CN102938102B - A kind of with batch processor across operation unit dispatching method - Google Patents

Abstract

The present invention relates to a kind of with batch processor across operation unit dispatching method, comprise the following steps: 1, definition three kinds of different structures pheromones; 2, initialization information element; 3,, according to processing sequence, the every procedure before the batch processing operation of each part is assigned to machine; 4, according to time sequencing, by the per pass Operation Sequencing on each machine; 5, by part batching, scheduling batch processing operation; 6,, according to processing sequence, the every procedure after the batch processing operation of each part is assigned to machine; 7, according to time sequencing, by the per pass Operation Sequencing on each machine; 8, according to the solution forming, lastest imformation element; If 9 cycle-indexes reach the upper limit, or several times optimal solution is unchanged continuously, finishes; Otherwise, turn the 2nd step. The present invention can solve in production process part across cell scheduling problem; (2) the batch processing operation in can processing production process and the integrated scheduling of non-batch processing operation, and ensure operational efficiency.

Description

A kind of with batch processor across operation unit dispatching method

Technical field

The present invention relates to a kind of dispatching method of manufacturing system, particularly a kind of with batch processor across operation unitDispatching method, belongs to advanced production control and the optimizing scheduling field of manufacturing.

Background technology

Cell Manufacture System (CellularManufacturingSystem, CMS) is group technology (GroupTechnology, GT) in typical case's application in the field of manufacture, embody the philosophic theory of lean production. In CMS, by machine according to zeroThe similitude of part technique is divided into groups, and forms the relatively independent unit of working ability, and each unit can complete one or more zeroThe production process of part family. But in actual production, due to product day by day in variation and unit production capacity limited, there is certainThe situation that some roads part operation of a little parts need to be processed on the machine of other unit. Meanwhile, on the other hand, consider lifeProduce the reasons such as economics, budget and spatial limitation, for buying additional machine infeasible in each unit. Because this class machine is logicalOften expensive, startup once expends larger, considers the reasons such as economics of production, budget and spatial limitation, can only be by itBe placed in certain specific unit. These equipment, as scarce resource, often can be processed the part of multiple part families (belowMiddle by this equipment be called share close reset standby, CriticalSharedMachine, CSM). In this case, need to be acrossThe peculiar part (Exceptionalparts, EP) that multiple unit just can complete, the transfer between unit has just formed across unitBranch problem (inter-cellmove). Shift and cause each unit independently not dispatch across unit, need between unit collaborativeArrange the processing sequence of part.

As far back as the early 1990s in last century, Garza and Smunt just point out owing to shifting and being difficult to avoid across unit, desirableCMS will be difficult to carry out, and must quantitative analysis shift the impact that production system is produced across unit, but most of research concentrate on alwaysHow to carry out unit structure and how to carry out in the problems such as unit internal control. Until in recent years, along with the enforcement of CMS is dark graduallyEnter and meet difficulty, just starting to be concerned across this problem of cell scheduling, correlative study can be divided into across flowing water unit with across workTwo types of industry unit.

Across flowing water cell scheduling aspect, Yang and Liao consider part mobile transfer between two unit at the most, and only moveMoving transfer once. The shortest in target taking flow-time, adopt branch and bound method and heuritic approach, solve work in each unitThe processing sequence of order. But, along with plant configuration constantly changes to unit manufacture, shift and be popular tendency across cell moving, andBecome increasingly complex, therefore part mobile branch problem between unit more than two more merits attention. The people such as SolimanpurConsider peculiar part mobile branch problem between unit more than two, taking makespan minimum as target, adopted heuristicAlgorithm solves in two steps across cell scheduling problem. The people such as Golmohammadi have considered mobile transfer between unit, adopt improvedThe interior part of ElectroMagnetism-like (EM-like) algorithm integrated solution part family dispatching sequence and each part familyDispatching sequence. The people such as Mosbah to be to minimize makespan, the employing that total completion date and standby time are optimization aimExtendedGrateDeluge (EGD) algorithm and heuritic approach solve has peculiar part in Cell Manufacture SystemScheduling problem. In order to solve the Multi-Objective Scheduling of part in unit and between unit, the people such as Gholipour propose a kind of baseIn the meta-heuristic algorithm of scatter searching.

Across operation unit aspect, the people such as Tang use scatter searching method to solve peculiar part across cell scheduling problem.The people such as Elmi on the basis of people's problem models such as Tang, the permission part of can reentrying, the discontinuous operation of a part is passableOn same machine, process; Adopt the method for simulated annealing, the people's such as coded system and Tang scheme is identical, utilizes adjacent simultaneouslyNearly structure optimization is finally separated. The people such as Xiao have considered the situation that part stochastic and dynamic arrives, and adopt the Agent association based on pheromonesBusiness's method solved under flexible path across operation unit scheduling problem.

The problem model that the research across cell scheduling problem is considered in the research of above-mentioned bibliography be all according to based onTraditional scheduling problem model, the scheduling model that production model (as Flow Shop or scheduling model, job shop) is divided.But in actual production, Machine Type produces material impact to production scheduling equally, this problem derives from crucial zero of vehicleThe actual production process of part.

The process of part not only needs the machining operations such as turnning and milling plane mill, also needs to anneal, quenches, returnsThe heat treatment steps such as fire, carburizing are to obtain mechanical property and the physical property of expection. Show crucial zero of vehicle according to findingIn the production process of part, there is 35% the existing machined order of part also to have heat treatment step. In general, conventionally machine is added to the stageSeparately consider with the scheduling of heat treatment stages, reason is that the heat treatment time of parts adds the time much larger than machine conventionally. But weFinding show: see on the whole, though the machined order time is shorter, but quantity is more, and therefore whole machine adds the stageTime accounts for 39% left and right of production overall process, and heat treatment stages accounts for 42% left and right, and the two is about the same; On the other hand, IIn investigation, find that at present reach thousands of minutes the process time of existing a part of complicated machined order, these phenomenons all makeMachine adds stage and heat treatment stages in most important optimization index---on the time, mentioned in the same breath. Equipment for Heating Processing is as oneClass CSM, is only placed in a certain discrete cell. Therefore different part families shift across unit the access cause of Equipment for Heating ProcessingAlso be the transfer between uniprocessor (machining equipment) and batch processor (Equipment for Heating Processing) simultaneously. Machined order need to meet machine(part is with for the moment for device unique constraints (a machine synchronization can only be processed a part) and part unique constraintsQuarter can only be by a machining), belong to solve job shop scheduling problems category; Heat treatment step need not meet machine unique constraints,Belong to lot size scheduling problem category. Classical job-shop scheduling problem is a NP-hard problem, and lot size scheduling problemIntroducing makes it more complicated. At present there is not yet achievement from the angle analysis of different device types across the research of cell scheduling problem sends outTable.

Summary of the invention

The object of the invention is for the deficiencies in the prior art, under across unit collaboration mode, for batch processorOperation unit finds efficient feasible dispatching method, to minimize completion date, to maximize batch processor utilization rate and minimumChanging the stand-by period between non-batch processing operation and batch processing operation is target, ensures Cell Manufacture System entirety fortune efficientlyDo.

The manufacturing system that the present invention considers is composed of multiple units, and has some the machines that working ability is different in each unitDevice, can complete the production of the part family of resemble process; In system, there is and only have a batch processor, be placed on one of them listIn unit; Uniprocessor once can only be processed a part, and batch processor can be processed multiple parts simultaneously; Manufacturing system is also fullBe enough to lower condition:

1) all parts arrived in zero moment;

2) batch processor has certain spatial limitation, and the dimensions of part is known, is being no more than the condition of spatial limitationUnder, the part of Same Part family can be processed simultaneously in batch processor;

3) operation that the process of part has order constraint by multiple tracks forms, and wherein each part has and only has togetherOperation need to complete in batch processor, and batch processing operation is not last procedure;

4) because machinery processing capacity between unit partly overlaps, cause part to there is flexible path, but for certain zeroThe certain working procedure of part, in each unit, have at the most one can processing machine and process time known;

5) be the maximum of all part batch processing times in current batch the process time of batch processing operation;

6) consider transfer time between unit, and ignore transfer time unit in, between different units transfer time because of distance withPart classification and difference;

7) consider the main time between different part families, and while ignoring the less important preparation of Same Part family insideBetween,, in the time processing continuously the part of Same Part family on uniform machinery, ignore time; Wherein main time is knownAnd do not change and change with scheduling sequence;

8) buffer pool size before batch processing equipment is enough large;

9) every machining part is all the non-preemptive type of non-interruption;

10) do not consider the situation of mechanical disorder, shortage of raw materials and operating personnel's vacancy.

For the manufacturing system that meets above condition, the invention provides a kind of adjusting across operation unit with batch processorDegree method, comprises the following steps:

The 1st step: be defined as follows the index shown in table and variable:

Table 1 index and variable

Meanwhile, the pheromones of three kinds of different structures of definition:

A) the pheromones structure in operation assignment

Select in the process of machine in operation, the matrix that defines O × M size represents pheromones, and wherein O represents workOrder sum, M represents machine sum, the element (O in matrix_ij, k) represent operation O_ijOn machine k, process corresponding pheromones denseDegree;

B) the pheromones structure in Operation Sequencing

When Operation Sequencing on every machine, the matrix that defines M O × O size represents pheromones, and wherein O representsOperation sum, the element (O in m matrix_ij, k) represent the upper operation O of machine m_ijK the letter that processing is corresponding on this machineCease plain concentration;

C) the pheromones structure in batch processor operation batching

In the process of batch processing operation batching, the matrix of definition N × N size represents pheromones, and wherein N represents partSum, element (i, j) represents that part i and part j are at same batch of corresponding pheromone concentration;

Select a part to add existing batch at every turn from optional parts list when batch processing operation batching, and batch numberOrder cannot be determined, therefore there is definition shown in formula (1):

Wherein, τ_i，bRepresent part i to add the pheromone concentration that batch b is corresponding, τ_i，kRepresent that part i and part k are sameOne batch of corresponding pheromone concentration, | B_b| represent B_bIn existing part number; If represent, batch b be sky to formula (1), zeroPart i add pheromone concentration that batch b is corresponding be part i respectively with batch b in existing part the pheromones of same batchConcentration sum, otherwise be definite value 1;

The 2nd step:

Initialize transfer distance, the part sum that must process and each between input machine information, dividing elements, unitThe technique information of part, then according to following explanation initialization information element:

A) initialize the pheromones in operation assignment

Wherein, τ_i，j，mRepresent operation O_ijOn machine m, process corresponding pheromone concentration, ε is that pheromone concentration is initialValue, is decided to be 0.01;

B) initialize the pheromones in Operation Sequencing

Wherein, τ on machine m_{m，i，j，k}Represent operation O_ijAt k the pheromone concentration that processing is corresponding, ε is that pheromones is denseDegree initial value, is decided to be 0.01;

C) initialize the batch processor operation pheromones in batches

Wherein, τ_i，kRepresent that part i and part j process corresponding pheromone concentration on same batch, ε is that pheromones is denseDegree initial value, is decided to be 0.01;

The 3rd step:

According to processing sequence, the every procedure before the batch processing operation of each part is assigned to machine, for oftenIndividual part is followed successively by the selected processing machine of every procedure according to the processing sequence of operation, and every selected probability of machine is:

${\mathrm{Pr}}_{i,j,m}=\frac{{\mathrm{\τ}}_{i,j,m}^{{\mathrm{\α}}_1}{\mathrm{\ρ}}_{i,j,m}^{{\mathrm{\β}}_1}}{{\mathrm{\Σ}}_{k=1}^M{\mathrm{\τ}}_{i,j,k}^{{\mathrm{\α}}_1}{\mathrm{\ρ}}_{i,j,k}^{{\mathrm{\β}}_1}}---\left(5\right)$

Wherein, Pr_i，j，mRepresent operation O_ijThe probability of processing on machine m, ρ_i，j，kRepresent corresponding heuristic information, α₁、β₁Represent respectively pheromone concentration, the shared weight of heuristic information;

Due to considered part across unit transfer time, therefore ρ_i，j，kBe defined as follows:

${\mathrm{\ρ}}_{i,j,k}=\frac{1}{P_{i,j,k}+{\mathrm{TT}}_i{D}_{m^{\′},k}}---\left(6\right)$

Wherein, P_i，j，kRepresent O_ijProcess time on machine k, TT_iD_m′，kRepresent part i unit distance transfer time withThe machine m ' of preceding working procedure processing is long-pending to the transfer distance of machine k, i.e. transfer time corresponding to part i; By heuristic informationFormula can be found out, preferential Choice and process time and the less machine of sum transfer time in the time that operation is assigned;

Obtain after probability that operation processes on every optional machine, with roulette algorithm at random selected certain machine addThis operation of work;

The 4th step:

According to time sequencing, by the per pass Operation Sequencing on each machine, on the basis of assigning in operation, determine everyProcessing sequencing and the initial time of operation on machine; Concrete grammar is:

For every machine, can dispatch operation from it, according to probability shown in following formula, select at random with roulette algorithmSelecting one procedure arrangement processes at the next one; Repeat this process, until all process steps is all scheduled;

${\mathrm{Pr}}_{m,i,j,k}=\frac{{\mathrm{\τ}}_{m,i,j,k}^{{\mathrm{\α}}_2}{\mathrm{\ρ}}_{m,i,j,k}^{{\mathrm{\β}}_2}}{{\mathrm{\Σ}}_{l=1}^O{\mathrm{\τ}}_{m,i,j,l}^{{\mathrm{\α}}_2}{\mathrm{\ρ}}_{m,i,j,l}^{{\mathrm{\β}}_2}}---\left(7\right)$

Wherein, Pr_{m，i，j，k}Represent the upper O of machine m_ijAt the probability of k processing, ρ_{m，i，j，l}Represent corresponding heuristic letterBreath, α₂、β₂Represent respectively pheromone concentration, the shared weight of heuristic information;

The heuristic information of Job Scheduling has considered the time across unit transfer time and machine of part, withTime part the time across unit transfer time and machine in fact can be overlapping, in the transfer process of partIn, corresponding machine can start to prepare for processing this part, therefore heuristic information defines suc as formula shown in (8):

Wherein, f_i，j-1Represent the time of the j-1 procedure process finishing of part i, D_m′，mRepresent from processing parts oneThe machine m ' of procedure is to the transfer distance of machine m, l_i，jThe affiliated part family of part for previous processing on marking machine:l_i，j=0 represent part i on machine on first processing or machine part and the part i of previous processing belong to Same PartFamily, l_i，j=1 represents that part and the part i of previous processing on machine do not belong to Same Part family, TE_m，i，jRepresent on machine mO_ijFinish time of last procedure, max (f_i，j-1+TT_iD_i，j，m，TE_m，i，j+l_i，jST_i，m)-TE_m，i，jRepresent in the past togetherOperation finishes to O_ijStart the real time of machining cost; From heuristic information formula, in the time of Operation Sequencing, preferentially adjustDegree can more early start the operation of processing;

In Operation Sequencing process, for any machine, it can dispatch operation is to be assigned in the operation of this machine,The set of the operation that the first operation of arbitrary part or last procedure have been sorted;

The 5th step:

By part batching, scheduling batch processing operation, i.e. the batch processing operation of which part is placed on batch processor by decision-makingProcess for same batch, concrete grammar is:

Step1. select the time of advent of part the earliest, add this batch;

Step2. upgrade candidate's part collection, candidate's part collection is defined as: if b is arbitrary batch, and its candidate's part collectionFor following set:

${\mathrm{CL}}_b=\left\{k\right|J_k=J_l,\&ForAll;l\&Element;B_b{\mathrm{andS}}_k<C_B-{\mathrm{\&Sigma;}}_{l\&Element;B_b}{S}_l{\mathrm{and\&Delta;WIS}}_b<0\}---\left(9\right)$

Be candidate's part collection of arbitrary batch, can not make with part Same Part family wherein and after adding this batch for allIn batch, the total size of part exceedes the set of the part of batch processor capacity or batch waste and free space increase;

For a batch of b of batch processor, the waste of b and free space WIS_bFor this batch of wasting space WS_bAnd the free timeSpace IS_bSum, Δ WIS_b" be the variable quantity in the WIS space of batch b; For the scheduling solution S of a batch processing operation, WIS (S)Represent waste and the free space of S, equal the WIS sum of all batches of S; WS and IS are defined as follows:

${\mathrm{WS}}_b=C_B\&CenterDot;P_b-{\mathrm{\&Sigma;}}_{i\&Element;B_b}{S}_i\&CenterDot;P_{i,m}---\left(10\right)$

IS_b＝C_B·(BS_b-BE_b-1)(11)

Wherein, BS_b，BE_bRepresent respectively the start and end time of batch b, P_i，mFor the batch processing operation of part i adds man-hourBetween, and in the time of b=0, make BE_b-1For initial time,

Therefore have:

$\begin{array}{c}{\mathrm{WIS}}_b={\mathrm{WS}}_b+{\mathrm{IS}}_b=C_B\&CenterDot;P_b-\underset{i\&Element;B_b}{\&Sigma;}{S}_i\&CenterDot;P_{iM}+C_B({\mathrm{BS}}_b-{\mathrm{BE}}_{b-1})\\ =C_B\&CenterDot;\left({\mathrm{BE}}_b-{\mathrm{BE}}_{b-1}\right)-{\mathrm{\&Sigma;}}_{i\&Element;B_b}{S}_i\&CenterDot;P_{i,m}\end{array}---\left(12\right)$

Step3. according to following formula calculated candidate part collection CL_bThe probability that middle part is selected, then calculates with rouletteMethod selects one of them part to add this batch, wherein CL_bDefined by formula (9);

Wherein, α₃、β₃Represent respectively pheromone concentration, the shared weight of heuristic information, Pr_i，bRepresent that part i adds a batch bProbability, ρ_i，bRepresent heuristic information, be defined as follows:

ΔWS_i，b＝WS_b′-WS_b＝+C_B·(P_b′-P_b)-S_iP_i，m(15)

B ' representative batch b add after part i batch, WS_b′Represent to add the wasting space of batch b ' after part i;

If Step4. candidate's part collection is not empty, goes to Step2, otherwise continue to carry out Step5;

If Step5. also have not yet batching of part, go to Step1, continue to set up next batch, otherwise finish;

The 6th step:

According to processing sequence, the every procedure after the batch processing operation of each part is assigned to machine, method is with the 3rdStep;

The 7th step:

According to time sequencing, by the per pass Operation Sequencing on each machine, method is with the 4th step;

The 8th step:

According to the solution forming, lastest imformation element, update rule is:

For in optimal solution set, each is separated,

If a) operation O_iiBe assigned to machine m,

τ_i，j，m＝(1-ρ)·τ_i，j，m+ρ·Δτ

If b) operation O_iiK processing on machine m,

τ_{m，i，j，k}＝(1-ρ)·τ_{m，i，j，k}+ρ·Δτ

If c) part i adds a batch b,

${\mathrm{\&tau;}}_{i,k}=(1-\mathrm{\&rho;})\&CenterDot;{\mathrm{\&tau;}}_{i,k}+\mathrm{\&rho;}\&CenterDot;\mathrm{\&Delta;}\mathrm{\&tau;},\&ForAll;k\&Element;B_b$

Wherein, τ_i，kRepresent that part i adds pheromone concentration corresponding to man-hour on same batch with part k, ρ represents informationElement volatility, Δ τ is pheromones renewal amount:

$\mathrm{\&Delta;}\mathrm{\&tau;}=Q\&CenterDot;({\mathrm{\&gamma;}}_1\frac{C_{max}^l}{C_{\max }^g}+{\mathrm{\&gamma;}}_2\frac{R^1}{R^g}+{\mathrm{\&gamma;}}_3\frac{{\mathrm{\&Delta;TW}}^l}{{\mathrm{\&Delta;TW}}^g})$

Q is pheromones renewal amount factor of influence, γ₁，γ₂，γ₃Be the weights of three optimization aim, represent respectively threeOptimization aim: the weight of stand-by period between maximum completion date, batch processor utilization rate, non-batch processing operation and batch processing operationVisual range degree, and γ₁+γ₂+γ₃＝1，Represent respectively total completion date of this scheduling solution, batch processingMachine utilization rate, the ratio of total waiting time and current optimal solution between non-batch processing operation and batch processing operation;

The 9th step:

If cycle-index reaches the upper limit, or several times optimal solution is unchanged continuously, finishes; Otherwise, turn the 2nd step.

The present invention adopts the pheromones update method of max-min ant system (MMAS), in the process of upgrading, all the timeKeep pheromone concentration in interval [τ_min，τ_max] interior (τ_minAnd τ_maxFor default definite value), the method is can avoiding method too earlyConverge on locally optimal solution.

Participate in upgrading only for this circulate in several locally optimal solutions (under be called optimal solution set close). Such renewalThe benefit of strategy is, makes the scope of ant group hunting concentrate near of more excellent solution, and don't can because of only use in circulation inningsPortion's optimal solution or globally optimal solution lastest imformation element and make convergence rate excessively slow.

Beneficial effect

The present invention is directed to across the solution proposing across operation unit scheduling problem with batch processor, mainly contain following3 beneficial effects:

(1) can solve in production process part across cell scheduling problem;

(2) the batch processing operation in can processing production process and the integrated scheduling of non-batch processing operation;

(3) operational efficiency is guaranteed.

Brief description of the drawings

Fig. 1 is solution flow chart.

Detailed description of the invention

Illustrate the preferred embodiment of the present invention below.

The present embodiment realizes according to the execution step in summary of the invention the band based on ant colony optimization algorithm that the present invention proposesHave batch processor across operation unit dispatching algorithm, as shown in Figure 1.

The present embodiment is carried out to following test simulation:

The present invention arranges as follows for the Cell Manufacture System of analog simulation: be provided with respectively 20 non-batch processors and 1Platform batch processor, under each machine, dividing elements is as shown in table 2. Non-batch processor is used 1～20 index, and batch processor is used 21Index.

Table 2 dividing elements

Annotation: in table, optional machine index M1-20 represents non-batch processor, and B21 represents batch processor.

The manufacturing process route of part is flexible, for same procedure, has the machine at least more than one unitDevice can be processed this procedure, and the corresponding processing time of different machines is generally not identical. The problem proposing for the present inventionModel, is provided with respectively 36 parts, and each part has different technological processes, and according to assumed condition, each part hasAnd only have batch processing operation one, all the other operations are non-batch processing operation, and batch processing operation is not first operation.

Corresponding can the generating by random algorithm by handling machine of every procedure of emulation experiment data. For non-batch of placeScience and engineering order, generates each 1～5 machine at random, lays respectively at different units. The size of part is choosing at random between 20～40Get, the process number of each part is chosen at random between 5～19. The displacement of different parts between unit is different, mobile distanceFrom choosing at random between 6～50, because the part that belongs to different part families switches the preparation producing on same machineTime is chosen at random between 1～10. Choose at random the process time of the non-batch processing operation of per pass between 2～50, batch processingChoose at random the process time of operation between 100～200.

The environment of emulation experiment is as follows:

(1) operating system: MicrosoftWindows732 position

(2)CPU：IntelCore2DuoCPUT64002.00GHz

(3) internal memory: 2G

(4) development environment: MicrosoftVisualStudio2010

(5) development language: C

Parameter in ACO algorithm has very large impact to its performance. Make parameter alpha₁＝α₂＝α₃=1, by adjusting differenceβ₁，β₂，β₃Value reflection pheromone concentration and the shared weight of heuristic information. In addition, pheromones volatile ratio ρ, pheromones is moreNew amount factor of influence Q, pheromone concentration maximum τ_maxAlso the performance of ACO is had to impact in various degree etc. parameter, therefore carry outThe reasonable value of parameter is determined in hereinafter described experiment. Because Q value is only relevant with the renewal amount of pheromone concentration, therefore only need determine QWith τ_maxThe reasonable value of ratio, therefore in when experiment by τ_maxBe decided to be constant 5, to simplify experimentation. In this paperIn algorithm, need in definite parameter and emulation experiment span in table 3.

The experimental design of table 3 parameters simulation

Parameter	Scope
		β1＝β2	(0.2，0.5，1，2)
β3	(0.01，0.1，0.5，1)
		Q/τ_max	(0.1，0.2，0.4)
ρ	(0.1，0.2，0.4)

For the parameter setting in table 3, when emulation, carry out contrast test. Different owing to defining 3 in problem modelRegulation goal, and wherein delivery date be one comparatively common and important in production reality, therefore, when emulation experiment is carried out,The weight of delivery date, batch processor utilization rate, total waiting time is set to respectively to 0.7,0.2,0.1, and evaluation algorithms uses notWhen the solution that same parameter obtains good and bad, use the weighted sum of the rank of the value of separating three corresponding regulation goals in all solutionsAs standard. Experimental result (limit by length, only list the preceding front 20 groups of data of weighting rank herein) as shown in table 4:

Table 4 parameters simulation experimental result

Can draw β from experimental result₁＝2，β₃=0.01, ρ=0.2, when Q/ τ _ max=0.2, the performance of the solution obtainingOptimum.

Therefore, user realizes time of the present invention, and suggestion arranges as follows to parameter:

Parameter	Span	Recommended value
			α1＝α2＝α3	1	1
β1＝β2	(0，4)	2
			β3	(0，4)	0.01
ρ	(0，1)	0.2
			τ_max	(1，10)	5
Q	(0，τ_max)	0.2*τ_max

The Performance Ratio of the present embodiment in, the performance of algorithm obtains based on above-mentioned recommended value.

In emulation experiment employing table 5, the combination of different candidate's part collection rule and (or) different heuristic rule is carried outContrast test (below use CACO method, CombinationalAntColonyOptimization, acute pyogenic infection of finger tip the present invention carriesThe algorithm going out).

In the time that batch processing operation is dispatched, adopt the concept of candidate's part collection, and used the inspiration of WS as ACO algorithmFormula information; In the time of the scheduling of other operations, the formula conduct that adopts the processing time and combine across unit transfer time, timeHeuristic information. For the performance of checking CACO method, contrived experiment is considered 2 kinds, and different candidate's part collection are regular and batching is heuristicInformation, a kind of operation is assigned heuristic information, and a kind of Operation Ordering Heuristics formula information is carried out group with CACOI-CACOIV algorithm respectivelyClose, above-mentioned 11 kinds of combined situation and algorithm CACO in this paper are contrasted. Specifically as shown in table 5.

The experimental design of table 5 contrast simulation

Below always to wait between maximum completion date, batch processor utilization rate, non-batch processing operation and batch processing operationTime three performance indications, the result of analysis emulation experiment. In order to carry out intuitively the Comprehensive Correlation of three performance indications, pointWhile analysing, calculate respectively the Gap value of other contrast combinations compared with CACO method in three performance indications, and according to 0.7,0.2,0.1Weight calculation weighting Gap value, and index using this Gap value as evaluation algorithms performance. Computing formula is as follows:

${\mathrm{Gap}}_i^M=\frac{{\mathrm{Makespan}}_i-{\mathrm{Makespan}}_{CACO}}{{\mathrm{Makespan}}_{CACO}}$

${\mathrm{Gap}}_i^U=\frac{{\mathrm{UseRate}}_{CACO}-{\mathrm{UseRate}}_i}{{\mathrm{UseRate}}_{CACO}}$

${\mathrm{Gap}}_i^W=\frac{{\mathrm{WaitTime}}_i-{\mathrm{WaitTime}}_{i_{CACO}}}{{\mathrm{WaitTime}}_{i_{CACO}}}$

${\mathrm{Gap}}_i=0.7\&CenterDot;{\mathrm{Gap}}_i^M+0.2\&CenterDot;{\mathrm{Gap}}_i^R+0.1\&CenterDot;{\mathrm{Gap}}_i^W$

In formula, Makespan, UseRate, WaitTime be the maximum completion date of three performance indications of acute pyogenic infection of finger tip, batch processing respectivelyBetween machine utilization rate, non-batch processing operation and batch processing operation,Represent respectively three performancesThe Gap value of a certain combination compared with CACO method in index, Gap_iFor its weighted value.

Can be found out by above-mentioned the simulation experiment result, in the present embodiment, employing FB still can be bright as candidate's part collectionAobvious lifting batch processor utilization rate, but can have a strong impact on all the other two performance indications, weighting Gap is many 10%～30%, performancePoor.

The completion date that the combination of CACOI+WIS+CACOIII+CACOIV solves is slightly better than the solution of CACO, but batch processorUtilization rate but has obvious decline, meanwhile, also has several combination, and in a certain performance indications, approach or be slightly better than CACO,But in all the other indexs, all there is certain gap. Can reach a conclusion in this case, candidate's part that CACO algorithm adoptsCollection strategy and heuristic information overall performance optimum.

It should be understood that present embodiment is instantiation of the invention process, should not be that the present invention protects modelThe restriction of enclosing. In the situation that not departing from spirit of the present invention and scope, foregoing is carried out equivalent amendment or changed allWithin should being included in the present invention's scope required for protection.

Claims (2)

1. With batch processor across an operation unit dispatching method, comprise the following steps:

   The 1st step: be defined as follows the index shown in table and variable:

   Table 1 index and variable

   Meanwhile, the pheromones of three kinds of different structures of definition:

   A) the pheromones structure in operation assignment

   Select in the process of machine in operation, the matrix that defines O × M size represents pheromones, and wherein O represents that operation is totalNumber, M represents machine sum, the element (O in matrix_ij, k) represent operation O_ijOn machine k, process corresponding pheromone concentration;

   B) the pheromones structure in Operation Sequencing

   When Operation Sequencing on every machine, the matrix that defines M O × O size represents pheromones, and wherein O represents operationSum, the element (O in m matrix_ij, k) represent the upper operation O of machine m_ijK the pheromones that processing is corresponding on this machineConcentration;

   C) the pheromones structure in batch processor operation batching

   In the process of batch processing operation batching, the matrix of definition N × N size represents pheromones, and wherein N represents that part is totalNumber, element (i, j) represents that part i and part j are at same batch of corresponding pheromone concentration;

   Select a part to add existing batch at every turn from optional parts list when batch processing operation batching, and batch number withoutMethod is determined, therefore there is definition shown in formula (1):

   Wherein, τ_i，bRepresent part i to add the pheromone concentration that batch b is corresponding, τ_i，kRepresent that part i and part k are with a collection ofThe pheromone concentration of inferior correspondence, | B_b| represent B_bIn existing part number; If represent, batch b is not empty to formula (1), and part i addsEnter pheromone concentration that batch b is corresponding and be part i respectively with batch b in existing part the pheromone concentration of same batch itWith, otherwise be definite value 1;

   The 2nd step:

   Initialize part sum and each part of transfer distance, palpus processing between input machine information, dividing elements, unitTechnique information, then according to following explanation initialization information element:

   A) initialize the pheromones in operation assignment

   Wherein, τ_i，j，mRepresent operation o_ijOn machine m, process corresponding pheromone concentration, ε is pheromone concentration initial value, is decided to be0.01；

   B) initialize the pheromones in Operation Sequencing

   Wherein, τ on machine m_{m，i，j，k}Represent operation o_ijAt k the pheromone concentration that processing is corresponding, ε is pheromone concentrationInitial value, is decided to be 0.01;

   C) initialize the batch processor operation pheromones in batches

   Wherein, τ_i，kRepresent that part i and part k process corresponding pheromone concentration on same batch, at the beginning of ε is pheromone concentrationInitial value, is decided to be 0.01;

   The 3rd step:

   According to processing sequence, the every procedure before the batch processing operation of each part is assigned to machine, for each zeroPart is followed successively by the selected processing machine of every procedure according to the processing sequence of operation, and every selected probability of machine is:

   Wherein, Pr_i，j，mRepresent operation o_ijThe probability of processing on machine m, ρ_i，j，kRepresent corresponding heuristic information, α₁、β₁PointDo not represent pheromone concentration, the shared weight of heuristic information;

   Due to considered part across unit transfer time, therefore ρ_i，j，kBe defined as follows:

   Wherein, P_i，j，kRepresent o_ijProcess time on machine k, TT_iD_m′，kRepresent part i unit distance Yu Qian transfer time roadThe machine m ' of operation processing is long-pending to the transfer distance of machine k, i.e. transfer time corresponding to part i; By heuristic information formulaCan find out preferential Choice and process time and the less machine of sum transfer time in the time that operation is assigned;

   Obtain after probability that operation processes on every optional machine, with roulette algorithm, selected certain machining should at randomOperation;

   The 4th step:

   According to time sequencing, by the per pass Operation Sequencing on each machine, on the basis of assigning in operation, determine every machineProcessing sequencing and the initial time of upper operation; Concrete grammar is:

   For every machine, can dispatch operation from it, according to probability shown in following formula, select one so that roulette algorithm is randomProcedure is arranged in next processing; Repeat this process, until all process steps is all scheduled;

   Wherein, Pr_{m，i，j，k}Represent the upper O of machine m_ijAt the probability of k processing, ρ_{m，i，j，l}Represent corresponding heuristic information, α₂、β₂Represent respectively pheromone concentration, the shared weight of heuristic information;

   The heuristic information of Job Scheduling has considered the time across unit transfer time and machine of part, the while zeroThe time across unit transfer time and machine of part in fact can be overlapping, in the transfer process of part, rightThe machine of answering can start to prepare for processing this part, therefore heuristic information defines suc as formula shown in (8):

   Wherein, f_i，j-1Represent the time of the j-1 procedure process finishing of part i, D_m′，mRepresent one work from processing partsThe machine m ' of order is to the transfer distance of machine m, l_i，jFor part family: l under the part of previous processing on marking machine_i，j＝0 represent part i on machine on first processing or machine part and the part i of previous processing belong to Same Part family, l_i，j=1 represents that part and the part i of previous processing on machine do not belong to Same Part family, TE_m，i，jRepresent the upper O of machine m_ijBeforeThe finish time of one procedure, max (f_i，j-1+TT_iD_i，j，m，TE_m，i，j+l_i，jST_i，m)-TE_m，i，jRepresent a procedure knot in the pastBundle is to o_ijStart the real time of machining cost; From heuristic information formula, in the time of Operation Sequencing, priority scheduling can be comparativelyEarly start the operation of processing;

   In Operation Sequencing process, for any machine, it can dispatch operation is to be assigned in the operation of this machine, arbitraryThe set of the operation that the first operation of part or last procedure have been sorted;

   The 5th step:

   By part batching, scheduling batch processing operation, i.e. the batch processing operation of which part is placed on the same of batch processor by decision-makingBatch process, concrete grammar is:

   Step1. select the time of advent of part the earliest, add this batch;

   Step2. upgrade candidate's part collection, candidate's part collection is defined as: if b is arbitrary batch, its candidate's part collection be withLower set:

   Be candidate's part collection of arbitrary batch, can not make with part Same Part family wherein and after adding this batch batch for allThe total size of middle part exceedes the set of the part of batch processor capacity or batch waste and free space increase;

   For a batch of b of batch processor, the waste of b and free space WIS_bFor this batch of wasting space WS_bAnd free spaceIS_bSum, Δ WIS_bFor the waste of batch b and the variable quantity of free space; For the scheduling solution S of a batch processing operation, WIS(S) represent waste and the free space of S, equal the waste of all batches and the free space sum of S; The wasting space of batch bWS_bWith free space IS_bBe defined as follows:

   IS_b＝C_B·(BS_b-BE_b-1)(11)

   Wherein, BS_b，BE_bRepresent respectively the start and end time of batch b, P_i，mFor the batch processing operation process time of part i,And in the time of b=0, make BE_b-1For initial time,

   Therefore have:

   Step3. concentrate the selected probability of part according to following formula calculated candidate part, then select it with roulette algorithmIn a part add this batch;

   Wherein, α₃、β₃Represent respectively pheromone concentration, the shared weight of heuristic information, Pr_i，bRepresent that part i adds that batch b's is generalRate,

   ΔWS_i，b＝WS_b‘-WS_b＝C_B·(P_b‘-P_b)-S_iP_i，m(15)

   B ' representative batch b add after part i batch, WS_b‘Represent to add the wasting space of batch b ' after part i;

   If Step4. candidate's part collection is not empty, goes to Step2, otherwise continue to carry out Step5;

   If Step5. also have not yet batching of part, go to Step1, continue to set up next batch, otherwise finish;

   The 6th step:

   According to processing sequence, the every procedure after the batch processing operation of each part is assigned to machine, method is with the 3rd step;

   The 7th step:

   According to time sequencing, by the per pass Operation Sequencing on each machine, method is with the 4th step;

   The 8th step:

   According to the solution forming, lastest imformation element, update rule is:

   For in optimal solution set, each is separated,

   If a) operation O_ijBe assigned to machine m,

   τ_i，j，m＝(1-ρ)·τ_i，j，m+ρ·Δτ

   If b) operation O_ijK processing on machine m,

   τ_{m，i，j，k}＝(1-ρ)·τ_{m，i，j，k}+ρ·Δτ

   If c) part i adds a batch b,

   Wherein, τ_i，kRepresent that part i adds pheromone concentration corresponding to man-hour on same batch with part k, ρ represents that pheromones wavesThe rate of sending out, Δ τ is pheromones renewal amount:

   Q is pheromones renewal amount factor of influence, γ₁，γ₂，γ₃Be the weights of three optimization aim, represent respectively three optimizationsTarget: the attention journey of stand-by period between maximum completion date, batch processor utilization rate, non-batch processing operation and batch processing operationDegree, and γ₁+γ₂+γ₃＝1，Represent respectively total completion date of this scheduling solution, batch processor profitBy rate, the ratio of total waiting time and current optimal solution between non-batch processing operation and batch processing operation;

   The 9th step:

   If cycle-index reaches the upper limit, or several times optimal solution is unchanged continuously, finishes; Otherwise, turn the 2nd step;

   In above step, the scope that arranges of customer parameter is:

   Parameter Span α₁＝α₂＝α₃ 1 β₁＝β₂ (0，4) β₃ (0，4) ρ (0，1) τ_max (1，10) Q (0，τ_max)

   τ _ max is a default definite value, and it has limited the maximum of pheromone concentration.
2. According to claim 1 a kind of with batch processor across operation unit dispatching method, it is characterized in that, according toCustomer parameter is set shown in following table:

   Parameter Recommended value α₁＝α₂＝α₃ 1 β₁＝β₂ 2 β₃ 0.01 ρ 0.2 τ_max 5 Q 0.2*τ_max