文档库 最新最全的文档下载
当前位置:文档库 › A Gaussian Mixture PHD Filter for Nonlinear Jump Markov Models

A Gaussian Mixture PHD Filter for Nonlinear Jump Markov Models

A Gaussian Mixture PHD Filter for Nonlinear Jump Markov Models
A Gaussian Mixture PHD Filter for Nonlinear Jump Markov Models

A Gaussian Mixture PHD Filter for Nonlinear Jump Markov Models

Ba-Ngu V o?,Ahmed Pasha?,Hoang Duong Tuan?

?Department of Electrical and Electronic Engineering

The University of Melbourne,Parkville,VIC3052,Australia

?School of Electrical Engineering and Telecommunications

The University of New South Wales,Sydney NSW2052,Australia

Email:bv@https://www.wendangku.net/doc/b4503022.html,.au,s.pasha@https://www.wendangku.net/doc/b4503022.html,.au,h.d.tuan@https://www.wendangku.net/doc/b4503022.html,.au

Abstract—The probability hypothesis density(PHD)?lter is an attractive approach to tracking an unknown,and time-varying number of targets in the presence of data association uncertainty,clutter,noise,and miss-detection.The PHD?lter has a closed form solution under linear Gaussian assumptions on the target dynamics and births.However,the linear Gaussian multi-target model is not general enough to accommodate maneuvering targets,since these targets follow jump Markov system models.In this paper,we propose an analytic imple-mentation of the PHD?lter for jump Markov system(JMS) multi-target model.Our approach is based on a closed form solution to the PHD?lter for linear Gaussian JMS multi-target model and the unscented https://www.wendangku.net/doc/b4503022.html,ing simulations, we demonstrate that the proposed PHD?ltering algorithm is effective in tracking multiple maneuvering targets.

I.I NTRODUCTION

In a multi-target environment,the number of targets changes with time due to targets appearing,disappearing, and it is not known which target generated which measure-ment.Tracking multiple maneuvering targets involves jointly estimating the number of targets and their states at each time step.This problem is extremely dif?cult due to noise,clutter and uncertainties in target maneuvers,data association,and detection.

While non-maneuvering target motion can be described by a?xed model,a combination of motion models that characterize different maneuvers may be needed to describe the motion of a maneuvering target.The jump Markov system(JMS)model,or multiple models,approach has proven to be an effective tool for single maneuvering target tracking[1].In this approach the target can switch between a set of models in a Markovian fashion.The jump Markov model approach can also be combined with traditional data association techniques such as joint probabilistic data as-sociation(JPDA)or multiple hypothesis tracking(MHT) to track multiple maneuvering targets.However,these data association-based approaches are computationally intensive in general.

Mahler’s Probability Hypothesis Density(PHD)?lter[2] circumvents the combinatorial computations that arise from data association while accommodating detection uncertainty, This work is supported in part by the discovery grant DP0345215 awarded by the Australian Research Council.Poisson false alarms,target motions and time-varying num-

ber of targets.The generic sequential Monte Carlo implemen-

tation of the PHD?lter[3]can accommodate any Markovian

target dynamics including JMS models.However,the main

drawbacks of the particle approach are the large number of

particles,and the unreliability of clustering techniques for

extracting state estimates[3],[4].A closed form solution to

the PHD recursion was proposed for linear Gaussian multi-

target models in[4],[5]and generalized to handle linear

Gaussian JMS(LGJMS)multi-target models in[6].

Although the LGJMS-PHD?lter[6]shows great promise,

many real world problems do not follow linear jump Markov

models.Moreover,at present,there is no tractable analytical

method for tracking multiple targets with nonlinear jump

Markov dynamics.In this paper we present a simple exten-

sion of the LGJMS-PHD recursion to handle nonlinear jump

Markov dynamics.This extension is based on an analytic ap-

proximation of the PHD recursion that combines the LGJMS-

PHD?lter[6]and the unscented transform[7].The resulting

multi-target?lter sidesteps the data association problem,

does not require gating,track initiation and termination,nor

clustering for extracting state estimates.

II.B ACKGROUND

A.Jump Markov System

A jump Markov system(JMS)can be described by a set

of parameterized state space models whose underlying pa-

rameters evolve with time according to a?nite state Markov

chain.Letξk∈R n and z k∈R m denote the kinematic state(e.g.target coordinates and velocity)and observation,

respectively,at time k.Suppose that r k∈M is the label of the model in effect at time k,where M denotes the(discrete) set of all model labels(also called modes).Then,the state dynamics and observation are described by the following state transition density and measurement likelihood:

?f

k|k?1(ξk|ξk?1,r k),

g k(z k|ξk,r k).

In addition,the modes follow a discrete Markov chain with

transition probability t k|k?1(r k|r k?1)and the transition of the augmented state vector x k=[ξT k,r k]T∈X=R n×M is governed by

f k|k?1(x k|x k?1)=?f k|k?1(ξk|ξk?1,r k)t k|k?1(r k|r k?1).

Proceedings of the 45th IEEE Conference on Decision & Control

Manchester Grand Hyatt Hotel

San Diego, CA, USA, December 13-15, 2006

ThB05.4

A linear Gaussian JMS (LGJMS)is a JMS with linear Gaussian models,i.e.conditioned on mode r k the state transition density and observation likelihood are given by ?f k |k ?1(ξk |ξk ?1,r k )=N (ξk ;F k ?1(r k )ξk ?1,Q k ?1(r k ))

g k (z k |ξk ,r k )=N (z k ;H k (r k )ξk ,R k (r k )).where N (·;m,Q )denotes a Gaussian density with mean m and covariance Q ,F k ?1(r k )and H k (r k )denote the transition and observation matrices of model r k respectively,Q k ?1(r k )and R k (r k )denote covariance matrices of the process noise and measurement noise,respectively.B.Random Finite Sets in Multi-target Tracking

In a multi-target scenario,suppose that x k,1,...,x k,N (k )∈X are the augmented states at time k ,where N (k )denotes the number of targets.At the next time step,some of these targets may die,new targets may appear and the surviving targets evolve to their new states.At the sensor,M (k )measurements z k,1,...,z k,M (k )∈R m are received at time k ,some of which are due to targets while the rest are clutter.Note that only some of the existing targets are detected by the sensor,and that the corresponding measurements are indistinguishable from clutter.Hence,the orders in which the states,and the measurements are listed bear no signi?cance.

Mahler’s ?nite set statistics (FISST)approach provides an elegant Bayesian formulation of the multi-target ?ltering problem by treating the ?nite sets of targets and observa-tions,at time k ,as the multi-target state and multi-target observation ,respectively [2]

X k ={x k,1,...,x k,N (k )}?X ,Z k

={z k,1,...,z k,M (k )}?R m .

To model uncertainty in multi-target states and observations,we appeal to the notion of a random ?nite set (RFS).An RFS on a state space X is simply a random variable taking values in the ?nite subsets of X [8].The intensity of an RFS on X is a non-negative function v on X such that v (x )is the instantaneous expected number of targets per unit volume at x .An RFS is Poisson if its cardinality distribution is Poisson with mean N =

v (x )dx and given a cardinality the elements of X are i.i.d.according to v/N .We refer the reader to [3],[4]for overviews on FISST and [2],[9]for comprehensive treatments.

C.The Probability Hypothesis Density Filter

The Probability Hypothesis Density (PHD)?lter is a multi-target ?lter avoids any data association computations derived from the RFS framework [2].The PHD ?lter propagates the posterior intensity of the RFS of targets in time,based on the following assumptions:

A.1Targets evolve in time and generate measurements independently of one another.

A.2The clutter RFS is Poisson and is independent of the measurements.

A.3The predicted multi-target RFS is Poisson.

Assumptions A.1and A.2are quite common in many multi-target tracking algorithms.The additional assumption A.3is a reasonable approximation in applications where interactions between targets are negligible [2].

The PHD propagation is a recursion consisting of a prediction step and an update step.Let v k |k ?1and v k denote the predicted intensity and posterior intensity at time k ,respectively.Then the PHD prediction is given by

v k |k ?1(x )=

p S,k |k ?1(x )f k |k ?1(x |x )v k ?1(x )dx + βk |k ?1(x |x )v k ?1(x )dx +γk (x ),(1)where it is understood that an integral with respect to a

discrete variable means a sum,and

f k |k ?1(·|x )=probability density of a target at time k,

given that its previous state is x ,

p S,k |k ?1(x )=probability that a target still exists at time

k given that its previous state is x ,

βk |k ?1(·|x )=intensity of the RFS of targets spawned at

time k by a target with previous state x ,γk (·)

=intensity of the birth RFS at time k.On arrival of a new multi-target measurement,the posterior

intensity v k is computed from the predicted intensity v k |k ?1via the PHD update :

v k (x )=[1?p D,k (x )]v k |k ?1(x )

+

z ∈Z

k

p D,k (x )g k (z |x )v k |k ?1(x )

κk (z )+ p D,k (x )g k (z |x )v k |k ?1(x )dx ,(2)where

Z k

=multi-target measurement at time k,g k (·|x )

=single-target measurement likelihood at

time k,

p D,k (x )=probability of detection given a state x at

time k,

κk (·)=intensity of the clutter RFS at time k.The PHD recursion is generally intractable due to the ‘curse of dimensionality’in numerical integration.A generic sequential Monte Carlo (SMC)implementation was proposed in [3].This so-called particle-PHD ?lter can accommodate targets with JMS dynamics,and has been used to track multiple maneuvering targets in [10],[11].However,the main drawbacks of the particle approach are the large number of particles,and the unreliability of clustering techniques for extracting state estimates [3],[4].The recently proposed Gaussian mixture PHD ?lter [4],[5]does not suffer from these drawbacks but is not general enough to handle targets with JMS dynamics.

III.U NSCENTED I MPLEMENTATION OF THE PHD F ILTER We present ?rst a JMS multi-target model.For clarity in the presentation of our analytic implementation of the PHD ?lter for JMS multi-target model,we review the analytic solution to the PHD recursion for linear Gaussian JMS (LGJMS)multi-target model proposed in [6].We then show

how the unscented transform is used to implement the PHD ?lter for nonlinear JMS multi-target model.

For notational convenience,Θis used to denote the ordered pair of mean and covariance(m,P)of a Gaussian distribution,i.e.N(x;Θ)=N(x;m,P).Given a linear Gaussian model z=Hx+v,where v is Gaussian noise with mean d and covariance matrix R,we use the notation?to denote the ordered triplet of model parameters(H,R,d), and L(x,z;?)=N(z;Hx+d,R)to denote the probability density at z.We also de?ne the operatorsΠandΨby Π(?,Θ)=(Hm+d,R+HP H T)(3)Ψ(z,?,Θ)=(m+K(z?d?Hm),(I?KH)P),(4) K=P H T(HP H T+R)?1.(5) A.JMS multi-target model

In addition to assumptions A.1-A.3,the JMS multi-target model,assumes:

A.4Each target follows a JMS model.

A.5The probabilities of target survival and target detec-tion are not functions of the kinematic state.

A.5The intensities of birth and spawn RFS take the following forms:

γk(ξ,r)=?γ(ξ)πk(r)(6)βk|k?1(ξ,r|ξ ,r )=?βk|k?1(ξ|ξ ,r )πk|k?1(r|r )(7) where?γk is the intensity of kinematic state births at time k, andπk(·|ξ)is the probability distribution of the modes for a

given birth with kinematic stateξat time k,?βk|k?1(·|ξ ,r ) is the intensity of kinematic states spawned at time k from [ξ T,r ]T,andπk|k?1(·|ξ,ξ ,r )is the probability distribu-tion of the mode for a given kinematic stateξ,spawned at time k from[ξ T,r ]T.

The JMS multi-target model is more general than those in standard multi-target tracking algorithms.Moreover,tra-ditional multi-target?ltering techniques are computationally intractable for a model of such generality.Most existing multiple maneuvering target tracking algorithms do not cater for births or spawnings.

B.Linear Gaussian JMS multi-target model

In the LGJMS multi-target model,each target follows a LGJMS model,i.e.the dynamics and measurement models for the kinematic state have the form:

?f

k|k?1(ξ|ξ ,r)=L(ξ ,ξ;?f,k|k?1(r)),

g k(z|ξ,r)=L(ξ,z;?g,k(r)),

where?f,k|k?1(r)=(F f,k?1(r),Q f,k?1(r),0)denotes the parameters of the linear target dynamics model conditioned on mode r,?g,k(r)=(H k(r),R k(r),0)denotes the param-eters of the linear observation model conditioned on mode r. In particular,conditional on mode r,F f,k?1(r)is the state transition matrix,Q f,k?1(r)is the process noise covariance matrix,H k(r)is the measurement matrix and R k(r)is the measurement noise covariance matrix.Additionally,the birth and spawning intensities of the kinematic states are Gaussian mixtures:

?γk(ξ)=

Jγ,k

i=1

w(i)γ,k N(ξ;Θ(i)γ,k),

k|k?1(ξ|ξ ,r )=

Jβ,k|k?1(r )

j=1

w(j)β,k|k?1(r )L(ξ ,ξ;?(j)β,k|k?1(r )),

where Jγ,k,Θ(i)γ,k=(m(i)γ,k,Q(i)γ,k),w(i)γ,k,Jβ,k|k?1(r ),?(j)β,k|k?1(r )=(F(j)β,k?1(r ),Q(j)β,k?1(r ),d(j)β,k?1(r )), w(j)β,k|k?1(r ),are given model parameters.

C.Closed form PHD recursion

Proposition1:For a LGJMS multi-target model,if the posterior intensity v k?1at time k?1has the form

v k?1(ξ ,r )=

J k?1(r )

i=1

w(i)k?1(r )N(ξ ;Θ(i)k?1(r )).(8)

Then the predicted intensity v k|k?1is given by

v k|k?1(ξ,r)=γk(ξ,r)+v f,k|k?1(ξ,r)+vβ,k|k?1(ξ,r),(9) where

vβ,k|k?1(ξ,r)=

r ,i,j

w(i,j)

β,k|k?1

(r,r )N(ξ;Θ(i,j)

β,k|k?1

(r )),

(10)

w(i,j)

β,k|k?1

(r,r )=πk|k?1(r|r )w(j)β,k|k?1(r )w(i)k?1(r ),(11)

Θ(i,j)

β,k|k?1

(r )=Π(?(j)β,k|k?1(r ),Θ(i)k?1(r )),(12) v f,k|k?1(ξ,r)=

r ,i

w(i)f,k|k?1(r,r )N(ξ;Θ(i)f,k|k?1(r,r )),

(13) w(i)f,k|k?1(r,r )=p S,k|k?1(r )t k|k?1(r|r )w(i)k?1(r ),(14)Θ(i)f,k|k?1(r,r )=Π(?f,k|k?1(r),Θ(i)k?1(r )).(15) Proposition2:For a LGJMS multi-target model,if the predicted intensity v k|k?1has the form

v k|k?1(ξ,r)=

J k|k?1(r)

i=1

w(i)k|k?1(r)N(ξ;Θ(i)k|k?1(r)).(16) Then the posterior intensity v k is given by

v k(ξ,r)=(1?p D,k(r))v k|k?1(ξ,r)+

z∈Z k

v g,k(ξ,r;z),(17) where

v g,k(ξ,r;z)=

i

w(i)g,k(r;z)N(ξ;Θ(i)g,k(r;z)),(18) w(i)g,k(r;z)=

p D,k(r)w(i)k|k?1(r)q(i)g,k(r;z)

κk(z)+

r,i

p D,k(r)w(i)k|k?1(r)q(i)g,k(r;z)

,(19) q(i)g,k(r;z)=N(z;Π(?g,k(r),Θ(i)k|k?1(r))),(20)Θ(i)g,k(r;z)=Ψ(z,?g,k(r),Θ(i)k|k?1(r)).(21)

Propositions 1and 2show how the intensities v k |k ?1and v k are analytically propagated in time under the l LGJMS multi-target model assumptions.The recursions for the means and covariances of v f,k |k ?1and v β,k |k ?1are the Kalman prediction and the recursive computations of the means and covariances of v D,k are the Kalman update.The PHD ?lter has a complexity of O (J k ?1|Z k |)where J k ?1is the number of Gaussian components representing v k ?1for a ?xed model r at time k ?1and |Z k |denotes the number measurements at time k .

These propositions also indicate that the number of com-ponents of the predicted and posterior intensity increases with time,which can be a problem in implementation.How-ever,this problem can be effectively handled by applying some simple pruning procedure [4],[5].Given the posterior intensity v k at time k

v k (ξ,r )=

J k (r )

i =1

w (i )k (r )N (ξ;Θ(i )

k (r )),

the peaks of the intensity are points of highest local con-centration of the expected number of targets.In order to extract the state of the targets from the posterior intensity at

time k ,an estimate of the number of targets ?N

k is needed.This number is simply J k (r )i =1w (i )

k (r )rounded to the nearest integer.The estimate of the multi-target state is the set of ?N k ordered pairs of means and modes (m (i )k

(r ),r )with the largest weights w (i )

k (r ),r ∈M ,i =1,...,J k (r ).

Propositions 1and 2can be extended to exponential mixture probability of survival and probability of detection as in [4].

D.The Unscented JMS-PHD ?lter

We consider the following form of single target dynamical and measurement model for a given mode r

ξk =?f,k ?1(ξk ?1,νf,k ?1,r ),(22)z k

=

h k (ξk ,εk ,r ),

(23)

where ?f,k ?1and h k are the non-linear mappings,and νf,k ?1and εk are independent zero-mean Gaussian noise processes with covariance matrices Q f,k ?1(r )and R k (r )respectively.In addition,the spawning intensity has the form

k |k ?1(ξ|ξ ,r )=J β,k |k ?1(r ) j =1

w (j )

β,k |k ?1(r )N (ξ;?β,k ?1(ξ ,r ),Q (j )β,k ?1(r )),(24)where J β,k |k ?1(r

),w (j )

β,k |k ?1(r ),

Q (j )

β,k ?1(r )

are given

model parameters,and ?(j )

β,k ?1(·,r )

is a nonlinear function on the kinematic state space R n

.

In single-target ?ltering,analytic approximations of the nonlinear Bayes ?lter include the extended Kalman (EK)?lter and the unscented Kalman (UK)?lter [7].The EK ?lter approximates the posterior density by a Gaussian,which is propagated in time by applying the Kalman recursions to local linearizations of the (nonlinear)mappings ?k and h k .The UK ?lter also approximates the posterior density

by a Gaussian,but instead of using the linearized model,it computes the Gaussian approximation of the posterior density at the next time step using the unscented transform.Details of the EK and UK ?lters are given in [12]and [7],respectively.

Following the development in Section III-C,it can be shown that the posterior intensity of the multi-target state propagated by the PHD recursions (1)-(2)is a weighted sum of various functions of ξ,many of which are non-Gaussian.In the same vein as UK ?lter,we can approximate each of these non-Gaussian constituent functions by a Gaussian.1)Prediction:To begin,suppose that the posterior inten-sity at time k ?1is approximated by

v k ?1(ξ

,r

)≈

J k ?1(r )

i =1

w (i )k ?1(r )N (ξ ;Θ(i )

k ?1(r )),

and that each mixture component N (·;Θ(i )

k ?1(r ))of the posterior intensity,has an associated set of sigma points 1{ξ( )k ?1(i,r )}L =0,with mean m (i )k ?1(r )and covariance P (i )

k ?1(r ).

The predicted intensity still has the same form as (9).Consider ?rst,v f,k |k ?1,the motion component of the pre-dicted intensity in (9).For each mode r of the state transition

generate a set of sigma points {ν( )

f,k ?1(r )}L =0,with mean 0and covariance Q f,k ?1(r ).Then,following the unscented Kalman prediction [7]we can approximate v f,k |k ?1by substituting into (13)

Θ(i )

f,k |k ?1(r,r )=(m (i )

k |k ?1(r,r ),P (i )

k |k ?1(r,r )),(25)

where

m (i )

k |k ?1(r,r )=11+L L

=0ξ( )

k |k ?1(i,r,r ),(26)ξ( )

k |k ?1(i,r,r )=

?f,k ?1(ξ( )k ?1(i,r ),ν( )

f,k ?1(i,r ),r ),

(27)

P (i )

k |k ?1(r,r )

=11+L L

=0

?ξ( )

k |k ?1(i,r,r )?ξ( )k |k ?1(i,r,r )T ,(28)?ξ( )k |k ?1

(i,r,r )=

ξ( )

k |k ?1(i,r,r )

?m (i )

k |k ?1(r,r ),

(29)

Similarly,to approximate v β,k |k ?1,the spawning com-ponent of (9),we generate a set of sigma points {ν( )β,k ?1(j,r )}L =0,with mean 0and covariance Q (j )β,k ?1(r ),for each (j,r )and substitute into (10)

Θ(i,j )

β,k |k ?1(r )=(m (i,j )

k |k ?1(r ),P (i,j )

k |k ?1(r )),

(30)

where

m (i,j )

k |k ?1(r )=11+L

L

=0ξ( )

k |k ?1(i,j,r ),(31)

ξ( )

k |k ?1(i,j,r )=

?β,k ?1(ξ( )

k ?1(i,r ))

+ν( )

β,k ?1(j,r ),(32)

P (i,j )

k |k ?1(r )

=11+L L

=0

?ξ( )

k |k ?1(i,j,r )?ξ( )k |k ?1(i,j,r )T ,(33)?ξ( )k |k ?1

(i,j,r )=

ξ( )

k |k ?1(i,j,r )

?m (i,j )

k |k ?1(r ),

(34)

1see

[7]and references therein for details on sigma points.

To complete the prediction step,we approximate the spontaneous birth intensityγk in(9)by a Gaussian mixture

Jγ,k

i=1

w(i)γ,k(r)N(ξ;Θ(i)γ,k).

2)Update:For the update step,suppose that the predicted intensity is approximated by

v k|k?1(ξ,r)≈J k|k?1(r)

i=1

w(i)k|k?1(r)N(ξ;Θ(i)k|k?1(r)),

and that each component has an associated set of sigma points{ξ( )k|k?1(i,r)}L =0,with mean m(i)k|k?1(r)and covari-ance P(i)k|k?1(r).

The updated posterior intensity still has the same form as

(17).The?rst term on the right hand side of(17)is just

a scaled version of v k|k?1.To approximate v g,k(ξ,r;z)in

(17),we generate a set of sigma points{ε( )k(r)}L =0,with mean0and covariance R k(r)for each mode r.Then,fol-lowing the unscented Kalman update[7]we can approximate v g,k(ξ,r;z)by substituting into(18)-(19)

q(i)g,k(r;z)=N(z;η(i)k|k?1(r),S(i)k(r)),(35)

Θ(i)g,k(r;z)=(m(i)k(r;z),P(i)k(r)),(36)

where

η(i)k|k?1(r)=

1

1+L

L

=0

z( )k|k?1(i,r),(37)

z( )k|k?1(i,r)=h k(ξ( )k|k?1(i,r), ( )k(r),r),(38)

S(i)k(r)=

1

1+L

L

=0

?z( )k|k?1(i,r)?z( )k|k?1(i,r)T,(39)

?z( )k|k?1(i,r)=z( )k|k?1(i,r)?η(i)k|k?1(r),(40) m(i)k(r;z)=m(i)k|k?1(r)?K(i)k(r)(z?η(i)k|k?1(r)),(41) P(i)k(r)=P(i)k|k?1(r)?K(i)k(r)G(i)k(r)T,(42) K(i)k(r)=G(i)k(r)[S(i)k(r)]?1,(43)

G(i)k(r)=

1

1+L

L

=0

?ξ( )

k|k?1

(i,r)?ξ( )k|k?1(i,r)T,(44)

?ξ( )

k|k?1

(i,r)=ξ( )k|k?1(i,r)?m(i)k|k?1(r).(45)

A similar implementation can be done with the EK ap-proach.In this case we approximate the nonlinear mappings ?f,k?1(·,νf,k?1,r),h k(·,εk,r)and?β,k?1(·,νβ,k?1,r )by their respective derivatives and apply Propositions1and2. However,calculating these derivatives may be tedious and error-prone.The unscented approach,on the other hand,does not suffer from these restrictions and can even be applied to models with discontinuities.For these reasons we omit the EK implementation in this paper.

IV.S IMULATION R ESULTS

Consider a two-dimensional scenario where aircrafts ap-pear in a surveillance region at different locations and times. The speed of the aircrafts is in the range Mach[0.45,0.6]. The kinematic stateξ=[p x,˙p x,p y,˙p y,?]T of each aircraft consists of position(p x,p y)in the horizontal plane, velocity(˙p x,˙p y)and turn rate?.Model r=1is the standard

linear constant velocity(CV)model as given in[1].Model r=2is the non-linear co-ordinated turn(CT)model[1] with an unknown turn rate?given by

F k?1(?,r=2)=

?

?

A2??A20

?A

2A20

001

?

?,

Q k?1(r=2)=

?

?

Σ2020

02Σ20

00?Σ2

?

?,

with

A2=

1sin?T

?

0cos?T

,?A2=

01?cos?T

?

0sin?T

,

Σ2=σ2v

2

T3/3T2/2

T2/2T

,?Σ2=T?σ2v

2

,

whereσv

2

=10m s?2and?σv

2

=0.5?s?2are the standard deviations of the noise for the linear and turn portions of the kinematic state during a level turn,respectively.The transition probabilities between the modes are given by

t k|k?1(r|r )

=

0.80.2

0.20.8

.

The aircrafts are observed by a bearing and range-only sensor in a region[?π,π]rad×[0,6×104]m

z k=

arctan(p x,k/p y,k)

p2x,k+p2y,k

+εk,

whereεk~N(·;0,R k)with R k=diag([σ2θ,σ2ρ]),σθ= 2×(π/180)rad s?1andσρ=20m.The interval be-tween the sensor measurements is T=5s.The probability that an aircraft survives at the next time step is taken as p S,k|k?1(r )=0.99and the probability that an aircraft is detected is p D,k(r)=0.98.Clutter is modelled as a Poisson RFS with intensityκk(z)=λc V U(z),where U(·)denotes a uniform density over the surveillance region,V the volume of the surveillance region andλc=4.167×10?4/πdenotes the average number of clutter returns per unit volume.

For simplicity target spawning is not considered.Consider a scenario where the surveillance region includes the?ve air-port locations at(?20,?20)km,(10,20)km,(30,?10)km, (?30,20)km and(?20,40)km.The spontaneous birth RFS is Poisson with intensity

γk(x)=0.1πk(r)

N(ξ;m(1)γ,Pγ)+N(ξ;m(2)γ,Pγ)+ N(ξ;m(3)γ,Pγ)+N(ξ;m(4)γ,Pγ)+N(ξ;m(5)γ,Pγ)

,

with

m (1)γ=

?2×104,

0,?2×104,

0,0 T

,m (2)γ

= 1×104,0,2×104

,0,0 T ,m (3)γ=

3×104,0,?1×104,0,0 T

,m (4)γ

= ?3×104,0,2×104

,0,0 T ,m (5)γ= ?2×104,0,4×104,0,0 T

,P γ=diag 103,200,103,200,0

,and the distribution of the models at birth is taken as

[πk (r )]=[0.80.2].

Fig.1.Target trajectories.‘?’–locations of target births;‘ ’–locations of target deaths (‘×’–location of sensor).

in Fig.2show that the ?lter successfully tracks the targets in clutter.Occasionally,the ?lter underestimates the number of aircrafts in the surveillance region and momentarily loses

in 3Monte Carlo runs is shown in Fig.3.Note that we apply the same pruning procedures and parameters as in [4],[5]for our implementation.

V.C ONCLUSIONS

This paper presents a PHD ?lter for tracking an unknown and time varying number of targets that follow (nonlinear)jump Markov systems models.The proposed algorithm elim-inates the need to perform data association gating,track initiation and termination.Simulation results in a nonlinear scenario with an unknown and time-varying number of ma-neuvering targets observed in clutter shows that the proposed PHD ?lter has promising performance .

In our approach the multiple models are not “interacting”.It is not clear how the PHD ?lter approach can be extended to interacting multiple models.This is an interesting problem in both theory and practice,which requires further investigation.

R EFERENCES

[1]Y .Bar-Shalom,X.-R.Li,and T.Kirubarajan,Estimation with Appli-cation to Tracking and Navigation .Wiley,2001.

[2]R.Mahler,“Multi-target Bayes ?ltering via ?rst-order multi-target

moments,”IEEE Trans.AES ,vol.39,no.4,pp.1152–1178,2003.[3] B.V o,S.Singh,and A.Doucet,“Sequential Monte Carlo

methods for multi-target ?ltering with random ?nite sets,”in IEEE Trans.AES ,vol.41,no.4,pp.1224–1245,2005,also:https://www.wendangku.net/doc/b4503022.html,.au/staff/bv/publications.html.

[4] B.V o and W.K.Ma,“The Gaussian mixture Probability Hypothesis

Density ?lter,”to appear in IEEE Trans.Signal Processing ,2006,also:https://www.wendangku.net/doc/b4503022.html,.au/staff/bv/publications.html.

[5]——,“A closed-form solution to the Probability Hypothesis Density

?lter,”in Proc.Int’l Conf.on Information Fusion,Philadelphia,2005.[6] A.Pasha,B.V o,H.D.Tuan,and W.K.Ma,“Closed form PHD

?ltering for Linear Jump Markov Models,”in Proc.Int’l Conf.on Information Fusion ,2006.

[7]S.J.Julier and J.K.Uhlmann,“Unscented ?ltering and nonlinear

estimation,”in Proc.IEEE ,vol.92,no.3,pp.401–422,2004.

[8] D.Daley and D.Vere-Jones,An Introduction to the Theory of Point

Processes .Springer-Verlag,1988.

[9]I.Goodman,R.Mahler,and H.Nguyen,Mathematics of Data Fusion .

Kluwer Academic Publishers,1997.

[10] B.V o and W.K.Ma,“Joint detection and tracking of multiple

maneuvering targets using random ?nite sets,”in Proc.ICARCV ,Kunming,China,2004.

[11]K.Punithakumar,T.Kirubarajan,and A.Sinha,“A multiple model

Probability Hypothesis Density ?lter for tracking maneuvering tar-gets,”in O.E.Drummond (ed.)Signal and Data Processing of Small Targets,Proc.SPIE ,vol.5428,pp.113–121,2004.

[12] B.D.Anderson and J.B.Moore,Optimal Filtering .Prentice-Hall,

New Jersey,1979.

瀑布水景工程计算方式

15米宽,6米高的人工瀑布,泵的流量要多大,怎样计算? 上水池32方,下水池是一个大湖南 假设瀑布的厚度为A米。那么可以算一下瀑布停止不动是瀑布的体积:15x6xA=90A,那么我们姑且算厚度A=1cm=0.01m,那么此时的体积是0.9立方。 根据瀑布的高度,水从6m处留下来的时间大约是0.6秒,那么此时的流量大概就是0.9x0.6=0.54立方/秒,即1944立方/小时。此时选泵就选流量2000吨/小时,扬程10m左右的泵,此时水泵的功率大概是110Kw左右。 当A=1mm=0.001m的时候,也根据这种算法,那么水泵的流量是194吨/小时。此时选泵就选流量200吨/小时,扬程10m左右的泵,此时水泵的功率大概是11-15Kw左右。 具体选什么泵可根据实况选择潜水泵,或者离心泵(选离心泵是应注意泵不能放在瀑布上方,因为离心泵没有那么高的吸程,放在上方时吸不上水的)。

水景园林给排水:浅谈景观瀑布设计 俗话说“水为庭院灵魂”,由此可见水在园林景观中的重要作用。水与周围景物结合,便会表现出或悠远宁静,或热情昂扬,或天真质朴,或灵动飞扬的意境.艺术地再造自然之魂.从而产生特殊的艺术感染力,使城市景观更添迷人的魅力。因此.景观瀑布作为水景形态之一,在城市景观设计中运用较多。这里,笔者仅就景观瀑布设计谈几点体会。 1 景观瀑布的分类 1.1 自然式瀑布.即模仿河床陡坎的形式,让水从陡坡处滚落下跌形成恢弘的瀑布景观。此类瀑布多用于自然景观与情趣的环境中 1.2 规则式瀑布.即强调落水的规则与秩序性,有着规整的人工构筑落水E1.可形成一级或多级跌落形式的瀑布景观此类瀑布多用于较为规整的建筑环境中。 1.3 斜坡瀑布,即落水由斜面滑落的瀑布景观。它的表面受斜坡表面质地、结构的影响.体现出较为平静、含蓄的意趣,适用于较为安静的场所。 2 景观瀑布的构成 一个完整的景观瀑布一般由背景、上游水源、落水口、瀑身、承瀑潭及溪流构成。其中,瀑身是观赏的主体。 3 景观瀑布的设计要素 3.1 水量 景观瀑布的形式与其上游水源的水量有着密切的关系,瀑布水量应满足景观瀑布的方案设计要求。供水量在lms/s左右时,瀑身可形成重落、离落、布落等形式;供水量在0.1m3/s左右时,瀑身可形成丝落、线落等形式。 3.2 水泵的选择 3.2.1 流量的选择 首先.根据前面提到的瀑布用水量估算表计算流量,再根据《建筑给水排水设计规范》GB50015—2003第3.1 1.9条计算设计循环流量。即:Qs=1.2Qc 式中:Qc-景观瀑布的设计循环流量,m3/h;

跌水水景流量设计

跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点.跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景??跌水跌水水景 在水景设计中,跌水水景是构成溪流、叠流、瀑布等水景的基本单元,具有动态和声响的效果,因而应用较广。 与静态水景不同,动态水景的水是流动的,其流动性一般用循环水泵来维持,水量过大则能耗大,长期运转费用高;水量过小则达不到预期的设计效果。因此,根据水景的规模确定适当的水流量十分重要。 1跌水水景的水力学特征及计算 跌水水景实际上是水力学中的堰流和跌水在实际生活中的应用,跌水水景设计中常用的堰流形式为溢流堰. 根据δ和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式: 当δ/H<0.67,为薄壁堰流;0.67<δ/H<2.5,为实用堰流;2.5<δ/H<10,为宽顶堰流; δ/H>10,为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。 当跌水水景的土建尺寸确定以后,首先要确定跌水水景流量Q,当水流从堰顶以一定的初速度v0落下时,它会产生一个长度为ld的水舌。若ld大于跌水台阶宽度lt,则水景水流会跃过跌水台阶;若ld太小,则有可能出现水景水舌贴着水景跌水墙而形成壁流。这两种情况的出现主要与跌水水景流量Q的大小有关,设计时应尽量选择一个恰当的跌水水景流量以避免上述现象的发生。 水景中的跌水水景设计(二) 跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点.跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景??跌水跌水水景 1.1跌水水景流量计算 根据水力学计算公式,一般宽顶堰自由出流的流量计算式为: Q=σc·m·b·(2g)0.5·H1.5=σc·M·b·H1.5 式中b——堰口净宽H——包括行进流速水头的堰前水头, H=H0+υ02/2g 式中υ0——行进流速m——自由溢流的流量系数,与堰型、堰高等边界条件有关σc——侧收缩系数 M=m·(2g)0.5当堰口为矩形时,侧收缩系数σc为1,上述计算式即简化为《给水排水设计手册》中的流量计算式: Q=m·b·(2g)0.5·H1.5=M·b·H1.5

跌水水景中设计中的计算

跌水水景中的计算实例 某宾馆根据其地形条件在大堂内设计一溢流式跌水景,为扇形结构,第一级跌水高度P为2.1 m,堰口为弧线形,长度b=14.65 m,堰顶宽δ=0.15 m,跌水台阶宽度l t =0.7m。 2.1 计算跌水流量Q 根据宾馆大堂环境的要求,跌水流量不须太大,因此,初始选定堰前水头H=0.2 kPa,根据堰流的出口形式,流量系数M=1 417.4,因此试算流量: 2.2 校核跌水水舌 l d 根据试算流量Q可求出跌水景溢流口的单宽流量: q=Q/b=4.007×10-3 m3/(s·m) 由此得 D=q2/(g·p3)=1.767 3×10-7 跌水水舌长度: l d =4.30×D0.27×P=0.136m 0.1

根据δ和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式: 当δ/H<0.67,为薄壁堰流;0.67<δ/H<2.5,为实用堰流;2.5<δ/H<10,为宽顶堰流; δ/H>10,为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。 当跌水水景的土建尺寸确定以后,首先要确定跌水水景流量Q,当水流从堰 顶以一定的初速度v 0落下时,它会产生一个长度为l d 的水舌。若l d 大于跌水台 阶宽度l t ,则水景水流会跃过跌水台阶;若l d 太小,则有可能出现水景水舌贴 着水景跌水墙而形成壁流。这两种情况的出现主要与跌水水景流量Q的大小有关,设计时应尽量选择一个恰当的跌水水景流量以避免上述现象的水景中的跌水水景设计(二) 1.1 跌水水景流量计算 根据水力学计算公式,一般宽顶堰自由出流的流量计算式为: Q=σ c ·m·b·(2g)0.5·H1.5=σ c ·M·b·H1.5 式中b——堰口净宽H——包括行进流速水头的堰前水头,H=H0+υ 2/2g 式中υ ——行进流速m——自由溢流的流量系数,与堰型、堰高等边界条件有关σc——侧收缩系数 M=m·(2g)0.5 当堰口为矩形时,侧收缩系数σc为1,上述计算式即简化为《给水排水设计手册》中的流量计算式: Q=m·b·(2g)0.5·H1.5=M·b·H1.5 上式中,M(或m)为流量系数,与堰的进口边缘形式有关;b为堰口净宽,为已知,因此要求出水景流量Q,关键要确定出堰前水景水头H,堰前水景水头一般先凭经验选定、试算。通常H的初试值可选为0.2~0.4 kPa,当水景堰口为直角时宜取上限,堰口为斜角或圆角时取下限。H初值选定后,根据上述计算式算

跌水水景流量设计

水景中的跌水水景设计(一) 跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点.跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景跌水跌水水景 在水景设计中,跌水水景是构成溪流、叠流、瀑布等水景的基本单元,具有动态和声响的效果,因而应用较广。 与静态水景不同,动态水景的水是流动的,其流动性一般用循环水泵来维持,水量过大则能耗大,长期运转费用高;水量过小则达不到预期的设计效果。因此,根据水景的规模确定适当的水流量十分重要。 1 跌水水景的水力学特征及计算 跌水水景实际上是水力学中的堰流和跌水在实际生活中的应用,跌水水景设计中常用的堰流形式为溢流堰. 根据δ和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式:当δ/H<0.67,为薄壁堰流;0.67<δ/H<2.5,为实用堰流;2.5<δ/H<10,为宽顶堰流; δ/H>10,为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。 当跌水水景的土建尺寸确定以后,首先要确定跌水水景流量Q,当水流从堰顶以一定的初速度v0落下时,它会产生一个长度为ld的水舌。若ld大于跌水台阶宽度lt,则水景水流会跃过跌水台阶;若ld太小,则有可能出现水景水舌贴着水景跌水墙而形成壁流。这两种情况的出现主要与跌水水景流量Q的大小有关,设计时应尽量选择一个恰当的跌水水景流量以避免上述现象的发生。 水景中的跌水水景设计(二) 跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点.跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景跌水跌水水景 1.1 跌水水景流量计算 根据水力学计算公式,一般宽顶堰自由出流的流量计算式为: Q=σc·m·b·(2g)0.5·H1.5=σc·M·b·H1.5 式中b——堰口净宽H——包括行进流速水头的堰前水头, H=H0+υ02/2g 式中υ0——行进流速m——自由溢流的流量系数,与堰型、堰高等边界条件有关σc——侧收缩系数 M=m·(2g)0.5当堰口为矩形时,侧收缩系数σc为1,上述计算式即简化为《给水排水设计手册》中的流量计算式: Q=m·b·(2g)0.5·H1.5=M·b·H1.5 上式中,M(或m)为流量系数,与堰的进口边缘形式有关;b为堰口净宽,为已知,因此要求出水景流量Q,关键要确定出堰前水景水头H,堰前水景水头一般先凭经验选定、试算。通常H的初试值可选为0.2~0.4 kPa,当水景堰口为直角时宜取上限,堰口为斜角或圆角时取下限。H初值选定后,根据上述计算式算出跌水水景流量Q,由于Q值为试算结果,还须根据跌水水景水舌的长度对Q的大小作进一步的校核和调整。 1.2 校核水景水舌长度 根据水力学的计算公式,溢流堰的跌落水景水舌长度为:

景观水景工程计算书

水景工程

第五章水景工程 导言:园林中最主要的造景法之一是什么? 水景工程中都包含哪些内容? 第三章水景工程 第一节水的功能及分类 ,涉及的内容有水体的类型,各种水景的布置,驳岸、护坡、喷泉等。 一、水体的功能 1.造景:水有三态液态:喷泉、瀑布、跌水。 气态:喷雾泉、创造仙境舞台。 固态:滑冰场、冰雕 2.改善小气吸收粉尘,改善环境卫生 3.有利于动植物的生长,特别是水生植物。 4.灌溉与消防 5.水上游乐,划船、游泳、垂钓、漂流 6.组织交通,水上游览 7.水能陶冶人的情操,提高人的修养 二、水系的构成 自然降雨→地表径流(泉水)→涧、溪→瀑布→潭→河→江→海 三、水源种类:

⑴市政给水,自来水(水质好) ⑵地下水 ⑶地表水 四、水体的形式与分类 1.按水体的形式分:水的形式与其所在环境有关。 ⑴自然式水体:边缘不规则,变化自然的水体。例如:河、湖、池、溪、涧等。 ⑵规则式水体:边缘规则,具有明显的轴线的水体,一般是几何形。 ⑶混合式水体:是规则式与不规则式两种交替穿插形成的水环境。 2.按水体的功能分: ⑴观赏性水体:叶饺装饰性水池,面积较小。 ⑵开展活动性水体:游泳馆、游船、垂钓。大规模综合性公园都属此类。 3.按水流状态分: ⑴静态水景:园林中成片汇集的水面,湖、塘、池等。 ⑵动态水景:流动的水,具有动感,溪、涧、瀑布、跌水等。

小结:本次课讲了三个方面 1.水的功能。 2.水的构成。 3.水体的分类。 思考题:1.水体的构成。 2.水体的分类。 引言:上节课我们学习了水景工程的基本知识,也就是水体的分类和功能,下面我们来学习水体中的重要水工措施: 驳岸与护坡。 第二节驳岸与护坡 园林水体要求有稳定、美观的水岸来维持陆地和水面有一定的面积比例,防止陆地被淹或水岸塌陷而扩大水面。因此在水体边缘必须建造驳岸与护坡。同时,作为水景组成的驳岸与护坡直接影响园景,必须从实用、经济、美观几个方面一起考虑。

跌水水景流量设计

跌水水景流量设计 集团企业公司编码:(LL3698-KKI1269-TM2483-LUI12689-ITT289-

跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点.跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景??跌水跌水水景 在水景设计中,跌水水景是构成溪流、叠流、瀑布等水景的基本单元,具有动态和声响的效果,因而应用较广。 与静态水景不同,动态水景的水是流动的,其流动性一般用循环水泵来维持,水量过大则能耗大,长期运转费用高;水量过小则达不到预期的设计效果。因此,根据水景的规模确定适当的水流量十分重要。 1跌水水景的水力学特征及计算 跌水水景实际上是水力学中的堰流和跌水在实际生活中的应用,跌水水景设计中常用的堰流形式为溢流堰. 根据δ和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式: 当δ/H<0.67,为薄壁堰流;0.67<δ/H<2.5,为实用堰流; 2.5<δ/H<10,为宽顶堰流; δ/H>10,为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。 当跌水水景的土建尺寸确定以后,首先要确定跌水水景流量Q,当水流从堰顶以一定的初速度v0落下时,它会产生一个长度为ld的水舌。若ld大于跌水台阶宽度lt,则水景水流会跃过跌水台阶;若ld太小,则有可能出现水景水舌贴着水景跌水墙而形成壁流。这两种情况的出现主要与跌水水景流量Q的大小有关,设计时应尽量选择一个恰当的跌水水景流量以避免上述现象的发生。 跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点.跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景??跌水跌水水景 1.1跌水水景流量计算 根据水力学计算公式,一般宽顶堰自由出流的流量计算式为:

跌水设计

一、概述 (一)定义 1、跌水:跌落的水,由于地形突然的高差变化而产生的水流现象。 2、瀑布:地形较大的落差变化,使平面的水流呈现直落或斜落的立面水流。 3、叠水:地形呈阶梯状的落差和地貌的凹凸变化,使水流呈现层叠流落而成水流现象 (二)跌水景观的功能 1、跌落的水携带空气中大量的氧进入河流,给水流中的动植物和微生物提供良好的生长条件。 2、飞溅的水花增加了空气湿度,过滤空气中的尘埃。 (三)跌水景观的形式种类 1、水立面形式:线状、点状、帘状、片状、散落状 2、落水方式:直落、飞落、叠落、滑落 3、跌落形式:直接入水式、溅落入水式、可视、可听,具有独特的景观效果。(四)不同形式的形成原因 1、地形的落差决定瀑布形成的高低和水声。 2、地貌的凹凸决定瀑布流落的形状。 3、水流量的多少决定瀑布落水的形式。 4、出水口的大小决定瀑布规模的宽窄。 二、跌水景观的设计要素 1、蓄容 蓄容水流的流量在1m3/s左右的瀑布可行成帘状,片状,和散落状;当仅有0.1m3/s的水流时,则呈现线状、点状。

蓄容分上下两个部位----底池蓄水和堰顶蓄水 2、出水口 (1)隐蔽式:将出水口隐藏在景观环境之中,让水流呈现自然瀑布的形状。(2)外露式:将出水口突显于景观之外,形成明显的人工瀑布造型。 (3)单点式:水流从单一出口跌落,形成单体瀑布。 (4)多点式:出水口以多点或阵列的方式布局,形成规模较大的瀑布景观。

3、瀑布水面 通过控制背景的凹凸肌理加强水面的细节表现,形成造型丰富、形式多样的瀑布景观。 三、叠水景观形式 1、叠水景观以水立面的变化为主要的表现形式。 2、叠水的形式 (1)水帘 水帘是由较大的落差和较宽水流面形成的叠水,控制水流量与出水口的形状将得到不同的水帘形态。 (2)洒落 流量较小的叠水,在较低水压下呈点状或线状跌落。 (3)涌流 涌流是有多层蓄水池不断被注满涌溢而出形成,水流量较大,叠水面呈面状跌落。 (4)管流 由外露式出水管以多种陈列方式形成叠水,水流呈线状。 (5)壁流 流水顺池壁流下,水面可随池壁呈多角度流落。 (6)阶梯式 由多层阶梯造型构成叠水景观。 (7)塔式 多层蓄水池由上至下,由小到大,呈环状倾流而下。 (8)错落式

跌水水景中设计中的计算

跌水水景中设计中的计 算 标准化管理处编码[BBX968T-XBB8968-NNJ668-MM9N]

跌水水景中的计算实例 某宾馆根据其地形条件在大堂内设计一溢流式跌水景,为扇形结构,第一级跌水高度P为2.1 m,堰口为弧线形,长度b=14.65 m,堰顶宽δ=0.15 m,跌水台阶宽度 l t =0.7m。 计算跌水流量Q 根据宾馆大堂环境的要求,跌水流量不须太大,因此,初始选定堰前水头H= kPa,根据堰流的出口形式,流量系数M=1 ,因此试算流量: 校核跌水水舌 l d 根据试算流量Q可求出跌水景溢流口的单宽流量: q=Q/b=×10-3 m3/(s·m) 由此得 D=q2/(g·p3)= 3×10-7 跌水水舌长度: l d =××P=0.136m

在水景设计中,跌水水景是构成溪流、叠流、瀑布等水景的基本单元,具有动态和声响的效果,因而应用较广。 与静态水景不同,动态水景的水是流动的,其流动性一般用循环水泵来维持,水量过大则能耗大,长期运转费用高;水量过小则达不到预期的设计效果。因此,根据水景的规模确定适当的水流量十分重要。 1 跌水水景的水力学特征及计算 跌水水景实际上是水力学中的堰流和跌水在实际生活中的应用,跌水水景设计中常用的堰流形式为溢流堰. 根据δ和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式: 当δ/H<,为薄壁堰流;<δ/H<,为实用堰流;<δ/H<10,为宽顶堰流; δ/H>10,为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。 当跌水水景的土建尺寸确定以后,首先要确定跌水水景流量Q,当水流从堰顶以 一定的初速度v 0落下时,它会产生一个长度为l d 的水舌。若l d 大于跌水台阶宽度l t ,则 水景水流会跃过跌水台阶;若l d 太小,则有可能出现水景水舌贴着水景跌水墙而形成壁流。这两种情况的出现主要与跌水水景流量Q的大小有关,设计时应尽量选择一个恰当的跌水水景流量以避免上述现象的 水景中的跌水水景设计(二)

跌水水景流量设计

跌水水景流量设计 Revised by Petrel at 2021

水景中的跌水水景设计(一) 跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点.跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景?跌水跌水水景 在水景设计中,跌水水景是构成溪流、叠流、瀑布等水景的基本单元,具有动态和声响的效果,因而应用较广。与静态水景不同,动态水景的水是流动的,其流动性一般用循环水泵来维持,水量过大则能耗大,长期运转费用高;水量过小则达不到预期的设计效果。因此,根据水景的规模确定适当的水流量十分重要。 1跌水水景的水力学特征及计算 跌水水景实际上是水力学中的堰流和跌水在实际生活中的应用,跌水水景设计中常用的堰流形式为溢流堰. 根据δ和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式: 当δ/H<0.67,为薄壁堰流;0.67<δ/H<2.5,为实用堰流;2.5<δ/H<10,为宽顶堰流;δ/H>10,为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。当跌水水景的土建尺寸确定以后,首先要确定跌水水景流量Q,当水流从堰顶以一定的初速度v0落下时,它会产生一个长度为ld的水舌。若ld大于跌水台阶宽度lt,则水景水流会跃过跌水台阶;若ld太小,则有可能出现水景水舌贴着水景跌水墙而形成壁流。这两种情况的出现主要与跌水水景流量Q的大小有关,设计时应尽量选择一个恰当的跌水水景流量以避免上述现象的发生。 水景中的跌水水景设计(二) 跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点.跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景?跌水跌水水景 1.1跌水水景流量计算根据水力学计算公式,一般宽顶堰自由出流的流量计算式为: Q=σc·m·b·(2g)0.5·H1.5=σc·M·b·H1.5式中b——堰口净宽H——包括行进流速水头的堰前水头,

水景设计中跌水水景的设计

水景设计中跌水水景的设计及计算 在水景设计中,跌水是构成溪流、叠流、瀑布等水景的基本单元,具有动态和声响的效果,因而应用较广。 与静态水景不同,动态水景的水是流动的,其流动性一般用循环水泵来维持,水量过大则能耗大,长期运转费用高;水量过小则达不到预期的设计效果。因此,根据水景的规模确定适当的水流量十分重要。 1 跌水水景的水力学特征及计算 跌水水景实际上是水力学中的堰流和跌水在实际生活中的应用,跌水水景设计中常用的堰流形式为溢流堰。 根据δ和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式: 当 δ/H<0.67,为薄壁堰流; ( δ:堰顶宽;H:堰前水头) 0.67<δ/H<2.5,为实用堰流; 2.5<δ/H<10,为宽顶堰流; δ/H>10,为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。  当跌水水景的土建尺寸确定以后,首先要确定跌水流量Q,当水流从堰顶以一定的初速度v0落下时,它会产生一个长度为l d的水舌。 若l d 大于跌水台阶宽度l t ,则水流会跃过跌水台阶;若l d 太小,则有

可能出现水舌贴着跌水墙而形成壁流。这两种情况的出现主要与跌水流量Q的大小有关,设计时应尽量选择一个恰当的流量以避免上述现象的发生。 1.1 跌水流量计算 根据水力学计算公式,一般宽顶堰自由出流的流量计算式为: Q=σc·m·b·(2g)0.5·H1.5=σc·M·b·H1.5 式中 b——堰口净宽 H——包括行进流速水头的堰前水头, 2/2g H=H0+υ ——行进流速 式中 υ m——自由溢流的流量系数,与堰型、堰高等边界条件有关 σc——侧收缩系数 M=m·(2g)0.5 当堰口为矩形时,侧收缩系数σc为1,上述计算式即简化为《给水排水设计手册》中的流量计算式: Q=m·b·(2g)0.5·H1.5=M·b·H1.5 上式中,M(或m)为流量系数,与堰的进口边缘形式有关;b为堰口净宽,为已知,因此要求出流量Q,关键要确定出堰前水头H,堰

跌水水景中设计中的计算定稿版

跌水水景中设计中的计 算 HUA system office room 【HUA16H-TTMS2A-HUAS8Q8-HUAH1688】

跌水水景中的计算实例 某宾馆根据其地形条件在大堂内设计一溢流式跌水景,为扇形结构,第一级跌水高度P为2.1 m,堰口为弧线形,长度b=14.65 m,堰顶宽δ=0.15 m,跌水台阶宽度l t=0.7m。 2.1 计算跌水流量Q 根据宾馆大堂环境的要求,跌水流量不须太大,因此,初始选定堰前水头H=0.2 kPa,根据堰流的出口形式,流量系数M=1 417.4,因此试算流量: 2.2 校核跌水水舌 l d根据试算流量Q可求出跌水景溢流口的单宽流量: q=Q/b=4.007×10-3 m3/(s·m) 由此得 D=q2/(g·p3)=1.767 3×10-7 跌水水舌长度: l d=4.30×D0.27×P=0.136m 0.1

水景中的跌水水景设计(一) 在水景设计中,跌水水景是构成溪流、叠流、瀑布等水景的基本单元,具有动态和声响的效果,因而应用较广。 与静态水景不同,动态水景的水是流动的,其流动性一般用循环水泵来维持,水量过大则能耗大,长期运转费用高;水量过小则达不到预期的设计效果。因此,根据水景的规模确定适当的水流量十分重要。 1 跌水水景的水力学特征及计算 跌水水景实际上是水力学中的堰流和跌水在实际生活中的应用,跌水水景设计中常用的堰流形式为溢流堰. 根据δ和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式: 当δ/H<0.67,为薄壁堰流;0.67<δ/H<2.5,为实用堰流;2.5<δ/H<10,为宽顶堰流; δ/H>10,为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。 当跌水水景的土建尺寸确定以后,首先要确定跌水水景流量Q,当水流从堰顶以一定的初速度v0落下时,它会产生一个长度为l d的水舌。若l d大于跌水台阶宽度l t,则水景水流会跃过跌水台阶;若l d太小,则有可能出现水景水舌贴着水景跌水墙而形成壁流。

跌水水景流量设计

跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点?跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景??跌水跌水水景 在水景设计中,跌水水景是构成溪流、叠流、瀑布等水景的基本单元,具有动态和声响的效果,因而应用较广。 与静态水景不同,动态水景的水是流动的,其流动性一般用循环水泵来维持,水量过大则能耗大,长期运转费用高;水量过小则达不到预期的设计效果。因此,根据水景的规模确定适当的水流量十分重要。 1跌水水景的水力学特征及计算 跌水水景实际上是水力学中的堰流和跌水在实际生活中的应用,跌水水景设计中常用的堰流形式为溢流堰. 根据S和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式: 当S /HvO.67为薄壁堰流;0.67< S /H<2.5为实用堰流;2.5< S /H<10为宽顶堰流; S /H>10为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。 当跌水水景的土建尺寸确定以后,首先要确定跌水水景流量Q,当水流从堰顶以一定的初速度vO落下时,它会产生一个长度为Id的水舌。若Id大于跌水台阶宽度It,则水景水流会跃过跌水台阶;若Id太小,则有可能出现水景水舌贴着水景跌水墙而形成壁流。这两种情况的出现主要与跌水水景流量Q的大小有关, 设计时应尽量选择一个恰当的跌水水景流量以避免上述现象的发生。 水景中的跌水水景设计(二) 跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点?跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景??跌水跌水水景 1.1跌水水景流量计算 根据水力学计算公式,一般宽顶堰自由出流的流量计算式为: Q=c c ? m- b ? (2g)0.5 ? H1.5= c c ? M- b ? H1.5 式中b——堰口净宽H ——包括行进流速水头的堰前水头, H=H0u 02/2g 式中u 0―-亍进流速m――自由溢流的流量系数,与堰型、堰高等边界条件有关 c c艸攵缩系数 M=m (2g)0.5当堰口为矩形时,侧收缩系数cc为1,上述计算式即简化为《给

各类水景计算

A景区一水力计算 ①跌泉流量计算Q=σc·m·b·(2g)0.5·H1.5 m取0.36,σc取1,初选H=0.2kpa,测量得L=9.0m,L t=1.6m,b=0.4m 则Q1=0.36*0.4*4.43*0.09=0.057 m3/s 单宽流量q=6.3*10-3 m3/(s·m),则D=40.11*10-6/0.63=6.37*10-5 L d=4.30*D0.27*p=0.127m,0.1< L d <2/3L t,,经校核,跌水景水舌长度lt在合理范围内,因此,选定的流量可作为选用跌水景循环水泵的依据。 ②同理求Q2 m取0.37,σc取1,初选H=0.2kpa,测量得L=8.4m,L t=1.6m,b=0.4m 则Q2=0.37*0.4*4.43*0.09=0.059 m3/s 单宽流量q=0.059/8.4=7.0 *10-3m3/(s·m),则D=4.9*10-5/0.63=7.78*10-5 水蛇L d=4.30*0.078*0.4=0.134m,可取 故Q=Q1+Q2=0.057+0.059=0.116 m3/s,考虑到二级跌泉接收部分一级的水量,故取Q=0.1 m3/s 即Q=100L/s,H0=21.65-20.84+1=1.81m h=0.065+0.025=0.09m,故扬程为1.9m ③同上求泵二,m取0.36,σc取1,初选H=0.2kpa,测量得L=6.0m,L t=1.6m,b=0.4m Q=0.057 m3/s 单宽流量q=0.057/6=0.0095=9.5*10-3 m3/(s·m),则D=14.33*10-5 L d=4.30*D0.27*p=0.158m可取,扬程H=1.81+0.045=1.86m A 景区二水力计算 同上L=7.8m,初选Q=0.057 m3/s 单宽流量q=7.3*10-3 m3/(s·m),则D=5.3*10-5/0.63=8.48*10-5 L d=4.30*D0.27*p=0.137m,符合要求 扬程H=1.29+0.045=1.33m 选泵IS100-80-125,流量60m3/h,扬程4m,功率1.5kw/h 跌水池 m取0.36,σc取1,初选H=0.2kpa,测量得L=5.5m,L t=1.5m,b=0.6m,p=0.85 Q=0.36*0.6*0.43*0.09=8.61*10-2 单宽流量q=1.56*10-2 m3/(s·m) ,则D=2.45*10-4/0.63=3.9*10-4 L d=4.30*D0.27*p=0.44m 0.1< L d <2/3L t故可取Q=86 L/s,Dn=200mm,h=0.09m,扬程H=1.62+0.09=1.71m,选泵为离心清水泵;XA65/13B 流量86.5 m3/s,扬程13.7m,功率7.5kw/h 卵石涌泉 涌泉高度0.4m,喷头选用YQ-201,额定流量6 m3/h,喷头直径DN25,个数15个,管材:钢衬塑复合管,总流量90 m3

跌水水景中设计中的计算

跌水水景中设计中的计 集团文件发布号:(9816-UATWW?MWUB?WUNN?INNUL?DQQTY?

跌水水景中的计算实例 某宾馆根据其地形条件在大堂内设计一溢流式跌水景,为扇形结构,第一级跌水高度 P为2. 1 m,堰口为弧线形,长度b=14. 65 m,堰顶宽6 =0. 15 m,跌水台阶宽度h二0. 7m。2.1计算跌水流量Q 根据宾馆大堂环境的要求,跌水流量不须太大,因此,初始选定堰前水头 H二0.2kPa,根据堰流的出口形式,流量系数M二1417. 4,因此试算流量: 2. 2校核跌水水舌 h根据试算流量Q可求出跌水景溢流口的单宽流量: q=Q/b=4. 007Xl(r ^/(s . m) 由此得 D=q7(g ? p5)=l. 7673 X10'7 跌水水舌长度: la=4. 3OXD O :7XP=O. 136m 0. Kl d<2/31t 经校核,跌水景水舌长度It在合理范围内,因此,选定的流量可作为选用跌水景循环水泵的依据。 水景中的跌水水景设计(一)

在水景设计中,跌水水景是构成溪流、叠流、瀑布等水景的基本单元,具有动态和声响的效果,因而应用较广。

与静态水景不同,动态水景的水是流动的,其流动性一般用循环水泵来维持,水量过大则能耗大,长期运转费用高;水量过小则达不到预期的设计效果。因此,根据水景的规模确定适当的水流量十分重要。 1跌水水景的水力学特征及计算 跌水水景实际上是水力学中的堰流和跌水在实际生活中的应用,跌水水景设计中常用的堰流形式为溢流堰. 根据S和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式:当6/H<0.67,为薄壁堰流;0. 67〈§/H〈2.5,为实用堰流;2. 5< 6/H<10,为宽顶堰流; 8/H>10,为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。 当跌水水景的土建尺寸确定以后,首先要确定跌水水景流量Q,当水流从堰顶以一定的初速度v。落下时,它会产生一个长度为b的水舌。若b大于跌水台阶宽度X,则水景水流会跃过跌水台阶;若b太小,则有可能出现水景水舌贴着水景跌水墙而形成壁流。这两种情况的出现主要与跌水水景流量Q的大小有关,设计时应尽量选择一个恰当的跌水水景流量以避免上述现象的 水景中的跌水水景设计(二) 1. 1跌水水景流量计算 根据水力学计算公式,一般宽顶堰自由出流的流量计算式为: Q= o e? m ? b ? (2g)0:? H15= o c? M ? b ? H1 °

冯晓东水景计算书

水景计算书编制人:冯晓东2014年5用10日

一、景观水钵专项计算 依据跌水花钵外形可确定水钵为溢流式跌水水景景,上下两级,由甲方购买成品及提供数据可知,第一级跌水高度P 为0.6 m ,堰口为矩形,单个堰宽0.05m ,堰口个数共计38个,第二级跌水高度P 为1.2m ,堰口为弧线形,堰口个数为30个,跌水台阶宽度l t =0.45m 。 根据水力学计算公式,宽顶堰自由出流的流量计算式为: 3c 3c 2H b M H g b m Q ???=????=σσ 式中 b ——堰口净宽 H ——包括行进流速水头的堰前水头, H=H0+υ02/2g 式中 υ0——行进流速

m ——自由溢流的流量系数,与堰型、堰高等边 界条件有关 =H σc ——侧收缩系数 g m M 2?=当堰口为矩形时,侧收缩系数σc 为1,上述计 算式简化为《给水排水设计手册》中的流量计算式: 332H b M H g b m Q ??=???= 校核水景水舌长度 P D l d 27 .030.4= 式中 D=q2/(g·p3) q--堰口单宽流量,q=Q/b ,m3/(s·m) p--跌水高度,1m g--重力加速度,9.81 m/s2 一、计算跌水流量Q 根据方案效果设计要求及现场情况环境的要求,跌水流量不须太大,因此,初始选定堰前水头H=0.15 kPa ,根据堰流的出口形式,流量系数M=1420,因此试算流量: 332H b M H g b m Q ??=???= =17.8m 3 /h 二、校核跌水水舌 l d 根据试算流量Q 可求出跌水景溢流口的单宽流量:

跌水水景中的计算实例

跌水水景中的计算实例 标签:水景跌水跌水水景2007-08-16 23:32 简介:为了更好的理解跌水水景中的水理计算,现以一工程为例. 关健字:水景跌水跌水水景 某宾馆根据其地形条件在大堂内设计一溢流式跌水景,为扇形结构,第一级跌水高度P 为2.1 m,堰口为弧线形,长度b=14.65 m,堰顶宽δ=0.15 m,跌水台阶宽度l t=0.7m。 2.1计算跌水流量Q 根据宾馆大堂环境的要求,跌水流量不须太大,因此,初始选定堰前水头H=0.2 kPa,根据堰流的出口形式,流量系数M=1 417.4,因此试算流量: 2.2校核跌水水舌 l d根据试算流量Q可求出跌水景溢流口的单宽流量: q=Q/b=4.007×10-3 m3/(s·m) 由此得 D=q2/(g·p3)=1.767 3×10-7 跌水水舌长度: l d=4.30×D0.27×P=0.136m 0.1

标签:水景跌水跌水水景2007-08-16 23:30 简介:跌水水景在小区水景中为常见的一种表现形式,水量控制是其中的一个关健点.跌水水景水量过大则能耗大,长期运转费用高;跌水水景水量过小则达不到预期的设计效果。 关健字:水景跌水跌水水景 在水景设计中,跌水水景是构成溪流、叠流、瀑布等水景的基本单元,具有动态和声响的效果,因而应用较广。 与静态水景不同,动态水景的水是流动的,其流动性一般用循环水泵来维持,水量过大则能耗大,长期运转费用高;水量过小则达不到预期的设计效果。因此,根据水景的规模确定适当的水流量十分重要。 1 跌水水景的水力学特征及计算 跌水水景实际上是水力学中的堰流和跌水在实际生活中的应用,跌水水景设计中常用的堰流形式为溢流堰. 根据δ和H的相对尺寸,堰流流态一般分为薄壁堰流、实用堰流、宽顶堰流等三种形式:当δ/H<0.67,为薄壁堰流;0.67<δ/H<2.5,为实用堰流;2.5<δ/H<10,为宽顶堰流; δ/H>10,为明渠水流,不是堰流。 跌水水景设计中,常用堰流形态为宽顶堰流。 当跌水水景的土建尺寸确定以后,首先要确定跌水水景流量Q,当水流从堰顶以一定的初速度v0落下时,它会产生一个长度为l d的水舌。若l d大于跌水台阶宽度l t,则水景水流会跃过跌水台阶;若l d太小,则有可能出现水景水舌贴着水景跌水墙而形成壁流。这两种情况的出现主要与跌水水景流量Q的大小有关,设计时应尽量选择一个恰当的跌水水景流量以避免上述现象的 水景中的跌水水景设计(二)

相关文档