WO2021068274A1 - 基于操作完工时间快速预测的集成电路生产线调度方法 - Google Patents
基于操作完工时间快速预测的集成电路生产线调度方法 Download PDFInfo
- Publication number
- WO2021068274A1 WO2021068274A1 PCT/CN2019/111449 CN2019111449W WO2021068274A1 WO 2021068274 A1 WO2021068274 A1 WO 2021068274A1 CN 2019111449 W CN2019111449 W CN 2019111449W WO 2021068274 A1 WO2021068274 A1 WO 2021068274A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- scheduling
- sub
- integrated circuit
- production line
- time
- Prior art date
Links
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 53
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000012545 processing Methods 0.000 claims abstract description 50
- 239000003016 pheromone Substances 0.000 claims abstract description 22
- 230000002040 relaxant effect Effects 0.000 claims abstract 2
- 238000004422 calculation algorithm Methods 0.000 claims description 22
- 238000000354 decomposition reaction Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 13
- 230000009977 dual effect Effects 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 4
- 238000012804 iterative process Methods 0.000 claims 1
- 238000005096 rolling process Methods 0.000 claims 1
- 238000005457 optimization Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 5
- 230000007704 transition Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 238000003754 machining Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 238000011112 process operation Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000002922 simulated annealing Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
- G05B19/41865—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by job scheduling, process planning, material flow
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32252—Scheduling production, machining, job shop
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Definitions
- the invention belongs to the fields of advanced manufacturing, automation and information, and specifically relates to an integrated circuit production line scheduling method based on rapid prediction of operation completion time.
- the integrated circuit manufacturing industry is a basic industry with strategic resources and international characteristics. It is the cornerstone and core of the information society. It penetrates and drives traditional industries and plays a key role in national security and national defense.
- the technical level of optimized scheduling of integrated circuit production lines has always been one of the bottlenecks restricting integrated circuit manufacturing enterprises to improve their competitiveness and the development of the integrated circuit manufacturing industry. With the fierce competition in the global market, integrated circuit production lines have shown the characteristics of low-volume high-mix, and the corresponding scheduling problems are of large scale (from production to product molding, dozens or even hundreds of channels are often required. It is difficult to optimize the solution due to its characteristics such as process steps) and complex constraints (including menu constraints, Setup time constraints, batch constraints, etc.). In recent years, this research direction has attracted the attention of academic and industrial circles at home and abroad.
- the present invention proposes an integrated circuit production line scheduling method based on rapid prediction of the operation completion time.
- Lot processing Process (called menu in integrated circuit manufacturing companies) is the link that connects operations and machines.
- l 1, 2,...,L ⁇ as the set of all menus, the menu for operating O ij is e ij ⁇ E, and the machine M k Machinable menu composition collection
- the operation O ij can be processed on the machine M k , and the corresponding processing time It is uniquely determined only by its processing machine and its menu.
- the batching capacity of the machine all machines can be divided into batch processing machines And serial processing machines On any serial processing machine, only one operation can be processed at each time. On any group of batch processing machines, multiple operations can be processed at each time. The simultaneous processing operations form a batch.
- the setup time required for the processing operation O ij is marked as u ij .
- the scheduling goal of the present invention is to minimize the average flow. over time, provided J i c i ⁇ 0 is a completion time of the last operation, the flow through time J i for c i -r i, the objective function is minimized
- the present invention needs to decompose the scheduling problem iteratively.
- the predicted value of the completion time of each operation in the scheduling problem needs to be used.
- the present invention proposes a method for quickly predicting the completion time of the operation based on the machine load .
- R ij is the set operation time of arrival O ij, i.e., the operation may begin early O ij processing time, and before the operation following the completion of time O i, j-1, then, the operation completion time O ij c ij can be processed according to its
- the waiting processing operation before the machine set M ij is predicted as follows:
- represents the number of elements in the set A.
- processing load equally distributed by the operation O ij to each machine in the set of processing machines M ij is:
- the machine front so that W k M k set to wait for the machining operation the theoretical load machine M k may be calculated according to the formula:
- the arrival time of the operation O im can be predicted as:
- the completion time c i of Lot J i can be predicted by the following formula based on the arrival time r ij of its j-th operation and the theoretical load ⁇ k of each machine:
- the present invention proposes an iterative decomposition method for scheduling problems, which decomposes the original scheduling problem into multiple sub-scheduling problems, and then solves the sub-scheduling problems .
- the iterative decomposition of the scheduling problem proposed by the present invention is carried out as follows:
- Step (1) Sub-scheduling problem construction
- stage (stage k) scheduled operation ie determines the operation starting time of processing
- Equation (9) is the objective function of scheduling: minimize the average flow time of the global prediction as the objective function of the sub-problem, J i is the predicted completion time is obtained by the formula (7), formula (10) given process constraints and path constraints Setup time, k (i, j-1 ) represents O i, j-1 processing machine, of formula (11) gives the non-interruptible constraint, and formula (12) gives the sequential processing sequence constraint of two operations on the same machine. If the operation O ij is processed before the operation O mn on the machine M k , then otherwise, Is a very large positive number. Equation (13) indicates that any machine can only process one menu at the same time.
- Equation (14) indicates that an operation can only be processed on one machine. If Lot J i is processed on machine M k at time t, then otherwise, Equation (15) gives the group batch constraint, and Equation (16) gives the start time of the first operation of each Lot in the subproblem, which is constrained by the completion time of the corresponding operation in the operation set B k of the determined processing start time; 2): Sub-scheduling problem adjustment
- the present invention adopts the sub-scheduling problem overlap mode to process the relationship between two consecutive sub-scheduling problems, suppose the scale of the overlap operation set R k is N r , and the set of batch processing machines that have been arranged for operation and processing in the sub-scheduling problem is Use the following steps to select some operations from the current stage sub-scheduling problem as overlap operations and move to the next stage sub-scheduling problem for solution:
- Step (2.3.1) sequentially select the operations with the largest processing start time s ij from H k and put them into the operation set R k so that the scale of the operation set R k reaches N r ;
- Step (2.3.2) Let step 2.3.1 move into all operations in the operation set R k , the operation corresponding to the smallest processing start time is Will be operated with All operations in the same batch are moved to the operation set R k ;
- the present invention proposes an ant colony algorithm based on dual pheromone to solve the sub-scheduling problem.
- the specific steps are as follows:
- Step (1) Determine the state transition probability in the dual pheromone ant colony algorithm
- Dual pheromone refers to Lot processing menu information and operation information.
- the ant colony algorithm needs to determine the state transition probability of the dual information office:
- ⁇ l,l' is the pheromone concentration of menu e l over menu e l'
- ⁇ l is heuristic information, using the DRLB that is to be processed on the current machine and has the operation of menu e l
- the average value is used as the heuristic information ⁇ l
- ⁇ is the pheromone factor
- ⁇ is the heuristic factor
- ⁇ ij is the heuristic information of operation O ij , which is used to guide the generation of the ant colony algorithm path, and the scheduling priority index given by DRLB is used as the heuristic information of operation O ij ⁇ ij , ⁇ and ⁇ are pheromone factor and heuristic factor respectively;
- Step (2) Determine the operation selection strategy in the dual pheromone ant colony algorithm
- ⁇ Menu selection from The menu set of the operation to be processed In, according to the menu selection probability given by formula (17), the processing menu is selected according to the way of roulette
- n l is The total number of operations waiting to be processed and with menu e l, for The batching capacity of, that is, the maximum total number of operations belonging to the menu e l can be processed at the same time, if Perform step (2.2.1), otherwise, perform step (2.2.2)
- Step (3) Pheromone update and initialization
- ⁇ b 1/f best
- f best is the objective function value corresponding to the global optimal solution S gs or the current iterative optimal solution S ls
- the objective function corresponding to the current iterative optimal solution S ls is used in the initial iteration of the algorithm Value and new pheromone to strengthen the local search ability of the algorithm, and in the subsequent iteration stage, by judging whether the global optimal solution S gs has not been improved for several consecutive generations, it is decided whether to use the global optimal solution S gs instead of the current iterative optimal
- the solution S ls updates the pheromone to strengthen the global search capability of the algorithm, so that the algorithm can jump from the local optimum; for the menu pheromone ⁇ l,l′ , when each sub-problem starts to be solved, it directly inherits the previous stage The menu pheromone at the end of the solution of the sub-problem to accelerate the convergence of the algorithm; while for the operation pheromone It needs to be
- the present invention has done a lot of simulation experiments. From the simulation results, it can be seen that the method proposed by the present invention can effectively improve the scheduling by minimizing the average flow time.
- the scheduling index of the target integrated circuit production line scheduling problem is a lot of simulation experiments.
- Figure 1 is a schematic diagram of the software and hardware architecture of the present invention in the integrated circuit production line scheduling.
- Figure 2 is a schematic flow diagram of an integrated circuit production line scheduling method based on rapid prediction of operation completion time.
- the scheduling method proposed by the present invention relies on the relevant data acquisition system, and is implemented by the integrated circuit production line scheduling system client and the scheduling server.
- a schematic diagram of the software and hardware architecture of the application of the present invention in the actual integrated circuit production line scheduling is shown in Figure 1.
- the implementation of the present invention is as follows:
- Step (1) Obtain the data corresponding to the integrated circuit production line scheduling problem
- Step (2) Establish an integrated circuit production line scheduling problem with minimizing the average flow time as the scheduling objective: use the method in the first section of the "Summary of the Invention" to construct an example of the integrated circuit production line scheduling problem to be solved.
- Step (3) Iterative decomposition of integrated circuit production line scheduling problems and solution of sub-scheduling problems
- the parameters of the ant colony algorithm are set as follows: the maximum number of iterations is set to 100; the maximum number of restarts is 10; The T_min that triggers the restart is 3; the colony size is 50; the pheromone factor and heuristic factors ⁇ and ⁇ are set to 2 and 3, respectively, and the forgetting factor ⁇ is set to 0.1.
- the present invention has done a lot of simulation tests.
- the running hardware environment is: P4 4GHz CPU, 2048M RAM, and the operating system is Windows 10. Due to space limitations, only some experimental results are listed.
- a different scale scheduling problem instance is formed for numerical calculation.
- the actual data contains 70 menus and 117 machines, of which 77 machines are batch processing machines and the rest are serial processing machines. Among the batch machines, the largest and smallest batch capacities are 8 and 2, respectively.
- 5 types of integrated circuit production line scheduling problem examples were randomly generated, including 500, 1000, 2000, 3000, and 5000 Lots.
- Each type contains 10 problem examples, named I1_1,..., I1_10, I2_1, ..., I2_10, ..., I5_1, ..., I5_10.
- the main parameters of the proposed ant colony algorithm based on predictive decomposition are set as follows:
- the sub-problem contains the operand N_s as 100; the overlap operation set R_k scale N_r is 10; the maximum number of ACO-D iterations is set to 100;
- the maximum number of restarts of ACO-D is 10; the T_min that triggers the restart is 3; the colony size is 50; the pheromone factor and heuristic factors ⁇ and ⁇ in ACO-D are set to 2 and 3, respectively, and the forgetting factor ⁇ is set to 0.1.
- Table 1 and Table 2 respectively list the application of the scheduling method (DM-IFP) proposed by the present invention and the typical integrated circuit production line scheduling algorithm in the literature on randomly generated scheduling problems with 3000 Lots and 5000 Lots. Compared. Comparison methods include:
- ACO (Guo C T, Jiang Z B, Zhang H, et al. Decomposition-based classified ant colony optimization algorithm for scheduling semiconductor wafer manufacturing system Computers & Industrial Engineering, 2012, 62(1):141-151)
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Manufacturing & Machinery (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- General Factory Administration (AREA)
Abstract
基于操作完工时间快速预测的集成电路生产线调度方法，属于先进制造、自动化和信息领域，其特征在于，针对以最小化平均流经时间为调度目标的集成电路生产线调度问题，首先通过松弛不可中断约束，提出一种基于机器负载的操作完工时间快速预测(CTP-ML)方法对操作完工时间进行预测，之后，根据各个操作的完工时间预测值，将集成电路生产线调度问题迭代分解为多个连续交迭的子调度问题，在每次迭代中，采用一种基于双信息素的蚁群算法(ACO-D)求解当前迭代阶段的子调度问题，并固定该子调度问题解的开始加工时间，将其余操作滚动到下一迭代子调度问题中。使用该调度方法可有效改善集成电路生产线的平均流经时间调度指标。
Description
相关申请的交叉引用
本申请要求于2019年10月10日提交的申请号为2019109575666，发明名称为“基于操作完工时间快速预测的集成电路生产线调度方法”的中国专利申请的优先权，其通过引用方式全部并入本公开。
本发明属于先进制造、自动化和信息领域，具体涉及一种基于操作完工时间快速预测的集成电路生产线调度方法。
集成电路制造产业是具有战略资源和国际特征的基础产业，是信息社会的基石和核心，对传统产业的具有渗透与带动作用，对国家安全与国防建设起到关键作用。而集成电路生产线优化调度技术水平一直是制约集成电路制造企业提高竞争力和集成电路制造产业发展的瓶颈之一。随着全球市场激烈的竞争，集成电路生产线呈现出多品种小批量(low-volume high-mix)特点，其相应的调度问题具有规模大(从投入生产到产品成型往往需要几十道甚至上百道工步)和约束复杂(包括菜单约束、Setup时间约束、组批约束等)等特点，其优化求解具有一定难度，该研究方向近年来一直受到国内外学术界和工业界的关注。
针对上述具有大规模特点的离散型生产过程调度问题，国内为学者们试图从问题分解的角度，降低求解规模，提高蚁群算法、遗传算法等群体智能优化算法的求解性能。主要包括基于机器的分解和基于时间的分解方法。机器分解的方法主要是将原调度问题涉及的机器根据一些属性(如瓶颈程度、机器所属流程等)进行分组，对不同的机器组采用不同的策略进行求解，然后协调不同分组之间的约束形成原调度问题的解。然而，由于不同子问题间存在约束，为获得问题的可行解往往需要对子问题反复迭代协调求解，其求解效率往往较低。另外，由于集成电路生产线存在多品种小批量及可重入特性，使得生产过程中瓶颈机器等特性会发生漂移，从而上述基于机器分解的方法难以直接应用于集成电路生产线调度问题。
发明内容
为解决以最小化平均流经时间为调度目标的集成电路生产线调度问题，本发明提出一种基于操作完工时间快速预测的集成电路生产线调度方法。
(一)以最小化平均流经时间为调度目标的集成电路生产线调度问题描述
设集成电路生产线调度问题包含I个Lot(加工晶圆批次)和K台机器，分别组成Lot集合J＝{J
i|i＝1，2，…，I}和机器集合M＝{M
k|k＝1，2，…，K}，r
i≥0为Lot J
i的释放时刻，Lot J
i(j＝1，2，…，n
i)由n
i个操作{O
ij|j＝1，2，…，n
i}组成，O＝{O
ij|i＝1，2，…，I；j＝1，2，…，n
i}为调度问题所有操作组成的集合，Lot的加工工艺(集成电路制造企业中称为菜单)是连接操作和机器的纽带，每个Lot的每个操作均具有某一菜单，而每台机器仅能加工具有某些菜单的操作(以下将机器加工具有某菜单的操作简称为机器加工某菜单)，设E＝{e
l|l＝1，2，…，L}为所有菜单的集合，操作O
ij的菜单为e
ij∈E，机器M
k可加工的菜单组成集合
从而当且仅当e
ij∈E
k时，操作O
ij可以在机器M
k上加工，相应的加工时间
仅由其加工机器和具有的菜单唯一确定，根据机器的组批能力，所有的机器可分为组批加工机器
和串行加工机器
在任一串行加工机器上，每个时刻仅能加工一个操作，在任一组批加工机器上，每一时刻可加工多个操作，同时加工的操作形成一批，能同时加工的最大操作数称为机器的组批能力，设B
lk为菜单e
l(l＝1，2，…，L)在机器M
k(k＝1，2，…，K)上的组批能力，不失一般性，对于串行加工机器M
k∈M
s和菜单e
l∈E
k，相应的组批能力B
lk＝1，另外，在机器M
k(k＝1，2，…，K)上，若即将加工操作的菜单e
l与上一个加工操作的菜单不同，则需要一个额外的Setup时间,令加工操作O
ij所需的Setup时间标记为u
ij，本发明涉及的调度目标是以最小化平均流经时间，设c
i≥0为J
i最后一个操作的完工时间，J
i的流经时间为c
i-r
i，则最小化目标函数为
(二)集成电路生产线调度问题中各操作完工时间快速预测
为求解上述集成电路生产线调度问题，本发明需将调度问题进行迭代分解，迭代分解中需要使用调度问题中各操作的完工时间预测值，本发明提出一种基于机器负载的操作完工时间快速预测方法。
设r
ij为操作O
ij的到达时间，即操作O
ij最早可开始加工时间，及前继操作O
i，j-1的完工时间，那么，操作O
ij的完工时间c
ij可根据其可加工机器集合M
ij前 的等待加工操作，即机器负载，按如下方法进行预测：
考虑到操作O
ij在其可加工机器集M
ij中每台机器M
k∈M
ij的加工时间
互不相同，将其在上述所有机器上的平均加工时间
来表征实际加工时间：
|A|表示集合A中元素的个数。
进一步，操作O
ij平均分配到其可加工机器集合M
ij中每台机器上的加工负载为：
那么，令W
k为机器M
k前等待加工操作的集合，则机器M
k的理论负载可根据下式计算：
根据式(3)，操作O
ij的完工时间c
ij可预测为：
进而，Lot J
i的第m(m＞j)个操作O
im的完工时间c
im可预测为：
那么，操作O
im的到达时间可预测为：
Lot J
i的完工时间c
i可根据其第j个操作的到达时间r
ij及各机器的理论负载Γ
k通过下式预测：
(三)集成电路生产线调度问题迭代分解
为求解上述集成电路生产线调度问题，基于第二节得到的操作完工时间预测值，本发明提出一种调度问题迭代分解方法，将原调度问题分解为多个子调度问题，再对子调度问题进行求解。本发明提出的调度问题迭代分解按如下步骤进行：
步骤(1)：子调度问题构造
根据Lot的工艺路径、时间窗口及操作完工时间预测值，从未调度操作 中选取当前可加工及部分将来可加工的操作构成子问题，子问题的调度目标为全局调度目标的预测值，设当前阶段(第k阶段)已调度操作(即已确定加工开始时间的操作)组成集合B
k,b
i(i＝1，2，…，I)为工件J
i中属于集合B
k的最大操作号，即
b
i＝max{j|O
ij∈B
k} (8)
然后，根据式(5)和(6)对不属于操作集B
k的操作O
ij(j＞b
i)的到达时间r
ij进行预测，并从其中选取N
s个
最小的操作组成子问题的操作集H
k，其余操作形成操作集合P
k，根据上述子问题的形成过程，第k阶段，子问题的数学规划模型可描述为：
式(9)为调度的目标函数：最小化全局预测的平均流经时间作为子问题的目标函数，
为J
i的预测完工时间，通过式(7)获得，式(10)给出了工艺路径约束和Setup时间约束，k(i，j-1)表示O
i，j-1的加工机器，式(11)给出了不可中断约束，式(12)给出了在同一台机器上两个操作的先后加工顺序约束,若在机器M
k上操作O
ij先于操作O
mn加工，则
否则，
是一极大的正数，式(13)表明任一台机器同一时刻只能加工一个菜单，若时刻t在机器M
k上加工的菜单为ρ
l，则
否则，
式(14)表明一个操作只能在一台机器上加工，若时刻t Lot J
i在机器M
k上加工，则
否 则，
式(15)给出了组批约束，式(16)给出子问题中每个Lot第一个操作的开始时间受已确定加工开始时间操作集B
k中相应操作完工时间的约束；步骤(2)：子调度问题调整
本发明采用子调度问题搭接的方式处理两个连续子调度问题间关系，设搭接操作集R
k规模为N
r，子调度问题中已安排操作加工的组批加工机器集合为
采用如下步骤从当前阶段子调度问题中选择部分操作作为搭接操作移入下一阶段子调度问题进行求解：
步骤(2.1)：令
步骤(2.2)：对Ω
b中每一台机器M
i∈Ω
b，若最后一批加工的操作所属的批次小于最大组批能力，则该批所有操作均为搭接操作，即将该批所有操作并入操作集R
k中；
步骤(2.3)：若R
k中所包含操作的总数小于N
r，则执行如下步骤，否则，结束流程：
步骤(2.3.1)：从H
k中顺序选择部分加工开始时间s
ij最大的操作放入操作集合R
k中，使得操作集R
k规模达到N
r；
步骤(2.3.2)：设步骤2.3.1移入操作集R
k中所有操作中，最小的加工开始时间对应的操作为
则将与操作
同批的所有操作移入到操作集R
k；
(四)子调度问题求解
针对在第三节中构建的子调度问题，本发明提出一种基于双信息素的蚁群算法对子调度问题进行求解，具体步骤如下：
步骤(1)：确定双信息素的蚁群算法中的状态转移概率
双信息素是指Lot加工菜单信息和操作信息，在蚁群算法需要确定双信息署的状态转移概率：
1)确定菜单状态转移概率
对于选定的机器
菜单e
l的选择概率如下式所示：
其中，
为被选择机器
待加工操作的所有菜单集合，τ
l，l′为 菜单e
l优先于菜单e
l′的信息素浓度，η
l为启发式信息，采用在当前机器待加工且具有菜单e
l的操作的DRLB平均值作为启发式信息η
l，α为信息素因子，β为启发式因子；
2)确定操作状态转移概率
对于选定的机器
和选定的菜单
操作O
ij的选择概率如下式所示：
其中，
为被选择机器
上选择加工操作具有的菜单，
为操作O
ij优先于操作O
mn的信息素浓度，η
ij为操作O
ij启发式信息，用来引导蚁群算法路径的生成，采用DRLB给出的调度优先级指标作为操作O
ij启发式信息η
ij，α和β分别为信息素因子和启发式因子；
步骤(2)：确定双信息素的蚁群算法中的操作选择策略
步骤(2.1)：若被选机器
为串行加工机器，按如下方法确定操作选择策略，否则，转到步骤(2.2)
●菜单选择：从
待加工操作具有的菜单集
中，根据式(17)给出的菜单选择概率，按照轮盘赌的方式选出加工菜单
●操作选择：从具有菜单
的操作中，根据式(18)给出的操作选择概率，按照轮盘赌的方式选出在机器
上加工的操作；
步骤(2.2)：若被选机器
为组批加工机器
从
选出操作数大于或等于机器
组批能力的菜单组成菜单集
其中，n
l为
等待加工且具有菜单e
l的操作总数，
为
的组批能力，即能同时加工属于菜单e
l的最大操作总数，若
执行步骤(2.2.1)，否则，执行步骤(2.2.2)
步骤(2.2.1)：
●菜单选择：根据式(17)给出的菜单选择概率，按照轮盘赌的方式从
中选出加工菜单
●操作选择：从具有菜单
的待加工操作中，根据式(18)给出的操作选择概率，按照轮盘赌的方式顺序选出
个操作组成一批在机器
上加工；
步骤(2.2.2)：
从
中选出操作数大于最小组批规模的菜单组成菜单集
为
的最小组批规模，即能同时加工属于菜单e
l的最小操作总数。根据式(17)给出的菜单选择概率，按照轮盘赌的方式从
中选出加工菜单
并选择具有菜单
的所有待加工操作组成一批在机器
上加工，若
空闲等待；
步骤(3)：信息素更新及初始化
在蚁群算法每次迭代完成后，需要对信息素τ
l，l′和
进行更新，设S
ls和S
gs分别为每次迭代的最优解和迄今为止的全局最优解，每次迭代后，若S
ls优于S
gs，则全局最优解S
gs被S
ls替换，信息素τ
l，l′和
分别采用式(19)和(20)更新
τ
l，l′＝(1-ρ)τ
l，l′+ρΔτ
b (19)
其中，Δτ
b＝1/f
best，f
best为全局最优解S
gs或当前迭代最优解S
ls对应的目标函数值，在算法初始迭代阶段采用当前迭代最优解S
ls对应的目标函数值跟新信息素，以加强算法的局部搜索能力，而在后续迭代阶段，通过判断全局最优解S
gs是否连续若干代未被改进，决定是否采用全局最优解S
gs代替当前迭代最优解S
ls对信息素进行更新，以加强算法的全局搜索能力，使算法能从局部最优跳跃出来；对于菜单信息素τ
l，l′，每一个子问题开始求解时，直接继承上一阶段子问题的求解结束时的菜单信息素，以加速算法的收敛；而对于操作信息素
在每个子问题求解前均需要重新初始化。
根据上述一种基于操作完工时间快速预测的集成电路生产线调度方法，本发明做了大量的仿真实验，从仿真结果可看出，本发明所提方法可有效改善以最小化平均流经时间为调度目标的集成电路生产线调度问题的调度指标。
图1是集成电路生产线调度中应用本发明的软硬件架构示意图。
图2是基于操作完工时间快速预测的集成电路生产线调度方法流程示意图。
本发明提出的调度方法依赖于相关数据采集系统，由集成电路生产线调 度系统客户端和调度服务器实现。在实际集成电路生产线调度中应用本发明的软硬件架构示意图如图1所示，本发明的实施方式如下：
步骤(1)：获取集成电路生产线调度问题对应的数据
包括Lot数量及释放时间、各Lot中操作数量及工艺菜单和加工时间、可用设备数量及各设备的释放时间、各设备可加工的工艺菜单等，并存储至调度数据库中。
步骤(2)：建立以最小化平均流经时间为调度目标的集成电路生产线调度问题：采用“发明内容”第一节的方法，构建待求解的集成电路生产线调度问题实例。
步骤(3)：集成电路生产线调度问题迭代分解及子调度问题求解
步骤(3.1)：操作完工时间快速预测：采用“发明内容”第二节的方法，对未确定调度开始时间的待调度的操作的完工时间进行预测。
步骤(3.2)：子调度问题构造：采用“发明内容”第三节步骤(1)的方法构造子调度问题。
步骤(3.3)：子调度问题求解：采用“发明内容”第四节的方法对子调度问题进行求解，蚁群算法的参数设定为：最大迭代次数设置为100；最大重新启动次数为10；触发重新启动的T_min为3；蚁群规模为50；信息素因子和启发式因子α和β分别设为2和3，遗忘因子ρ设置为0.1。
步骤(3.3)：子问题调整：采用“发明内容”第三节步骤(2)的方法，根据步骤(3.3)确定的子调度问题解(各操作的开始加工时间)，对子调度问题进行调整。
步骤(3.4)：判断所有操作的开始建工时间是否均确定了，若是，则结束；否则，返回步骤(3.1)。
根据上述所提出的基于操作完工时间快速预测的集成电路生产线调度方法，本发明做了大量的仿真试验，运行的硬件环境为：P4 4GHz CPU，2048M RAM，操作系统为Windows10。由于篇幅所限，仅列出部分实验结果。首先，基于实际集成电路制造企业的实际生产数据形成不同规模的调度问题实例进行数值计算。实际数据中包含70个菜单和117台机器，其中77台机器为组批加工机器，其余的为串行加工机器。组批机器中，最大和最小的组批能力分别为8和2。根据上述实际数据，随机生成了5类集成电路生产线调度问 题实例，分别包含500、1000、2000、3000和5000个Lot，每类包含10个问题实例，分别命名为I1_1、…、I1_10、I2_1、…、I2_10、…、I5_1、…、I5_10。所提出基于预测分解的蚁群算法(简称DM-IFP)的主要参数设置如下：子问题包含操作数N_s为100；搭接操作集R_k规模N_r为10；ACO-D最大迭代次数设置为100；ACO-D最大重新启动次数为10；触发重新启动的T_min为3；蚁群规模为50；ACO-D中信息素因子和启发式因子α和β分别设为2和3，遗忘因子ρ设置为0.1。
表1和表2分别列出了本发明所提出的调度方法(DM-IFP)与文献中的典型集成电路生产线调度算法在随机产生的具有3000个Lot和5000个Lot的调度问题实例上的应用对比。对比方法包括：
1)ACO(Guo C T,Jiang Z B,Zhang H,et al.Decomposition-based classified ant colony optimization algorithm for scheduling semiconductor wafer fabrication system.Computers&Industrial Engineering,2012,62(1):141-151)
2)GA(Noroozi A,Mokhtari H,Kamal Abadi I N.Research on computational intelligence algorithms with adaptive learning approach for scheduling problems with batch processing machines[J].Neurocomputing,2013,101:190-203)
3)SA(Chou F,Wang H,Chang P.A simulated annealing approach with probability matrix for semiconductor dynamic scheduling problem[J].Expert Systems with Applications,2008,35(4):1889-1898)
表1具有3000个Lot的调度问题实例
表2具有5000个Lot的调度问题实例
Claims (1)
- 基于操作完工时间快速预测的集成电路生产线调度方法，其特征在于，该方法针对以最小化平均流经时间为调度目标的集成电路生产线调度问题，首先通过松弛不可中断约束，提出一种基于机器负载的操作完工时间快速预测(CTP-ML)方法对操作完工时间进行预测，之后，根据各个操作的完工时间预测值，将集成电路生产线调度问题迭代分解为多个连续交迭的子调度问题，在每次迭代中，采用一种基于双信息素的蚁群算法(ACO-D)求解当前迭代阶段的子调度问题，并固定该子调度问题解的开始加工时间，将其余操作滚动到下一迭代子调度问题中；所述方法在计算机上依次按如下步骤实现：步骤(1)：获取集成电路生产线相关数据基于集成电路生产线上的数据采集系统，获取集成电路生产线调度问题相关的数据，具体包括Lot数量及释放时间、各Lot中操作数量及工艺菜单和加工时间、可用设备数量及各设备的释放时间、各设备可加工的工艺菜单等，并存储至调度数据库中；步骤(2)：根据调度数据库中的相关信息，建立以最小化平均流经时间为调度目标的集成电路生产线调度问题实例；步骤(3)：集成电路生产线调度问题迭代分解及子调度问题求解步骤(3.1)：松弛集成电路生产线调度问题中的不可中断约束，采用基于机器负载的操作完工时间的快速预测(CTP-ML)方法，以获取未确定加工开始时间操作的近似完工时间；步骤(3.2)：子调度问题构造，采用滚动时域方法，通过限定子问题包含的操作数，对调度问题进行顺序迭代搭接式分解，每一迭代过程固定部分操作的加工开始时间，部分操作滚动到下一子问题参与资源竞争；根据Lot的工艺路径、时间窗口及操作完工时间预测值，从未调度操作中选取当前可加工及部分将来可加工的操作构成子问题；步骤(3.3)：子调度问题求解，采用一种基于双信息素的蚁群算法(ACO-D)对子调度问题进行求解，蚁群算法的参数设定为：最大迭代次数设置为100；最大重新启动次数为10；触发重新启动的T_min为3；蚁群规模为50；信息素因子和启发式因子α和β分别设为2和3，遗忘因子ρ设置为0.1；步骤(3.4)：子调度问题调整，采用子调度问题搭接的方式处理两个连 续子调度问题间关系，在本阶段的子调度问题中选取部分操作作为搭接操作移入下一阶段子调度问题中。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910957566.6A CN110673560B (zh) | 2019-10-10 | 2019-10-10 | 基于操作完工时间快速预测的集成电路生产线调度方法 |
| CN201910957566.6 | 2019-10-10 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2021068274A1 true WO2021068274A1 (zh) | 2021-04-15 |
Family
ID=69081276
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2019/111449 WO2021068274A1 (zh) | 2019-10-10 | 2019-10-16 | 基于操作完工时间快速预测的集成电路生产线调度方法 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110673560B (zh) |
| WO (1) | WO2021068274A1 (zh) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113341896A (zh) * | 2021-06-07 | 2021-09-03 | 电子科技大学 | 面向离散制造的动态集成车间调度与装配序列规划方法 |
| CN115186495A (zh) * | 2022-07-18 | 2022-10-14 | 广东工业大学 | 一种基于关键工序增强优化的流水线快速优化方法 |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113093724B (zh) * | 2021-02-24 | 2022-09-06 | 上海工程技术大学 | 一种基于改进蚁群算法的agv路径规划方法 |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101788819A (zh) * | 2010-03-08 | 2010-07-28 | 清华大学 | 大规模生产过程一种基于迭代式分解和流松驰的调度方法 |
| CN103116324A (zh) * | 2012-12-17 | 2013-05-22 | 清华大学 | 基于指标预测和在线学习的微电子生产线调度方法 |
| CN103246969A (zh) * | 2013-05-07 | 2013-08-14 | 北京邮电大学 | 一种物流调配的实现方法和装置 |
| CN104460590A (zh) * | 2014-10-23 | 2015-03-25 | 北京化工大学 | 一种半导体生产线多产品工件合并方法 |
| US20170185928A1 (en) * | 2015-12-28 | 2017-06-29 | Sap Se | Data analysis for scheduling optimization with multiple time constraints |
| CN110086855A (zh) * | 2019-03-29 | 2019-08-02 | 江苏开拓信息与系统有限公司 | 基于蚁群算法的Spark任务智能感知调度方法 |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105741181A (zh) * | 2016-02-03 | 2016-07-06 | 四川索牌科技股份有限公司 | 一种不同并行机混合流水线车间调度方法 |
-
2019
- 2019-10-10 CN CN201910957566.6A patent/CN110673560B/zh active Active
- 2019-10-16 WO PCT/CN2019/111449 patent/WO2021068274A1/zh active Application Filing
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101788819A (zh) * | 2010-03-08 | 2010-07-28 | 清华大学 | 大规模生产过程一种基于迭代式分解和流松驰的调度方法 |
| CN103116324A (zh) * | 2012-12-17 | 2013-05-22 | 清华大学 | 基于指标预测和在线学习的微电子生产线调度方法 |
| CN103246969A (zh) * | 2013-05-07 | 2013-08-14 | 北京邮电大学 | 一种物流调配的实现方法和装置 |
| CN104460590A (zh) * | 2014-10-23 | 2015-03-25 | 北京化工大学 | 一种半导体生产线多产品工件合并方法 |
| US20170185928A1 (en) * | 2015-12-28 | 2017-06-29 | Sap Se | Data analysis for scheduling optimization with multiple time constraints |
| CN110086855A (zh) * | 2019-03-29 | 2019-08-02 | 江苏开拓信息与系统有限公司 | 基于蚁群算法的Spark任务智能感知调度方法 |
Non-Patent Citations (2)
| Title |
|---|
| HAO JING-HUA, LIU MIN, LIU YI-ZHOU, CHENG WU, RUI ZHANG: "Decomposition-based Optimization Algorithm for Complex Parallel Machine Scheduling Problems", CONTROL ENGINEERING OF CHINA, vol. 12, no. 6, 1 January 2005 (2005-01-01), pages 520 - 526, XP055798908 * |
| MIN LIU, MING-YU DONG, CHENG WU: "An objective decomposing method based on IBMDC for solving complex production scheduling problem", MACHINE LEARNING AND CYBERNETICS, 2002. PROCEEDINGS. 2002 INTERNATIONA L CONFERENCE ON NOV. 4-5, 2002, PISCATAWAY, NJ, USA,IEEE, vol. 3, 4 November 2002 (2002-11-04) - 5 November 2002 (2002-11-05), pages 1676 - 1679, XP010805089, ISBN: 978-0-7803-7508-6, DOI: 10.1109/ICMLC.2002.1167499 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113341896A (zh) * | 2021-06-07 | 2021-09-03 | 电子科技大学 | 面向离散制造的动态集成车间调度与装配序列规划方法 |
| CN113341896B (zh) * | 2021-06-07 | 2022-08-05 | 电子科技大学 | 面向离散制造的动态集成车间调度与装配序列规划方法 |
| CN115186495A (zh) * | 2022-07-18 | 2022-10-14 | 广东工业大学 | 一种基于关键工序增强优化的流水线快速优化方法 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110673560A (zh) | 2020-01-10 |
| CN110673560B (zh) | 2021-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021068274A1 (zh) | 基于操作完工时间快速预测的集成电路生产线调度方法 | |
| Wang et al. | A bi-population based estimation of distribution algorithm for the flexible job-shop scheduling problem | |
| Zhou et al. | Hyper-heuristic coevolution of machine assignment and job sequencing rules for multi-objective dynamic flexible job shop scheduling | |
| US8527080B2 (en) | Method and system for managing process jobs in a semiconductor fabrication facility | |
| Sang et al. | Intelligent factory many-objective distributed flexible job shop collaborative scheduling method | |
| Guo et al. | Decomposition-based classified ant colony optimization algorithm for scheduling semiconductor wafer fabrication system | |
| Zhang et al. | Multi-objective discrete grey wolf optimizer for solving stochastic multi-objective disassembly sequencing and line balancing problem | |
| Zhang et al. | A genetic algorithm and tabu search for multi objective flexible job shop scheduling problems | |
| Zhang et al. | A Q-learning-based hyper-heuristic evolutionary algorithm for the distributed flexible job-shop scheduling problem with crane transportation | |
| Lin et al. | A simulation-based optimization approach for a semiconductor photobay with automated material handling system | |
| US20230315953A1 (en) | Using deep reinforcement learning for time constraint management at a manufacturing system | |
| Zhang et al. | An improved imperialist competitive algorithm based photolithography machines scheduling | |
| Duan et al. | A hybrid estimation of distribution algorithm for distributed permutation flowshop scheduling with flowline eligibility | |
| Xu et al. | Human-Robot collaboration multi-objective disassembly line balancing subject to task failure via multi-objective artificial bee colony algorithm | |
| Quan et al. | Multi-objective optimization scheduling for manufacturing process based on virtual workflow models | |
| Cao et al. | Surrogate-assisted symbiotic organisms search algorithm for parallel batch processor scheduling | |
| Wu et al. | Multi-objective discrete brainstorming optimizer for stochastic disassembly line balancing problem subject to disassembly failure | |
| Guo et al. | Two-phase imperialist competitive algorithm for energy-efficient flexible job shop scheduling | |
| Wu et al. | Optimizing job release and scheduling jointly in a reentrant hybrid flow shop | |
| Laili et al. | DSAC-Configured Differential Evolution for Cloud–Edge–Device Collaborative Task Scheduling | |
| Mhasawade et al. | A survey of hybrid metaheuristics to minimize makespan of job shop scheduling problem | |
| Hung et al. | Sensitivity search for the rescheduling of semiconductor photolithography operations | |
| He et al. | Memetic Algorithm for Dynamic Joint Flexible Job Shop Scheduling with Machines and Transportation Robots | |
| Quan et al. | Multi-objective evolutionary scheduling based on collaborative virtual workflow model and adaptive rules for flexible production process with operation reworking | |
| Liu et al. | Probabilistic memetic algorithm for flowshop scheduling |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19948738 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 19948738 Country of ref document: EP Kind code of ref document: A1 |