CN106508018B - Flying quality " outlier " detection and elimination method based on airplane motion equation - Google Patents

Abstract

The present invention relates to a kind of flying quality " outlier " detection and elimination method based on airplane motion equation, belongs to flight mechanics and technical field of information processing, whether there is for check flight data and differ greatly " outlier " with other measured values.In order to overcome art methods to depend only on itself measurement data and caused " outlier " erroneous judgement, the present invention proposes a kind of progressively prediction " outlier " threshold determination based on airplane motion equation and elimination method, change speed of the method according to state, state of flight and parameter are progressively predicted according to the output response for effectively simplifying, is often walked according to the threshold determination for setting and is rejected in measurement " outlier "；As in actual measurement, fast state or parameter presence " outlier " probability are bigger than slow variable, and output response has the precision in engineering, and method proposed by the present invention is more effective to rejecting flying quality " outlier ".

Description

Flying quality " outlier " detection and elimination method based on airplane motion equation

Technical field

The present invention relates to a kind of flying quality " outlier " detection and elimination method, more particularly to based on airplane motion equation Flying quality " outlier " is detected and elimination method.

Background technology

In real aircraft airborne measurement and control system, test volume is usually due to vibration, coding, measuring apparatus error itself etc. Reason causes measurement data inaccurate, particularly can usually there is some or multiple jumps in the measurement such as acceleration, angular speed Point, these hops differ greatly with other measured values, " outlier " in commonly referred to as measuring.These " outlier " are although individual Count few, but calculating etc. analyzed to state of flight estimation, navigation, aeroplane performance because amplitude is larger and bring larger error, As a result the even actual state of flight of substantial deviation.Therefore, people are in the processing procedure to flying quality, it is necessary to pick Except these " outlier ", to ensure the accuracy of test result.

Rejecting the method for " outlier " mainly has following several：

One is the statistical property according to observation, set by calculating standard deviation etc. threshold value to " outlier " carry out judge and Reject, conventional method has Vladimir Romanovskiy (Romanovsky) criterion, Dixon (Dixon) criterion, Xiao Weile (Chauvenet) special criterion of criterion, Grubbs test method, Lay etc..

Two be, by certain information processing and feature extracting method, " outlier " to be judged and rejected, conventional method Mainly there are small wave converting method, the method handled based on time series feature extracting method, based on output signal etc..

Three be that the estimate produced according to certain method of estimation is judged with the statistical property of observation difference.

First kind method is easily caused erroneous judgement because without the information included in utilization observation, confidence level is not high； Although Equations of The Second Kind method adaptability is stronger, algorithm is numerous and diverse, causes the added burden of system, is not easy to work online； 3rd class method, which is mostly based on to be filtered under optimal situation, detects outlier, to outlier detectability in the case of non-optimal filtering Deficiency, moreover, not carrying out simplifying the projected relationship for providing flight parameter to airplane motion equation.

The content of the invention

In order to overcome the shortcomings of that existing flying quality " outlier " detection and elimination method have erroneous judgement, the present invention provides one Flying quality " outlier " detection and elimination method kind based on airplane motion equation, this method is according to the change of detection state Speed, responds progressively to predict state of flight and parameter according to effectively simplified output, often walks according to the threshold value set Judge and reject in measurement " outlier ".

The technical solution adopted for the present invention to solve the technical problems：A kind of flying quality based on airplane motion equation is " wild Value " detection and elimination method, are characterized in comprising the following steps：

(a) when carrying out speed " outlier " and rejecting, the speed increment predicted value between two sampled points is made to be：

Δ V=gT { n_x(k)cos[α(k)]+n_z(k)sin[α(k)]}

As Δ V >=0, V₀(k)+0.8ΔV≤V₀(k+1)≤V₀(k) without " outlier " during+1.2 Δ V；

As Δ V ＜ 0, V₀(k)+0.8ΔV≥V₀(k+1)≥V₀(k) without " outlier " during+1.2 Δ V；

If the measured value of speed increment is excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that not having " outlier "；

(b) when carrying out the air-flow angle of attack " outlier " and rejecting, the air-flow angle of attack incremental forecasting value between two sampled points is made to be：

Δ α=T { q (k)+g [1+n_z(k)]/V₀(k)]}

As the Δs of Δ α >=0, α (k)+0.8 α≤α (k+1)≤Δ α of α (k)+1.2 without " outlier "；

As Δ α ＜ 0, the Δs of α (k)+0.8 α >=α (k+1) >=Δ α of α (k)+1.2 without " outlier "；

If the measured value of air-flow angle of attack increment is excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that Do not have " outlier "；

(c) when carrying out yaw angle " outlier " and rejecting, the yaw angle incremental forecasting value between two sampled points is made to be：

As the Δs of Δ β >=0, β (k)+0.8 β≤β (k+1)≤Δ β of β (k)+1.2 without " outlier "；

As Δ β ＜ 0, the Δs of β (k)+0.8 β >=β (k+1) >=Δ β of β (k)+1.2 without " outlier "；

If the measured value of sideslip angle increment is excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that not having Have " outlier "；

(d) when carrying out three Eulerian angles " outlier " and rejecting, the angle of pitch incremental forecasting value between two sampled points is made to be：

As the Δs of Δ θ >=0, θ (k)+0.8 θ≤θ (k+1)≤Δ θ of θ (k)+1.2 without " outlier "；

As Δ θ ＜ 0, the Δs of θ (k)+0.8 θ >=θ (k+1) >=Δ θ of θ (k)+1.2 without " outlier "；

The roll angle incremental forecasting value between two sampled points is made to be：

WhenWhen without " outlier "；

WhenWhen without " outlier "；

The yaw angle incremental forecasting value between two sampled points is made to be：

As the Δs of Δ ψ >=0, ψ (k)+0.8 ψ≤ψ (k+1)≤Δ ψ of ψ (k)+1.2 without " outlier "；

As Δ ψ ＜ 0, the Δs of ψ (k)+0.8 ψ >=ψ (k+1) >=Δ ψ of ψ (k)+1.2 without " outlier "；

If the angle of pitch, roll angle, the measured value for angle increment of going off course are excessive beyond predicted value, need to carry out " outlier " Reject, otherwise it is assumed that there is no " outlier "；

(e) when carrying out three angular speed " outlier " rejectings, order

Three angular speed predicted values of sample point are：

Make respectivelyAnd measured value π (k)=p_m(k), r_m(k), q_m(k)

WhenWhen,When without " outlier "；

WhenWhen,When without " outlier "；

If three angular speed measured values are excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that not having " outlier "；

(f) when " outlier " for carrying out three acceleration is rejected, order

Three acceleration predicted values of sample point are：

Make respectivelyAnd measured value

WhenWhen,When without " outlier "；

WhenWhen,When without " outlier "；

If three acceleration measured values are excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that there is no " wild Value "；

In above formula, T is the sampling period；G is acceleration of gravity.

The beneficial effects of the invention are as follows：This method is rung according to the change speed of detection state according to effectively simplified output It should come progressively to predict state of flight and parameter, often walk according to the threshold determination set and reject in measurement " outlier "；By Fast state or parameter are bigger than slow variable in the presence of " outlier " possibility in being measured in reality, and output response has Precision in engineering, method proposed by the present invention is more effective to rejecting flying quality " outlier ".

The present invention is elaborated with reference to the accompanying drawings and examples.

Brief description of the drawings

Fig. 1 is flying quality " outlier " detection and elimination method flow chart of the present invention based on airplane motion equation.

Fig. 2 is flying quality " outlier " detection and elimination method medium velocity " outlier " of the present invention based on airplane motion equation Reject flow chart.

Fig. 3 is flying quality " outlier " detection of the present invention based on airplane motion equation and the angle of attack " outlier " in elimination method Reject flow chart.

Fig. 4 is that flying quality " outlier " detection of the present invention based on airplane motion equation and yaw angle in elimination method are " wild Value " rejects flow chart.

Fig. 5 is that flying quality " outlier " detection of the present invention based on airplane motion equation and yaw angle in elimination method are " wild Value " rejects flow chart.

Embodiment

Reference picture 1~5, in flying quality " outlier " detection and elimination method of the present invention based on airplane motion equation, T is Sampling period；G is acceleration of gravity, is comprised the following steps that：

The first step, speed " outlier " is rejected.

The speed increment predicted value between two sampled points is made to be：

Δ V=gT { n_x(k)cos[α(k)]+n_z(k)sin[α(k)]}

As Δ V >=0, V₀(k)+0.8ΔV≤V₀(k+1)≤V₀(k) without " outlier " during+1.2 Δ V；

As Δ V ＜ 0, V₀(k)+0.8ΔV≥V₀(k+1)≥V₀(k) without " outlier " during+1.2 Δ V；

If the measured value of speed increment is excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that not having " outlier "；

Second step, the air-flow angle of attack " outlier " is rejected.

The air-flow angle of attack incremental forecasting value between two sampled points is made to be：

Δ α=T { q (k)+g [1+n_z(k)]/V₀(k)]}

As the Δs of Δ α >=0, α (k)+0.8 α≤α (k+1)≤Δ α of α (k)+1.2 without " outlier "；

As Δ α ＜ 0, the Δs of α (k)+0.8 α >=α (k+1) >=Δ α of α (k)+1.2 without " outlier "；

If the measured value of air-flow angle of attack increment is excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that Do not have " outlier "；

3rd step, yaw angle " outlier " is rejected.

The yaw angle incremental forecasting value between two sampled points is made to be：

As the Δs of Δ β >=0, β (k)+0.8 β≤β (k+1)≤Δ β of β (k)+1.2 without " outlier "；

As Δ β ＜ 0, the Δs of β (k)+0.8 β >=β (k+1) >=Δ β of β (k)+1.2 without " outlier "；

If the measured value of sideslip angle increment is excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that not having Have " outlier "；

4th step, three Eulerian angles " outlier " are rejected.

The angle of pitch incremental forecasting value between two sampled points is made to be：

As the Δs of Δ θ >=0, θ (k)+0.8 θ≤θ (k+1)≤Δ θ of θ (k)+1.2 without " outlier "；

As Δ θ ＜ 0, the Δs of θ (k)+0.8 θ >=θ (k+1) >=Δ θ of θ (k)+1.2 without " outlier "；

The roll angle incremental forecasting value between two sampled points is made to be：

WhenWhen without " outlier "；

WhenWhen without " outlier "；

The yaw angle incremental forecasting value between two sampled points is made to be：

As the Δs of Δ ψ >=0, ψ (k)+0.8 ψ≤ψ (k+1)≤Δ ψ of ψ (k)+1.2 without " outlier "；

As Δ ψ ＜ 0, the Δs of ψ (k)+0.8 ψ >=ψ (k+1) >=Δ ψ of ψ (k)+1.2 without " outlier "；

If the angle of pitch, roll angle, the measured value for angle increment of going off course are excessive beyond predicted value, need to carry out " outlier " Reject, otherwise it is assumed that there is no " outlier "；

5th step, three angular speed " outlier " are rejected.Order

Three angular speed predicted values of sample point are：

Make respectivelyAnd measured value π (k)=p_m(k), r_m(k), q_m(k)

WhenWhen,When without " outlier "；

WhenWhen,When without " outlier "；

If three angular speed measured values are excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that not having " outlier "；

6th step, " outlier " of three acceleration is rejected.Order

Three acceleration predicted values of sample point are：

Make respectivelyAnd measured value

WhenWhen,When without " outlier "；

WhenWhen,When without " outlier "；

If three acceleration measured values are excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that not having " outlier ".

Claims (1)

1. it is a kind of based on airplane motion equation flying quality " outlier " detection and elimination method, it is characterised in that including with Lower step：

(a) when carrying out speed " outlier " and rejecting, the speed increment predicted value between two sampled points is made to be：

Δ V=gT { n_x(k)cos[α(k)]+n_z(k)sin[α(k)]}

As Δ V >=0, V₀(k)+0.8ΔV≤V₀(k+1)≤V₀(k) without " outlier " during+1.2 Δ V；

As Δ V ＜ 0, V₀(k)+0.8ΔV≥V₀(k+1)≥V₀(k) without " outlier " during+1.2 Δ V；

If the measured value of speed increment is excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that not having " outlier "；

(b) when carrying out the air-flow angle of attack " outlier " and rejecting, the air-flow angle of attack incremental forecasting value between two sampled points is made to be：

Δ α=T { q (k)+g [1+n_z(k)]/V₀(k)]}

As the Δs of Δ α >=0, α (k)+0.8 α≤α (k+1)≤Δ α of α (k)+1.2 without " outlier "；

As Δ α ＜ 0, the Δs of α (k)+0.8 α >=α (k+1) >=Δ α of α (k)+1.2 without " outlier "；

If the measured value of air-flow angle of attack increment is excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that Do not have " outlier "；

(c) when carrying out yaw angle " outlier " and rejecting, the yaw angle incremental forecasting value between two sampled points is made to be：

$\mathrm{\Δ}\mathrm{\β}=T\{\frac{1}{V_0\left(k\right)}{\mathrm{gn}}_y\left(k\right)-r\left(k\right)cos\[\mathrm{\α}\left(k\right)\]+p\left(k\right)sin\[\mathrm{\α}\left(k\right)\]\}$

As the Δs of Δ β >=0, β (k)+0.8 β≤β (k+1)≤Δ β of β (k)+1.2 without " outlier "；

As Δ β ＜ 0, the Δs of β (k)+0.8 β >=β (k+1) >=Δ β of β (k)+1.2 without " outlier "；

If the measured value of sideslip angle increment is excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that not having Have " outlier "；

(d) when carrying out three Eulerian angles " outlier " and rejecting, the angle of pitch incremental forecasting value between two sampled points is made to be：

As the Δs of Δ θ >=0, θ (k)+0.8 θ≤θ (k+1)≤Δ θ of θ (k)+1.2 without " outlier "；

As Δ θ ＜ 0, the Δs of θ (k)+0.8 θ >=θ (k+1) >=Δ θ of θ (k)+1.2 without " outlier "；

The roll angle incremental forecasting value between two sampled points is made to be：

WhenWhen without " outlier "；

WhenWhen without " outlier "；

The yaw angle incremental forecasting value between two sampled points is made to be：

As the Δs of Δ ψ >=0, ψ (k)+0.8 ψ≤ψ (k+1)≤Δ ψ of ψ (k)+1.2 without " outlier "；

As Δ ψ ＜ 0, the Δs of ψ (k)+0.8 ψ >=ψ (k+1) >=Δ ψ of ψ (k)+1.2 without " outlier "；

If the angle of pitch, roll angle, the measured value for angle increment of going off course are excessive beyond predicted value, need to carry out " outlier " Reject, otherwise it is assumed that there is no " outlier "；

(e) when carrying out three angular speed " outlier " rejectings, order

Three angular speed predicted values of sample point are：

Make respectivelyAnd measured value π (k)=p_m(k), r_m(k), q_m(k)

WhenWhen,When without " outlier "；

WhenWhen,When without " outlier "；

If three angular speed measured values are excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that not having " outlier "；

(f) when " outlier " for carrying out three acceleration is rejected, order

$\left\{\begin{array}{cc}\stackrel{\·}{u}\left(k\right)\≈\frac{7u(k+1)-8u\left(k\right)+u(k-1)}{12T}& u\left(i\right)=V_0\left(i\right)cos\[\mathrm{\α}\left(i\right)\]cos\[\mathrm{\β}\left(i\right)\],(i=k-1,k,k+1)\\ \stackrel{\·}{v}\left(k\right)\≈\frac{7v(k+1)-8v\left(k\right)+v(k-1)}{12T}& v\left(i\right)=V_0\left(i\right)\sin \[\mathrm{\β}\left(i\right)\],(i=k-1,k,k+1)\\ \stackrel{\·}{w}\left(k\right)\≈\frac{7w(k+1)-8w\left(k\right)+w(k-1)}{12T}& w\left(i\right)=V_0\left(i\right)\sin \[\mathrm{\α}\left(i\right)\]cos\[\mathrm{\β}\left(i\right)\],(i=k-1,k,k+1)\end{array}\right.$

Three acceleration predicted values of sample point are：

Make respectivelyAnd measured value

WhenWhen,When without " outlier "；

WhenWhen,When without " outlier "；

If three acceleration measured values are excessive beyond predicted value, need to carry out " outlier " rejecting, otherwise it is assumed that there is no " outlier "；

In above formula, T is the sampling period；G is acceleration of gravity.