CN103324983B - A kind of isomorphism identification method of mechanism kinematics chain based on immune genetic hybrid algorithm - Google Patents

Abstract

The invention discloses a kind of isomorphism identification method of mechanism kinematics chain based on immune genetic hybrid algorithm, comprise the steps: 1) topological diagram of its correspondence is formed according to the structure of mechanism kinematics chain；2) genetic coding of adjacency matrix is set according to topological diagram；3) two topological diagram adjacency matrix to be determined are taken, it is determined that object function；4) initial antibodies population is randomly generated；Calculate the fitness of each individuality, and utilize fitness size that it is ranked up；5) it is carried out optimum maintaining strategy；6) concentration of each individuality, total individual number in individual bulk concentration=close individual sum/colony are calculated；7) antibody population obtained in step 5) is carried out optimum Crossover Strategy；8) carry out mutation operation, adopt inverse operators to realize；9) local searching operator is combined with genetic algorithm；10) form immune genetic hybrid algorithm, carry out the Isomorphism Identification computing of mechanism kinematics chain.Instant invention overcomes the shortcoming that genetic algorithm easily converges to local optimum.

Description

A kind of isomorphism identification method of mechanism kinematics chain based on immune genetic hybrid algorithm

Technical field

The present invention relates to isomorphism identification method of mechanism kinematics chain, refer in particular to a kind of for isomorphism identification method of mechanism kinematics chain in Creative Mechanism Design and intelligent CAD.

Background technology

Owing to, in Creative Mechanism Design and intelligent CAD, the research of mechanism structure is very important, it may be assumed that how to select required type in the kinematic chain being made up of multiple components and kinematic pair.And this problem is exactly MECHANISM KINEMATICS CHAIN ISOMORPHISM IDENTIFICATION at all, along with the proposition of this proposition, the expert of countries in the world this respect and scholar propose a series of solution.As in 2000, Feng Chun etc. are published in the paper of " Machine Design and research " supplementary issue P40～41 page in 2000 to propose genetic algorithm is applied to MECHANISM KINEMATICS CHAIN ISOMORPHISM IDENTIFICATION one straightforward procedure, but the method makes its population diversity existed in calculating process be deteriorated due to oversimplification, difficulty converges to global extremum, slack-off being even difficult to of algorithm computing late convergence is restrained；Chinese patent ZL200910184138.0 " a kind of isomorphism identification method of mechanism kinematics chain based on pseudo-hybridization hybrid genetic algorithm ", although improve genetic algorithm in the slack-off problem of computing late convergence, but also without the problem that the population diversity well solving exist in calculating process is deteriorated.

Summary of the invention

It is an object of the invention to the deficiency for overcoming existing method for designing to exist, better solve that existing Creative Mechanism Design cycle length, cost be high and also inefficient problem.Creative Mechanism Design is promoted to develop to efficient and low cost direction and provide a kind of isomorphism identification method of mechanism kinematics chain based on immune genetic hybrid algorithm.

The technical scheme is that and adopt following steps successively:

1) topological diagram of its correspondence is formed according to the structure of mechanism kinematics chain；

2) genetic coding of adjacency matrix is set according to topological diagram；

3) two topological diagram adjacency matrix to be determined are taken, it is determined that object function；

4) randomly generating initial antibodies population, Population Size is M；Calculate the fitness of each individuality, and utilize fitness size that it is ranked up；If just then preserving the individuality of relevance grade optimum for colony in specific one private variable, forwarding step 6) to, otherwise, continuing；

5) antibody population of step 4) is carried out optimum maintaining strategy, its method is as follows: judge t for the relevance grade of optimum individual in colony P (t) whether more than the relevance grade of optimum individual in private variable, if more than, optimum individual in private variable is replicated portion in colony P (t), deleting worst individuality therein and preserve colony P(t simultaneously) in optimum individual in above-mentioned private variable, substitute original numerical value；If less than, only optimum individual in private variable is replicated portion and arrive in colony P (t), delete worst individuality therein simultaneously；

6) calculating the concentration of each individuality, total individual number in individual bulk concentration=close individual sum/colony, individual bulk concentration is big, then suppress this antibody, allow for close number of individuals and tail off, thus maintain the multiformity of individual in population；Select the individuality that fitness is high and concentration is relatively low and carry out clone's duplication, and the individuality that fitness low concentration is higher is suppressed；Antagonist group performs to select and replicate operation；

7) antibody population obtained in step 5) is carried out optimum Crossover Strategy: selecting t for the 10%*M(M at the most end of the fitness in colony P (t) is population number) individual antibody and the optimum individual P being saved in variable^*T () is intersected respectively, the method for intersection is: the individual P selected_i(t) and P^*T () is put in a copulation pond, i=1～0.1M, according to selected Crossover Strategy (i.e. pseudo-crossover process), two individualities selected is carried out intersection operation, obtains a pair offspring individual P_i(t) and P^*T (), then with individual P_iT () substitutes the individual P in colony_i(t) and by individuality P^*T () is given up to fall；

8) in order to avoid being absorbed in local optimum, adopt inverse operators that antibody each in above-mentioned antibody population is carried out mutation operation；

9) local searching operator is combined with genetic algorithm；

10) form optimal choice criss-cross inheritance hybrid algorithm, carry out the Isomorphism Identification computing of mechanism kinematics chain.

The invention has the beneficial effects as follows:

1, in the present invention, Immunity Operator and genetic algorithm is utilized to combine, it is ensured that algorithm is in the multiformity of running colony.Overcome the shortcoming that genetic algorithm easily converges to local optimum；

2, in the present invention, utilize the feature of optimum maintaining strategy that the optimum individual occurred so far not only will not be made to be chosen, intersect and mutation operation is lost and destroys, it is ensured that algorithmic statement is to the probability of globally optimal solution.Overcome the shortcoming that genetic algorithm is difficult to restrain in the algorithm later stage；

3, in the present invention, the feature utilizing optimum Crossover Strategy is so that evolving in the individual quick direction higher to fitness that relevance grade is relatively low, accelerates the algorithm convergence rate to global optimum.Overcome the shortcoming that genetic algorithm convergence rate is slow.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of the present invention.

Detailed description of the invention

The idiographic flow of inventive algorithm describes as follows:

Step 1: by formula C=TAT^T, it is the initial population P(0 of M that conversion T randomly generates a scale)；

Step 2: calculate target function value and the fitness of each individuality, stores individuality maximum for fitness in locally optimal solution to private variable；

Step 3: if just for population, then forwarding step 5) to；Otherwise, continue；

Step 4: determine the relevance grade of each individuality, and press the sequence of fitness size；Carry out optimum maintaining strategy, its method is as follows: judge P(t) in the relevance grade of optimum individual whether more than the relevance grade of optimum individual in private variable, if more than, optimum individual in private variable is replicated portion in P (t), deleting worst individuality therein and preserve P(t simultaneously) in optimum individual in above-mentioned private variable, substitute original numerical value；If less than, only optimum individual in private variable is replicated portion and arrive in P (t), delete worst individuality therein simultaneously；

Step 5: define according to the similarity of antibody and concentration, calculate the concentration of each antibody；

Step 6: select the individuality that fitness is high and concentration is relatively low and carry out clone's duplication, and the individuality that fitness low concentration is higher is suppressed；Antagonist group performs to select and replicate operation；

Step 7: antagonist group carries out optimum Crossover Strategy: selecting t for the 10%*M(M at the most end of the fitness in colony P (t) is population number) individual antibody and the optimum individual P being saved in variable^*T () is intersected respectively, the method for intersection is: the individual P selected_i(t) and P^*T () is put in a copulation pond, i=1～0.1M, according to selected Crossover Strategy (, i.e. pseudo-crossover process), two individualities selected carries out intersecting operation, obtains a pair offspring individual P_i(t) and P^*T (), then with individual P_iT () substitutes the individual P in colony_i(t) and by individuality P^*T () is given up to fall；

Step 8: in order to avoid being absorbed in local optimum, adopts inverse operators that antibody each in above-mentioned antibody population is carried out mutation operation；

Step 9: antagonist group performs Local Search；

Step 10: find optimal solution, two kinematic chain isomorphisms；Otherwise, tripe systems.

Wherein in step 2, object function and fitness function computational methods are as follows:

Taking the adjacency matrix that A, B are two topological diagram G1, G2 to be determined, object function takes:

$\min f\left(x\right)=\underset{i,j=0}{\overset{d-1}{\mathrm{\Σ}}}|C_{(i,j)}-B_{(i,j)}|,$

C=TAT^T,

In formula, d is number of vertex, is also individual lengths, T=P₁P₂……P_n, wherein P_iFor matrix

A is the adjacency matrix of mechanism topological diagram G1,

B is the adjacency matrix of mechanism topological diagram G2,

The matrix that C adjacency matrix A obtains after carrying out line translation and rank transformation through transformation matrix T.

Affinity judgement schematics is as follows:

F(x)=1/f(x)

Can extrapolate the value of F (x) (0,1] between, it is seen that f (x) is more little, and F (x) is more big

The wherein Local Search in step 9, for individual x=(x₁,x₂,…,x_d) algorithm is as follows:

Step 1: randomly select down two elements being designated as i, j；

Step 2: it is exactly an original individual neighbours x' that exchange the two element obtains new individuality；

Step 3: compare the fitness F of x and x', if F (x') > F (x), then x=x'；Otherwise give up x'.

Claims (1)

1., based on an isomorphism identification method of mechanism kinematics chain for immune genetic hybrid algorithm, it is characterized in that adopting successively following steps:

1) topological diagram of its correspondence is formed according to the structure of mechanism kinematics chain；

2) genetic coding of adjacency matrix is set according to topological diagram；

3) two topological diagram adjacency matrix to be determined are taken, it is determined that object function；

4) randomly generating initial antibodies population, Population Size is M；Calculate the fitness of each individuality, and utilize fitness size that it is ranked up；If just then preserving the individuality of relevance grade optimum for colony in specific one private variable, forward step 6 to), otherwise, continue；

5) to step 4) antibody population carry out optimum maintaining strategy, its method is as follows: if t=1, namely it is initial population, as long as individuality optimum in initial population is saved in private variable, if t > 1, then must judge t for the relevance grade of optimum individual in colony P (t) whether more than the relevance grade of optimum individual in private variable, if more than, optimum individual in private variable is replicated portion and arrives in colony P (t), delete optimum individual in worst individuality therein and preservation colony P (t) simultaneously and substitute original numerical value in above-mentioned private variable；If less than, only optimum individual in private variable is replicated portion and arrive in colony P (t), delete worst individuality therein simultaneously；

6) calculating the concentration of each individuality, total individual number in individual bulk concentration=close individual sum/colony, individual bulk concentration is big, then suppress this antibody, allow for close number of individuals and tail off, thus maintain the multiformity of individual in population；Select the individuality that fitness is high and concentration is relatively low and carry out clone's duplication, and the individuality that fitness low concentration is higher is suppressed；Antagonist group performs to select and replicate operation；

7) to step 5) in the antibody population that obtains carry out optimum Crossover Strategy: select t for minimum 10%*M the antibody of the fitness in colony P (t) and the optimum individual P being saved in variable^*T () is intersected respectively, the method for intersection is: the individual P selected_i(t) and P^*T () is put in a copulation pond, i=1～0.1M, according to selected Crossover Strategy, i.e. and pseudo-crossover process, two individualities selected carry out intersecting operating, obtains a pair offspring individual P_i(t) and P^*T (), then with individual P_iT () substitutes the individual P in colony_i(t) and by individuality P^*T () is given up to fall；

8) in order to avoid being absorbed in local optimum, adopt inverse operators that antibody each in above-mentioned antibody population is carried out mutation operation；

9) local searching operator is combined with genetic algorithm；Local Search is for individual x=(x₁,x₂,…,x_d) step be:

Step 1: randomly select down two elements being designated as i, j；

Step 2: it is exactly an original individual neighbours x ' that exchange the two element obtains new individuality；

Step 3: compare the fitness F of x and x ', if F (x ') > F (x), then x=x '；Otherwise give up x '；

10) form optimal choice criss-cross inheritance hybrid algorithm, carry out the Isomorphism Identification computing of mechanism kinematics chain.