Aero-engine fault classification method based on sparrow algorithm and hierarchical clustering
By combining sparrow algorithm and hierarchical clustering method, the classification of aircraft engine faults is optimized, and the problems of poor noise data processing capabilities and high computational complexity in the existing technology are solved, more efficient fault feature recognition and classification are achieved, and fault detection is improved accuracy and reliability.
Patent Information
- Application Number
- CN202411119739.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-08-15
AI Technical Summary
The existing aircraft engine fault diagnosis technology has problems such as poor processing capabilities, high missed detection rate, high computational complexity, difficulty in determining parameters and insufficient interpretation, making it difficult to effectively identify fault characteristics, affecting the accuracy and efficiency of fault classification.
Using a method based on sparrow algorithm and hierarchical clustering, the original data matrix is constructed by selecting aircraft engine state evaluation indicators, normalized preprocessing and clustering analysis are carried out, and combined with sparrow algorithm optimization, the optimal fault classification results are iteratively obtained.
It improves the accuracy and efficiency of fault detection, reduces the possibility of missed detection and false alarms, enhances the reliability of fault classification, and provides strong technical support for the maintenance and health management of aircraft engines.
Smart Images

Figure CN119066514A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to aircraft engine fault classification, in particular to an aircraft engine fault classification method based on sparrow algorithm and hierarchical clustering. Background Art
[0002] In the earliest days of aircraft engine maintenance, only when there was a problem with the working condition of the engine or when it was returned to the factory for overhaul could we know whether some parts inside the engine had failed. In order to ensure the working safety of aircraft engines and the flight safety of aircraft, the maintenance of engines has long adopted a regular maintenance and inspection system. That is, after the engine has been used for a period of time, no matter what the condition is, the engine is regularly repaired. In this way, the length of the repair time directly affects the safety and economic efficiency of the engine. If the time is too long, the engine failure rate will increase, and even cause accidents, resulting in significant economic and social losses; if the time is too short, unnecessary replacement of parts (components) will be caused due to maintenance, which is neither economical nor reasonable. Therefore, in order to extend the life of the engine as much as possible under the premise of safety and reliability, it is necessary to strengthen the research and application of fault diagnosis technology with the goal of condition-based maintenance.
[0003] Since typical modern systems have a large number of instruments and meters, large systems will generate a large amount of data. Although a lot of information can be obtained from these processes, it is difficult for operators or engineers to directly evaluate the process operation status from the observed data. The advantage of multivariate statistical analysis is that it can analyze noisy and highly correlated measurement data, compress high-dimensional data into low-dimensional subspaces, and obtain important information from them.
[0004] Multivariate statistical analysis methods include principal component analysis (PCA), kernel principal component analysis (Kernel Principal Components Analysis), independent principal component analysis (Independent Principal Component Analysis), partial least squares (PLS) and Fisher classification analysis (FDA). Although they are widely used in dimension reduction, feature extraction and pattern recognition, they each have some limitations. These shortcomings mainly include restrictions on assumptions about data distribution (such as the assumption of linear structure in PCA and FDA), sensitivity to outliers, difficulty in determining optimal parameters (such as the number of principal components), computational complexity (especially on large-scale data sets), and limited ability to handle nonlinear relationships. In addition, these methods may require a lot of prior knowledge to select appropriate models and parameters, and in some cases, the interpretability of the model may be insufficient. Summary of the invention
[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide an aircraft engine fault classification method based on the sparrow algorithm and hierarchical clustering, which can effectively identify fault characteristics from complex operating data, improve the reliability of fault classification, and reduce the possibility of missed detection and false alarm.
[0006] The object of the present invention is achieved through the following technical scheme: a method for classifying aircraft engine faults based on sparrow algorithm and hierarchical clustering, comprising the following steps:
[0007] S1. Selecting evaluation indicators that characterize the state of the aircraft engine;
[0008] S2. Obtain the original data of various indicators of multiple fault types of aircraft engines as samples and construct an original data matrix;
[0009] S3. performing normalization preprocessing on the original data matrix to obtain preprocessed data;
[0010] S4. Perform cluster analysis on the processed data and classify the preprocessed data using a hierarchical clustering method;
[0011] S5. Optimize the clustering process using the sparrow algorithm and iterate to obtain the optimal fault classification effect.
[0012] The beneficial effects of the present invention are as follows: the present invention combines hierarchical clustering and sparrow algorithm optimization to improve the accuracy and efficiency of fault detection. The present invention can effectively identify fault characteristics from complex operating data, improve the reliability of fault classification, reduce the possibility of missed detection and false alarm, and provide strong technical support for the maintenance and health management of aircraft engines. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] Figure 1 is a flow chart of the method of the present invention;
[0014] Figure 2 To solve the parameter matrix flow chart;
[0015] Figure 3 It is the hierarchical clustering process diagram;
[0016] Figure 4 To optimize the clustering result graph;
[0017] Figure 5 is the silhouette coefficient plot. DETAILED DESCRIPTION
[0018] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings, but the protection scope of the present invention is not limited to the following.
[0019] Since aircraft engines may contain very strong noise during operation, the ability of basic statistical analysis to extract features is relatively poor and the missed detection rate is relatively high in processes containing strong noise. This method combines hierarchical clustering and sparrow algorithm optimization to improve the accuracy and efficiency of fault detection. The present invention can effectively identify fault features from complex operating data, improve the reliability of fault classification, reduce the possibility of missed detection and false alarm, and provide strong technical support for the maintenance and health management of aircraft engines.
[0020] like Figure 1 As shown, an aircraft engine fault classification method based on sparrow algorithm and hierarchical clustering includes the following steps:
[0021] S1. Selecting evaluation indicators that characterize the state of the aircraft engine;
[0022] Engine parameter matrix establishment
[0023] In the embodiments of the present application, in order to understand the various parameters that affect the healthy operation status of the aircraft engine throughout its entire cycle, it is necessary to conduct a systematic study of the aircraft engine system. However, obtaining the operating parameters of the aircraft engine through experimental methods is costly and time-consuming, which is not conducive to further in-depth research. Therefore, it is necessary to establish a suitable aircraft engine model to obtain the parameters involved in the operation of the aircraft engine and the related parameter matrix from the model. The aircraft engine model provides an economical and efficient tool that can better understand the operating mechanism of the aircraft engine and help identify the key factors that affect the healthy status of the engine. By establishing such a model, the present invention can effectively reduce research costs, shorten research cycles, and provide a reliable basis for fault classification of aircraft engines.
[0024] Establishment of aircraft engine model: Based on the volumetric method and component characteristics, a dynamic mathematical model of a certain type of high bypass ratio turbofan engine is established, which can describe the real dynamic characteristics of the aircraft engine. The turbofan aircraft engine object selects a certain type of high bypass ratio dual-shaft separate exhaust turbofan engine. The establishment of the dynamic model of the engine object adopts the modeling idea based on component characteristics and volumetric method. The engine object components include atmospheric conditions, inlet duct components, fan components, outer duct components, low-pressure compressor, compressor components, combustion chamber components, high-pressure turbine components, low-pressure turbine components, mixing chamber, tail nozzle (including inner nozzle and outer duct nozzle), high-pressure rotor components, and low-pressure rotor components. The whole process realizes the establishment of the aircraft engine model through a series of gas path, aerodynamic thermodynamics and rotor dynamics calculation methods.
[0025] Linearized parameter matrix: The engine model is the basis for establishing the engine parameter matrix. Corresponding to each environmental parameter point and operating condition, in the embodiment of the present application, the linearized parameters under this condition can be solved by linearization, and a linearized parameter matrix set can be established by this method. The specific process of solving the parameter matrix is as follows: Figure 2 .
[0026] Design point parameters such as N1, N2 speed and design point fuel supply rate enter the engine model through the linearization module. At the same time, external environmental parameters (altitude, Mach number, temperature) also enter the engine model. After the engine model runs through a series of inputs, all dependent variables are output and enter the Newton-Raphson solver at the same time, by configuring the initial conditions and parameter selection conditions of the solver. Through a series of iterative cycles, when the error is less than the error set by the solver, all independent variables are output and enter the engine model. At the same time, this process converts the nonlinear engine model into a linear model. At the same time, the module composed of the engine model and the solver inputs data into the linearization module. The linearization module outputs a series of parameters related to the health status of the engine, including temperature, pressure, speed, stall margin, and fuel parameters. The parameters are processed to form a parameter matrix.
[0027] The aircraft engine component-level model obtained by the component method and the volume method has high accuracy, but the aircraft engine it represents is essentially a system with strong nonlinearity. In order to obtain the linearized parameter matrix of the aircraft engine, it is necessary to use a solver to obtain a specific matrix. Here, the widely used Newton-Raphson theory is used to solve the parameter matrix.
[0028] The Newton-Raphson method is an iterative numerical method for solving nonlinear equations with numerical approximation. The basic idea is to approximate the root of the equation through continuous iteration. Assume that the equation to be solved is f(x) = 0, select an initial approximate solution x0, and then iterate to make f(x i ) approaches 0. In each iteration, by i The tangent is used to update the approximate solution until the convergence condition is met. The Newton-Raphson method has quadratic convergence, and the error at each step decreases at a quadratic rate. The convergence speed depends on the choice of the initial approximate solution and the properties of the equation f(x). In some cases, the convergence speed may be slow or non-convergence may occur, and adjustments or other methods may be required.
[0029] In fact, in this application, as long as the evaluation indicators involved in the engine health status evaluation can be given in advance, the evaluation indicator data can be measured, simulated, or obtained according to the above method. The specific acquisition method is not limited. The key lies in the selection of the evaluation indicators and the specific evaluation process, specifically:
[0030] Determination of aircraft engine health status parameter set
[0031] According to the establishment of the above-mentioned aircraft engine parameter matrix, by establishing a suitable aircraft engine performance calculation process, the present invention can obtain the parameters involved in the full-cycle operation of the aircraft engine from the model, and can simulate the full-cycle operation of the real aircraft engine by changing the input conditions of the performance calculation, and obtain the required related parameters by the method of linearizing the parameter matrix. The aircraft engine performance calculation and related parameters provide a reliable basis for the aircraft engine fault classification method. Next, the present invention constructs an aircraft engine health status parameter set for monitoring and evaluating the performance and status of the aircraft engine, and making auxiliary strategies for the next step of operation, maintenance, and decision-making.
[0032] Aircraft engine health status parameters are key indicators used to evaluate and monitor engine status and performance. They provide important information about the engine's operating conditions, helping to identify potential problems in a timely manner and take preventive or repair measures to ensure safe, reliable and efficient operation of the engine. The aircraft engine health status parameter set is determined by defining the following health status parameters:
[0033] (1) Temperature health status parameters
[0034] There are many temperature parameters in aircraft engines, including turbine inlet temperature (TIT), turbine outlet temperature (TOT), combustion chamber temperature, etc. These parameters can reflect the operating temperature of various components inside the engine and are important indicators for evaluating the performance of aircraft engines. Temperature parameters reflect the thermal state of the gas flow after the turbine and directly affect the power output and efficiency of the engine. At the same time, temperature parameters have a direct impact on the life and reliability of the engine.
[0035] Excessive temperature will increase thermal stress, accelerate the wear and aging of engine parts, and thus reduce the life of the engine. At the same time, excessive temperature may cause engine failure or even cause an accident, so timely detection and handling of temperature anomalies is an important part of ensuring aviation safety.
[0036] Among them, the temperature parameter that can directly reflect the combustion efficiency of aircraft engines is the temperature after the combustion chamber. However, due to the high temperature after the combustion chamber, the requirements for temperature sensors are high. Usually, the outlet temperature of the turbine is selected as the temperature monitoring parameter. The control of the temperature after the turbine is also closely related to fuel efficiency. By reasonably controlling the temperature after the turbine, the combustion efficiency can be improved, thereby reducing fuel consumption and reducing aviation operating costs. By monitoring and controlling the temperature after the turbine, the performance of the engine can be optimized and ensured to operate within the design parameter range. In the present invention, the exhaust temperature T after the turbine is selected. 45 As a health status parameter of aircraft engines.
[0037] (2) Stress health status parameters
[0038] Aircraft engine pressure parameters are important parameters for evaluating the health status of aircraft engines. There are many pressure parameters in aircraft engines, such as low-pressure compressor outlet pressure (LPCOP), high-pressure compressor outlet pressure (HPCOP), turbine inlet pressure (TIP), turbine outlet pressure (TOP), etc. These parameters reflect the air pressure conditions in various components of the engine and have an important impact on the performance and efficiency of the engine.
[0039] The high-pressure compressor after-pressure can directly reflect the health of the engine air path, engine performance and efficiency. Appropriate high-pressure compressor after-pressure can ensure that the compressor works near the design point, thereby ensuring the performance and efficiency of the engine. At the same time, by maintaining an appropriate after-pressure level, the workload of the compressor can be reduced and the fuel consumption of the engine can be reduced. This is of great significance for reducing aviation operating costs and reducing the impact on the environment.
[0040] Reasonable control of the compressor after-pressure can reduce the wear and fatigue of the compressor components and extend the service life of the engine. Too high a pressure may cause excessive stress on the components and accelerate wear, while too low a pressure may cause unstable operation of the components and affect their service life. At the same time, appropriate high-pressure compressor after-pressure can ensure the stable operation of the compressor under various operating conditions. Too high or too low pressure may cause problems such as compressor stall or overheating, affecting the stability and reliability of the engine. Effective monitoring and control of the compressor after-pressure is one of the keys to ensure the safe operation of the engine. By monitoring the high-pressure compressor after-pressure, the health status of the aircraft engine can be directly detected. In the present invention, the high-pressure compressor outlet pressure Ps3 is selected as the health status parameter of the aircraft engine.
[0041] (3) Speed parameters
[0042] The speed of an aircraft engine is one of the important indicators to measure the operating status of the engine, such as the high-pressure shaft speed, low-pressure shaft speed, fan speed, etc. The change in speed can reflect the changes in engine load, power output and working status. The speed is directly related to the performance of the engine. The appropriate set point speed can ensure that components such as the compressor and turbine operate near the designed working point, thereby maximizing the efficiency and performance of the engine. By adjusting the speed, the engine's thrust output, fuel efficiency and response speed can be optimized to meet the needs of different flight phases.
[0043] The stability and balance of the high-pressure shaft speed is crucial to the operating stability of the engine. Excessive speed fluctuations will cause increased engine vibration and may cause equipment damage or system failure. Abnormal changes in the high-pressure shaft speed may cause serious engine failures or even cause aircraft failure. Therefore, continuous monitoring and control of the high-pressure shaft speed is one of the important measures to ensure aviation safety. By timely detecting and handling speed abnormalities, safety issues such as engine stall and excessive wear can be prevented, ensuring flight safety.
[0044] At the same time, the high-pressure shaft speed has an important impact on the life of each key component of the engine. Too high or too low speed will lead to increased wear of components, thereby shortening their service life. By reasonably controlling the high-pressure shaft speed, the wear and fatigue of components can be reduced, the service life of the engine can be extended, and the maintenance cost can be reduced. In the present invention, the high-pressure shaft speed N2 is selected as the health status parameter of the aircraft engine.
[0045] (4) Stall margin parameters
[0046] Aircraft engine stall margin is an important parameter for evaluating engine performance and safety, including fan stall margin (SM Fan ), low pressure compressor stall margin (SM LPC ), high pressure compressor stall margin (SM HPC ) and so on. Stall is a serious engine failure that may cause the engine to be unable to provide sufficient thrust or completely lose thrust. Ensuring sufficient stall margin is one of the keys to ensuring safe aircraft operation. The adequacy of stall margin directly affects the engine's ability to respond to emergencies and improves flight safety.
[0047] Secondly, the stall margin reflects the stability and reliability of the aircraft engine in the face of external disturbances. A larger stall margin means that the engine is more stable and can better cope with changes in external conditions, thereby improving the performance of the engine. By ensuring sufficient stall margin, the engine's thrust output, fuel efficiency and response speed can be optimized.
[0048] Stall margin is also related to the engine's adaptability to different environmental conditions. In special environments such as extreme climates or high altitudes, aircraft engines may be subject to more severe external airflow disturbances, so a larger stall margin is required to ensure stable engine operation. Sufficient stall margin can improve the engine's environmental adaptability and ensure that it can operate reliably under various operating conditions.
[0049] A larger stall margin can reduce fatigue and stress on the blades and extend their service life. At the same time, the adequacy of the stall margin can also reduce the engine failure rate and reduce maintenance costs. Therefore, reasonable design and guarantee of the stall margin are of great significance to the maintenance and life management of the engine. In the present invention, the fan stall margin (SM Fan ), low pressure compressor stall margin (SM LPC ), high pressure compressor stall margin (SM HPC )Three stall margins are used as health status parameters of aircraft engines.
[0050] (5) Fuel parameters
[0051] 1)Specific fuel consumption (T sfc )
[0052] Aircraft engine specific fuel consumption (T sfc , Thrust Specific Fuel Consumption) refers to the amount of fuel consumed by the engine to produce unit thrust, usually in pounds / hour or kilograms / Newton (lb / hr or kg / N). sfc It is of great significance to the performance and operation of aircraft engines. sfc It is one of the important indicators for evaluating engine economy. sfc This means that the engine consumes less fuel to produce a certain amount of thrust.
[0053] Engine T sfc Directly affects the range and endurance of the aircraft. sfc The engine can provide a longer range with a given fuel capacity, or consume less fuel with the same range, thereby increasing the aircraft's range. sfc The engine consumes less fuel per unit thrust, which correspondingly reduces the emission of carbon dioxide and other emissions generated by fuel combustion, which helps to reduce the impact of aircraft on the environment and meets environmental protection requirements. sfc The change of can directly reflect the health status of the aircraft engine, so the present invention selects T sfc As a health status parameter of aircraft engines.
[0054] 2) Fuel consumption ratio
[0055] Aviation engine fuel flow refers to the amount of fuel consumed by the engine while it is running, usually expressed as fuel mass or volume per unit time. Fuel flow is of great significance to the performance and operation of aircraft engines, and fuel flow directly affects the range and endurance of the aircraft. With a given fuel capacity, lower fuel flow means that the aircraft can fly longer distances or longer. The study and optimization of fuel flow helps to improve the design and manufacture of engines and enhance their performance.
[0056] During the operation cycle of an aircraft engine, the health status of the aircraft engine can be monitored by comparing its fuel consumption ratio under a certain operating state. In this paper, a fuel consumption ratio is defined as the health status parameter of the aircraft engine, as shown in Formula 1.
[0057]
[0058] Where η is the fuel consumption ratio of the aircraft engine, W f-PLA1 、N 2-PLA1 The fuel flow and low pressure shaft speed in PLA1 state; W f-PLA2 、N 2-PLA2 It is the fuel flow and low pressure shaft speed in PLA2 state. PLA is the abbreviation of Power Lever Angle. PLA1 refers to the angle when the throttle lever angle is in the first state position. For general research on aircraft engines, it is usually 40°; PLA2 refers to the angle when the throttle lever angle is in the second state position. For general research on aircraft engines, it is usually 80°; other angles can also be used to study some other operating states.
[0059] The monitoring and analysis of aircraft engine health parameters are usually performed through flight data recording systems or ground test equipment. Continuous monitoring and analysis of these parameters can help predict potential failures, optimize maintenance plans, and improve the reliability and safety of aircraft engines.
[0060] Based on the aircraft engine performance calculation, determine the parameters that need to be used as input. These parameters include engine operating status, external environmental conditions, fuel characteristics, etc. Define the health status parameters to be monitored and evaluated. These parameters should be able to reflect the performance, status and health of the engine, such as temperature gradient, pressure loss, vibration amplitude, fuel efficiency, etc. Collect data and ensure the quality and reliability of the data, and perform preprocessing to eliminate noise and outliers. Predict the output parameters based on the input parameters and evaluate the accuracy and generalization ability of the model using a validation data set. Perform model tuning and improvement as needed. Extract health status parameters using real-time or historical data. Embed the model into a real-time monitoring system to calculate health status parameters in real time at runtime, or apply the model to historical data to obtain previously unknown health information.
[0061] The health status parameter set is optimized according to actual needs and application scenarios to ensure that the health status parameter set can accurately reflect the working status and health status of the engine. The established health status parameter set is applied to the real-time monitoring system to monitor the engine status in real time and provide timely feedback and alarms. At the same time, the collected real-time data is used to continuously optimize and improve the health status parameter set.
[0062] Such a health status parameter set based on aircraft engine performance calculation is established to monitor and evaluate the performance and health of the engine, and provide support for maintenance and preventive maintenance.
[0063] S2. Obtain the original data of various indicators of multiple fault types of aircraft engines as samples and construct an original data matrix;
[0064] The present invention divides the faults into six categories according to the actual operating status of the engine: no fault (healthy), low-pressure turbine fault, low-pressure compressor fault, high-pressure compressor fault, fan fault, and combustion chamber fault. The description corresponding to each fault category is shown in Table 1.
[0065] Table 1 Engine fault types and descriptions
[0066]
[0067] Assume that a total of n aircraft engine samples are obtained, and each aircraft engine sample includes the original data of m indicators, forming the original data matrix X:
[0068]
[0069] Among them, x ij Represents the original data of the jth indicator of the ith aircraft engine.
[0070] S3. performing normalization preprocessing on the original data matrix to obtain preprocessed data;
[0071] Each element of the original data matrix is standardized and preprocessed to obtain the preprocessed matrix P = (p ij ) n×m :
[0072]
[0073] In the formula, is the maximum value of the monitoring data of the jth evaluation indicator; is the minimum value of the monitoring data of the jth evaluation indicator; p ij Indicates x ij The normalized values obtained by standardization preprocessing.
[0074] The scale and dimension of data features will affect the clustering results, causing the clustering results to be biased towards features with large scales. Preprocessing can also handle some cases of missing data and outliers, which can effectively ensure the clustering quality.
[0075] S4. Perform cluster analysis on the processed data and classify the preprocessed data using a hierarchical clustering method;
[0076] Fault clustering is an analytical process that aims to identify and classify different fault modes. Its effectiveness is based not only on whether the clustering indicators can capture the key features of the fault object in detail, but also on whether the relative importance of these indicators can be correctly evaluated to ensure that the weight distribution matches the actual fault situation. Although the selection principles for fault clustering indicators are relatively clear at present, following these principles can establish a set of indicators that can fully reveal the characteristics of faults, but the accurate determination of clustering indicator weights is still the key to ensure the accuracy and practicality of clustering analysis results. Therefore, how to reasonably determine the weights of these indicators so that they are close to the actual fault situation has become a core challenge in fault clustering analysis. Weight is a relative concept and a measure of the degree of influence on the health status of the object to be evaluated. If the weight value of a certain evaluation indicator is larger, it means that this indicator has a greater impact on the health status of the object to be evaluated, and vice versa.
[0077] Hierarchical clustering
[0078] Hierarchical clustering is an adaptive data analysis technique that reveals the intrinsic connections between data points by constructing a multi-level clustering structure. This method does not require the number of clusters to be specified in advance, but dynamically explores the hierarchy of the data through gradual merging or splitting operations. The results of hierarchical clustering are usually presented in the form of a dendrogram, which intuitively shows the aggregation process of data points and enables observers to identify natural groupings in the data. It is particularly effective in dealing with data sets with complex structures and is widely used in many fields such as bioinformatics, market segmentation, social network analysis, etc. It is favored by data scientists for its intuitiveness and flexibility. The following is the principle of hierarchical clustering:
[0079] S401. Initialization: The preprocessed matrix P = (p ij ) n×m Each row of is regarded as a sample, and each sample is regarded as an initial cluster;
[0080] S402. Calculate similarity: Calculate the distance between each pair of clusters:
[0081] The distance between each pair of clusters is calculated as follows:
[0082] First, for each pair of clusters, the centroids are calculated respectively, where the centroid refers to the mean of all samples in the cluster;
[0083] Then find the distance between the two clusters: calculate the distance between the centroids of the two clusters; the distance calculation method includes Euclidean distance or Manhattan distance;
[0084] S403. Merge the nearest clusters: select a pair of clusters with the smallest distance to merge and form a new cluster;
[0085] S404. Update distance: Update the distance between all clusters according to the merged clusters;
[0086] S405. Repeat steps S402 to S404 until a predetermined number of clusters is reached.
[0087] S5. Optimize the clustering process using the sparrow algorithm and iterate to obtain the optimal fault classification result.
[0088] Sparrow Search Algorithm (SSA) is a new type of swarm intelligence optimization algorithm proposed in 2020, mainly inspired by the foraging behavior and anti-predation behavior of sparrows. It simulates the strategies and behaviors of sparrows in the foraging process to find the optimal solution. The main principles of the SSA algorithm are as follows:
[0089] (1) Simulation of sparrow foraging behavior: The SSA algorithm is optimized by simulating the foraging behavior of sparrows. Sparrows choose different positions and directions to find food when foraging. The SSA algorithm regards the solution space in the optimization problem as the distribution of food, and individuals search for the optimal solution in the solution space, which is similar to the behavior of sparrows in the foraging process.
[0090] (2) Sparrow individuals and populations: In the SSA algorithm, each individual represents a solution, and they search for the optimal solution through their own behavior and interaction with other individuals. The population consists of multiple individuals, which work together to improve the search effect.
[0091] (3) Flight and position update: In the SSA algorithm, individuals search the solution space by flying and updating their positions. Each individual will fly according to its own position and speed in each iteration. The direction and distance of the flight are affected by the individual and global optimal solutions. The position of the individual will be updated according to the results of the flight to gradually approach the optimal solution.
[0092] (4) Fitness evaluation: In the SSA algorithm, the fitness of each individual is evaluated based on the value of the objective function. The better the fitness value, the closer the individual's position is to the optimal solution.
[0093] (5) Knowledge transfer and update: In the SSA algorithm, individuals transfer and update knowledge to improve the search effect. Individuals can update their own behavior strategies by observing the positions and fitness values of other individuals.
[0094] By simulating the foraging behavior of sparrows, the interaction between individuals and populations, flight and position updates, fitness evaluation, and knowledge transfer and update, the SSA algorithm can search for optimal solutions or solutions close to the optimal solution in the solution space. It has certain global search capabilities and convergence, and is suitable for solving various optimization problems.
[0095] S501. Initialize the sparrow algorithm: Each sparrow corresponds to a clustering threshold, which is used as a predetermined number of clusters in the hierarchical clustering algorithm. When the number of clusters obtained by the hierarchical clustering reaches the clustering threshold, the clustering algorithm stops;
[0096] S502. For each sparrow, execute the hierarchical clustering algorithm to obtain the clustering result and evaluate the clustering effect:
[0097] Based on the hierarchical clustering, the silhouette coefficient of each generated cluster is calculated as an evaluation indicator of clustering quality:
[0098] A1. For any sample point:
[0099] First, calculate the distance between the current sample point and other sample points in its cluster, and take the average value as the cohesion; the distance calculation method includes Euclidean distance or Manhattan distance;
[0100] Then find the cluster closest to the cluster where the current sample point is located as the nearest cluster of the current sample point; then calculate the distance between the sample point and each point in the nearest cluster, and take the average as the separation degree; the distance calculation method includes Euclidean distance or Manhattan distance;
[0101] Subtract the cohesion from the separation, and then divide it by the maximum value between the two to get the silhouette coefficient of the sample point;
[0102] A2. Repeat step A1 for each sample point, and then take the average value of the silhouette coefficients of all sample points to represent the clustering quality of the entire data set;
[0103] S503. Through the sparrow search, within the local search range, find a new solution with a higher silhouette coefficient. If a new solution with a higher silhouette coefficient is found, the optimal solution is updated; the local search range refers to the upper and lower limits of the threshold; wherein the new solution and the optimal solution refer to the clustering threshold corresponding to the sparrow;
[0104] S504. Repeat step S503 until the termination condition is met, where the race condition is: the number of iterations is reached and the silhouette coefficient does not change;
[0105] Using the optimal solution, all samples are hierarchically clustered to obtain the clustering result, which is the optimal fault classification result.
[0106] In the embodiment of the present application, Table 2 shows the engine fault classification results of the fault statistics method proposed by the present invention for different types of fault aircraft engine operation data samples, and the evaluation classification results are consistent with the actual fault state of the engine model. Faults 1 to 5 in the table are low-pressure turbine fault, low-pressure compressor fault, high-pressure compressor fault, fan fault, and combustion chamber fault, respectively.
[0107] Table 2 Classification results of engine faults under test
[0108]
[0109] By constructing a simulation model of an aircraft engine, the present invention can efficiently simulate data that is consistent with actual operation. The acquisition of these data does not rely on costly and risky actual tests and has a high degree of credibility. Through a carefully designed preprocessing process, combined with the advanced classification technology proposed by the present invention, in-depth statistical analysis of potential engine failures can be performed. This innovative method not only overcomes the shortcomings of traditional maintenance strategies in engine failure classification, but also significantly reduces maintenance costs and improves the efficiency and accuracy of decision-making.
[0110] The method provides a new perspective for aircraft engine fault management. It can accurately diagnose potential engine faults by using high-quality data obtained through simulation models and combining advanced classification technology. This method not only improves the reliability of engine operation and reduces the risk of failure, but also shows its excellent performance and application potential in fault prediction, maintenance plan formulation and maintenance cost optimization. Through the method of the present invention, aircraft engine fault management becomes more scientific and efficient, providing strong support for the improvement of aviation safety and operational efficiency.
[0111] The following describes the embodiments of the present invention under different fault types conditions simulated by the aircraft engine model. For a certain type of high bypass ratio turbofan engine, a corresponding aircraft engine model is established according to its component characteristics and cavity effect, and the faults of each component are simulated and fault data is collected as samples of the embodiments of the present invention.
[0112] Select the high-pressure rotor speed N2, the exhaust temperature after the turbine T 45 , unit fuel consumption rate T sfc , high pressure compressor mass flow W, combustion chamber oil-gas ratio FAR, enthalpy value Ht, fuel consumption ratio W f / N2, 7 parameters are used as fault type classification features.
[0113] The possible performance parameter correspondence is shown in the following table, where the number of performance indicators is m (i = 1…, m), and the original data of performance indicators corresponding to x1-x6 are N2, T 45 、T sfc ,W,FAR,Ht,W f / N2, the number of engine samples is n (j=1...,n), and the original data in the present invention uses x ij describe.
[0114]
[0115] The possible performance parameter correspondence is shown in the following table, where the number of performance indicators is m (i = 1…, m), and the performance indicator standardized data corresponding to p1-p6 are N2, T 45 、T sfc ,W,FAR,Ht,W f / N2, the number of engine samples is n (j=1...,n), and the standardized data in the present invention uses p ij describe.
[0116]
[0117] First, the above 7 parameters are normalized and preprocessed. The preprocessing results are shown in Tables 3 to 7.
[0118] Table 3 Normalized data of aircraft engine characteristic parameters (low-pressure turbine failure)
[0119]
[0120]
[0121] Table 4 Normalized data of aircraft engine characteristic parameters (low-pressure compressor failure)
[0122]
[0123] Table 5 Normalized data of aircraft engine characteristic parameters (high pressure compressor failure)
[0124]
[0125] Table 6 Normalized data of aircraft engine characteristic parameters (fan failure)
[0126]
[0127] Table 7 Normalized data of aircraft engine characteristic parameters (combustion chamber failure)
[0128]
[0129] According to the fault classification method process of the present application, engine fault classification is carried out using fault samples collected by the simulation model.
[0130] According to the health parameter matrix, the features that best represent the engine performance are selected, and the collected fault data samples are preprocessed. Then, hierarchical clustering analysis and sparrow algorithm iterative optimization are performed to finally obtain the classification results. The classification results are shown in Table 2, which are consistent with the actual status.
[0131] The number of sample categories is set to 2 to 10, and the silhouette coefficients are {0.7193, 0.8093, 0.9266, 0.8750, 0.9979, 0.9999, 0.9820, 0.9622, 0.9821} respectively. It can be seen that when the fault category is 6, the silhouette coefficient is the largest, and the sparrow algorithm iteration has reached the optimal. The invented civil aviation engine fault statistical method can accurately classify faults.
[0132] The above is only a preferred embodiment of the present invention, and the scope of the rights of the present invention cannot be limited by this. Unless otherwise mentioned, the elements or features may be considered as optional. Each element or feature may be practiced without being combined with other elements or features. The order of operations described in the embodiments of the present invention may be rearranged. Some structures of any embodiment may be included in another embodiment, and may be replaced by the corresponding structure of another embodiment. For those of ordinary skill in the art, without departing from the principles of the present invention, several improvements and changes may also be made, and these improvements and changes are also considered as the scope of protection of the present invention.
Claims
1. An aircraft engine fault classification method based on sparrow algorithm and hierarchical clustering, characterized by: The following steps are involved: S1. Selecting evaluation indicators that characterize the state of the aircraft engine; S2. Obtain the original data of various indicators of multiple fault types of aircraft engines as samples and construct an original data matrix; S3. performing normalization preprocessing on the original data matrix to obtain preprocessed data; S4. Perform cluster analysis on the processed data and classify the preprocessed data using a hierarchical clustering method; S5. Optimize the clustering process using the sparrow algorithm and iterate to obtain the optimal fault classification effect.
2. The method for classifying aircraft engine faults based on the sparrow algorithm and hierarchical clustering according to claim 1 is characterized in that: In step S1, the evaluation indexes of the aircraft engine status include: (1) Temperature health parameter: The exhaust temperature after the turbine of the aircraft engine is selected. 45 ; (2) Pressure health parameter: The high pressure compressor outlet pressure Ps3 of the aircraft engine is selected; (3) Speed parameter: select the high pressure shaft speed N2 of the aircraft engine; (4) Stall margin parameter: Use the aircraft engine fan stall margin SM Fan , low pressure compressor stall margin SM LPC , High pressure compressor stall margin SM HPC Three stall margins; (5) Fuel parameters: Select the specific fuel consumption rate T of the aircraft engine. sfc and fuel consumption ratio η; where, W f-PLA1 、N 2-PLA1 The fuel flow and low pressure shaft speed in PLA1 state; W f-PLA2 、N 2-PLA2 The fuel flow rate and low-pressure shaft speed in PLA2 state.
3. The method for classifying aircraft engine faults based on the sparrow algorithm and hierarchical clustering according to claim 1 is characterized in that: The step S2 comprises: Assume that a total of n aircraft engine samples are obtained, and each aircraft engine sample includes the original data of m indicators, forming the original data matrix X: Among them, x ij Represents the original data of the jth indicator of the ith aircraft engine.
4. The method for classifying aircraft engine faults based on the sparrow algorithm and hierarchical clustering according to claim 1 is characterized in that: The step S3 comprises: Each element of the original data matrix is standardized and preprocessed to obtain the preprocessed matrix P = (p ij ) n×m : In the formula, is the maximum value of the monitoring data of the jth evaluation indicator; is the minimum value of the monitoring data of the jth evaluation indicator; p ij Indicates x ij The normalized values obtained by standardization preprocessing.
5. The method for classifying aircraft engine faults based on the sparrow algorithm and hierarchical clustering according to claim 1 is characterized in that: The step S4 comprises: S401. Initialization: The preprocessed matrix P = (p ij ) n×m Each row of is regarded as a sample, and each sample is regarded as an initial cluster; S402. Calculate similarity: Calculate the distance between each pair of clusters: The distance between each pair of clusters is calculated as follows: First, for each pair of clusters, the centroids are calculated respectively, where the centroid refers to the mean of all samples in the cluster; Then find the distance between the two clusters: calculate the distance between the centroids of the two clusters; the distance calculation method includes Euclidean distance or Manhattan distance; S403. Merge the nearest clusters: select a pair of clusters with the smallest distance to merge and form a new cluster; S404. Update distance: Update the distance between all clusters according to the merged clusters; S405. Repeat steps S402 to S404 until a predetermined number of clusters is reached.
6. The method for classifying aircraft engine faults based on the sparrow algorithm and hierarchical clustering according to claim 1, characterized in that: Described step S5 comprises: S501. Initialize the sparrow algorithm: Each sparrow corresponds to a clustering threshold, which is used as a predetermined number of clusters in the hierarchical clustering algorithm. When the number of clusters obtained by the hierarchical clustering reaches the clustering threshold, the clustering algorithm stops; S502. For each sparrow, execute the hierarchical clustering algorithm to obtain the clustering result and evaluate the clustering effect: Based on the hierarchical clustering, the silhouette coefficient of each generated cluster is calculated as an evaluation indicator of clustering quality: A1. For any sample point: First, calculate the distance between the current sample point and other sample points in its cluster, and take the average value as the cohesion; the distance calculation method includes Euclidean distance or Manhattan distance; Then find the cluster closest to the cluster where the current sample point is located as the nearest cluster of the current sample point; then calculate the distance between the sample point and each point in the nearest cluster, and take the average as the separation degree; the distance calculation method includes Euclidean distance or Manhattan distance; Subtract the cohesion from the separation, and then divide it by the maximum value between the two to get the silhouette coefficient of the sample point; A2. Repeat step A1 for each sample point, and then take the average value of the silhouette coefficients of all sample points to represent the clustering quality of the entire data set; S503. Through the sparrow search, within the local search range, find a new solution with a higher silhouette coefficient. If a new solution with a higher silhouette coefficient is found, the optimal solution is updated; the local search range refers to the upper and lower limits of the threshold; wherein the new solution and the optimal solution refer to the clustering threshold corresponding to the sparrow; S504. Repeat step S503 until the termination condition is met, where the race condition is: the number of iterations is reached and the silhouette coefficient does not change; Using the optimal solution, all samples are hierarchically clustered to obtain the clustering result, which is the optimal fault classification result.
Citation Information
Patent Citations
Aero-engine health state recognition system and method based on data mining
CN112381965A
Group recommendation method based on sparrow search optimization clustering
CN112883282A
WSN clustering routing method based on improved fuzzy C-means clustering
CN115843085A
Water plant substation electrical equipment fault diagnosis method based on bi-kmean-LDA-KELM
CN117688994A
Feature amplification-based aero-engine fault diagnosis method and system
CN117988988A
