shifts of frequencies to the left for all datasets when imbalances occur in the UAV propeller with an average difference of 8.93%. The shifts in the vibration frequencies are an uneven distribution of mass or aerodynamic forces on the propeller blades. As the propeller rotates, it creates an uneven distribution of centrifugal force, resulting in different vibration characteristics.

Figure 4.

Frequency values obtained from the ADXL335 accelerometer in the ground mode.

Table 4.

Frequencies of different propeller conditions acquired from the accelerometer in the ground mode.

Dataset	Frequency
Dataset	Healthy propeller	Imbalanced propeller
1	57.7174 Hz	52.6978 Hz
2	57.0707 Hz	51.3794 Hz
3	56.5309 Hz	52.1484 Hz
4	56.0075 Hz	52.1118 Hz
5	56.3336 Hz	49.9878 Hz
Average	56.732 Hz	51.6665 Hz

Table 5 shows the maximum amplitude and root mean square (RMS) values acquired from the ADXL335 accelerometer for the healthy and imbalanced propeller cases. The maximum amplitude values are unreliable in detecting the imbalance case. However, the RMS values can be utilized as there are huge differences (minimum 82%) between healthy and imbalanced propellers. Utilizing the root mean square (RMS) values instead of the vibration frequency values can expedite the detection process.

Table 5.

Time-domain vibration data in the ground mode.

Dataset	Healthy propeller		Imbalanced propeller
Dataset	Maximum Amplitude	RMS	Maximum Amplitude	RMS
1	−0.16 g	0.0318 g	0.17 g	0.0877 g
2	0.29 g	0.0402 g	0.14 g	0.073 g
3	0.2 g	0.0306 g	0.16 g	0.0872 g
4	0.19 g	0.0353 g	0.17 g	0.0871 g
5	0.17 g	0.0312 g	−0.17 g	0.0856 g
Average	0.2 g	0.0338 g	0.16 g	0.0841 g

As mentioned before, the vibration-based propeller imbalanced detection under different environments is also conducted, and the findings can be observed in Figure 5. The imbalanced propeller can still be detected when the detection process is conducted outdoors or in the presence of wind with a speed of 3.3 m/s. There are shifts of frequencies to the left when imbalances occur in the UAV propeller, with a difference of 12.4% (windy environment) and 6.7% (outdoor environment).

Figure 5.

Frequency values obtained under different environments during the ground mode.

4.1.2. Thrust mode

Based on Table 6, for the healthy propeller condition, the frequencies obtained from the ADXL335 accelerometer show an average increase of 71.57% compared to the ground mode. The shift in the frequency values is due to a higher rotational speed of the motor, which produces greater forces that can cause the motor and its surroundings to vibrate more. Figure 6 shows the 13.42% decrease in the frequency value obtained from the ADXL335 accelerometer for an imbalanced propeller compared to the healthy propeller. The vibration results obtained are in agreement with Mircea et al. (2011), in which the vibration due to imbalance increases with rotational frequency Mohanta et al. (2017). Similar to the ground mode, it can be observed that the data-driven method employing the ADXL335 accelerometer can detect imbalances in the UAV propeller.

Table 6.

Frequencies of different propeller conditions acquired from the accelerometer in the thrust mode.

Dataset	Frequency
Dataset	Healthy propeller	Imbalanced propeller
1	98.6583 Hz	83.32 Hz
2	99.796 Hz	82.892 Hz
3	97.632 Hz	89.68 Hz
4	99.252 Hz	89.68 Hz
5	91.328 Hz	83.524 Hz
Average	97.3327 Hz	85.8192 Hz

Figure 6.

Frequency values obtained from the ADXL335 accelerometer in the thrust mode.

The time-domain data also can be utilized to distinguish the healthy and imbalanced propellers, as demonstrated in Table 7. The difference is that in the thrust mode, the maximum amplitude value is sufficient to detect imbalance. There are average increases of 211.96% and 365.53% in terms of maximum amplitude and RMS values, respectively, when an imbalance occurs. The vibration frequencies obtained under different environments can be observed in Figure 7. The imbalanced propeller can also be detected outdoors or in a windy condition during the thrust mode.

Table 7.

Time-domain data vibration data in the thrust mode.

Dataset	Healthy propeller		Imbalanced propeller
Dataset	Maximum amplitude	RMS	Maximum amplitude	RMS
1	0.37 g	0.1087 g	1.22 g	0.5641 g
2	−0.4 g	0.1094 g	1.05 g	0.404 g
3	−0.38 g	0.1043 g	1.29 g	0.3095 g
4	±0.38 g	0.1083 g	1.29 g	0.6886 g
5	−0.44 g	0.1085 g	1.27 g	0.5499 g
Average	0.39 g	0.1078 g	1.22 g	0.5 g

Figure 7.

Frequency values obtained under different environments during the thrust mode.

4.4. Acoustic-based propeller imbalanced detection

Similar to the vibration-based results, the acoustic results are divided into ground and thrust modes. The experiments are also repeated five times, and the data are labeled Datasets 1, 2, 3, 4, and 5. However, the obtained data consists of analog voltage readings ranging from 0 to 1023, rather than frequency data.

4.2.1. Ground mode

Figure 8 shows the time-domain representation of acoustic data for healthy and imbalanced propellers in the ground mode. The maximum amplitudes for all five datasets are in the range of 600 to 900, with an average amplitude value of 84.25. For the imbalanced propeller condition, the maximum amplitudes surpass 900, while the average amplitude value drops to 78.46. Based on the results, as shown in Figure 8, both maximum amplitude and average amplitude values can be utilized to detect the imbalanced UAV propeller.

Figure 8.

Acoustic data obtained in the ground mode.

The maximum amplitudes of acoustic data for a healthy propeller surpass 900 under a windy or noisy environment, as depicted in Figure 9. The average amplitudes for the healthy and faulty propellers increase to 115.6 and 90.5, respectively. Compared to the normal environment (absent of wind), maximum amplitude values cannot be utilized to detect the imbalanced propeller.

Figure 9.

Acoustic data obtained under different environments (ground mode).

4.2.2. Thrust mode

Figure 10 shows the time-domain representation of acoustic data for a healthy propeller in the thrust mode. The maximum amplitudes for all five datasets surpass 900, with an average amplitude value of 354.96. For the imbalanced propeller condition, the maximum amplitudes surpass 900, while the average amplitude value drops to 276.07, as depicted in Figure 10. The increase in the amplitude values is due to the higher noise produced during the thrust mode compared to the ground mode. It is recommended to only use the average amplitude value for the fault detection process since occasionally healthy propellers can display maximum amplitude beyond 900, which can also be identified as an imbalanced propeller. Under different environments, the acoustic data can be observed in Figure 11.

Figure 10.

Acoustic data obtained in the thrust mode.

Figure 11.

Acoustic data obtained under different environments (thrust mode).

5. Visual-based propeller imbalanced detection

5.1. Ground mode

Figure 12 shows the vibration frequencies obtained from the Silky EvCam camera for different propeller conditions in the ground mode. For a healthy propeller, the camera detected a frequency of 58 Hz, where there is a small average percentage difference of 2.19% between the measurements obtained from the camera and accelerometer in terms of frequency. For an imbalanced propeller, the camera records a vibration frequency of 47 Hz, specifically at the motor arm and the motor. Referring to Table 4, the average percentage difference between the frequencies obtained from the camera and accelerometer is 9.93%.

Figure 12.

Frequency values obtained from vibration camera (ground mode).

There is a 19% discrepancy between frequency values obtained from the Silky EvCam camera between healthy and imbalanced propellers. The discrepancy is slightly reduced to 5.56% when the detection process is conducted outdoors. As shown in Figure 13, the frequencies obtained for the healthy and imbalanced propeller cases are 54 H and 51 Hz, respectively. However, the frequencies remain the same in a noisy environment, which is similar to Figure 12.

Figure 13.

Frequency values obtained from vibration camera in outdoor environment (ground mode).

5.2. Thrust mode

Referring to Figure 14, for a healthy propeller condition, the vibration frequency estimated by the Silky EvCam camera is 85 Hz, a 46.55% increase compared to the frequency obtained in the ground mode. Based on Table 6, the average percentage difference between the frequencies obtained from the camera and accelerometer is 14.51%. For the imbalanced propeller, Figure 14 shows a 9.41% decrease in the frequency value compared to the healthy propeller condition. The vibration frequency estimated by the camera is 77 Hz, a slight decrease (10.28%) compared to the average frequency obtained from accelerometer data. Similar to the ground mode, it can be observed that the visual-based approach utilizing the Silky EvCam camera can detect imbalances in the UAV propeller during the thrust mode.

Figure 14.

Frequency values obtained from vibration camera (thrust mode).

Additionally, the imbalanced propeller can also be detected by the proposed visual-based approach in windy and outdoor environments. In the windy environment (indoors), the healthy and imbalanced propeller frequencies are similar to those of Figure 14. However, the frequencies drop by 2 Hz when the detection process is conducted outdoors, as depicted in Figure 15.

Figure 15.

Frequency values obtained from vibration camera in outdoor environment (thrust mode).

6. Discussion

Vibration and acoustic are some of the parameters that can be utilized to detect faults in a UAV. Based on the results presented in the previous sections, analyzing the vibration data measured from the ADXL335 accelerometer can detect an imbalance in the UAV propeller in different environments during the ground and thrust modes. The visual-based approach using the Silky EvCam camera can also detect an imbalanced propeller during both modes, as there is a shift of vibration frequency when imbalances are detected. The imbalance propeller can also be detected using the acoustic-based data-driven approach. However, based on Figure 10, the imbalanced propeller cannot be detected during the thrust mode because the maximum amplitudes between the healthy and imbalanced propellers are not reliable enough to distinguish between the two conditions. Nevertheless, the imbalanced propeller can still be identified during the thrust mode by utilizing average amplitude values instead of relying solely on maximum amplitudes.

Although the data-driven and visual-based approaches can be used to detect the imbalanced propeller, there are some advantages and limitations possessed by each technique. The detection process performed via the acoustic-based approach needs to be performed in environments with low noise. This is essential to prevent disruptions caused by the environment, as the noise can interfere with the accuracy of the results. Referring to Figures 9 and 11, additional disturbances in the form of noise produced by the wind will disrupt the acoustic data, whereas the vibration data are unaffected. Therefore, the maximum amplitude values cannot be utilized to detect the imbalanced propeller. Additionally, the proposed visual-based method is limited by the camera cable’s length, making it unsuitable to be implemented when the UAV is flying outdoors.

Based on 150 new samples of imbalanced propeller data (100 data per sample), the vibration-based method achieved 100% accuracy when the detection process was conducted in the ground modes without any disturbance. These accuracies are achieved by utilizing the RMS values of each sample. Similarly, the average amplitude values of 150 acoustic data samples are applied for the acoustic-based technique, and the accuracy acquired for the ground and thrust modes are 54% and 93%, respectively.

Regarding generalization capability, the proposed vibration-based data-driven and visual-based methods will not be affected by the wind conditions (up to 3.3 m/s wind speed). However, as mentioned before, a disturbance in the form of wind will reduce the performance of the acoustic-based technique, which can be seen in Figure 11. Furthermore, the proposed methods are only applicable to the current propeller dimension. Changing the propeller with propellers with different dimensions will affect the fault detection output.

7. Conclusion

The comparative analysis between data-driven and visual-based approaches in detecting the imbalanced UAV propeller has been presented in this study. Findings show that both approaches can detect the imbalanced UAV propeller in the ground and thrust modes. An imbalance in the UAV propeller will shift the vibration frequency to a lower value, as well as increase the maximum amplitude of the acoustic data. Nevertheless, in terms of accuracy, both the vibration-based and visual-based techniques outperform the acoustic-based approach. The latter exhibits an accuracy of only 54% in the ground mode and 93% in the thrust mode. In the future, the comparative analysis can be further implemented on other fault cases, such as motor failure, broken propeller, and cracks.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study is supported by CREST under Grant 304/PELECT/6050424/C121.

ORCID iD

Mohamad H. Mohd Ghazali

References

Altinors

Yol

Yaman

(2021) A sound based method for fault detection with statistical feature extraction in uav motors. Applied Acoustics 183: 108325.

Balestrieri

Daponte

De Vito

, et al. (2021) Sensors and measurements for uav safety: an overview. Sensors 21(24): 8253.

Banerjee

Okolo

Moore

(2020) In-flight detection of vibration anomalies in unmanned aerial vehicles. Journal of Nondestructive Evaluation, Diagnostics and Prognostics of Engineering Systems 3(4): 1–8.

Bondyra

Gasior

Gardecki

, et al. (2017) Fault diagnosis and condition monitoring of uav rotor using signal processing, In: 2017 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA). Poznan, Poland, 20-22 Sept. 2017, 233–238. IEEE.

de Jesus Rangel-Magdaleno

Ureña-Ureña

Hernández

, et al. (2018) Detection of unbalanced blade on uav by means of audio signal. In: IEEE International Autumn Meeting on Power, Electronics and Computing. Ixtapa, Mexico, 14-16 Nov. 2018, pp. 1–5. IEEE.

Ghalamchi

Mueller

(2018) Vibration-based propeller fault diagnosis for multicopters, In: 2018 International Conference on Unmanned Aircraft Systems (ICUAS), TX, USA, 12 Jun, 2018, 1041–1047. IEEE.

Ghalamchi

Jia

Mueller

(2019) Real-time vibration-based propeller fault diagnosis for multicopters. IEEE/ASME Transactions on Mechatronics 25(1): 395–405.

Ghazali

MHM

Rahiman

(2022) An investigation of the reliability of different types of sensors in the real-time vibration-based anomaly inspection in drone. Sensors 22(16): 6015.

Ghazali

MHM

Azmin

Rahiman

(2022) Drone implementation in precision agriculture—a survey. International Journal of Emerging Technology and Advanced Engineering 12: 67–77.

10.

Gupta

Shukla

(2018) Application of drone for landslide mapping, dimension estimation and its 3d reconstruction. Journal of the Indian Society of Remote Sensing 46: 903–914.

11.

Iannace

Ciaburro

Trematerra

(2019) Fault diagnosis for uav blades using artificial neural network. Robotics 8(3): 59.

12.

Kellermann

Biehle

Fischer

(2020) Drones for parcel and passenger transportation: a literature review. Transportation Research Interdisciplinary Perspectives 4: 100088.

13.

Lee

, et al. (2021) Fault classification and diagnosis of uav motor based on estimated nonlinear parameter of steady-state model. International Journal of Mechanical Engineering and Robotics Research 10(1): 22–31.

14.

Martinez-Alpiste

Golcarenarenji

Wang

, et al. (2021) Search and rescue operation using uavs: a case study. Expert Systems with Applications 178: 114937.

15.

Medeiros

Lima Filho

Ramos

JGG

, et al. (2018) A novel approach for speed and failure detection in brushless dc motors based on chaos. IEEE Transactions on Industrial Electronics 66(11): 8751–8759.

16.

Mircea

Radu

Danut

(2011) Critical analysis of vibration sources of hydro aggregates in operating regime. Sustainable Energy 2: 1–8.

17.

Mohanta

Chelliah

Allamsetty

, et al. (2017) Sources of vibration and their treatment in hydro power stations-a review. Engineering science and Technology, an international journal 20(2): 637–648.

18.

Pechan

Sescu

(2015) Experimental study of noise emitted by propeller’s surface imperfections. Applied Acoustics 92: 12–17.

19.

Pourpanah

Zhang

, et al. (2018) Anomaly detection and condition monitoring of uav motors and propellers. In: 2018 IEEE SENSORS, NJ, USA. IEEE, pp. 1–4.

20.

Saied

Lussier

Fantoni

, et al. (2017) Fault diagnosis and fault-tolerant control of an octorotor uav using motors speeds measurements. IFAC-PapersOnLine 50(1): 5263–5268.

21.

Siean

Vatavu

Vanderdonckt

(2021) Taking that perfect aerial photo: a synopsis of interactions for drone-based aerial photography and video, In: ACM International Conference on Interactive Media Experiences, 21-23. June 2021, New York, USA, Association for Computing Machinery(ACM).

22.

Yang

Geng

Wen

, et al. (2021) An intelligent quadrotor fault diagnosis method based on novel deep residual shrinkage network. Drones 5(4): 133–146.

23.

Zhang

Zhao

Wang

, et al. (2021) Fault detection and identification method for quadcopter based on airframe vibration signals. Sensors 21(2): 581–597.