A Mining Method, Medium and System for Wind Tunnel Health Monitoring Data

Abstract

The present invention provides a method, medium and system for mining wind tunnel health monitoring data, belonging to the technical field of wind tunnel health monitoring data mining and processing. It includes collecting multiple sensing signals to construct a monitoring data stream, performing filtering and noise reduction processing to obtain preprocessed data and an abnormal data set, starting the GPU and CPU processing streams to receive the preprocessed data and the abnormal data set, performing component decomposition to obtain a steady-state data matrix, a dynamic data matrix, an abnormal feature matrix, a parameter data vector and an abnormal parameter vector, establishing a data mining feature equation set, using the wind tunnel operation state evaluation equation set to calculate the wind tunnel operation state evaluation result, generating a wind tunnel operation feature matrix and a wind tunnel state warning index, and outputting a warning signal according to the comparison result between the warning index and a preset health threshold. It solves the technical problem that currently, it is impossible to quickly and efficiently mine and analyze wind tunnel health detection data from both the steady-state and dynamic aspects, resulting in the inability to better discover potential fault signs.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind tunnel health monitoring data mining and processing. Specifically, it relates to a method, medium and system for mining wind tunnel health monitoring data. Background Art

[0002] With the rapid development of aerospace technology, wind tunnel tests play an irreplaceable role in aircraft R & D, aerodynamic shape optimization and other fields. As a large-scale precision experimental device, the operating state of the wind tunnel directly affects the accuracy and reliability of test data. Traditional wind tunnel monitoring methods mainly rely on the threshold judgment of single sensor data and cannot comprehensively reflect the health state of the wind tunnel system. Currently, the following technical means are mainly adopted for wind tunnel health monitoring:

[0003] The first category is the fault diagnosis method based on expert experience. This method establishes an expert knowledge base and formulates fault judgment rules in combination with historical operation data. However, due to the complexity of the wind tunnel system, it is difficult to cope with changing operating conditions solely relying on empirical rules, and potential faults cannot be detected in a timely manner.

[0004] The second category is the single parameter monitoring method. Sensors such as temperature, pressure, and vibration are arranged at key positions to monitor the operating parameters of the equipment in real time. This method is simple to operate, but there is an information island phenomenon, and the correlation relationship between parameters cannot be mined, resulting in one-sided monitoring results.

[0005] The third category is the statistical analysis method. Statistical means are used to analyze historical data and establish a model of parameter change rules. However, this type of method has high requirements for data quality and is difficult to adapt to the dynamic change characteristics of wind tunnel test conditions.

[0006] Due to the extremely high real-time requirement and large amount of wind tunnel monitoring data, the above three methods all have the technical problem that they cannot quickly and efficiently mine and analyze wind tunnel health detection data from both steady-state and dynamic aspects, resulting in the inability to better discover potential fault signs. Summary of the Invention

[0007] In view of this, the present invention provides a method, medium and system for mining wind tunnel health monitoring data, which can solve the technical problem that the prior art cannot quickly and efficiently mine and analyze wind tunnel health detection data from both steady-state and dynamic aspects, resulting in the inability to better discover potential fault signs.

[0008] The present invention is implemented as follows:

[0009] The first aspect of the present invention provides a method for mining wind tunnel health monitoring data, including collecting multiple sensing signals to construct a monitoring data stream, performing filtering and noise reduction processing on the monitoring data stream to obtain preprocessed data and an abnormal data set, starting the GPU and CPU processing streams to receive the preprocessed data and the abnormal data set, performing component decomposition to obtain a steady-state data matrix, a dynamic data matrix, an abnormal feature matrix, a parameter data vector, and an abnormal parameter vector, establishing a data mining feature equation set based on the steady-state data matrix, the dynamic data matrix, and the abnormal feature matrix, calculating the wind tunnel operation state evaluation result by using the wind tunnel operation state evaluation equation set, generating a wind tunnel operation feature matrix and a wind tunnel state warning index, and outputting a warning signal according to the comparison result between the wind tunnel state warning index and a preset health threshold.

[0010] On the basis of the above technical solution, the method for mining wind tunnel health monitoring data of the present invention can be further improved as follows:

[0011] Among them, the multiple sensing signals include a temperature sensing signal, a vibration sensing signal, a pressure sensing signal, a sound sensing signal, a current sensing signal, a rotational speed sensing signal, a power sensing signal, a flow sensing signal, and a noise sensing signal.

[0012] Further, the CUDA parallel processing stream of the GPU receives the preprocessed data; the input stream of the CPU receives the abnormal data set.

[0013] Further, the preprocessed data is subjected to component decomposition to obtain the steady-state data matrix, the dynamic data matrix, and the parameter data vector; the abnormal data set is subjected to component decomposition to obtain the abnormal feature matrix and the abnormal parameter vector.

[0014] Further, the wind tunnel operation state evaluation equation set includes a state feature equation, a dynamic correlation equation, a steady-state mapping equation, and an abnormal influence equation.

[0015] Further, the inputs of the state feature equation include the temperature sensing signal and the pressure sensing signal in the steady-state data matrix, the dynamic data matrix, and the parameter data vector, and the output is the operation state feature vector in the wind tunnel operation state evaluation result; the inputs of the dynamic correlation equation include the rotational speed sensing signal and the flow sensing signal in the dynamic data matrix, the abnormal feature matrix, and the parameter data vector, and the output is the operation parameter dynamic feature vector in the wind tunnel operation state evaluation result.

[0016] Further, the input of the steady-state mapping equation includes the power sensing signal and current sensing signal in the steady-state data matrix and the parameter data vector, and the output is the steady-state feature vector of the operating parameters in the wind tunnel operating state evaluation result; the input of the abnormal influence equation includes the vibration sensing signal and noise sensing signal in the abnormal feature matrix and the abnormal parameter vector, and the output is the abnormal influence feature vector in the wind tunnel operating state evaluation result.

[0017] Further, the wind tunnel operation characteristic data is calculated by the wind tunnel operation state evaluation equation set to obtain the wind tunnel operation state evaluation result; the wind tunnel operation state evaluation result is used to generate the wind tunnel operation characteristic matrix and the wind tunnel state warning index.

[0018] The second aspect of the present invention provides a computer-readable storage medium, wherein program instructions are stored in the computer-readable storage medium, and when the program instructions run on a computer, they are used to execute the above-mentioned method for mining wind tunnel health monitoring data.

[0019] The third aspect of the present invention provides a system for mining wind tunnel health monitoring data, which includes the above-mentioned computer-readable storage medium.

[0020] Compared with the prior art, the beneficial effects of the method, medium and system for mining wind tunnel health monitoring data provided by the present invention are as follows:

[0021] First, through the parallel processing architecture of GPU and CPU cooperation, efficient processing of massive monitoring data is achieved. The CUDA parallel computing technology is used to process the preprocessed data, and at the same time, the CPU is used to process the abnormal data set, improving the data processing efficiency and laying a foundation for real-time monitoring.

[0022] Second, an innovative component decomposition method is proposed to decompose the monitoring data into a steady-state data matrix, a dynamic data matrix and a parameter data vector, realizing multi-dimensional characterization of the wind tunnel operation state. This decomposition method can effectively extract data features and provide a reliable basis for subsequent state evaluation.

[0023] Third, a complete set of wind tunnel operation state evaluation equations is constructed, including a state feature equation, a dynamic correlation equation, a steady-state mapping equation and an abnormal influence equation. These equations describe the wind tunnel operation characteristics from different angles and realize a comprehensive evaluation of the system state.

[0024] Fourth, an early warning mechanism based on multi-source data is established. Through the wind tunnel operation characteristic matrix and the state warning index, timely discovery and warning of abnormal states are realized. The early warning mechanism takes into account the characteristics of various sensing signals and their correlation relationships, improving the accuracy of early warning.

[0025] In summary, the present invention solves the technical problem that the prior art cannot quickly and efficiently mine and analyze wind tunnel health detection data from both steady-state and dynamic aspects, resulting in the inability to better discover potential fault signs. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flowchart of the method of the present invention;

[0027] Figure 2 is a characteristic distribution diagram of the steady-state data matrix in Embodiment 2;

[0028] Figure 3 is a change process diagram of the early warning index H over time in Embodiment 2. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0030] As Figure 1 shown, it is a flowchart of a method for mining wind tunnel health monitoring data provided by the present invention. This method includes the following steps: S10. Collect temperature sensing signals, vibration sensing signals, pressure sensing signals, sound sensing signals, current sensing signals, rotational speed sensing signals, power sensing signals, flow sensing signals, and noise sensing signals during the operation of the wind tunnel, and construct a monitoring data stream;

[0031] S20. Perform filtering and noise reduction processing on the monitoring data stream to obtain preprocessed data, and exclude abnormal band data in the preprocessed data during the filtering and noise reduction process, denoted as an abnormal data set;

[0032] S30. Start the CUDA parallel processing stream of the GPU to receive the preprocessed data; start the input stream of the CPU to receive the abnormal data set;

[0033] S40. Decompose the preprocessed data to obtain a steady-state data matrix, a dynamic data matrix, and a parameter data vector; decompose the abnormal data set to obtain an abnormal feature matrix and an abnormal parameter vector;

[0034] S50. Establish a data mining feature equation set based on the steady-state data matrix, the dynamic data matrix, and the abnormal feature matrix to obtain wind tunnel operation feature data;

[0035] S60. Use the wind tunnel operation feature data to calculate the wind tunnel operation status evaluation result based on the wind tunnel operation status evaluation equation set;

[0036] S70. Generate a wind tunnel operation characteristic matrix and a wind tunnel status warning index based on the evaluation result of the wind tunnel operation status;

[0037] S80. Output a warning signal according to the comparison result between the wind tunnel status warning index and the preset health threshold.

[0038] The wind tunnel operation status evaluation equation set includes a status characteristic equation, a dynamic correlation equation, a steady-state mapping equation, and an abnormal influence equation:

[0039] The status characteristic equation is used to establish the correlation relationship between the wind tunnel operation parameters. The inputs include the steady-state data matrix, the dynamic data matrix, and the temperature sensing signal and pressure sensing signal in the parameter data vector. The output is the operation status characteristic vector in the wind tunnel operation status evaluation result;

[0040] The dynamic correlation equation is used to analyze the dynamic change law of the wind tunnel operation parameters. The inputs include the dynamic data matrix, the abnormal characteristic matrix, and the rotational speed sensing signal and flow sensing signal in the parameter data vector. The output is the operation parameter dynamic characteristic vector in the wind tunnel operation status evaluation result;

[0041] The steady-state mapping equation is used to establish the steady-state characteristic mapping of the wind tunnel operation parameters. The inputs include the steady-state data matrix and the power sensing signal and current sensing signal in the parameter data vector. The output is the operation parameter steady-state characteristic vector in the wind tunnel operation status evaluation result;

[0042] The abnormal influence equation is used to evaluate the influence degree of abnormal data on the wind tunnel operation status. The inputs include the abnormal characteristic matrix and the vibration sensing signal and noise sensing signal in the abnormal parameter vector. The output is the abnormal influence characteristic vector in the wind tunnel operation status evaluation result.

[0043] The following describes the specific implementation manners of the above steps in detail:

[0044] The specific implementation manner of step S10 is as follows: First, a series of sensors installed at key positions of the wind tunnel, such as temperature sensors, vibration sensors, pressure sensors, sound sensors, current sensors, rotational speed sensors, power sensors, flow sensors, and noise sensors, are used to collect various monitoring data during the operation of the wind tunnel in real time. These monitoring data include temperature sensing signals, vibration sensing signals, pressure sensing signals, sound sensing signals, current sensing signals, rotational speed sensing signals, power sensing signals, flow sensing signals, and noise sensing signals, constituting a comprehensive monitoring data stream. The mathematical models of these sensing signals are as described in the claims and include various typical characteristics such as periodic changes, attenuation, and accumulation. By collecting these signal data, the operation status of the wind tunnel can be comprehensively reflected.

[0045] The specific implementation of step S20 is as follows: perform filtering and noise reduction on the monitored data stream collected in step S10. First, use a band-pass filter to filter the original sensor signal data to remove high-frequency noise and low-frequency drift, obtaining preprocessed data. During this process, a reasonable threshold also needs to be set to screen and eliminate the abnormal band data in the preprocessed data, forming an abnormal data set. The abnormal bands in the preprocessed data may indicate abnormal conditions in the wind tunnel and need to be separately extracted for analysis. The filtering and noise reduction threshold can be adjusted according to the data characteristics of different sensors, generally set to 1.5 - 2 times the normal value. Through this step, stable preprocessed data and the abnormal data set that needs to be analyzed key can be effectively extracted from the original monitored data.

[0046] The specific implementation of step S30 is as follows: First, start the CUDA parallel processing stream of the GPU to receive the preprocessed data after filtering and noise reduction; at the same time, start the input stream of the CPU to receive the selected abnormal data set. CUDA parallel processing can greatly improve the analysis and processing speed of a large amount of preprocessed data, while the serial processing of the CPU is more suitable for the individual analysis of the abnormal data set. Such an asynchronous parallel architecture can make full use of hardware resources and improve the overall computing efficiency of the system.

[0047] The specific implementation of step S40 is as follows: perform feature component decomposition on the preprocessed data and the abnormal data set respectively. For the preprocessed data, use the wavelet decomposition algorithm to decompose it into a steady-state data matrix, a dynamic data matrix, and a parameter data vector. Among them, the steady-state data matrix reflects the basic operating characteristics of the system, the dynamic data matrix reflects the instantaneous change characteristics of the system, and the parameter data vector contains characteristic quantities such as the reference values and amplitudes of each monitoring parameter. For the abnormal data set, use the same wavelet decomposition algorithm to decompose it into an abnormal feature matrix and an abnormal parameter vector, which reflect the characteristic quantities of the system under abnormal conditions. These feature components provide a basis for subsequent data mining and state assessment.

[0048] The specific implementation of step S50 is as follows: Based on the steady-state data matrix, dynamic data matrix, and abnormal feature matrix obtained in the previous step, a set of data mining feature equations are established. This set of equations includes a state feature equation, a dynamic correlation equation, a steady-state mapping equation, and an abnormal influence equation. The state feature equation describes the instantaneous change characteristics and cumulative effects of the system state; the dynamic correlation equation describes the dynamic coupling relationship between system parameters and introduces a second derivative term to reflect the acceleration change; the steady-state mapping equation describes the steady-state characteristics of system parameters and their relationship with temperature deviation; the abnormal influence equation describes the influence of abnormal vibration and noise on the system state and introduces a second derivative term of abnormal energy to reflect the mutation characteristics. This set of feature equations links various physical mechanisms contained in the monitoring data and provides a mathematical model basis for subsequent state evaluation.

[0049] The specific implementation of step S60 is as follows: Using the data mining feature equation set established in step S50 and combining with the steady-state data matrix, dynamic data matrix, and abnormal feature matrix obtained in step S40, the state evaluation result of the wind tunnel operation is calculated. The state evaluation result includes an operation state feature vector, an operation parameter dynamic feature vector, an operation parameter steady-state feature vector, and an abnormal influence feature vector. These feature vectors comprehensively describe the current operation state of the wind tunnel and provide a basis for subsequent state warning. The undetermined coefficients in the state evaluation equation set can be obtained through training and optimization using the particle swarm optimization algorithm with a complete data set collected under standard working conditions.

[0050] The specific implementation of step S70 is as follows: According to the wind tunnel operation state evaluation result calculated in step S60, two key outputs are generated: one is the wind tunnel operation feature matrix, which comprehensively reflects the steady-state characteristics, dynamic characteristics, and abnormal characteristics of various operation parameters of the wind tunnel; the other is the wind tunnel state warning index, which is a comprehensive warning index obtained by weighted synthesis of the operation state feature vector, dynamic feature vector, steady-state feature vector, and abnormal influence feature vector. These two outputs provide an important basis for subsequent state warning. The threshold of the wind tunnel state warning index can be initially set to 1.2 - 1.5 times the normal value according to engineering experience and can be further optimized and adjusted according to actual applications.

[0051] The specific implementation of step S80 is as follows: Compare the wind tunnel status warning index generated in step S70 with a pre-set health threshold. If the warning index exceeds the health threshold, then output the corresponding warning signal. The setting of the health threshold can refer to the following principles: For the operating state feature vector and the dynamic feature vector, it can be set to 1.2 - 1.5 times the normal value; for the steady-state feature vector, it can be set to 1.5 - 2 times the normal value; for the abnormal influence feature vector, it can be set to 2 - 3 times the normal value. These thresholds can be dynamically adjusted and optimized according to the actual operating conditions. Once the warning signal is output, it means that an abnormal situation has occurred in the wind tunnel, and corresponding maintenance measures need to be taken in a timely manner.

[0052] The following will describe in detail the calculation processes or equations involved in the present invention.

[0053] 1. The specific representations of the respective sensing signals are as follows:

[0054] ;

[0055] ;

[0056] ;

[0057] ;

[0058] ;

[0059] ;

[0060] ;

[0061] ;

[0062] ;

[0063] In the formula:

[0064] is the time variable;

[0065] is the reference value of each signal;

[0066] is the amplitude of each signal;

[0067] is the frequency of each signal;

[0068] is the phase angle;

[0069] is the temperature attenuation coefficient;

[0070] are the harmonic components of each signal;

[0071] is the rotational speed drift term;

[0072] is the voltage signal;

[0073] is the power factor;

[0074] is the noise random term;

[0075] Each parameter is obtained through a sensor calibration experiment.

[0076] 2. The steady-state data matrix is specifically expressed as follows:

[0077] ;

[0078] In the formula:

[0079] are the steady-state components of each signal;

[0080] is the sampling time point;

[0081] The steady-state components are obtained through wavelet decomposition.

[0082] 3. The dynamic data matrix is specifically expressed as follows:

[0083] ;

[0084] In the formula: is the dynamic component of each signal, obtained through wavelet decomposition.

[0085] 4. The abnormal feature matrix is specifically expressed as follows:

[0086] ;

[0087] In the formula:

[0088] is the abnormal vibration signal;

[0089] is the abnormal noise signal;

[0090] is the abnormal energy signal, ;

[0091] is the integration time window.

[0092] 5. The state characteristic equation is specifically expressed as follows:

[0093] ;

[0094] In the formula:

[0095] is the undetermined weight coefficient, determined by the particle swarm optimization algorithm;

[0096] is the error correction term;

[0097] This equation takes into account the instantaneous change characteristics and cumulative effects of the system state.

[0098] 6. The dynamic correlation equation is specifically expressed as follows:

[0099] ;

[0100] In the formula:

[0101] is the undetermined weight coefficient;

[0102] is the eigenvalue;

[0103] is the error correction term;

[0104] This equation introduces a second derivative term to describe the acceleration change of the system dynamic characteristics.

[0105] 7. The steady-state mapping equation is specifically expressed as follows:

[0106] ;

[0107] In the formula:

[0108] is the undetermined weight coefficient;

[0109] is the imaginary unit;

[0110] is the error correction term;

[0111] This equation takes into account the quadratic integral effect of the temperature deviation.

[0112] 8. The abnormal influence equation is specifically expressed as follows:

[0113] ;

[0114] In the formula:

[0115] is the weight coefficient to be determined;

[0116] is the error correction term;

[0117] The second derivative of the abnormal energy is introduced into this equation to reflect the mutation characteristics of the abnormal state.

[0118] All the coefficients to be determined are obtained through the following steps: collect a complete data set under the standard working conditions of the wind tunnel and conduct classification and annotation; decompose the signal using wavelet transform; optimize the coefficients using the particle swarm optimization algorithm; and verify the accuracy of the model using the cross-validation method.

[0119] Reasons for choosing the above functional relationships: use trigonometric functions to describe periodic change characteristics; use exponential functions to describe attenuation characteristics; introduce integral terms to reflect cumulative effects; introduce derivative terms to reflect change rates and accelerations; use complex exponentials to describe phase relationships; use logarithmic functions to compress the data range; and use quadratic terms to reflect non-linear characteristics.

[0120] The derivation process and parameter sources of each equation are explained in detail below.

[0121] 1. Derivation process of the original signal modeling:

[0122] Through the analysis of the wind tunnel operation mechanism, the composition of each sensing signal includes the following characteristics:

[0123] (1) Periodic change: use sine functions to describe the basic period characteristics;

[0124] (2) Attenuation characteristics: use exponential functions to describe;

[0125] (3) Harmonic components: use Fourier series expansion;

[0126] (4) Random perturbations: introduce noise terms;

[0127] For example, the temperature signal The derivation steps are as follows:

[0128] Step 1: Determine the reference temperature , which is obtained through calibration in the no-load test;

[0129] Step 2: Analyze the periodic change characteristics to obtain the term;

[0130] Step 3: Consider the temperature attenuation law and introduce the term;

[0131] Step 4: Determine the , , , parameter values through least squares fitting.

[0132] 2. Steady-state data matrix Derivation process:

[0133] Step 1: Perform wavelet transform on the original signal;

[0134] Step 2: Select the wavelet basis function and determine the decomposition level;

[0135] Step 3: Extract the low-frequency component as the steady-state feature;

[0136] Step 4: Construct the steady-state data matrix.

[0137] 3. State feature equation Derivation process:

[0138] Step 1: Based on the first law of thermodynamics, establish the energy balance equation;

[0139] Step 2: Introduce the time partial derivative term to describe the state change rate;

[0140] Step 3: Consider the integral effect of temperature and pressure;

[0141] Step 4: Add the error correction term.

[0142] Optimization effect: Can reflect the instantaneous change of the system state; Considers the energy accumulation effect; Has strong anti-interference ability.

[0143] 4. Dynamic correlation equation Derivation process:

[0144] Step 1: Based on the principle of fluid mechanics, establish the dynamic characteristic equation;

[0145] Step 2: Consider the quadratic term effect of rotational speed;

[0146] Step 3: Introduce the cubic term of flow rate to reflect the turbulence characteristics;

[0147] Step 4: Add the second derivative term to describe the acceleration change.

[0148] Parameter determination:

[0149] Obtained by optimizing through genetic algorithm;

[0150] 、 Optimized through particle swarm algorithm;

[0151] Determined through simulated annealing algorithm.

[0152] 5. Steady-state mapping equation Derivation process:

[0153] Step 1: Establish the power factor mapping relationship;

[0154] Step 2: Introduce complex exponential to describe the phase relationship;

[0155] Step 3: Add the second integral term of the temperature deviation;

[0156] Step 4: Optimize the error correction term.

[0157] 6. Abnormal influence equation Derivation process:

[0158] Step 1: Analyze the coupling mechanism of vibration and noise;

[0159] Step 2: Establish the abnormal energy calculation model;

[0160] Step 3: Introduce logarithmic compression and square root terms;

[0161] Step 4: Add the second derivative to reflect the mutation characteristics.

[0162] Parameter optimization:

[0163] (1) 、 Obtained through support vector regression;

[0164] (2) Determined through cross-validation;

[0165] (3) Determine the range through error analysis.

[0166] The existing methods for mining wind tunnel health monitoring data mainly have problems such as low processing efficiency and insufficient monitoring accuracy. Traditional monitoring methods often use single-parameter thresholds or basic statistical analysis methods for fault diagnosis, which cannot effectively reflect the complex coupling relationships among multiple parameters in the wind tunnel system. At the same time, due to the huge amount of data generated during the operation of the wind tunnel and the need for real-time processing and analysis, the traditional serial computing mode can no longer meet the requirements of real-time monitoring. In addition, neural networks or deep learning methods are generally used for state recognition in the existing technology. Although these methods have strong non-linear fitting capabilities, the training process is time-consuming, and the real-time computing complexity is high, making it difficult to meet the requirements of rapid response. The monitoring methods based on pattern recognition require a complete fault pattern library to be established in advance, have poor recognition ability for newly emerging fault patterns, and the feature extraction process has a large amount of calculation. More importantly, the existing technology lacks a special processing mechanism for abnormal data and often uses simple methods such as deletion or smoothing. This approach not only loses important fault feature information but also may lead to misjudgment.

[0167] The present invention proposes an innovative method for mining health monitoring data of a wind tunnel. By establishing a parallel processing architecture that coordinates GPU and CPU, efficient data processing is achieved. This method assigns computationally intensive temperature, vibration, pressure, and sound signals to the CUDA parallel stream processing of the GPU, while handing over control-related current and rotation speed signals to the CPU, giving full play to the advantages of heterogeneous computing systems. In terms of signal processing, the present invention establishes a complete mathematical model for multi-source signals, considering not only the periodicity, attenuation characteristics, harmonic components, and random perturbations of the signals, but also achieving precise separation of the steady-state components and dynamic components through wavelet transform. Especially in the processing of abnormal data, an abnormal energy evaluation model is innovatively established, improving the accuracy of abnormal detection.

[0168] The most remarkable feature of the present invention lies in the design of four core matrix calculation equations. The state characteristic equation, based on the principle of energy balance and by introducing a time partial derivative term, can accurately describe the dynamic change characteristics of the system state. The dynamic correlation equation integrates fluid mechanics characteristics, considers nonlinear effects, and effectively reflects the complex coupling relationship between system parameters. The steady-state mapping equation uses complex variable functions to describe the phase relationship, improving the accuracy of steady-state characteristic identification. The abnormal influence equation establishes a coupling model between vibration and noise, providing a reliable basis for abnormal state identification. These equations are implemented through matrix parallel calculation, greatly improving the calculation efficiency.

[0169] In terms of parameter optimization, the present invention uses a variety of intelligent optimization algorithms to determine the weight coefficients, introduces an adaptive mechanism to dynamically adjust the parameters, and establishes an error correction mechanism to improve the calculation accuracy. This method not only improves the early warning accuracy rate of the system, but also significantly reduces the false alarm rate. Through the efficient data processing achieved by matrix calculation, memory usage is optimized by using sparse matrix storage, parallel efficiency is improved by matrix block calculation, and data transmission overhead is reduced by pipeline processing. The establishment of a multi-level early warning mechanism enables the system to monitor the changing trend of abnormal energy in real time, dynamically evaluate the evolution direction of the system state, and timely discover potential fault hazards.

[0170] The present invention demonstrates remarkable technical effects in practical applications. Through the coordinated processing of GPU and CPU and optimized matrix calculation, the data processing speed is greatly improved, and the system response time is significantly shortened. The fault mode recognition method based on multi-source information fusion improves the early warning accuracy rate of the system and reduces the false alarm rate. The abnormal data processing mechanism improves the anti-interference ability of the system, multiple error corrections ensure the reliability of the calculation results, and adaptive parameter adjustment enhances the adaptability of the system. The equations have clear physical meanings, facilitating fault diagnosis, and the modular design and standardized interfaces endow the system with good maintainability and scalability. These improvements enable the present invention to effectively solve the real-time and accuracy problems in wind tunnel health monitoring and have important engineering application values.

[0171] The second aspect of the present invention provides a computer-readable storage medium, in which program instructions are stored, and when the program instructions run on a computer, they are used to execute the above-mentioned method for mining wind tunnel health monitoring data.

[0172] The third aspect of the present invention provides a system for mining wind tunnel health monitoring data, which includes the above-mentioned computer-readable storage medium.

[0173] Specifically, the principle of the present invention is as follows: At the data acquisition level, a comprehensive data acquisition network is constructed by arranging various sensors such as temperature, vibration, and pressure. This multi-source data acquisition method conforms to the complexity characteristics of the wind tunnel system and can comprehensively reflect the system operation status. The arrangement positions and sampling frequencies of various sensors are determined according to the characteristics of the wind tunnel to ensure the representativeness and integrity of the data.

[0174] At the data preprocessing level, filtering and noise reduction techniques are used to process the original data, and at the same time, abnormal data is identified and separated. This processing method is based on signal processing theory, and through frequency domain analysis and statistical feature extraction, the data quality is effectively improved. The separation of abnormal data is based on statistical significance tests to ensure the accurate extraction of abnormal characteristics.

[0175] At the parallel computing level, the preprocessed data is processed based on the CUDA architecture of the GPU, and at the same time, the CPU is used to process the abnormal data set. This division of labor and cooperation mode makes full use of the hardware resources and improves the computing efficiency. The GPU is suitable for processing large-scale matrix operations, while the CPU is more suitable for processing abnormal data analysis with complex logic.

[0176] At the feature extraction level, through the component decomposition method, the data is decomposed into steady-state, dynamic, and parameter features. This decomposition is based on time series analysis theory and can effectively capture different operation characteristics of the system. The steady-state features reflect the basic state of the system, the dynamic features reflect the change law of the system, and the parameter features reflect the key indicators of the system.

[0177] At the state evaluation level, an evaluation system including four core equations is constructed. The state feature equation is based on parameter correlation analysis, the dynamic correlation equation is based on time series pattern recognition, the steady-state mapping equation is based on feature mapping theory, and the abnormal influence equation is based on abnormal detection principle. These equations complement each other and jointly constitute a complete evaluation framework.

[0178] The technical solution of the present invention has strict logic and scientific nature. Each processing link is connected to form a complete data mining system, which can effectively solve the problem of insufficient data mining in wind tunnel health monitoring.

[0179] A specific Embodiment 1 of the present invention is provided below. The specific implementation of each step in this Embodiment 1 is described in detail as follows: The specific implementation of step S10 is: First, by installing a series of sensors at key positions of the wind tunnel, including temperature sensors, vibration sensors, pressure sensors, sound sensors, current sensors, rotational speed sensors, power sensors, flow sensors, and noise sensors, various monitoring data during the operation of the wind tunnel are collected in real time. These monitoring data include temperature sensing signals , vibration sensing signals , pressure sensing signals , sound sensing signals , current sensing signals , rotational speed sensing signals , power sensing signals , flow sensing signals , and noise sensing signals . These sensing signals can be described by the following mathematical models:

[0180] ;

[0181] ;

[0182] ;

[0183] ;

[0184] ;

[0185] ;

[0186] ;

[0187] ;

[0188] ;

[0189] Among them, is the time variable; is the reference value of each signal; are the amplitudes of each signal; is the frequency of each signal; is the phase angle; is the temperature attenuation coefficient; are the harmonic components of each signal; is the rotational speed drift term; is the voltage signal; is the power factor; is the noise random term. These parameters can be obtained through sensor calibration experiments. By collecting these signal data, the operating conditions of the wind tunnel can be comprehensively reflected.

[0190] The specific implementation of step S20 is as follows: First, a digital filter is used to filter and denoise the monitoring data collected in step S10. Specifically, a band-pass filter is used to filter the original sensor signal data to remove high-frequency noise and low-frequency drift, obtaining preprocessed data. During this process, a reasonable threshold also needs to be set to screen and eliminate abnormal band data in the preprocessed data, forming an abnormal data set. The abnormal bands in the preprocessed data may indicate the fault conditions existing in the wind tunnel and need to be separately extracted for analysis. The threshold for filtering and denoising can be adjusted according to the data characteristics of different sensors, generally set to 1.5 - 2 times the normal value. Through this step, stable preprocessed data and the abnormal data set that needs to be analyzed key can be effectively extracted from the original monitoring data.

[0191] The specific implementation of step S30 is as follows: First, start the CUDA parallel processing stream of the GPU to receive the preprocessed data after filtering and denoising; at the same time, start the input stream of the CPU to receive the screened abnormal data set. The CUDA parallel processing stream can greatly improve the analysis and processing speed of a large amount of preprocessed data, while the serial input stream of the CPU is more suitable for analyzing the abnormal data set one by one. Such an asynchronous parallel architecture can make full use of hardware resources and improve the overall computing efficiency of the system.

[0192] The specific implementation of step S40 is as follows: Feature component decomposition is performed on the preprocessed data and the abnormal data set respectively. For the preprocessed data, the Discrete Wavelet Transform (DWT) algorithm is used to decompose it into a steady-state data matrix , a dynamic data matrix and a parameter data vector. Among them, the steady-state data matrix reflects the basic operating characteristics of the system and can be expressed as:

[0193] ;

[0194] The dynamic data matrix reflects the instantaneous change characteristics of the system and can be expressed as:

[0195] ;

[0196] The parameter data vector contains characteristic quantities such as the reference values and amplitudes of each monitoring parameter. For the abnormal data set, the same DWT algorithm is used to decompose it into an abnormal feature matrix and an abnormal parameter vector, which can be expressed as:

[0197] ;

[0198] Among them, is the abnormal vibration signal, is the abnormal noise signal, is the abnormal energy signal, is the integration time window. These characteristic components provide a basis for subsequent data mining and state evaluation.

[0199] The specific implementation of step S50 is: based on the steady-state data matrix , dynamic data matrix and abnormal characteristic matrix obtained in the previous step, a set of data mining characteristic equations are established. This set of equations includes:

[0200] State characteristic equation:

[0201] ;

[0202] Among them, is the undetermined weight coefficient, is the error correction term. This equation takes into account the instantaneous change characteristics and cumulative effects of the system state.

[0203] Dynamic correlation equation:

[0204] ;

[0205] Among them, is the undetermined weight coefficient, is the eigenvalue, is the error correction term. This equation introduces a second derivative term to describe the acceleration change of the system dynamic characteristics.

[0206] Steady-state mapping equation:

[0207] ;

[0208] Among them, is the undetermined weight coefficient, is the imaginary unit, is the error correction term. This equation takes into account the quadratic integral effect of temperature deviation.

[0209] Abnormal influence equation:

[0210] ;

[0211] Among them, is the undetermined weight coefficient, is the error correction term. This equation introduces the second derivative of abnormal energy to reflect the mutation characteristics of abnormal states.

[0212] This set of characteristic equations relates various physical mechanisms contained in the monitoring data and provides a mathematical model basis for subsequent state assessment. All undetermined coefficients can be obtained through training and optimization using the particle swarm optimization algorithm with a complete data set collected under standard operating conditions.

[0213] The specific implementation of step S60 is: using the data mining characteristic equation set established in step S50, combined with the steady-state data matrix obtained in step S40 , the dynamic data matrix and the abnormal feature matrix , calculate the state assessment result of the wind tunnel operation. The state assessment result includes:

[0214] The operation state feature vector ;

[0215] The dynamic feature vector of operation parameters ;

[0216] The steady-state feature vector of operation parameters ;

[0217] The abnormal influence feature vector ;

[0218] These feature vectors comprehensively describe the current operation state of the wind tunnel and provide a basis for subsequent state warning.

[0219] The specific implementation of step S70 is: according to the wind tunnel operation state assessment result calculated in step S60, generate two key outputs:

[0220] One is the wind tunnel operation feature matrix, which comprehensively reflects the steady-state features, dynamic features and abnormal features of various operation parameters of the wind tunnel;

[0221] The other is the wind tunnel state warning index, which is a comprehensive warning index obtained by weighted synthesis of the operation state feature vector , the dynamic feature vector , the steady-state feature vector and the abnormal influence feature vector .

[0222] The threshold of the wind tunnel state warning index can be initially set to 1.2 - 1.5 times the normal value according to engineering experience and can be further optimized and adjusted according to actual applications.

[0223] The specific implementation of step S80 is: compare the wind tunnel state warning index generated in step S70 with a pre-set health threshold. If the warning index exceeds the health threshold, output the corresponding warning signal. The setting of the health threshold can refer to the following principles:

[0224] For the operating state feature vector and the dynamic feature vector , it can be set to 1.2 to 1.5 times the normal value;

[0225] For the steady-state feature vector , it can be set to 1.5 to 2 times the normal value;

[0226] For the abnormal influence feature vector , it can be set to 2 to 3 times the normal value.

[0227] These thresholds can be dynamically adjusted and optimized according to the actual operating conditions. Once the warning signal is output, it means that the wind tunnel has an abnormal condition and corresponding maintenance measures need to be taken in a timely manner.

[0228] To better understand and implement the present invention, the following provides Example 2 of a specific application scenario of the present invention: A certain aviation research institute is conducting an aerodynamic performance test on a new type of aircraft and needs to use the large-scale parallel-flow wind tunnel of the institute for the test. In view of the complex structure and harsh working environment of the new aircraft, the researchers decided to adopt the wind tunnel health monitoring data mining method proposed by the present invention to comprehensively monitor and intelligently warn the operating conditions of the wind tunnel.

[0229] First, temperature sensors, vibration sensors, pressure sensors, sound sensors, current sensors, rotational speed sensors, power sensors, flow sensors, and noise sensors were respectively installed at key positions such as the inlet section, test section, and outlet section of the wind tunnel. These sensors collected and output the following monitoring signals in real time:

[0230] Temperature sensing signal :

[0231] ;

[0232] Among them, is the reference temperature, is the temperature amplitude, is the temperature change frequency, is the temperature attenuation coefficient, is the temperature attenuation coefficient.

[0233] Vibration sensing signal :

[0234] ;

[0235] Among them, is the amplitude, is the vibration frequency, is the amplitude, is Sub - harmonic component

[0236] Pressure sensing signal :

[0237] ;

[0238] Wherein, is the reference pressure, is the pressure amplitude, is the pressure change frequency, is Sub - harmonic component

[0239] Sound sensing signal :

[0240] ;

[0241] Wherein, is the amplitude of the th infrasonic component, is the frequency of the th infrasonic component, is the phase of the th infrasonic component.

[0242] Current sensing signal :

[0243] ;

[0244] Wherein, is the reference current, is the current amplitude, is the current change frequency, is Sub - harmonic component

[0245] Rotational speed sensing signal :

[0246] ;

[0247] Wherein, is the reference rotational speed, is the rotational speed amplitude, is the rotational speed change frequency, is the rotational speed drift term.

[0248] Power sensing signal :

[0249] ;

[0250] Wherein, is the voltage signal, is the power factor.

[0251] Flow sensing signal :

[0252] ;

[0253] where, is the reference flow rate, is the flow amplitude, is the flow change frequency, is the

[0254] Noise sensing signal :

[0255] ;

[0256] where, is the amplitude of the th noise component, is the frequency of the th noise component, is the phase of the th noise component, is the random noise term.

[0257] The parameters of the above-mentioned sensing signals are all determined in advance through calibration experiments.

[0258] After collecting about 10 minutes of initial monitoring data, the researchers initiated the data preprocessing process. First, a band-pass filter was used to filter each sensing signal to remove high-frequency noise and low-frequency drift, obtaining preprocessed data. Then, an abnormal band detection was performed on the preprocessed data, setting the threshold to 1.5 times the normal value, and the data segments exceeding the threshold were excluded to form an abnormal data set. At this time, the preprocessed data and the abnormal data set were ready for subsequent feature decomposition and data mining analysis.

[0259] Next, the researchers initiated the CUDA parallel processing stream of the GPU to perform feature decomposition on the preprocessed data; at the same time, they initiated the serial input stream of the CPU to perform feature decomposition on the abnormal data set. Using the discrete wavelet transform (DWT) algorithm, the preprocessed data was decomposed into a steady-state data matrix and a dynamic data matrix , specifically as follows:

[0260] Steady-state data matrix :

[0261] ;

[0262] Dynamic data matrix :

[0263] ;

[0264] For the abnormal data set, the DWT algorithm obtains the abnormal feature matrix :

[0265] ;

[0266] Among them, is the abnormal vibration signal, is the abnormal noise signal, is the abnormal energy signal. As Figure 2 shown, the characteristic distribution of the steady-state data matrix is displayed through a three-dimensional surface, reflecting the correlation between monitoring parameters.

[0267] With these characteristic components, the researchers began to establish a set of data mining characteristic equations proposed by the present invention.

[0268] First is the state characteristic equation:

[0269] ;

[0270] Among them, is the undetermined weight coefficient, is the error correction term. is the integration time window. This equation considers the instantaneous change characteristics and cumulative effect of the system state.

[0271] Dynamic correlation equation:

[0272] ;

[0273] Among them, is the undetermined weight coefficient, is the eigenvalue, is the error correction term. This equation introduces a second derivative term to describe the acceleration change of the system dynamic characteristics.

[0274] Steady-state mapping equation:

[0275] ;

[0276] Among them, is the undetermined weight coefficient, is the imaginary unit, is the error correction term. This equation considers the quadratic integral effect of temperature deviation.

[0277] Abnormal influence equation:

[0278] ;

[0279] Among them, is the weight coefficient to be determined, is the error correction term. The second derivative of the abnormal energy is introduced in this equation to reflect the mutation characteristics of the abnormal state.

[0280] Through calculation, the researchers obtained the evaluation results of various states of the wind tunnel operation:

[0281] Operation state eigenvector ;

[0282] Dynamic characteristic vector of operation parameters ;

[0283] Steady-state characteristic vector of operation parameters ;

[0284] Abnormal influence eigenvector ;

[0285] By weighting and synthesizing these eigenvectors, the wind tunnel state warning index is obtained. The preset health threshold is , so the current warning index exceeds the health threshold. As Figure 3 shown, it shows the change process of the warning index H over time, and marks the health threshold line, clearly showing the over-threshold alarm area.

[0286] The researchers immediately checked the wind tunnel operation characteristic matrix and found that there were large abnormalities in parameters such as temperature, rotational speed, and noise. Through further analysis of the abnormal characteristic matrix , a certain vibration sensor in the outlet section of the wind tunnel was located to have a fault, resulting in abnormal phenomena such as temperature increase, rotational speed fluctuation, and noise intensification.

[0287] Based on the above analysis results, the researchers immediately stopped the aerodynamic test and arranged for maintenance personnel to check and repair the vibration sensor in the outlet section. After a short shutdown, after the wind tunnel returned to the normal operation state, the aerodynamic test was successfully carried out.

[0288] This practice fully verified the effectiveness of the wind tunnel health monitoring data mining method proposed by the present invention. Through comprehensive analysis of multi-source monitoring data, this method can not only timely detect the abnormal conditions of the wind tunnel and accurately locate the fault location, but also provide a strong guarantee for the safe operation of the wind tunnel by generating reliable state warning indicators. Compared with the traditional single-sensing signal fault diagnosis, this intelligent warning system based on big data analysis has higher accuracy and robustness.

[0289] It should be noted that the detailed explanations of the variables involved in the present invention are shown in Table 1 below.

[0290] Table 1 Variable Explanation Table

[0291]

[0292] As described above, this is only a specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention.

Claims

1. A method for mining health monitoring data of a wind tunnel, characterized in that, It includes collecting multiple sensing signals to construct a monitoring data stream, performing filtering and noise reduction processing on the monitoring data stream to obtain preprocessed data and an abnormal data set, starting the GPU and CPU processing streams to receive the preprocessed data and the abnormal data set, performing component decomposition to obtain a steady-state data matrix, a dynamic data matrix, an abnormal feature matrix, a parameter data vector, and an abnormal parameter vector, establishing a wind tunnel operating state evaluation equation set based on the steady-state data matrix, the dynamic data matrix, and the abnormal feature matrix, calculating a wind tunnel operating state evaluation result using the wind tunnel operating state evaluation equation set, generating a wind tunnel operation feature matrix and a wind tunnel state warning index, and outputting a warning signal according to the comparison result between the wind tunnel state warning index and a preset health threshold; The wind tunnel operating state evaluation equation set includes: a state feature equation, the input of which includes temperature sensing signals, pressure sensing signals in the steady-state data matrix, the dynamic data matrix, and the parameter data vector, and the output of which is an operating state feature vector in the wind tunnel operating state evaluation result; a dynamic correlation equation, the input of which includes rotational speed sensing signals, flow sensing signals in the dynamic data matrix, the abnormal feature matrix, and the parameter data vector, and the output of which is an operating parameter dynamic feature vector in the wind tunnel operating state evaluation result; a steady-state mapping equation, the input of which includes power sensing signals, current sensing signals in the steady-state data matrix and the parameter data vector, and the output of which is an operating parameter steady-state feature vector in the wind tunnel operating state evaluation result; an abnormal influence equation, the input of which includes vibration sensing signals, noise sensing signals in the abnormal feature matrix and the abnormal parameter vector, and the output of which is an abnormal influence feature vector in the wind tunnel operating state evaluation result.

2. The mining method of wind tunnel health monitoring data according to claim 1, characterized in that, The multiple sensing signals include temperature sensing signals, vibration sensing signals, pressure sensing signals, sound sensing signals, current sensing signals, rotational speed sensing signals, power sensing signals, flow sensing signals, and noise sensing signals.

3. The mining method of wind tunnel health monitoring data according to claim 2, characterized in that The CUDA parallel processing stream of the GPU receives the preprocessed data; the input stream of the CPU receives the abnormal data set.

4. A method for mining wind tunnel health monitoring data according to claim 3, characterized in that The preprocessed data is decomposed by components to obtain the steady-state data matrix, the dynamic data matrix, and the parameter data vector; the abnormal data set is decomposed by components to obtain the abnormal feature matrix and the abnormal parameter vector.

5. A method for mining health monitoring data of a wind tunnel according to claim 4, characterized in that, The wind tunnel operation feature data is calculated by the wind tunnel operating state evaluation equation set to obtain the wind tunnel operating state evaluation result; the wind tunnel operating state evaluation result is used to generate the wind tunnel operation feature matrix and the wind tunnel state warning index.

6. A computer-readable storage medium, characterized in that, Program instructions are stored in the computer-readable storage medium, and when the program instructions run on a computer, they are used to execute a method for mining wind tunnel health monitoring data according to any one of claims 1-5.

7. A mining system for wind tunnel health monitoring data, characterized in that, It includes the computer-readable storage medium according to claim 6.

Images