Fault diagnosis method, device and system for aircraft

Abstract

The invention discloses a fault diagnosis method, device and system for an aircraft, and the method comprises the steps: obtaining a plurality of vibration signals, the plurality of vibration signals come from a plurality of vibration signal collection parts, the plurality of vibration signal collection parts are disposed on a fixed platform, the aircraft is fixed on the fixed platform, and the plurality of vibration signal collection parts are connected with the fixed platform; the plurality of vibration signal acquisition parts are used for acquiring the plurality of paths of vibration signals under the condition that a rotor system of the aircraft is started; and performing fault diagnosis on the rotor system of the aircraft based on the acquired multiple paths of vibration signals. Therefore, fault diagnosis of the rotor system of the aircraft is realized.

Description

Technical Field

[0001] This application relates to the field of fault diagnosis technology, and more specifically, to a fault diagnosis method, apparatus and system for an aircraft. Background Technology

[0002] With the rapid development of drone technology, quadcopter drones have been widely used in logistics delivery, security monitoring, power line inspection, environmental monitoring, and other fields. As the core power component of a drone, the rotor's operational status directly determines its flight safety. If the rotor experiences malfunctions such as broken rotor blades or cracks, it can lead to anything from loss of flight control and mission interruption to, in severe cases, drone crashes, resulting in equipment damage, personnel injuries, or the leakage of confidential data (such as drones used for monitoring stations). Therefore, timely and accurate fault diagnosis of the rotor system is a core requirement for ensuring the safe operation of drones.

[0003] Therefore, how to diagnose faults in rotor systems has become a technical problem that needs to be solved in this field. Summary of the Invention

[0004] In view of this, this application proposes a method, apparatus and system for diagnosing aircraft faults, so as to realize fault diagnosis of the rotor system of an aircraft.

[0005] In a first aspect, this application provides a fault diagnosis method for an aircraft, the fault diagnosis method comprising: acquiring multiple vibration signals, wherein the multiple vibration signals come from multiple vibration signal acquisition components, the multiple vibration signal acquisition components are mounted on a fixed platform, the aircraft is fixed on the fixed platform, the multiple vibration signal acquisition components acquire the multiple vibration signals when the rotor system of the aircraft is started; and performing fault diagnosis on the rotor system of the aircraft based on the acquired multiple vibration signals.

[0006] Optionally, based on the acquired multi-channel vibration signals, fault diagnosis of the aircraft is performed, including: recombining the acquired multi-channel vibration signals to construct a two-dimensional vibration signal matrix; determining the probability of each fault type based on the constructed two-dimensional vibration signal matrix and a trained neural network model; and performing fault diagnosis of the aircraft based on the determined probability of each fault type.

[0007] Optionally, the acquired multi-channel vibration signals are recombined to construct a two-dimensional vibration signal matrix, including: dividing each vibration signal into multiple segments of equal length; and splicing the multiple segments corresponding to the multi-channel vibration signals to construct the two-dimensional vibration signal matrix.

[0008] Secondly, this application also provides a fault diagnosis device for an aircraft, comprising: an acquisition module for acquiring multiple vibration signals, wherein the multiple vibration signals come from multiple vibration signal acquisition components, the multiple vibration signal acquisition components are mounted on a fixed platform, the aircraft is fixed on the fixed platform, and the multiple vibration signal acquisition components acquire the multiple vibration signals when the rotor system of the aircraft is started; and a diagnosis module for performing fault diagnosis on the rotor system of the aircraft based on the acquired multiple vibration signals.

[0009] Optionally, based on the acquired multi-channel vibration signals, fault diagnosis of the aircraft is performed, including: recombining the acquired multi-channel vibration signals to construct a two-dimensional vibration signal matrix; determining the probability of each fault type based on the constructed two-dimensional vibration signal matrix and a trained neural network model; and performing fault diagnosis of the aircraft based on the determined probability of each fault type.

[0010] Optionally, the acquired multi-channel vibration signals are recombined to construct a two-dimensional vibration signal matrix, including: dividing each vibration signal into multiple segments of equal length; and splicing the multiple segments corresponding to the multi-channel vibration signals to construct the two-dimensional vibration signal matrix.

[0011] Thirdly, this application also provides a fault diagnosis system for an aircraft, which performs fault diagnosis based on the above-mentioned fault diagnosis method. The fault diagnosis system includes: a fixed platform for supporting the aircraft; a fixed component for fixing the aircraft when it is located on the fixed platform; and multiple vibration signal acquisition components installed on the fixed platform.

[0012] Optionally, the fault diagnosis system further includes a calculation unit for performing fault diagnosis based on the fault diagnosis method described above.

[0013] Optionally, the computing unit is mounted on the fixed platform, and the computing unit and the aircraft are located on opposite sides of the fixed platform.

[0014] Optionally, the plurality of vibration signal acquisition components includes four vibration signal acquisition components.

[0015] Optionally, the fixed platform has a rectangular cross-section, and the four vibration signal acquisition components are distributed at the four corners of the fixed platform.

[0016] Optionally, the fixing component includes: a main body portion, the first end of which is a limiting hook, and the second end of which is threaded; and a nut, the second end of which passes through the fixing platform and is fixed by the nut.

[0017] Fourthly, this application also provides a machine-readable storage medium storing instructions that cause a machine to perform the above-described fault diagnosis method.

[0018] According to the technical solution of this application, multiple vibration signal acquisition components and the aircraft are fixed on a fixed platform. The multiple vibration signals acquired by the multiple vibration signal acquisition components can reflect the vibration status of the aircraft's rotor system. The vibration status of the rotor system can reflect the fault status of the rotor system. Thus, when the aircraft's rotor system starts up, multiple vibration signal acquisition components acquire multiple vibration signals, and fault diagnosis of the aircraft's rotor system is performed based on the multiple vibration signals, thereby realizing fault diagnosis of the aircraft's rotor system.

[0019] Other features and advantages of this application will be described in detail in the following detailed description section. Attached Figure Description

[0020] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application, and the illustrative embodiments and descriptions thereof are used to explain this application. In the drawings: Figure 1 This is a flowchart of a fault diagnosis method for an aircraft according to a preferred embodiment of this application; Figure 2 This is a schematic diagram of sliding sampling according to a preferred embodiment of this application; Figure 3 This is a structural block diagram of a fault diagnosis device for an aircraft according to a preferred embodiment of this application; Figure 4 This is a schematic diagram of the structure of a fault diagnosis system for an aircraft according to a preferred embodiment of this application; Figure 5 This is a top view of a fault diagnosis system for an aircraft according to a preferred embodiment of this application. Detailed Implementation

[0021] The technical solution of this application will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] In a first aspect, embodiments of this application provide a method for diagnosing aircraft faults.

[0023] Figure 1 This is a flowchart of a fault diagnosis method for an aircraft according to a preferred embodiment of this application. The fault diagnosis method is used to monitor the operational status and identify faults in the rotor system of an aircraft under fixed platform conditions. The aircraft may preferably be a quadcopter unmanned aerial vehicle (UAV). Figure 1 As shown, the fault diagnosis method may include the following:

[0024] In step S10, multiple vibration signals are acquired. These multiple vibration signals originate from multiple vibration signal acquisition components mounted on a fixed platform. The aircraft is fixed to the fixed platform via a fixing component to establish a stable vibration transmission path after the rotor system is activated. When the aircraft's rotor system is activated, the multiple vibration signal acquisition components simultaneously acquire the multiple vibration signals. Optionally, in this embodiment, the vibration signal acquisition components may be accelerometers, and the multiple vibration signals include the output signals of multiple accelerometers mounted on the fixed platform. In this embodiment, the multiple vibration signals may be the output signals of four accelerometers, used to cover the fundamental frequency, harmonics, and fault impact components of the quadcopter rotor system.

[0025] In step S11, fault diagnosis is performed on the aircraft's rotor system based on the acquired multi-channel vibration signals. For example, the probability of each fault type can be determined based on the acquired multi-channel vibration signals, and fault diagnosis can be performed according to the determined probabilities. In this embodiment, the fault types may include, but are not limited to, different types of faults such as normal state, rotor blade breakage, rotor crack, rotor imbalance, and motor failure.

[0026] Optionally, in this embodiment of the application, fault diagnosis of the aircraft based on the acquired multi-channel vibration signals may include the following:

[0027] The acquired multi-channel vibration signals are reassembled to construct a two-dimensional vibration signal matrix. Based on the constructed two-dimensional vibration signal matrix and a trained neural network model, the probability of each fault type is determined. Based on the determined probabilities of each fault type, fault diagnosis of the aircraft is performed. The two-dimensional vibration signal matrix is ​​input into the pre-trained neural network model, and temporal texture features, high-frequency impact information, and cross-channel correlation features are extracted through multi-scale convolution, feature flattening, and fully connected classification modules. Finally, the probability value of each fault type is output. If the probability of a certain abnormal type exceeds a preset threshold, it can be determined that there is a corresponding fault in the rotor system, and further prompt information or abnormal events can be output.

[0028] Optionally, in the embodiments of this application, recombining the acquired multi-channel vibration signals to construct a two-dimensional vibration signal matrix may include the following:

[0029] Each vibration signal is divided into multiple segments of equal length. The length of each segment and the total number of segments can be determined based on specific circumstances. One possible strategy is to ensure that each segment contains local vibration modes of multiple micro-cycles of the rotor, thereby enhancing feature representation capabilities. For example, by continuously acquiring vibration signals for 0.1 seconds at a sampling rate of 20kHz, resulting in a vibration sequence of 2000 points per channel, the approximately 2000 points acquired in 0.1 seconds can be divided into 20 segments, each 100 points long, ensuring that each segment contains local vibration modes of multiple micro-cycles of the rotor and enhancing feature representation capabilities.

[0030] Multiple segments corresponding to the various vibration signals are spliced ​​together to construct a two-dimensional vibration signal matrix. Specifically, the segments corresponding to the various vibration signals are spliced ​​together in a preset order to obtain the two-dimensional vibration signal matrix. This two-dimensional matrix can simultaneously express the periodic vibration modes of the rotor system, the platform-aircraft coupling characteristics, and the fault impact characteristics.

[0031] In this embodiment, the method of segmenting and splicing multiple vibration signals enables spatial structure encoding of data segments of finite duration. This allows the two-dimensional vibration signal matrix to simultaneously reflect temporal local patterns and multi-channel structural responses, providing richer input formats for neural networks. The equal-length slicing method ensures alignment of signals from different channels in the temporal dimension, thereby guaranteeing consistency in feature construction.

[0032] For example, there are four vibration signals. According to the above division method, each of the four vibration signals generates 20 sub-segments, forming a total of 80 signal segments of equal length. These segments are then spliced ​​together in a preset order along the row direction to obtain a two-dimensional vibration signal matrix with a size of 80×100. This matrix is ​​used to simultaneously express the propagation mode of aircraft vibration on different structural paths, the platform-aircraft coupling characteristics, and the texture structure of the rotor periodic signal.

[0033] Optionally, in this embodiment, the neural network model is a fault classification model trained on vibration data. The neural network model may include a multi-scale convolutional feature extraction module, a feature flattening module, and a fully connected classification module. The multi-scale convolutional feature extraction module may include a three-layer multi-scale convolutional structure. The constructed two-dimensional vibration signal matrix is ​​input into the trained neural network model to extract periodic features, local texture features, and high-frequency impact features, generating a high-dimensional feature vector. This high-dimensional feature vector, after being processed by a classification layer, is used to output the probability of each fault type in the aircraft rotor system. In this embodiment, the fault types may include, but are not limited to, normal operation, rotor breakage, rotor crack, rotor imbalance, motor failure, and other different types of faults.

[0034] In this embodiment, the neural network model can use different convolution kernel scales to extract vibration features of different frequency components. Large-scale convolution kernels are used to capture the rotor fundamental frequency and the gradual change mode of the structure, while small-scale convolution kernels are used to capture the high-frequency impact mode caused by the fault, thereby realizing the fusion of multi-scale features of the vibration signal.

[0035] In this application's embodiments, the core idea of ​​the fault diagnosis method is to utilize the temporal correlation of vibration signals to improve feature distinguishability through structured reconstruction, and to perform efficient feature compression and classification in a low-dimensional space. The technical solution provided in this application's embodiments is illustrated below using four vibration signals as an example.

[0036] Vibration signals were continuously acquired for 0.1 seconds at a sampling rate of 20kHz, resulting in a vibration sequence of 2000 points per channel. Therefore, the dimensions of the original input signal are: The vibration channels at four support points cover the main vibration modes of the avionics structure (fixed platform). The 2000 points are derived from the calculation of the sampling rate and acquisition time: 2000 = 20000Hz × 0.1s. A 20kHz sampling rate can simultaneously cover the fundamental mechanical frequency of the rotor (approximately 50Hz), harmonic components (100Hz, 150Hz, etc.), and high-frequency impact components (1–10kHz) caused by blade cracks and damage. The fundamental mechanical frequency of the UAV rotor is determined by the motor speed; one cycle of 50Hz is approximately 20ms. Therefore, 0.1 seconds of data contains approximately five complete rotor cycles, which can fully cover the periodic structural characteristics.

[0037] To enable the neural network model to simultaneously focus on the structural features of the vibration signal at different time scales, this application divides each 2000-point signal into 20 sub-segments, each with 100 points, resulting in a segmentation matrix: 1×2000→20×100, the physical meaning of which is shown in Table 1.

[0038] Table 1 The four signals are segmented and then spliced ​​together along the row direction to obtain: The matrix consists of four vibration signals, each with 20 local time segments, totaling 80 rows, representing the joint response pattern of the nest structure under 80 local vibration windows; and 100 columns representing the details of the original vibration waveform for each local window. This matrix can be viewed as a "vibration image" containing periodic structural texture, local waveform changes, and impact components, providing structured input for subsequent convolutional networks.

[0039] In this embodiment, the multi-scale convolutional feature extraction module can employ a three-layer multi-scale convolutional structure, targeting the large-scale periodic features of the vibration signal (related to the fundamental frequency), the medium-scale texture features (related to harmonics and load disturbances), and the small-scale transient impact features (related to minor impacts from cracks). The size of each convolutional kernel is determined by the rotor fundamental frequency (50Hz, period approximately 20ms), the length of the reconstructed segment (100 points), and the system vibration frequency band design, thus possessing a clear physical meaning. The following layered descriptions illustrate this.

[0040] The first convolutional layer, Conv1, extracts the periodic vibration structure of the rotor. Input The kernel size is (4, 100), as shown in Table 2.

[0041] Table 2 Therefore, Conv1 primarily identifies macroscopic characteristics such as the rotor's fundamental frequency, harmonics, overall periodic vibration shape, and motor uniformity, and outputs... As shown in Table 2.

[0042] Table 2 The second convolutional layer, Conv2, extracts mesoscale texture vibration features. (Input...) The kernel size is (2, 50), as shown in Table 3.

[0043] Table 3 Output As shown in Table 4.

[0044] Table 4 The third convolutional layer, Conv3, extracts transient impacts and high-frequency weak features. (Input...) The kernel size is (1,25), as shown in Table 5.

[0045] Table 5 Output As shown in Table 6.

[0046] Table 6 For the feature flattening module and the fully connected classification module, the input features are the features output from the third layer. It is a multi-scale vibration feature map of the current operating state of the UAV, including periodic features (fundamental frequency correlation), mid-frequency texture and harmonic structure, and high-frequency transient impact features.

[0047] For the feature flattening module, after flattening, we get D = 10 × 12 × C3, as shown in Table 7.

[0048] Table 7 For fully connected classification modules (FC) Where D is the "health and fault state vector" of the intelligent diagnostic model, and K is the probability of the fault category, such as normal state, rotor blade breakage, rotor crack, rotor imbalance, motor failure, etc.

[0049] The fully connected layer and Softmax map the health vector to the probability of different failure modes, thus achieving the final diagnosis.

[0050] Detailed parameters for each module are shown in Table 8.

[0051] Table 8 Optionally, in this embodiment of the application, fault diagnosis of the aircraft based on the determined probability of each fault type can be performed by pre-setting thresholds. Specifically, if the probability of a certain fault type exceeds the preset threshold, the rotor system can be determined to be in a corresponding abnormal state, and a warning can be generated or protective measures can be implemented.

[0052] When a fault is detected in the rotor system, fault prompts can be output, fault logs can be recorded, or diagnostic results can be transmitted to a host computer for subsequent maintenance, data analysis, or model retraining.

[0053] Optionally, in this embodiment, before performing fault diagnosis on the rotor system of the aircraft based on the acquired multi-channel vibration signals, the multi-channel vibration signals can be preprocessed, including noise reduction, filtering, normalization, and time synchronization calibration of the multi-channel vibration signals. This processing improves the usability of the vibration signals and the stability of subsequent feature construction. Specifically, the vibration signals can be preprocessed in the frequency domain or time-frequency domain, including but not limited to noise reduction, filtering, and normalization, to enhance the observability of periodic vibration modes and fault impact components, thereby improving signal quality and enhancing fault characteristics.

[0054] Alternatively, in embodiments of this application, the neural network model can be trained according to the following:

[0055] Set an appropriate sampling rate and sample multiple vibration signals. Simulate typical faults such as rotor blade breakage and cracks. Fix the aircraft on a fixed platform and power it on to start rotating to the set speed.

[0056] The multiple vibration signals are preprocessed separately, including at least one of the following: 1) Standardization (StandardScaler). 2) Filtering: Bandpass filtering is applied to the fundamental frequency of the propeller rotation speed, ensuring that the collected vibration signals mainly retain frequency components related to the fundamental frequency and its harmonics, while filtering out non-rotor-related frequency components. 3) Framing and data augmentation processing, including but not limited to sample mixing and noise introduction methods, to expand the training sample distribution and enhance the model's adaptability to signal fluctuations.

[0057] Construct a fault signal dataset. Select samples using a sliding window combined with multi-statistic scoring. Sliding sampling can employ a 50% overlap in window step size, such as... Figure 2 As shown.

[0058] The collected samples include the time-domain and frequency-domain characteristics of the vibration signals. The time-domain and frequency-domain indicators of the vibration signals include variance, kurtosis, skewness, spectral entropy, kurtosis, standard deviation, etc., as shown in Table 9.

[0059] Table 9 Typical samples were selected by combining Min–Max normalization of statistics with empirical weights. Fault labels were created for the selected samples, and a fault signal dataset was constructed to train the neural network model. For example, the format of the fault label could be "speed gear & fault type", such as "low speed & Normal" or "low speed & One propeller Broken (No.1)".

[0060] Optionally, the fault diagnosis method also includes: outputting fault prompt information or recording fault data when it is determined that the aircraft has a fault.

[0061] The technical solution provided by the embodiments of this application can utilize multiple vibration signals under a fixed platform, enhance the vibration feature expression capability through two-dimensional matrix recombination, and combine with multi-scale neural networks to achieve high-precision fault diagnosis of the aircraft rotor system, thereby improving the reliability of aircraft structural health monitoring.

[0062] Secondly, embodiments of this application also provide an aircraft fault diagnosis device.

[0063] Figure 3 This is a structural block diagram of an aircraft fault diagnosis device according to a preferred embodiment of this application. The aircraft can be a quadcopter unmanned aerial vehicle (UAV). Figure 3As shown, the fault diagnosis device includes an acquisition module 10 and a diagnosis module 20. The acquisition module 10 acquires multiple vibration signals from multiple vibration signal acquisition components mounted on a fixed platform. The aircraft is fixed to the fixed platform, and the multiple vibration signal acquisition components acquire the multiple vibration signals when the aircraft's rotor system is activated. The diagnosis module 20 performs fault diagnosis on the aircraft's rotor system based on the acquired multiple vibration signals.

[0064] Optionally, based on the acquired multi-channel vibration signals, fault diagnosis of the aircraft is performed, including: reconstructing the acquired multi-channel vibration signals to construct a two-dimensional vibration signal matrix; determining the probability of each fault type based on the constructed two-dimensional vibration signal matrix and a trained neural network model; and performing fault diagnosis of the aircraft based on the determined probability of each fault type.

[0065] Optionally, the acquired multi-channel vibration signals are recombined to construct a two-dimensional vibration signal matrix, including: dividing each vibration signal into multiple segments of equal length; and splicing the multiple segments corresponding to the multi-channel vibration signals to construct a two-dimensional vibration signal matrix.

[0066] The specific working principle and benefits of the fault diagnosis device provided in this application are similar to those of the fault diagnosis method provided in this application, and will not be repeated here.

[0067] Optionally, the diagnostic module 20 can upload the diagnostic results to an external host computer system or store them locally for use in maintaining decision-making, recording historical data, or for subsequent model updates.

[0068] The technical solution provided by the embodiments of this application adopts a fixed platform and multi-point vibration acquisition method, and uses two-dimensional vibration matrix construction and intelligent neural network model analysis to achieve high-precision and high-reliability fault diagnosis of aircraft rotor system. Compared with traditional single-channel or one-dimensional signal processing methods, it can significantly improve the robustness and accuracy of fault identification.

[0069] Thirdly, this application also provides a fault diagnosis system for an aircraft. The fault diagnosis system performs fault diagnosis based on the above-described fault diagnosis method and constructs a complete fault detection platform for a fixed aircraft rotor system. The fault diagnosis system includes a fixed platform, fixed components, and multiple vibration signal acquisition components. The fixed platform is used to support the aircraft; the fixed components are used to fix the aircraft when it is placed on the fixed platform; and the multiple vibration signal acquisition components are mounted on the fixed platform to collect vibration signals of the aircraft during rotor system operation.

[0070] Optionally, the fault diagnosis system further includes a computing unit for performing fault diagnosis based on the fault diagnosis method described in the above embodiments.

[0071] Optionally, the computing unit can be mounted on a fixed platform, with the computing unit and the aircraft located on opposite sides of the fixed platform, so as to facilitate a compact system structure, simple wiring, and reduced signal transmission interference.

[0072] Optionally, the multiple vibration signal acquisition components may include four vibration signal acquisition components. For example, the multiple vibration signal acquisition components may include four acceleration sensors.

[0073] Optionally, the fixed platform has a rectangular cross-section, and four vibration signal acquisition components are distributed at the four corners of the fixed platform to obtain the vibration response at different locations on the platform. For example, the fixed platform has a rectangular cross-section, and four sensors are fixed at the four corners of the platform.

[0074] Four vibration signal acquisition units are positioned at the four corners of the rectangular fixed platform, covering the main modal response paths of the platform structure. This enhances the sampling effect of vibration characteristics in different orientations of the aircraft and increases the spatial diversity of vibration acquisition. After the aircraft rotor system is started, the acquisition module simultaneously acquires four acceleration signals through a synchronous sampling circuit, ensuring that the multiple signals are aligned on the time axis. This allows subsequent data reconstruction to accurately reflect the propagation characteristics of vibration along different structural paths.

[0075] Optionally, the fixing component includes a main body and a nut. The first end of the main body forms a limiting hook structure for engaging with the aircraft landing gear or fuselage for positioning; the second end of the main body has a threaded structure that passes through the fixing platform and is secured by a nut for stable locking.

[0076] Figure 4 This is a schematic diagram of the structure of an aircraft fault diagnosis system according to a preferred embodiment of this application. Figure 4 As shown, the fault diagnosis system mainly includes a fixed platform 1, four acceleration sensors 2, fixed components 4, and a computing unit 5.

[0077] Fixed platform 1 is used to support aircraft 3. For example... Figure 4 As shown, the fixed platform 1 is a rectangular metal platform to ensure platform rigidity and stability while providing a good vibration transmission path. Four accelerometers 2 are mounted on the fixed platform 1 to synchronously collect four vibration signals when the rotor system of the aircraft 3 starts. The four accelerometers 2 are respectively installed at the four corners of the rectangular fixed platform, corresponding to the four support points of the frame, to cover the main structural modal positions of the platform and improve the spatial integrity of the vibration response. Optionally, the accelerometers can be single-axis accelerometers.

[0078] In this embodiment, single-axis (z-axis direction) accelerometers are arranged at the four corners of the fixed platform 1. Their technical advantages are mainly reflected in the following points: 1) Highly targeted, directly addressing key coupling paths. Vibrations on the fixed platform (aluminum plate) mainly originate from force transmission perpendicular to the plate surface—including propeller thrust fluctuations and the z-axis component of unbalanced centrifugal force. Aligning the sensor axis with the Z-axis maximizes the capture of vibration energy generated by these fault excitations, avoiding interference from other directions and significantly improving the signal-to-noise ratio of the fault signal. 2) Corner placement, covering the lowest-order modal peaks. Placing single-axis accelerometers at the four corners of the fixed platform 1 aligns with the mode shapes of its inherent vibration modes. The lowest-order mode often forms a "belly" at the center of the plate surface, while the four corners are the "belly" positions with the largest amplitude. This allows for effective monitoring of the first and second-order modal peaks caused by different fault excitations, ensuring comprehensive acquisition of various fundamental and harmonic components.

[0079] The aircraft 3 is fixed to the fixed platform 1 by the fixing component 4. For example... Figure 4 and Figure 5 As shown, the fixing component 4 includes a main body and a nut. The first end of the main body is a bent limiting hook, which is used to contact the landing gear of the aircraft 3 and provide lateral limiting constraint. The second end of the main body has a threaded structure, which passes through the fixing platform 1 and is fastened by the nut to provide preload in the vertical direction, so that the aircraft can be reliably fixed to the surface of the fixing platform. This structure ensures that the aircraft maintains a stable attitude when the rotor is running at high speed and maintains the repeatability of vibration signal acquisition. In this embodiment, the number of fixing components 4 can be determined according to specific circumstances, such as... Figure 5 As shown, a total of four fixed components are used.

[0080] The calculation unit 5 is used to perform fault diagnosis based on four vibration signals collected by four accelerometers 2. For example... Figure 4 As shown, the computing unit 5 is installed below the fixed platform 1 and isolated from the platform by a vibration damping bracket to prevent the vibration of the processing unit itself from affecting the signal judgment. The computing unit 5 can perform vibration signal preprocessing, two-dimensional vibration signal matrix construction, neural network inference and output the probability of each fault type according to the fault diagnosis method described in the above embodiment, and finally realize the automated fault diagnosis of the aircraft rotor system.

[0081] like Figure 4 As shown, the computing unit 5 and the aircraft 3 are located on opposite sides of the fixed platform 1, that is, the computing unit 5 is located at the bottom of the fixed platform 1 and the aircraft 3 is located at the top of the fixed platform 1, which isolates the data acquisition and computing parts spatially and improves the stability of the system operation.

[0082] Optionally, this system can not only perform real-time fault diagnosis, but also upload the diagnosis results to the ground station or host computer via the communication interface for task recording, automatic maintenance strategy implementation, model updates, and long-term structural health monitoring.

[0083] Optionally, the system can also store and re-analyze vibration data under different operating conditions, including different rotor speeds, single-propeller or multi-propeller removal conditions, and motor failure simulation conditions, in order to enrich the training dataset and improve the robustness of the neural network model in real-world scenarios.

[0084] In this embodiment, the system enables automatic takeoff and landing of the UAV and real-time operational status monitoring. During status monitoring, the fixed platform serves as a test platform. The aircraft is fixed to the test platform, and accelerometers are installed at the four corners of the platform. The aircraft's rotor is activated, and vibration signals are collected through the accelerometers. The signals are then analyzed to diagnose the rotor. Simultaneously, the fixed platform is equipped with an edge computing unit for online processing of vibration signals.

[0085] The edge computing unit can be installed in the center below the fixed platform and secured with a metal bracket and vibration damping isolation components. The computing unit connects to accelerometers located at the four corners of the fixed platform, and its integrated ADC or external data acquisition module (acquisition card) synchronously acquires multiple vibration signals in real time. After the aircraft lands and is fixed to the platform, the edge computing unit automatically triggers preprocessing steps such as vibration data acquisition, bandpass filtering, normalization, and sliding window framing. Based on a pre-built lightweight neural network, it performs online inference to quickly identify typical faults such as rotor blade breakage, cracks, and motor jamming. The diagnostic results are returned to the nest control module and cloud platform for automatic determination of whether to allow takeoff again.

[0086] Fixed platforms can be deployed in logistics centers or monitoring stations for 24 / 7 unmanned inspection. Flight and diagnostic commands can be issued via a remote cloud platform.

[0087] After completing its mission, the aircraft autonomously returns to the fixed platform and lands there. It is then rigidly attached to the platform using a hook or mechanical gripper, effectively utilizing the platform's idle time window to perform rapid inspections of the aircraft. This significantly improves overall operational efficiency and system-wide operational capabilities. If a potential fault is detected, the system will automatically lock takeoff clearance and issue an alarm.

[0088] In the embodiments of this application, the aircraft can be a quadcopter drone, typically a commercial / industrial grade rotary-wing drone, and any rotary-wing drone that can be rigidly fixed to a fixed platform is acceptable.

[0089] In this application, a rigid connection assumption is adopted, in which the aircraft and the fixed platform are rigidly connected by a diagonal hook and are regarded as a single rigid body structure, and the vibration can be transmitted directly without attenuation.

[0090] In this application, the fundamental frequency of the propeller and its harmonics generate clear vibrational spectral peaks through the steel and aluminum structure of the fixed platform.

[0091] In this application, standard signal samples can be obtained based on the following: Under identical conditions, using a test platform identical to the fault condition, the same platform mounting method, sensor arrangement, and different rotational speeds, 10-second vibration data are collected for each component, using intact, defect-free blades and motors.

[0092] To remove the influence of random noise, the sampling was repeated at least 3 times at each speed setting, and the sample with the highest signal-to-noise ratio was taken as the "standard sample".

[0093] In this embodiment of the application, the standard signal samples are used as normal state samples in the training process of the neural network model to participate in the construction of the training dataset for learning the vibration characteristic distribution of the aircraft rotor system under healthy operating conditions.

[0094] Characterization of Standard Samples. To verify and describe the health status of standard signal samples from a physical mechanism perspective, spectral characteristics, propeller fundamental frequency peak value, peak amplitude and width, and statistical indicators are used for characterization. Propeller fundamental frequency peak value: In the FFT spectrum, the fundamental frequency and its integer multiples of harmonic peaks (e.g., 50Hz, 100Hz, 150Hz…) should appear one-to-one with the propeller rotational speed. Peak amplitude and width: The fundamental frequency peak of a healthy propeller typically has a controllable amplitude and narrow spectral peak, while the corresponding peak of a faulty propeller will show spikes, jitter, or broadening. Statistical indicators: The mean of the time-domain standard signal is close to 0; the spectral entropy and bandwidth in the frequency domain are at low levels, indicating that the signal components are concentrated in a few harmonic peaks.

[0095] Other characteristic parameters. Based on the above core physical characterization, auxiliary characteristic parameters are further introduced to describe the statistical stability and distribution characteristics of vibration signals. These time-domain / frequency-domain indicators include variance, kurtosis, skewness, spectral entropy, kurtosis, and standard deviation, which are used to characterize the dispersion, non-Gaussianity, and complexity of vibration signals under different fault conditions.

[0096] The aforementioned core characterization features and auxiliary statistical features are mainly used for standard sample construction, data consistency verification, and quality assessment during model training, and are not directly used as input to the neural network model; the input to the neural network model is a two-dimensional vibration signal matrix obtained by reconstructing multiple vibration signals.

[0097] The fault diagnosis system provided by the embodiments of this application can fix the aircraft in a controlled structural environment. By having vibration acquisition components arranged in multiple positions on the fixed platform and intelligent processing units work together, it can achieve rapid, stable and high-precision diagnosis of rotor system faults, thereby improving the safety, reliability and maintenance efficiency of the UAV system.

[0098] Fourthly, this application also provides a machine-readable storage medium storing instructions that cause a machine to perform the fault diagnosis method described in the above embodiments.

[0099] The preferred embodiments of this application have been described in detail above. However, this application is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this application, various simple modifications can be made to the technical solution of this application, and these simple modifications all fall within the protection scope of this application.

[0100] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, this application will not describe the various possible combinations separately.

[0101] Furthermore, various different implementations of this application can be combined in any way, as long as they do not violate the spirit of this application, they should also be regarded as the content disclosed in this application.

Claims

1. A fault diagnosis method for an aircraft, characterized in that, The fault diagnosis method includes: Acquiring multiple vibration signals, wherein the multiple vibration signals originate from multiple vibration signal acquisition components mounted on a fixed platform, the aircraft being fixed on the fixed platform, and the multiple vibration signal acquisition components acquiring the multiple vibration signals when the aircraft's rotor system is activated; and Based on the acquired multi-channel vibration signals, fault diagnosis is performed on the rotor system of the aircraft.

2. The fault diagnosis method according to claim 1, characterized in that, Based on the acquired multi-channel vibration signals, fault diagnosis is performed on the aircraft, including: The acquired multi-channel vibration signals are recombined to construct a two-dimensional vibration signal matrix; Based on the constructed two-dimensional vibration signal matrix and the trained neural network model, the probability of each fault type is determined; and Based on the determined probability of each fault type, the aircraft is subjected to fault diagnosis.

3. The fault diagnosis method according to claim 2, characterized in that, The acquired multi-channel vibration signals are reassembled to construct a two-dimensional vibration signal matrix, including: Each vibration signal is divided into multiple segments of equal length; and The two-dimensional vibration signal matrix is ​​constructed by splicing together multiple sub-segments corresponding to the multiple vibration signals.

4. A fault diagnosis device for an aircraft, characterized in that, The fault diagnosis device includes: An acquisition module is used to acquire multiple vibration signals, wherein the multiple vibration signals come from multiple vibration signal acquisition components, which are mounted on a fixed platform. The aircraft is fixed on the fixed platform, and the multiple vibration signal acquisition components acquire the multiple vibration signals when the aircraft's rotor system is activated; and The diagnostic module is used to perform fault diagnosis on the rotor system of the aircraft based on the acquired multi-channel vibration signals.

5. A fault diagnosis system for an aircraft, characterized in that, The fault diagnosis system performs fault diagnosis based on the fault diagnosis method according to any one of claims 1-3. The fault diagnosis system includes: A fixed platform is used to support the aircraft; A fixing component for securing the aircraft when it is located on the fixing platform; Multiple vibration signal acquisition components are mounted on the fixed platform.

6. The fault diagnosis system according to claim 5, characterized in that, The fault diagnosis system also includes: A computing unit is used to perform fault diagnosis based on the fault diagnosis method according to any one of claims 1-3.

7. The fault diagnosis system according to claim 6, characterized in that, The computing unit is mounted on the fixed platform, and the computing unit and the aircraft are located on opposite sides of the fixed platform.

8. The fault diagnosis system according to claim 5, characterized in that, The plurality of vibration signal acquisition components includes four vibration signal acquisition components; Preferably, the fixed platform has a rectangular cross-section, and the four vibration signal acquisition components are distributed at the four corners of the fixed platform.

9. The fault diagnosis system according to claim 5, characterized in that, The fixing component includes: The main body has a first end that is a limiting hook and a second end that is threaded. The second end of the main body is secured by the nut after passing through the fixing platform.

10. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions for causing the machine to perform the fault diagnosis method according to any one of claims 1-3.

Images