CN103761581A - Method for civil aircraft flight deck human-computer interface comprehensive evaluation - Google Patents

Abstract

The invention provides a method for civil aircraft flight deck human-computer interface comprehensive evaluation. The method comprises the steps that an evaluation index of a human-computer interface is selected through a comprehensive method or an analysis method or a composite algorithm method or an index attribute grouping method; a civil aircraft flight deck human-computer interface comprehensive evaluation index system is determined; grading is conducted on each index of civil aircraft flight deck human-computer interface evaluation and an initial grading value of the index is determined; data collection is conducted on a given human-computer interface sample, prepared inspection data are substituted, and a relative error of an output actual value and a predicted value is calculated; the accuracy of the relative error is inspected, data of each index of the civil aircraft flight deck human-computer interface to be evaluated are input and are calculated through an established BP nerve network module, and a final evaluation result is obtained. The method can conduct objective and accurate evaluation on the civil aircraft flight deck human-computer interface.

Description

Civil aircraft driving cabin Human-Machine Interface Comprehensive Evaluation method

Technical field

The present invention relates to civil aircraft driving cabin man-machine interface the Automation Design technology.

Background technology

Man-machine interface, as a multidisciplinary crossing domain that relates to the subjects such as computer science, artificial intelligence, ergonomics, computational linguistics, cognitive science, social psychology, is being brought into play important function served as bridge in the information interchange between human and computer, environment.Whether the information interaction passage of human and computer and environment is natural, smooth and easy, efficient, accurate, for improving the availability of man-machine interface and user, experience, for excavating personnel's job interest and potential, minimizing personnel work fatigue and misoperation, enhance productivity and economic benefit all most important.Civil aircraft driving cabin man-machine interface, as the tie that connects aircraft system and pilot, becomes pilot and flight system and carries out mutual main media, and pilot completes aerial mission by controlling man-machine interface.Traditional aircraft cockpit Human Machine Interface is mainly paid close attention to the Man-Machine Engineering Problems relevant to hardware device space layout, and the cognition of less consideration interface information and interaction problems.Along with developing rapidly of high speed processing chip, computer graphics techniques and multimedia technology, use the computer generalization display device of high-resolution giant-screen and touch-sensitive formula control interface to replace the inexorable trend that traditional mechanical type interface has become aircraft cockpit design of new generation.There is the aobvious control techniques of flexible data digitizing comprehensive and graphics display capability and for optimizing man-machine interface and man-machine interaction, provide condition on the one hand, but then, if the design to the information content at interface, Information Organization and display mode is unreasonable, by increasing pilot's cognitive load, affect flight safety.Therefore people more and more pay attention to the design of driving cabin man-machine interface, in design process, the mankind's the imagination is also more and more bold, but the application of everything is all the degree based on human-machine interface evaluation, whether really rationally, comprehensive evaluation man-machine interface how, be a difficult problem be also one must research major issue.

Human-machine interface evaluation belongs to the category of systems engineering, its essence is large system is evaluated, and is in conjunction with the feature of institute's descriptive system, to construct an energy and from a plurality of visual angles and level, to reflect objectively the integral level of system.Civil aircraft driving cabin human-machine interface evaluation is that the cognition according to relevant design criteria and pilot self judges whether the design of the matching relationship of pilot in man-machine interface and relevant display, controller and accessory (as seat etc.) meets the requirement of ergonomics, carry out human-machine interface evaluation and can help to judge the state of overall and each inscape of existing man-machine interface, also can find whereby the problem existing in design, to feed back in time design department, modify.Because relating between many factors and each factor, civil aircraft driving cabin there is complex relationship, and factor in common evaluation method often just rule of thumb or select from the related data of collecting, this produces the large key element of not high, the man-machine ring three of confidence level by the selection course that causes factor of evaluation and is difficult to the problems such as the some factors all listing, list and human-machine interface evaluation have nothing to do.Therefore, make human-computer interface system reach rational requirement, must have a suitable evaluation means.At present, the evaluation of civil aircraft driving cabin man-machine interface has obtained sufficient application and development abroad, the existing sleeve forming of Boeing and Air Passenger company and supporting human-machine interface evaluation method, but have no open report in view of technique sensitive.The domestic evaluation about man-machine interface adopts the method for qualitative analysis and virtual emulation more, and quantitative analysis and research are less, does not form the evaluation method of system synthesis.

Summary of the invention

For overcome prior art evaluation procedure exist the subjective uncertainty of the randomness of decision process, the personnel that participate in evaluation and electing with and understanding on the deficiencies such as ambiguity, the present invention proposes a kind of civil aircraft driving cabin Human-Machine Interface Comprehensive Evaluation method based on BP neural network (Back-Propagation Neural Network).

The technical solution adopted for the present invention to solve the technical problems comprises the following steps:

1) use overall approach, analytic approach, composite algorithm or Criterion Attribute method of grouping are chosen the evaluation index of man-machine interface, evaluation index comprises total arrangement, demonstration information, display frame, controller and materiality principle, function proximity, frequency of utilization principle, use Ordinal Consistency, visual angle rationality, accuracy, information content, the high-lighting of important information, information Comprehensible, information is acceptable, the easy perceptibility of danger signal, picture layout's rationality, color rationality, the rationality of character design, the rationality of icon design, the form size of scale instrument, the rationality of meter dial design, the science of element limit operation position, function division definition, rationality and handling comfort that important executor is placed,

2) determine civil aircraft driving cabin Human-Machine Interface Comprehensive Evaluation index system, comprising: E-civil aircraft driving cabin human-machine interface evaluation, B ₁-total arrangement, B ₂-demonstration information, B ₃-display frame, B ₄-controller, C ₁-materiality principle, C ₂-function proximity, C ₃-frequency of utilization principle, C ₄-use Ordinal Consistency, C ₅-visual angle rationality, C ₆-accuracy, C ₇-information content, C ₈the high-lighting of-important information, C ₉-information Comprehensible, C ₁₀-information is acceptable, C ₁₁the easy perceptibility of-danger signal, C ₁₂-picture layout rationality, C ₁₃-color rationality, C ₁₄the rationality of-character design, C ₁₅the rationality of-icon design, C ₁₆the form size of-scale instrument, C ₁₇the rationality of-meter dial design, C ₁₈the science of-element limit operation position, C ₁₉-function division definition, C ₂₀the rationality of-important executor placement, C ₂₁-handling comfort;

3) adopt 5 grades of comment collection excellent, good, in, poor, the bad indices to civil aircraft driving cabin human-machine interface evaluation carries out grade classification, determines its initial score value;

4) according to civil aircraft driving cabin human-machine interface evaluation system, given man-machine interface sample is carried out to data acquisition, the marking of each index is input, and evaluation result is output, and the learning sample using N the data sample gathering as BP neural network

k=1,2,3 ..., n;

5) by learning sample input vector X _klength n determine that network input layer number is n, by learning sample output vector

length m determine that network output layer nodes is m; Determine network number of plies L and each the number of hidden nodes, wherein, L>=3, the nodes of l layer is n ^(l), and n ⁽¹⁾=n, n ^(L)=m;

Define each interlayer connection weight matrix, l layer connects the connection weight matrix of l+1 layer l=1,2,3 ..., L-1, the element value of each connection weight matrix of initialization;

6) input permissible error ε and learning rate η, initialization iterative computation number of times t=1, learning sample sequence number k=1; Get k learning sample

x _k=(x _1k, x _2k..., x _nk),

by Xk, carry out forward-propagating calculating;

First calculate the output of each node of input layer $O_{\mathrm{jk}}^{\left(l\right)}=f\left(x_{\mathrm{jk}}\right)j=\mathrm{1,2},...,n;$

And the input of successively calculating each node in each layer $I_{\mathrm{jk}}^{\left(l\right)}=\underset{i=1}{\overset{n^{(l-1)}}{\mathrm{\Σ}}}{\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}{O}_{\mathrm{ik}}^{(l-1)}$ And output $O_{\mathrm{jk}}^{\left(l\right)}=f\left(I_{\mathrm{jk}}^{\left(l\right)}\right);$

Finally calculate the error of each output node of output layer $y_{\mathrm{jk}}=O_{\mathrm{jk}}^{\left(L\right)},E_{\mathrm{jk}}=\frac{1}{2}{(y_{\mathrm{jk}}^{*}-y_{\mathrm{jk}})}^2;$

If the arbitrary sample k value to N learning sample, has E _jk≤ ε, j=1,2 ..., m, learning process finishes; Otherwise, carry out error back propagation, revise each connection weight matrix;

First revise L-1 layer hidden layer to the connection weight matrix of output layer $\begin{array}{c}{\mathrm{\δ}}_{\mathrm{jk}}^{\left(L\right)}=-(y_{\mathrm{jk}}^{*}-y_{\mathrm{jk}})f^{\′}\left(I_{\mathrm{jk}}^{\left(L\right)}\right)\\ \mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}\left(t\right)=\mathrm{\η}{\mathrm{\δ}}_{\mathrm{jk}}^{(L-1)}{O}_{\mathrm{ik}}^{\left(L\right)}\\ {\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}(t+1)={\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}\left(t\right)+\mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}\left(t\right);\end{array}$

$\begin{array}{c}{\mathrm{\δ}}_{\mathrm{jk}}^{\left(l\right)}=f^{\′}\left(I_{\mathrm{jk}}^{\left(l\right)}\right)\underset{q=1}{\overset{n^{(l+1)}}{\mathrm{\Σ}}}{\mathrm{\δ}}_{\mathrm{qk}}^{(l+1)}{\mathrm{\ω}}_{\mathrm{jq}}^{\left(l\right)}\\ \mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}\left(t\right)=-\mathrm{\η}{\mathrm{\δ}}_{\mathrm{jk}}^{\left(l\right)}{O}_{\mathrm{ik}}^{(l-1)}\\ {\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}(t+1)={\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}\left(t\right)+\mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}\left(t\right);\end{array}$

Finally make k and t add one, re-start study; When error amount is less than predefined determined value, circulation finishes;

7) by ready check data substitution, calculate the actual value of output and the relative error of desired value, check its accuracy, when meeting the demands, can be used to evaluate the man-machine interface of civil aircraft driving cabin;

8) by the indices data input of civil aircraft driving cabin man-machine interface to be evaluated, by the BP neural network model establishing, calculate, draw final evaluation result.

The invention has the beneficial effects as follows: this method can be made objective and accurate evaluation to civil aircraft driving cabin man-machine interface, for the Automation Design of man-machine interface lays the foundation, relatively to there is significant progress and application widely with existing method.

Accompanying drawing explanation

Fig. 1 is civil aircraft driving cabin Human-Machine Interface Comprehensive Evaluation index system structural representation;

Fig. 2 is network error curve map.

Embodiment

The present invention includes following steps:

1. choose evaluation index

The way of thinking single or that be combined with overall approach, analytic approach, composite algorithm and four kinds of structure indexs of Criterion Attribute method of grouping is chosen the evaluation index of man-machine interface.Choose total arrangement, demonstration information, display frame, controller and materiality principle, function proximity, frequency of utilization principle, use Ordinal Consistency, visual angle rationality, accuracy, information content, the high-lighting of important information, information Comprehensible, information is acceptable, the easy perceptibility of danger signal, picture layout's rationality, color rationality, the rationality of character design, the rationality of icon design, the form size of scale instrument, the rationality of meter dial design, the science of element limit operation position, function division definition, the rationality that important executor is placed, these 25 indexs of handling comfort are as the evaluation index of civil aircraft driving cabin man-machine interface.

2. determine assessment indicator system

" hierarchy of objectivies formula " and " Factor Decomposition formula " this two type systematics Structural Tectonics mode of employing, and carry out structure optimization from four aspects such as completeness, rationality, validity, reliabilities, determine civil aircraft driving cabin Human-Machine Interface Comprehensive Evaluation index system as shown in Figure 1.This assessment indicator system is by 4 two-level index and 21 three grades of index constitutes, being wherein expressed as follows of various indexs: E-civil aircraft driving cabin human-machine interface evaluation, B ₁-total arrangement, B ₂-demonstration information, B ₃-display frame, B ₄-controller, C ₁-materiality principle, C ₂-function proximity, C ₃-frequency of utilization principle, C ₄-use Ordinal Consistency, C ₅-visual angle rationality, C ₆-accuracy, C ₇-information content, C ₈the high-lighting of-important information, C ₉-information Comprehensible, C ₁₀-information is acceptable, C ₁₁the easy perceptibility of-danger signal, C ₁₂-picture layout rationality, C ₁₃-color rationality, C ₁₄the rationality of-character design, C ₁₅the rationality of-icon design, C ₁₆the form size of-scale instrument, C ₁₇the rationality of-meter dial design, C ₁₈the science of-element limit operation position, C ₁₉-function division definition, C ₂₀the rationality of-important executor placement, C ₂₁-handling comfort.

3. pair indices carries out initial score

Adopt 5 grades of comment collection excellent, good, in, poor, bad (0-59 is bad, and 60-69 is for poor, and during 70-79 is, 80-89 is good, and 90-100 is excellent), the indices of civil aircraft driving cabin human-machine interface evaluation is carried out to grade classification, determine its initial score value.

4. gather and evaluate sample

Civil aircraft driving cabin human-machine interface evaluation system index structure according to providing in Fig. 1, carries out data acquisition to given man-machine interface sample, and the marking of each index is input, and evaluation result is output.And the learning sample using N the data sample gathering as BP neural network

, k=1,2,3 ..., n.

5. set up BP neural network structure model

By learning sample input vector X _klength n determine that network input layer number is n, by learning sample output vector

length m determine that network output layer nodes is m; Determine the network number of plies (L>=3) and each the number of hidden nodes.Wherein, the nodes of l layer is n ^(l), and n ⁽¹⁾=n, n ^(L)=m.

Define each interlayer connection weight matrix, the connection weight matrix that l layer connects l+1 layer is:

The element value of each connection weight matrix of initialization.

6. pair BP neural network is learnt

Input permissible error ε and learning rate η, initialization iterative computation number of times t=1, learning sample sequence number k=1.Get k learning sample

x _k=(x _1k, x _2k..., x _nk),

by X _kcarry out forward-propagating calculating.

First calculating each node of input layer is output as:

$O_{\mathrm{jk}}^{\left(l\right)}=f\left(x_{\mathrm{jk}}\right)j=\mathrm{1,2},...,n---\left(1\right)$

And the input and output of successively calculating each node in each layer are:

$I_{\mathrm{jk}}^{\left(l\right)}=\underset{i=1}{\overset{n^{(l-1)}}{\mathrm{\Σ}}}{\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}{O}_{\mathrm{ik}}^{(l-1)}\text{---}\left(3\right)$

$O_{\mathrm{jk}}^{\left(l\right)}=f\left(I_{\mathrm{jk}}^{\left(l\right)}\right)---\left(4\right)$

L=1 wherein, 2 ..., L-1; J=1,2 ..., n ^(l).

The error of finally calculating each output node of output layer (L layer) is:

$y_{\mathrm{jk}}=O_{\mathrm{jk}}^{\left(L\right)}---\left(5\right)$

$E_{\mathrm{jk}}=\frac{1}{2}{(y_{\mathrm{jk}}^{*}-y_{\mathrm{jk}})}^{2}j=\mathrm{1,2},...,m---\left(6\right)$

If the arbitrary sample k value to N learning sample, has E _jk≤ ε, j=1,2 ..., m, learning process finishes; Otherwise, carry out error back propagation, revise each connection weight matrix.

First revise L-1 layer hidden layer to the connection weight matrix of output layer (L layer):

${\mathrm{\δ}}_{\mathrm{jk}}^{\left(L\right)}=-(y_{\mathrm{jk}}^{*}-y_{\mathrm{jk}})f^{\′}\left(I_{\mathrm{jk}}^{\left(L\right)}\right)---\left(7\right)$

$\mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}\left(t\right)=\mathrm{\η}{\mathrm{\δ}}_{\mathrm{jk}}^{(L-1)}{O}_{\mathrm{ik}}^{\left(L\right)}\text{---}\left(8\right)$

${\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}(t+1)={\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}\left(t\right)+\mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}\left(t\right)---\left(9\right)$

Wherein, j=1,2 ..., m; I=1,2 ..., n ^(L-1).

Then oppositely successively revise the connection weight matrix that connects each hidden layer:

${\mathrm{\δ}}_{\mathrm{jk}}^{\left(l\right)}=f^{\′}\left(I_{\mathrm{jk}}^{\left(l\right)}\right)\underset{q=1}{\overset{n^{(l+1)}}{\mathrm{\Σ}}}{\mathrm{\δ}}_{\mathrm{qk}}^{(l+1)}{\mathrm{\ω}}_{\mathrm{jq}}^{\left(l\right)}---\left(10\right)$

$\mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}\left(t\right)=-\mathrm{\η}{\mathrm{\δ}}_{\mathrm{jk}}^{\left(l\right)}{O}_{\mathrm{ik}}^{(l-1)}---\left(11\right)$

${\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}(t+1)={\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}\left(t\right)+\mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}\left(t\right)---\left(12\right)$

Wherein, l=L-1 ..., 2,1; J=1,2 ..., n ^(l); I=1,2 ..., n ^(l-1).

Finally make k=k+1, t=t+1 re-starts study.When error amount is less than predefined determined value, circulation finishes, and learning process is also just through with.

7. check BP neural network model

By ready check data substitution, calculate the actual value of output and the relative error of desired value, check its accuracy, when meeting the demands, can be used to evaluate the man-machine interface of civil aircraft driving cabin.

8. draw evaluation result

By the indices data input of civil aircraft driving cabin man-machine interface to be evaluated, by the BP neural network model establishing, calculate, draw final evaluation result.

Below in conjunction with drawings and Examples, the present invention is further described, the present invention includes but be not limited only to following embodiment.

1. the civil aircraft driving cabin human-machine interface evaluation index of choosing by overall approach, analytic approach, composite algorithm and Criterion Attribute method of grouping, is included as 25 altogether.

2. adopt " hierarchy of objectivies formula " and " Factor Decomposition formula " two kinds of system Structural Tectonics modes to set up civil aircraft driving cabin human-machine interface evaluation index system, as shown in Figure 1.

3. adopt 5 grades of comment collection excellent, good, in, poor, the bad indices to civil aircraft driving cabin human-machine interface evaluation carries out initial score, as shown in table 1.

The scoring of table 1 human-machine interface evaluation indices

4. gather human-machine interface evaluation sample, choose ten samples (X1-X10) as training sample, three samples (Y1-Y3) are made test samples, and the data of collection are as shown in table 2.Because gathered data provide with marking form, there is unitarity, thus can be directly divided by 100 when carrying out standardization processing, the percentages drawing can be used to train.

Table 2BP neural network learning sample

?	X1	X2	X3	X4	X5	X6	X7	X8	X9	X10	Y1	Y2	Y3
														C1	93	94	98	88	81	96	86	95	88	89	86	82	92
C2	81	88	95	83	73	88	84	85	86	87	85	89	91
														C3	77	86	94	78	91	93	87	94	85	91	97	94	88

C4	75	82	93	88	95	91	92	92	84	86	92	78	89
														C5	95	90	96	89	78	98	89	90	89	92	84	93	97
C6	88	87	89	78	88	92	75	96	87	75	87	77	88
														C7	77	86	90	76	83	95	73	90	90	84	77	86	87
C8	84	75	91	89	76	89	69	93	84	86	90	91	86
														C9	89	84	87	90	96	87	89	89	91	92	93	81	85
C10	81	86	80	73	93	83	84	86	83	77	84	72	91
														C11	86	78	83	91	66	90	93	84	78	76	91	94	92
C12	90	79	78	82	72	88	67	82	89	73	76	85	89
														C13	90	88	76	75	84	86	81	83	85	68	83	76	93
C14	91	93	92	71	76	76	89	85	76	89	82	87	84
														C15	92	89	84	72	82	66	88	76	91	85	73	75	83
C16	87	81	95	94	73	69	85	73	93	83	86	76	81
														C17	78	77	90	76	79	89	76	91	83	74	71	77	78
C18	89	90	85	67	65	75	93	86	87	90	88	84	79
														C19	87	83	88	88	61	74	84	81	88	85	91	87	94
C20	86	77	82	84	83	89	86	77	85	67	84	72	73
														C21	66	75	80	80	78	81	87	64	71	84	85	91	88
OUT	86	85	89	81	79	86	85	86	86	83	85	83	87

5. according to the civil aircraft driving cabin human-machine interface evaluation index system structure of having set up, set up BP neural network structure model, adopt three layers of BP neural network to assess the man-machine interface of civil aircraft driving cabin, according to the content in Fig. 1, input layer is chosen 21 nodes, and output layer node is 1.Selection for hidden layer, generally by testing model evaluating ability, determine, between repeatedly computing comparative experiments data and the model calculated value that obtains through identification, error square chooses, adopt trial and error, hidden layer is chosen to different nodes at every turn, the training effect of supervising network respectively, take that to train iterations and output valve be evaluation criteria, finally determines that hidden layer node number is 10.Because the training algorithm of BP neural network can self-adaptation be determined network structure, do not need in advance the initial weight of evaluation index to be carried out to assignment, only need to utilize the evaluation sample having gathered to train network, make e-learning, understand implicit weight rule in training data.

6. setting maximum train epochs is 10000, and the minimum error values of network is 0.000001, selects the data of this 10 personal-machine interface rating sample of X1-X10 as the input of BP neural network, according to the BP neural network structure model training of setting up.

Wherein, a is the data of each index in front ten personal-machine interface samples:

a=[0.93,0.94,0.98,0.88,0.81,0.96,0.86,0.95,0.88,0.89;0.81,0.88,0.95,0.83,0.73,0.88,0.84,0.85,0.86,0.87;0.77,0.86,0.94,0.78,0.91,0.93,0.87,0.94,0.85,0.91;0.75,0.82,0.93,0.88,0.95,0.91,0.92,0.92,0.84,0.86;0.95,0.90,0.96,0.89,0.78,0.98,0.89,0.90,0.89,0.92;0.88,0.87,0.89,0.78,0.88,0.92,0.75,0.96,0.87,0.75;0.77,0.86,0.90,0.76,0.83,0.95,0.73,0.90,0.90,0.84;0.84,0.75,0.91,0.89,0.76,0.89,0.69,0.93,0.84,0.86;0.89,0.84,0.87,0.90,0.96,0.87,0.89,0.89,0.91,0.92;0.81,0.86,0.80,0.73,0.93,0.83,0.84,0.86,0.83,0.77;0.86,0.78,0.83,0.91,0.66,0.90,0.93,0.84,0.78,0.76;0.90,0.79,0.78,0.82,0.72,0.88,0.67,0.82,0.89,0.73;0.90,0.88,0.76,0.75,0.84,0.86,0.81,0.83,0.85,0.68;0.91,0.93,0.92,0.71,0.76,0.76,0.89,0.85,0.76,0.89;0.92,0.89,0.84,0.72,0.82,0.66,0.88,0.76,0.91,0.85;0.87,0.81,0.95,0.94,0.73,0.69,0.85,0.73,0.93,0.83;0.78,0.77,0.90,0.76,0.79,0.89,0.76,0.91,0.83,0.74;0.89,0.90,0.85,0.67,0.65,0.75,0.93,0.86,0.87,0.90;0.87,0.83,0.88,0.88,0.61,0.74,0.84,0.81,0.88,0.85;0.86,0.77,0.82,0.84,0.83,0.89,0.86,0.77,0.85,0.67;0.66,0.75,0.80,0.80,0.78,0.81,0.87,0.64,0.71,0.84]

B is the evaluation result of output:

b=[0.860873,0.849957,0.888184,0.812415,0.794577,0.855947,0.847210,0.862129,0.855973,0.832882]

A in training sample and b respectively as mentioned above, carry out after error calculating, and network error curve as shown in Figure 2.

Result shows, when training, has carried out after 215 times, and error has reached the least error setting.The input of X1-X10 Zhe10Ge sample index data can be obtained to actual Output rusults, and after expected results, find that relative error is little.

7. after training finishes, using the data of this 3 personal-machine interface rating sample of Y1-Y3 gathering as test samples, bring BP neural network into and test, assay is as shown in table 3.

Table 3 Sample result

Test samples

Desired value

Actual value

Absolute error

Relative error (%)

Y1	85.1420	86.7573	1.6153	1.8972
					Y2	83.3562	85.5183	2.1621	2.5938
Y3	86.5141	88.1643	1.6502	1.9074

As can be seen from the above table, relative error between desired value and actual value is all less than 3%, and the sequence of the actual value of exporting and desired value is also in full accord, therefore can think that this network model has superior generalization ability, can be used for new civil aircraft driving cabin man-machine interface to evaluate.

8. the input of the indices data using the initial score in the 3rd step as civil aircraft driving cabin man-machine interface to be evaluated, final result is 83.92, belongs to good level.

Claims (1)

1. a civil aircraft driving cabin Human-Machine Interface Comprehensive Evaluation method, is characterized in that comprising the steps:

1) use overall approach, analytic approach, composite algorithm or Criterion Attribute method of grouping are chosen the evaluation index of man-machine interface, evaluation index comprises total arrangement, demonstration information, display frame, controller and materiality principle, function proximity, frequency of utilization principle, use Ordinal Consistency, visual angle rationality, accuracy, information content, the high-lighting of important information, information Comprehensible, information is acceptable, the easy perceptibility of danger signal, picture layout's rationality, color rationality, the rationality of character design, the rationality of icon design, the form size of scale instrument, the rationality of meter dial design, the science of element limit operation position, function division definition, rationality and handling comfort that important executor is placed,

2) determine civil aircraft driving cabin Human-Machine Interface Comprehensive Evaluation index system, comprising: E-civil aircraft driving cabin human-machine interface evaluation, B ₁-total arrangement, B ₂-demonstration information, B ₃-display frame, B ₄-controller, C ₁-materiality principle, C ₂-function proximity, C ₃-frequency of utilization principle, C ₄-use Ordinal Consistency, C ₅-visual angle rationality, C ₆-accuracy, C ₇-information content, C ₈the high-lighting of-important information, C ₉-information Comprehensible, C ₁₀-information is acceptable, C ₁₁the easy perceptibility of-danger signal, C ₁₂-picture layout rationality, C ₁₃-color rationality, C ₁₄the rationality of-character design, C ₁₅the rationality of-icon design, C ₁₆the form size of-scale instrument, C ₁₇the rationality of-meter dial design, C ₁₈the science of-element limit operation position, C ₁₉-function division definition, C ₂₀the rationality of-important executor placement, C ₂₁-handling comfort;

3) adopt 5 grades of comment collection excellent, good, in, poor, the bad indices to civil aircraft driving cabin human-machine interface evaluation carries out grade classification, determines its initial score value;

4) according to civil aircraft driving cabin human-machine interface evaluation system, given man-machine interface sample is carried out to data acquisition, the marking of each index is input, and evaluation result is output, and the learning sample using N the data sample gathering as BP neural network

, k=1,2,3 ... n;

5) by learning sample input vector X _klength n determine that network input layer number is n, by learning sample output vector

length m determine that network output layer nodes is m; Determine network number of plies L and each the number of hidden nodes, wherein, L>=3, the nodes of l layer is n ^(l), and n ⁽¹⁾=n, n ^(L)=m;

Define each interlayer connection weight matrix, l layer connects the connection weight matrix of l+1 layer

l=1,2,3 ..., L-1, the element value of each connection weight matrix of initialization;

6) input permissible error ε and learning rate η, initialization iterative computation number of times t=1, learning sample sequence number k=1; Get k learning sample

xk=(x _1k, x _2k..., x _nk),

by X _kcarry out forward-propagating calculating;

First calculate the output of each node of input layer $O_{\mathrm{jk}}^{\left(l\right)}=f\left(x_{\mathrm{jk}}\right)j=\mathrm{1,2},...,n;$

And the input of successively calculating each node in each layer $I_{\mathrm{jk}}^{\left(l\right)}=\underset{i=1}{\overset{n^{(l-1)}}{\mathrm{\Σ}}}{\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}{O}_{\mathrm{ik}}^{(l-1)}$ And output $O_{\mathrm{jk}}^{\left(l\right)}=f\left(I_{\mathrm{jk}}^{\left(l\right)}\right);$

Finally calculate the error of each output node of output layer $y_{\mathrm{jk}}=O_{\mathrm{jk}}^{\left(L\right)},E_{\mathrm{jk}}=\frac{1}{2}{(y_{\mathrm{jk}}^{*}-y_{\mathrm{jk}})}^2;$

If the arbitrary sample k value to N learning sample, has E _jk≤ ε, j=1,2 ..., m, learning process finishes; Otherwise, carry out error back propagation, revise each connection weight matrix;

First revise L-1 layer hidden layer to the connection weight matrix of output layer $\begin{array}{c}{\mathrm{\δ}}_{\mathrm{jk}}^{\left(L\right)}=-(y_{\mathrm{jk}}^{*}-y_{\mathrm{jk}})f^{\′}\left(I_{\mathrm{jk}}^{\left(L\right)}\right)\\ \mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}\left(t\right)=\mathrm{\η}{\mathrm{\δ}}_{\mathrm{jk}}^{(L-1)}{O}_{\mathrm{ik}}^{\left(L\right)}\\ {\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}(t+1)={\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}\left(t\right)+\mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(L-1)}\left(t\right);\end{array}$

Then oppositely successively revise the connection weight matrix that connects each hidden layer $\begin{array}{c}{\mathrm{\δ}}_{\mathrm{jk}}^{\left(l\right)}=f^{\′}\left(I_{\mathrm{jk}}^{\left(l\right)}\right)\underset{q=1}{\overset{n^{(l+1)}}{\mathrm{\Σ}}}{\mathrm{\δ}}_{\mathrm{qk}}^{(l+1)}{\mathrm{\ω}}_{\mathrm{jq}}^{\left(l\right)}\\ \mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}\left(t\right)=-\mathrm{\η}{\mathrm{\δ}}_{\mathrm{jk}}^{\left(l\right)}{O}_{\mathrm{ik}}^{(l-1)}\\ {\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}(t+1)={\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}\left(t\right)+\mathrm{\Δ}{\mathrm{\ω}}_{\mathrm{ij}}^{(l-1)}\left(t\right);\end{array}$

Finally make k and t add one, re-start study; When error amount is less than predefined determined value, circulation finishes;

7) by ready check data substitution, calculate the actual value of output and the relative error of desired value, check its accuracy, when meeting the demands, can be used to evaluate the man-machine interface of civil aircraft driving cabin;

8) by the indices data input of civil aircraft driving cabin man-machine interface to be evaluated, by the BP neural network model establishing, calculate, draw final evaluation result.

Images