CN102566431A - Method for dispatching and controlling fuzzy parameter operation workshop under dynamic disturbance based on improved genetic algorithm - Google Patents

Abstract

The invention discloses a method for solving the dispatching and controlling of an operation workshop under a complex production environment based on an improved genetic algorithm. The method comprises the following steps that: (1) a dispatching model of the operation workshop with a fuzzy parameter can be set, and a maximized satisfaction degree and a minimized satisfaction degree of all the products are used as a target function; and (2) the target function can be solved by adopting the improved genetic algorithm, (2.1) a code based on a working procedure can be adopted, (2.2) an initial population can be generated by operating a G&T algorithm repeatedly, (2.3) a unity can be evaluated by adopting the target function as a fitness function, (2.4) the operations of selecting and intersecting can be carried out, (2.5) a reversal variation technology can be adopted,(2.6) a population of each group can be combined and can be continuously evolved until being astringed when the unity of each group is astringed to a certain degree, and (2.7) a preset maximum evolution algebra Nmax is used as a stop condition, a solution which is the best one at present is used as an optimal solution, and thus a dispatching scheme of the operation workshop under the complex production environment can be obtained. The method disclosed by the invention has the advantages of good accuracy, better reliability and strong practicability.

Description

A kind of based on the obfuscation parameter job shop dispatch control method under the dynamic disturbances of improving genetic algorithm

Technical field

The present invention relates to job shop scheduling controlling technical field, job shop dispatch control method under especially a kind of solution complicated production environment.

Background technology

In recent years to JSP (job-shop scheduling problem (Job-Shop Scheduling Problem; JSP)) research; Broken through original operational research category; All there is the scholar to carry out fruitful research work in fields such as management science, kybernetics, artificial intelligence, Industrial Engineering, systems engineerings, thereby promotes the development and the fusion of various optimized Algorithm with domain knowledge separately.Along with development of advanced manufacturing, the implication of job-shop scheduling problem is expanded to some extent, has increased attributes such as randomness, dynamic, uncertainty, binding character, multiple goal property, and itself and practical condition are more approaching.

There is following problem mostly in research to job-shop scheduling problem at present:

(1) regards the parameters in the production run as deterministic exact value.The research work majority in past concentrates on confirms the type job-shop scheduling problem, and promptly the process time of workpiece and delivery date all are known quantitative really.And in fact, because various enchancement factors,, seldom can obtain accurate process time and delivery date like influences such as mechanical disorder, operator's skill level, environmental parameters, people can only estimate the variation range at process time or delivery date.Therefore, manufacturing parameter representes that with non-perfect number more accords with production is actual in the scheduling, promptly fuzzy job-shop scheduling problem.

(2) production system is used as the system of a static state, is ignored the appearance of various emergency situations in the actual production process.Interrupt failure conditions owing to the former scheduling that workpiece to be processed continuously arrives, machine tool damages suddenly or quilt is repaired, the appearance of various situations such as the workman is sick, rush order produces, should take in the scheduling, promptly dynamic job shop scheduling.

Existing job shop scheduling research is often simplified production models greatly, and is far away with the actual production situation distance, not exclusively is fit to working condition and changes frequent, working conditions change manufacturing enterprise greatly.

Summary of the invention

In order to overcome the deficiency that accuracy is relatively poor, reliability is relatively poor, practicality is relatively poor of existing job shop dispatching technique, the present invention provide that a kind of accuracy is good, reliability is better, practical based on the obfuscation parameter job shop dispatch control method under the dynamic disturbances of improving genetic algorithm.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of based on the obfuscation parameter job shop dispatch control method under the dynamic disturbances of improving genetic algorithm, said control method may further comprise the steps:

1), confirms the objective function of obfuscation parameter job shop scheduling

The objective function of fuzzy job-shop scheduling problem is: ask one to satisfy z ^*Workpiece processing order, make:

z ^*＝max(z′) (4)

$\left\{\begin{array}{c}{z}^{\&prime;}=y_1{\mathrm{<figure-callout\; id="1"\; label="z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">z</figure-callout>}}_1+y_2{\mathrm{<figure-callout\; id="2"\; label="z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">z</figure-callout>}}_2\\ \mathrm{<figure-callout\; id="1"\; label="max\; z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">max</figure-callout>}{z}_{}{}_1=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i{\mathrm{AI}}_i\\ \mathrm{<figure-callout\; id="2"\; label="max\; z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">max</figure-callout>}{z}_{}{}_2={\mathrm{AI}}_{\min }=\underset{i=1,\mathrm{\&Lambda;n}}{\min }{\mathrm{AI}}_i\\ \underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i=1\\ {\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">y</figure-callout>}}_1+{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">y</figure-callout>}}_2=1\end{array}\right.---(4-1)$

2), adopt the said objective function of improved genetic algorithm for solving, detailed process is following:

(2.1) coding: the workpiece in the scheduling window is adopted the coding based on operation;

(2.2) generation of initial population: repeatedly move the G&T algorithm and produce an initial population;

(2.3) adopt objective function individuality to be estimated as fitness function;

(2.4) selection and interlace operation:

Adopt interlace operation from two parents, to produce a filial generation, specific as follows:

Step 2.4.1. selects the 1st road operation of all workpiece, adds set C; Suppose that machine can process any a plurality of operation at synchronization, respectively operate O among the set of computations C _IjkThe fuzzy deadline of ∈ C, note is done

According to the comparison criterion of Triangular Fuzzy Number, confirm the operation O of least confusion deadline among the set C _I*j*k*, will with the operation O _I*j*k*Use same machine M _K*Operation O _Ijk*∈ C constitutes set G.Because operation O _Ijk*With operation O _I*j*k*Running time on have overlappingly, set G is conflict set;

Step 2.4.2. chooses body one by one wantonly with 50% equal probabilities from two parent individualities; In the set G that conflicts, select to have the operation of least confusion deadline, use the individual replacement of the parent of selecting and be expressed as

Step 2.4.3. is with the selected benchmark that is operating as among the step 2.4.2., through the EC of other operations in the set of the big operation of getting of fuzzy number update inconsistency successively ¹, EC ²And EC ³Value is removed operation from set C

Its next process is added set C, and calculate its corresponding fuzzy deadline;

Obtained a new offspring individual through above operation, carried out c above operation and obtain c new offspring individual.

For keeping the individuality that has merit in (c+2) individuals of forming by c filial generation and 2 parents, the individuality that when hereditary next time, remains with two of following method selections:

A. in c filial generation individuality, selecting to have the individuality of maximum target functional value, also is that local rank is selected;

B. in remaining (c+1) individuals, select to have the individuality of maximum target functional value;

(2.5) mutation operation: adopt the counter-rotating variation;

(2.6) population structure: when the individuality of each group all converges to some spending, each group population is merged continuation evolve up to convergence;

(2.7) with predefined maximum evolutionary generation N _MaxAs stop condition, with best separating so far as optimum solution;

3) when dynamic event takes place, dispatch, concrete steps are following again:

Step 1. initialization scheduling times n=0, the time t that begins to dispatch ₀=0, the maximum number Lw of workpiece puts into window with workpiece to be processed in the given window, and establishing the workpiece window is N _w, having accomplished the workpiece collection is N _c

Workpiece in the step 2. pair current window is carried out and is improved genetic algorithm, selects to make the optimum chromosome of objective function as scheduling result, produces scheduling scheme;

When step 3. takes place as following dynamic event, carry out once scheduling again, establish dynamic event t takes place to do constantly _c

1. start once scheduling again as inserting when single workpiece arrives, change step 4;

2. when the workpiece expiration time changes, start once scheduling again, change step 4;

3. when the machine breakdown incident takes place, start once scheduling again, change step 4;

Step 4. is revised item capable of using, with completed operation from workpiece window N _wMove into and accomplish workpiece collection N _c, produce the scheduling scheme of having accomplished operation; The order of not accomplishing operation in advance, with the residue operation as the new process of not dispatching:

1. will insert single workpiece and add workpiece window N _wAnd do not accomplish operation and form new operation to be processed, carry out scheduling again.If when slotting single workpiece arrived, certain machine was being processed certain procedure, then the automatically delaying certain hour is carried out scheduling again after this procedure is accomplished;

When 2. the workpiece expiration time changes, upgrade the workpiece information in the workpiece window, carry out scheduling again;

3. during machine breakdown, the workpiece of just on this machine, processing is shifted out the workpiece window, remaining workpiece is carried out scheduling again; When simulation clock runs to the residue workpiece and needs the operation in this machining, this workpiece is shifted out the workpiece window, the residue workpiece is carried out scheduling again; When machine is repaired, the workpiece that shifts out the workpiece window is moved into again, carry out scheduling again;

Repeat step 4 according to scheduling events again, accomplish, produce final scheduling scheme up to all workpiece processing.

Technical conceive of the present invention is: on the basis of at present domestic and international job-shop scheduling problem As-Is analysis; Consider uncertainty machined parameters and dynamic disturbances factor in the actual production scheduling process; The ecotopia that job-shop scheduling problem is limited from strictness is extended to the complicated production environment of reality, makes it have dirigibility and practicality.

This paper is defined as the job shop scheduling research under the complicated production environment with the job shop scheduling of machined parameters uncertainty and the generation of dynamic disturbance incident.More meet the actual condition of production just because of the scheduling of the workshop under the complicated production environment, therefore study the job-shop scheduling problem under the complicated production environment, have more important significance for theories and practical value.

Job-shop scheduling problem is a typical NP-hard difficult problem, and mostly there is following problem in the research of existing job-shop scheduling problem: the one, regard various machined parameters in the production system as deterministic exact value; The 2nd, production system is used as the system of a static state, ignored the various emergency situations in the actual process.The actual demand of this paper production function and management under the complicated production environment; Consider the influence of factors such as machined parameters uncertainty exact value in the production run, dynamic disturbances, carried out the research work of the job-shop scheduling problem under the complicated production environment based on improving genetic algorithm.

Genetic algorithm be a kind of use for reference highly-parallel that organic sphere natural selection and evolutionary mechanism grow up, at random, self-adaptive search algorithm, all obtained challenging achievement in fields such as problem solving, optimization and search, Based Intelligent Control, pattern-recognition and artificial lifes.Especially suitable for the insoluble complicacy of those traditional searching algorithms particularly with nonlinear problem.At present, evolutionary computation has become new research direction---computational intelligence with artificial neural network, fuzzy system theory.Likewise, in the Job Shop Scheduling field, also become the focus of research based on the research of genetic algorithm even computational intelligence.

The Fuzzy Set Theory basis: set theory is at first founded 19 end of the centurys by Germany mathematics man Qiao Kangtuo (George Contor).So-called set generally is meant all of object with certain particular community.The object (or things) of forming set is called element, and these elements are corresponding to any one object on the territory, and the relation between the set can only be the relation that belongs to or do not belong to.Progress and science and technology development along with human society; People have been fully recognized that the set theory of utilizing Contor is to describe complicated and diversified real world fully accurately; This just requires the set theory of Contor is expanded; Be about to the value whether an object belong to certain characteristic function of a set and not only be restricted to 0 and 1 liang of number, and should be any one value that to get between 0 and 1.Only be about to 0, and among the 1} fundamental function of value be extended for can be in [0,1] subordinate function of value.

The G&T algorithm: the concrete steps of G&T algorithm are following:

Step 1. is selected the 1st road operation (operation) of all workpiece, adds set C;

Step 2. supposition machine can be processed any a plurality of operation at synchronization, respectively operates O among the set of computations C _IjkThe fuzzy deadline of ∈ C, note is done

Step 3. is confirmed the operation O of least confusion deadline among the set C according to the comparison criterion of Triangular Fuzzy Number _I*j*k*, will with the operation O _I*j*k*Use same machine M _K*Operation O _Ijk*∈ C constitutes set G.Because operation O _Ijk*With operation O _I*j*k*Running time on have overlappingly, therefore gather G and be called the conflict set;

Step 4. is selected an operation at random from set G

The EC of step 5. selected benchmark that is operating as in the above step, the add operation through big operation of getting of fuzzy number (to the adjustment of processing start time of operation) and fuzzy number (to the adjustment of time that machines of operation) other operations during update inconsistency is gathered successively ¹, EC ²And EC ³Value (because every machine synchronization can only be processed a workpiece, so must adjust) to the determined Fuzzy Time of step 2.From set C, remove operation its next process is added set C, and calculate its corresponding fuzzy deadline.

Step 6. repeating step 3-5 is up to all operation completion of processing.

Just can obtain a chromosome like this through above step, carry out the G&T algorithm n time, just can obtain n bar chromosome.Active scheduling collection can be the search volume of problem be limited to this initial population, thereby the operation efficiency of genetic algorithm can be improved as genetic algorithm.

Simulated annealing: the simulated annealing solution procedure that adopts here is described below:

Step 1. is selected to produce a scheduling scheme (current separating) in the 4th step of G&T algorithm at random, uses X ^cExpression, calculating target function value F (x ^c).The setting initial temperature is T ₀

Step 2. is with the current X that separates of corresponding matrix representation ^cThe workpiece processing order of every machine, select a machine M at random _kSelect two workpiece on this machine at random and exchange.For example, in 3 * 3 the scheduling problem, as the 1st workpiece (J of machine 2 ₂) and the 3rd workpiece (J ₃) when being selected, the result after the exchange is just shown in Fig. 4 (b).

After step 3. exchanges through above-mentioned workpiece, obtain new workpiece processing order, the individuality conflict in the 4th step of G&T algorithm will obtain dismissing, and produce a new scheduling scheme (separating).If separating before separating of obtaining and workpiece exchange is different, then establishes current separating and be one and separate X, and get into next step; If not, returning for the 2nd step selects a new exchange right;

Step 4. is then accepted exchange, and is made X if the current target function value of separating is optimized ^c=X.If not, whether accept exchange with following substep decision;

1. calculate current goal functional value F (x) and target function value F (x ^c) difference DELTA f, calculating probability value exp (Δ f/T);

2. generate the random number between (0,1), and compare with exp (Δ f/T);

3. if exp (Δ f/T)＞random (0,1) then accepts exchange, make X ^c=X; Otherwise, exchange and will do not accepted;

, exchange gets into next step when accepting.Otherwise it is right to return the next exchange of the 2nd step searching.

Step 5. is carried out the equilibrium state check.Whether inspection is enough little through the variation of the target function value that certain number of times exchange back obtains.The number of times of equilibrium state check is defined as from generation to generation.Carry out the equilibrium state check by four following steps.

1. repeating step 2 to 4, reach quantity from generation to generation up to exchange.Carry out following operation then;

(2) calculation of the current generation of the average value of the objective function

, and after an exchange after the average value of the objective function

3 Comparison and

The relative error

with the given tolerance values ε size;

4. when relative error during less than tolerance value; Promptly

then thinks and reaches equilibrium state in this temperature, gets into next step.Otherwise, remove calculating from generation to generation, get back to the 2nd step repetition workpiece exchange flow process;

Step 6. is confirmed the decay temperature of simulated annealing, T _New=α * T _Old, (α is the simulated annealing attenuation coefficient, generally gets about 0.8-0.9)

If step 7. exchange logarithm reaches predetermined quantity, algorithm finishes, and obtains the scheduling problem approximate solution.

Repeat above operation and stop up to algorithm, obtain separate in the separating of selection optimal objective function value.

Beneficial effect of the present invention mainly shows: the machined parameters of uncertainty is represented to complicated production environment in (1) with fuzzy number.Based on the satisfaction notion, be objective function with the minimum satisfaction of single workpiece with all product customer satisfactions, set up the job shop scheduling model of obfuscation parameter.

(2) the G&T algorithm is applied to the generation of genetic algorithm initial chromosome, the obfuscation parameter model of being set up is found the solution, through with the comparison of simulated annealing, verified the validity and the stability of algorithm, model is had finds the solution effect preferably.

(3) set up dynamic dispatching model based on the obfuscation parameter; Three kinds of dynamic disturbances incidents such as urgent arrival, machine breakdown and reparation of inserting single workpiece, workpiece expiration time change have been gone through; With improving genetic algorithm model is verified, and it is good and bad to have compared the scheduling scheme of carrying out scheduling again and unexecuted scheduling again after dynamic event takes place.

(4) be example with the TYDQ machine shop,, set up the dynamic dispatching model of obfuscation parameter in conjunction with the workshop practical condition, through with the comparison of primitive scheduling scheme, obtained result preferably.

Description of drawings

Fig. 1 is that similarity is calculated synoptic diagram.

Fig. 2 is the synoptic diagram of upset mutation operator.

Fig. 3 is the synoptic diagram of population structure.

Fig. 4 is the workpiece processing order and the synoptic diagram that exchanges of 3 * 3 scheduling problems, wherein is somebody's turn to do, (a) for before the workpiece exchange; (b) for after the workpiece exchange.

Embodiment

Below in conjunction with accompanying drawing the present invention is further described.

With reference to Fig. 1～Fig. 4, a kind of based on the obfuscation parameter job shop dispatch control method under the dynamic disturbances of separating that improves genetic algorithm, said control method may further comprise the steps:

1), confirms the objective function of obfuscation parameter job shop scheduling

Satisfy customer demand and client have different satisfactions to different product situation in order more relevantly to react job shop scheduling maximization; The job shop scheduling model of obfuscation parameter is proposed; Is objective function to maximize all product customer satisfactions with the minimum satisfaction of maximization, promptly

$\mathrm{<figure-callout\; id="1"\; label="max\; z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">max</figure-callout>}{z}_{}{}_1=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i{\mathrm{AI}}_i---\left(1\right)$

$\mathrm{<figure-callout\; id="2"\; label="max\; z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">max</figure-callout>}{z}_{}{}_2={\mathrm{AI}}_{\min }=\underset{i=1,\mathrm{\&Lambda;n}}{\min }\mathrm{AIi}---\left(2\right)$

Wherein, W=(w ₁, w ₂, Λ, w _n) be the weight coefficient of each product satisfaction at delivery date, w ₁The significance level that the expression client requires product i delivery date, Objective function z ₁, z ₂Represent all product customer satisfactions and minimum satisfaction respectively.

Order

z′＝y ₁z ₁+y ₂z ₂ (3)

Y=(y ₁, y ₂) be z ₁And z ₂Weight coefficient, y ₁+ y ₂=1.

Like this, the objective function of fuzzy job-shop scheduling problem is: ask one to satisfy z ^*Workpiece processing order, make:

z ^*＝max(z′) (4)

$\left\{\begin{array}{c}{z}^{\&prime;}=y_1{\mathrm{<figure-callout\; id="1"\; label="z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">z</figure-callout>}}_1+y_2{\mathrm{<figure-callout\; id="2"\; label="z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">z</figure-callout>}}_2\\ \mathrm{<figure-callout\; id="1"\; label="max\; z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">max</figure-callout>}{z}_{}{}_1=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i{\mathrm{AI}}_i\\ \mathrm{<figure-callout\; id="2"\; label="max\; z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">max</figure-callout>}{z}_{}{}_2={\mathrm{AI}}_{\min }=\underset{i=1,\mathrm{\&Lambda;n}}{\min }{\mathrm{AI}}_i\\ \underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i=1\\ {\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">y</figure-callout>}}_1+{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">y</figure-callout>}}_2=1\end{array}\right.---(4-1)$

2), improved genetic algorithm

(2.1) coding

Reasonably designing encoding mechanism has very big influence to the quality and benefits of genetic algorithm, when carrying out genetic algorithm encoding, must consider chromosomal legitimacy, feasibility, validity and the integrality that the problem solution space is characterized.

Here adopt coding based on operation, the problem of 3 workpiece, 3 machines for example, putative chromosomal is [1 23 11 322 3], wherein 1 represents workpiece J ₁, 2 represent workpiece J ₂, 3 represent workpiece J ₃Because each workpiece has three process, so each workpiece occurs in chromosome three times.The order that the corresponding workpiece of the operation of workpiece occurs in chromosome.1 represent workpiece J like first ₁The 1st procedure, represent workpiece J for second 1 ₁The 2nd procedure.Represent with Triangular Fuzzy Number the process time of each operation.

During Code And Decode, because of dispatching the thought that has adopted the rolling window in dynamic workshop, so the Code And Decode here all is to carry out to the workpiece in the scheduling window.

(2.2) generation of initial population

Repeatedly move the G&T algorithm and can produce an initial population.Because the 4th step of its G&T is to select at random, is the diversity that guarantees initial population, require the similarity of initial population can not be too big.Through the research of Sakawa, when individual similarity 0.8 or following the time, the initial population of generation has maximum stability.Here require the individual similarity of initial population can not be greater than 0.8.

JSP with 4 * 4 is that example is introduced individual calculation of similarity degree method.Workpiece processing sequence table on every machine of individual 1 and individual 2 is shown in Fig. 3-4.For individuality 1, on machine 1, the right of priority of workpiece is respectively 1,2,4,3.For individuality 2, the workpiece right of priority on the machine 1 is 2,1,4,3.Contrast two individuals with the workpiece on machine order, the workpiece that has an identical precedence relationship with workpiece 1 on the machine 1 is 3 and 4, so assignment is 2.In like manner, the workpiece that has an identical precedence relationship with workpiece 2 on the machine 1 is 3 and 4, so assignment 2.Have 1,2 and 4 with workpiece that workpiece 3 has an identical precedence relationship on the machine 1, so an assignment 3, the workpiece that has an identical precedence relationship with workpiece 4 on the machine 1 is 1,2 and 3, assignment 3.Workpiece assignment on the machine 1 adds up to 2+2+3+3=10 like this.Same method is handled other machine, and computation process is as shown in Figure 1.Through this method, the similarity that can obtain individuality 1 and individuality 2 is 0.917.

(2.3) fitness function design

Fitness function is used for individuality is estimated, and also is simultaneously the foundation of optimizing process development.Here adopt objective function individuality to be estimated as fitness function.

(2.4) selection and interlace operation

Selection operation is used to avoid the loss of effective gene, makes the big individuality of fitness be able to bigger probability survival to of future generation, thereby improves global convergence and counting yield.Interlace operation is used to produce a new individuality, and the merit of parent is inherited in the assurance filial generation as far as possible.Based on the G&T algorithm, the interlace operation below adopting produces a filial generation from two parents.

Step 2.4.1. carries out the 1st to 3 step of G&T algorithm, obtains set C and conflict set G;

Step 2.4.2. chooses body one by one wantonly with 50% equal probabilities from two parent individualities.In the set G that conflicts, select to have the operation of least confusion deadline, use the individual replacement of the parent of selecting and be expressed as

Step 2.4.3. carries out the 5th and the 6th step of G&T algorithm.

Obtained a new offspring individual through above operation, carried out c above operation and just can obtain c new offspring individual.

For keeping the individuality that has merit in (c+2) individuals of forming by c filial generation and 2 parents, the individuality that when hereditary next time, remains with two of following method selections.

A. in c filial generation individuality, selecting to have the individuality of maximum target functional value, also is that local rank is selected;

B. in remaining (c+1) individuals (2 parent with (c-1) individual filial generation) select to have the individuality of maximum target functional value.

Above-mentioned intersection and selection operation, because bigger c value means bigger progeny population, the generation probability of good filial generation will become than higher.Yet because these filial generations are to be produced by identical parent, the similarity between filial generation also will become than higher.Therefore in this algorithm, c is set at 5.

(2.5) mutation operation

No longer evolve when offspring's fitness that interlace operation produces, do not develop toward the good aspect, just mean the precocity convergence of algorithm.Mutation operation is exactly to overcome this situation to a certain extent.Here, adopt more common counter-rotating variation (Inversion Mutation), promptly select two gene locations at random, the gene of this position is exchanged each other.For example, choosing gene location for the individual parent1 of parent is that 1 and 6 gene exchanges reaching the purpose of variation, thereby obtains making a variation offspring individual son1, as shown in Figure 2.

(2.6) population structure

In service in genetic algorithm, must prevent to get into locally optimal solution.For preventing to converge to local solution, the population building method that adopts Sakawa to propose ^[71,72]Suppose to have 3 groups of populations, every group produces 10 individuals based on similarity.When the individuality of each group all converges to some spending, each group population is merged continuation evolve up to convergence, as shown in Figure 3.

(2.7) algorithm end condition

Owing in practical problems, use genetic algorithm, never allow to let it ceaselessly search for down, the optimum solution of problem is also unknown usually simultaneously, and therefore must design some approximate convergence criterions stops the algorithm process.Here with predefined maximum evolutionary generation N _MaxAs stop condition, with best separating so far as optimum solution.

3), dispatches again dynamic event when taking place:

Step 1. initialization scheduling times n=0, the time t that begins to dispatch ₀=0, the maximum number L of workpiece in the given window _w, workpiece to be processed is put into window, establishing the workpiece window is N _w, having accomplished the workpiece collection is N _c

Workpiece in the step 2. pair current window is carried out and is improved genetic algorithm, selects to make the optimum chromosome of objective function as scheduling result, produces scheduling scheme;

When step 3. takes place as following dynamic event, carry out once scheduling again, establish dynamic event t takes place to do constantly _c

1. start once scheduling again as inserting when single workpiece arrives, change step 4;

2. when the workpiece expiration time changes, start once scheduling again, change step 4;

3. when the machine breakdown incident takes place, start once scheduling again, change step 4;

Step 4. is revised item capable of using, with completed operation from workpiece window N _wMove into and accomplish workpiece collection N _c, produce the scheduling scheme of having accomplished operation; The order of not accomplishing operation shifts to an earlier date, with remaining operation as the new process of not dispatching.

1. will insert single workpiece and add workpiece window N _wAnd do not accomplish operation and form new operation to be processed, carry out scheduling again.If when slotting single workpiece arrived, certain machine was being processed certain procedure, then the automatically delaying certain hour is carried out scheduling again after this procedure is accomplished;

When 2. the workpiece expiration time changes, upgrade the workpiece information in the workpiece window, carry out scheduling again;

3. during machine breakdown, the workpiece of just on this machine, processing is shifted out the workpiece window, remaining workpiece is carried out scheduling again.When simulation clock runs to the residue workpiece and needs the operation in this machining, this workpiece is shifted out the workpiece window, the residue workpiece is carried out scheduling again.When machine is repaired, the workpiece that shifts out the workpiece window is moved into again, carry out scheduling again.

Repeat step 4 according to scheduling events again, accomplish, produce final scheduling scheme up to all workpiece processing.

This programme to the dynamic disturbances situation in actual job workshop, has considered that the workpiece expiration time changes, inserts three dynamic events such as single workpiece arrival and machine breakdown on the basis of last one scheme-obfuscation parameter scheduling model; Set up the dynamic dispatching model of obfuscation parameter; Find the solution with the improvement genetic algorithm, calculate and checking with example at last, and compare with the scheme of not dispatching again; The result shows that model can embody the actual dynamic event in workshop, and produces scheduling scheme preferably.

The practice of this method:

Step 1. is not used M ₁, M ₂..., M ₁₁Represent various production machines.

Step 2. is represented the time of different product at the line workstation with fuzzy number.

Step 3. is represented the installation and removal time of different product with fuzzy number.

Step 4. is represented the process time of the corresponding machinery of different product with fuzzy number.

Step 5. is represented the oil sealing vanning time of different product with fuzzy number.

Step 6. counting yield completion date.

Step 7. obtains product delivery date (sky make into hour).

The foundation of step 8. scheduling model

Before setting up the instance scheduling model, confirm some basic parameters earlier:

t _Ih-workpiece line workstation time;

t _Ip-workpiece processing the time (comprising set-up time and take-down time);

t _If-oil sealing vanning the time;

The dynamic dispatching model of setting up like this under the obfuscation parameter of machine shop is following:

z ^*＝max(z′) (1)

$\left\{\begin{array}{c}{z}^{\&prime;}=y_1{\mathrm{<figure-callout\; id="1"\; label="z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">z</figure-callout>}}_1+y_2{\mathrm{<figure-callout\; id="2"\; label="z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">z</figure-callout>}}_2\\ \mathrm{<figure-callout\; id="1"\; label="max\; z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">max</figure-callout>}{z}_{}{}_1=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i{\mathrm{AI}}_i\\ \mathrm{<figure-callout\; id="2"\; label="max\; z"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">max</figure-callout>}{z}_{}{}_2={\mathrm{AI}}_{\min }=\underset{i=1,\mathrm{\&Lambda;n}}{\min }\mathrm{AIi}\\ {\mathrm{AI}}_i=\frac{\mathrm{area}({\tilde{C}}_i\&cap;{\tilde{D}}_i)}{\mathrm{area}{\tilde{C}}_i}\\ {\tilde{C}}_i=t_{\mathrm{ih}}+t_{\mathrm{ip}}+t_{\mathrm{if}}\\ \underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i=1\\ {\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">y</figure-callout>}}_1+{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="HDA0000131223250000011.png"\; state="\{\{state\}\}">y</figure-callout>}}_2=1\end{array}\right.---(4-2)$

In this model; Wherein, general machining the time is consistent in and the preceding surface model.Be that the time of machining has been carried out specific processing in instance model, be divided into line time, machining time and oil sealing vanning time.In the model, y ₁Be made as 0.7, y ₃Be made as 0.3, expression enterprise pursues the satisfaction maximum of integral product emphatically.

Step 9. is confirmed the satisfaction evaluation value of client to each product.

Step 10. is confirmed each product customer satisfaction weight coefficient.

Step 11. dynamic event is described

11.1 confirm the damage and the reparation obfuscation time of machine

11.2 promptly insert single workpiece arrival time ambiguityization

The emulation of step 13. model and interpretation of result.

When system take place to insert single workpiece arrival dynamic event, optimal objective function value was 0.058 if do not dispatch then again, and again after the scheduling optimal objective function value be 0.181.

When system's generation machine breakdown and reparation dynamic event, optimal objective function value is 0.093 if do not dispatch then again, and scheduling back optimal objective function value is 0.234 again.

The model of present embodiment can obtain the scheduling scheme of good embodiment dynamic perfromance under the situation of higher target function value.Machine breakdown and reparation, slotting single workpiece arrive and can both in scheduling scheme, take into account, and the production that has met enterprise is actual.

Claims (1)

1. one kind based on the obfuscation parameter job shop dispatch control method under the dynamic disturbances of improving genetic algorithm, and it is characterized in that: said control method may further comprise the steps:

1), confirms the objective function of obfuscation parameter job shop scheduling

The objective function of fuzzy job-shop scheduling problem is: ask one to satisfy z ^*Workpiece processing order, make:

z ^*＝max(z′) (4)

$\left\{\begin{array}{c}{z}^{\&prime;}=y_1{z}_1+y_2{z}_2\\ \max z_1=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i{\mathrm{AI}}_i\\ \max z_2={\mathrm{AI}}_{\min }=\underset{i=1,\mathrm{\&Lambda;n}}{\min }{\mathrm{AI}}_i\\ \underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i=1\\ y_1+y_2=1\end{array}\right.---(4-1)$

2), adopt the said objective function of improved genetic algorithm for solving, detailed process is following:

(2.1) coding: the workpiece in the scheduling window is adopted the coding based on operation;

(2.2) generation of initial population: repeatedly move the G&T algorithm and produce an initial population;

(2.3) adopt objective function individuality to be estimated as fitness function;

(2.4) selection and interlace operation:

Adopt interlace operation from two parents, to produce a filial generation, specific as follows:

Step 2.4.1. selects the 1st road operation of all workpiece, adds set C; Suppose that machine can process any a plurality of operation at synchronization, respectively operate O among the set of computations C _IjkThe fuzzy deadline of ∈ C, note is done According to the comparison criterion of Triangular Fuzzy Number, confirm the operation O of least confusion deadline among the set C _I*j*k*, will with the operation O _I*j*k*Use same machine M _K*Operation O _Ijk*∈ C constitutes set G.Because operation O _Ijk*With operation O _I*j*k*Running time on have overlappingly, set G is conflict set;

Step 2.4.2. chooses body one by one wantonly with 50% equal probabilities from two parent individualities; In the set G that conflicts, select to have the operation of least confusion deadline, use the individual replacement of the parent of selecting and be expressed as

Step 2.4.3. is with the selected benchmark that is operating as among the step 2.4.2., through the EC of other operations in the set of the big operation of getting of fuzzy number update inconsistency successively ¹, EC ²And EC ³Value is removed operation from set C

Its next process is added set C, and calculate its corresponding fuzzy deadline;

Obtained a new offspring individual through above operation, carried out c above operation and obtain c new offspring individual.

For keeping the individuality that has merit in (c+2) individuals of forming by c filial generation and 2 parents, the individuality that when hereditary next time, remains with two of following method selections:

A. in c filial generation individuality, selecting to have the individuality of maximum target functional value, also is that local rank is selected;

B. in remaining (c+1) individuals, select to have the individuality of maximum target functional value;

(2.5) mutation operation: adopt the counter-rotating variation;

(2.6) population structure: when the individuality of each group all converges to some spending, each group population is merged continuation evolve up to convergence;

(2.7) with predefined maximum evolutionary generation N _MaxAs stop condition, with best separating so far as optimum solution;

3) when dynamic event takes place, dispatch, concrete steps are following again:

Step 1. initialization scheduling times n=0, the time t that begins to dispatch ₀=0, the maximum number Lw of workpiece puts into window with workpiece to be processed in the given window, and establishing the workpiece window is N _w, having accomplished the workpiece collection is N _c

Workpiece in the step 2. pair current window is carried out and is improved genetic algorithm, selects to make the optimum chromosome of objective function as scheduling result, produces scheduling scheme;

When step 3. takes place as following dynamic event, carry out once scheduling again, establish dynamic event t takes place to do constantly _c

1. start once scheduling again as inserting when single workpiece arrives, change step 4;

2. when the workpiece expiration time changes, start once scheduling again, change step 4;

3. when the machine breakdown incident takes place, start once scheduling again, change step 4;

Step 4. is revised item capable of using, with completed operation from workpiece window N _wMove into and accomplish workpiece collection N _c, produce the scheduling scheme of having accomplished operation; The order of not accomplishing operation in advance, with the residue operation as the new process of not dispatching:

1. will insert single workpiece and add workpiece window N _wAnd do not accomplish operation and form new operation to be processed, carry out scheduling again.If when slotting single workpiece arrived, certain machine was being processed certain procedure, then the automatically delaying certain hour is carried out scheduling again after this procedure is accomplished;

When 2. the workpiece expiration time changes, upgrade the workpiece information in the workpiece window, carry out scheduling again;

3. during machine breakdown, the workpiece of just on this machine, processing is shifted out the workpiece window, remaining workpiece is carried out scheduling again; When simulation clock runs to the residue workpiece and needs the operation in this machining, this workpiece is shifted out the workpiece window, the residue workpiece is carried out scheduling again; When machine is repaired, the workpiece that shifts out the workpiece window is moved into again, carry out scheduling again;

Repeat step 4 according to scheduling events again, accomplish, produce final scheduling scheme up to all workpiece processing.

Images