CN103034754B - The data processing subdivided modeling method of the Decoupling Mode of body lightening design - Google Patents

Abstract

A data processing subdivided modeling method for the Decoupling Mode of body lightening design, the steps include: response function to be decoupled into non-coupled item and single order coupling terms; Preliminary judgement needs the item number of tectonic model; Construct each non-coupled item, and determine whether non-linear; Repeat above-mentioned steps, until construct complete by all non-coupled Xiang Douyi; Obtain the preliminary approximate model be made up of pure non-coupled item and by approximate model compared with true model, identify whether to there is single order coupling terms, if there is single order coupling terms, then need to identify the variable combination intercoupled, and by the corresponding coupling terms of approximate model technical construction; Repeat above-mentioned steps, until all single order coupling terms have identified complete all; Obtain global approximation model and be optimized, if do not meet the condition of convergence, entering iterative step.The principle of the invention is simple, can meet the structure of the higher-dimension approximate model of requirement of engineering, ensure solving precision, improves approximate model optimized algorithm efficiency.

Description

The data processing subdivided modeling method of the Decoupling Mode of body lightening design

Technical field

The present invention is mainly concerned with Body Optimal Design field, refers in particular to a kind of data processing subdivided modeling method being applicable to the Decoupling Mode of body lightening design, and it is mainly used in building the approximate model structure for body lightening design.

Background technology

" safety, energy-saving and environmental protection " are three large themes of contemporary automotive development, and safety problem (i.e. crash-worthiness problem) occupy the first.At present, noticeable under the background that energy problem rises steadily in international oil price, according to data, China's motor vehicle fuel consumption has accounted for more than 30% of national total oil consumption.Therefore, extremely urgent problem is exactly farthest reduce the consumption of petroleum resources, as far as possible the arriving of alleviating energy crisis, for the research and extension of substitute energy is raced against time.Simultaneously, automobile is every year to the airborne release approximately harmful gas of several hundred million tons, account for more than 60% of atmosphere pollution, high risks is caused to the healthy of people and living environment, and automobile is also acknowledged as " number one killer " of non-renewable energy resources and atmospheric pollution.

Under the object that " body lightening " designs just is to guarantee the prerequisite of car body strength, rigidity, alleviate the quality of body frame as much as possible, so not only can reduce the consumption of steel and fuel oil, can also discharge, improve the speed of a motor vehicle, improve automobile start and braking ability by decreasing pollution further, and effectively can reduce vibration & noise, increase automobile and highway serviceable life; Simultaneously, the design of " automotive light weight technology " is absolutely not simply by its miniaturization, first the original performance of automobile should be kept unaffected, targetedly should alleviate the weight of automobile self, ensure the security of running car, crash-worthiness, vibration resistance and comfortableness again as much as possible, ensure that the cost of automobile itself should be tried one's best not to be enhanced, the comprehensive competitive power guaranteed and strengthen product.Therefore, automotive light weight technology has become the increasingly serious problem in Car design field, and how reasonably designing car body structure, effective layout buffering energy-absorbing structure of good performance are to meet the heat subject that vehicle collision resistant requirement has become numerous scientist research.

In order to break through this bottleneck, " Approximate Model Method " become a kind of can the method for balance quality and efficiency.It is being widely used in engineering optimization in recent years.The essence of " Approximate Model Method " is exactly utilize to approach approximation technique and carry out to discrete data the prediction that matching or interpolation realize unknown point response, replaces computer simulation model that is complicated, that have a large amount of degree of freedom in engineering design to carry out analytical calculation by simple mathematical model.But along with the lifting of design problem complicacy, the quantity of Optimal Parameters also increases considerably thereupon, how efficiently setting up the approximate model tectonic system that can meet engineering precision and robustness requirement becomes problem demanding prompt solution.

The approximate model constructing technology of current main flow mainly comprises Quadratic response method (2ndPolynomialResponseSurfacemethod, 2ndPR), Moving Least (movingleastsquare, MLS), Kriging, RBF interpolation method, MultivariateAdaptiveRegressionSplines(MARS), LeastInterpolatingPolynomials(LIP), inductivelearning(IL) based on the method such as HDMR (HighDimensionalModelRepresentation) of RBF.By the test to these functions, Kriging, RBF, MLS and RBF-HDMR are considered to the higher modeling technique of precision, and RBF-HDMR is also very effective to higher-dimension problem.But Kriging and RBF belongs to interpolation technique, be difficult to the impact overcoming noise, only when input and output are determined, the practicality of this class model of guarantee.What is more important: this class model is all be based upon on empiric risk criterion, therefore its Generalization Capability is difficult to ensure.

In recent years, Support vector regression (Supportvectorregression, the SVR) technology of structure based minimax risk criterion becomes current study hotspot, and is successfully applied to engineering optimization field.But along with the increase of design parameter, SVR method needs to expand the sample points in design space on a large scale equally, needs the direct problem of carrying out long-time computing for those, counting yield becomes the principal element that SVR is difficult to extensively promote.Therefore, above algorithm is for small-scale computational problem normally effective (within 20 parameters).But along with going deep into of research, the calculating scale of research object and the reinforcement of complicacy, the multiparameter application how approximate modelling technique being brought up to Practical Project is the major issue that engineering optimization field is badly in need of solving.

The end of the sixties in last century, the former Russian scholar Ivaknenko used the evolutionism principle of multilayer neural network principle and the survival of the fittest; with K-G(Kolmogorov-Gabor) polynomial expression be modeling basis; propose a kind of GMDH method (GroupMethodofDataHandling, data processing group technology) of identification Complex Nonlinear System.For Unknown Nonlinear Systems, the high-order K-G polynomial expression that classical GMDH algorithm commonly uses Volterra discrete function describes.Its main thought produces first generation intermediate candidate model by the combined crosswise of various simple initial input (partial model), from first generation intermediate candidate model, select optimum again some combine and produce second generation intermediate candidate model, repeat such generation, select and the process of genetic evolution, model complexity is constantly increased, until select optimum complexity model.For this reason, GMDH algorithm is screened by multilayer, with the simple continuous combined approximation of reference function in local, to obtain model enough complicated on the whole.In theory, input parameter can be increased to 1000 by GMDH method, and the design variable of the model namely optimized can extend to 1000.GMDH is through development for many years, also reach its maturity, it can describe the complex nonlinear relation between each factor preferably, predict the outcome less with actual value error, there is pattern-recognition compared with additive method and adopt the advantage such as self-organization and automatic screening, in complicated economy modeling, obtain good effect.Usually, quadratic polynomial is the most frequently used selection of GMDH, but a lot of scholar adopts different patterns to carry out revising to GMDH and perfect, from low order probability model to order Probability model Approach by inchmeal.In recent years, the application of GMDH progressively combines with fuzzy set theory, artificial neural network etc., the further expansion that made it have.

GMDH method also Shortcomings, as variable number is many, the large and intermediate variable of calculated amount is many, brings difficulty to the choose reasonable of intermediate variable; By quadratic polynomial as partial expression, the error produced in multivariate situation is larger; It is poor that GMDH fits non-linear strong data precision, uses the polynomial expression of more multi-level or more complicated high-order likely to enter into the blind alley of over-fitting if pursue precision; Need for direct problem the nonlinear model carrying out calculating consuming time, iterations is too much, thus causes the inefficiency of tectonic model; Contained by final expression formula, variable is many, can not according to defects such as mathematical model are optimized.

Summary of the invention

The technical problem to be solved in the present invention is just: the technical matters existed for prior art, the invention provides a kind of principle is simple, can meet the higher-dimension approximate model of requirement of engineering structure, guarantee solving precision, improve the data processing subdivided modeling method of the Decoupling Mode of the body lightening design of approximate model optimized algorithm efficiency.

For solving the problems of the technologies described above, the present invention by the following technical solutions:

A data processing subdivided modeling method for the Decoupling Mode of body lightening design, the steps include:

(1) response function (lightweight objective function) is decoupled into non-coupled item and single order coupling terms;

(2) preliminary judgement needs the item number of tectonic model;

(3) construct each non-coupled item, and determine whether non-linear; If non-linear, then on corresponding straight line, continue sampling, until meet accuracy requirement;

(4) above-mentioned steps (3) is repeated, until construct complete by all non-coupled Xiang Douyi;

(5) the preliminary approximate model be made up of pure non-coupled item is obtained and by approximate model compared with true model f (x), identify whether to there is single order coupling terms; If there is single order coupling terms, then need to identify the variable combination intercoupled, and by the corresponding coupling terms of approximate model technical construction;

(6) above-mentioned steps (5) is repeated, until all single order coupling terms have identified complete all;

(7) obtain global approximation model and be optimized, if do not meet the condition of convergence, entering iterative step (3).

As a further improvement on the present invention:

Adopt in above-mentioned steps (1) based on HDMR theory and response function is decoupled into non-coupled item and single order coupling terms.

Adopting GMDH(GroupMethodofDataHandling in above-mentioned steps (2)) theory tentatively judges the item number needing tectonic model.

Approximate model technology is adopted to carry out the structure of non-coupled item in above-mentioned steps (3).

Adopt three standards as the interpretational criteria of approximate model, if X _i(i=1,2...m) is that m of stochastic generation in design domain obeys equally distributed test sample book point:

（1）R ²；

$R^2=1-\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{(f\left(X_i\right)-\hat{f}\left(X_i\right))}^2}{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{(f\left(X_i\right)-\stackrel{\‾}{f}\left(X_i\right))}^2}$

Wherein, the mean value of output function at m test sample book point; R ²reflect the precision of an approximate model on the whole, R ²value more close to 1, then approximate model is more accurate;

（2）RAAE；

$\mathrm{RAAE}=\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}|f\left(X_i\right)-\hat{f}\left(X_i\right)|}{m*\mathrm{STD}}$

Wherein, STD represents standard deviation, with R ²the same, this index reflects the precision of approximate model on the whole, and the value of RAAE is more close to 0, then approximate model is more accurate;

（3）RMAE；

$\mathrm{RMAE}=\frac{\max \left(\right|f\left(X_1\right)-\hat{f}\left(X_1\right)|,...,|f\left(X_m\right)-\hat{f}\left(X_m\right)\left|\right)}{\mathrm{STD}}$

RMAE is a local indexes, and RMAE describes the error of certain local field of design space, and the value of RMAE is the smaller the better.

Compared with prior art, the invention has the advantages that:

1. the present invention can meet the structure of the higher-dimension approximate model of requirement of engineering, under the prerequisite ensureing solving precision, and the extensive efficiency improving approximate model optimized algorithm.Approximate model optimization method with current main flow is different, and its main innovation sets up the optimization method of multiparameter problem.By based on function space decoupling technology and GMDH theory, by a small amount of sample, and then comparatively accurate spatial mappings relation can be set up, and be optimized, obtain sane optimum results.In theory, method of the present invention may be used for any modeling problem.

2. adopt the method for the invention described above, clear and definite function analytic expression can be obtained; The self-organization GMDH network synthesis thought of neural network and statistical modeling, can provide the result that functional expression is expressed, or even other modeling method unapproachable multivariate high order regression equation.

3. adopt the method for the invention described above, modeling process Self-organizing Science can be realized.Namely do not need any original hypothesis statistical models and common neural net model establishing process, often need modeling person rule of thumb to do some to mode input variable and model structure and suppose in advance, then find out through repetition test and be satisfied with model.GMDH network then allows input variable up to a hundred, a large amount of model to be selected is successively produced again with a large amount of variablees, algorithm looks for input item explained variable being had to materially affect according to data-driven, self-organization generates optimum network structure, reduces the impact of modeling person's subjective factor as far as possible.

4. adopt the method for the invention described above, can predict complicacy and high precision; When small sample or data noise larger, common neural network can produce the over-fitting to noise, reduce extensive function, and the optimum complex characteristics of GMDH network ensure that it can from that be similar to, uncertain, or even make a policy in conflicting knowledge environment, also because it avoids the over-fitting of model structure and not enough matching simultaneously, model more close to the truth of system, thus has higher predicting reliability.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of the inventive method.

The schematic flow sheet of Fig. 2 the present invention in embody rule example.

Fig. 3 is the present invention's schematic diagram as 8 of design variable parts in embody rule example.

Fig. 4 is the car load distortion energy-absorbing schematic diagram after the present invention carries out optimal design in embody rule example.

Fig. 5 is the schematic diagram of the present invention's initial designs and optimal design B post accelerating curve in embody rule example.

Embodiment

Below with reference to Figure of description and specific embodiment, the present invention is described in further details.

As depicted in figs. 1 and 2, the data processing subdivided modeling method of the Decoupling Mode of body lightening design of the present invention, its flow process is:

(1) response function is decoupled into non-coupled item and single order coupling terms; Consider that the sensitivity of response function to higher-order Coupled Variable item is more weak, the present invention is by HDMR(highdimensionmodelrepresentation) based on theory, only response function is decoupled into non-coupled item and single order coupling terms.Compare with current method, can build in model and can control number of samples in real time according to susceptibility, significantly improve the efficiency of modeling;

(2) preliminary judgement needs the item number of tectonic model; In order to improve modeling efficiency, adopt GMDH theory tentatively to judge the item number needing tectonic model, compare with current HDMR method, GMDH theory has Self-organization, can increase substantially the structure efficiency of modeling;

(3) construct each non-coupled item, and determine whether non-linear; If non-linear, then on corresponding straight line, continue sampling, until meet accuracy requirement; In the present embodiment, the SVR approximate model technology that sane performance can be adopted higher is to carry out the structure of non-coupled item.With traditional based on polynomial response surface (Responsesurfacemethod, RSM), Kriging method is compared, and SVR model has higher robustness, under the prerequisite ensureing precision, can ensure the robustness of model to greatest extent.

(4) above-mentioned steps (3) is repeated, until construct complete by all non-coupled Xiang Douyi;

(5) the preliminary approximate model be made up of pure non-coupled item is obtained and by approximate model compared with true model f (x), identify whether to there is single order coupling terms; If there is single order coupling terms, then need to identify the variable combination intercoupled, and by the corresponding coupling terms of approximate model technical construction, therefore the essence of model can show and embodies;

(6) above-mentioned steps (5) is repeated, until all single order coupling terms have identified complete all;

(7) obtain global approximation model and be optimized, if do not meet the condition of convergence, entering iterative step (3).

Adopt the method for the invention described above, clear and definite function analytic expression can be obtained; The self-organization GMDH network synthesis thought of neural network and statistical modeling, can provide the result that functional expression is expressed, or even other modeling method unapproachable multivariate high order regression equation.

Adopt the method for the invention described above, modeling process Self-organizing Science can be realized.Namely do not need any original hypothesis statistical models and common neural net model establishing process, often need modeling person rule of thumb to do some to mode input variable and model structure and suppose in advance, then find out through repetition test and be satisfied with model.GMDH network then allows input variable up to a hundred, a large amount of model to be selected is successively produced again with a large amount of variablees, algorithm looks for input item explained variable being had to materially affect according to data-driven, self-organization generates optimum network structure, reduces the impact of modeling person's subjective factor as far as possible.

Adopt the method for the invention described above, can predict complicacy and high precision; When small sample or data noise larger, common neural network can produce the over-fitting to noise, reduce extensive function, and the optimum complex characteristics of GMDH network ensure that it can from that be similar to, uncertain, or even make a policy in conflicting knowledge environment, also because it avoids the over-fitting of model structure and not enough matching simultaneously, model more close to the truth of system, thus has higher predicting reliability.

In the present embodiment, adopt three general in the world criterions as the interpretational criteria of approximate model:

If X _i(i=1,2...m) is that m of stochastic generation in design domain obeys equally distributed test sample book point:

（1）R ²

$R^2=1-\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{(f\left(X_i\right)-\hat{f}\left(X_i\right))}^2}{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{(f\left(X_i\right)-\stackrel{\‾}{f}\left(X_i\right))}^2}$

Wherein, the mean value of output function at m test sample book point.R ²reflect the precision of an approximate model on the whole, R ²value more close to 1, then approximate model is more accurate.

（2）RAAE

$\mathrm{RAAE}=\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}|f\left(X_i\right)-\hat{f}\left(X_i\right)|}{m*\mathrm{STD}}$

Wherein, STD represents standard deviation, with R ²the same, this index reflects the precision of approximate model on the whole, and the value of RAAE is more close to 0, then approximate model is more accurate.

（3）RMAE

$\mathrm{RMAE}=\frac{\max \left(\right|f\left(X_1\right)-\hat{f}\left(X_1\right)|,...,|f\left(X_m\right)-\hat{f}\left(X_m\right)\left|\right)}{\mathrm{STD}}$

RMAE is a local indexes, and RMAE describes the error of certain local field of design space, and therefore, the value of RMAE is the smaller the better.

By with embody rule example, the method in the present invention and traditional Approximate Model Method are compared below.

First, get 3 higher-dimension trial functions (seeing the following form 1), adopt one group of identical training sample point (computational costs is identical) to adopt HDMR-GMDH and GMDH two kinds of methods to carry out modeling respectively for same function and compare their precision, result of calculation is if following table 2(is in order to reflect the performance of the inventive method more objectively, and the data in table are the mean value of calculating 100 times).

Table 1

Table 2

From data analysis in table, for the HDMR-GMDH model of 3 trial functions, R ²be worth close to 1, RAAE value all between 0 and 0.12, RMAE value all below 0.15, HDMR-GMDH model can reflect the characteristic of true model preferably as can be seen here, and fitting precision is higher.And for 3 GMDH models, R ²be worth all very low (less than 0.3, even occur negative value), and RMAE value all higher (being greater than 3), fitting result is undesirable.

By above analytic explanation, for higher-dimension problem, based on same one group of training sample, adopt the approximate model that GMDH Method Modeling obtains, precision is poor, can not as the approximate model of true model, but, adopt HDMR-GMDH method but can obtain degree of approximation higher, can be used for the approximate model of Optimization analyses.

Secondly, higher-dimension trial function is adopted:

$f\left(x\right)=\underset{i=1}{\overset{d-1}{\mathrm{\Σ}}}[{\left(x_i^2\right)}^{(x_{i+1}^2+1)}+{\left(x_{i+1}^2\right)}^{(x_i^2+1)}],0\≤x_i\≤1$

Test the efficiency of HDMR-GMDH method.

In above formula, dimension d gets d=10 respectively, 20,30,50 different values such as grade, suppose the sample points of every one dimension be 7(after tested, substantially can meet accuracy requirement).As can be seen here, adopt the sample points needed for HDMR-GMDH Method Modeling be polynomial expression (and not as classic method exponentially) growth along with the growth of problem dimension d, thus increased substantially the efficiency of Modeling and optimization.

In a concrete Optimal Example for body lightening, two subsystems be made up of two finite element models, are that car load vehicle body just hits finite element model respectively, carry out cae analysis respectively to it.Wherein, car load is just hitting finite element model and is being made up of 196396 unit and 200351 nodes.Just hitting in simulation process in finite element, vehicle fixes rigid wall with the speed impacts of 50km/h, the collision simulation process of whole system completes in 100ms, and design object is making collision safety performance and comfortableness is improved while, makes the quality of vehicle body little as far as possible.

The thickness of slab first defining front part of vehicle 8 energy absorbing components (see Fig. 3) is used for just hitting optimization as 8 design variables, and the design value scope of all design variables is as table 3shown in.Consider car body B post acceleration peak value and 8 parts quality and for response.

Design parameter: T ₁-T ₈

Objective function:

$\begin{array}{cc}\min & m\\ s.t.& {\mathrm{\&alpha;}}_B<40g\\ & \mathrm{Eng}>1.2\&times;{10}^{8}J\\ & T_L\&le;T_1-T_8\&le;T_U\end{array}---\left(1\right)$

In formula, m be 8 parts in Fig. 3 quality and; T ₁-T ₈, T _l, T _ube respectively the sheet thickness of 8 energy absorbing components and upper and lower limit thereof; α _bfor car body B post acceleration, and Eng is the size of energy in absorbing.

Result of calculation is as shown in table 3 below:

Table 3

Design variable and response

Initial value

Lower limit

The upper limit

Optimal value

T 1/mm	1.0	0.7	2.0	1.1
					T 2/mm	1.0	0.8	2.0	0.9
T 3/mm	0.8	0.5	1.5	0.7
					T 4/mm	1.2	1.0	3.0	1.6
T 5/mm	1.5	1.0	3.0	1.0
					T 6/mm	1.0	0.8	2.0	1.2
T 7/mm	0.8	0.5	1.5	1.3
					T 8/mm	1.2	1.0	3.0	1,6
m/g	8524			9414
					αB /g	46.8			37.9
frq/hz	26.0			27.5

According to the energy absorption characteristics optimizing front and back, as shown in Figure 4, after optimizing, the distortion energy-absorbing modes of car load is better, has good security performance, simultaneously also satisfied corresponding constraint.Body structure is safely the basis evaluating vehicle safety, and the important indicator evaluating body structure safety is the size of car load B post acceleration peak value, automobile is just hitting security legislation (CMVDR294 regulation) regulation B post acceleration peak value can not give B post accelerating curve comparison diagram before and after optimization more than 60g, Fig. 5.As shown in Figure 5, initial designs B post acceleration peak value is close to 50g, and security performance is not ideal, and the peak value of the acceleration plots after optimizing then has larger reduction, and the B post acceleration peak value after optimization is about 36g, reduces about 28% than initial designs.Meet the requirement that automobile is just hitting security legislation preferably, also make the structural safety of vehicle obtain larger improvement simultaneously, meet the requirement of safety Design.

Below be only the preferred embodiment of the present invention, protection scope of the present invention be not only confined to above-described embodiment, all technical schemes belonged under thinking of the present invention all belong to protection scope of the present invention.It should be pointed out that for those skilled in the art, some improvements and modifications without departing from the principles of the present invention, should be considered as protection scope of the present invention.

Claims (4)

1. a data processing subdivided modeling method for the Decoupling Mode of body lightening design, is characterized in that step is:

(1) response function is decoupled into non-coupled item and single order coupling terms;

(2) preliminary judgement needs the item number of tectonic model;

(3) construct each non-coupled item, and determine whether non-linear; If non-linear, then on corresponding straight line, continue sampling, until meet accuracy requirement;

(4) above-mentioned steps (3) is repeated, until construct complete by all non-coupled Xiang Douyi;

(5) the preliminary approximate model be made up of pure non-coupled item is obtained and by approximate model compared with true model f (x), identify whether to there is single order coupling terms; If there is single order coupling terms, then need to identify the variable combination intercoupled, and by the corresponding coupling terms of approximate model technical construction;

(6) above-mentioned steps (5) is repeated, until all single order coupling terms have identified complete all;

(7) obtain global approximation model, adopt following whole three standards as the interpretational criteria of approximate model, if do not meet the condition of convergence, enter iterative step (3);

If X _im the equally distributed test sample book point of obedience of stochastic generation in design domain, wherein i=1,2...m;

(a)R ²；

$R^2=1-\frac{\underset{i=1}{\overset{m}{\&Sigma;}}{\left(f\right(X_i)-\hat{f}(X_i\left)\right)}^2}{\underset{i=1}{\overset{m}{\&Sigma;}}{\left(f\right(X_i)-\stackrel{\&OverBar;}{f}(X_i\left)\right)}^2}$

Wherein, the mean value of output function at m test sample book point; R ²reflect the precision of an approximate model on the whole, R ²value more close to 1, then approximate model is more accurate;

(b)RAAE；

$RAAE=\frac{\underset{i=1}{\overset{m}{\&Sigma;}}|f\left(X_i\right)-\hat{f}\left(X_i\right)|}{m*STD}$

Wherein, STD represents standard deviation, with R ²the same, this index reflects the precision of approximate model on the whole, and the value of RAAE is more close to 0, then approximate model is more accurate;

(c)RMAE；

$RMAE=\frac{max\left(\right|f\left(X_1\right)-\hat{f}\left(X_1\right)|,\mathrm{...},|f\left(X_m\right)-\hat{f}\left(X_m\right)\left|\right)}{STD}$

RMAE is a local indexes, and RMAE describes the error of certain local field of design space, and the value of RMAE is the smaller the better.

2. the data processing subdivided modeling method of the Decoupling Mode of body lightening design according to claim 1, is characterized in that: adopt in above-mentioned steps (1) based on HDMR theory and response function is decoupled into non-coupled item and single order coupling terms.

3. the data processing subdivided modeling method of the Decoupling Mode of body lightening design according to claim 1, is characterized in that: adopt GMDH theory tentatively to judge to need the item number of tectonic model in above-mentioned steps (2).

4. the data processing subdivided modeling method of the Decoupling Mode of body lightening design according to claim 1, is characterized in that: adopt approximate model technology to carry out the structure of non-coupled item in above-mentioned steps (3).