The design and dispatching method of constrained forecast controller under a kind of aero-engine multi-state
Technical field
The present invention provides the designs and dispatching method of constrained forecast controller under a kind of aero-engine multi-state, belong to
System control and emulation field in Aerospace Propulsion Theory and Engineering.
Background technique
With the development of aero engine technology, aero-engine operating condition is increasingly complicated, and control system should protect
Demonstrate,prove engine and from a kind of working condition be steadily rapidly transitioned into another working condition, prevent again engine enter excess revolutions,
The abnormalities such as overtemperature, surge/stall.Switch the method combined to improve traditional linear PID controllers with Min-Max
Intrinsic conservative, Prediction and Control Technology because its can directly handle input and output constraint be applied to engine control system
Design can omit Min-Max switching, simplify original controller architecture.However, it is contemplated that aeroengine operation status is more, flight item
The wide feature of part, only the constrained forecast controller under single operating condition is it is difficult to ensure that promising result under all situations, according to existing
Document, designing the predictive controller for meeting the various operating conditions demands of engine generally has following several method: first is that passing through and being
System discrimination method, which constantly corrects prediction model, makes it match with aero-engine virtual condition, and still, there are some operating conditions not
Meeting can identification condition, it is possible to deteriorate identified parameters;Second is that nominal constrained forecast controller is designed under certain flying condition,
By the aero-engine principle of similitude by under the conditions of the parameter conversion to design point of other flying conditions, still, using conventional
Similarity criterion might not be applicable in all types engine;Third is that the flight envelope of aero-engine and working condition are carried out
It divides, nominal constrained forecast controller is designed to each subregion, while the scheduling designed between many constrained forecast controllers is patrolled
Volume, the control effect of this method is somewhat dependent on the reasonability of scheduling scheme.So far, without patent disclosure
Scheduling scheme of the constrained forecast controller under full flight envelope, multiple working conditions under multi-state.
Summary of the invention
In order to guarantee that aero-engine is attained by desired revolving speed under entire flight envelope, full working condition, while not
There is the problems such as overtemperature, excess revolutions, stall, surge, the present invention proposes constrained forecast controller under a kind of aero-engine multi-state
Design and the double-deck dispatching method.
The design and dispatching method of constrained forecast controller under a kind of aero-engine multi-state of the present invention, control system tool
There is double-layer structure, first layer is flight envelope dispatch layer, and scheduling parameter is flying height and Mach number, using fuzzy membership side
Method distributes the weight of nominal prediction controller, obtains the control amount under current flight conditions, multiple nominal operation states;The second layer
It is working condition dispatch layer, scheduling parameter is revolving speed, uses linear interpolation method to the control amount that first layer obtains, determines current
Final control amount under operating condition.
The design and dispatching method of constrained forecast controller under a kind of aero-engine multi-state, steps are as follows:
Step 1. determines the nominal operating condition of aero-engine
Aero-engine is in different flying height H, Mach number Ma, revolving speed NfIt flies under operating condition, corresponds to different start
Machine linear dynamic model;N is chosen in flight envelope1A nominal dot chooses N under working condition2A nominal dot, then correspond to N1N2
Kind nominal operating condition, wherein the nominal dot under working condition is from slow train NfTo maximum rating NfBetween choose at equal intervals, in flight envelope
Nominal dot determines that method is as follows:
According to aero-engine working principle, the output of aero-engine is only the function of flying height H and Mach number Ma,
And engine intake total temperature T1With stagnation pressure P1It is the function of H and Ma again, calculation formula is as follows:
As H≤11km
As H > 11km
Therefore, engine linear dynamic model and T1、P1Directly related, following formula (3) divides flight envelope,
Wherein, T10、P10And T1x、P1xNominal dot and import total temperature and stagnation pressure to reconnaissance in flight envelope are respectively referred to, if to
The root mean square of the variable quantity of the total temperature and stagnation pressure of reconnaissance and nominal dot is no more than ε, is considered as the engine of the state of flight to be selected
Linear dynamic model changes less compared with the model of nominal dot, incorporates into as same subregion, and a sub-regions have a mark
Claim point;The selection of ε size directly affects the control effect of system and the quantity of nominal dot, determines ε using trial and error procedure: to a certain
ε value arbitrarily selects state point to be emulated, if good results, amplifies ε and continue to verify in the boundary position of subregion, otherwise,
ε is reduced to continue to verify;
Step 2. designs N1N2Constrained forecast controller under the nominal operating condition of kind
For aero-engine demand for control, conventional linear restriction predictive control algorithm is improved, with not Tongfang
Formula processing tracking output quantity and constraint output quantity;
Using the aero-engine separate manufacturing firms model at certain nominal operating condition as prediction model:
Pass through augmented stateForm, formula (4) are expressed as augmented state form:
Interference, component degradation, it is non-linear due to, prediction model and present engine state are not exactly the same, exist
Model mismatch introduces feedback element and is modified, defines current k moment engine reality output yp(k) it is exported with prediction modelError bePrediction model outputY is exported including trackingt(k) and limitation exports yl
(k), prediction correction outputIt indicates are as follows:
Wherein, correction coefficient hi, i=1,2 ..., nyIt is selected between 0-1, takes h1=1, remaining hi<1;
Aero-engine control target is turn up to be adjusted under input and output constraint to desired value, and guarantee good dynamic
State quality, performance indicator formula (7):
Wherein, yr(k+j) be engine tracking output quantity expectation reference locus;WithPoint
Not Biao Shi the prediction of engine tracking amount and amount of restraint correct output;Indicate the changing value of control amount to be optimized,
Its weight coefficient λ is positive definite matrix;nyAnd nuRefer to prediction time domain and control time domain;U in constraint equationmax、uminWith Δ umax、Δ
uminIt is the maximin constraint of control amount and its rate of change respectively;yl maxAnd yl minRefer to the maximin of output quantity about
Beam;
It brings formula (4)~(6) into performance indicator formula (7), then constitutes the quadratic programming problem of a with constraint conditions, often
A sampling instant calls quadratic programming majorized function quadprog to be solved in Matlab, and by the first of control sequence
A amount Δ u (k) acts on controlled device;
Step 3. designs the double-deck dispatching method and coordinates N1N2A constrained forecast controller
The double-deck dispatching method is as follows: first layer is flight envelope dispatch layer, and scheduling parameter is H and Ma, using fuzzy membership
Degree method distributes N1The weight of a nominal prediction controller, obtains current flight conditions, N2Control under a nominal operation state
Amount;The second layer is working condition dispatch layer, and scheduling parameter is Nf, linear interpolation method is used to the control amount that first layer obtains,
Determine the final control amount under current working;It is specific as follows:
A. flight envelope dispatch layer
First layer is used as scheduling parameter with (H, Ma), and to Mr. Yu's nominal operation state, definition current flight conditions are (Hx,
Max), parameter (T can be surveyed by obtaining it by formula (1) or (2)1x、P1x), nominal state of flight 1,2 ..., N1Parameter respectively indicate
ForAccordingly the output quantity of constrained forecast controller isDefinition
Wherein,Respectively indicate current flight conditions and N1The close degree of a nominal state of flight point,
Its value is smaller to show that state is closer;
If J1Value be 0, enable the control amount W at current flight conditionsfx=Wf1,Similarly, if not being 0,
It enablesW at this timefxIs defined as:
The control amount of flight envelope dispatch layer, non-nominal flying condition will be by the defeated of all nominal constrained forecast controllers
It indicates, changes with flying condition out, weight shared by each nom inalcontroller gradually changes, so that control amount consecutive variations;
B. working condition dispatch layer
The second layer is with revolving speed NfIt is obtained under current flight conditions as scheduling parameter by flight envelope dispatch layer, N2A mark
Claim the control amount under working condition;For current working status Nfx, there are Nfk<Nfx<Nf(k+1), wherein NfkAnd Nf(k+1)Indicate with
NfxK-th adjacent and+1 nominal operation state of kth;WithIndicate kth under current flight conditions
Control amount under a and+1 nominal operation state of kth, the linear interpolation dispatching method that working condition dispatch layer uses are as follows:
Pass through the double-deck dispatching method rational management N1N2Constrained forecast controller under a nominal operating condition, obtains current flight
Control amount u under condition, working conditioncmd。
Beneficial effects of the present invention:
(1) under a kind of aero-engine multi-state proposed by the present invention predictive controller dispatching method, by flight envelope
Layer degree of membership dispatching method and working condition layer linear interpolation dispatching method blend, to constrained forecast controls many under different operating conditions
Device processed is scheduled, and expands the limitation protection control range of single constrained forecast controller under certain operating condition, it can be achieved that non-nominal work
Transition state control under stable state control under condition and the working condition wide variation in flight envelope.
(2) present invention can be in the constrained forecast controller for nominal operating condition design, directly consideration aero-engine control
Amount processed and its rate constraint, the output constraints such as temperature, surge margin, so that the Min-Max switch logic in conventional method is omitted,
It greatly simplifies the structure of former controller and avoids and saturation problem is integrated by switching bring.
(3) the double-deck dispatching method proposed by the present invention has expansibility, is applicable not only to many nominal prediction controllers,
The other types controller being also applied under multiple operating conditions.
Detailed description of the invention
Fig. 1 is the control area in flight envelope.
Fig. 2 is that the division of control area and nominal dot are chosen in flight envelope.
Fig. 3 is the double-deck dispatching method schematic diagram.
Fig. 4 is the fuel flow response under aeroengine operation status wide variation.
Fig. 5 is the rotating speed response under aeroengine operation status wide variation.
Fig. 6 is the temperature-responsive under aeroengine operation status wide variation.
Fig. 7 is the surge margin response under aeroengine operation status wide variation.
Fig. 8 is aero-engine flying condition variation track.
Fig. 9 is the rotating speed response under aero-engine flying condition wide variation.
Figure 10 is the fuel flow response (subregion scheduling method) of flight envelope dispatch layer.
Figure 11 is the fuel flow response (degree of membership scheduling method) of flight envelope dispatch layer.
Figure 12 is the fuel flow response (subregion scheduling method) of working condition dispatch layer.
Figure 13 is the fuel flow response (degree of membership scheduling method) of working condition dispatch layer.
Specific embodiment
Below in conjunction with attached drawing and technical solution, a specific embodiment of the invention is further illustrated.
The present embodiment is the design and dispatching method of constrained forecast controller under a kind of aero-engine multi-state.It is specific detailed
Thin design procedure is as follows:
Step 1. first divides flight envelope.According to aero-engine working principle, to certain control law,
Aero-engine output (such as HP&LP Rotor revolving speed and turbine blow down ratio) is only the function of flying height H and Mach number Ma,
And engine intake total temperature T1With stagnation pressure P1It is the function of height and Mach number again, calculation formula is as follows:
As H≤11km
As H > 11km
Therefore, engine linear model and T1、P1Directly related, following formula (3) divides flight envelope:
Wherein, T10、P10Index claims the import total temperature and stagnation pressure of state of flight, T1x、P1xIt is state to be selected in flight envelope
Total temperature and stagnation pressure, if the root mean square of the variable quantity of total temperature and stagnation pressure to reconnaissance and nominal dot be no more than ε, be considered as the state
The linear dynamic model of point changes less compared with the model of nominal dot, can incorporate into as same subregion, sub-regions tool
There is a nominal dot.The selection of ε size directly affects the control effect of system and the quantity of nominal dot, using trial and error procedure come really
Determine ε: to a certain ε value, arbitrarily selecting the state point situation of rather harsh (that is, in subregion) to carry out in the boundary position of subregion
Emulation, if good results, amplifies ε and continues to verify, and otherwise, reduces ε and continues to verify.By attempting, for duct big in this example
Than two shaft turbofan engine, ε≤0.2 is met the requirements.
Different operating condition (height H, Mach number Ma, revolving speed Nf) different engine linear models is corresponded to, therefore, if flight
N is chosen in envelope curve1A nominal dot chooses N under working condition2A nominal dot just shares N1N2Kind operating condition.Assuming that only considering flight
A control area (Fig. 1) in envelope curve, can divide three sub-regions by the above method, and obtain the nominal of each subregion
State of flight point, as shown in Figure 2.This example considers that working condition is 80%Nf~104%NfBetween great transition state process, between waiting
Every three working condition nominal dot 86%N of selectionf, 92%NfAnd 98%Nf, consider further that three kinds of flying condition marks in control area
Claim point, shares 9 kinds of operating conditions.
Step 2. separately designs corresponding constrained forecast controller according to engine demand for control, to above-mentioned 9 kinds of operating conditions, side
Method is as follows.Using the aero-engine separate manufacturing firms model at certain operating condition as prediction model:
Pass through augmented stateTraditional quadrature advantage is incorporated in linear prediction control by form,
At this point, formula (4) can be expressed as augmented state form:
In real engine control, usually due to interference, component degradation, non-linear etc., prediction model and current hair
Motivational state is not exactly the same, that is, there is model mismatch, needs to introduce feedback element to correct.Define current k moment engine
Reality output yp(k) it is exported with prediction model(including tracking output yt(k) and limitation exports yl(k)) error isPrediction correction outputIt may be expressed as:
Wherein, correction coefficient hi, i=1,2 ..., nyIt is selected between 0-1, usually takes h1=1, remaining hi<1。
Aero-engine control target is turn up to be adjusted under input and output constraint to desired value, and need to guarantee good
Dynamic quality, i.e. overshoot are small, response is fast, oscillation is less and steady reliable.A kind of performance indicator has following form:
Wherein, yr(k+j) be the following j step from current time reference locus, the phase of corresponding engine tracking output quantity
Hope responding trajectory;WithRespectively indicate the tracking amount and limit amount of the following j step from current time
Prediction correct output;Indicate that the prediction walked from future at current time i input variable quantity, weight coefficient λ are positive definite
Matrix;The limit upper value n of summation symbolyAnd nuRespectively indicate prediction time domain and control time domain, and n under normal circumstancesy≥nu.About
U in beam conditional expressionmaxAnd uminRespectively indicate the maximin limitation of control amount, Δ umaxWith Δ uminRespectively indicate control
Measure the maximin limitation of rate of change, mainly for characterization control device reality output by executing agency's extreme position and
The limitation such as rate of change;yl maxAnd yl minIt is the maximin constraint of output quantity respectively.
It brings formula (4)-(6) into performance indicator formula (7), may make up the quadratic programming problem of a with constraint conditions, often
A sampling instant calls quadratic programming majorized function quadprog to be solved in Matlab, and by the first of control sequence
A amount Δ u (k) acts on controlled device.
Step 3. considers 9 nominal operating points (in such as Fig. 3 shown in " * ") under flying condition and working condition, in step 2
On the basis of designed 9 nominal constrained forecast controllers, the double-deck scheduling scheme is designed, is coordinated at above-mentioned many nominal operating conditions
Constrained forecast controller, pass sequentially through flight envelope dispatch layer and working condition dispatch layer, realize steady under non-nominal operating condition
State and transition state control.Fig. 3 is with operating condition (H, Ma, Nf) three parameters are coordinate, illustrate the double-deck dispatching method of the invention: non-
Control amount and similar 1 at nominal m operating condition, 2,3,4,5,6 operating conditions corresponding 6 nominal constrained forecast controllers are related, first
Flight envelope dispatch layer is first passed through, determines operating condition m1(being determined by 4,5,6 nom inalcontrollers) and m2It (is determined by 1,2,3 nom inalcontroller
Control amount calmly);Then by working condition dispatch layer, by m1And m2Determine the final controlling value under m operating condition.
A. flight envelope dispatch layer
First layer is used as scheduling parameter with (H, Ma).With working condition 92%N in Fig. 3fFor illustrate flight envelope in adopt
Dispatching method.In 92%NfUnder working condition, for current flight conditions (Hx, Max), being obtained by formula (2) or (3) can
Survey parameter (T1x、P1x), the parameter of nominal state of flight 1,2,3 is expressed as (T11、P11), (T12、P12), (T13、P13), phase
The output quantity for answering constrained forecast controller is Wf1, Wf2, Wf3.Definition
Wherein, J1, J2, J3Respectively indicate the phase short range of current flight conditions and 3 nominal state of flight points in flight envelope
Degree, value is smaller to show that state is closer.
If J1Value be 0, enable Wfx=Wf1, J2And J3Similarly.If three is not 0, Q is enabled1=1/J1, Q2=1/J2, Q3
=1/J3, then current flight conditions (Hx, Max) at W to be askedfxIt may be expressed as:
If unknown point is located at the subregion that nominal dot 1 is covered, for Wf1CoefficientWith nominal dot 1
State is closer, Q1Bigger, the coefficient is also bigger, makes Wf1It plays a leading role, and is actually consistent.Assuming that unknown point and nominal dot 1
It is almost overlapped, then J1→ 0, Q1→ ∞, at this timeThen there is Wfx≈Wf1.When unknown point is located at nominal dot 2,3
It can also be led to the same conclusion in subregion.
The control amount of flight envelope dispatch layer, non-nominal flying condition will be by the defeated of all nominal constrained forecast controllers
It indicates, changes with flying condition out, weight shared by each nom inalcontroller gradually changes, so that control amount consecutive variations.
B. working condition dispatch layer
The second layer is with revolving speed NfAs scheduling parameter.For current working status Nfx, there are Nfk<Nfx<Nf(k+1), wherein Nfk
And Nf(k+1)Expression and NfxK-th adjacent and+1 nominal operation state of kth.By above-mentioned flight envelope dispatch layer, can be worked as
Control amount under preceding flying condition, under k-th and+1 nominal operation state of kthWithThis layer of institute
The linear interpolation dispatching method of use are as follows:
For this example, m is obtained by first layer1(Hx, Max, 86%Nf) and m2(Hx, Max, 92%Nf) control amount at operating condition,
The control amount u under final m operating condition is obtained by interpolationcmd。
In order to further illustrate the effect of dispatching method double-deck in the present embodiment, by two groups of emulation experiments, to verify this
The validity of method in invention.
(1) aeroengine operation status wide variation
Fig. 4-Fig. 7 is aero-engine in height H=11km, Mach number Ma=0.8 flying condition, 80%Nf- 104%Nf
Control effect under (4200r/min-5200r/min) working condition wide variation.As shown in Figure 5, based on the double-deck dispatching party
The constrained forecast controller of method can adjust turn up to desired value under input and output constraint, and transition state process is almost without super
It adjusts and regulating time is short, dynamic property is good.As shown in figure 4, control amount rate constraint acts as first during acceleration and deceleration
With control amount is increasedd or decreased with the amplitude of each control period 0.03kg/s, until touching other limitations.By Fig. 6, Fig. 7
It is found that shown limitation output quantity turbine-exit temperature T45With surge margin smHPC in the whole process all in respective constrained line
It is interior.The simulation example illustrates the case where double-deck dispatching method can cope with working condition wide variation.
(2) aero-engine flying condition wide variation
Fig. 8-Figure 13 is aero-engine in working condition 90%Nf, flying condition (H, Ma) be widely varied under control
Effect.The advantage of middle flight envelope layer scheduling method to illustrate the invention, this example to the degree of membership dispatching method of proposition with it is simpler
Single direct subregion dispatching method is compared, and wherein sub-district domain scheduling refers to if flying condition is in certain sub-regions
It is interior, then it is worked by the nominal prediction controller in the subregion.
Fig. 8 is flying condition variation track, and comparative diagram 2 is it is found that flight path crosses over two sub-regions.Pass through first first
The flight envelope dispatch layer of layer obtains the control amount at " m1 " and " m2 " operating condition, dispatches layer line further according to the working condition of the second layer
Property interpolation method obtains final control amount Wf.By taking " m2 " as an example, Figure 10 is subregion scheduling method, is determined by nom inalcontroller 1 or 2
Determine control amount, about at the 89s moment, causes controller to switch because nominal prediction controller 1 and 2 joins control, cause to control
Amount is jumped, and it is unstable (shown in Fig. 9) to directly result in rotating speed response;Figure 11 is degree of membership scheduling method of the invention, by three
A nominal prediction controller 1,2,3 codetermines control amount, control amount can consecutive variations, speed dynamic better effect (Fig. 9 institute
Show).Figure 12 and Figure 13 is responded by the working condition dispatch layer fuel flow of linear interpolation method, and Figure 12 shows flight envelope
The jump of dispatch layer control amount will have a direct impact on final control amount WfAlso it jumps.
This example has specifically carried out simulation and analysis to every layer of control amount variation in the double-deck dispatching method, it is known that flight envelope layer
Degree of membership dispatching method and working condition layer linear interpolation dispatching method can co-ordination, become control amount continuously
Change, control target is realized with preferable dynamic effect.