Gas circuit fault diagnosis method and system in aero-engine starting process

Abstract

The invention discloses a gas circuit fault diagnosis method and system in the aero-engine starting process, and relates to the technical field of aero-engine fault diagnosis, and the method comprises the following steps: obtaining an acoustic signal, a vibration signal, temperature data and structural deformation data of an engine; inputting the temperature data into a pre-constructed thermal-mechanical coupling model, and outputting thermal deformation theoretical data; and performing difference calculation on the thermal deformation theoretical data and the structural deformation data to obtain a structural anomaly difference. According to the scheme, through a time-frequency domain joint compensation method, comprehensive compensation of a sound wave propagation path, material absorption characteristics and elastic modulus changes is achieved, the accuracy of the acoustic signal representation structure state is remarkably improved, the interference of engine structure deformation on the acoustic signal propagation characteristics is effectively eliminated, and the reliability of the engine is improved. The acoustic features can accurately reflect the real state of the gas path system, and fault feature misjudgment or missed judgment caused by insufficient signal compensation is avoided.

Description

Technical Field

[0001] This invention relates to the field of aero-engine fault diagnosis technology, specifically to a method and system for diagnosing air circuit faults during the aero-engine start-up process. Background Technology

[0002] As the core power plant of aircraft, the operational safety and reliability of aero engines have always been a key focus in the aviation field. With the continuous increase in the complexity of aero engine structures and the continuous improvement of performance indicators, the health status of the engine's air path has a significant impact on the engine's start-up safety, working efficiency, and even the successful completion of the entire flight mission. Because the engine's air path system is subjected to complex coupling effects of multiple physical fields such as heat, force, and vibration at different stages, it generates structural deformation data, which has a significant impact on the engine's performance and safety during operation.

[0003] However, in the process of implementing the technical solution of the invention in the embodiments of this application, it was found that the above-mentioned technology has at least the following technical problems:

[0004] Existing technologies mainly utilize acoustic and vibration signals from different stages of engine startup, combined with signal processing techniques, to locate faults and improve fault detection capabilities. However, they do not consider thermo-mechanical coupling of temperature data and structural deformation data, making it difficult to obtain a real-time structural model of the engine. Furthermore, they do not correct the propagation path and signal characteristics of acoustic signals based on the real-time structural model to eliminate the influence of structural deformation on acoustic signals. In addition, they do not apply intermediate data such as structural anomaly differences generated during thermo-mechanical coupling to the fusion process of acoustic and vibration signals, affecting the accuracy of fault diagnosis and resulting in a lack of specificity in the generated diagnostic reports. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for diagnosing air circuit faults during the start-up process of an aero-engine, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] In a first aspect, the present invention discloses a method for diagnosing air circuit faults during the start-up process of an aero-engine, comprising the following steps:

[0008] Acquire acoustic signals, vibration signals, temperature data, and structural deformation data of the engine;

[0009] The temperature data is input into a pre-constructed thermo-mechanical coupling model, and theoretical data of thermal deformation are output.

[0010] The difference between the theoretical thermal deformation data and the structural deformation data is calculated to obtain the structural anomaly difference.

[0011] The engine's three-dimensional model is invoked, and a real-time structural model is generated by combining the structural deformation data and the thermal deformation theoretical data.

[0012] The actual propagation path of the acoustic signal is established based on the real-time structural model, and the set of correction parameters for the acoustic signal is calculated based on the actual propagation path.

[0013] The acoustic signal is multidimensionally compensated according to the set of correction parameters to obtain a structure-corrected acoustic signal.

[0014] Using the structural anomaly difference as a weighting factor, the structurally corrected acoustic signal and vibration signal are weighted and fused to obtain a multimodal fusion feature vector;

[0015] Anomaly detection is performed on the multimodal fusion feature vector, and a diagnostic report is output based on the anomaly detection results.

[0016] Secondly, this invention discloses a gas path fault diagnosis system for the start-up process of an aero-engine, comprising:

[0017] The data acquisition module is used to acquire acoustic signals, vibration signals, temperature data, and structural deformation data.

[0018] The structural anomaly difference calculation module is used to input the temperature data into a pre-built thermo-mechanical coupling model and output theoretical data of thermal deformation.

[0019] The difference between the theoretical thermal deformation data and the structural deformation data is calculated to obtain the structural anomaly difference.

[0020] The real-time structural model generation module is used to call the engine's three-dimensional model and generate a real-time structural model by combining the structural deformation data and the thermal deformation theoretical data.

[0021] An acoustic signal correction module is used to establish the actual propagation path of the acoustic signal based on the real-time structural model, and to calculate the correction parameter set of the acoustic signal based on the actual propagation path.

[0022] The acoustic signal is multidimensionally compensated according to the set of correction parameters to obtain a structure-corrected acoustic signal.

[0023] The multimodal fusion feature vector calculation module is used to perform weighted fusion of the structural modified acoustic signal and vibration signal with the structural anomaly difference as a weighting factor to obtain a multimodal fusion feature vector;

[0024] The diagnostic report output module is used to perform anomaly detection on the multimodal fusion feature vector and output a diagnostic report based on the anomaly detection results.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] 1. This solution achieves comprehensive compensation for sound wave propagation path, material absorption characteristics, and elastic modulus changes through a time-frequency domain joint compensation method. This significantly improves the accuracy of acoustic signal characterization of structural state, effectively eliminates the interference of engine structural deformation on acoustic signal propagation characteristics, and enables acoustic features to accurately reflect the true state of the air circuit system, avoiding misjudgment or omission of fault features due to insufficient signal compensation.

[0027] 2. This scheme introduces structural anomaly difference as a dynamic weighting factor, which can automatically identify abnormal regions and enhance the contribution of their corresponding features in the fusion process. It solves the feature fusion deviation problem caused by ignoring structural state differences in traditional methods, strengthens the feature representation capability of abnormal regions, and enables the fused feature vector to more accurately reflect the actual fault state of the engine air circuit, thereby improving the diagnostic accuracy of subsequent anomaly detection links.

[0028] 3. This solution integrates multimodal vectors of structural deformation, acoustic compensation, and vibration characteristics, combined with threshold judgment and correlation tracing, to achieve simultaneous processing of fault location and cause analysis. This solves the problem of lack of specificity in diagnostic reports in existing technologies and can accurately identify the fault location and its physical cause. Attached Figure Description

[0029] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0030] Figure 1 This is a flowchart of the steps of a method for diagnosing air circuit faults during the start-up process of an aero-engine according to the present invention.

[0031] Figure 2 A schematic diagram illustrating the process of correcting acoustic signals using the output structure provided by this invention;

[0032] Figure 3 This is a flowchart illustrating the output resonance feature data provided by the present invention.

[0033] Figure 4 A schematic diagram of the output diagnostic report provided by the present invention;

[0034] Figure 5 This invention provides a schematic diagram of the module functions of an air circuit fault diagnosis system for the start-up process of an aero-engine. Detailed Implementation

[0035] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0036] Application Overview:

[0037] In existing technologies, fault diagnosis of aero-engine air circuits mainly relies on acoustic and vibration signal analysis, but does not fully consider the coupling effect of temperature and structural deformation. During engine startup, thermal stress and mechanical loads work together to cause structural deformation. Existing methods lack real-time structural models and cannot accurately correct the acoustic signal propagation path, resulting in signal characteristic distortion. For example, during the startup phase of a certain type of engine, thermal deformation was not modeled, and the acoustic signal propagation delay error was not compensated, leading to misjudgment of the resonance frequency, which in turn affects the accuracy of fault location.

[0038] To address the aforementioned issues, the study found that the dynamic correlation between structural deformation and temperature distribution is a key factor affecting acoustic signals. By analyzing the impact of thermo-mechanical coupling on material properties, it proposes inputting temperature data into a thermo-mechanical coupling model to predict theoretical data of thermal deformation, and comparing this data with structural deformation data to generate structural anomaly differences. Furthermore, it recognizes that structural anomaly differences can reflect the degree of structural anomalies, and uses them as a weighting factor to fuse multi-source signals, thereby improving the characterization ability of fault features.

[0039] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] Example 1:

[0041] Please see Figure 1-Figure 4 A method for diagnosing air circuit faults during the start-up process of an aero-engine, comprising the following steps:

[0042] Acquire acoustic signals, vibration signals, temperature data, and structural deformation data of the engine;

[0043] Input the temperature data into the pre-built thermo-mechanical coupling model, and output the theoretical data of thermal deformation. The specific calculation formula is as follows:

[0044]

[0045] In the formula, Represents theoretical data on thermal deformation. Indicates the original length. Indicates the coefficient of thermal expansion. Represents temperature change data. Indicates force, Indicates the elastic modulus;

[0046] The difference between the theoretical data of thermal deformation and the data of structural deformation is calculated to obtain the structural anomaly difference.

[0047] The engine's 3D model is called up, and a real-time structural model is generated by combining structural deformation data and thermal deformation theoretical data.

[0048] The actual propagation path of the acoustic signal is established based on the real-time structural model, and the set of correction parameters for the acoustic signal is calculated based on the actual propagation path.

[0049] The acoustic signal is multidimensionally compensated based on the set of correction parameters to obtain the structure-corrected acoustic signal.

[0050] Using the structural anomaly difference as a weighting factor, the structurally corrected acoustic signal and vibration signal are weighted and fused to obtain a multimodal fusion feature vector;

[0051] Anomaly detection is performed on the multimodal fusion feature vector, and a diagnostic report is output based on the anomaly detection results.

[0052] Among them, the thermo-mechanical coupling model refers to the computational model of the interaction between the temperature field and the stress field established by the finite element method. Specifically, it can be implemented using ANSYS Mechanical APDL software and is used to predict the theoretical deformation of engine components based on the temperature gradient.

[0053] Structural anomaly difference refers to the deviation between the measured structural deformation and the theoretical prediction value. It can be calculated by point-by-point coordinate difference and is used to characterize the degree of local structural anomaly.

[0054] The real-time structural model refers to a three-dimensional geometric model that integrates measured deformation and theoretical deformation. Specifically, it can be implemented by updating the spatial node coordinates using CATIA software, and is used to reflect the real-time deformation state of the engine structure.

[0055] The correction parameter set includes time delay, amplitude, and spectral compensation parameters, which can be calculated from the propagation path length variation and material parameters to eliminate the influence of structural deformation on acoustic signal propagation.

[0056] The specific implementation process is as follows:

[0057] First, acoustic signals, vibration signals, temperature data, and structural deformation data are collected throughout the engine start-up process.

[0058] After data acquisition, temperature data is input into a pre-built thermo-mechanical coupling model, and theoretical thermal deformation data is output. Then, the difference between the theoretical thermal deformation data and the structural deformation data is calculated to obtain the structural anomaly difference, which reflects the potential abnormal deformation that cannot be explained by thermal load.

[0059] By calling the engine's three-dimensional model and combining it with structural deformation data and thermal deformation theoretical data, the three-dimensional model is reconstructed to obtain a real-time structural model. Based on this, the actual propagation path of acoustic signals under complex air paths is established. The correction parameter set of acoustic signals is calculated using this propagation path. The acoustic signals are then compensated in multiple dimensions according to the correction parameter set to obtain structurally corrected acoustic signals, which significantly improves the sensitivity and accuracy of acoustic diagnosis to minute structural changes.

[0060] Furthermore, the structural anomaly difference is used as a weighting factor to perform weighted fusion of the structural correction acoustic signal and the vibration signal to obtain a multimodal fusion feature vector. This feature not only integrates the acoustic and vibration anomaly information, but also adaptively amplifies the weak signals related to the structural anomaly difference through weighting, thereby enabling early amplification and perception of potential hazards such as micro-leakage.

[0061] Finally, anomaly detection is performed on the multimodal fusion feature vector, and a diagnostic report is output based on the anomaly detection results.

[0062] Through the above technical solutions, the present invention can effectively eliminate the interference of normal environmental and structural deformation, highlight the detection capability of minute anomalies, improve the accuracy of multimodal data fusion, and improve the accuracy of the final diagnostic report.

[0063] This application further proposes the following specific methods for acquiring acoustic signals, vibration signals, temperature data, and structural deformation data:

[0064] The acoustic signals of the engine are collected in real time through a distributed microphone array; the acoustic signals include sound speed, housing and air passage resonance sound signals during the start-up process;

[0065] The vibration signals of the engine are acquired in real time through a multi-axis acceleration sensor array; the vibration signals include multi-axis acceleration vector data, resonant frequency, amplitude, and energy distribution characteristics at each stage of startup;

[0066] Engine temperature data is collected in real time through a sensor network;

[0067] The engine's structural deformation data is collected in real time using displacement sensors; the structural deformation data includes strain, real-time displacement measurements, and the evolution of structural deformation at each stage of startup.

[0068] All the data collected above were given corresponding spatial location information and timestamp tags.

[0069] Among them, the distributed microphone array refers to a data acquisition system consisting of multiple microphones arranged in a spatial distribution. Specifically, it can be implemented using a ring or linear array layout. It is used to capture acoustic signals at different locations on the surface of the engine casing and inside the air passage, solving the problem that a single sensor cannot cover complex sound field distributions.

[0070] A multi-axis accelerometer array refers to a measuring device containing multiple triaxial accelerometers, which can be implemented using piezoelectric or MEMS sensors. It is used to synchronously collect vibration data of different parts of the engine in the X, Y, and Z axes, thus solving the problem that traditional single-axis sensors cannot fully characterize complex vibration modes.

[0071] Displacement sensors are sensing devices that can measure structural deformation. Specifically, they can be implemented using laser displacement gauges or fiber optic grating sensors. They are used to acquire deformation data of key engine components under thermo-mechanical coupling in real time, solving the problem that traditional offline detection cannot reflect the dynamic deformation process.

[0072] Spatial location information refers to the installation location data of the sensor in the engine's three-dimensional coordinate system. Specifically, it can be achieved by using CAD model coordinate mapping to establish the correspondence between the measurement data and the engine's physical structure, thus solving the problem of spatial matching of multi-source data.

[0073] A timestamp tag is a precise time mark that is synchronously recorded when data is collected. It can be implemented using a GPS synchronized clock or the IEEE 1588 protocol to ensure the temporal synchronization of multi-sensor data and solve the problem of data phase deviation caused by asynchronous acquisition.

[0074] Specifically, during the engine start-up phase, a distributed microphone array continuously collects acoustic signals at a sampling rate of no less than 10 kHz, and the spatial coordinates of each microphone are pre-calibrated through 3D modeling; a multi-axis accelerometer array collects vibration signals at a sampling rate of no less than 5 kHz, and the installation position of each sensor corresponds to the 3D model through spatial coordinate mapping; temperature data is collected through a sensor network with a sampling interval of no more than 100 ms; displacement sensors collect structural deformation data at a sampling rate of no less than 1 kHz; all collected data are marked through a unified spatiotemporal reference system, for example, the spatial coordinate accuracy of each data point is controlled within ±0.5 mm, and the time synchronization error does not exceed 10 μs.

[0075] Through the above technical solutions, this application effectively solves the problem of spatiotemporal matching of multi-source heterogeneous data, ensures the spatial correspondence between acoustic signals and structural deformation, and enables subsequent thermal deformation theoretical calculations to accurately reflect actual working conditions. Through high-density sensor layout and high-speed data acquisition, the physical field characteristics of transient changes during startup are fully captured, providing data support for establishing an accurate real-time structural model. The introduction of spatiotemporal tags enables multimodal data to be fused and analyzed in a unified coordinate system, significantly improving the accuracy of fault feature extraction.

[0076] This application further proposes to call upon the engine's three-dimensional model and combine it with structural deformation data and thermal deformation theoretical data to generate a real-time structural model, specifically including:

[0077] Call the engine's 3D model and obtain information about each spatial node within the 3D model;

[0078] The structural deformation data is mapped to the corresponding spatial nodes in the 3D model using spatial coordinates.

[0079] The thermal deformation theoretical data is mapped to the corresponding spatial nodes in the three-dimensional model through spatial coordinates, and coupled with the structural deformation data for calculation to obtain the comprehensive deformation of each spatial node.

[0080] The spatial nodes are corrected based on the comprehensive deformation to obtain the new coordinates of each spatial node;

[0081] Based on the new coordinates of each spatial node, the three-dimensional model of the engine is reconstructed to obtain the real-time structural model.

[0082] Among them, the spatial node information of the three-dimensional model refers to the set of discretely distributed geometric coordinate points in the three-dimensional model, which can be realized by finite element mesh generation technology, and is used to locate the spatial correspondence between structural deformation data and thermal deformation theoretical data.

[0083] Spatial coordinate mapping refers to matching the physical space data collected by the sensor with the coordinate system of the three-dimensional model. Specifically, it can be achieved by using coordinate transformation algorithms to ensure the spatial consistency of the deformation data in the model.

[0084] Among them, coupled calculation refers to superimposing theoretical data of thermal deformation with measured structural deformation data. Specifically, it can be implemented by weighted average method or nonlinear superposition model to eliminate the error interference of a single data source.

[0085] Among them, the comprehensive deformation refers to the total deformation of the spatial node under the action of thermo-mechanical coupling. Specifically, it can be calculated by vector synthesis method and is used to characterize the real-time deformation state of the engine structure.

[0086] The specific implementation process is as follows: call the three-dimensional model of the engine, obtain its set of spatial nodes, and associate the collected structural deformation data and thermal deformation theoretical data with the corresponding spatial nodes of the three-dimensional model through spatial coordinate mapping method. For each node, the structural deformation data and thermal deformation theoretical data are coupled (such as linear superposition or weighted coupling) to calculate the comprehensive deformation of each node. Based on this, the coordinates of all nodes of the three-dimensional model are corrected, and the three-dimensional model of the engine is reconstructed in real time to obtain a real-time structural model that reflects the real structural state during the start-up process.

[0087] Through the above technical solution, this application solves the problem of structural model distortion caused by the failure to consider thermo-mechanical coupling in the prior art, and realizes dynamic updating of the engine three-dimensional model by fusing multi-source deformation data; this provides a reliable structural benchmark for the accurate correction of the subsequent acoustic signal propagation path.

[0088] This application further proposes that after generating the real-time structural model, it also includes generating simulated sound field distribution data by combining temperature data, specifically including:

[0089] Temperature data is mapped to each spatial node of the real-time structural model to obtain temperature distribution data on the real-time structural model.

[0090] Based on temperature distribution data, the material parameters of each spatial node in the real-time structural model are corrected.

[0091] Based on the real-time structural model, combined with temperature distribution data and the corrected material parameters of each spatial node, the boundary conditions for the engine sound field simulation are set.

[0092] The acoustic finite element method was used to perform numerical calculations of the simulated sound field of the engine casing and the internal structure of the air passage. The results of the numerical calculations of the simulated sound field were used as the simulated sound field distribution data of the casing and the internal structure of the air passage at different spatial nodes.

[0093] Among them, the acoustic finite element method refers to discretizing the engine structure into finite elements for numerical calculation of the sound field. For example, it can be implemented using software such as COMSOL or ANSYS to simulate the sound field propagation characteristics inside the casing and air passage.

[0094] The specific implementation process is as follows:

[0095] The collected temperature data is accurately mapped to each spatial node of the real-time structural model through an interpolation algorithm to obtain a fine-grained temperature distribution field. Unlike existing technologies that only use average temperature or a small number of measuring point temperatures to correct the structural model, this invention achieves high-precision fusion of spatially distributed temperature and real-time structural model, which greatly improves the realism of structural modeling.

[0096] Based on temperature distribution data, the material parameters of each spatial node in the real-time structural model are corrected point by point using temperature sensitivity formulas for material physical properties (such as empirical formulas for the change of elastic modulus, density, and sound velocity with temperature). In this way, the influence of temperature fluctuations on the acoustic properties of materials is fully considered in the modeling stage, dynamically reflecting the influence of the actual environment on the acoustic properties of materials, and significantly improving the accuracy of sound field simulation.

[0097] Furthermore, based on the corrected real-time structural model, the temperature distribution data and the corrected material parameters of each spatial node are used together to set the boundary conditions for the acoustic field finite element simulation. The acoustic finite element method is used to perform full-space acoustic field numerical calculations on the engine casing and the internal structure of the air passage, and to obtain the simulated acoustic field distribution data of different spatial nodes, including information such as intrinsic frequency, sound pressure level, and resonance mode.

[0098] Through the above technical solution, this application realizes the dynamic coupling of temperature field and sound field simulation, which can accurately simulate the sound field distribution characteristics of the engine under thermal load, help identify abnormal acoustic signals caused by local overheating or material performance degradation, and improve the detection sensitivity of thermal deformation-related faults.

[0099] This application further proposes establishing the actual propagation path of the acoustic signal based on a real-time structural model, and calculating the correction parameter set of the acoustic signal based on the actual propagation path, specifically including:

[0100] Based on the real-time structural model, the material absorption coefficient, deformation of each part, and elastic modulus of the material of the engine casing and internal structure of the air passage are obtained.

[0101] For each spatial node, based on the three-dimensional model, the shortest feasible propagation path of the acoustic signal in the corresponding structural state of the three-dimensional model is identified, and this path is taken as the original propagation path of the acoustic signal.

[0102] For each spatial node, based on the real-time structural model, the shortest feasible propagation path 2 of the acoustic signal in the corresponding structural state of the real-time structural model is identified and used as the actual propagation path of the acoustic signal.

[0103] Adjust the sound velocity along the actual propagation path of the corresponding spatial node based on the material's elastic modulus;

[0104] The actual propagation path and the original propagation path of the same spatial node are processed to obtain a set of corrected parameters for the acoustic signal, specifically including:

[0105] The difference between the actual propagation path and the original propagation path of the same spatial node is calculated to obtain the path length change value;

[0106] The time delay correction parameter is obtained by dividing the path length change value and the sound speed after the actual propagation path of the corresponding spatial node.

[0107] Nonlinear calculations are performed on the path length variation and the material absorption coefficient corresponding to the spatial nodes to obtain the amplitude correction parameters.

[0108] Nonlinear calculations are performed on the path length variation and the material elastic modulus corresponding to the spatial nodes to obtain the spectrum correction parameters.

[0109] The set of correction parameters includes time delay correction parameters, amplitude correction parameters, and spectrum correction parameters.

[0110] Among them, the material absorption coefficient refers to the energy attenuation coefficient caused by the material properties during the propagation of sound waves. Specifically, it can be realized by material acoustic test experimental data or simulation model calibration value, which is used to quantify the attenuation effect of structural deformation on the sound wave amplitude.

[0111] The elastic modulus of a material refers to the stress-strain ratio during the elastic deformation stage. It can be measured by a dynamic mechanical analyzer or calculated using a thermodynamic model based on temperature distribution. It is used to reflect the degree of influence of structural deformation on sound velocity.

[0112] The path length change value refers to the difference in spatial geometric length between the actual propagation path and the original propagation path. Specifically, it can be calculated using three-dimensional coordinate difference or path topology comparison algorithm, and is used to characterize the offset of the sound wave propagation path caused by structural deformation.

[0113] Nonlinear computation refers to the nonlinear mapping of input variables through exponential functions, polynomial functions, or neural network models. Specifically, it can be implemented using pre-calibrated empirical formulas or data-driven models to establish a nonlinear relationship between path length changes and correction parameters.

[0114] Specifically, during engine startup, a real-time structural model is generated by fusing thermal deformation theory data and structural deformation data. The geometry and material properties of its spatial nodes are dynamically adjusted with temperature changes. The calculation of the actual propagation path of the acoustic signal is based on the three-dimensional topology of the real-time structural model, and the shortest feasible path is identified through a path search algorithm. The change in path length reflects the geometric difference in the sound wave propagation path caused by structural deformation. Combined with the dynamically adjusted sound velocity parameter based on the material's elastic modulus, the time delay correction parameter can be accurately calculated. The nonlinear relationship between the material absorption coefficient and the change in path length is used to calculate the amplitude correction parameter to compensate for the sound wave energy attenuation caused by changes in material properties. The nonlinear relationship between the material's elastic modulus and the change in path length is used to calculate the spectral correction parameter to correct the sound wave frequency shift caused by changes in structural stiffness. Thus, the correction parameter set can dynamically compensate for the acoustic signal in three dimensions: time domain, frequency domain, and energy domain.

[0115] Through the above technical solution, this application can eliminate the influence of structural deformation on the acoustic signal propagation path during engine start-up, accurately compensate for signal delay, amplitude and spectral distortion caused by changes in material properties, provide accurate acoustic signal data for subsequent multimodal feature fusion, and thus improve the accuracy of air path fault diagnosis.

[0116] This application further proposes to perform multidimensional compensation on the acoustic signal based on a set of correction parameters to obtain a structure-corrected acoustic signal, specifically including:

[0117] Based on the time delay correction parameters, the acoustic signal is interpolated and corrected on the time axis;

[0118] The spectrum of the interpolated acoustic signal is obtained by fast Fourier transform.

[0119] Based on the spectral correction parameters, the spectrum is shifted along the frequency axis as a whole;

[0120] Based on the amplitude correction parameter, the amplitude of each frequency point in the shifted spectrum is compensated by multiplication.

[0121] The multidimensional compensated acoustic signal is obtained by inverse fast Fourier transform of the spectrum after multiplication compensation, and is used as the structure-corrected acoustic signal.

[0122] Among them, the time delay correction parameter refers to the parameter used to compensate for the time delay difference caused by the change in the propagation path length of the acoustic signal. Specifically, it can be achieved by using linear interpolation or spline interpolation algorithms to dynamically resample the acoustic signal on the time axis. Its function is to eliminate the sound wave propagation time offset caused by structural deformation.

[0123] Spectrum correction parameters are parameters used to correct the shift in acoustic frequency characteristics caused by changes in the elastic modulus of materials. Specifically, they can be achieved by shifting the entire spectrum to a fixed frequency band using a frequency domain translation algorithm. Their function is to compensate for the influence of structural deformation on the acoustic frequency propagation characteristics.

[0124] Amplitude correction parameters are parameters used to compensate for differences in sound wave energy attenuation caused by changes in the material's absorption coefficient. Specifically, they can be achieved by adjusting the gain of each frequency component using a frequency domain amplitude multiplier. Their function is to restore the abnormal attenuation of sound wave energy caused by structural deformation.

[0125] Specifically, the acoustic signal is first interpolated and corrected according to the time delay correction parameter to eliminate the sound wave propagation time difference caused by structural deformation; then, the spectrum of the acoustic signal is obtained by fast Fourier transform, and the spectrum is shifted along the frequency axis according to the spectrum correction parameter to achieve frequency adaptation under the dynamic characteristics of the structure; the shifted spectrum is then multiplied and compensated for the amplitude of each frequency point according to the amplitude correction parameter to restore the acoustic energy; finally, the compensated spectrum is restored to the time domain signal by inverse fast Fourier transform to obtain the multidimensional compensated structure-corrected acoustic signal.

[0126] Through the above technical solution, this application effectively eliminates the interference of engine structural deformation on acoustic signal propagation characteristics, enabling acoustic features to accurately reflect the real state of the air circuit system; this multi-dimensional compensation method solves the feature distortion problem caused by the traditional single compensation mode, provides a high-precision acoustic data foundation for subsequent multi-modal feature fusion, and avoids misjudgment or omission of fault features due to insufficient signal compensation.

[0127] This application further proposes that after obtaining the structure-corrected acoustic signal, it also includes performing joint spectral analysis on the simulated sound field distribution data and the structure-corrected acoustic signal at the same spatial node to output resonance characteristic data, specifically including:

[0128] Acquire the structurally corrected acoustic signal and simulated sound field distribution data of the same spatial node;

[0129] The structurally corrected acoustic signal and the simulated sound field distribution data were denoised and normalized respectively.

[0130] Time-frequency analysis was used to extract the spectral characteristics of the processed structurally corrected acoustic signal and the simulated sound field distribution data, respectively.

[0131] The spectral characteristics of the processed structurally corrected acoustic signal and the simulated sound field distribution data are matched one-to-one with the spatial node coordinates and the same frequency and spatial node coordinates as indices.

[0132] The spectral characteristics of the processed structure-corrected acoustic signal and the simulated sound field distribution data are subjected to joint spectral analysis in the frequency domain, and the joint spectral analysis results of the two at each frequency point are output.

[0133] By integrating the results of the joint spectrum analysis, a spatial-frequency distribution matrix is ​​obtained;

[0134] Feature extraction is performed on the spatial-frequency distribution matrix to obtain resonance feature data.

[0135] Among them, joint spectrum analysis refers to the correlation operation between the actual measured spectrum of the structure-corrected acoustic signal and the theoretically predicted spectrum of the simulated sound field distribution data in the frequency domain. Specifically, it can be achieved by coherence function calculation or cross-spectral density analysis to discover the resonant frequency deviation between the actual measurement and the theoretical prediction.

[0136] The spatial-frequency distribution matrix refers to a matrix data structure that constructs the frequency response features corresponding to different spatial nodes according to three-dimensional coordinates and frequency values. Specifically, it can be implemented using tensor storage to integrate multi-dimensional acoustic feature information.

[0137] Resonance feature data refers to the abnormal resonance frequencies, amplitude abrupt change points and their spatial distribution information screened from the space-frequency distribution matrix by feature extraction algorithms. Specifically, it can be achieved by principal component analysis or singular value decomposition methods, and is used to characterize the abnormal vibration modes of the engine casing and air passage structure.

[0138] The specific implementation process is as follows:

[0139] The structurally corrected acoustic signal and the simulated sound field distribution data are denoised and normalized respectively to eliminate the influence of environmental noise and signal amplitude differences, and to ensure the accuracy of subsequent spectral feature extraction.

[0140] The short-time Fourier transform method is used to extract spectral features from the processed structure-corrected acoustic signal and the simulated sound field distribution data respectively. To ensure the correspondence between spatial nodes, the two sets of spectral features are paired one by one using the spatial node coordinates and the same frequency as the index, which improves the accuracy of fault micro-area localization.

[0141] In the joint spectrum analysis stage, the spectral characteristics of the structure-corrected acoustic signal and the simulated sound field distribution data are jointly analyzed in the frequency domain. The correlation coefficient method is used to compare the response of the same spatial node at different frequencies and output the joint spectrum analysis results of each frequency point. Then, the joint analysis results of all spatial nodes and frequencies are integrated to construct the space-frequency distribution matrix.

[0142] Finally, principal component analysis was used to process the spatial-frequency distribution matrix and extract resonance characteristic data that reflects spatial distribution characteristics and resonance properties.

[0143] Through the above technical solution, this application can effectively eliminate the interference of structural deformation on acoustic signal analysis, accurately identify abnormal resonance phenomena caused by real faults, and achieve precise location of faults through spatial-frequency distribution characteristics, significantly improving the detection sensitivity and location accuracy of micro-leakage and local structural anomalies, and reducing the false alarm rate.

[0144] This application further proposes a method for fusing acoustic and vibration features using structural anomaly differences as weighting factors to obtain a multimodal fusion feature vector, specifically including:

[0145] Based on the structural anomaly difference, a weighting factor is assigned to each spatial node, and the specific calculation formula is as follows:

[0146]

[0147] In the formula, Indicates the weighting factor. Indicates the structural anomaly difference. This represents the maximum value among the structural anomaly differences. This represents the minimum value among the structural anomaly differences;

[0148] The acoustic, vibration, and resonance characteristic data were normalized separately.

[0149] The normalized acoustic, vibration, and resonance feature data are weighted and fused according to weighting factors to obtain a multimodal fusion feature vector. The specific calculation formula is as follows:

[0150]

[0151] In the formula, This represents a multimodal fusion feature vector. This represents the acoustic characteristics after normalization. This represents the base weights of the acoustic features after normalization. This represents the vibration characteristics after normalization. The basic weights representing the vibration characteristics after normalization. This represents the resonance feature data after normalization. This represents the base weights of the resonance feature data after normalization.

[0152] The weighting factor refers to the feature fusion coefficient that is dynamically adjusted based on the structural anomaly difference. Specifically, it can be implemented using a linear weighting or a non-linear weighting function. This factor can enhance the feature contribution of the anomaly region.

[0153] Weighted fusion refers to the process of superimposing multiple features according to their weights to generate a new feature vector. Specifically, it can be achieved by matrix weighted summation or vector concatenation. By fusing multimodal data, the representational ability of features is improved.

[0154] Specifically, firstly, the corresponding weighting factor is calculated based on the structural anomaly difference of each spatial node. Then, the acoustic, vibration, and resonance feature data are normalized to eliminate the dimensional differences between different sensor data. Finally, the normalized feature data are combined with the corresponding weighting factor to generate a multimodal fusion feature vector through weighted fusion.

[0155] Through the above technical solution, this application achieves effective fusion of multimodal features, enhances the feature representation capability of abnormal regions, and enables the fused feature vector to more accurately reflect the actual fault state of the engine air circuit, thereby improving the diagnostic accuracy of subsequent abnormal detection links.

[0156] This application further proposes anomaly detection for multimodal fused feature vectors, and outputs a diagnostic report based on the anomaly detection results, specifically including:

[0157] Call the standard feature vector and calculate the Euclidean distance between the multimodal fusion feature vector and the standard feature vector;

[0158] Determine if the Euclidean distance is greater than a preset threshold; if so, determine if the multimodal fusion feature vector is an abnormal feature vector.

[0159] Extract spatial node information of the multimodal fusion feature vector corresponding to the abnormal feature vector;

[0160] Based on the spatial node information, relevant data are retrieved, including: structural anomaly difference, acoustic characteristics, vibration characteristics, and temperature data;

[0161] For this spatial node, calculate the correlation coefficient between structural anomaly difference, acoustic characteristics, vibration characteristics and temperature data;

[0162] Based on the calculation results of the correlation coefficient, output the diagnostic conclusion for this spatial node;

[0163] Based on the diagnostic node and spatial node information, a diagnostic report is output.

[0164] The standard feature vector refers to the benchmark dataset of multimodal fusion feature vectors under normal engine conditions. It can be generated through cluster analysis of historical operating data or simulation models and is used to compare with real-time features.

[0165] Euclidean distance refers to the geometric distance between two vectors in a multidimensional space. It can be calculated by taking the square root of the sum of the squares of the differences in each dimension of the vectors and is used to quantify the degree of deviation between real-time features and the standard state.

[0166] The preset threshold refers to the critical value for judging abnormal states, which can be dynamically adjusted by statistical analysis methods combined with engine models, such as setting it based on three times the standard deviation of historical fault data.

[0167] The correlation coefficient refers to the degree of association between different physical quantities. It can be calculated using the Pearson correlation coefficient or the grey relational analysis method and is used to identify the coupling relationship of the root cause of anomalies.

[0168] Specifically, the standard feature vector trained based on historical working conditions big data under healthy conditions is called, and the Euclidean distance between the current multimodal fusion feature vector of each spatial node and its corresponding standard feature vector is calculated; if the Euclidean distance is greater than a preset threshold (such as 2.0), the multimodal fusion feature vector is determined to be an abnormal feature vector, and the spatial node information of the multimodal fusion feature vector corresponding to the abnormal feature vector is extracted.

[0169] Based on spatial node information, relevant data are retrieved, including structural anomaly differences, acoustic characteristics, vibration characteristics, and temperature data, to construct a multi-physical quantity dataset for each node. By calculating the correlation coefficients (such as Pearson correlation coefficients) between structural anomaly differences and acoustic, vibration, and temperature data, it is analyzed whether the anomaly is a structure-temperature linked anomaly or an independent acoustic / vibration micro-leakage anomaly. Finally, based on the strength of the correlation, a diagnostic conclusion (such as "structure-temperature induced anomaly" or "suspected micro-leakage") is automatically output, and a diagnostic report is automatically generated based on all anomaly nodes and their diagnostic conclusions.

[0170] Through the above technical solution, this application solves the problem of the lack of specificity in the diagnostic reports of the prior art. Through multi-dimensional data fusion and anomaly tracing mechanism, it can accurately identify the fault location and its physical cause, such as distinguishing between acoustic anomalies caused by thermal deformation and vibration anomalies caused by mechanical wear, thereby generating a diagnostic report with clear maintenance guidance value.

[0171] To better understand the above embodiments, an application scenario example is given as follows:

[0172] The study focused on the health monitoring of the gas path during the start-up process of a certain type of aero-engine.

[0173] The test engine is equipped with 32 distributed microphones (sampling rate 12.8kHz), 24 triaxial accelerometers (sampling rate 6.4kHz), 16 temperature sensors (sampling interval 50ms), and 16 fiber optic displacement gauges (sampling rate 2kHz). All data are synchronized at the hardware level via the IEEE 1588 PTP protocol (error <5μs), and the spatial coordinate mapping accuracy is better than 0.2mm.

[0174] During startup, all sensor data are collected in real time, and spatial nodes and timestamps are added uniformly. Temperature data is mapped to the engine's 3D model (approximately 6400 nodes) through an interpolation algorithm and input into the thermo-mechanical coupling model constructed by ANSYS Mechanical APDL, outputting the theoretical values ​​of thermal deformation for each spatial node. Structural deformation data is mapped to the corresponding spatial nodes through spatial coordinates and the difference is obtained from the theoretical values ​​of thermal deformation to obtain structural anomaly differences, which are used to identify local anomalies that cannot be explained by thermal loads.

[0175] Automatically reconstruct the real-time structural model (CATIA plugin) and dynamically correct the coordinates of all spatial nodes. Based on the real-time structural model, for each spatial node, calculate the actual propagation path of the acoustic signal under the real-time structural model based on the shortest path algorithm, and combine the elastic modulus and absorption coefficient of the node material to output three types of correction parameters: time delay, amplitude, and spectrum. After the acoustic signal is processed by time delay interpolation, spectrum shifting, and amplitude compensation, a structurally corrected acoustic signal is generated. Subsequently, the main frequency, harmonic energy, and spectral kurtosis of the structurally corrected acoustic signal are extracted and normalized together with the vibration characteristics of the accelerometer (RMS, main mode frequency, energy distribution, etc.) and the node resonance characteristics (extracted by joint spectrum analysis).

[0176] Using the structural anomaly difference as a weighting factor (dynamic normalization), the above features are weighted and fused to obtain a multimodal fusion feature vector for each spatial node. The standard feature vector library under healthy operating conditions is called periodically (per second) to calculate the Euclidean distance. Spatial nodes exceeding the threshold (3.0) are judged as "abnormal". For abnormal spatial nodes, their structural anomaly difference, acoustic, vibration, temperature and other data are retrieved and the correlation coefficient is calculated. If the correlation between structural anomaly and acoustic and vibration is low (<0.5), it is judged as "suspected micro-leakage"; if the correlation is high and the temperature is abnormal, it is judged as "thermal induced structural anomaly".

[0177] For example, at 4000 seconds, node #17 (gas path corner segment) detected an anomaly in the Euclidean distance between the multimodal fusion feature vector and the standard feature vector, which was 4.1. The structural anomaly difference of this node was 0.22 mm (far greater than the normal average of 0.03 mm), the acoustic dominant frequency drifted by 70 Hz, the vibration RMS increased by 23%, the temperature changed by 0.8°C, the structure-acoustic and structure-vibration correlation coefficients were 0.26 and 0.21, respectively, and the structure-temperature correlation was 0.13. Based on comprehensive analysis, the system automatically output the conclusion of "high-risk micro-leakage" and located the specific spatial node in the diagnostic report, listing all abnormal quantities and their physical causal relationships.

[0178] Example 2:

[0179] Please see Figure 5 A fault diagnosis system for the air circuit during the start-up process of an aircraft engine, comprising:

[0180] The data acquisition module is used to acquire acoustic signals, vibration signals, temperature data, and structural deformation data.

[0181] The structural anomaly difference calculation module is used to input the temperature data into a pre-built thermo-mechanical coupling model and output theoretical data of thermal deformation.

[0182] The difference between the theoretical thermal deformation data and the structural deformation data is calculated to obtain the structural anomaly difference.

[0183] The real-time structural model generation module is used to call the engine's three-dimensional model and generate a real-time structural model by combining the structural deformation data and the thermal deformation theoretical data.

[0184] An acoustic signal correction module is used to establish the actual propagation path of the acoustic signal based on the real-time structural model, and to calculate the correction parameter set of the acoustic signal based on the actual propagation path.

[0185] The acoustic signal is multidimensionally compensated according to the set of correction parameters to obtain a structure-corrected acoustic signal.

[0186] The multimodal fusion feature vector calculation module is used to perform weighted fusion of the structural modified acoustic signal and vibration signal with the structural anomaly difference as a weighting factor to obtain a multimodal fusion feature vector;

[0187] The diagnostic report output module is used to perform anomaly detection on the multimodal fusion feature vector and output a diagnostic report based on the anomaly detection results.

[0188] This embodiment has the same technical effects as Embodiment 1.

[0189] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0190] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for diagnosing air circuit faults during the start-up process of an aero-engine, characterized in that, Includes the following steps: Acquire acoustic signals, vibration signals, temperature data, and structural deformation data of the engine; The temperature data is input into a pre-constructed thermo-mechanical coupling model, and theoretical data of thermal deformation are output. The difference between the theoretical thermal deformation data and the structural deformation data is calculated to obtain the structural anomaly difference. The engine's three-dimensional model is invoked, and a real-time structural model is generated by combining the structural deformation data and the thermal deformation theoretical data. The actual propagation path of the acoustic signal is established based on the real-time structural model, and the set of correction parameters for the acoustic signal is calculated based on the actual propagation path. The acoustic signal is multidimensionally compensated according to the set of correction parameters to obtain a structure-corrected acoustic signal. Using the structural anomaly difference as a weighting factor, the structurally corrected acoustic signal and vibration signal are weighted and fused to obtain a multimodal fusion feature vector; Anomaly detection is performed on the multimodal fusion feature vector, and a diagnostic report is output based on the anomaly detection results.

2. The method for diagnosing air circuit faults during the start-up process of an aero-engine according to claim 1, characterized in that: Acquiring acoustic signals, vibration signals, temperature data, and structural deformation data specifically includes: The acoustic signals of the engine are acquired in real time through a distributed microphone array; the acoustic signals include sound speed, housing and air passage resonance sound signals during the start-up process; The vibration signals of the engine are acquired in real time through a multi-axis acceleration sensor array; the vibration signals include multi-axis acceleration vector data, resonant frequency, amplitude, and energy distribution characteristics at each stage of startup; Engine temperature data is collected in real time through a sensor network; The engine's structural deformation data is collected in real time using displacement sensors; the structural deformation data includes strain, real-time displacement measurements, and the evolution of structural deformation at each stage of startup. All the data collected above were given corresponding spatial location information and timestamp tags.

3. The method for diagnosing air circuit faults during the start-up process of an aero-engine according to claim 1, characterized in that: The process of calling upon the engine's 3D model and combining it with the structural deformation data and the theoretical data of thermal deformation to generate a real-time structural model specifically includes: Call the engine's 3D model and obtain the information of each spatial node within the 3D model based on the 3D model; The structural deformation data is mapped to the corresponding spatial nodes in the three-dimensional model using spatial coordinates. The thermal deformation theoretical data is mapped to the corresponding spatial nodes in the three-dimensional model through spatial coordinates, and coupled with the structural deformation data for calculation to obtain the comprehensive deformation amount of each spatial node. The spatial nodes are corrected based on the comprehensive deformation to obtain the new coordinates of each spatial node; Based on the new coordinates of each spatial node, the three-dimensional model of the engine is reconstructed to obtain the real-time structural model.

4. The method for diagnosing air circuit faults during the start-up process of an aero-engine according to claim 3, characterized in that: After generating the real-time structural model, the process also includes generating simulated sound field distribution data based on the temperature data, specifically including: The temperature data is mapped to each spatial node of the real-time structural model to obtain the temperature distribution data on the real-time structural model. Based on the temperature distribution data, the material parameters of each spatial node in the real-time structural model are corrected. Based on the real-time structural model, combined with temperature distribution data and the corrected material parameters of each spatial node, the boundary conditions for the engine sound field simulation are set. The acoustic finite element method was used to perform numerical calculations of the simulated sound field of the engine casing and the internal structure of the air passage. The results of the numerical calculations of the simulated sound field were used as the simulated sound field distribution data of the casing and the internal structure of the air passage at different spatial nodes.

5. The method for diagnosing air circuit faults during the start-up process of an aero-engine according to claim 3, characterized in that: Establishing the actual propagation path of the acoustic signal based on the real-time structural model, and calculating the correction parameter set of the acoustic signal based on the actual propagation path specifically includes: For each spatial node, based on the three-dimensional model, the shortest feasible propagation path of the acoustic signal in the corresponding structural state of the three-dimensional model is identified and used as the original propagation path of the acoustic signal. For each spatial node, based on the real-time structural model, the shortest feasible propagation path 2 of the acoustic signal in the corresponding structural state of the real-time structural model is identified and used as the actual propagation path of the acoustic signal. The actual propagation path and the original propagation path of the same spatial node are processed to obtain the set of correction parameters for the acoustic signal; The set of correction parameters includes time delay correction parameters, amplitude correction parameters, and spectrum correction parameters.

6. The method for diagnosing air circuit faults during the start-up process of an aero-engine according to claim 5, characterized in that: The acoustic signal is multidimensionally compensated according to the set of correction parameters to obtain the structure-corrected acoustic signal, specifically including: Based on the time delay correction parameters, the acoustic signal is interpolated and corrected on the time axis; The spectrum of the interpolated acoustic signal is obtained by fast Fourier transform. Based on the aforementioned spectrum correction parameters, the spectrum is shifted entirely along the frequency axis. Based on the amplitude correction parameters, multiplication compensation is performed on the amplitude of each frequency point of the shifted spectrum; The multidimensional compensated acoustic signal is obtained by inverse fast Fourier transform of the spectrum after multiplication compensation, and is used as the structure-corrected acoustic signal.

7. A method for diagnosing air circuit faults during the start-up process of an aero-engine according to claim 4, characterized in that: After obtaining the structure-corrected acoustic signal, the method further includes performing joint spectral analysis on the simulated sound field distribution data and the structure-corrected acoustic signal at the same spatial node to output resonance characteristic data, specifically including: Acquire the structurally corrected acoustic signal and simulated sound field distribution data of the same spatial node; The structurally corrected acoustic signal and the simulated sound field distribution data are respectively subjected to denoising and normalization processing; Time-frequency analysis was used to extract the spectral characteristics of the processed structurally corrected acoustic signal and the simulated sound field distribution data, respectively. The spectral characteristics of the processed structurally corrected acoustic signal and the simulated sound field distribution data are matched one-to-one with the spatial node coordinates and the same frequency and spatial node coordinates as indices. The spectral characteristics of the processed structure-corrected acoustic signal and the simulated sound field distribution data are subjected to joint spectral analysis in the frequency domain, and the joint spectral analysis results of the two at each frequency point are output. By integrating the results of the joint spectrum analysis, a spatial-frequency distribution matrix is ​​obtained; Feature extraction is performed on the spatial-frequency distribution matrix to obtain resonance feature data.

8. A method for diagnosing air circuit faults during the start-up process of an aero-engine according to claim 7, characterized in that: Using the structural anomaly difference as a weighting factor, the acoustic and vibration features are fused to obtain a multimodal fusion feature vector, specifically including: Based on the structural anomaly difference, a weighting factor is assigned to each spatial node; The acoustic, vibration, and resonance feature data are respectively normalized. The normalized acoustic features, vibration features, and resonance features are weighted and fused according to the weighting factors to obtain a multimodal fusion feature vector.

9. A method for diagnosing air circuit faults during the start-up process of an aero-engine according to claim 1, characterized in that: Anomaly detection is performed on the multimodal fused feature vector, and a diagnostic report is output based on the anomaly detection results, specifically including: Call the standard feature vector and calculate the Euclidean distance between the multimodal fusion feature vector and the standard feature vector; If the Euclidean distance is greater than a preset threshold, then the multimodal fusion feature vector is determined to be an abnormal feature vector. Extract spatial node information of the multimodal fusion feature vector corresponding to the abnormal feature vector; Based on the spatial node information, relevant data are retrieved, including: structural anomaly difference, acoustic characteristics, vibration characteristics, and temperature data; For this spatial node, calculate the correlation coefficient between structural anomaly difference, acoustic characteristics, vibration characteristics and temperature data; Based on the calculation results of the correlation coefficient, output the diagnostic conclusion for this spatial node; A diagnostic report is output based on the diagnostic node and the spatial node information.

10. A gas path fault diagnosis system for the start-up process of an aero-engine, characterized in that, include: The data acquisition module is used to acquire acoustic signals, vibration signals, temperature data, and structural deformation data. The structural anomaly difference calculation module is used to input the temperature data into a pre-built thermo-mechanical coupling model and output theoretical data of thermal deformation. The difference between the theoretical thermal deformation data and the structural deformation data is calculated to obtain the structural anomaly difference. The real-time structural model generation module is used to call the engine's three-dimensional model and generate a real-time structural model by combining the structural deformation data and the thermal deformation theoretical data. An acoustic signal correction module is used to establish the actual propagation path of the acoustic signal based on the real-time structural model, and to calculate the correction parameter set of the acoustic signal based on the actual propagation path. The acoustic signal is multidimensionally compensated according to the set of correction parameters to obtain a structure-corrected acoustic signal. The multimodal fusion feature vector calculation module is used to perform weighted fusion of the structural modified acoustic signal and vibration signal with the structural anomaly difference as a weighting factor to obtain a multimodal fusion feature vector; The diagnostic report output module is used to perform anomaly detection on the multimodal fusion feature vector and output a diagnostic report based on the anomaly detection results.

Images