CN103963938B - A kind of based on oblique rudder ship anti-pitchstabilizer intelligent adaptive Predictive Control System - Google Patents

Abstract

The present invention relates to a kind of oblique rudder ship anti-pitchstabilizer control method.Based on oblique rudder ship anti-pitchstabilizer intelligent adaptive Predictive Control System, swing including hanging down with pitching reference information maker, input reference locus maker, information constrained system, controlled quentity controlled variable fuse information maker, controlled quentity controlled variable estimator, the fuse information maker of association's state, association's state sequence estimator, association's state fusion quantity of information against time series initial value estimator, association's state against time series initial value estimator, sensing system, controller.The present invention is applicable to oblique rudder ship anti-pitchstabilizer and so has non-linear, the system of coupling, time variation, and control accuracy is high, and robustness is good.

Description

A kind of based on oblique rudder ship anti-pitchstabilizer intelligent adaptive Predictive Control System

Technical field

The present invention relates to a kind of oblique rudder ship anti-pitchstabilizer control method.

Background technology

Oblique rudder ship anti-pitchstabilizer multi-variable system, due to non-linear, strong coupling, time stickiness and Parameter uncertainties etc. Characteristic so that it is the design of control system has certain challenge.

Ship stabilization and one of important subject that control is Naval Architecture and Ocean Engineering.At present rollstabilization aspect achieved with Many achievements in research.In terms of boats and ships pitching stabilization, achievement in research is little, and they are to use half submerged body and fin the pitching stabilization such as to combine. Rudder stabilization has many advantages compared with stabilizer, tank stabilizer.It can utilize existing rudder and steerable system on ship, only needs Increase a few components such as controller, sensor, not only economy but also practical.Premise is that boats and ships are to liking Twin Rudders or many rudders.

Document " Chen Hongli, Zhao Xiren, Ye Kui, Peng Xiuyan. the LQG technique study that boats and ships pitching stabilization controls. Harbin Institute of Engineering College journal .2004,25 (4): 407-411 " have ignored the coupling between rolling, pitching in, utilize ∏ type hydroplane pitching stabilization, And it is to carry out linear quadratic gaussian (LQG) for permanent equation under Longitudinal Movement of Ship typical case's sea situation to control.The present invention uses 45 ° tilt the flap-type rudders installed, it is possible to achieve rolling, the uneoupled control of pitching, improve lift of rudder coefficient, and pitching subtract when shaking Do not produce driftage control moment.

Many scholars, by research to rudder stabilization for many years, find that the uncertain problem of boats and ships motion model becomes left and right One of key factor of rudder stabilization technology application success or failure.

The present invention utilizes strip theory according to the tank experiments test result of hydrodynamic force coefficient under typical case's sea situation, it is established that with The ship control intelligent adaptive model of ship course (angle), the speed of a ship or plane and sea condition real-time change, vertical in conjunction with subtracting based on oblique rudder ship Shake device information fusion forecasting and control the robustness to improve controller to suppress ship model uncertainty to pitching stabilization effect Impact.Thus ensure that the safe and reliable operation of oblique rudder ship anti-pitchstabilizer.

Summary of the invention

Present invention aims to oblique rudder ship anti-pitchstabilizer and there is serious coupling with non-linear, it is provided that be a kind of Be suitable for variable working condition based on tandem Combination neural network model, the oblique rudder ship anti-pitchstabilizer of good control performance can be obtained Self-adaptation control method.

The object of the present invention is achieved like this:

Based on oblique rudder ship anti-pitchstabilizer intelligent adaptive Predictive Control System, swing raw with pitching reference information including hanging down Grow up to be a useful person, input reference locus maker, information constrained system, controlled quentity controlled variable fuse information maker, controlled quentity controlled variable estimator, association's state Fuse information maker, association state sequence estimator, association state fusion quantity of information against time series initial value estimator, association state Inverse time series initial value estimator, sensing system, controller.Input reference locus maker receives to hang down and swings with pitching with reference to letter Obtain after the data of breath maker and sensing system inputting reference locus, by inverse to association's state for input reference locus transmitted in parallel Time series initial value estimator and association's state sequence estimator；Controlled quentity controlled variable fuse information maker receives the number of information constrained system Obtain the fuse information amount controlled after according to, the fuse information amount of control is passed to controlled quentity controlled variable estimator；Sequence between association's state inverse time Row initial value estimator receives association's state fusion quantity of information and obtains between association's state inverse time after the data of time series initial value estimator Sequence initial estimate amount, issues association's state sequence estimator, simultaneously association's state sequence by association's state against time series initial estimate amount Estimated sequence between the inverse time that row estimator obtains association's state after receiving the data of the fuse information maker of association's state, controlled quentity controlled variable is estimated Optimum control amount sequence is obtained after estimated sequence data between the inverse time that gauge receives association's state；Last controller is by control instruction Being sent to the actuator steering wheel of boats and ships, the pitching realizing waterborne vessel controls.

The oblique rudder ship anti-pitchstabilizer intelligent adaptive forecast Control Algorithm of the present invention can also include:

It is that boats and ships hang down and swing and pitch angle that described oblique rudder ship anti-pitchstabilizer vertical swings with pitching reference information maker Reference information is supplied to definite value (expected value).

The sensing system of described oblique rudder ship anti-pitchstabilizer refers to, under certain sea condition, gathers the actual boat of boats and ships To, the speed of a ship or plane, rudder angle, hang down and swing displacement transducer with pitch angle, accelerometer, gyroscope, log, boat-carrying theodolite.

The input reference locus maker of described oblique rudder ship anti-pitchstabilizer is

${Y^{*}}_j={\mathrm{\Ω}}^{j-k}{Y}_k+{(1-\mathrm{\Ω})}^{j-k}{\stackrel{\‾}{R}}_k$

In formula, Y_kFor the boats and ships of sensing system collection actual hang down swing, pitch angle, Y_jFor to Y_k(j-k) forward step pre- Survey,Swinging and pitching reference information set-point for hanging down, Ω ∈ (0,1) is softening coefficient, and choosing of Ω need to consider that system responds Aggregate balancing between characteristic and robustness.

The information constrained system of described oblique rudder ship anti-pitchstabilizer, comprises following 3 information representation formulas:

The hard constraint information representation formula determined by system state equation is

${\stackrel{\‾}{X}}_j=\stackrel{\‾}{\mathrm{\Φ}}{\stackrel{\‾}{X}}_{j-1}+\stackrel{\‾}{\mathrm{\Ψ}}{u}_{j-1}+\stackrel{\‾}{\mathrm{\Γ}}{W}_{f,j-1}$

In formulaFor system state equation coefficient matrix, W_f,j-1It is I for zero-mean and covariance matrix_2×1White noise.

The soft-constraint information representation formula determined by association's state estimation error is:

${\widehat{\stackrel{\‾}{X}}}_j={\stackrel{\‾}{X}}_j+E_j$

E in formula_jFor zero-mean and covariance matrix it isWhite noise.

By requiring that controlling the least the determined soft-constraint information representation formula of energy is:

0=u_j-1+n_j-1

N in formula_j-1It is R for zero-mean and covariance matrix_j-1 ^-1White noise.

By the soft-constraint information representation formula requiring that system output tracking input reference locus is determined

$Y_j^{*}=Y_j+{\stackrel{\‾}{M}}_j$

In formulaFor zero-mean and covariance matrix in order toWhite noise.

The controlled quentity controlled variable fuse information maker of described oblique rudder ship anti-pitchstabilizer is

In formulaCovariance matrix.

Association's state fusion quantity of information of described oblique rudder ship anti-pitchstabilizer against time series initial value estimator is:

$P_{k+N}^{-1}={\stackrel{\‾}{D}}^T{Q}_{k+N}\stackrel{\‾}{D}+{\stackrel{\‾}{\mathrm{\Φ}}}^T{(I_{n\×n}/\mathrm{\λ}+\stackrel{\‾}{\mathrm{\Ψ}}{R}_{k+N}^{-1}{\stackrel{\‾}{\mathrm{\Ψ}}}^T)}^{-1}\stackrel{\‾}{\mathrm{\Φ}}$

In formulaFor systematic observation matrix, λ is sufficiently small positive nonzero number, I_n×nFor the unit matrix of n × n dimension, Q_jForCovariance matrix, V_jFor systematic survey noise covariance matrix.

Association's state of described oblique rudder ship anti-pitchstabilizer against time series initial value estimator is

${\widehat{\stackrel{\‾}{X}}}_{k+N}=P_{k+N}\{-{\stackrel{\‾}{\mathrm{\Φ}}}^T{(I_{n\×n}/\mathrm{\λ}+\stackrel{\‾}{\mathrm{\Ψ}}{R}_{k+N}^{-1}{\stackrel{\‾}{\mathrm{\Φ}}}^T)}^{-1}\stackrel{\‾}{\mathrm{\Ψ}}+{\stackrel{\‾}{D}}^T{Q}_{k+N}{Y}_{k+N}^{*}\}$

The fuse information maker of association's state of described oblique rudder ship anti-pitchstabilizer is

In formula,

Association's state sequence estimator of described oblique rudder ship anti-pitchstabilizer is

${\widehat{\stackrel{\‾}{X}}}_j=P_j[{\stackrel{\‾}{\mathrm{\Φ}}}^T{O}_j^{-1}({\widehat{\stackrel{\‾}{X}}}_{j+1}-\stackrel{\‾}{\mathrm{\Ψ}})+{\stackrel{\‾}{D}}^T{Q}_j{Y}_j^{*}]$

J=k+1 in formula ..., k+N.

The controlled quentity controlled variable estimator of described oblique rudder ship anti-pitchstabilizer is

The beneficial effects of the present invention is:

The present invention uses 45 ° to tilt the flap-type rudder installed, and establishes based on strip theory according to hydrodynamic force under typical case's sea situation The hydrodynamic force nonlinear adaptive model of mind (tandem Combination neural network model) of the tank experiments test result of coefficient.According to Instrument on ship, can record course angle, the speed of a ship or plane and the sea condition in change in real time, utilizes hydrodynamic force self-adapting intelligent model to count online Calculate the hydrodynamic force coefficient of oblique rudder ship anti-pitchstabilizer, constantly update oblique rudder ship anti-pitchstabilizer model, then by nerve Network self-adapting model is combined with PREDICTIVE CONTROL, forecast model output error caused by procedure parameter slow time-varying can obtain and Time correction, thus improve the dynamic property of system, it is adaptable to oblique rudder ship anti-pitchstabilizer so has non-linear, coupling Property, the system of time variation, control accuracy is high, and robustness is good.

Accompanying drawing explanation

Fig. 1 is the structured flowchart of oblique rudder ship anti-pitchstabilizer intelligent adaptive information fusion Predictive Control System；

Fig. 2 is oblique rudder control figure；

Fig. 3 is generalized regression nerve networks structure chart；

Fig. 4 is that optimal stopping rule optimizes network program block diagram；

Fig. 5 extends system block diagram；

Fig. 6 is tandem combination neural net structure chart；

Fig. 7 is for controlling effect relative value.

Specific implementation method

Below in conjunction with the accompanying drawings the present invention is described further.

The intelligence that the present invention is to provide a kind of oblique rudder ship anti-pitchstabilizer based on information fusion optimal estimation is adaptive Answer Predictive Control System.Under certain sea condition, sensing system gathers boats and ships actual heading, the speed of a ship or plane, rudder angle, hangs down and swings and pitching Angle, hangs down in conjunction with boats and ships and swings and pitching reference information maker, generate input reference locus；By Quadratic Optimal Control performance indications Function is converted into be followed the tracks of input and controls the soft-constraint information of energy system, the hard constraint information of coupling system state equation, All information of fused controlling amount, obtain the association's status switch under rolling optimization performance indications and control the optimal estimation of sequence； Evaluate the control effect of controlled quentity controlled variable according to control performance evaluation index, and revise current control according to Actual Control Effect of Strong Amount, finally selects the control input of the oblique rudder ship anti-pitchstabilizer of optimum, passes to controller and generate control instruction, drive rudder Machine, the pitching stabilization realizing waterborne vessel controls.

The operation principle of the present invention is: the core of the present invention is that oblique rudder ship anti-pitchstabilizer intelligent adaptive is advanced The design of control system.This system is oblique rudder ship anti-pitchstabilizer intelligent adaptive integrated model and information fusion PREDICTIVE CONTROL The combination of strategy.

Oblique rudder ship anti-pitchstabilizer intelligent adaptive Predictive Control System is many based on intelligent adaptive integrated model Step prediction device is the following output of controlled system under future reference track effect, and according to rolling optimization criterion, combining information merges Predictive control strategy, obtains control performance evaluation index, evaluates the control effect of controlled quentity controlled variable according to control performance evaluation index, and Revising current controlled quentity controlled variable according to Actual Control Effect of Strong, the control finally selecting the oblique rudder ship anti-pitchstabilizer of optimum is defeated Enter.As it is shown in figure 1, oblique rudder ship anti-pitchstabilizer intelligent adaptive advanced control method of the present invention, mainly by based on intelligence certainly The link compositions such as the multistep prediction device of adaptation integrated model and information fusion PREDICTIVE CONTROL.In Fig. 1,J for state Step predictive value, wherein 1≤j≤N+1, N is prediction step；Y^*(k+j) it is reference locus, tandem combination neural net and oblique rudder ship Oceangoing ship anti-pitchstabilizer composition intelligent adaptive integrated model.

The first step utilizes oblique rudder to realize the uneoupled control in ship rolling, pitching and course

Rudder and horizontal plane installation at 45 °, during roll damping (Fig. 2 (a)), left rudder guide margin upwards time, right standard rudder guide margin is downward, left The pitchmoment that right standard rudder causes is in opposite direction, cancels out each other；Pitching subtracts when shaking (Fig. 2 (b)), and left and right rudder guide margin in the same direction, only produces Vertically upward make a concerted effort, in the horizontal direction left and right rudder produce component cancel out each other, thus, rolling control and pitching control realization Natural decoupling. pitching simultaneously subtracts and does not produce driftage control moment when shaking.

The foundation of the second step Longitudinal Movement of Ship differential equation

The Longitudinal Movement of Ship differential equation is set up based on strip theory

Document [Chen Hongli. study [D] based on π type rudder ship longitudinal direction multivariate Random Control Method. Harbin: Harbin Engineering University Ph.D. Dissertation, 2004.12] utilize strip theory, according to Newton's second law, establish and make at irregular wave Boats and ships under with hang down coupled motions (Longitudinal Movement of Ship) differential equation swung with pitching:

$\begin{array}{c}(a_{33}+\mathrm{\Δ})\stackrel{\·\·}{z}+b_{33}\stackrel{\·}{z}+c_{33}z+a_{35}\stackrel{\·\·}{\mathrm{\θ}}+b_{35}\stackrel{\·}{\mathrm{\θ}}+c_{35}\mathrm{\θ}=F_R+F_{w3}\\ a_{53}\stackrel{\·\·}{z}+b_{53}\stackrel{\·}{z}+c_{53}z+(I_5+a_{55})\stackrel{\·\·}{\mathrm{\θ}}+b_{55}\stackrel{\·}{\mathrm{\θ}}+c_{55}\mathrm{\θ}=F_R{X}_R+M_{w3}\end{array}---\left(1\right)$

In formula:

Z represents that the vertical of hull swings, m

θ represents the pitching of hull, rad

Pitching the moment of inertia I₅,kg.m²

Displacement Δ, kg

F_RRepresent lift of rudder, kg.m/s²

X_RRepresent that lift of rudder center is to the fore-and-aft distance of hull center of gravity, m

F_w3Represent that wave hangs down and swing perturbed force, kg.m/s²

M_w5Represent wave pitching disturbance torque, kg.m²/s²

a₃₃Expression additional mass, kg,

a₃₅Expression moment of mass, kg.m,

a₅₃Expression moment of mass, kg.m,

a₅₅Represent rotary inertia, kg.m²,

b₃₃Represent damping force coefficient,

b₃₅Represent damping force coefficient, $b_{35}=V{\∫}_{H_s}^{H_b}{A}_{33}^{\′}\mathrm{dx}-{\∫}_{H_s}^{H_b}{N}_{22}^{\′}\mathrm{xdx}$

c₃₃Represent recuperability coefficient,

c₃₅Represent recuperability coefficient, $c_{35}=V{\∫}_{H_s}^{H_b}{N}_{33}^{\′}\mathrm{dx}-\mathrm{\ρg}{\∫}_{H_s}^{H_b}{B}_{22}^{*\′}\mathrm{xdx}$

b₅₃Represent damping moment coefficient,

b₅₅Represent damping moment coefficient,

c₅₃Represent restoring moment coefficient,

c₅₅Represent restoring moment coefficient, $c_{55}=\mathrm{\ρg}{\∫}_{H_s}^{H_b}{B^{*}x}^2\mathrm{dx}-V{\∫}_{H_s}^{H_b}{N}_{22}^{\′}\mathrm{xdx}-V^2{\∫}_{H_S}^{H_b}{A}_{33}^{\′}\mathrm{dx}$

B^*Represent boats and ships locality cross-sectional width, m

V represents speed of the ship in metres per second, kn

H_bRepresent the coordinate of bow, m

H_sRepresent the coordinate of stern, m

A′₃₃Represent the vertical additional mass in cross section,

N′₂₂Represent boats and ships local heave damped coefficient,

Can be calculated by Griem collection of illustrative plates.

Wherein a_ij, b_ij, c_ij(i, j=3 or 5) are referred to as hydrodynamic force coefficient, wherein c_ijCan be by being calculated and being Constant.a_ij,b_ijUse w_ijRepresent, obtain by formula (2):

$w_{\mathrm{ij}}=\left(\underset{k=1}{\overset{30}{\mathrm{\Σ}}}\sqrt{2s\left({\mathrm{\ω}}_k\right){\mathrm{d\ω}}_k}\right|w_{\mathrm{ijk}}\left|\right)/\underset{k=1}{\overset{30}{\mathrm{\Σ}}}\sqrt{2s\left({\mathrm{\ω}}_k\right){\mathrm{d\ω}}_k}---\left(2\right)$

According to strip theory, the regular wave disturbance sink experiment at the different speed of a ship or plane, course and different frequency is utilized to measure Data calculate to obtain w_ijk。

S (ω) is irregular wave wave spectrum, uses 15^thThe two-parameter spectrum of ITTC,Its Middle H_1/3There are adopted wave height, T₁Averagely spending null cycle, they are by different sea condition values.

Utilizing formula (2) to calculate hydrodynamic force coefficient is to change with sea condition, the speed of a ship or plane and course but the typical case sea unrelated with frequency Hydrodynamic force coefficient a under condition_ij,b_ij.In order to control Ship Motion Attitude precisely in real time, one need to be set up based on boat Speed, sea condition and the boats and ships motion 3 D intelligent model of course angle real-time change, the hydrodynamic force coefficient in the equation of motion is three-dimensional space Continually varying nonlinear function between.Approaching these Nonlinear Curved quickly and accurately is to set up the key of model of mind.

2. the structure of generalized regression nerve networks and principle

Utilize radially base neuron and linear neuron can set up generalized regression nerve networks (Generalized Regression Neural Network-GRNN), the network of this form is frequently used for function approximation.GRNN structure such as Fig. 3 Shown in.This network includes that two-layer, intermediate layer (hidden layer) are radial direction basic unit, and output layer is linear layer.

GRNN shown in Fig. 3 has Q group input vector, and the element number often organizing vector is m, and there is n radially base in intermediate layer Neuron.

In order to avoid crossing the phenomenon of study, improve training speed and the generalization ability of GRNN, utilize optimal stopping rule to net Network is optimized.

Flow chart when Fig. 4 is for utilizing optimal stopping rule, wherein parameter: training set T, checking collection C, current authentication error Errornew, previous validation error Erroroed, weight w_ij,Enumerator count, fault-tolerant thresholding S, frequency of training flag bit X0, frequency of training k, training error E, error cost function ε, cycle-index R, initial weightTry to achieve by gradient descent method Weights variation delta w_ij。

3. the foundation of Longitudinal Movement of Ship hydrodynamic force model of mind

Boats and ships longitudinal direction hydrodynamic force coefficient is pressed sea condition, speed of a ship or plane refinement packet, uses generalized regression nerve networks to set up hydrodynamic(al) The tandem Combination neural network model of force coefficient.

3rd step, based on oblique rudder ship anti-pitchstabilizer information fusion PREDICTIVE CONTROL.

Oblique rudder ship pitching stabilization separate manufacturing firms model based on wave disturbance modeling

Utilize Longitudinal Movement of Ship differential equation (1), set up state equation and observational equation and discretization (takes sampling week Phase T_s=0.2 second), utilize hydrodynamic force model of mind to obtain hydrodynamic force coefficientOr 5, substitute into state equation and try to achieve The eigenvalue of system, it is judged that the stability of system, can control, controllability.

As follows based on oblique rudder ship pitching stabilization discrete time state equation

X_k=Φ X_k-1+Ψu_k-1+ΓW_k-1 (3)

Y_k=DX_k+V_k (4)

In formula: $X={\left[\begin{array}{cccc}{x}_1& x_2& x_3& x_4\end{array}\right]}^T={\left[\begin{array}{cccc}z& \stackrel{\·}{z}& \mathrm{\θ}& \stackrel{\·}{\mathrm{\θ}}\end{array}\right]}^T,$ U=F_R, Y=[x₁ x₃]^T, $D=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& 0& 1& 0\end{array}\right),$ V is two-dimensional measurement noise, W^T=[F_w3 M_w5] it is the random seaway disturbance to boats and ships, regard system interference as, subscript k is for strain mutually The value moment after amount discretization.

The spectrum density of wave disturbance (including perturbed force and disturbing moment) is a narrow-band spectrum, is considered a kind of tool There is the coloured noise of Gauss distribution, wave disturbance power and disturbing moment are composed the output regarding the formed filter that white noise drives as (output signal at this moment is no longer real wave disturbance, but the two has identical statistical property, is the most all zero-mean, Identical covariance and the coloured noise of same spectrum density), as shown in Figure 5.Wherein, w₃T () is wave disturbance power forming filter G₃ The input white noise signal of (s), w₅T () is wave disturbance moment forming filter G₅S the input white noise signal of (), they are equal Value is zero, and variance is 1；Y_fT () is to have the coloured noise of identical statistical property with wave disturbance, it is original that it is used for " replacement " Wave disturbance W (t),

The formed filter of boats and ships longitudinal direction perturbed force and disturbing moment is chosen as follows by wave, (ship control for convenience Systematic study and design and reduce the exponent number of institute's ships control system, the order of G (s) should not obtain the highest):

$G_i\left(s\right)=\frac{y_i\left(s\right)}{w_i\left(s\right)}=\frac{b_{i1}s}{s^2+a_{i1}+a_{i2}},i=\mathrm{3,5}---\left(5\right)$

Utilize wave disturbance to compose, obtain a by least square method of recursion_i1,a_i2,b_i1, i=3,5.

Transmission function above is converted to state space form, i.e.

${\stackrel{\·}{m}}_i=A_i{m}_i+C_i{w}_i---\left(6\right)$

y_i=H_im_i (7)

In formula: $A_i=\left[\begin{array}{cc}0& 1\\ -a_{i2}& -a_{i1}\end{array}\right],C_i=\left[\begin{array}{c}0\\ 1\end{array}\right],$ H_i=[0 b_i1], i=3,5.

I=3,5 is associated with

$\begin{array}{c}\stackrel{\·}{M}=A_{f}M+C_f{W}_f\\ Y_f=H_{f}M\end{array}---\left(8\right)$

In formula: M^T=[m₃ m₅], $W_f^T=\left[\begin{array}{cc}{w}_3& w_5\end{array}\right],Y_f^T=\left[\begin{array}{cc}{y}_3& y_5\end{array}\right],$ A_f=diag [A₃ Α₅],

Fig. 5 is equivalent to the series connection (referred to as extending system) of original system and formed filter, and after extension, the disturbance of new system is all White noise, i.e. its input is dimensional Gaussian white noise sequence, and output remains the measured value of state z and θ, and extension system is represented by

$\begin{array}{c}\stackrel{\·}{\stackrel{\‾}{X}}=\stackrel{\‾}{A}\stackrel{\‾}{X}+\stackrel{\‾}{B}u+\stackrel{\‾}{C}{W}_f\\ Y=\stackrel{\‾}{H}\·\stackrel{\‾}{X}+V\end{array}---\left(9\right)$

In formula: $\stackrel{\‾}{X}=\left[\begin{array}{c}X\\ M\end{array}\right],\stackrel{\‾}{A}=\left[\begin{array}{cc}A& {\mathrm{CH}}_f\\ 0& A_f\end{array}\right],\stackrel{\‾}{B}=\left[\begin{array}{c}B\\ 0\end{array}\right],\stackrel{\‾}{C}=\left[\begin{array}{c}0\\ C_f\end{array}\right],\stackrel{\‾}{H}=\left[\begin{array}{cc}H& 0\end{array}\right],$ Q_wff =diag [1 1].

Equation after discretization is

${\stackrel{\‾}{X}}_k=\stackrel{\‾}{\mathrm{\Φ}}{\stackrel{\‾}{X}}_{k-1}+\stackrel{\‾}{\mathrm{\Ψ}}{u}_{k-1}+\stackrel{\‾}{\mathrm{\Γ}}{W}_{f,k-1}---\left(10\right)$

$Y_k=\stackrel{\‾}{D}{\stackrel{\‾}{X}}_k+V_k---\left(11\right)$

Oblique rudder ship pitching stabilization information fusion PREDICTIVE CONTROL

If object function is:

$J=E[\underset{j=k+1}{\overset{k+N}{\mathrm{\Σ}}}{(Y_j-{Y^{*}}_j)}^2+\underset{j=k+1}{\overset{k+N_u}{\mathrm{\Σ}}}{\left({\mathrm{Ru}}_{j-1}\right)}^2]---\left(12\right)$

Wherein, N is prediction step, N_uFor controlling step-length, R is for controlling weighter factor, Y_jFor to Y_k(j-k) forward step pre- Survey, Y^* _jFor the input reference locus after filtered device softening, it is taken as::

${Y^{*}}_j={\mathrm{\Ω}}^{j-k}{Y}_k+{(1-\mathrm{\Ω})}^{j-k}{\stackrel{\‾}{R}}_k---\left(13\right)$

In formula,Swinging and pitching reference information set-point for hanging down, Ω ∈ (0,1) is softening coefficient, and choosing of Ω need to be examined Consider the aggregate balancing between system response characteristic and robustness.

The purpose that information fusion controls is by merging all information about controlled quentity controlled variable, estimating optimum control amount sequence Row, are presented herein below the information fusion SEQUENTIAL ALGORITHM solving optimum control amount sequence.

About controlled quentity controlled variable, comprise following 3 information representation formulas:

The hard constraint information representation formula determined by system state equation is

${\stackrel{\‾}{X}}_j=\stackrel{\‾}{\mathrm{\Φ}}{\stackrel{\‾}{X}}_{j-1}+\stackrel{\‾}{\mathrm{\Ψ}}{u}_{j-1}+\stackrel{\‾}{\mathrm{\Γ}}{W}_{f,j-1}---\left(14\right)$

2) the soft-constraint information representation formula determined by association's state estimation error is:

${\widehat{\stackrel{\‾}{X}}}_j={\stackrel{\‾}{X}}_j+E_j---\left(15\right)$

3) by requiring that controlling the least the determined soft-constraint information representation formula of energy is:

0=u_j-1+n_j-1 (16)

E in formula_jFor zero-mean and covariance matrix it isWhite noise, n_j-1It is R for zero-mean and covariance matrix_j-1 ^-1 White noise.

Simultaneous formula (14) and (15) can obtain

${\widehat{\stackrel{\‾}{X}}}_j=\stackrel{\‾}{\mathrm{\Φ}}{\stackrel{\‾}{X}}_{j-1}+\stackrel{\‾}{\mathrm{\Ψ}}{u}_{j-1}+\stackrel{\‾}{\mathrm{\Γ}}{W}_{f,j-1}+E_j---\left(17\right)$

Formula (17) is further converted to

${\widehat{\stackrel{\‾}{X}}}_j=\stackrel{\‾}{\mathrm{\Φ}}{\stackrel{\‾}{X}}_{j-1}+\stackrel{\‾}{\mathrm{\Ψ}}{u}_{j-1}+{\mathrm{\ω}}_j---\left(18\right)$

ω in formula_jFor zero-mean and covariance matrix it isWhite noise

Controlled quentity controlled variable fuse information maker is

Controlled quentity controlled variable estimator is

About association's state against time series, comprise 4 information representation formulas altogether:, information table Reach in formula (14)～(16) and performance index function formula (12) by requiring that system output tracking input reference locus is determined Soft-constraint information representation formula

$T_j^{*}=Y_j+{\stackrel{\‾}{M}}_j---\left(21\right)$

In formulaFor zero-mean and covariance matrix it isWhite noise.

Simultaneous information representation formula (16), (18), in formula (20), the information fusion about association's status switch is estimated as

${\widehat{\stackrel{\‾}{X}}}_{j+1}-{\mathrm{\ω}}_{j+1}=\stackrel{\‾}{\mathrm{\Φ}}{\stackrel{\‾}{X}}_j+\stackrel{\‾}{\mathrm{\Ψ}}-\stackrel{\‾}{\mathrm{\Ψ}}{n}_j---\left(22\right)$

I.e.

${\widehat{\stackrel{\‾}{X}}}_j=\stackrel{\‾}{\mathrm{\Φ}}{\stackrel{\‾}{X}}_j+\stackrel{\‾}{\mathrm{\Ψ}}+{\stackrel{\‾}{V}}_j---\left(23\right)$

In formula,It is O for zero-mean and covariance matrix_jWhite noise, and meet

Secondly, simultaneous output equation formula (11) and information representation formula (21) obtain

In formulaFor zero-mean and covariance matrix for Q_j ^-1White noise.

About association's stateInformation representation formula include formula (23) and (25), the fuse information maker of association's state is

Then association's state sequence estimator is

${\widehat{\stackrel{\‾}{X}}}_j=P_j[{\stackrel{\‾}{\mathrm{\Φ}}}^T{O}_j^{-1}({\widehat{\stackrel{\‾}{X}}}_{j+1}-\stackrel{\‾}{\mathrm{\Ψ}})+{\stackrel{\‾}{D}}^T{Q}_j{Y}_j^{*}]---\left(27\right)$

J=k+1 in formula ..., k+N.

Association's status information and input reference locus information after following N step are the most unknown, i.e. assist State fusion estimationThen assist stateAbout association's stateQuantity of information beNow about association's state Fuse information amount beBut, due toPositive semidefinite matrix often, therefore, for ensureing letter Breath amount is nonsingular, takes (k+N+1) moment association's stateQuantity of informationWherein λ be sufficiently small non-zero just Number, I_n×nFor the unit matrix of n × n dimension, now, association's state fusion quantity of information against time series initial value estimator is

$P_{k+N}^{-1}={\stackrel{\‾}{D}}^T{Q}_{k+N}\stackrel{\‾}{D}+{\stackrel{\‾}{\mathrm{\Φ}}}^T{(I_{n\×n}/\mathrm{\λ}+\stackrel{\‾}{\mathrm{\Ψ}}{R}_{k+N}^{-1})}^{-1}\stackrel{\‾}{\mathrm{\Φ}}---\left(28\right)$

Association's state against time series initial value estimator is

${\widehat{\stackrel{\‾}{X}}}_{k+N}=P_{k+N}\{-{\stackrel{\‾}{\mathrm{\Φ}}}^T{(I_{n\×n}/\mathrm{\λ}+\stackrel{\‾}{\mathrm{\Ψ}}{R}_{k+N}^{-1}{\stackrel{\‾}{\mathrm{\Ψ}}}^T)}^{-1}\stackrel{\‾}{\mathrm{\Ψ}}+{\stackrel{\‾}{D}}^T{Q}_{k+N}{Y}_{k+N}^{*}\}---\left(29\right)$

1 uses generalized regression nerve networks to set up the tandem Combination neural network model of hydrodynamic force coefficient.

Before carrying out the training of neutral net, find out the erroneous point of experimental data first with a neutral net, i.e. The functional relationship of one neuron network simulation experimental data of training, utilizes neutral net output valve to substitute erroneous point thus to data It is modified, ensures that the Coefficient Space of model is close with reality.Owing to each input variable exists certain difference on the order of magnitude, Thus every one-dimensional input data must be done normalized.

As a example by 442T surface ship, the boats and ships longitudinal direction hydrodynamic force coefficient data that actual experiment measures, the most corresponding typical case Sea condition grade 3 grades, 4 grades, 5 grades, typical case the speed of a ship or plane 6 save, 12 joint, 18 joint, 24 joint, typical case course angle from 0 °～180 °, take every 15 ° One sub-value, altogether 13 data points, altogether 156 (=3 × 4 × 13) group data.

By to the actual checking computations of conventional method for normalizing and the observation of data distribution after processing, finally determine choose following One group of formula as normalization and the basic formula of renormalization: $\left\{\begin{array}{c}x=0.1+0.8(q-q_{\min })/(q_{\max }-q_{\min })\\ y=q_{\min }+(q_{\max }-q_{\min })(x-0.1)/0.8\end{array}\right.$

Wherein, q_max,q_minRepresenting the maximum in sample and minima respectively, q represents the actual value of sample, and x, y are respectively Represent the sample after normalization and the sample after renormalization.

By 156 groups of data refinement packets, with hydrodynamic force coefficient a₃₃As a example by, as shown in Figure 6, i.e. divide three grades of composition tandem combinations Formal modeling, in figure, GRNNi_j_k, i represent that i-stage, j represent that Hai Qing, k represent the speed of a ship or plane.

This network is divided three grades to form cascade thereof by 10 GRNN, and wherein the first order is made up of 4 GRNN, its input variable Being 1 neuron (course angle), output variable is the most all 1 (fixing sea condition, the fixing speed of a ship or plane, the hydrodynamic(al) corresponding to course angle Force coefficient)；The second level is made up of 4 GRNN, and its input variable is 2 neurons, and output variable is 2 neurons or 1 god Through unit's (input and output are all the hydrodynamic force coefficients corresponding to course angle)；The third level is made up of 2 GRNN, and its input variable is 2 Individual neuron, output variable is 1 neuron (being all the hydrodynamic force coefficient corresponding to course angle).

Test samples takes every kind of sea condition (3 grades, 4 grades, 5 grades) respectively, during the every kind of speed of a ship or plane (6 joints, 12 joints, 18 joints, 24 joints) 15 °, 120 °, 165 ° totally 36 groups, other situation (every kind of sea condition, during every kind of speed of a ship or plane 0 °, 30 °, 45 °, 60 °, 75 °, 90 °, 105 °, 135 °, 150 °, 180 °) it is training sample totally 120 groups.

Network is through training, inspection, and final tandem combination neural net institute established model maximum relative error is 0.95%.

Application optimal stopping rule is when modeling GRNN, and we are provided with a fault-tolerant thresholding S, prevents when certain or some Checking sample (signal) temporarily causes validation error and becomes big because of burst error, is mistakenly considered network and comes into excessively plan Closing, in this enforcement, fault-tolerant thresholding S takes 3.

The intelligent adaptive PREDICTIVE CONTROL of oblique rudder ship anti-pitchstabilizer

In enforcement, (20) formula state estimationObtain according to EKF filter method.Kalman filter recursion Equation is shown in formula (30)～formula (35).

$\begin{array}{c}\widehat{\stackrel{\‾}{X}}\left(0\right)=E\left[\stackrel{\‾}{X}\left(0\right)\right]={\mathrm{\μ}}_0\\ P\left(0\right)=\mathrm{Cov}[\stackrel{\‾}{X}\left(0\right)-{\mathrm{\μ}}_0,\stackrel{\‾}{X}\left(0\right)-{\mathrm{\μ}}_0]\end{array}---\left(30\right)$

${\widehat{\stackrel{\‾}{X}}}_{k/k-1}=\stackrel{\‾}{\mathrm{\Φ}}{\widehat{\stackrel{\‾}{X}}}_{k-1}+{\stackrel{\‾}{\mathrm{\Ψ}}u}_{k-1}---\left(31\right)$

$P_{k/k-1}=\stackrel{\‾}{\mathrm{\Φ}}{P}_{k-1}{\stackrel{\‾}{\mathrm{\Φ}}}^T+\stackrel{\‾}{\mathrm{\Γ}}{Q}_{\mathrm{wff}}{\stackrel{\‾}{\mathrm{\Γ}}}^T---\left(32\right)$

$K_k=P_{k/k-1}{\stackrel{\‾}{D}}^T{[\stackrel{\‾}{D}{P}_{k/k-1}{\stackrel{\‾}{D}}^T+Q_w]}^{-1}---\left(33\right)$

${\widehat{\stackrel{\‾}{X}}}_k={\widehat{\stackrel{\‾}{X}}}_{k/k-1}+K_k[Y_k-\stackrel{\‾}{D}{\widehat{\stackrel{\‾}{X}}}_{k/k-1}]---\left(34\right)$

$\begin{array}{c}{P}_k=[I-K_k\stackrel{\‾}{D}]P_{k/k-1}{[I-K_k\stackrel{\‾}{D}]}^T\\ +K_k{Q}_w{K_k}^T\end{array}---\left(35\right)$

Initial value takesMeasure noise V_kCovariance matrix

Q_vv=diag [20.3 × 10^-4 2.25×10^-6]

Operational configuration parameter (Hai Qing, the speed of a ship or plane, course angle) information inputs as tandem combination neural net, by tandem group Close neutral net and carry out identification oblique rudder ship anti-pitchstabilizer intelligent adaptive model.

Estimate with the intelligent adaptive integrated model that tandem combination neural net forms based on oblique rudder ship anti-pitchstabilizer Device, by merging all information about controlled quentity controlled variable, is provided optimal control sequence by controlled quentity controlled variable estimator.

During actual control, the fusion the most only solving current k moment control law is estimatedThe most only by sequence between the controlled quentity controlled variable inverse time Last component of rowAddition system, mainly comprising the following steps of algorithm:

Putting k=0, system brings into operation；

Initial parameter: softening coefficient Ω, it was predicted that step number N, maximum iteration time d, end time K is set；

Input system Setting signalWeight matrix Q is set_j,R_k,R_j, j=k+1, k+2 ..., k+N, by kalman Filtering recurrence equation is asked

Put iterations i=1；

The fusion solving association's status switch according to the inverse time orientation of formula (26)～(29) is estimatedAnd quantity of information

The fusion solving k moment controlled quentity controlled variable according to formula (19) and (20) is estimated

If iterations i >=d, then perform step 8)；Otherwise, i=i+1, return and ask step 5)；

If meeting k >=K, then system is out of service, otherwise k=k+1, returns step 3).

The present invention takes N=3, N_u=2, λ=10^-4, softening coefficient Ω=0.2, K=800, R=10^-2, Q=diag [3,900].

Table 1 is statistical result (4 grades of speed of a ship or plane 18kn course angles of sea condition of the control effect relative value of 1000 second time period 120o), pitching stabilization controls effect and reaches 74%, and hard over angle is 11o.

This example demonstrates that, owing to sea condition is complicated and changeable, the impact of the factor such as wind-engaging, wave and stream, the water of oblique rudder ship simultaneously The coefficient of impact changes constantly, and the present invention uses 45 ° to tilt the flap-type rudder installed, and establishes based on strip theory according to typical case sea Hydrodynamic force nonlinear adaptive model of mind (the tandem combination neural net mould of the tank experiments test result of hydrodynamic force coefficient under condition Type).According to instrument on ship, course angle, the speed of a ship or plane and the sea condition in change can be recorded in real time, utilize hydrodynamic force self-adapting intelligent model Online calculate the hydrodynamic force coefficient of oblique rudder ship anti-pitchstabilizer, constantly update oblique rudder ship anti-pitchstabilizer model, then will Neutral net adaptive model is combined with PREDICTIVE CONTROL, and forecast model output error caused by procedure parameter slow time-varying can obtain To revise timely, thus improve the dynamic property of system.Result shows that this example control method has good parameter robustness.

Claims (3)

1., based on an oblique rudder ship anti-pitchstabilizer intelligent adaptive Predictive Control System, swing and pitching reference information including hanging down Maker (1), input reference locus maker (2), information constrained system (3), controlled quentity controlled variable fuse information maker (4), control Measure estimator (5), the fuse information maker (6) of association's state, association's state sequence estimator (7), assist state fusion quantity of information inverse Time series initial value estimator (8), association's state are against time series initial value estimator (9), sensing system (10), controller (11), it is characterised in that: input reference locus maker (2) receives to hang down swings and pitching reference information maker (1) and sensor Obtain after the data of system (10) inputting reference locus, reference locus transmitted in parallel will be inputted to association's state against time series initial value Estimator (9) and association's state sequence estimator (7)；Controlled quentity controlled variable fuse information maker (4) receives the number of information constrained system (3) Obtain the fuse information amount controlled after according to, the fuse information amount of control is passed to controlled quentity controlled variable estimator (5)；Between association's state inverse time Sequence initial value estimator (9) receives association's state fusion quantity of information and obtains assisting shape after the data of time series initial value estimator (8) Association's state, against time series initial estimate amount, is issued association's state sequence estimator (7) against time series initial estimate amount by state, with Time association's state sequence estimator (7) receive the data of the fuse information maker (6) of association state after obtain association's state inverse time between Estimated sequence, obtains optimum control amount sequence after estimated sequence data between the inverse time that controlled quentity controlled variable estimator (5) receives association's state, Pass to controller (11) and generate control instruction；Control instruction is sent to the actuator rudder of boats and ships by last controller (11) Machine, the pitching stabilization realizing waterborne vessel controls.

One the most according to claim 1 based on oblique rudder ship anti-pitchstabilizer intelligent adaptive Predictive Control System, its It is characterised by: it is that boats and ships hang down and swing and pitch angle that described oblique rudder ship anti-pitchstabilizer vertical swings with pitching reference information maker Reference information is supplied to definite value；The sensing system of described oblique rudder ship anti-pitchstabilizer refers to, collection boats and ships actual heading, The speed of a ship or plane, rudder angle, hang down and swing displacement transducer with pitch angle, gyroscope, log, boat-carrying theodolite.

One the most according to claim 1 based on oblique rudder ship anti-pitchstabilizer intelligent adaptive Predictive Control System, its It is characterised by: the input reference locus maker of described oblique rudder ship anti-pitchstabilizer is:

${Y^{*}}_j={\mathrm{\Ω}}^{j-k}{Y}_k+{(1-\mathrm{\Ω})}^{j-k}{\stackrel{\‾}{R}}_k$

In formula, Y_kFor the boats and ships of sensing system collection actual hang down swing, pitch angle, Y_jFor to Y_k(j-k) forward step prediction, For hang down swing and pitching reference information set-point, Ω ∈ (0,1) is softening coefficient, and Ω choose need to consider system response characteristic and Aggregate balancing between robustness.