Aero-engine test-bed performance parameter synchronous acquisition and analysis system based on multi-sensor fusion

Abstract

The application discloses a kind of based on multi-sensor fusion's aero-engine test bed performance parameter synchronous acquisition and analysis system;System is through the hierarchical linkage of sensor layer, data acquisition layer, data processing layer and analysis decision layer, integrates multiple types of high-precision sensors, adopts the dual time alignment mechanism of hardware synchronous trigger and software compensation combination, fuses the multi-modal algorithm of Kalman filter, SVM classifier, deep neural network (DNN) and D-S evidence reasoning, with double-layer inverse Broyden iteration and online learning adaptive model, realize the accurate synchronous acquisition of engine temperature, pressure, vibration and other key parameters, real-time analysis and fault early warning.System is compatible with multiple aviation buses, with redundant transmission and anti-interference design, visual interface supports multidimensional data linkage and playback, can adapt to the engine test demand under different conditions, provides reliable technical support for aero-engine research and development, maintenance and performance optimization.

Description

Technical Field

[0001] This invention relates to the field of aero-engine testing and condition monitoring technology, specifically to a system for synchronous acquisition and analysis of aero-engine test bench performance parameters based on multi-sensor fusion. Background Technology

[0002] As the core power unit of aircraft, the performance reliability and operational safety of aero-engines directly determine the flight quality of aircraft. Test bench testing is a key link in the research, development, finalization, and maintenance of aero-engines. By accurately monitoring and analyzing various performance parameters of the engine under different operating conditions, the rationality of the design scheme can be verified, the health of the operating status can be assessed, and potential failure risks can be identified, providing data support for engine performance optimization and predictive maintenance. As aero-engines develop towards higher thrust-to-weight ratios, higher reliability, and longer lifespans, their internal working mechanisms are becoming increasingly complex, placing higher demands on the parameter acquisition accuracy, synchronization, data processing capabilities, and analysis and decision-making efficiency of the test bench testing system.

[0003] Existing performance parameter acquisition and analysis systems for aero-engine test benches have numerous technical limitations, making it difficult to meet current testing requirements for high precision, high reliability, and intelligence. Firstly, the accuracy of parameter synchronization is insufficient. Existing systems mostly rely on single hardware triggers to achieve sensor synchronization, failing to fully consider the impact of signal transmission delays, differences in sensor sampling frequencies, and environmental interference on time alignment. This results in significant timestamp discrepancies among multi-source parameters, failing to accurately reflect the dynamic correlations between parameters and consequently affecting the accuracy of performance evaluation and fault diagnosis. Secondly, the data fusion algorithm design is simplistic. Most systems only use simple weighted averaging or single filtering algorithms to process multi-source data, making it difficult to handle the nonlinear characteristics and uncertainties of heterogeneous parameters such as temperature, pressure, and vibration. This hinders the effective extraction of deep correlation features from the data, leading to low accuracy in fault mode identification and making it difficult to achieve accurate early warning of minor faults.

[0004] Furthermore, existing systems mostly use fixed-parameter performance analysis models, failing to consider engine performance degradation under different test conditions and after long-term operation. The model parameters cannot be corrected in real time, resulting in significant deviations between the calculated performance parameters and the actual operating conditions, thus failing to provide accurate data support for engine performance optimization. At the same time, the fault warning mechanism is not perfect. Existing systems mostly trigger alarms based on single parameter thresholds, without combining the fusion characteristics of multi-source parameters and historical data to build a comprehensive health status assessment system. This easily leads to false alarms or missed alarms, making it difficult to support the implementation of predictive maintenance.

[0005] Furthermore, the existing systems suffer from insufficient transmission compatibility and reliability. Most systems only support a single aviation bus protocol, which cannot adapt to the hardware interface requirements of different engine test benches. Moreover, they lack effective redundant transmission design, making them prone to data loss in the event of bus failure. The visualization and data playback functions are relatively simple, only able to display curves of single parameters, and do not support multi-dimensional parameter linkage analysis and panoramic playback of the test process, which is not conducive to in-depth data mining and problem tracing. Finally, the system's environmental adaptability needs to be improved. The existing sensor packaging and signal processing circuits lack anti-interference and high-temperature resistance design. In the complex environment of high temperature and strong electromagnetic interference on the test bench, problems such as signal distortion and sensor failure are prone to occur, affecting the continuity of the test process and the reliability of the data.

[0006] The aforementioned technical limitations make it difficult for existing systems to achieve high-precision synchronous acquisition and intelligent analysis of aero-engine performance parameters, which restricts the testing efficiency of test benches and the level of intelligence in engine research and development and maintenance. Therefore, there is an urgent need for a multi-sensor fusion acquisition and analysis system with high synchronization, strong adaptability and high reliability to solve the shortcomings of existing technologies. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a system for synchronous acquisition and analysis of performance parameters of aero-engine test benches based on multi-sensor fusion, which solves the problems mentioned in the background.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solution: a synchronous acquisition and analysis system for performance parameters of an aero-engine test stand based on multi-sensor fusion, comprising a sensor layer, a data acquisition layer, a data processing layer and an analysis and decision layer cascaded in sequence;

[0009] The sensor layer is used to collect key performance parameters of the aero-engine and output multi-source raw sensor signals. The data acquisition layer receives the raw sensor signals, achieves time alignment of multi-source signals through hardware synchronization triggering and software compensation, and transmits them to the data processing layer after signal conditioning and analog-to-digital conversion. The data processing layer preprocesses the synchronized digital signals and performs multi-sensor fusion calculations to extract feature parameters. The analysis and decision-making layer completes engine performance evaluation, fault warning, and data visualization based on the feature parameters. All levels realize data interaction and command transmission through the aviation bus.

[0010] Optionally, the sensor layer includes a temperature monitoring module, a pressure monitoring module, a vibration monitoring module, a speed monitoring module, and a gas path analysis module;

[0011] The temperature monitoring module uses a combination of a K-type thermocouple and a PT100 resistance thermometer to cover the high-temperature areas of the combustion chamber and turbine blades; the pressure monitoring module uses a piezoelectric pressure sensor for pressure acquisition in the intake duct, compressor, and combustion chamber; the vibration monitoring module uses a piezoelectric accelerometer to detect rotor imbalance and bearing failure; the speed monitoring module uses a combination of a magnetoelectric sensor and a Hall effect sensor to achieve non-contact measurement of high and low pressure rotor speeds; and the gas path analysis module uses laser absorption spectroscopy and a spectral analysis module to achieve ppm-level detection of O2 and CO2 concentrations in the exhaust gas and to evaluate combustion efficiency.

[0012] Optionally, the data acquisition layer includes a synchronization triggering unit, a signal conditioning unit, a high-precision ADC module, and a bus transmission unit. The synchronization triggering unit generates a GPS-PPS global trigger signal through an FPGA to synchronize the sampling clocks of all sensors. The signal conditioning unit integrates an anti-aliasing filter and a linear compensation circuit to perform noise reduction, amplification, and linearization processing on the original sensor signal. The high-precision ADC module is a 24-bit analog-to-digital converter that converts the conditioned analog signal into a digital signal. The bus transmission unit adopts the ARINC664 / AFDX aviation bus protocol and achieves fault-tolerant transmission through a dual-redundant network. The data acquisition layer also includes a timestamp unit and an interpolation calibration unit. The timestamp unit adds a precise timestamp to each digital signal, and the interpolation calibration unit uses a cubic spline interpolation algorithm to fill the signal gaps caused by sampling frequency differences. Specifically, given a known node... Corresponding function value In the interval cubic spline interpolation function satisfy ,and , , The interpolation results are obtained by solving the following system of equations:

[0013]

[0014] Optionally, the data processing layer includes a preprocessing unit and a multi-sensor fusion unit; the preprocessing unit uses the Grubbs criterion to remove outliers, performs digital filtering through a Butterworth filter, and performs cold junction compensation on the thermocouple signal, with the compensation formula being:

[0015]

[0016] in, For actual measurement of electromotive force (mV). The thermocouple outputs an electromotive force (mV). Cold end temperature The corresponding electromotive force (mV); the multi-sensor fusion unit integrates a Kalman filter algorithm for optimal estimation of dynamic parameters, and its state equation and observation equation are as follows:

[0017]

[0018]

[0019] in, for Time-state vector Here is the state transition matrix. To control the input matrix, To control the input vector, For process noise, Let k be the observation vector at time k. For the observation matrix, To account for observation noise; through Kalman filtering prediction-update iteration, the prediction is: ;renew: The optimal state estimate is obtained. .

[0020] Optionally, the multi-sensor fusion unit also integrates an SVM classifier, a DNN model, and a DS evidence reasoning algorithm; the SVM classifier constructs a state monitoring feature library based on the ISO-13374 standard, and maps a high-dimensional feature space through a radial basis function kernel, the formula of which is:

[0021]

[0022] in, Kernel function parameters , For vectors and The optimal classification hyperplane is obtained by solving a convex quadratic programming problem using the Euclidean distance, enabling the identification of fault modes such as bearing wear and surge. The DNN model includes an input layer, three hidden layers, and an output layer. The input layer dimension is consistent with the fusion feature dimension, and the hidden layers use the ReLU activation function. The output layer uses the Sigmoid function to output the predicted values ​​of nonlinear parameters, and the model loss function is:

[0023]

[0024] in, These are the model parameters (weights and biases). For the sample size, For the first The true value of each sample For the first The predicted value of each sample; the DS evidence reasoning algorithm uses the Kalman filter estimation result, SVM classification result, and DNN prediction result as independent evidence, and uses the mass function Assign trust For identification framework A subset of, satisfying and Combined with Dempster's synthesis rules By integrating evidence from multiple sources, the final fault diagnosis result is obtained.

[0025] Optionally, the analysis and decision-making layer includes a performance evaluation unit, a fault early warning unit, and a visualization unit; the performance evaluation unit, based on a component-level model of a twin-shaft turbojet engine, solves the common working equations using a two-layer inverse Broyden iterative method to calculate the total temperature and total pressure parameters of key sections; the inner iterative formula is:

[0026]

[0027] in, This is a vector composed of the compressor pressure ratio and the turbine pressure ratio. For the first Jacobian matrix of the next iteration It is the residual function;

[0028] The outer iteration formula is:

[0029]

[0030] in, For health parameter vectors, Step size factor ( ), For the health parameter Jacobian matrix, The performance evaluation unit is a performance deviation function; it also integrates an online learning module to periodically update the DNN model weights using stochastic gradient descent. ( (Learning rate), adapting to different operating conditions and performance degradation states.

[0031] Optionally, the fault warning unit includes a threshold setting module and a health index calculation module; the threshold setting module sets a safe threshold range for each performance parameter based on engine design parameters and historical test data, and triggers a real-time alarm when the parameter exceeds the threshold; the health index calculation module constructs a health index based on multi-source fusion features and using a weighted summation formula.

[0032]

[0033] in, To the number of features to be fused, For the first The weights of each feature, For the first Normalized values ​​of each feature ( The fault early warning unit combines the trend of health index changes with historical data, and uses a linear regression model. ( For degradation rate, A performance degradation model is established for the intercept to achieve predictive early warning of potential faults.

[0034] Optionally, the visualization unit is built on the LabVIEW platform and includes a parameter curve display module, a 3D model dynamic simulation module, and a data playback module. The parameter curve display module displays time series curves of parameters such as temperature, pressure, and vibration in real time, and supports multi-parameter comparison and correlation analysis on the same screen. The 3D model dynamic simulation module is based on the engine's 3D model and synchronously displays the operating status and parameter distribution of key components. The data playback module supports the playback of test data along the time axis, and allows for customization of playback speed and parameter filtering conditions, enabling panoramic traceability of the testing process.

[0035] Optionally, all sensors in the sensor layer adopt an anti-interference packaging design. The temperature monitoring module and pressure monitoring module adopt ceramic coating packaging and air purging cooling system, while the vibration monitoring module and speed monitoring module adopt shielded shell packaging. The circuits of the data acquisition layer and data processing layer adopt differential signal transmission and grounding optimization design to reduce the impact of electromagnetic interference and radio frequency interference on the signal.

[0036] Optionally, the aviation bus is also compatible with CAN bus and RS485 bus protocols, and the conversion between different protocols is realized through the bus adapter module; the data processing layer also includes a data storage unit, which uses a solid-state hard disk array to store raw data and processing results, with a storage capacity of not less than 1TB, supports data indexing and long-term traceability by timestamp, working condition type and parameter category, and provides an API interface to support secondary analysis.

[0037] This invention provides a system for synchronous acquisition and analysis of performance parameters of aero-engine test bench based on multi-sensor fusion, which has the following advantages:

[0038] This multi-sensor fusion-based system for synchronous acquisition and analysis of performance parameters of aero-engine test benches effectively solves the problems of low synchronization accuracy, weak data processing capabilities, poor model adaptability, and insufficient reliability of existing aero-engine test bench parameter acquisition and analysis systems through hierarchical architecture design and multi-technology integration and innovation. It has significant technical advantages and application value.

[0039] This invention employs a hardware synchronization mechanism that generates GPS-PPS global trigger signals using FPGA, combined with a software compensation strategy of timestamp alignment and cubic spline interpolation. This significantly reduces the impact of signal transmission delay and sampling frequency differences on time alignment, ensuring that the time synchronization accuracy of multi-source sensor parameters reaches the microsecond level, laying the foundation for accurate analysis of the dynamic correlation between parameters. Secondly, this invention integrates a multi-modal algorithm system of Kalman filtering, SVM classifier, DNN, and DS evidence reasoning, achieving differentiated processing for the characteristics of different types of parameters: Kalman filtering suppresses noise interference of dynamic parameters through optimal estimation, SVM classifier achieves accurate identification of fault modes based on a standardized fault feature library, DNN model mines the intrinsic correlation of nonlinear parameters through deep network structure, and DS evidence reasoning effectively integrates multi-source fault features. The collaborative work of multiple algorithms not only improves the comprehensiveness and accuracy of data processing, but also enhances the system's adaptability to heterogeneous data under complex working conditions.

[0040] Meanwhile, the dual-layer inverse Broyden iterative algorithm and online learning mechanism proposed in this invention realize the dynamic adaptive correction of the performance evaluation model. The inner layer iteration ensures the calculation accuracy by solving the component-level parameter relationship, while the outer layer iteration corrects the health parameters in real time for engine performance degradation. Combined with the online weight update of the DNN model, the system can quickly adapt to different test conditions and changes in engine operating status, avoiding performance evaluation bias caused by a fixed model and providing more accurate data support for engine performance optimization. This invention constructs a dual early warning mechanism of "threshold alarm + health index evaluation". It constructs a health index (HI) through multi-source parameter fusion features and establishes a performance degradation trend model by combining historical data. It can not only respond to parameter over-limit faults in real time, but also predict potential fault risks in advance, providing a basis for predictive maintenance decisions.

[0041] This invention adopts a dual-redundant ARINC664 / AFDX avionics bus transmission architecture, coupled with shielded cables, differential signal transmission, and grounding optimization anti-interference design, ensuring the stability and integrity of data transmission in complex environments; at the same time, it is compatible with multiple avionics bus protocols and sensor interfaces, and can flexibly adapt to the modification and upgrade needs of different types of aero-engine test benches, reducing system deployment costs.

[0042] In summary, this invention, through multi-technology collaborative innovation, has achieved an intelligent upgrade of the entire process of performance parameter acquisition, processing, analysis, and decision-making for aero-engine test benches. This not only improves the accuracy and reliability of test data but also enhances the system's adaptability to complex operating conditions and environments. It provides comprehensive technical support for aero-engine R&D iteration, fault diagnosis, and predictive maintenance, and is of great significance for promoting the intelligent development of aero-engine testing technology. Detailed Implementation

[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0044] This invention provides a technical solution: a synchronous acquisition and analysis system for performance parameters of an aero-engine test bench based on multi-sensor fusion, comprising a sensor layer, a data acquisition layer, a data processing layer, and an analysis and decision layer cascaded in sequence;

[0045] The sensor layer is used to collect key performance parameters of the aero-engine and output multi-source raw sensor signals. The sensor layer includes a temperature monitoring module, a pressure monitoring module, a vibration monitoring module, a speed monitoring module, and a gas path analysis module. The temperature monitoring module uses a combination of K-type thermocouples and PT100 resistance thermometers. The K-type thermocouples have a measurement range of -200℃ to 1250℃, and the PT100 resistance thermometers have a measurement range of -50℃ to 1600℃, covering high-temperature areas such as the combustion chamber and turbine blades. The pressure monitoring module uses a piezoelectric pressure sensor with a measurement range of 0-5 MPa. The system is used for pressure acquisition in the intake duct, compressor, and combustion chamber; the vibration monitoring module uses a piezoelectric accelerometer with a sensitivity of 100mV / g and a frequency response of 0.5Hz-10kHz to detect rotor imbalance and bearing failure; the speed monitoring module uses a combination of a magnetoelectric sensor and a Hall effect sensor, with the Hall effect sensor measuring a range of 0-20,000RPM, enabling non-contact measurement of high and low pressure rotor speeds; the gas path analysis module uses laser absorption spectroscopy and a spectral analysis module to achieve ppm-level detection of O2 and CO2 concentrations in the exhaust gas and to evaluate combustion efficiency.

[0046] All sensors in the sensor layer adopt an anti-interference packaging design. The temperature monitoring module and pressure monitoring module adopt ceramic coating packaging and air purging cooling system, while the vibration monitoring module and speed monitoring module adopt shielded shell packaging. The circuits of the data acquisition layer and data processing layer adopt differential signal transmission and grounding optimization design to reduce the impact of electromagnetic interference and radio frequency interference on the signal.

[0047] The data acquisition layer receives raw sensor signals, achieves time alignment of multi-source signals through hardware synchronization triggering and software compensation, and transmits them to the data processing layer after signal conditioning and analog-to-digital conversion. The data acquisition layer includes a synchronization triggering unit, a signal conditioning unit, a high-precision ADC module, and a bus transmission unit. The synchronization triggering unit generates a GPS-PPS global trigger signal through an FPGA to achieve clock synchronization of all sensor sampling, with a time alignment accuracy ≤1μs. The signal conditioning unit integrates an anti-aliasing filter and a linear compensation circuit to perform noise reduction, amplification, and linearization processing on the raw sensor signals. The precision ADC module is a 24-bit analog-to-digital converter with a synchronous sampling rate ≥100kHz, converting conditioned analog signals into digital signals. The bus transmission unit adopts the ARINC664 / AFDX avionics bus protocol, with a transmission rate of 100Mbps and a deterministic delay ≤50μs, achieving fault-tolerant transmission through a dual-redundant network. The data acquisition layer also includes a timestamp unit and an interpolation calibration unit. The timestamp unit adds a precise timestamp to each digital signal, and the interpolation calibration unit uses a cubic spline interpolation algorithm to fill signal gaps caused by sampling frequency differences. Specifically, given a known node... Corresponding function value In the interval cubic spline interpolation function satisfy ,and , , The interpolation results are obtained by solving the following system of equations:

[0048]

[0049] The data processing layer preprocesses and performs multi-sensor fusion calculations on the synchronized digital signal to extract feature parameters. The data processing layer includes a preprocessing unit and a multi-sensor fusion unit. The preprocessing unit uses the Grubbs criterion to remove outliers, performs digital filtering using a Butterworth filter, and performs cold junction compensation on the thermocouple signal. The compensation formula is as follows:

[0050]

[0051] in, For actual measurement of electromotive force (mV). The thermocouple outputs an electromotive force (mV). Cold end temperature The electromotive force (mV) corresponding to (°C); the multi-sensor fusion unit integrates a Kalman filter algorithm for optimal estimation of dynamic parameters, and its state equation and observation equation are as follows:

[0052]

[0053]

[0054] in, for Time-state vector Here is the state transition matrix. To control the input matrix, To control the input vector, For process noise (mean 0, covariance 0), Gaussian white noise). Let k be the observation vector at time k. For the observation matrix, The observation noise (mean 0, covariance ) Gaussian white noise); prediction-update iteration via Kalman filtering,

[0055] predict:

[0056]

[0057]

[0058] renew:

[0059]

[0060]

[0061]

[0062] Obtain the optimal state estimate ;

[0063] The multi-sensor fusion unit also integrates an SVM classifier, a DNN model, and a DS evidence reasoning algorithm; the SVM classifier constructs a state monitoring feature library based on the ISO-13374 standard, and maps a high-dimensional feature space through a radial basis function kernel function, the formula of which is:

[0064]

[0065] in, Kernel function parameters , For vectors and The optimal classification hyperplane is obtained by solving a convex quadratic programming problem using the Euclidean distance, enabling the identification of fault modes such as bearing wear and surge. The DNN model includes an input layer, three hidden layers, and an output layer. The input layer dimension is consistent with the fusion feature dimension, and the hidden layers use the ReLU activation function. The output layer uses the Sigmoid function to output the predicted values ​​of nonlinear parameters, and the model loss function is:

[0066]

[0067] in, For model parameters, For the sample size, For the first The true value of each sample For the first The predicted value of each sample; the DS evidence reasoning algorithm uses the Kalman filter estimation result, SVM classification result, and DNN prediction result as independent evidence, and uses the mass function Assign trust For identification framework A subset of, satisfying and Combined with Dempster's synthesis rules By integrating evidence from multiple sources, the final fault diagnosis result is obtained;

[0068] The analysis and decision-making layer completes engine performance evaluation, fault warning, and data visualization based on characteristic parameters. Data interaction and command transmission between levels are achieved through an aviation bus. The analysis and decision-making layer includes a performance evaluation unit, a fault warning unit, and a visualization unit. The performance evaluation unit, based on a component-level model of a twin-shaft turbojet engine, solves the common working equations using a two-layer inverse Broyden iterative method to calculate the total temperature and pressure parameters of key sections. The inner iterative formula is:

[0069]

[0070] in, This is a vector composed of the compressor pressure ratio and the turbine pressure ratio. For the first Jacobian matrix of the next iteration The residual function (common working equation deviation) is used; the outer iteration formula is:

[0071]

[0072] in, For health parameter vectors (expiratory air flow, mechanical efficiency, etc.) Step size factor ( ), For the health parameter Jacobian matrix, The performance evaluation unit is defined as a performance deviation function (the deviation between calculated and measured values). It also integrates an online learning module that periodically updates the DNN model weights using stochastic gradient descent. ( (Learning rate), adapting to different operating conditions and performance degradation states;

[0073] The fault warning unit includes a threshold setting module and a health index calculation module. The threshold setting module sets safe threshold ranges for each performance parameter based on engine design parameters and historical test data, triggering a real-time alarm when a parameter exceeds the threshold. The health index calculation module constructs a health index based on multi-source fusion features and a weighted summation formula.

[0074]

[0075] in, To the number of features to be fused, For the first The weights of each feature (satisfying) (Determined through the Analytic Hierarchy Process) For the first Normalized values ​​of each feature ( The fault early warning unit combines the trend of health index changes with historical data, and uses a linear regression model. ( For degradation rate, A performance degradation model is established for the intercept to achieve predictive early warning of potential faults;

[0076] The visualization unit is built on the LabVIEW platform and includes a parameter curve display module, a 3D model dynamic simulation module, and a data playback module. The parameter curve display module displays time series curves of parameters such as temperature, pressure, and vibration in real time, and supports multi-parameter comparison and correlation analysis on the same screen. The 3D model dynamic simulation module is based on the engine's 3D model and synchronously displays the operating status and parameter distribution of key components. The data playback module supports the playback of test data along the time axis, and allows for customization of playback speed and parameter filtering conditions, enabling panoramic traceability of the testing process.

[0077] The aviation bus is also compatible with CAN bus and RS485 bus protocols, and the conversion between different protocols is realized through the bus adapter module; the data processing layer also includes a data storage unit, which uses a solid-state hard disk array to store raw data and processing results, with a storage capacity of not less than 1TB. It supports data indexing and long-term traceability by timestamp, working condition type and parameter category, and provides API interface to support secondary analysis.

[0078] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A system for synchronous acquisition and analysis of performance parameters of an aero-engine test stand based on multi-sensor fusion, characterized in that, It includes a cascaded sensor layer, a data acquisition layer, a data processing layer, and an analysis and decision-making layer; The sensor layer is used to collect key performance parameters of the aero-engine and output multi-source raw sensor signals. The data acquisition layer receives the raw sensor signals, achieves time alignment of multi-source signals through hardware synchronization triggering and software compensation, and transmits them to the data processing layer after signal conditioning and analog-to-digital conversion. The data processing layer preprocesses the synchronized digital signals and performs multi-sensor fusion calculations to extract feature parameters. The analysis and decision-making layer completes engine performance evaluation, fault warning, and data visualization based on the feature parameters. All levels realize data interaction and command transmission through the aviation bus.

2. The system according to claim 1, characterized in that, The sensor layer includes a temperature monitoring module, a pressure monitoring module, a vibration monitoring module, a speed monitoring module, and a gas path analysis module; The temperature monitoring module uses a combination of a K-type thermocouple and a PT100 resistance thermometer to cover the high-temperature areas of the combustion chamber and turbine blades; the pressure monitoring module uses a piezoelectric pressure sensor for pressure acquisition in the intake duct, compressor, and combustion chamber; the vibration monitoring module uses a piezoelectric accelerometer to detect rotor imbalance and bearing failure; the speed monitoring module uses a combination of a magnetoelectric sensor and a Hall effect sensor to achieve non-contact measurement of high and low pressure rotor speeds; and the gas path analysis module uses laser absorption spectroscopy and a spectral analysis module to achieve ppm-level detection of O2 and CO2 concentrations in the exhaust gas and to evaluate combustion efficiency.

3. The system according to claim 1, characterized in that, The data acquisition layer includes a synchronization triggering unit, a signal conditioning unit, a high-precision ADC module, and a bus transmission unit. The synchronization triggering unit generates a GPS-PPS global trigger signal via an FPGA to synchronize the sampling clocks of all sensors. The signal conditioning unit integrates an anti-aliasing filter and a linear compensation circuit to reduce noise, amplify, and linearize the original sensor signals. The high-precision ADC module is a 24-bit analog-to-digital converter that converts the conditioned analog signals into digital signals. The bus transmission unit adopts the ARINC664 / AFDX aviation bus protocol and achieves fault-tolerant transmission through a dual-redundant network. The data acquisition layer also includes a timestamp unit and an interpolation calibration unit. The timestamp unit adds a precise timestamp to each digital signal, and the interpolation calibration unit uses a cubic spline interpolation algorithm to fill the signal gaps caused by sampling frequency differences. Specifically, given a known node... Corresponding function value In the interval cubic spline interpolation function satisfy ,and , , The interpolation results are obtained by solving the following system of equations:

4. The system according to claim 1, characterized in that, The data processing layer includes a preprocessing unit and a multi-sensor fusion unit. The preprocessing unit uses the Grubbs criterion to remove outliers, performs digital filtering using a Butterworth filter, and performs cold junction compensation on the thermocouple signal. The compensation formula is as follows: ； in, For actual measurement of electromotive force (mV). The thermocouple outputs an electromotive force (mV). Cold end temperature The corresponding electromotive force (mV); the multi-sensor fusion unit integrates a Kalman filter algorithm for optimal estimation of dynamic parameters, and its state equation and observation equation are as follows: ； ； in, for Time-state vector Here is the state transition matrix. To control the input matrix, To control the input vector, For process noise, Let k be the observation vector at time k. For the observation matrix, To account for observation noise; through Kalman filtering prediction-update iteration, the prediction is: ;renew: The optimal state estimate is obtained. .

5. The system according to claim 4, characterized in that, The multi-sensor fusion unit also integrates an SVM classifier, a DNN model, and a DS evidence reasoning algorithm. The SVM classifier constructs a state monitoring feature library based on the ISO-13374 standard and maps a high-dimensional feature space through a radial basis function kernel. The kernel function formula is: ； in, Kernel function parameters , For vectors and The optimal classification hyperplane is obtained by solving a convex quadratic programming problem using the Euclidean distance, enabling the identification of fault modes such as bearing wear and surge. The DNN model includes an input layer, three hidden layers, and an output layer. The input layer dimension is consistent with the fusion feature dimension, and the hidden layers use the ReLU activation function. The output layer uses the Sigmoid function to output the predicted values ​​of nonlinear parameters, and the model loss function is: ； in, These are the model parameters (weights and biases). For the sample size, For the first The true value of each sample For the first The predicted value of each sample; the DS evidence reasoning algorithm uses the Kalman filter estimation result, SVM classification result, and DNN prediction result as independent evidence, and uses the mass function Assign trust For identification framework A subset of, satisfying and Combined with Dempster's synthesis rules By integrating evidence from multiple sources, the final fault diagnosis result is obtained.

6. The system according to claim 1, characterized in that, The analysis and decision-making layer includes a performance evaluation unit, a fault early warning unit, and a visualization unit. The performance evaluation unit, based on a component-level model of a twin-shaft turbojet engine, solves the common working equations using a two-layer inverse Broyden iterative method to calculate the total temperature and pressure parameters of key sections. The inner iterative formula is: ； in, This is a vector composed of the compressor pressure ratio and the turbine pressure ratio. For the first Jacobian matrix of the next iteration It is the residual function; The outer iteration formula is: ； in, For health parameter vectors, Step size factor ( ), For the health parameter Jacobian matrix, The performance evaluation unit is a performance deviation function; it also integrates an online learning module to periodically update the DNN model weights using stochastic gradient descent. ( (Learning rate), adapting to different operating conditions and performance degradation states.

7. The system according to claim 6, characterized in that, The fault warning unit includes a threshold setting module and a health index calculation module. The threshold setting module sets safe threshold ranges for each performance parameter based on engine design parameters and historical test data, triggering a real-time alarm when a parameter exceeds the threshold. The health index calculation module constructs a health index based on multi-source fusion features and a weighted summation formula. ； in, To the number of features to be fused, For the first The weights of each feature For the first Normalized values ​​of each feature ( The fault early warning unit combines the trend of health index changes with historical data, and uses a linear regression model. ( For degradation rate, A performance degradation model is established for the intercept to achieve predictive early warning of potential faults.

8. The system according to claim 6, characterized in that, The visualization unit is built on the LabVIEW platform and includes a parameter curve display module, a 3D model dynamic simulation module, and a data playback module. The parameter curve display module displays time series curves of parameters such as temperature, pressure, and vibration in real time, and supports multi-parameter comparison and correlation analysis on the same screen. The 3D model dynamic simulation module is based on the engine's 3D model and synchronously displays the operating status and parameter distribution of key components. The data playback module supports the playback of test data along the time axis, and allows for customization of playback speed and parameter filtering conditions, enabling panoramic traceability of the testing process.

9. The system according to claim 1, characterized in that, All sensors in the sensor layer adopt an anti-interference packaging design. The temperature monitoring module and pressure monitoring module adopt ceramic coating packaging and air purging cooling system, while the vibration monitoring module and speed monitoring module adopt shielded shell packaging. The circuits of the data acquisition layer and data processing layer adopt differential signal transmission and grounding optimization design to reduce the impact of electromagnetic interference and radio frequency interference on the signal.

10. The system according to claim 1, characterized in that, The aviation bus is also compatible with CAN bus and RS485 bus protocols, and the conversion between different protocols is realized through the bus adapter module; the data processing layer also includes a data storage unit, which uses a solid-state hard disk array to store raw data and processing results, with a storage capacity of not less than 1TB, supports data indexing and long-term traceability by timestamp, working condition type and parameter category, and provides API interface to support secondary analysis.

Images