CN105528675B - A kind of production distribution scheduling method based on ant group algorithm - Google Patents

Abstract

The invention discloses a kind of production distribution scheduling method based on ant group algorithm, it is modeled for the discrepant dispensing combined dispatching problem that produces of volume and production time, then solved by a kind of ant group algorithm that improves, thus obtain the prioritization scheme of a set of production distribution scheduling, it is substantially reduced total operating cost of target position manufacturing enterprise with this, improves the efficiency of operation of enterprises.The present invention is simultaneous for particular problem, gives the program development scheme of practicality, makes enterprise's high efficiency acquisition of energy be suitable for its optimal scheduling scheme.

Description

A kind of production distribution scheduling method based on ant group algorithm

Technical field

The invention belongs to supply chain field, a kind of production distribution scheduling method based on ant group optimization.

Background technology

Under Vehicles Collected from Market environment, the core competitiveness of manufacturing enterprise is no longer simple manufacturing capacity, but supply Chain operating capability, manufacturing enterprise needs buying, produces and link of providing and delivering carries out overall control, by production system and logistics system Carry out combined optimization, the maximization of overall economic benefit could be obtained, the competitiveness of enterprise.

Combined dispatching is the class optimization method towards supply chain, uses the normal form of accurately scheduling, in design supply chain The combined dispatching scheme of each link, it is achieved the optimization of enterprise-wide economic benefit, thus the service level of enterprise.

Research to associating scheduling problem at present all concentrates under traditional mode of production pattern, in this kind of production model, and one Equipment can process an operation or the operation of a collection of fixed qty simultaneously, but in reality industry, difference batch production mould Formula has the character of both production models concurrently, and more increasingly complex than this two class, applies more extensive.And traditional dispatching method is big Many consider how to reduce production cost simply, do not consider how that the totle drilling cost making production dispensing combined dispatching is minimum, Therefore Production requirement instantly can not be adapted to.

Summary of the invention

The present invention is the weak point in order to overcome prior art to exist, it is provided that a kind of production based on ant group optimization is provided and delivered Dispatching method, to the optimization of enterprise-wide economic benefit can be realized, it is thus possible to reduce entreprise cost, the service effect of enterprise Rate.

The present invention solves that technical problem adopts the following technical scheme that

A kind of feature producing distribution scheduling method based on ant group optimization of the present invention is to carry out as follows:

Step 1, assuming to exist n batch needs and produce and provide and deliver, the volume of equipment producing each batch is designated as B；Dispensing The vehicle volume of each batch is designated as V；Constituted one batch of set by described n batch, be designated as U={b₁,b₂,…,b_k,…,b_n, b_kTable Show that kth is criticized；And kth is criticized b_kSize be designated as S_k；Kth is criticized b_kProduction time be designated as T_k；Same car will be added In carry out providing and delivering all batches be designated as a group；1≤k≤n；

Step 2, equivalently-sized the criticizing gathered described batch in U are divided into a class, thus obtain a classification；It is designated as W= {w₁,w₂,…,w_z,…,w_a, w_zRepresent the z classification；Described the z classification w_zIn batch total be designated as f_z；

Step 3, the parameters of initialization ant group algorithm, including: m represents the m Formica fusca, and initializes m=1；M represents Formica fusca sum, L represent iterations, and initialize L=1；L_maxRepresent maximum iteration time；

Step 4, defined variable are l, and initialize l=1；Definition kth criticizes b_kIdentifier be flag_k, and initialize flag_k=0；

Step 5, create the l group of m Formica fusca of the L time iterationWith the l corresponding candidate listAnd make the m Formica fusca of the L time iteration can be assigned to described n batch in different groups provide and deliver；And L The m Formica fusca of secondary iteration completes the vehicle fleet that packet is used to all batches

Step 6, m+1 is assigned to m, and returns step 5 and perform, until m=M, thus obtain the institute of the L time iteration There is M Formica fusca that all batches complete the vehicle fleet set that packet is used

Step 7, from described vehicle fleet setIn choose minima As the locally optimal solution of the L time iteration, it is designated as π_L；

Step 8, the i-th candidate utilizing formula (1) to update the L time iteration criticize b_i' join the l group with jthIn candidate criticize b_jPheromone between 'Thus obtain the pheromone of the L+1 time iteration

In formula (1), ρ represents the evaporation rate of pheromone；m_ij(L) represent that in the L time iteration, i-th candidate criticizes b_i' and jth Individual join the l groupIn candidate criticize b_j' it is assigned to the number of times of same group；And have:

Step 9, formula (3) and formula (4) is utilized to judgeWhether meet pheromone concentration and limit interval [τ_min,τ_max], If meeting, then retain the pheromone of the L+1 time iterationAnd perform step 11；Otherwise, step 10 is performed:

${\mathrm{\τ}}_{max}\≤\frac{1}{(1-\mathrm{\ρ})f\left({\mathrm{\π}}^{*}\right)}---\left(3\right)$

${\mathrm{\τ}}_{\min }=\frac{{\mathrm{\τ}}_{max}(1-\sqrt[a]{0.05})}{(a/2-1)\sqrt[a]{0.05}}---\left(4\right)$

In formula (3) and formula (4), π^*Represent the minima in current acquired all locally optimal solutions；

If step 10Then willIt is assigned toIfThen willCompose Value is given

Step 11；L+1 is assigned to L, it is judged that L ＜ L_maxWhether set up, if setting up, returning step 4 and performing, otherwise completing L_maxSecondary iteration, and obtain globally optimal solution π_best, it is L_maxThe minima of all locally optimal solutions in secondary iteration；With the overall situation Optimal solution π_bestCorresponding distribution project is as optimum distribution project；

Step 12, in optimum distribution project each group batch will carry out descending sort according to the production time, it is thus achieved that row Sequence result is as series-produced order, thus obtains Optimal Production and dispensing combined dispatching scheme.

The feature of the production scheduling method of porcelain calcine technology based on ant group optimization of the present invention lies also in, described In step 5, the m Formica fusca of the L time iteration is to be assigned to described n batch in different group as follows provide and deliver:

Step 5.1, defined variable f；

Step 5.2, initialization z=1；

Step 5.3, initialization f=1；

Step 5.4, judge described the z classification w_zIn select the identifier flag of f crowd_fWhether it is 1；If 1, then Represent that f batch is complete packet, and perform step 5.5；Otherwise, by the z classification w_zIn select f batch and join institute State the l candidate listIn, and perform step 5.6；

Step 5.5, f+1 is assigned to f, and returns step 5.4 and perform, until f=f_zAfter till, perform step 5.6；

Step 5.6, z+1 is assigned to z, and returns step 5.3 and perform；Until after z=a, thus obtain to be updated The l candidate listRemember described the l candidate list to be updatedIn candidate criticize as { b₁′,b₂′,…, b′_i,…,b_a′}；1≤i≤a；

Step 5.7, from described candidate list to be updatedCrowd b that middle selection size is maximum_key' criticize as key and to add Enter to described the l groupAnd key is criticized b_key' identifier flag '_keyIt is set to 1；Then delete from candidate list and close Key criticizes b_key′；

Step 5.8, formula (5) is utilized to obtain the m Formica fusca by described the l candidate list to be updatedMiddle i-th Candidate criticizes b_i' join the l groupCandidate probabilityThus obtain the m Formica fusca and all candidates are criticized join The l groupCandidate probability set

$P_{i1}^m=\left\{\begin{array}{cc}\frac{{\mathrm{\θ}}_{i1}^{\mathrm{\α}}{\mathrm{\η}}_{i1}^{\mathrm{\β}}}{{\mathrm{\Σ}}_{b_{i^{\′}}\∈X_1^{\′\left(L\right)\left(m\right)}}{\mathrm{\θ}}_{i1}^{\mathrm{\α}}{\mathrm{\η}}_{i1}^{\mathrm{\β}}}& {b_i}^{\′}\∈X_1^{\′\left(L\right)\left(m\right)}\\ 0& else\end{array}\right.---\left(5\right)$

In formula (5), α is the weight of pheromone, and β is the weight of heuristic information；θ_ilRepresent that i-th candidate criticizes b_i' can add Enter to the l groupThe expecting degree of middle dispensing；η_ilRepresent and i-th candidate is criticized b_i' the l group can be joinedJoin The heuristic information sent；And have:

${\mathrm{\θ}}_{i1}=\frac{1}{\left|G_1^{\left(L\right)\left(m\right)}\right|}{\mathrm{\Σ}}_{j\∈G_1^{\left(L\right)\left(m\right)}}{\mathrm{\τ}}_{ij}\left(L\right)---\left(6\right)$

${\mathrm{\η}}_{i1}=\frac{S_i}{{\mathrm{\Σ}}_{i=1}^n{S}_i}---\left(7\right)$

In formula (6), τ_ijRepresent that the i-th candidate of the L time iteration criticizes b_i' join the l group with jthIn Candidate criticize b_jPheromone between ', 1≤i ≠ j≤a；S_iRepresent that i-th candidate criticizes b '_iSize；

Step 5.9, from described candidate probability setIn select Maximum alternative probability Corresponding Maximum alternative is criticized, and is designated as b '_max；The most described Maximum alternative criticizes b '_maxSize be designated as s_max；

Step 5.10, described Maximum alternative is criticized b '_maxJoin described the l groupAnd Maximum alternative is criticized b′_maxIdentifier flag '_maxIt is set to 1；

Step 5.11, described vehicle volume V is deducted described Maximum alternative criticize b '_maxSize s_max, it is thus achieved that residue vehicle Volume, is designated as

Step 5.12, according to described residue vehicle volumeFrom described the l candidate list to be updatedIn Delete size more than described residue vehicle volumeCandidate criticize；Thus obtain the l candidate list of renewal

Step 5.13, obtain described Maximum alternative and criticize b '_maxCorresponding classification, is designated as w_max；

Step 5.14, judge that described Maximum alternative criticizes b '_maxAt corresponding classification w_maxIn whether be f_zIndividual batch；If It is, then from the l the candidate list updatedThe described Maximum alternative of middle deletion criticizes b '_max；Otherwise, from the l the time updated Select tableThe described Maximum alternative of middle deletion criticizes b '_max, and described Maximum alternative is criticized b '_maxCorresponding classification w_maxIn f_max+ 1 batch of the l the candidate list joining described renewalIn, thus obtain the l the candidate list again updated

Step 5.15, with described the l the candidate list again updatedAs the l candidate list to be updatedAnd return step 5.8 order execution, until the l candidate list to be updatedTill sky, will the L time The l group of m Formica fusca of iterationFill it up with；Thus obtain the l group of m Formica fusca of the L time iterationDistribution project；

Step 5.16, the identifier judging n batch are the most all 1, if being all 1, then it represents that complete the m of the L time iteration The distribution to described n batch of Formica fusca；And l is assigned toOtherwise, l+1 is assigned to l, and returns step 5 order and hold OK.

Compared with the prior art, the present invention has the beneficial effect that:

1, the present invention is under typical difference is produced in batches pattern, the production of research manufacturing enterprise, dispensing two benches associating Scheduling problem, the ant group algorithm improved by employing, under difference is produced in batches pattern, carry out first against product the most in batches Classification, then proposes candidate list based on batch size classes, by the method setting up candidate list for each Formica fusca, draws each The packet scheme of individual Formica fusca；Recycling pheromone updating rule, updates pheromone, it is achieved that successive ignition, finally obtains optimum Solve；Compared to other method；Ant group algorithm, on the degree of optimization of available time and result, is that a kind of performance is the most excellent Change the approximate data manufacturing span；Solve the problem that in reality industry, job batching produces, packet is provided and delivered so that obtained Optimal solution achieves the optimization of enterprise-wide economic benefit, reduces energy consumption, provides cost savings, and improves the service water of enterprise Flat.

2, the present invention carries out Pheromone update by the optimal solution obtaining iteration each time, solves search procedure too fast Focus on this and solve problem around；If the optimal solution obtained since first time iteration is updated, very Algorithmic statement may be caused too early；And the optimal solution of iteration has the biggest difference after iteration each time each time, so more The optimal solution of new iteration each time, it is possible to make the pheromone of more solution strengthen, thus obtain more optimal solution.

3. the present invention uses candidate list strategy to find optimal solution；In traditional ant group algorithm, as long as volume is less than The operation of vehicle volume, can select into candidate list, so the substantial amounts of feasible solution；But during actual entrucking, very Many batches of sizes are the same, if identical for volume size criticizing is classified as a class, the categorical measure criticized can never exceed The total quantity criticized, the porcelain base of same size is criticized the most, and batch number of candidate list is the fewest, which reduces the quantity of feasible solution, Effectively reduce the time of solving.

Accompanying drawing explanation

Fig. 1 is that the present invention produces dispensing span schematic diagram.

Detailed description of the invention

In the present embodiment, a kind of production distribution scheduling method based on ant group optimization, is to have for volume and production time The production dispensing combined dispatching problem of difference is modeled, and is then solved by a kind of ant group algorithm that improves, thus obtains The prioritization scheme of a set of production distribution scheduling, is substantially reduced total operating cost of target position manufacturing enterprise with this, improves enterprise's fortune Make efficiency；It is simultaneous for particular problem, gives the program development scheme of practicality, make enterprise high efficiency acquisition can be suitable for it Optimal scheduling scheme.Specifically, it is to carry out as follows:

Step 1, assuming to exist n batch needs and produce and provide and deliver, the volume of equipment producing each batch is designated as B；Dispensing The vehicle volume of each batch is designated as V；Constituted one batch of set by n batch, be designated as U={b₁,b₂,…,b_k,…,b_n, b_kRepresent the K batch；And kth is criticized b_kSize be designated as S_k；Kth is criticized b_kProduction time be designated as T_k；Same for addition car will be entered All batches of row dispensing are designated as a group；1≤k≤n；

Step 2, equivalently-sized batch in crowd set U is divided into a class, thus obtains a classification；It is designated as W={w₁, w₂,…,w_z,…,w_a, w_zRepresent the z classification；The z classification w_zIn batch total be designated as f_z；In traditional ant group algorithm, As long as volume is less than the operation of place capacity, can select into candidate list, so the substantial amounts of feasible solution.But in reality During ceramic roasting, the volume size of a lot of potteries is the same, if porcelain base identical for volume size is classified as a class, porcelain The categorical measure of base can never exceed the total quantity of porcelain base, and the porcelain base of same size is the most, and the porcelain base number of candidate list is the fewest, Thus can reduce the quantity of feasible solution, it is possible to effectively reduce solving the time.

Step 3, the parameters of initialization ant group algorithm, including: m represents the m Formica fusca, and initializes m=1；M represents Formica fusca sum, L represent iterations, and initialize L=1；L_maxRepresent maximum iteration time；

Step 4, defined variable are l, and initialize l=1；Definition kth criticizes b_kIdentifier be flag_k, and initialize flag_k=0；

Step 5, create the l group of m Formica fusca of the L time iterationWith the l corresponding candidate listAnd make the m Formica fusca of the L time iteration can be assigned to n batch in different groups provide and deliver；And change for the L time The m Formica fusca in generation completes the vehicle fleet that packet is used to all batches

Step 5.1, defined variable f；

Step 5.2, initialization z=1；

Step 5.3, initialization f=1；

Step 5.4, judge the z classification w_zIn select the identifier flag of f crowd_fWhether it is 1；If 1, then it represents that F batch is complete packet, and performs step 5.5；Otherwise, by the z classification w_zIn select f batch and join l Candidate listIn, and perform step 5.6；

Step 5.5, f+1 is assigned to f, and returns step 5.4 and perform, until f=f_zAfter till, perform step 5.6；

Step 5.6, z+1 is assigned to z, and returns step 5.3 and perform；Until after z=a, thus obtain to be updated The l candidate listRemember the l candidate list to be updatedIn candidate criticize as { b₁′,b₂′,…,b ′_i,…,b_a′}；1≤i≤a；

Step 5.7, from candidate list to be updatedCrowd b that middle selection size is maximum_key' criticize as key and to join The l groupAnd key is criticized b_key' identifier flag '_keyIt is set to 1；Then from candidate list, delete key batch b_key′；

Step 5.8, formula (1) is utilized to obtain the m Formica fusca by the l candidate list to be updatedMiddle i-th candidate Criticize b_i' join the l groupCandidate probabilityThus obtain the m Formica fusca and all candidates are criticized join l Individual groupCandidate probability set

$P_{i1}^m=\left\{\begin{array}{cc}\frac{{\mathrm{\θ}}_{i1}^{\mathrm{\α}}{\mathrm{\η}}_{i1}^{\mathrm{\β}}}{{\mathrm{\Σ}}_{b_{i^{\′}}\∈X_1^{\′\left(L\right)\left(m\right)}}{\mathrm{\θ}}_{i1}^{\mathrm{\α}}{\mathrm{\η}}_{i1}^{\mathrm{\β}}}& {b_i}^{\′}\∈X_1^{\′\left(L\right)\left(m\right)}\\ 0& else\end{array}\right.---\left(1\right)$

In formula (1), α is the weight of pheromone, and β is the weight of heuristic information；θ_ilRepresent that i-th candidate criticizes b_i' can add Enter to the l groupThe expecting degree of middle dispensing, pheromone τ_ij(L) represent that i-th candidate criticizes b_i' criticize b with j-th candidates_j′ It is arranged expecting degree in the same set.Owing to lot number mesh and operation place group number are uncertain, therefore can not directly use information Element τ_ij(L), thus defined variable θ_ilPheromone is carried out indirect utilization；η_ilRepresent and i-th candidate is criticized b_i' l can be joined GroupThe heuristic information of dispensing；And have:

${\mathrm{\θ}}_{i1}=\frac{1}{\left|G_1^{\left(L\right)\left(m\right)}\right|}{\mathrm{\Σ}}_{j\∈G_1^{\left(L\right)\left(m\right)}}{\mathrm{\τ}}_{ij}\left(L\right)---\left(2\right)$

${\mathrm{\η}}_{i1}=\frac{S_i}{{\mathrm{\Σ}}_{i=1}^n{S}_i}---\left(3\right)$

In formula (2), τ_ijRepresent that the i-th candidate of the L time iteration criticizes b_i' join the l group with jthIn Candidate criticize b_jPheromone between ', 1≤i ≠ j≤a；S_iRepresent that i-th candidate criticizes b '_iSize；

Step 5.9, from candidate probability setIn select Maximum alternative probabilityInstitute is right The Maximum alternative answered is criticized, and is designated as b '_max；Then Maximum alternative criticizes b '_maxSize be designated as s_max；

Step 5.10, Maximum alternative is criticized b '_maxJoin the l groupAnd Maximum alternative is criticized b '_maxMark Symbol flag '_maxIt is set to 1；

Step 5.11, vehicle volume V is deducted Maximum alternative criticize b '_maxSize s_max, it is thus achieved that residue vehicle volume, it is designated as

Step 5.12, according to residue vehicle volumeFrom the l candidate list to be updatedMiddle deletion size More than residue vehicle volumeCandidate criticize；Thus obtain the l candidate list of renewal

Step 5.13, acquisition Maximum alternative criticize b '_maxCorresponding classification, is designated as w_max；

Step 5.14, judge that Maximum alternative criticizes b '_maxAt corresponding classification w_maxIn whether be f_zIndividual batch；The most then From the l the candidate list updatedMiddle deletion Maximum alternative criticizes b '_max；Otherwise, from the l the candidate list updatedMiddle deletion Maximum alternative criticizes b '_max, and Maximum alternative is criticized b '_maxCorresponding classification w_maxIn f_max+ 1 batch adds Enter to the l the candidate list updatedIn, thus obtain the l the candidate list again updated

Step 5.15, with the l the candidate list again updatedAs the l candidate list to be updatedAnd return step 5.8 order execution, until the l candidate list to be updatedTill sky, will the L time The l group of m Formica fusca of iterationFill it up with；Thus obtain the l group of m Formica fusca of the L time iterationDistribution project；

Step 5.16, the identifier judging n batch are the most all 1, if being all 1, then it represents that complete the m of the L time iteration The distribution to n batch of Formica fusca；And l is assigned toOtherwise, l+1 is assigned to l, and returns step 5 order execution.

At the l the candidate list updatedIn, if i-th candidate criticizes b_i' criticize b with j-th candidates_j' meet S_i ＜ S_jAnd S_iT_i＞ S_jT_j, select i-th candidate to criticize b_i' add current group than selecting j-th candidates to criticize b_j' more sky can be reduced Free space, and select i-th candidate to criticize b_iThe residual capacity of ' rear current group is bigger, the l candidate listIn optional Lot number is the most more, therefore selects i-th candidate to criticize b_iThe current group of ' addition is more excellent.

Step 6, m+1 is assigned to m, and returns step 5 and perform, until m=M, thus obtain the institute of the L time iteration There is M Formica fusca that all batches complete the vehicle fleet set that packet is used

Step 7, from vehicle fleet setIn choose minima conduct The locally optimal solution of the L time iteration, is designated as π_L；

Step 8, the i-th candidate utilizing formula (4) to update the L time iteration criticize b_i' join the l group with jthIn candidate criticize b_jPheromone between 'Thus obtain the pheromone of the L+1 time iteration

In formula (4), ρ represents the evaporation rate of pheromone；m_ij(L) represent that in the L time iteration, i-th candidate criticizes b_i' and jth Individual join the l groupIn candidate criticize b_j' it is assigned to the number of times of same group；And have:

If the optimal solution obtained since first time iteration is updated, it is likely that cause search procedure mistake Fast focuses on this solution around, makes algorithmic statement too early；And the optimal solution of iteration has after iteration each time each time The biggest difference, so updating the optimal solution of iteration each time, it is possible to makes the pheromone of more solution strengthen.Rotation plan can also be used Slightly being updated, the optimal solution i.e. obtained iteration each time carries out Pheromone update, the most often through corresponding iteration, to certainly The optimal solution obtained since Yi Ci is through the renewal of row primary information element.

Step 9, formula (6) and formula (7) is utilized to judgeWhether meet pheromone concentration and limit interval [τ_min,τ_max], If meeting, then retain the pheromone of the L+1 time iterationAnd perform step 11；Otherwise, step 10 is performed:

${\mathrm{\τ}}_{max}\≤\frac{1}{(1-\mathrm{\ρ})f\left({\mathrm{\π}}^{*}\right)}---\left(6\right)$

${\mathrm{\τ}}_{\min }=\frac{{\mathrm{\τ}}_{max}(1-\sqrt[a]{0.05})}{(a/2-1)\sqrt[a]{0.05}}---\left(7\right)$

In formula (6) and formula (7), π^*Represent the minima in current acquired all locally optimal solutions；

If step 10Then willIt is assigned toIfThen willAssignment GiveThe concentration of pheromone is limited to by ant group algorithmBetween, to reduce the difference between the pheromone of feasible solution Different.

Step 11；L+1 is assigned to L, it is judged that L ＜ L_maxWhether set up, if setting up, returning step 4 and performing, otherwise completing L_maxSecondary iteration, and obtain globally optimal solution π_best, it is L_maxThe minima of all locally optimal solutions in secondary iteration；With the overall situation Optimal solution π_bestCorresponding distribution project is as optimum distribution project, in batches with delivery process as shown in Figure 1；

Step 12, in optimum distribution project each group batch will carry out descending sort according to the production time, it is thus achieved that row Sequence result is as series-produced order, thus obtains Optimal Production and dispensing combined dispatching scheme.

Claims (2)

1. a production distribution scheduling method based on ant group optimization, is characterized in that carrying out as follows:

Step 1, assuming to exist n batch needs and produce and provide and deliver, the volume of equipment producing each batch is designated as B；Provide and deliver each The vehicle volume criticized is designated as V；Constituted one batch of set by described n batch, be designated as U={b₁,b₂,…,b_k,…,b_n, b_kRepresent the K batch；And kth is criticized b_kSize be designated as S_k；Kth is criticized b_kProduction time be designated as T_k；Same for addition car will be entered All batches of row dispensing are designated as a group；1≤k≤n；

Step 2, equivalently-sized the criticizing gathered described batch in U are divided into a class, thus obtain a classification；It is designated as W={w₁, w₂,…,w_z,…,w_a, w_zRepresent the z classification；Described the z classification w_zIn batch total be designated as f_z；

Step 3, the parameters of initialization ant group algorithm, including: m represents the m Formica fusca, and initializes m=1；M represents Formica fusca Sum, L represent iterations, and initialize L=1；L_maxRepresent maximum iteration time；

Step 4, defined variable are l, and initialize l=1；Definition kth criticizes b_kIdentifier be flag_k, and initialize flag_k =0；

Step 5, create the l group of m Formica fusca of the L time iterationWith the l corresponding candidate listAnd make the m Formica fusca of the L time iteration can be assigned to described n batch in different groups provide and deliver；And L The m Formica fusca of secondary iteration completes the vehicle fleet that packet is used to all batches

Step 6, m+1 is assigned to m, and returns step 5 and perform, until m=M, thus obtain all M of the L time iteration Formica fusca completes the vehicle fleet set that packet is used to all batches

Step 7, from described vehicle fleet setIn choose minima as The locally optimal solution of L iteration, is designated as π_L；

Step 8, the i-th candidate utilizing formula (1) to update the L time iteration criticize b_i' join the l group with jthIn Candidate criticize b_jPheromone τ between '_ij(L), thus obtain the pheromone τ of the L+1 time iteration_ij(L+1):

${\mathrm{\τ}}_{ij}(L+1)=(1-\mathrm{\ρ}){\mathrm{\τ}}_{ij}\left(L\right)+m_{ij}\left(L\right)\▿{\mathrm{\τ}}_{ij}---\left(1\right)$

In formula (1), ρ represents the evaporation rate of pheromone；m_ij(L) represent that in the L time iteration, i-th candidate criticizes b_i' and jth is Join the l groupIn candidate criticize b_j' it is assigned to the number of times of same group；And have:

Step 9, formula (3) and formula (4) is utilized to judge τ_ij(L+1) whether meet pheromone concentration and limit interval [τ_min,τ_max], if full Foot, then retain the pheromone τ of the L+1 time iteration_ij, and perform step 11 (L+1)；Otherwise, step 10 is performed:

${\mathrm{\τ}}_{max}\≤\frac{1}{(1-\mathrm{\ρ})f\left({\mathrm{\π}}^{*}\right)}---\left(3\right)$

${\mathrm{\τ}}_{\min }=\frac{{\mathrm{\τ}}_{max}(1-\sqrt[a]{0.05})}{(a/2-1)\sqrt[a]{0.05}}---\left(4\right)$

In formula (3) and formula (4), π^*Represent the minima in current acquired all locally optimal solutions；

If step 10 τ_ij(L+1)≥τ_max, then by τ_maxIt is assigned to τ_ij(L+1)；If τ_ij(L+1)≤τ_min, then by τ_minIt is assigned to τ_ij(L+1)；

Step 11；L+1 is assigned to L, it is judged that L ＜ L_maxWhether set up, if setting up, returning step 4 and performing, otherwise completing L_maxSecondary Iteration, and obtain globally optimal solution π_best, it is L_maxThe minima of all locally optimal solutions in secondary iteration；With global optimum Solve π_bestCorresponding distribution project is as optimum distribution project；

Step 12, in optimum distribution project each group batch will carry out descending sort according to the production time, it is thus achieved that sequence knot Fruit is as series-produced order, thus obtains Optimal Production and dispensing combined dispatching scheme.

Production distribution scheduling method based on ant group optimization the most according to claim 1, is characterized in that, in described step 5, The m Formica fusca of the L time iteration is to be assigned to described n batch in different group as follows provide and deliver:

Step 5.1, defined variable f；

Step 5.2, initialization z=1；

Step 5.3, initialization f=1；

Step 5.4, judge described the z classification w_zIn select the identifier flag of f crowd_fWhether it is 1；If 1, then it represents that F batch is complete packet, and performs step 5.5；Otherwise, by the z classification w_zIn select f batch and join described the L candidate listIn, and perform step 5.6；

Step 5.5, f+1 is assigned to f, and returns step 5.4 and perform, until f=f_zAfter till, perform step 5.6；

Step 5.6, z+1 is assigned to z, and returns step 5.3 and perform；Until after z=a, thus obtain l to be updated Individual candidate listRemember described the l candidate list to be updatedIn candidate criticize as { b₁,b₂,…,b_i,…, b_a}；1≤i≤a；

Step 5.7, from described candidate list to be updatedCrowd b that middle selection size is maximum_key' criticize as key and to join Described the l groupAnd key is criticized b_key' identifier flag '_keyIt is set to 1；Then from candidate list, delete key batch b_key′；

Step 5.8, formula (5) is utilized to obtain the m Formica fusca by described the l candidate list to be updatedMiddle i-th candidate Criticize b_i' join the l groupCandidate probabilityThus obtain the m Formica fusca and all candidates are criticized join l GroupCandidate probability set

$P_{il}^m=\left\{\begin{array}{cc}\frac{{\mathrm{\θ}}_{il}^{\mathrm{\α}}{\mathrm{\η}}_{il}^{\mathrm{\β}}}{{\mathrm{\Σ}}_{b_{i^{\′}}\∈X_l^{\′\left(L\right)\left(m\right)}}{\mathrm{\θ}}_{il}^{\mathrm{\α}}{\mathrm{\η}}_{il}^{\mathrm{\β}}}& b_i\∈{X_l}^{\left(L\right)\left(m\right)}\\ 0& else\end{array}\right.---\left(5\right)$

In formula (5), α is the weight of pheromone, and β is the weight of heuristic information；θ_ilRepresent that i-th candidate criticizes b_i' can join The l groupThe expecting degree of middle dispensing；η_ilRepresent and i-th candidate is criticized b_i' the l group can be joinedDispensing Heuristic information；And have:

${\mathrm{\θ}}_{il}=\frac{1}{\left|G_l^{\left(L\right)\left(m\right)}\right|}{\mathrm{\Σ}}_{j\∈G_l^{\left(L\right)\left(m\right)}}{\mathrm{\τ}}_{ij}\left(L\right)---\left(6\right)$

${\mathrm{\η}}_{il}=\frac{S_i}{{\mathrm{\Σ}}_{i=1}^n{S}_i}---\left(7\right)$

In formula (6), τ_ijRepresent that the i-th candidate of the L time iteration criticizes b_i' join the l group with jthIn time Choosing batch b_jPheromone between ', 1 ＜ i ≠ j≤a；S_iRepresent that i-th candidate criticizes b_i' size；

Step 5.9, from described candidate probability setIn select Maximum alternative probabilityCorresponding Maximum alternative criticize, be designated as b '_max；The most described Maximum alternative criticizes b '_maxSize be designated as s_max；

Step 5.10, described Maximum alternative is criticized b '_maxJoin described the l groupAnd Maximum alternative is criticized b '_max's Identifier flag '_maxIt is set to 1；

Step 5.11, described vehicle volume V is deducted described Maximum alternative criticize b '_maxSize s_max, it is thus achieved that residue vehicle volume, It is designated as

Step 5.12, according to described residue vehicle volumeFrom described the l candidate list to be updatedMiddle deletion Size is more than described residue vehicle volumeCandidate criticize；Thus obtain the l candidate list X of renewal_l ^(L)(m)；

Step 5.13, obtain described Maximum alternative and criticize b '_maxCorresponding classification, is designated as w_max；

Step 5.14, judge that described Maximum alternative criticizes b '_maxAt corresponding classification w_maxIn whether be f_zIndividual batch；The most then from The l the candidate list X updated_l ^(L)(m)The described Maximum alternative of middle deletion criticizes b '_max；Otherwise, from the l the candidate list X updated_l ^(L)(m) The described Maximum alternative of middle deletion criticizes b '_max, and described Maximum alternative is criticized b '_maxCorresponding classification w_maxIn f_max+ 1 batch adds Enter the l candidate list X to described renewal_l ^(L)(m)In, thus obtain the l the candidate list X again updated_l ^(L)(m)；

Step 5.15, with described the l the candidate list X again updated_l ^(L)(m)As the l candidate list to be updated And return step 5.8 order execution, until the l candidate list to be updatedTill sky, will the L time iteration The l group of m Formica fuscaFill it up with；Thus obtain the l group of m Formica fusca of the L time iterationJoin Send scheme；

Step 5.16, the identifier judging n batch are the most all 1, if being all 1, then it represents that complete the m ant of the L time iteration The ant distribution to described n batch；And l is assigned toOtherwise, l+1 is assigned to l, and returns step 5 order execution.