CN105321042A - Genetic algorithm-based advanced plan scheduling system and method - Google Patents

Abstract

The present invention relates to a genetic algorithm-based advanced plan scheduling system and method. The system comprises a basic setting module, a data management module and a scheduling module. Optimal solutions are successively generated from potential solution populations by using the characteristics of mutation and population evolution in a genetic algorithm, so as to solve the scheduling problems of machining workshops of aviation enterprises.

Description

A kind of high-level plan program system based on genetic algorithm and method

Technical field

The present invention relates to a kind of high-level plan program system based on genetic algorithm and method, belong to computer control managing technical field.

Background technology

The military enterprises such as Aeronautics and Astronautics belong to typical discrete manufacturing business, and Workshop Production generally has product structure complexity, and parts are many, multi items, short run, and batch product and scientific research Flexible production, affect the factor of production many, the features such as plan difficulty is large.In actual production, staff planners need according to batch product task, scientific research mission, other fragmentary tasks, need to ensure that situation, technological preparation situation, material supply situation, personnel's situation etc. carry out arrangement and the scheduling of Production Scheduling Problem in conjunction with inventories, device resource situation, frock tool.Because actual conditions are complicated, the factor of impact plan is many, information acquisition data volume is large, a lot of workshop management software is greatly mainly with some rules pre-defined, as preferential in the shortest process time, First Come First Served, delivery date priority scheduling the earliest, the search and the sequence that have " emphasis " formula is carried out, for Workshop Production Management person provides dispose by plan decision-making by this didactic rule.But there is in actual production process the complicacy of production factors usually, affect the randomness of plan factor, the binding character of working condition, and solve the feature such as multiple goal of site problems.Therefore, utilize general regular operational method, on fast search optimal case, the aspects such as multiple-objection optimization can not solve a NP difficult problem for this certified JobShop type well.

The method of high-level plan scheduling has a lot, mainly comprises the scheduling method based on planning, constraint planning CP, branch boundary B & B, decomposition technique DT, tabu search and genetic algorithm.Scheduling method based on planning implements simple and convenient, but rule is single, cannot realize the production scheduling of multiple constraint; Although constraint planning CP take into account various constraints in reality, it is mainly used for rule and goes out infeasible scheme and cut down a large amount of search volumes, and the rule base required is larger, cannot ensure to obtain optimum solution; Branch boundary B & B avoids and enumerates, and makes primal problem be divided into less subproblem, but calculates consumption internal memory consuming time; Decomposition technique DT can reduce optimization scale, the solution be comparatively satisfied with in finite time, but limits to more, is such as limited to the length and Moving Unit etc. of form.Tabu search and genetic algorithm similar, but the initial solution of tabu search is comparatively large to algorithm search performance impact, more difficult execution.Genetic algorithm has good robustness, can obtain comparatively ideal solution in np problem at Finite-time convergence.

In the research of Production Planning Scheduling, genetic algorithm also has more application.Tan Hui etc. (2005) and Zhou Xin etc. (2010) have studied the production scheduling problem under distributed production model, and the result of case verification shows that genetic algorithm can meet the scheduling requirement of distributed enterprise well.But the main constraints in these researchs is the factory being distributed in diverse location, very large with the constraint condition difference of discrete type manufacturing shop.Fan Yingli etc. (2002) solve the production planning problem of Reconfigurable Manufacturing System in conjunction with genetic algorithm and tabu search; Li Xiu etc. (2001) to have studied in batch production genetic algorithm in the application of Production Scheduling Problem.These two researchs all demonstrate the advantage that genetic algorithm solves production planning problem, but one is laid particular stress on enterprise's production schedule, lays particular stress on Production Scheduling Problem for one, do not inquire into the application of genetic algorithm at production scheduling.

Chinese patent application 200710200861.4, denomination of invention: production schedule automatic arrangement program system and method, applicant: great Fujin accurate industry (Shenzhen) Ltd.Which disclose a kind of production schedule automatic arrangement program system and method, provide main frame and the database of system, after the information such as order, Work station capacity that in main frame reading database, plan scheduling is relevant, automatically calculate and export Production Planning Scheduling result.But how this patent also not mentionedly realizes Production Planning Scheduling, the algorithm of the scheduling that also do not clarify a plan.

Chinese patent application 201410506043.7, denomination of invention: material production scheduling simulator, applicant: Wuhan Iron & Steel (Group) Corp..Which disclose a kind of material production scheduling simulator, provide unified data-interface, achieve the wireless transmission between Scheduling Algorithm server and database, terminal device.What this patent solved is hardware integration problem, does not relate to the algorithm of plan scheduling.

Summary of the invention

The technology of the present invention is dealt with problems: for the Workshop Production feature of the military enterprises such as Aeronautics and Astronautics, for the problem that plan scheduling difficulty is large, overcome the deficiencies in the prior art, a kind of high-level plan program system based on genetic algorithm and method are provided, utilize variation, Evolution of Population characteristic in genetic algorithm, in potential solution population, successively produce optimal solution plan, solve the scheduling problem of Aviation Enterprise workshop.

The technology of the present invention solution: based on the high-level plan program system of genetic algorithm, comprises basis instrument module, data management module and plan scheduling module three part, wherein:

Basis instrument module: carry out the material essential information in workshop, personal information, equipment and equipment group information, the work input of calendar information and maintenance, setting; Operating right setting is carried out to the application personnel in system; This module provides the support of Back ground Information and system cloud gray model information for data management module and plan scheduling module;

Data management module: incoming task and precedence information, processing technology information, operation resource information and inventory information.Wherein processing technology information comprises process information, operation information in man-hour, operation machining cell information, the operation configuration information of processing, and operation material requested information; Operation resource information comprises equipment and equipment group information, operation machining needs frock, tool information;

Plan scheduling module: the work calendar information in reading system basis instrument module, reads the task in data management module and precedence information, processing technology information, operation resource information and inventory information.Input schduling parameter N, T, Pc, Pm, N represent Population Size, and T represents iterations, and Pc represents crossover probability, and Pm represents mutation probability, call Scheduling Algorithm, export scheduling result, i.e. the Gantt chart of device resource dimension and task dimension.

Based on the high-level plan scheduling method of genetic algorithm, comprise preliminary scheduling and constraint adjustment two parts, wherein:

Preliminary scheduling is:

(1) schduling parameter N is inputted, T, Pc, Pm read the work calendar information in basis instrument module, read the task in data management module and precedence information, processing technology information, operation resource information and inventory information, be task coding, stochastic generation scale is the initial population A of N;

(2) every chromosomal Proper treatment of bar in initial population A is calculated, using the shortest total elapsed time as evaluation criterion, using the result that obtains after chromosomal for every bar total elapsed time gets inverse as this chromosomal appropriateness value;

(3) (computing formula is as follows to give different select probability Ps for the chromosomal appropriateness value of every bar, wherein i represents certain chromosome, its appropriateness value is designated as fi), adopt the mode of roulette from initial population A, select N bar chromosome, generate new population A ＇;

$Ps=f_i/\underset{i=1}{\overset{N}{\Σ}}{f}_i$

(4) according to the crossover probability Pc of input, the parent chromosome in population A ＇ is intersected between two, and the chromosome after intersection needs to carry out legitimacy inspection, and legal chromosome retains, and illegal chromosome abandons;

(5) according to the mutation probability Pm of input, the rearrangement of gene section is carried out to parent chromosome, needs equally after generating new chromosome to carry out legitimacy inspection;

(6) repeat this process T time, obtain the optimum solution of preliminary scheduling;

Constraint is adjusted to:

Occur inserting single, incur loss through delay or material supply changes time, need correspondingly to increase task, adjustment task priority information, processing technology information, operation resource information and inventory information, again scheduling obtain the scheduling result meeting current production status.

The present invention's advantage is compared with prior art:

(1) the most of problem all np problems in scheduling field, current method mainly based on planning and exhaustive method, can only solve small-scale scheduling problem, and it is comparatively slow to converge to optimum solution, and distance practical application also has larger problem.The present invention, by the high-level plan scheduling technology based on genetic algorithm, utilizes variation, Evolution of Population characteristic in genetic algorithm, in potential solution population, successively produces optimal solution plan, solve the scheduling problem of Aviation Enterprise workshop.

(2) the present invention is taking into full account that various resource content is (as the equipment in basis instrument module and equipment group, work calendar etc., order, processing technology, operation resource, stock etc. in data management module) situation, and combine relevant parameter configuration, use genetic algorithm, to the production task automatic arrangement program in workshop, for the production plan decision-making of workshop plan personnel every day provides scientific basis.Based on the high-level plan scheduling technology of genetic algorithm in the present invention, the utilize Local Search of specific priority rule or the sequence overcome in the past forms so-called " optimal case " shortcoming, and this optimal case is the best answer of non-globalize obtained under given conditions.And utilize genetic algorithm, by the interaction between colony, keep the information searched, this is incomparable based on the optimization method of single search procedure, in addition, selection, intersection that genetic algorithm uses, these three operators that make a variation are all random operations, instead of appointment is established rules then really.

Accompanying drawing explanation

Fig. 1 is composition frame chart of the present invention;

Fig. 2 is the realization flow figure of basis instrument module in Fig. 1;

Fig. 3 is the realization flow figure of data management module in Fig. 1;

Fig. 4 is the realization flow figure planning arranging module in Fig. 1;

Fig. 5 is the realization flow figure using genetic algorithm scheduling;

Fig. 6 is the convergence map using genetic algorithm scheduling;

Fig. 7 is the output Gantt chart (equipment dimension) of specific embodiment.

Embodiment

As shown in Figure 1, the present invention includes basis instrument module, data management module, plan scheduling module.

As shown in Figure 2, being implemented as follows of basis instrument module in the present invention:

1) material of workshop or plant produced processing is arranged, comprise material code, title, Zhu Zhi production division, and other attribute information;

2) information of personnel in system is arranged, comprise personnel's coding, personnel's attribute, and be subordinate to the information settings such as department;

3) the teams and groups' information comprised in production division is arranged, comprise teams and groups' coding, teams and groups' title, and the setup of attribute of teams and groups;

4) arrange the work calendar of workshop or factory, comprise the appointment of working day and nonworkdays, different departments can arrange different work calendars;

5) operating right setting is carried out to systematic difference personnel, comprise system administration authority, data manipulation authority, business function authority.

As shown in Figure 3, being implemented as follows of data management module in the present invention:

1) input order or the task in workshop, or obtain these orders or mission bit stream by interface from ERP system or MES system;

2) input the technique information of workshop process component, comprise the process information of processing, operation information in man-hour, operation machining cell information, operation configuration information, and operation material requested information;

3) input operation resource information, comprise equipment and equipment group information, frock, tool information that operation machining needs;

4) input the production inventory information in workshop or obtained the inventory information in workshop by interface from ERP system or MES system.

As shown in Figure 4, being implemented as follows of arranging module is planned in the present invention:

1) the work calendar information of production division in basis instrument module is read, as shown in table 3 equipment group break time schedule;

2) mission bit stream and task priority information in data management module is read, as shown in table 8;

3) the processing technology information in data management module is read, as table 1 task-operation-equipment group table, table 2 task-operation-time, table 5 operation starting material use shown in table and table 6 operation-starting material schedule of quantities;

4) the operation resource information in workshop in data management module is read, contained by table 4 equipment group shown in device data table;

5) inventory information in workshop in data management module is read, as shown in table 7 materials procurement table;

6) carry out automatic arrangement program by genetic algorithm, comprise and generate initial population, calculating moderate function, selection operation, interlace operation and mutation operation five steps, as shown in Figure 5.The first, for task is encoded, the initial population A of stochastic generation scale N=50; The second, calculate every chromosomal Proper treatment of bar in population A, using the shortest total elapsed time as evaluation criterion, using the result that obtains after chromosomal for every bar total elapsed time gets inverse as this chromosomal appropriateness value; 3rd, give different probability functions for the chromosomal appropriateness value of every bar, adopt the mode of roulette from population A, select 50 chromosomes, generate new population A ＇; (computing formula is as follows to give different select probability Ps for the chromosomal appropriateness value of every bar, wherein i represents certain chromosome, its appropriateness value is designated as fi), adopt the mode of roulette from initial population A, select N bar chromosome, generate new population A ＇; 4th, (adjust according to actual production with the crossover probability of 0.6, crossover probability value 0.6 in present case), parent chromosome in population A ＇ is intersected between two, chromosome after intersection needs to carry out legitimacy inspection, legal chromosome retains, and illegal chromosome abandons; 5th, with 0.01 for mutation probability (adjusting according to actual production, mutation probability value 0.01 in present case), the rearrangement of gene section is carried out to parent chromosome, needs equally after generating new chromosome to carry out legitimacy inspection; 6th, repeat this process 50 times, finally obtain the scheduling result of one group of optimum.

$Ps=f_i/\underset{i=1}{\overset{N}{\Σ}}{f}_i$

7) scheduling result exports Gantt chart from the dimension of device resource, and can view the plan of every platform equipment in every day each period, as shown in Figure 7, black block wherein represents the unavailable time section of machine.Scheduling result is:

Chromosome (coding)=[3,4,4,1,2,5,2,3,3,3,5,1,6,4,1,2,2,4,1,6,6,3,2,5,1], T=54.

Wherein numeral 1,2,3,4,5 and 6 represents different tasks, and task goes out to represent i-th operation of this task for i-th time, and decoded result is:

Chromosome (decoding)=[(J3,1), (J4,1), (J4,2), (J1,1), (J2,1), (J5,1), (J2,2), (J3,2), (J3,3), (J3,4), (J5,2), (J1,2), (J6,1), (J4,3), (J1,3), (J2,3), (J2,4), (J4,4), (J6,2), (J6,3), (J3,5), (J2,5), (J5,3), (J1,5)], T=54.

As can be seen from scheduling convergence map (as shown in Figure 6), during T=40, scheduling result restrains, and namely obtains optimum solution.

Task quantity is 6, operation maximal value 5 needed for every single task, and equipment group number is 5, and chronomere is 1.

Table 1 task-operation-equipment group table

Table 2 task-operation-timetable

Table 3 equipment group break time schedule

Group_num	Start_time	End_time
			Q21-4(M1)	0	16
Q21-4(M1)	40	45
			R5(M2)	0	10
X51(M3)	35	40
			X51(M3)	80	200
X51(M3)	250	450
			X51(M3)	900	950
SCX-73(M4)	30	40

Number of devices scale contained by table 4 equipment group

Group_num	Number
		Q21-4(M1)	2
R5(M2)	2
		X51(M3)	2

SCX-73(M4)	1
		CM6140(M5)	1

Table 5 operation-starting material use table

Table 6 operation-starting material schedule of quantities

Table 7 materials procurement table

Material	Quantity	Arrival Time
			R28966	65	0
R22269	75	0
			R19910	70	0
R21883	80	0
			R28966	100	10
R21883	25	20
			R22269	30	35

Table 8 task priority list

Task_num	Priority
		LJ586962(1)	1

LJ586961(2)	1
		LJ586949(3)	2
LJ586998(4)	2
		LJ586943(5)	2
LJ586922(6)	2

Claims (2)

1., based on the high-level plan program system of genetic algorithm, comprise basis instrument module, data management module and plan scheduling module three part, wherein:

Basis instrument module: carry out the material essential information in workshop, personal information, equipment and equipment group information, the work input of calendar information and maintenance, setting; Operating right setting is carried out to the application personnel in system; This module provides the support of Back ground Information and system cloud gray model information for data management module and plan scheduling module;

Data management module: incoming task and precedence information, processing technology information, operation resource information and inventory information, wherein processing technology information comprises process information, operation information in man-hour, operation machining cell information, the operation configuration information of part, and operation material requested information; Operation resource information comprises equipment and equipment group information, operation machining needs frock, tool information;

Plan scheduling module: the work calendar information in reading system basis instrument module, reads the task in data management module and precedence information, processing technology information, operation resource information and inventory information; Input schduling parameter N, T, Pc, Pm, N represent Population Size, and T represents iterations, and Pc represents crossover probability, and Pm represents mutation probability, call Scheduling Algorithm, export scheduling result, i.e. the Gantt chart of device resource dimension and task dimension.

2. based on the high-level plan scheduling method of genetic algorithm, it is characterized in that: comprise preliminary scheduling and constraint adjustment two parts, wherein: preliminary scheduling is:

(1) schduling parameter N is inputted, T, Pc, Pm, read the work calendar information in basis instrument module, read the task in data management module and precedence information, processing technology information, operation resource information and inventory information, be task coding, stochastic generation scale is the initial population A of N;

(2) every chromosomal Proper treatment of bar in initial population A is calculated, using the shortest total elapsed time as evaluation criterion, using the result that obtains after chromosomal for every bar total elapsed time gets inverse as this chromosomal appropriateness value;

(3) (computing formula is as follows to give different select probability Ps for the chromosomal appropriateness value of every bar, wherein i represents certain chromosome, its appropriateness value is designated as fi), adopt the mode of roulette from initial population A, select N bar chromosome, generate new population A ＇;

$Ps=f_i/\underset{i=1}{\overset{N}{\Σ}}{f}_i$

(4) according to the crossover probability Pc of input, the parent chromosome in population A ＇ is intersected between two, and the chromosome after intersection needs to carry out legitimacy inspection, and legal chromosome retains, and illegal chromosome abandons;

(5) according to the mutation probability Pm of input, the rearrangement of gene section is carried out to parent chromosome, needs equally after generating new chromosome to carry out legitimacy inspection;

(6) repeat this process T time, obtain the optimum solution of preliminary scheduling;

Constraint is adjusted to:

Occur inserting single, incur loss through delay or material supply changes time, need correspondingly to increase task, adjustment task priority information, processing technology information, operation resource information and inventory information, again scheduling obtain the scheduling result meeting current production status.