CN105159242A - Optimization scheduling method of textile fabrics dyeing process - Google Patents

Abstract

The invention relates to an optimization scheduling method of a textile fabrics dyeing process, and belongs to the technical field of intelligent optimization scheduling of a chemical production process. First of all, an optimization object is optimized through determining a fabrics dyeing process scheduling model and the optimization object by use of an optimization scheduling method based on a genetic-estimation of distribution algorithm (GA-EDA), wherein the scheduling model is established according to the maximum completion time of fabrics processed on each dyeing device, and at the same time, the optimization object is minimized maximum completion time. According to the invention, the expression of a fabrics dyeing process in chemical production is clear and accurate; the GA-EDA is enabled to have a global and local search capability; the GA-EDA can effectively overcome disadvantages in local search; and guidance is provided for a search direction by full use of excellent individual information, the search width and depth of the algorithm are enabled to be reasonably balanced, and the searching optimization capability of the algorithm is enhanced.

Description

A kind of Optimization Scheduling of weaving face fabric dyeing course

Technical field

The present invention relates to a kind of Optimization Scheduling of weaving face fabric dyeing course, belong to chemical process intelligent optimization dispatching technique field.

Background technology

Along with maintaining sustained and rapid growth of global economy, the development degree of process industry has become the important indicator weighing National Industrial level.It is one of most important ingredient of process industry that weaving face fabric is produced.By carrying out corresponding physics and chemistry processing to weaving face fabric, and then realize the appreciation of fabric value.In weaving face fabric production run, the occasion of dying operation application is quite wide.Dyeing course is usually used in the output of final fabric product, and the direct quality to product is produced material impact by the speed of the process of dyeing fabric and quality, significant to the optimization and upgrading of whole system structure.Therefore, in weaving face fabric production, the Optimized Operation of dyeing course has important research value.

Dyeing to fabric is common operation, and because the physicochemical property of fabric there are differences, some fabric can complete processing through once dyeing, and some fabric need to carry out secondary or repeatedly dying operation just can complete processing; Meanwhile, in order to prevent the cross pollution between product, every platform dyeing installation, after processing a kind of fabric, needs certain setup times to carry out cleaning and adjusting, and could continue to process another fabric, and setup times depends on the processing sequence between fabric; In addition, processing (equipment) unit of actual dying operation is often made up of the LPT device of multiple stage isomery, the working ability of these equipment is not quite similar, and fabric need be processed according to the equipment that the selecting factors such as the physicochemical property of self, volume, quality are suitable.This process is exactly typical isomerism parallel machine scheduling problem.The same with other production scheduling problems, the optimizing index of isomerism parallel machine scheduling problem mainly comprises Maximal Makespan, drags phase product number, average flowing-through time etc., wherein with Maximal Makespan (makespan or C _max) use the most extensive.Isomerism parallel machine scheduling problem belongs to NP-Complete problem, and its solution space exponentially increases with the increase of problem scale.Therefore, have higher reality and theory value to the research of isomerism parallel machine Scheduling Problem algorithm, the design that can be relevant Chemical Manufacture optimization system provides practical guidance.

Because dyeing fabric process scheduling problem belongs to NP complete category, traditional mathematic programming methods and Heuristic construction method cannot ensure majorization of solutions quality.Therefore, the present invention devises a kind of based on heredity-Estimation of Distribution Algorithm (GeneticAlgorithm-EstimationofDistributionAlgorithm, GA-EDA) Optimization Scheduling, can obtain the approximate optimal solution of weaving face fabric dyeing course scheduling problem within a short period of time.

Summary of the invention

The invention provides a kind of Optimization Scheduling of weaving face fabric dyeing course, for solving the problem obtaining the approximate optimal solution of dyeing fabric process scheduling problem in Chemical Manufacture within a short period of time.

The Optimization Scheduling of weaving face fabric dyeing course of the present invention is achieved in that

First by determining dyeing fabric process scheduling model and optimization aim, and the Optimization Scheduling based on heredity-Estimation of Distribution Algorithm GA-EDA is used to be optimized optimization aim; Wherein scheduling model according on every platform dyeing installation process fabric Maximal Makespan set up, optimization aim is C for minimizing Maximal Makespan simultaneously _max:

$\begin{array}{c}{C}_{\max }\left(\mathrm{\π}\right)=\max \{St\left({\mathrm{\π}}_{T_1}^{T\left(1\right)}\right)+P\left({\mathrm{\π}}_{T_1}^{T\left(1\right)}\right),St\left({\mathrm{\π}}_{T_2}^{T\left(2\right)}\right)+P\left({\mathrm{\π}}_{T_2}^{T\left(2\right)}\right),\mathrm{...},\\ St\left({\mathrm{\π}}_{T_m}^{T\left(m\right)}\right)+P\left({\mathrm{\π}}_{T_m}^{T\left(m\right)}\right)\}\end{array}---\left(1\right)$

$\begin{array}{c}St\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=St\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+P\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+S({\mathrm{\π}}_{k-1}^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)}),\\ pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=0,k=1,\mathrm{...},T_i,i=1,\mathrm{...},m\end{array}---\left(2\right)$

$\begin{array}{c}St\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=\max \{St\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+P\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+S({\mathrm{\π}}_{k-1}^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)}),\\ St\left({\mathrm{\π}}_{pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)}^{T(pre\_m({\mathrm{\π}}_k^{T\left(i\right)}))}\right)+P\left({\mathrm{\π}}_{pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)}^{T(pre\_m({\mathrm{\π}}_k^{T\left(i\right)}))}\right)\},\end{array}---\left(3\right)$

$\begin{array}{c}pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)\≠0,k=1,\mathrm{...},T_i,i=1,\mathrm{...},m\\ C_{\max }\left({\mathrm{\π}}^{*}\right)=\underset{\mathrm{\π}\∈\Π}{\min }{C}_{\max }\left(\mathrm{\π}\right)\end{array}--\left(4\right)$

${\mathrm{\π}}^{*}=\arg \left\{C_{max}\left(\mathrm{\π}\right)\right\}\→min,\∀\mathrm{\π}\∈\Π---\left(5\right)$

If product set is N, production equipment set is M, and wherein workpiece or product number are n, and number of devices is m.Often kind of product j (j ∈ (1 ..., n)) need stg _jroad manufacturing procedure, S _t=stg ₁∪ stg ₂∪ ... ∪ stg _nrepresent the set that the process number of all products is formed, the different operations of identical product need to process sequentially; All process steps can only be processed by the equipment meeting processing constraint in set M; The process time of product is relevant with process equipment, any equipment i (i ∈ (1 ..., m)) and synchronization can only process a kind of product; Be with sequence to be correlated with setup times between different product, setup times depends on processing sequence, and the setup times between identical product is 0.

Order for total process number of all products, π=[π ₁, π ₂..., π _tS] (π _j∈ (1 ..., n), j=1 ..., TS) and be the arrangement based on operation (product in this arrangement is assigned in individual device according to certain rule and processing constraint from left to right and processes) of a n to be processed product or workpiece, T _iit is the operation sum that i-th equipment is processed ${\mathrm{\π}}^{T\left(i\right)}=\[{\mathrm{\π}}_1^{T\left(i\right)},{\mathrm{\π}}_2^{T\left(i\right)},...,{\mathrm{\π}}_{T_i}^{T\left(i\right)}\]({\mathrm{\π}}_k^{T\left(i\right)}\∈(1,...,n),k=1,...,T_i)$ For the arrangement based on operation of institute's processing work on i-th equipment, for process time $(k=0,...,T_i,P({\mathrm{\π}}_0^{T\left(i\right)})=0),$ for with between setup times $(k=1,...,T_i,S({\mathrm{\π}}_0^{T\left(i\right)},{\mathrm{\π}}_1^{T\left(i\right)})=0,$ As k > 1 and ${\mathrm{\π}}_{k-1}^{T\left(i\right)}={\mathrm{\π}}_k^{T\left(i\right)}$ Time $S({\mathrm{\π}}_{k-1}^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)})=0),$ for beginning process time $(k=0,...,T_i,St({\mathrm{\π}}_0^{T\left(i\right)})=0),$ for a front machining device number (k=1 ..., T _i, when add man-hour first ), for front time processing at equipment in position from left to right (k=1 ..., T _i, when add man-hour first ), the target of optimization is find a π in the set Π of all model sequencing ^*, make completion date C the earliest _max(π) minimum;

Wherein, formula (1) to formula (3) is completion date C the earliest _max(π) computing formula, formula (4) and formula (5) expression find optimal sequencing π in the set Π of all model sequencing ^*, make C _max(π) minimum.The target of scheduling is find a π in the set Π of all fabric to be dyed sequences ^*, make Maximal Makespan C _max(π) minimum.

The concrete steps of the described Optimization Scheduling based on heredity-Estimation of Distribution Algorithm are as follows:

Step1, coded system: carry out coding π=[π with fabric to be dyed sequence ₁, π ₂..., π _tS]; Wherein TS represents total process number of all products;

The initialization of Step2, probability matrix and population:

Step2.1: initialization probability matrix: first need the initial probabilistic model or the matrix P that are configured to training _ori(gen)=[p _ij] _{n × TS}, initialized P _ori(gen) the allelic parameter probability valuing in all gene position is equal, i.e. p _ij=1/n (i=1 ..., n, j=1 ..., TS), initialized probability matrix P _ori(gen=1) form is such as formula shown in (6), and wherein gen represents the algebraically of Evolution of Population;

Step2.2: initialization population: GA-EDA adopts random fashion initialization population, and namely algorithm produces popsize individuality at random, and forming population scale is the initial population pop of popsize _candi(gen=1);

The update mechanism of Step3, probability matrix: GA-EDA uses history optimum individual update probability matrix, if π _local(gen)=π _{local_1}(gen) ..., π _{local_TS}(gen) for population is at the history optimum individual in gen generation, LR is learning rate, then probability matrix P _matrix(gen) following steps are adopted to upgrade:

Step3.1: establish x=π _{local_j}(gen), p _xj(gen)=p _xj(gen)+LR, wherein, j=1 .., TS;

Step3.2: probability normalization: $p_{wj}\left(gen\right)=p_{wj}\left(gen\right)/{\mathrm{\Σ}}_{y=1}^n{p}_{yj}\left(gen\right),$ Wherein, w=1 .., n, j=1 ..., TS.

Step4, generation new population:

Step4.1: produce candidate population: candidate population pop _candi(gen) be made up of two parts: (1) candidate population pop _candi(gen) in the individuality of e% according to the method for roulette to EDA probability matrix P _matrix(gen) sampling is formed; (2) individuality of all the other g% is by previous generation population pop _candi(gen-1) in, the individuality of the front g% that adaptation value is best is formed;

Step4.2: produce new population: by pop _candi(gen) order performs the genetic manipulation of GA, forms new population

Step5, end condition: the maximum iteration time of setting end condition is 200, if met, then export " optimum individual "; Otherwise go to Step3, iterate, until meet end condition.

Described population scale is set to popsize=50, learning rate LR=0.1, e%=0.6, g%=0.4.

The invention has the beneficial effects as follows:

The present invention proposes scheduling model and the dispatching method of dyeing fabric process in the Chemical Manufacture minimized under maximum completion index, make the expression of dyeing fabric process in Chemical Manufacture accurately clear, dispatching method is rationally effective; Adopt the training and operation method based on GA to obtain initialized probability matrix, improve the effective information accumulation of probability model at the algorithm evolution initial stage; By the combination of GA and EDA, GA-EDA is made to have overall situation and partial situation's search capability; EDA is made can effectively to overcome the deficiency on Local Search; The information being conducive to making full use of excellent individual comes guidance search direction, and then makes the search width of algorithm and the degree of depth obtain reasonable balance, and the optimizing ability of algorithm is enhanced;

Optimized Operation scheme based on GA-EDA proposed by the invention can effective dyeing fabric process scheduling problem in Chemical Manufacture.

Accompanying drawing explanation

Fig. 1 is the process flow diagram in the present invention.

Embodiment

Embodiment 1: as shown in Figure 1, a kind of Optimization Scheduling of weaving face fabric dyeing course, first by determining dyeing fabric process scheduling model and optimization aim, and the Optimization Scheduling based on heredity-Estimation of Distribution Algorithm GA-EDA is used to be optimized optimization aim; Wherein scheduling model according on every platform dyeing installation process fabric Maximal Makespan set up, optimization aim is C for minimizing Maximal Makespan simultaneously _max:

$\begin{array}{c}{C}_{\max }\left(\mathrm{\π}\right)=\max \{St\left({\mathrm{\π}}_{T_1}^{T\left(1\right)}\right)+P\left({\mathrm{\π}}_{T_1}^{T\left(1\right)}\right),St\left({\mathrm{\π}}_{T_2}^{T\left(2\right)}\right)+P\left({\mathrm{\π}}_{T_2}^{T\left(2\right)}\right),\mathrm{...},\\ St\left({\mathrm{\π}}_{T_m}^{T\left(m\right)}\right)+P\left({\mathrm{\π}}_{T_m}^{T\left(m\right)}\right)\}\end{array}---\left(1\right)$

$\begin{array}{c}St\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=St\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+P\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+S({\mathrm{\π}}_k^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)}),\\ pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=0,k=1,\mathrm{...},T_i,i=1,\mathrm{...},m\end{array}---\left(2\right)$

$\begin{array}{c}St\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=\max \{St\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+P\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+S({\mathrm{\π}}_{k-1}^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)}),\\ St\left({\mathrm{\π}}_{pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)}^{T(pre\_m({\mathrm{\π}}_k^{T\left(i\right)}))}\right)+P\left({\mathrm{\π}}_{pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)}^{T(pre\_m({\mathrm{\π}}_k^{T\left(i\right)}))}\right)\},\end{array}---\left(3\right)$

$\begin{array}{c}pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)\≠0,k=1,\mathrm{...},T_i,i=1,\mathrm{...},m\\ C_{\max }\left({\mathrm{\π}}^{*}\right)=\underset{\mathrm{\π}\∈\Π}{\min }{C}_{\max }\left(\mathrm{\π}\right)\end{array}---\left(4\right)$

${\mathrm{\π}}^{*}=\arg \left\{C_{max}\left(\mathrm{\π}\right)\right\}\→min,\∀\mathrm{\π}\∈\Π---\left(5\right)$

If product set is N, production equipment set is M, and wherein workpiece or product number are n, and number of devices is m.Often kind of product j (j ∈ (1 ..., n)) need stg _jroad manufacturing procedure, S _t=stg ₁∪ stg ₂∪ ... ∪ stg _nrepresent the set that the process number of all products is formed, the different operations of identical product need to process sequentially; All process steps can only be processed by the equipment meeting processing constraint in set M; The process time of product is relevant with process equipment, any equipment i (i ∈ (1 ..., m)) and synchronization can only process a kind of product; Be with sequence to be correlated with setup times between different product, setup times depends on processing sequence, and the setup times between identical product is 0.

Order for total process number of all products, π=[π ₁, π ₂..., π _tS] (π _j∈ (1 ..., n), j=1 ..., TS) and be the arrangement based on operation (product in this arrangement is assigned in individual device according to certain rule and processing constraint from left to right and processes) of a n to be processed product or workpiece, T _iit is the operation sum that i-th equipment is processed ${\mathrm{\π}}^{T\left(i\right)}=\[{\mathrm{\π}}_1^{T\left(i\right)},{\mathrm{\π}}_2^{T\left(i\right)},...,{\mathrm{\π}}_{T_i}^{T\left(i\right)}\]({\mathrm{\π}}_k^{T\left(i\right)}\∈(1,...,n),k=1,...,T_i)$ For the arrangement based on operation of institute's processing work on i-th equipment, for process time $(k=0,...,T_i,P({\mathrm{\π}}_0^{T\left(i\right)})=0),$ for with between setup times $(k=1,...,T_i,S({\mathrm{\π}}_0^{T\left(i\right)},{\mathrm{\π}}_1^{T\left(i\right)})=0,$ As k > 1 and ${\mathrm{\π}}_{k-1}^{T\left(i\right)}={\mathrm{\π}}_k^{T\left(i\right)}$ Time $S({\mathrm{\π}}_{k-1}^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)})=0),$ for beginning process time $(k=0,...,T_i,St({\mathrm{\π}}_0^{T\left(i\right)})=0),$ for a front machining device number (k=1 ..., T _i, when add man-hour first ), for front time processing at equipment in position from left to right (k=1 ..., T _i, when add man-hour first ), the target of optimization is find a π in the set Π of all model sequencing ^*, make completion date C the earliest _max(π) minimum;

Wherein, formula (1) to formula (3) is completion date C the earliest _max(π) computing formula, formula (4) and formula (5) expression find optimal sequencing π in the set Π of all model sequencing ^*, make C _max(π) minimum.The target of scheduling is find a π in the set Π of all fabric to be dyed sequences ^*, make Maximal Makespan C _max(π) minimum.

The concrete steps of the described Optimization Scheduling based on heredity-Estimation of Distribution Algorithm are as follows:

Step1, coded system: carry out coding π=[π with fabric to be dyed sequence ₁, π ₂..., π _tS]; Wherein TS represents total process number of all products;

The initialization of Step2, probability matrix and population:

Step2.1: initialization probability matrix: first need the initial probabilistic model or the matrix P that are configured to training _ori(gen)=[p _ij] _{n × TS}, initialized P _ori(gen) the allelic parameter probability valuing in all gene position is equal, i.e. p _ij=1/n (i=1 ..., n, j=1 ..., TS), initialized probability matrix P _ori(gen=1) form is such as formula shown in (6), and wherein gen represents the algebraically of Evolution of Population;

Step2.2: initialization population: GA-EDA adopts random fashion initialization population, and namely algorithm produces popsize individuality at random, and forming population scale is the initial population pop of popsize _candi(gen=1);

The update mechanism of Step3, probability matrix: GA-EDA uses history optimum individual update probability matrix, if π _local(gen)=π _{local_1}(gen) ..., π _{local_TS}(gen) for population is at the history optimum individual in gen generation, LR is learning rate, then probability matrix P _matrix(gen) following steps are adopted to upgrade:

Step3.1: establish x=π _{local_j}(gen), p _xj(gen)=p _xj(gen)+LR, wherein, j=1 .., TS;

Step3.2: probability normalization: $p_{wj}\left(gen\right)=p_{wj}\left(gen\right)/{\mathrm{\Σ}}_{y=1}^n{p}_{yj}\left(gen\right),$ Wherein, w=1 .., n, j=1 ..., TS.

Step4, generation new population:

Step4.1: produce candidate population: candidate population pop _candi(gen) be made up of two parts: (1) candidate population pop _candi(gen) in the individuality of e% according to the method for roulette to EDA probability matrix P _matrix(gen) sampling is formed; (2) individuality of all the other g% is by previous generation population pop _candi(gen-1) in, the individuality of the front g% that adaptation value is best is formed;

Step4.2: produce new population: by pop _candi(gen) order performs the genetic manipulation of GA, forms new population

Step5, end condition: the maximum iteration time of setting end condition is 200, if met, then export " optimum individual "; Otherwise go to Step3, iterate, until meet end condition.

Described population scale is set to popsize=50, learning rate LR=0.1, e%=0.6, g%=0.4.

Embodiment 2: as shown in Figure 1, a kind of Optimization Scheduling of weaving face fabric dyeing course, first by determining dyeing fabric process scheduling model and optimization aim, and the Optimization Scheduling based on heredity-Estimation of Distribution Algorithm GA-EDA is used to be optimized optimization aim; Wherein scheduling model according on every platform dyeing installation process fabric Maximal Makespan set up, optimization aim is C for minimizing Maximal Makespan simultaneously _max:

$\begin{array}{c}{C}_{\max }\left(\mathrm{\π}\right)=\max \{St\left({\mathrm{\π}}_{T_1}^{T\left(1\right)}\right)+P\left({\mathrm{\π}}_{T_1}^{T\left(1\right)}\right),St\left({\mathrm{\π}}_{T_2}^{T\left(2\right)}\right)+P\left({\mathrm{\π}}_{T_2}^{T\left(2\right)}\right),\mathrm{...},\\ St\left({\mathrm{\π}}_{T_m}^{T\left(m\right)}\right)+P\left({\mathrm{\π}}_{T_m}^{T\left(m\right)}\right)\}\end{array}---\left(1\right)$

$\begin{array}{c}St\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=St\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+P\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+S({\mathrm{\π}}_k^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)}),\\ pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=0,k=1,\mathrm{...},T_i,i=1,\mathrm{...},m\end{array}---\left(2\right)$

$\begin{array}{c}St\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=\max \{St\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+P\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+S({\mathrm{\π}}_{k-1}^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)}),\\ St\left({\mathrm{\π}}_{pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)}^{T(pre\_m({\mathrm{\π}}_k^{T\left(i\right)}))}\right)+P\left({\mathrm{\π}}_{pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)}^{T(pre\_m({\mathrm{\π}}_k^{T\left(i\right)}))}\right)\},\end{array}---\left(3\right)$

$\begin{array}{c}pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)\≠0,k=1,\mathrm{...},T_i,i=1,\mathrm{...},m\\ C_{\max }\left({\mathrm{\π}}^{*}\right)=\underset{\mathrm{\π}\∈\Π}{\min }{C}_{\max }\left(\mathrm{\π}\right)\end{array}---\left(4\right)$

${\mathrm{\π}}^{*}=\arg \left\{C_{max}\left(\mathrm{\π}\right)\right\}\→min,\∀\mathrm{\π}\∈\Π---\left(5\right)$

If product set is N, production equipment set is M, and wherein workpiece or product number are n, and number of devices is m.Often kind of product j (j ∈ (1 ..., n)) need stg _jroad manufacturing procedure, S _t=stg ₁∪ stg ₂∪ ... ∪ stg _nrepresent the set that the process number of all products is formed, the different operations of identical product need to process sequentially; All process steps can only be processed by the equipment meeting processing constraint in set M; The process time of product is relevant with process equipment, any equipment i (i ∈ (1 ..., m)) and synchronization can only process a kind of product; Be with sequence to be correlated with setup times between different product, setup times depends on processing sequence, and the setup times between identical product is 0.

Order for total process number of all products, π=[π ₁, π ₂..., π _tS] (π _j∈ (1 ..., n), j=1 ..., TS) and be the arrangement based on operation (product in this arrangement is assigned in individual device according to certain rule and processing constraint from left to right and processes) of a n to be processed product or workpiece, T _iit is the operation sum that i-th equipment is processed ${\mathrm{\π}}^{T\left(i\right)}=\[{\mathrm{\π}}_1^{T\left(i\right)},{\mathrm{\π}}_2^{T\left(i\right)},...,{\mathrm{\π}}_{T_i}^{T\left(i\right)}\]({\mathrm{\π}}_k^{T\left(i\right)}\∈(1,...,n),k=1,...,T_i)$ For the arrangement based on operation of institute's processing work on i-th equipment, for process time $(k=0,...,T_i,P({\mathrm{\π}}_0^{T\left(i\right)})=0),$ for with between setup times $(k=1,...,T_i,S\left({\mathrm{\π}}_0^{T\left(i\right)},{\mathrm{\π}}_1^{T\left(i\right)}\right)=0,$ As k > 1 and ${\mathrm{\π}}_{k-1}^{T\left(i\right)}={\mathrm{\π}}_k^{T\left(i\right)}$ Time $S({\mathrm{\π}}_{k-1}^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)})=0),$ for beginning process time $(k=0,...,T_i,St({\mathrm{\π}}_0^{T\left(i\right)})=0),$ for a front machining device number (k=1 ..., T _i, when add man-hour first ), for front time processing at equipment in position from left to right (k=1 ..., T _i, when add man-hour first ), the target of optimization is find a π in the set Π of all model sequencing ^*, make completion date C the earliest _max(π) minimum;

Wherein, formula (1) to formula (3) is completion date C the earliest _max(π) computing formula, formula (4) and formula (5) expression find optimal sequencing π in the set Π of all model sequencing ^*, make C _max(π) minimum.The target of scheduling is find a π in the set Π of all fabric to be dyed sequences ^*, make Maximal Makespan C _max(π) minimum.

The concrete steps of the described Optimization Scheduling based on heredity-Estimation of Distribution Algorithm are as follows:

Step1, coded system: carry out coding π=[π with fabric to be dyed sequence ₁, π ₂..., π _tS]; Wherein TS represents total process number of all products;

The initialization of Step2, probability matrix and population:

Step2.1: initialization probability matrix: first need the initial probabilistic model or the matrix P that are configured to training _ori(gen)=[p _ij] _{n × TS}, initialized P _ori(gen) the allelic parameter probability valuing in all gene position is equal, i.e. p _ij=1/n (i=1 ..., n, j=1 ..., TS), initialized probability matrix P _ori(gen=1) form is such as formula shown in (6), and wherein gen represents the algebraically of Evolution of Population;

Step2.2: initialization population: GA-EDA adopts random fashion initialization population, and namely algorithm produces popsize individuality at random, and forming population scale is the initial population pop of popsize _candi(gen=1);

The update mechanism of Step3, probability matrix: GA-EDA uses history optimum individual update probability matrix, if π _local(gen)=π _{local_1}(gen) ..., π _{local_TS}(gen) for population is at the history optimum individual in gen generation, LR is learning rate, then probability matrix P _matrix(gen) following steps are adopted to upgrade:

Step3.1: establish x=π _{local_j}(gen), p _xj(gen)=p _xj(gen)+LR, wherein, j=1 .., TS;

Step3.2: probability normalization: $p_{wj}\left(gen\right)=p_{wj}\left(gen\right)/{\mathrm{\Σ}}_{y=1}^n{p}_{yj}\left(gen\right),$ Wherein, w=1 .., n, j=1 ..., TS.

Step4, generation new population:

Step4.1: produce candidate population: candidate population pop _candi(gen) be made up of two parts: (1) candidate population pop _candi(gen) in the individuality of e% according to the method for roulette to EDA probability matrix P _matrix(gen) sampling is formed; (2) individuality of all the other g% is by previous generation population pop _candi(gen-1) in, the individuality of the front g% that adaptation value is best is formed;

Step4.2: produce new population: by pop _candi(gen) order performs the genetic manipulation of GA, forms new population

Step5, end condition: the maximum iteration time of setting end condition is 200, if met, then export " optimum individual "; Otherwise go to Step3, iterate, until meet end condition.

Described population scale is set to popsize=50, learning rate LR=0.1, e%=0.6, g%=0.4.

In order to verify this patent put forward validity and the robustness of GA-EDA algorithm, GA-EDA and standard EDA is contrasted.

First, adopt the test problem of the different scales of stochastic generation, the setup times between the process time of product and any two kinds of products, produces integer in [1,100] and [1,20] by being uniformly distributed respectively at random; Manufacturing procedure number required for various product produces integer in [1,5] at random by being uniformly distributed; The operation processing constraint stochastic generation of various product, all process steps at least can be processed on an equipment.N × m combination for testing comprises: 20 × 3,20 × 5,30 × 3,30 × 5,40 × 5,40 × 10,50 × 10,50 × 20,60 × 20,70 × 10,80 × 30 and 100 × 20.The optimum configurations of GA-EDA is as follows: population scale popsize=50, and training algebraically is 200, learning rate LR=0.1, e%=0.6, g%=0.4.All algorithms and test procedure realize by Delphi7.0 coding, and operating system is Windows7, and CPU frequency is 2.0GHz, inside saves as 2GB.Often kind of algorithm all independently reruns 20 times, and wherein AVG represents optimal value average, and SD represents standard deviation, and T (s) represents the averaging time that algorithm runs.

As known from Table 1, except in problem 60 × 20,70 × 10 and 80 × 30, the AVG of EDA is dominant (but the working time of EDA is obviously longer), the AVG of GA-EDA is better than the AVG of other scale EDA, this demonstrates the validity of GA-EDA.

The target function value of trying to achieve in the different problem scale situation of table 1

By reference to the accompanying drawings the specific embodiment of the present invention is explained in detail above, but the present invention is not limited to above-mentioned embodiment, in the ken that those of ordinary skill in the art possess, various change can also be made under the prerequisite not departing from present inventive concept.

Claims (3)

1. an Optimization Scheduling for weaving face fabric dyeing course, is characterized in that: first by determining dyeing fabric process scheduling model and optimization aim, and uses the Optimization Scheduling based on heredity-Estimation of Distribution Algorithm GA-EDA to be optimized optimization aim; Wherein scheduling model according on every platform dyeing installation process fabric Maximal Makespan set up, optimization aim is C for minimizing Maximal Makespan simultaneously _max:

$\begin{array}{c}{C}_{\max }\left(\mathrm{\π}\right)=\max \{St\left({\mathrm{\π}}_{T_1}^{T\left(1\right)}\right)+P\left({\mathrm{\π}}_{T_1}^{T\left(1\right)}\right),St\left({\mathrm{\π}}_{T_2}^{T\left(2\right)}\right)+P\left({\mathrm{\π}}_{T_2}^{T\left(2\right)}\right),\mathrm{...},\\ St\left({\mathrm{\π}}_{T_m}^{T\left(m\right)}\right)+P\left({\mathrm{\π}}_{T_m}^{T\left(m\right)}\right)\}\end{array}---\left(1\right)$

$\begin{array}{c}St\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=St\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+P\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+S({\mathrm{\π}}_k^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)}),\\ pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=0,k=1,\mathrm{...},T_i,i=1,\mathrm{...},m\end{array}---\left(2\right)$

$\begin{array}{c}St\left({\mathrm{\π}}_k^{T\left(i\right)}\right)=\max \{St\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+P\left({\mathrm{\π}}_{k-1}^{T\left(i\right)}\right)+S({\mathrm{\π}}_{k-1}^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)}),\\ St\left({\mathrm{\π}}_{pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)}^{T(pre\_m({\mathrm{\π}}_k^{T\left(i\right)}))}\right)+P\left({\mathrm{\π}}_{pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)}^{T(pre\_m({\mathrm{\π}}_k^{T\left(i\right)}))}\right)\},\\ pre\_k\left({\mathrm{\π}}_k^{T\left(i\right)}\right)\≠0,k=1,\mathrm{...},T_i,i=1,\mathrm{...},m\end{array}---\left(3\right)$

$C_{max}\left({\mathrm{\π}}^{*}\right)=\underset{\mathrm{\π}\∈\Π}{\min }{C}_{max}\left(\mathrm{\π}\right)---\left(4\right)$

${\mathrm{\π}}^{*}=\arg \left\{C_{max}\left(\mathrm{\π}\right)\right\}\→min,\∀\mathrm{\π}\∈\Π---\left(5\right)$

If product set is N, production equipment set is M, and wherein workpiece or product number are n, and number of devices is m.Often kind of product j (j ∈ (1 ..., n)) need stg _jroad manufacturing procedure, S _t=stg ₁∪ stg ₂∪ ... ∪ stg _nrepresent the set that the process number of all products is formed, the different operations of identical product need to process sequentially; All process steps can only be processed by the equipment meeting processing constraint in set M; The process time of product is relevant with process equipment, any equipment i (i ∈ (1 ..., m)) and synchronization can only process a kind of product; Be with sequence to be correlated with setup times between different product, setup times depends on processing sequence, and the setup times between identical product is 0.

Order for total process number of all products, π=[π ₁, π ₂..., π _tS] (π _j∈ (1 ..., n), j=1 ..., TS) and be the arrangement based on operation (product in this arrangement is assigned in individual device according to certain rule and processing constraint from left to right and processes) of a n to be processed product or workpiece, T _iit is the operation sum that i-th equipment is processed ${\mathrm{\π}}^{T\left(i\right)}=\[{\mathrm{\π}}_1^{T\left(i\right)},{\mathrm{\π}}_2^{T\left(i\right)},...,{\mathrm{\π}}_{T_i}^{T\left(i\right)}\]({\mathrm{\π}}_k^{T\left(i\right)}\∈(1,...,n),k=1,...,T_i)$ For the arrangement based on operation of institute's processing work on i-th equipment, for process time $(k=0,...,T_i,P({\mathrm{\π}}_0^{T\left(i\right)})=0),$ for with between setup times $(k=1,...,T_i,S({\mathrm{\π}}_0^{T\left(i\right)},{\mathrm{\π}}_1^{T\left(i\right)})=0,$ As k > 1 and ${\mathrm{\π}}_{k-1}^{T\left(i\right)}={\mathrm{\π}}_k^{T\left(i\right)}$ Time $S({\mathrm{\π}}_{k-1}^{T\left(i\right)},{\mathrm{\π}}_k^{T\left(i\right)})=0),$ for beginning process time $(k=0,...,T_i,St({\mathrm{\π}}_0^{T\left(i\right)})=0),$ for a front machining device number (k=1 ..., T _i, when add man-hour first for front time processing at equipment in position from left to right (k=1 ..., T _i, when add man-hour first the target optimized is find a π in the set Π of all model sequencing ^*, make completion date C the earliest _max(π) minimum;

Wherein, formula (1) to formula (3) is completion date C the earliest _max(π) computing formula, formula (4) and formula (5) expression find optimal sequencing π in the set Π of all model sequencing ^*, make C _max(π) minimum.The target of scheduling is find a π in the set Π of all fabric to be dyed sequences ^*, make Maximal Makespan C _max(π) minimum.

2. the Optimization Scheduling of weaving face fabric dyeing course according to claim 1, is characterized in that: the concrete steps of the described Optimization Scheduling based on heredity-Estimation of Distribution Algorithm are as follows:

Step1, coded system: carry out coding π=[π with fabric to be dyed sequence ₁, π ₂..., π _tS]; Wherein TS represents total process number of all products;

The initialization of Step2, probability matrix and population:

Step2.1: initialization probability matrix: first need the initial probabilistic model or the matrix P that are configured to training _ori(gen)=[p _ij] _{n × TS}, initialized P _ori(gen) the allelic parameter probability valuing in all gene position is equal, i.e. p _ij=1/n (i=1 ..., n, j=1 ..., TS), initialized probability matrix P _ori(gen=1) form is such as formula shown in (6), and wherein gen represents the algebraically of Evolution of Population;

Step2.2: initialization population: GA-EDA adopts random fashion initialization population, and namely algorithm produces popsize individuality at random, and forming population scale is the initial population pop of popsize _candi(gen=1);

The update mechanism of Step3, probability matrix: GA-EDA uses history optimum individual update probability matrix, if π _local(gen)=π _{local_1}(gen) ..., π _{local_TS}(gen) for population is at the history optimum individual in gen generation, LR is learning rate, then probability matrix P _matrix(gen) following steps are adopted to upgrade:

Step3.1: establish x=π _{local_j}(gen), p _xj(gen)=p _xj(gen)+LR, wherein, j=1 .., TS;

Step3.2: probability normalization: wherein, w=1 .., n, j=1 ..., TS.

Step4, generation new population:

Step4.1: produce candidate population: candidate population pop _candi(gen) be made up of two parts: (1) candidate population pop _candi(gen) in the individuality of e% according to the method for roulette to EDA probability matrix P _matrix(gen) sampling is formed; (2) individuality of all the other g% is by previous generation population pop _candi(gen-1) in, the individuality of the front g% that adaptation value is best is formed;

Step4.2: produce new population: by pop _candi(gen) order performs the genetic manipulation of GA, forms new population

Step5, end condition: the maximum iteration time of setting end condition is 200, if met, then export " optimum individual "; Otherwise go to Step3, iterate, until meet end condition.

3. the Optimization Scheduling of weaving face fabric dyeing course according to claim 2, is characterized in that: described population scale is set to popsize=50, learning rate LR=0.1, e%=0.6, g%=0.4.