A kind of control system and controlling method excavating the acceleration potential of aeroengine
Technical field
The invention belongs to aeroengine and accelerate control technique field, relate to a kind of aeroengine accelerating process performance seeking control system, be specifically related to a kind of control system and the controlling method of excavating the acceleration potential of aeroengine.
Background technique
Now, the maneuverability requirement of the military fighter plane of high-performance to motor is more and more higher, and this is just objectively having higher requirement to motor acceleration-controlled system.In engine accelerating course, if can motor power be risen to maximum within the time short as far as possible, aircraft will be made to have better mobility, thus obtain better operational superiority.
Aeroengine is an extremely complicated thermal machine system, and it has the characteristics such as strong nonlinearity, multivariable, strong coupling.Along with the development of aero engine technology, the operating range of aeroengine is increasing, also more and more higher to the requirement of the acceleration performance of aeroengine, as fast etc. in wished that the thrust of time for acceleration short, motor promotes.For realizing above-mentioned target, require that modern aeroengine acceleration-controlled system possesses online optimizing control ability.And traditional mechanical hydraulic-pressure type controller have employed the method for subtense angle independent design and control, be difficult to realize vehicle air-conditioning.And along with the development of the present computer technology, based on motor Full Authority Digital electronic controller (FullAuthorityDigitalElectronicControl, FADEC), aeroengine can be realized and accelerate online optimizing and control.Aeroengine accelerating process Optimized-control Technique, for raising aero-engine performance, makes aircraft obtain better air fighting advantage, has great importance.
At 20 century 70s, linear-quadratic optimal control is applied to aeroengine and accelerates to control by existing American scholar.And when linear-quadratic optimal control being applied to the optimization control of aeroengine accelerating process, there is distinct disadvantage, linearization process need be carried out to engine mathematical model as adopted during the method, which results in model accuracy to reduce, control effects is had a greatly reduced quality, and robustness is poor, less model error just may cause system unstable.At present external disclosed about the research data of aeroengine accelerating process optimization control considerably less, and be mostly the data of some science popularization, some critical technical data are also unexposed, are difficult to recognize its core technology.
The domestic scholar of having devises with linear programming method, SQP (SequentialQuadraticProgramming, SQP) algorithm, genetic algorithm (GeneticAlgorithm, GA) scheduling algorithm are that the aeroengine of system optimizing control accelerates vehicle air-conditioning system.But there is following defect in these methods:
(1) in these methods, the outstanding advantages of linear programming method and SQP algorithm is that required amount of calculation is little, real-time is good, but when using these methods, need the nonlinear mathematical model using theory of similarity piecemeal linearization process motor in full flight envelope, this causes aeroengine model to occur linearized stability, causes its precision poor.In addition, these two kinds of methods are very responsive for initial solution, if initial solution arranges improper, easily cause algorithm to be absorbed in locally optimal solution and even do not restrain.
(2), when adopting the intelligent algorithm such as genetic algorithm, without the need to carrying out linearization to engine mathematical model and its effect of optimization is good, but the shortcoming of this kind of intelligent algorithm is to need a large amount of calculating just can converge to last solution, therefore its computational efficiency is low, poor real.
To sum up, not yet occur at present accelerating for aeroengine that online optimizing controls, the model error that engine mockup linearization process is brought can be avoided, there is again strict global convergence ability, not to the optimized algorithm of the characteristics such as initial solution sensitivity, high optimization efficiency (algorithm real-time can reach application request).
Summary of the invention
The object of the invention is to solve the problem, a kind of control system and the controlling method of excavating the acceleration potential of aeroengine are provided, this system directly adopts aeroengine nonlinear mathematical models as airborne model, any linearization process is not done to it, to improve the confidence coefficient of optimization control effect.
For achieving the above object, the present invention realizes by the following technical solutions:
Excavate a control system for the acceleration potential of aeroengine, comprise aeroengine, for control aeroengine Full Authority Digital electronic controller and for providing the airborne computer of optimal control policy for Full Authority Digital electronic controller; Aeroengine is provided with speed probe, temperature transducer and pressure transducer, the output terminal of each sensor is connected with airborne computer with Full Authority Digital electronic controller respectively; The output terminal of airborne computer is connected with Full Authority Digital electronic controller; Full Authority Digital electronic controller controls aeroengine according to optimal control policy by electromagnetic valve component.
The non-linear Real time mathematical model of processor module, optimal controller and motor is provided with in above-mentioned airborne computer; Described optimal controller and the non-linear Real time mathematical model of motor interact.
Above-mentioned optimal controller comprises acceleration optimal-search control algoritic module, optimization control object module, security constraint module and physical constraint module; Described acceleration optimal-search control algoritic module by calling the non-linear Real time mathematical model of motor, and according to optimization control object module, security constraint module and physical constraint module for Full Authority Digital electronic controller provides optimal control policy.
Above-mentioned electromagnetic valve component is connected with aeroengine with main fuel control unit respectively by geometric position adjusting part.
The invention also discloses a kind of controlling method excavating the acceleration potential of aeroengine, comprise the following steps:
1) to main combustion chamber fuel flow WFM, the nozzle of aircraft engine tail throat opening area A of aeroengine
8, aerial engine fan inlet guide vane corner α
1, aeroengine high-pressure compressor is adjustable stator blade corner α
2carry out floating-point encoding, stochastic generation one is by several individual colonies formed, and described individuality is by WFM, the A after encoding respectively
8, α
1and α
2composition;
2) optimal controller calls motor non-linear Real time mathematical model calculation procedure 1) in each ideal adaptation degree of obtaining, and according to fitness size order alignment step 1 from high to low) in the individuality that obtains, according to putting in order, select from high to low rank front 5% ~ 40% individuality, select from low to high simultaneously rank rear 5% ~ 40% individuality;
By the rank selected front 5% ~ 40% individuality directly remain into the next generation, simultaneously direct by rank rear 5% ~ 40% individuality remove from colony;
From rank Stochastic choice " parents " the individuality of front 5% ~ 40%, then carry out interlace operation, generate new individuality, with the new individuality generated, replace the individuality be removed;
3) being provided with body is one by one: X=[x
1, x
2, x
3, x
4]
t, x
1, x
2, x
3, x
4represent engine parameter WFM, A of this individuality of composition respectively
8, α
1and α
2if it is at x
k(k=1,2,3 or 4) point morphs, x
kspan be [x
kmin, x
kmax], wherein, x
kminand x
kmaxbe respectively x under the condition of existing physical unit
kthe minimum value that can get and maximum value, then the x in individuality
kby x '
kreplace, according to designed non-uniform mutation operating method, after morphing,
In formula, r is that computer random produces a numerical value, and this value is 0 or 1; γ is a random numbers of obeying non-uniform probability distribution in [0,1] scope, γ (x
kmax-x
kmin) α is for being applied to x
kon a noise signal, its average is 0; α is the standard deviation factor of this noise signal, as x '
kbe less than x
kminor be greater than x
kmaxtime, make its value be respectively x
kminand x
kmax; x
knoise standard deviation be Sd, then:
Sd=α(x
kmax-x
kmin);
4) through step 2) and step 3) operation after, form population of future generation, population of future generation carry out step 2 again) and the operation of step 3), then form population of new generation, repeatedly carry out, until evolutionary generation G>=G
tafterwards, genetic algorithm stops, and obtain the individuality that fitness is the highest, wherein, G is current evolutionary generation, G
tfor maximum evolutionary generation, 3≤G
t≤ 100;
5) individuality that fitness step 4) obtained is the highest inputs in standard SQP algorithm, and it can be used as the initial point of SQP algorithm to ask for final optimal control parameter, then inputed to by optimal control parameter in the Full Authority Digital electronic controller of aeroengine, Full Authority Digital electronic controller controls aeroengine according to optimal control parameter.
Above-mentioned steps 2) in the defining method of fitness size as follows:
In the individuality that step 1) produces, when individuality meets security constraint and physical constraint condition, the optimization control desired value J ' in each control cycle
iless individuality, its fitness is higher; When individuality does not meet security constraint and physical constraint, then the more close fitness meeting the individuality of security constraint and physical constraint condition is higher.
Above-mentioned steps 2) in optimization control desired value J ' in each control cycle
idefining method as follows:
Optimization control desired value J is:
In above formula, FN represents that the thrust that motor produces, T are total acceleration control time, and K is constant, its value is set to: K=1.5FN herein
max, FN
maxfor aeroengine is under maximum rating, when adopting Full Authority Digital electronic controller, producible maximum thrust.
Optimization control desired value J is handled as follows:
(1) in (3) formula, t
i=t
i-1+ τ, t
k=T, t are the optimization control time, t
ibe the initial time of i-th control cycle, FN
ibe the motor power of i-th control cycle, τ is the time that each unit control cycle continues;
The optimization control objective function in each control cycle is obtained according to formula (3):
minJ′
i=|K-FN
i|τ,(i=0,1,2,3...k)(4)
τ is normal number, makes | K-FN
i| for minimum, then the optimization control desired value J ' in each control cycle
ialso be minimum, therefore formula (4) become following form:
minJ′
i=|K-FN
i|,(i=0,1,2,3...k)。
The time T of will speed up is divided into limited unit control cycle, and setting τ is the time that each unit control cycle continues, then in i-th control cycle:
WFM
i, A
8i, α
1i, α
2isubmeter represents WFM, A
8, α
1, α
2at the parameter quantities of i-th control cycle; Δ WFM
i, Δ A
8i, Δ α
1i, Δ α
2irepresent WFM, A respectively
8, α
1, α
2the increment of i-th control cycle;
Above-mentioned steps 2) in, physical constraint and security constraint as follows:
In above formula, WFM
minand WFM
maxrepresent the minimum of WFM and KB limit, A respectively
8minand A
8maxrepresent A respectively
8minimum and KB limit, Δ WFM
maxwith Δ A
8maxrepresent respectively | Δ WFM
i| with | Δ A
8i| KB limit, FAR
maxrepresent KB limit, the n of main combustion chamber oil-gas ratio FAR
1maxrepresent rotational speed of lower pressure turbine rotor n
1kB limit, n
2maxrepresent high pressure rotor rotating speed n
2kB limit, P
31maxrepresent blower outlet stagnation pressure P
31kB limit, T
41maxrepresent turbine inlet temperature T
41kB limit, SMF
minrepresent minimum limit value, the SMC of fan stability margin SMF
minrepresent the minimum limit value of gas compressor stability margin SMC.
Above-mentioned steps 2) in the concrete steps of interlace operation as follows:
(1) two individualities in continuous random selecting colony match, until all individualities to be selected have matched, and then random selecting point of intersection on every a pair individuality;
(2) chromosome at the point of intersection place of the individuality of every a pair random pair is exchanged.
Compared with prior art, the present invention has following beneficial effect:
The present invention excavates the control system of the acceleration potential of aeroengine, directly using aeroengine nonlinear mathematical models as airborne model, the optimal control action quantity that Full Authority Digital electronic controller inputs according to optimal controller controls aeroengine, significantly promote and optimize computational efficiency, thus improve the real-time accelerating optimal-search control system.This aeroengine is accelerated optimal-search control system and fully can be excavated motor and accelerate potentiality, General Promotion engine acceleration Fa Dongjicongzhidingdituilizhuantaianquanxunsudiguodudaozhidinggaotuili energy, can many-side promote China's military fighter aircraft mobility, make it obtain air fighting advantage.Relative to traditional motor acceleration-controlled system, the present invention on the basis ensureing engine health work, significantly can reduce the time for acceleration and General Promotion motor power.
The present invention excavates the controlling method of the acceleration potential of aeroengine, on the basis of the model error avoiding engine mockup linearization process to bring, compared to current existing control algorithm, designed system optimizing control can on the basis ensureing optimization control effect, after introducing security constraint, significantly can promote the job security of engine accelerating course; Simultaneously, this controlling method is directly for the nonlinear mathematical model of aeroengine, there is global convergence ability, to the acceleration system optimizing control of the good characteristics such as initial solution sensitivity and higher optimization efficiency, compared with traditional control method, the present invention can avoid engine mockup linearized stability, significantly elevator system effect of optimization confidence coefficient.
Accompanying drawing explanation
Fig. 1 is the schematic diagram that aeroengine of the present invention accelerates optimal-search control system;
Fig. 2 is the fundamental diagram of GA-SQP hybrid optimization control algorithm of the present invention.
Embodiment
Below in conjunction with bright be described further of accompanying drawing to this:
See Fig. 1, a kind of control system excavating the acceleration potential of aeroengine of the present invention, it comprise aeroengine, for control aeroengine Full Authority Digital electronic controller and for providing the airborne computer of optimal control policy for Full Authority Digital electronic controller; Aeroengine is provided with speed probe, temperature transducer and pressure transducer, the output terminal of each sensor is connected with the non-linear Real time mathematical model of Full Authority Digital electronic controller and motor respectively; The non-linear Real time mathematical model of processor module, optimal controller and motor is provided with in airborne computer; Above-mentioned optimal controller and the non-linear Real time mathematical model of motor interact, and the output terminal of optimal controller is connected with Full Authority Digital electronic controller; Optimal controller comprises acceleration optimal-search control algoritic module, optimization control object module, security constraint module and physical constraint module; Described acceleration optimal-search control algoritic module by calling the non-linear Real time mathematical model of motor, and according to optimization control object module, security constraint module and physical constraint module for Full Authority Digital electronic controller provides optimal control policy; Full Authority Digital electronic controller controls aeroengine by electromagnetic valve component, and electromagnetic valve component is connected with aeroengine with main fuel control unit respectively by geometric position adjusting part.
The controlling method of the acceleration potential of the excavation aeroengine that the present invention is a kind of, adopt GA-SQP hybrid algorithm, it is made up of genetic algorithm and sequential quadratic programming algorithm, wherein, the Operational Limits of the GA accelerated for aeroengine in the GA-SQP hybrid algorithm of optimal-search control comprises population scale P, maximum evolutionary generation G
t, copy probability P
r, mutation probability P
mcomprise the following steps with each operating process of the GA in standard deviation factor α, GA-SQP hybrid algorithm:
1) processor module calls the non-linear Real time mathematical model of aeroengine, to main combustion chamber fuel flow WFM, the nozzle of aircraft engine tail throat opening area A of aeroengine
8, aerial engine fan inlet guide vane corner α
1, aeroengine high-pressure compressor is adjustable stator blade corner α
2carry out floating-point encoding, stochastic generation one is by several individual colonies formed, and described individuality is by WFM, the A after encoding respectively
8, α
1and α
2composition;
2) optimal controller calls motor non-linear Real time mathematical model calculation procedure 1) in each ideal adaptation degree of obtaining, and according to fitness size order alignment step 1 from high to low) in the individuality that obtains, according to putting in order, select from high to low rank front 5% ~ 40% individuality, select from low to high simultaneously rank rear 5% ~ 40% individuality;
By the rank selected front 5% ~ 40% individuality directly remain into the next generation, simultaneously direct by rank rear 5% ~ 40% individuality remove from colony;
From rank Stochastic choice " parents " the individuality of front 5% ~ 40%, then carry out interlace operation, generate new individuality, with the new individuality generated, replace the individuality be removed.
Wherein, the defining method of fitness size is as follows:
In the individuality that step 1) produces, when individuality meets security constraint and physical constraint condition, the optimization control desired value J ' in each control cycle
iless individuality, its fitness is higher; When individuality does not meet security constraint and physical constraint, then the more close fitness meeting the individuality of security constraint and physical constraint condition is higher.
Wherein, the optimization control desired value J ' in each control cycle
idefining method as follows:
Optimization control desired value J is:
In above formula, FN represents that the thrust that motor produces, T are total acceleration control time, and K is constant, its value is set to: K=1.5FN herein
max, FN
maxfor aeroengine is under maximum rating, when adopting Full Authority Digital electronic controller, producible maximum thrust.
Optimization control desired value J is handled as follows:
(1) in (3) formula, t
i=t
i-1+ τ, t
k=T, t are the optimization control time, t
ibe the initial time of i-th control cycle, FN
ibe the motor power of i-th control cycle, τ is the time that each unit control cycle continues;
The optimization control objective function in each control cycle is obtained according to formula (3):
minJ′
i=|K-FN
i|τ,(i=0,1,2,3...k)(4)
τ is normal number, makes | K-FN
i| for minimum, then the optimization control desired value J ' in each control cycle
ialso be minimum, therefore formula (4) become following form:
minJ′
i=|K-FN
i|,(i=0,1,2,3...k)。
The time T of will speed up is divided into limited unit control cycle, and setting τ is the time that each unit control cycle continues, then in i-th control cycle:
WFM
i, A
8i, α
1i, α
2isubmeter represents WFM, A
8, α
1, α
2at the parameter quantities of i-th control cycle; Δ WFM
i, Δ A
8i, Δ α
1i, Δ α
2irepresent WFM, A respectively
8, α
1, α
2the increment of i-th control cycle;
Physical constraint and security constraint as follows:
In above formula, WFM
minand WFM
maxrepresent the minimum of WFM and KB limit, A respectively
8minand A
8maxrepresent A respectively
8minimum and KB limit, Δ WFM
maxwith Δ A
8maxrepresent respectively | Δ WFM
i| with | Δ A
8i| KB limit, FAR
maxrepresent KB limit, the n of main combustion chamber oil-gas ratio FAR
1maxrepresent rotational speed of lower pressure turbine rotor n
1kB limit, n
2maxrepresent high pressure rotor rotating speed n
2kB limit, P
31maxrepresent blower outlet stagnation pressure P
31kB limit, T
41maxrepresent turbine inlet temperature T
41kB limit, SMF
minrepresent minimum limit value, the SMC of fan stability margin SMF
minrepresent the minimum limit value of gas compressor stability margin SMC.
The concrete steps of interlace operation are as follows:
(1) two individualities in continuous random selecting colony match, until all individualities to be selected have matched, and then random selecting point of intersection on every a pair individuality;
(2) chromosome at the point of intersection place of the individuality of every a pair random pair is exchanged.
3) being provided with body is one by one: X=[x
1, x
2, x
3, x
4]
t, x
1, x
2, x
3, x
4represent engine parameter WFM, A of this individuality of composition respectively
8, α
1and α
2if it is at x
k(k=1,2,3 or 4) point morphs, x
kspan be [x
kmin, x
kmax], wherein, x
kminand x
kmaxbe respectively x under the condition of existing physical unit
kthe minimum value that can get and maximum value, then the x in individuality
kby x '
kreplace, according to designed non-uniform mutation operating method, after morphing,
In formula, r is that computer random produces a numerical value, and this value is 0 or 1; γ is a random numbers of obeying non-uniform probability distribution in [0,1] scope, γ (x
kmax-x
kmin) α is for being applied to x
kon a noise signal, its average is 0; α is the standard deviation factor of this noise signal, as x '
kbe less than x
kminor be greater than x
kmaxtime, make its value be respectively x
kminand x
kmax; x
knoise standard deviation be Sd, then:
Sd=α(x
kmax-x
kmin);
4) through step 2) and step 3) operation after, form population of future generation, population of future generation carry out step 2 again) and the operation of step 3), then form population of new generation, repeatedly carry out, until evolutionary generation G>=G
tafterwards, genetic algorithm stops, and obtain the individuality that fitness is the highest, wherein, G is current evolutionary generation, G
tfor maximum evolutionary generation, 3≤G
t≤ 100;
5) individuality that fitness step 4) obtained is the highest inputs in standard SQP algorithm, and it can be used as the initial point of SQP algorithm to ask for final optimal control parameter, then inputed to by optimal control parameter in the Full Authority Digital electronic controller of aeroengine, Full Authority Digital electronic controller controls aeroengine according to optimal control parameter.
Working principle of the present invention is:
The control system that the present invention excavates the acceleration potential of aeroengine comprises optimal controller, Full Authority Digital electronic controller, airborne motor nonlinear mathematical model, electromagnetic valve component, geometric position adjusting part, main fuel control unit and sensor.Wherein, optimal controller forms by accelerating optimal-search control algorithm, optimization control target, the security constraint of motor and physical constraint, and this system optimizing control is a kind of GA-SQP hybrid algorithm merged by genetic algorithm and sequential quadratic programming algorithm; Sensor comprises: speed probe, temperature transducer, pressure transducer etc.; The signal of sensor measurement comprises: the air pressure signal of motor height rotor speed signal, engine portion partial cross-section and fuel gas temperature signal, jet nozzle throat opening area, aerial engine fan inlet guide vane corner and the adjustable stator blade corner of aeroengine high-pressure compressor etc.
Optimal controller is by accelerating optimal-search control algorithm, optimization control target, engine health constraint and physical constraint composition, itself and the non-linear Real time mathematical model of airborne motor mutually exchange data and calculate the optimal control action quantity that aeroengine accelerates to control, and this optimal control action quantity signal is input to Full Authority Digital electronic controller, Full Authority Digital electronic controller sends control action amount according to this optimal control action quantity signal and drives electromagnetic valve component action, make geometric position adjusting part and the running of main fuel control unit, thus (comprising: the main combustion chamber fuel flow WFM of aeroengine to the aeroengine input controlled quentity controlled variable corresponding with the optimal control action quantity that optimal controller exports, nozzle of aircraft engine tail throat opening area A
8, aerial engine fan inlet guide vane corner α
1, aeroengine high-pressure compressor is adjustable stator blade corner α
2).Sensor is connected with aeroengine, and the actual parameter of sensor measurement motor is also fed back in the non-linear Real time mathematical model of motor and Full Authority Digital electronic controller.
See the fundamental diagram that Fig. 2, Fig. 2 are GA-SQP hybrid optimization control algorithm of the present invention.In figure, G represents current evolutionary generation, G
trepresent maximum evolutionary generation.This optimization control process can be described below
1) carry out Initialize installation: make control cycle number i=0, the current evolutionary generation G of genetic algorithm is 1.
2) genetic algorithm is utilized to produce a colony at random.
3) fitness of each individuality in colony is calculated.By selecting operating process, reservation has the individuality of better fitness and is removed from colony by the individuality with poor fitness.After this, the more outstanding individuality that Stochastic choice retains, by interlace operation, produces " offspring " of these excellent individual and replaces the individuality that is removed with these " offsprings ".Finally by mutation operation, the characteristic of some individuality in colony is changed, thus forms new individuality.
4) after the operations of step 3), can form population of future generation, population of future generation carries out 3 again) described by operations, then form population of new generation.So repeatedly carry out, until evolutionary generation reaches the maximum evolutionary generation set in advance, genetic algorithm stops, and exports current optimal solution as last solution, i.e. " quasi-optimal solution ".
5) last solution that genetic algorithm exports after stopping is the quasi-optimal solution of first control cycle, after obtaining the quasi-optimal solution of first control cycle, this quasi-optimal solution is inputed in standard SQP algorithm, and it can be used as the initial point of SQP algorithm to ask for final optimal control parameter, then optimal control parameter is inputed to aeroengine.After this, the solution of second control cycle need be asked for again.Acquiring method is: initialization genetic algorithm, makes current evolutionary generation G be 1, then repeats step 2) to step 4), obtain the solution of second control cycle, so circulate, the optimal control parameter of new control cycle is constantly provided to aeroengine.
6), after aeroengine accelerating process terminates, the computing of GA-SQP hybrid algorithm terminates.The Optimal Curve of the control parameters of engine combined by the last solution of this k control cycle that Optimal Control System exports, is the geometric locus of the controlled variable with time for acceleration change.