CN106094570B - A kind of aero-engine complete machine health evaluating method under variable working condition based on this distance of operating mode's switch and paddy - Google Patents

Abstract

A health assessment method for aero-engines under variable operating conditions based on operating condition identification and Tanimoto distance. Aeroengines work in a complex and changeable environment, and the changing operating conditions of the engine during use make the performance data not decline. Trend, the evaluation results of health assessment methods based on single model are not accurate enough. The present invention proposes a health assessment method of an aero-engine under variable working conditions based on working condition identification and Tanimoto distance. First, the current working condition of the aero-engine is identified; secondly, the corresponding health assessment model is selected according to the identified working condition , to evaluate the health status of the system; finally, combine the health indices under different working conditions to obtain the health index series under variable working conditions. The method proposed by the invention can avoid the interference of the change of working conditions and provide the true health state of the aeroengine.

Description

An aero-engine machine health under variable operating conditions based on operating condition identification and Tanimoto distance health assessment method

technical field

The invention relates to the technical field of aero-engine health management, in particular to a method for evaluating the health of an aero-engine under variable operating conditions based on operating condition identification and Tanimoto distance.

Background technique

Aeroengines are very important to aircraft, and their performance directly determines the flight safety of the aircraft. Due to the high cost of aircraft engines, airlines do not frequently replace engines when there is a problem with the engine, but solve the problem through maintenance and repair. Since the 1980s, the global aviation industry has developed rapidly, but the accompanying flight safety issues have become more and more prominent. Statistics show that among all aircraft flight accidents, accidents caused by aeroengine failure account for about 60%. Therefore, how to effectively prevent aero-engine failures has become the top priority of ensuring flight safety.

The structure of the aero-engine is very complicated, and the working environment is extremely harsh, so it is easy to break down without maintenance. However, the cost of repairing aero-engines is very high. Due to cost budget considerations, it is impossible for airlines to conduct a comprehensive overhaul before each flight takes off, and often adopts the strategy of "condition-based maintenance". "Condition-based maintenance" is also called "condition-based maintenance", which is to judge the health status of the aero-engine at the current stage. If the health status of the aircraft engine is good, it will not be repaired; It is overhauled. In practice, the health status of an aero-engine is often evaluated based on the performance parameters of the aero-engine, such as engine exhaust temperature, fuel flow, and turbine speed. When the monitored performance parameters are abnormal, engineers can determine that there is a problem with the engine. Therefore, monitoring the performance parameters of aero-engines and assessing their health status is crucial for guiding "condition-based maintenance". The faults of most aeroengines are gradually exposed through the accumulation of time, and the current health status of the system can be evaluated by studying the changing rules of performance parameters. When the health is lower than the warning value, it can be judged that the aero-engine is likely to fail in the near future, and maintenance and repair measures should be taken quickly. From the above statement, it can be seen that the health assessment of aero-engine is of great significance to assist the "condition-based maintenance".

However, aeroengines work under complex and changeable working conditions, and the changing working conditions cover up the real degradation law of performance parameters. Due to the interference of changing working conditions, pure performance parameters cannot reflect the health status of the system.

Contents of the invention

The object of the present invention is: in order to overcome the above-mentioned defect of prior art, the present invention proposes a kind of aero-engine whole machine health evaluation method under variable working condition based on working condition identification and Tanimoto distance, firstly identify the working condition of the current state, and then Selecting the evaluation model under specific working conditions to evaluate the health status can avoid the interference of working conditions and give the real health status of the engine.

The technical solution adopted in the present invention is: a method for evaluating the health of an aero-engine complete machine under variable working conditions based on working condition identification and Tanimoto distance, comprising the following steps:

The first step is to identify the operating conditions of the engine. First, use all working condition parameters of the training set data for cluster analysis to obtain the cluster center and cluster radius of each working condition; then calculate the distance between the working condition parameters of each state and each cluster center, and identify the working condition of the system ;

In the second step, the health assessment of a single working condition is performed based on the Tanimoto distance measure. First, the normal state space of the normal operating state of the aeroengine is obtained; then, the Tanimoto distance between the current state and the normal state is calculated; finally, the distance is converted into a health index to quantify the health state of the system.

The third step is to integrate the assessment results of different health assessment models. The health assessment results under different working conditions are integrated into a new time series according to the original time order, and the system health index series is obtained.

The advantage of the present invention compared with prior art is:

(1) On the basis of working condition identification, multiple evaluation models are used to evaluate the system status under different working conditions, and the evaluation results are not disturbed by working conditions, which is more accurate and objective;

(2) Tanimoto distance can not only characterize the length difference between two vectors, but also characterize their angle difference. Compared with Euclidean distance and angle cosine, it can measure the difference between two vectors more comprehensively.

Description of drawings

Figure 1 is a schematic diagram of the health assessment process of an aero-engine machine;

Fig. 2 is a flow chart of K-means clustering;

Figure 3 is a flow chart of working condition identification;

Figure 4 is a schematic diagram of health assessment based on Tanimoto distance;

Fig. 5 is a structural diagram of a turbofan engine simulation model;

Fig. 6 is a schematic diagram of the sensor monitoring data of the training set 1#;

Fig. 7 is a schematic diagram of the monitoring data of the training set 2# engine sensor;

Fig. 8 is the working condition clustering result diagram;

Figure 9 is a schematic diagram of the real-time working conditions of the engine, wherein Figure 9(a) is the training set 1# engine, and Figure 9(b) is the training set 2# engine;

Fig. 10 is a schematic diagram of the health assessment results, wherein Fig. 10(a) is the training set #1 engine, and Fig. 10(b) is the training set #2 engine.

detailed description

The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

The present invention is based on working condition identification and Tanimoto distance health assessment method for aero-engine complete machine under variable working conditions. The specific steps are as follows:

1. Working condition identification based on K-means clustering

The K-means clustering algorithm was proposed by MacQue, which is widely used in data mining and is one of the classic clustering algorithms. Let X={X ₁ ,X ₂ ,...,X _n } be a known data set, X ₁ ,X ₂ ,...,X _n in X are n data objects and each data object is N-dimensional, that is, _{Xi = (x i1 ,} x _i2 , . . . , x _iN ). The K-means clustering algorithm is to find a set C={C ₁ ,C ₂ ,...,C _K }={(c ₁₁ ,c ₁₂ ,...,c _1N ),( c ₂₁ ,c ₂₂ ,...,c _2N ),...,(c _K1 ,c _K2 ,...,c _KN )} such that the objective function

Among them, n _i is classified as the number of data object points of class C _i , and d(C _i , X _j ) represents the Euclidean distance between the cluster center and the data object, which is defined as follows:

The core idea of the K-means clustering algorithm is to divide the data set into K classes that make the objective function reach the minimum value. The specific steps are shown in Figure 2.

The process of working condition identification is shown in Figure 3. The cluster centers and radii of each working condition are obtained through the K-means clustering algorithm using all working condition data in the training set. Calculate the distance between the real-time working condition parameters and each cluster center, and the closer one is the working condition.

2. Health assessment based on Tanimoto distance

After the working conditions are identified, it is necessary to establish a health assessment model based on the state data of each working condition, and then evaluate each state. The essence of health assessment is the process of transforming multidimensional parameters into one-dimensional health index, and the health index is between 0 and 1. Health assessment includes two core steps, ie multivariate data fusion and data standardization, and the present invention uses distance measure method to realize multivariate data fusion.

The Tanimoto Distance measure can simultaneously represent the angle and relative distance information between two points (states). For some datasets, the relative lengths of the vectors contain valuable information; for others, the angle between the vectors measures how close those data objects are. The Euclidean distance measure reflects the distance between vector lengths, regardless of the angle between the vectors; while the cosine distance measure ignores the length of the vectors. Therefore, the present invention selects Tanimoto distance to comprehensively measure the gap between different states of the engine. The Tanimoto distance measure formula between two n-dimensional state vectors a=(a ₁ ,a ₂ ,...,a _n ) and b=(b ₁ ,b ₂ ,...,b _n ) is:

In the formula, Defined as Tanimoto Coefficient, Tanimoto Coefficient measures the degree of similarity between two spaces or vectors.

The principle of health assessment based on Tanimoto distance is shown in Figure 4. First, the normal state space of the engine system is established, then the Tanimoto distance between the state to be evaluated and the reference space is calculated, and finally the current health index of the system is obtained by using the normalization formula. The formula is:

In the formula, D _i is Tanimoto distance, and a is a smoothing parameter, which is used to adjust the sensitivity of health assessment.

The health indexes obtained by using multiple evaluation models under different working conditions have the same range and scale, and the health indexes are comparable. Merge them in the order of the original time stamps to obtain a unified health index time under variable working conditions sequence, complete a health status assessment.

3. Case verification

The invention verifies the proposed health assessment model through the performance degradation simulation data of civil turbofan engines provided by the NASA research center. The simulation model is established in the CMAPSS environment developed by NASA Army Research Laboratory. CMAPSS can simulate the operation of a 90,000-pound thrust class engine model at different altitudes, Mach numbers, and sea-level temperatures. The thrust of the engine is controlled by the throttle lever angle (TRA). , which can be regarded as a working condition parameter. By modifying the flow rate and efficiency of each rotating part in CMAPSS, a total of 13 health state parameters can simulate the five major rotating parts of the engine, namely the fan, low-pressure compressor, high-pressure compressor, high-pressure turbine, and low-pressure turbine. The failure and performance degradation of the engine machine impact. CMAPSS can simultaneously monitor and record 58 performance variables of the simulated system. The gas circuit structure of the turbofan engine is shown in Figure 5, and the physical meanings and symbols of the 14 input quantities of CMAPSS are shown in Table 1.

Table 1 CMAPSS input variables

Saxena et al. used the CMAPSS model to conduct a simulation test in 2008, and monitored 21 sensor parameters in 58 system outputs in real time. The monitoring parameters mainly consist of temperature, pressure and speed at different positions in the engine gas circuit. In addition, the state parameters also include the height, Mach number and throttle lever angle (thrust) that characterize the engine working state. Table 2 gives the physical meaning and symbols of 21 sensor parameters.

Table 2 CMAPSS output parameters

serial number symbol describe unit 1 T2 Fan inlet temperature °R 2 T24 Low pressure compressor outlet temperature °R 3 T30 High pressure compressor outlet temperature °R 4 T50 Low pressure turbine outlet temperature °R 5 P2 Fan outlet pressure psia 6 P15 duct pressure psia 7 P30 High pressure compressor outlet pressure psia 8 Nf speed of the fan rpm 9 Nc core speed rpm 10 epr pressure ratio (P50/P2) 11 Ps30 High pressure compressor outlet static pressure psia 12 phi fuel flow ratio pps/psi 13 NRf Corrected fan speed rpm 14 NRc Corrected core speed rpm 15 BPR Bypass ratio -- 16 far B Combustion air ratio of the burner -- 17 htBleed Exhaust enthalpy -- 18 Nf_dmd Required fan speed rpm 19 PCNfR_dmd Required corrected fan speed rpm 20 W31 High pressure turbine coolant flow rate lbm/s twenty one W32 Low Pressure Turbine Coolant Flow Rate lbm/s

NASA provides 4 performance degradation datasets with different simulation settings, which come from 4 independent simulation tests, and each set of simulation tests has different settings such as working conditions and failure modes. Dataset 1 and Dataset 2 contain one failure mode, High Pressure Compressor Degradation (HPC), Dataset 3 and Dataset 4 contain two failure modes of High Pressure Compressor Degeneration (HPC) and Fan Degeneration (Fan). Dataset 1 and Dataset 3 contain one working condition pattern, and Dataset 2 and Dataset 4 contain 6 working condition patterns. Table 3 gives the details of the 4 datasets.

Table 3 Experimental setup information of the dataset

Each dataset contains a test set and a training set. The data in the training set is whole-life and can be used to develop models for health assessment and lifespan prediction. The present invention uses the training data of dataset #2 to test and validate the proposed health assessment method.

Figures 6 and 7 show the monitoring data of the sensors in the training set 1# and 2# engines during the performance degradation process. The two engines are working under variable conditions. From the analysis of the curves, it can be seen that the complex condition changes make all sensors None of the parameters can represent the performance decline trend.

Using all the data in the training set, the cluster centers and cluster radii of the six working conditions are obtained through K-means clustering, as shown in Table 4, and the clustering diagram of the working conditions is shown in Figure 8. For the data to be evaluated, the distance between the real-time working condition parameters and the six cluster centers is firstly calculated, and the one with the closest distance is judged as the working condition.

Table 4 Cluster center and cluster radius

The operating condition profiles of the training set 1# and 2# engines are identified, and the results are shown in Figure 9.

Taking the training set 1# and 2# engines as examples, the proposed variable-condition health assessment method is tested and verified. The results of the health assessment are shown in Figure 10.

It can be seen from Figure 10 that as the number of cycles increases, the health index of the aero-engine is getting smaller and smaller, and the health status is degraded. It is possible to set the threshold of health to issue an over-limit alarm, prompting maintenance personnel to arrange "condition-based maintenance" in advance, which can not only avoid engine failure during use, but also reduce the high cost of planned maintenance.

The technologies known in the art involved in the present invention are not described in detail.

Claims (1)

1. a method for evaluating the health of an aero-engine complete machine under variable operating conditions based on operating condition identification and Tanimoto distance, is characterized in that: comprises the following steps: The first step is to identify the operating conditions of the aero-engine: first, use the flight altitude, flight speed, and throttle stick angle of the aero-engine training set data to perform cluster analysis to obtain the cluster center and cluster radius of the operating conditions ; Then calculate the normalized Euclidean distance between the working condition parameters of the current state and each cluster center, and identify the current working condition state of the system; The second step is to evaluate the health state of each single working condition based on the Tanimoto distance measure: first, get the health baseline of the aero-engine under normal operating conditions; then, calculate the Tanimoto distance between the current operating state and the health baseline; finally, the The obtained Tanimoto distance is normalized to the health degree CV value, which is used to quantify the health degree of the system; The third step is to integrate the evaluation results of different health evaluation models to obtain the health degradation curve: the evaluation results under different working conditions are integrated into a new time series according to the original time sequence, and the health state degradation time series of the system is obtained; Based on working condition identification and Tanimoto distance, the aero-engine health assessment method under variable working conditions is based on working condition identification, and uses multiple health assessment models with comprehensive Tanimoto distance and normalization methods to evaluate the health of different working conditions. The health status of the system, the evaluation results are not disturbed by the change of working conditions, and are more objective and accurate; The Tanimoto distance can not only represent the length difference between the eigenvectors corresponding to different health states, but also represent the angle difference between the health eigenvectors. , its measurement of differences in health characteristics is more comprehensive and closer to objective reality, and the evaluation results based on Tanimoto distance are more credible.