Double-rotor bolt connection online monitoring method based on multi-rotating-speed FFT

Abstract

The invention belongs to the technical field of aero-engine structure health monitoring and fault diagnosis, and particularly relates to a double-rotor bolt connection online monitoring method based on multi-rotating-speed FFT, which comprises the steps of key rotating speed determination, data acquisition and training set construction, FFT enhanced neural network model construction and training, and state recognition and abnormity positioning. By monitoring the vibration signal of the intermediate fulcrum double-rotor system, real-time judgment and abnormal positioning of the bolt connection state of the rotor system can be realized without disassembling the engine.

Description

Technical Field

[0001] The present invention belongs to the technical field of aircraft engine structural health monitoring and fault diagnosis, and in particular relates to an online monitoring method for dual-rotor bolt connections based on multi-speed FFT. Background Art

[0002] Bolted connections are widely used in the dual-rotor systems of aircraft engines, primarily for connecting multiple components and transferring loads. As the number of cycles increases, the bolted connection structure will inevitably be damaged by multiple loads such as unbalanced excitation, thermal loads, and centrifugal loads, leading to changes in its mechanical properties. These changes can significantly affect the dynamic characteristics of the rotor system, altering the rotor's bending stiffness and internal damping characteristics, leading to increased or abnormal vibrations in the rotor system at high speeds, which is detrimental to the safe operation of the aircraft engine. Therefore, timely and accurate diagnosis of the status of the rotor connection structure has become particularly important, especially in the context of digital transformation and the growing demand for intelligent maintenance. The demand for intelligent diagnosis of bolted connection structures is becoming increasingly urgent.

[0003] Currently, there are various testing methods for bolt loosening or preload changes. Ultrasonic testing, a mature and commonly used method, calculates the bolt preload by measuring the propagation time and echo signal amplitude of ultrasonic waves. Additionally, the axial force of the bolt can be identified by the variation in modal frequency. A rotor bolt preload identification method based on multi-order modal frequency variations is also available. Based on fractal contact theory and using PZT (piezoelectric material) excitation, combined with mechanical impedance analysis, the looseness of the bolt connection can be determined. These methods typically provide a detailed and accurate assessment of the bolt connection status based on the local frequency response of the bolt.

[0004] However, these methods are limited in that they typically require disassembly of the engine for testing, and most focus on local detection, making them difficult to apply directly to monitoring an engine while it is running. Therefore, there is a need for an intelligent monitoring method that can determine the state of bolted connections and locate anomalies by monitoring the vibration signals of the support points without disassembling the engine. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides an online monitoring method for dual-rotor bolt connections based on multi-speed FFT, which can determine the bolt connection status and locate abnormalities by monitoring the support vibration signal without disassembling the engine.

[0006] The present invention is implemented by providing a method for online monitoring of dual-rotor bolt connections based on multi-speed FFT, comprising the following steps: Step S0, critical speed determination: define the two rotors in the intermediate bearing dual-rotor system as the first rotor and the second rotor, the speed frequencies of the first rotor and the second rotor are f1 and f2 respectively, determine the combined resonance state point caused by the asymmetric rotor stiffness, and symmetrically select multiple speed points around the combined resonance state point as the center, and use them together with the combined resonance state point as monitoring speed points; Step S1, data collection and training set construction: Based on the monitoring speed points determined in step S0, a signal segment of each monitoring speed point is extracted from the actual test data. The length of each signal segment is L. The extracted signal segment is combined with the corresponding bolt connection status label (such as normal, loose, loose location, etc.) to construct a labeled sample set, which is further divided into a training set and a test set. Step S2, FFT-enhanced neural network model construction and training: Construct an FFT-enhanced neural network model, extract frequency domain information of each monitoring speed point at different speeds through fast Fourier transform (FFT), and input spectrum data in the frequency band of 0 to 2 or 3 times the maximum speed into the neural network model for deep feature extraction and classification. Use the training set constructed in step S1 to train the neural network model, optimize model parameters, determine the optimal neural network model, and test the neural network model using the test set. Step S3, state identification and abnormality location: The vibration data to be diagnosed is reconstructed according to the monitoring speed and monitoring speed points to ensure that the format is consistent with the training set samples. The trained neural network model is then input and the classification results output by the neural network model are used to determine the state of the bolt connection in the intermediate bearing dual-rotor system.

[0007] Preferably, in step S0, if there is a modal frequency fn in the intermediate bearing dual-rotor system, for the case where the bolt connection of the first rotor is loose, the speed state satisfying 2f1-f2=fn or 2f1+f2=fn is selected as the combined resonance state point; for the case where the bolt connection of the second rotor is loose, the speed state satisfying 2f2-f1=fn or 2f2+f1=fn is selected as the combined resonance state point; M monitoring speed points are symmetrically selected around the combined resonance state point to ensure that the dynamic characteristics of the intermediate bearing dual-rotor system under different working conditions are covered.

[0008] Further preferably, in step S1, the specific operation is: according to the M monitoring speed points selected in step S0, for the P vibration measurement points at each speed, respectively intercept a signal of length L, a total of N samples, thereby forming a four-dimensional tensor N×M×L×P, and the bolt connection status label corresponding to each sample is a unique hot encoding, corresponding to the number of classifications C, so the bolt connection status label is a two-dimensional tensor N×C.

[0009] Further preferably, in step S2, the specific operation is: the constructed FFT-enhanced neural network model processes each vibration signal in the sample through fast Fourier transform, and intercepts the spectrum data of length Q in each vibration signal, splices the spectrum data of each speed and each monitoring speed point together, and forms a vector of length M×P×Q for each sample. The input layer dimension of the neural network model is M×P×Q, and feature extraction and classification are performed through a multi-layer network structure. The output layer dimension is the corresponding classification number C of the unique hot encoding, and the neural network model is trained through the training set to optimize the model parameters.

[0010] Further preferably, in step S3, the specific operation is: for the intermediate bearing dual-rotor system to be diagnosed, vibration data is collected and reconstructed according to the format of the training set to form a data tensor with a shape of 1×M×L×P, the reconstructed vibration data is input into the trained neural network model, and the status of the bolt connection in the intermediate bearing dual-rotor system is judged by the output result of the neural network model.

[0011] Compared with the prior art, the advantages of the present invention are: 1) Real-time monitoring without engine disassembly: By monitoring the vibration signal of the intermediate bearing dual-rotor system, the present invention can determine the bolt connection status in real time while the engine is running, without disassembling the engine, significantly improving detection efficiency and maintenance convenience.

[0012] 2) High-precision identification and positioning: The neural network model based on multi-speed FFT enhancement can extract deep features from frequency domain information, accurately identify the bolt connection status (such as normal, loose, loose position, etc.), and accurately locate the location of the abnormality, providing a reliable basis for subsequent maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 Schematic diagram of a dual-rotor system with an intermediate bearing according to an embodiment of the present invention; Figure 2 Schematic diagram of an FFT-enhanced neural network model according to an embodiment of the present invention; Figure 3 Schematic diagram of the confusion matrix of bolt connection status recognition results according to an embodiment of the present invention. DETAILED DESCRIPTION

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0015] like Figure 1 As shown, the intermediate bearing dual rotor system of the present invention consists of a first rotor, namely the low-pressure rotor 1, and a second rotor, namely the high-pressure rotor 2. The high-pressure rotor 2 is supported on the low-pressure rotor 1 through an intermediate bearing 3. The rotational frequency of the low-pressure rotor 1 is denoted as f1, and the rotational frequency of the high-pressure rotor 2 is denoted as f2. In the complete machine state, there are three vibration measuring points, namely the air intake casing measuring point 4, the intermediate casing measuring point 5, and the turbine casing measuring point 6. The high-pressure rotor 2 is provided with five rows of bolt connections, which are denoted as J1, J2, J3, J4, and J5 from front to back. This embodiment mainly identifies the bolt connection status of the high-pressure rotor 2.

[0016] Specifically, this embodiment provides a method for online monitoring of dual-rotor bolt connections based on multi-speed FFT, and the specific steps are as follows: Step S0, critical speed determination: Through dual-rotor modal analysis, the high-pressure rotor pitch modal frequency fn = 100 Hz is obtained. It is determined that when f1 = 72 Hz and f2 = 86 Hz, 2f2 - f1 = fn. Therefore, the combined resonance state point is determined to be (72 Hz, 86 Hz). With this point as the center, M = 5 speed points are selected: (68 Hz, 82 Hz), (70 Hz, 84 Hz), (72 Hz, 86 Hz), (74 Hz, 88 Hz), and (76 Hz, 90 Hz).

[0017] Step S1, Data Collection and Training Set Construction: Based on the key speeds determined in Step S0, signal segments at each vibration measurement point at each key speed are extracted from the real test run data. The data sampling frequency is 4000 Hz, and the length of each signal segment is L = 2000. Bolt connection states are categorized into six categories: loose bolts in the first row (J1), loose bolts in the second row (J2), loose bolts in the third row (J3), loose bolts in the fourth row (J4), loose bolts in the fifth row (J5), and no loose bolts (J0). Labels are represented using a 1×6 one-hot encoding. A total of 150 samples are collected, resulting in a training set with a data shape of 150×5×2000×3, and labels with a data shape of 150×6. The extracted signal segments are combined with the corresponding bolt connection state labels to construct a labeled training set. Twenty-five unlabeled samples are also collected as a test set.

[0018] Step S2, FFT enhanced neural network model construction and training: Construct an FFT enhanced neural network model, such as Figure 2As shown in the figure, a fast Fourier transform (FFT) was used to extract frequency domain information at each measurement point at different speeds, and key spectrum data with a length of Q = 70 was intercepted from 60 to 200 Hz. The key frequency band data for each speed and measurement point were concatenated to form a vector of length M × P × Q = 1050 for each sample. The neural network consists of four layers, with dimensions of 1050 -> 64, 64 -> 32, 32 -> 32, and 32 -> 6, respectively. The model was trained using the training set, and the network parameters were optimized to achieve sufficiently high accuracy on the training set. The model was then tested on the test set.

[0019] Step S3, state identification and abnormality location: For the dual-rotor system to be diagnosed, vibration data is collected and reconstructed according to the training set format, with a total of 25 samples, forming a data tensor with a shape of 25×5×2000×3. This data is input into the trained neural network model, and the state of the bolt connection is obtained from the model output and compared with the preset state. The accuracy of the recognition result is given by the confusion matrix, as shown in Figure 3 As shown, it can be seen that the present invention has a 100% recognition accuracy for each category, and can accurately identify the bolt connection status through vibration signals in the state of the whole machine.

[0020] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dual-rotor bolt connection online monitoring method based on multi-speed FFT, characterized in that: The steps include: Step S0, critical speed determination: define the two rotors in the intermediate bearing dual-rotor system as the first rotor and the second rotor, the speed frequencies of the first rotor and the second rotor are f1 and f2 respectively, determine the combined resonance state point caused by the asymmetric rotor stiffness, and symmetrically select multiple speed points around the combined resonance state point as the center, and use them together with the combined resonance state point as monitoring speed points; Step S1, data collection and training set construction: Based on the monitoring speed points determined in step S0, extract signal segments for each monitoring speed point from the actual test data. Each signal segment has a length of L. The extracted signal segments are combined with the corresponding bolt connection status labels to construct a labeled sample set, which is further divided into a training set and a test set. Step S2, FFT-enhanced neural network model construction and training: Construct an FFT-enhanced neural network model, extract frequency domain information of each monitoring speed point at different speeds through fast Fourier transform, and input spectrum data in the frequency band of 0 to 2 or 3 times the maximum speed into the neural network model for deep feature extraction and classification. Use the training set constructed in step S1 to train the neural network model, optimize model parameters, determine the optimal neural network model, and test the neural network model using the test set; Step S3, state identification and abnormality location: The vibration data to be diagnosed is reconstructed according to the monitoring speed and monitoring speed points to ensure that the format is consistent with the training set samples. The trained neural network model is then input and the classification results output by the neural network model are used to determine the state of the bolt connection in the intermediate bearing dual-rotor system.

2. The method for online monitoring of dual-rotor bolt connection based on multi-speed FFT according to claim 1, characterized in that: In step S0, if there is a modal frequency fn in the intermediate bearing dual-rotor system, for the case where the bolt connection of the first rotor is loose, the speed state satisfying 2f1-f2=fn or 2f1+f2=fn is selected as the combined resonance state point; for the case where the bolt connection of the second rotor is loose, the speed state satisfying 2f2-f1=fn or 2f2+f1=fn is selected as the combined resonance state point; M monitoring speed points are symmetrically selected around the combined resonance state point to ensure that the dynamic characteristics of the intermediate bearing dual-rotor system under different working conditions are covered.

3. The method for online monitoring of dual-rotor bolt connection based on multi-speed FFT according to claim 2, characterized in that: In step S1, the specific operation is: according to the M monitoring speed points selected in step S0, for each P vibration measurement point at each speed, a signal of length L is intercepted, with a total of N samples, thereby forming a four-dimensional tensor N×M×L×P. The bolt connection status label corresponding to each sample is a one-hot encoding corresponding to the number of categories C, so the bolt connection status label is a two-dimensional tensor N×C.

4. The method for online monitoring of dual-rotor bolt connection based on multi-speed FFT according to claim 3, characterized in that: In step S2, the specific operation is as follows: the constructed FFT-enhanced neural network model processes each vibration signal in the sample through fast Fourier transform, and intercepts the spectrum data of length Q in each vibration signal, splices the spectrum data of each speed and each monitoring speed point together, and forms a vector of length M×P×Q for each sample. The input layer dimension of the neural network model is M×P×Q, and feature extraction and classification are performed through a multi-layer network structure. The output layer dimension is the corresponding classification number C of the unique hot encoding. The neural network model is trained through the training set to optimize the model parameters.

5. The method for online monitoring of dual-rotor bolt connection based on multi-speed FFT according to claim 3, characterized in that: In step S3, the specific operation is: for the intermediate bearing dual-rotor system to be diagnosed, vibration data is collected and reconstructed according to the format of the training set to form a data tensor with a shape of 1×M×L×P, the reconstructed vibration data is input into the trained neural network model, and the status of the bolt connection in the intermediate bearing dual-rotor system is judged according to the output result of the neural network model.

Images