CN119164659A

CN119164659A - An adaptive threshold aeroengine sensor fault diagnosis method based on measurement parameters

Info

Publication number: CN119164659A
Application number: CN202410873613.XA
Authority: CN
Inventors: 高思捷; 韩鑫浩; 鲁峰; 刘康飞; 周鑫; 方怡丹
Original assignee: Nanjing University of Aeronautics and Astronautics
Current assignee: Nanjing University of Aeronautics and Astronautics
Priority date: 2024-07-01
Filing date: 2024-07-01
Publication date: 2024-12-20

Abstract

The invention discloses a fault diagnosis method of an adaptive threshold aeroengine sensor based on measured parameters, which comprises the steps of adopting OSELM algorithm to establish an analytical redundancy model of an engine input signal and a sensor signal, designing a univariate controller based on model-free adaptive control, designing a tracker of a baseline model on the basis, outputting and calculating a fault threshold of the sensor fault diagnosis by utilizing the baseline model, obtaining the sensor fault threshold with adaptive capacity in a certain working condition range in an envelope, and giving out a fault judgment basis. The invention solves the problem that the sensor fault diagnosis threshold value is difficult to determine under the conditions of variable working environment parameters and random noise in the traditional sensor fault method. The method uses historical data of health sensors to train and update an engine baseline model. For offset faults and drift faults, the new algorithm achieves higher fault detection rates, lower false alarm rates, and smaller minimum detection fault magnitudes. The system can adapt to changeable working environments and long-term service life, and provides more accurate and reliable diagnosis basis for early warning of faults of the engine sensor.

Description

Adaptive threshold aeroengine sensor fault diagnosis method based on measurement parameters

1. Technical field

The invention belongs to the technical field of aeroengine fault diagnosis, and particularly relates to a self-adaptive threshold value aeroengine sensor fault diagnosis method based on measurement parameters.

2. Background art

As a main means for the engine to acquire operating information, the accuracy of the sensor measurement information is critical to the proper operation of the engine control system and the health management system. Aeroengine sensor elements are mostly operated in harsh environments of high temperature, high pressure, high speed, intense vibration and gas corrosion. Thus, the sensor is a component that is prone to failure. Once the sensor fails, the engine performance will drop substantially, the engine will be damaged, the flight safety will be compromised and catastrophic accidents will occur. Therefore, timely and effective fault diagnosis plays an indispensable role in improving the reliability and safety of the control system. The need for high reliability and low cost of aeroengines places higher demands on engine health management technology, and therefore more intelligent and autonomous diagnostic, prognostic and health management systems must be developed. For sensor fault diagnosis, a reasonable diagnostic criteria is required. The main difficulty of fault diagnosis is to design a threshold setting scheme, which can effectively distinguish faults and false alarms and balance the relationship between fault diagnosis detection rate and false alarm rate. Therefore, it is of great importance to study and design a threshold setting method for sensor fault diagnosis.

The traditional threshold setting method needs to carry out statistical analysis on a large amount of historical data of the health sensor, and a series of experiments are also needed to obtain a diagnosis threshold value when the sensor works normally. The method considers a data model-based method to calculate an adaptive threshold for engine sensor failure. In the model-free adaptive control (MFAC) method, a new dynamic linearization method and a new concept called Pseudo Partial Derivative (PPD) are used for discrete-time nonlinear systems. And establishing an equivalent dynamic linearization data model at each dynamic working point of the closed-loop system, designing a controller based on the equivalent dynamic linearization data model, and carrying out theoretical analysis on the control system, thereby realizing the self-adaptive control of the nonlinear system. The MFAC method has the following advantages, which makes it more suitable for control problems of practical systems. First, MFAC is a data-driven control method that relies solely on real-time measurement data of the controlled system and not on any mathematical model information of the controlled system. Secondly, the MFAC method is simple, small in calculated amount, easy to implement and high in robustness. MFAC has significant advantages for strongly non-linear systems with model uncertainty like aeroengines. Under the control of the MFAC method, an equivalent dynamic linearization data model is built on-line at each dynamic operating point, which model can represent the operating characteristics of the operating point. The MFAC-based method establishes a health tracker to track the output of the aircraft engine in a healthy state. Based on the data model established by the MFAC method, the adaptive threshold of the sensor is deduced.

When a sensor fails, a data model based on engine measurement parameters cannot represent the normal output of the sensor. Therefore, it is necessary to establish sensor redundancy to characterize the output of a normal sensor. OSELM can make full use of the currently acquired data, recursively update the output weights of the neural network according to samples at different times, and update network parameters online. From this feature, a mapping between the variable to be estimated and other signals can be established by OSELM. During engine maintenance, the data model is updated online based on the sensor output signals. In order to obtain a reliable sensor fault diagnosis threshold value, improve the accuracy of sensor fault diagnosis and reduce the false alarm rate, an adaptive threshold value aeroengine sensor fault diagnosis method based on measurement parameters is provided, according to the method, the theoretical threshold value of sensor fault diagnosis is calculated by utilizing the characteristic that a model-free self-adaptive control is used for establishing a dynamic data model at each working point, a basis is provided for the sensor to judge faults, and the reliability of an engine control system can be improved.

3. Summary of the invention

Aiming at the technical problems, the invention provides a sensor fault diagnosis method of an adaptive threshold aeroengine based on measurement parameters, which solves the problem that the sensor fault diagnosis threshold is difficult to determine under the conditions of variable working environment parameters and random noise in the traditional sensor fault method. The controller usually uses the rotating speed as a control quantity, so that the fault diagnosis of the rotating speed sensor is particularly important, the fault diagnosis of the engine sensor is suitable for changeable working environments and long-term service life, and the fault diagnosis of the engine sensor has positive promotion effects on improving the accuracy of the fault diagnosis of the sensor, reducing the false alarm rate and improving the engine control system.

The technical scheme adopted by the invention is as follows:

An adaptive threshold aeroengine sensor fault diagnosis method based on measured parameters is characterized by comprising the following steps:

And A), respectively constructing an aeroengine sensor analysis redundancy model by adopting an OS-ELM algorithm group according to the historical data of the selected engine sensor and the rest sensors.

Step B) designing an MFAC controller, the tracker based on the baseline model in step (a) being designed by the MFAC controller.

And C) calculating a sensor fault threshold value by using a baseline model tracker, and designing an aero-engine sensor fault judgment criterion.

Further, in the step a), according to the historical data of the selected engine sensor and the rest sensors, the specific steps of constructing an analysis redundancy model of the aeroengine sensor by adopting an OS-ELM algorithm set are as follows:

Step A1), selecting a sensor y ₁(t),...,y_n(t),y_k (t) of an engine needing to establish a prediction model as output quantity, y ₁(t),...y_k-1(t),y_k+1(t),...,y_n (t) and fuel flow And according to the selected input output quantity, a sensor prediction model constructed by using an OS-ELM algorithm is as follows:

Where f ₁,…,f_n can be approximated by training the OS-ELM algorithm, p is the embedding dimension, and p=2 since the aeroengine can be approximated as a 2-order object.

Step A2), generating sensor training samples by using the engine component level model, wherein the acquisition of the sensor signals and the online training of the predictor are performed simultaneously. In each sampling period, the predictor collects the output of each sensor and groups the output according to different prediction units, takes the measured value of the sensor to be estimated as expected output, takes the measured signals of the other sensors and the historical signals thereof as input, and constructs training samples. The samples are trained online by the OS-ELM model, and an accurate sensor baseline model is obtained according to the repeated updating of new samples.

Further, the step B) of designing the MFAC controller, and the step of designing the tracker based on the baseline model in the step (a) by using the MFAC controller is as follows:

And B1), selecting the control quantity and the controlled quantity of the MFAC controller designed for the engine, and respectively solving a control law and a PPD estimation algorithm by utilizing a control input criterion function and a PPD estimation standard function to obtain the MFAC-based engine controller.

The control rule is as follows:

If it is Or (b)

Wherein σ ₁＞0,σ₂ >0 is a weight coefficient for limiting and controlling the magnitude of the input signal variation, y _d (k+1) is a positive number where the desired output signal epsilon is sufficiently small; η ₂ e (0, 1) is the step size factor; An estimate of the pseudo-partial derivative phi (k); Is that Is the initial value of (a).

Step B2), designing a nonlinear tracker based on the MFAC controller for tracking the obtained baseline model.

The designed tracker is as follows:

wherein ρ is the feedback gain, For the estimation of the error, Δu _N (k) is the engine control output variation amount when the sensor is normal.

Further, the step C) calculates a sensor fault threshold by using a baseline model tracker, and the specific steps for designing the fault judgment criterion of the sensor of the aero-engine are as follows:

step C1), defining a system for normal operation of the engine as follows:

y_N(k+1)＝y_N(k)+φ_N ^T(k)Δu_N(k)

Substituting it into tracker

Wherein the method comprises the steps of

Step C2) is scaled to obtain

Wherein the method comprises the steps of

Step C3) introducing a sliding window mechanism, further reducing the threshold value and the minimum detection fault amplitude

Where n is the sliding window length for adjusting the utilized historical data time span.

Wherein the method comprises the steps of

Step C4) obtaining a sensor fault threshold value and a sensor fault judgment criterion:

the sensor failure threshold is:

d_{f_ad}(k)＝(1-w)d_{f_slid}(k)

w=1/(1+e^-q(k))

wherein std is the standard deviation of the sensor noise, Is the error between the sensor measurement and the baseline model prediction.

The sensor fault judgment criterion is as follows:

A common logic for sensor fault diagnosis is to analyze whether the prediction residual exceeds a given threshold. In the present method, when At this time, a sensor failure of the aircraft engine may be detected at k. So far, the design of the sensor fault diagnosis method based on the self-adaptive threshold value is finished.

The self-adaptive threshold value aeroengine sensor fault diagnosis method based on the measured parameters has the advantages that compared with the prior art, the technical scheme has the following technical effects:

(1) A large number of statistics and simulation experiments are not needed to be carried out on the sensor historical data like the traditional threshold setting method, a dynamic linear model is established on line at each working point by utilizing the sensor historical data, and a theoretical self-adaptive threshold of diagnosis is calculated according to the data model.

(2) The OSELM algorithm provides analysis redundancy for sensors used in the turboprop engine, has good prediction accuracy in acceleration, deceleration and steady state processes, and improves the reliability of a high-pressure turbine rotating speed sensor baseline model.

(3) The threshold setting method can be suitable for a wider engine working range, and the threshold can be adjusted according to sensor data under various engine working conditions. Therefore, it has flexible and convenient working characteristics.

(4) Aiming at the micro-bias faults of the engine sensor, a threshold correction coefficient is introduced, so that the threshold setting method still keeps good false alarm rate and fault detection rate.

4. Description of the drawings

FIG. 1 is a block diagram of an adaptive threshold aircraft engine sensor fault diagnosis based on measured parameters of the present invention.

FIG. 2 is a graph of simulated fuel flow.

FIG. 3 is a graph comparing predicted and actual output curves of a turboprop.

Fig. 4 is a control effect diagram of the MFAC controller.

Fig. 5 is a graph of tracking error of the tracker in case of sensor failure.

Fig. 6 is an Nh sensor micro bias fault diagnosis.

Fig. 7 is an Nh sensor drift fault diagnostic diagram.

5. Detailed description of the preferred embodiments

The following describes the embodiments of the present invention further with reference to the drawings.

The invention discloses a self-adaptive threshold value aeroengine sensor fault diagnosis method based on measured parameters, which is characterized by comprising the following steps of:

Step A1), selecting a sensor y ₁(t),...,y_n(t),y_k (t) of an engine needing to establish a prediction model as output quantity, y ₁(t),...y_k-1(t),y_k+1(t),...,y_n (t) and fuel flowAnd according to the selected input output quantity, a sensor prediction model constructed by using an OS-ELM algorithm is as follows:

The control rule is as follows:

If it is

The designed tracker is as follows:

Step C1), defining a system for normal operation of the engine as follows:

y_N(k+1)＝y_N(k)+φ_N ^T(k)Δu_N(k)

Substituting it into tracker

Wherein the method comprises the steps of

Step C2) is scaled to obtain

Wherein the method comprises the steps of

the sensor failure threshold is:

d_{f_ad}(k)＝(1-w)d_{f_slid}(k)

w=1/(1+e^-q(k))

The sensor fault judgment criterion is as follows:

In order to verify the effectiveness of the self-adaptive threshold aeroengine sensor fault diagnosis method based on the measured parameters, the invention carries out the baseline model output precision simulation of the engine take-off and acceleration and deceleration processes in a certain envelope and the digital simulation of the high-pressure compressor rotating speed sensor fault diagnosis in the MATLAB environment.

The invention adopts a nonlinear component model of a certain type of three-rotor turboprop engine as a research object. The model is built by an object-oriented programming method, comprises important parts of an aeroengine such as an air inlet channel, a fan, a gas compressor, a combustion chamber, a gas turbine, a power turbine, a tail nozzle and the like, and is easy to call in a MATLAB environment.

And constructing an aeroengine sensor analysis redundancy model by adopting an OS-ELM algorithm group according to the selected historical data of the engine sensor and the rest sensors.

Firstly, selecting high-pressure compressor outlet pressure, high-pressure compressor outlet temperature and power turbine outlet pressure to construct predictors, and using OSELM as a signal predictor, wherein the expression of each sensor signal prediction model is as follows:

n _h predictor:

P ₃ predictor:

t ₃ predictor:

P ₅ predictor:

historical sensor data for the in-service turboprop is stored in a database. The OSELM algorithm trains the predictive model for each sensor using the sensor data in the database to arrive at a baseline model for the engine.

The experiment adopts an open loop control process, which is beneficial to eliminating the influence of other interferences. Given the fuel flow inputs as shown in fig. 2, output conditions and performance parameters are obtained. The simulation lasted 80 seconds, with the fuel flow rate varying from 40% to 100%, going through acceleration, steady state, and deceleration processes in sequence. The set engine operating environment is h=0km, ma=0. All sensors were normalized. To simulate the working environment of the engine in the actual process, gaussian noise with a mean value of 0 and a variance of 0.003 ² were added to the sensor output, respectively. For each sensor prediction model, the number of hidden layer neurons is set to 21 and the activation function is a sigmoid function. To evaluate the accuracy of the sensor predictor, RMSE was selected as an evaluation index for the algorithmic prediction performance. The training and testing accuracy is shown in table 1. The prediction curves of the prediction model are given in fig. 3.

TABLE 1 training and testing accuracy of turboprop data sets

Bias faults and drift faults are taken as typical fault modes according to fault characteristics. The following model was used for simulation.

Bias failure:

Wherein y _t is the sensor output, For the sensor true value, Δy _t is the offset, t is the current time, t ₀ is the time of failure occurrence, and 1 (·) is the unit step function.

Drift failure:

Where ε _t is the drift rate.

The sensor fault diagnosis simulation is simulated using simulation conditions in the baseline model. The parameters of the MFAC are set as follows, initial values of the adaptation lawDesign parameter σ ₁＝2,σ₂＝1,η₁＝0.3,η₂＝1,ε＝10^-5, feedback gain of tracker ρ=0.3, length of sliding window n=5. The experiments include bias fault simulation and drift fault simulation. The relevant fault parameters are set as deltay _t＝0.008,ε_t =0.002/s. To evaluate the adaptive threshold setting method presented herein, the performance of the adaptive threshold and two fixed threshold methods were compared. The two fixed thresholds selected are d _max and d _3*std. The threshold d _max is the maximum value under normal conditions. The d _3*std threshold is 3 times the standard deviation of the noise.

The effect of the MFAC control in closed loop control according to a given desired speed is shown in fig. 4, and it can be seen from fig. 4 that the MFAC controller can be quickly adjusted to the reference value over a large operating range and that the steady state error during simulation is close to zero. The adjustment time of the engine acceleration and deceleration dynamic process is also shorter, and satisfactory dynamic performance is obtained, which ensures that the high-pressure gas turbine rotational speed can accurately run to the target rotational speed and achieve the ideal power of the turboprop engine. The variation of the control amount is also within a reasonable range.

FIG. 5 shows the tracking effect of an MFAC-based tracker on a baseline model under both bias sensor failure and drift sensor failure. It can be seen from the figure that the tracking error is within 1%, and the tracker has high estimation accuracy. Thus, the theoretical threshold calculated by the tracker may represent the range of the sensor under normal conditions.

FIG. 6 shows a speed profile of the engine from 88% gas turbine speed to the design point speed. Within 60-70 seconds, the speed sensor has a deviation fault. For small bias faults with a fault amplitude of 0.008, the method has a very high fault detection rate.

In fig. 7, when a drift fault occurs at 60 seconds, the adaptive threshold is difficult to detect a sensor fault due to the small amplitude of the fault at the beginning. As the fault time increases, the fault magnitude of the drift fault becomes progressively larger. When the fault magnitude reaches 0.00328, the adaptive threshold begins to detect a sensor fault.

Table 2 shows the specific performance indicators for the three methods.

Table 2 sensor fault diagnosis performance index

Therefore, the self-adaptive threshold aeroengine sensor fault diagnosis method based on the measured parameters has good sensor fault diagnosis precision.

The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.

Claims

1. A method for diagnosing faults of aircraft engine sensors based on adaptive thresholds of measurement parameters, characterized in that it comprises the following steps:

Step A) Based on the historical data of the selected engine sensor and the remaining sensors, an aircraft engine sensor analytical redundancy model is constructed using the OS-ELM algorithm group.

Step B) Design an MFAC controller based on the tracker of the baseline model in step (A).

Step C) Calculate the sensor fault threshold using the baseline model tracker and design the aero-engine sensor fault judgment criteria.

2. The method for fault diagnosis of aircraft engine sensors based on adaptive thresholds of measurement parameters according to claim 1 is characterized in that: in step A), the aircraft engine sensor analytical redundancy model is constructed by using the OS-ELM algorithm group according to the historical data of the selected engine sensor and the remaining sensors respectively. The specific steps are as follows:

Step A1), select the sensors y ₁ (t), ..., _yn (t), y _k (t) of the engine for which a prediction model needs to be established as output quantities, y ₁ (t), ...y _k-1 (t), y _k+1 (t), ..., _yn (t) and the fuel flow As input; according to the selected input and output, the sensor prediction model constructed by OS-ELM algorithm is:

Among them, f ₁ , ..., f _n can be approximated by training the OS-ELM algorithm, p is the embedding dimension, and since the aircraft engine can be approximated as a second-order object, p=2.

Step A2) uses the engine component-level model to generate sensor training samples. The acquisition of sensor signals and the online training of the predictor are carried out simultaneously. In each sampling period, the predictor collects the outputs of each sensor and arranges and groups them according to different prediction units. The sensor measurement value to be estimated is used as the expected output, and the remaining sensor measurement signals and their historical signals are used as input to construct the training samples. The samples are trained online by the OS-ELM model, and recursively updated according to the continuously acquired new samples to obtain an accurate sensor baseline model.

3. The method for fault diagnosis of an aircraft engine sensor based on an adaptive threshold of measurement parameters according to claim 1, characterized in that: the MFAC controller is designed in step B), and the tracker of the baseline model is designed based on the MFAC controller. The specific steps are as follows:

Step B1), selecting the control variable and the controlled variable of the MFAC controller designed for the engine, using the control input criterion function and the PPD estimation standard function to solve the control law and the PPD estimation algorithm respectively, and obtaining the MFAC-based engine controller.

Its control law is:

if or

Among them, σ ₁ >0, σ ₂ >0 are weight coefficients used to limit and control the amplitude of the input signal change; y _d (k+1) is the expected output signal ε is a sufficiently small positive number; η ₂ ∈(0,1) is the step size factor; is the estimate of the pseudo partial derivative φ(k); for The initial value of .

Step B2), designing a nonlinear tracker based on the MFAC controller to track the obtained baseline model.

The designed trackers are:

Where ρ is the feedback gain, is the estimation error, and Δu _N (k) is the change in engine control output when the sensor is normal.

4. The intelligent multivariable control method for an aircraft engine with online correction according to claim 1 is characterized in that: the step C) uses a baseline model tracker to calculate the sensor fault threshold and designs an aircraft engine sensor fault judgment criterion. The specific steps are as follows:

Step C1), define the system for normal operation of the engine as:

y _N (k+1)＝y _N (k)+φ _N ^T (k)Δu _N (k)

Substituting this into the tracker gives

in

Step C2) After scaling, the

in

Step C3) introduces a sliding window mechanism to further reduce the threshold value and the minimum detection fault amplitude

Where n is the length of the sliding window, which is used to adjust the time span of the historical data used.

The sensor failure thresholds are:

d _{f_ad} (k)=(1-w)d _{f_slid} (k)

w＝1/(1+e ^-q(k) )

Where std is the standard deviation of sensor noise, is the error between the sensor measurement value and the baseline model prediction value. The sensor fault judgment criterion is:

The common logic of sensor fault diagnosis is to analyze whether the prediction residual exceeds a given threshold. When , the sensor failure of the aircraft engine can be detected at k.