CN111832917A - Fixed point output-oriented distributed manufacturing scheduling method and device - Google Patents

Abstract

The invention discloses a fixed point output-oriented distributed manufacturing scheduling method and a fixed point output-oriented distributed manufacturing scheduling device, wherein the method comprises the following steps: constructing a task structure measurement p model; constructing an equipment distribution state measurement q model; based on the p model and the q model, performing reverse allocation scheduling according to the reverse sequence of the procedures, wherein the scheduling process comprises allocating a termination procedure to termination equipment; acquiring an alternative process set of the assigned processes and an alternative equipment set of the alternative processes; calculating the priority O of each alternative procedure v and the alternative equipment m thereof_v,m(ii) a Selecting the minimum priority O_v,mV and m with corresponding values, and distributing the working procedure v to the equipment m for processing; and repeating the processes until all the procedures are distributed. The invention provides a reverse scheduling method, simplifies the scheduling process, reduces the total scheduling time, optimizes the two aspects of tasks and equipment structures, can better ensure the fixed-point output constraint of the tasks, reduces the scheduling procedures, and mostDistributed manufacturing efficiency is greatly improved.

Description

Fixed point output-oriented distributed manufacturing scheduling method and device

Technical Field

The invention belongs to the technical field of distributed manufacturing, and particularly relates to a fixed-point output-oriented distributed manufacturing scheduling method and device.

Background

The production scheduling problem is an important problem in actual industrial production, and has been widely noticed and researched by scholars as early as over seventy years. Depending on the modeling, the problem can be formulated as a goal optimization problem that satisfies a range of continuous and discrete constraints. At present, even a small-scale production scheduling problem is difficult to obtain an optimal solution. With the change of market demands and the improvement of production equipment, the same type of equipment can realize the processing of multiple types of processes, so that the processes are not bound with the equipment, and the equipment can be selected from a certain equipment set for processing. To meet this Scheduling requirement, Bruker et al further propose Flexible Job Scheduling (FJSP). Flexible job scheduling is also an NP-hard extension of the traditional production scheduling problem. In order to obtain the optimal solution and the approximately optimal solution of the two problems, researchers have proposed a large number of heuristic algorithms, the most well-known algorithm of which is the tabu search algorithm proposed by Watson et al, and in recent years, a series of improved algorithms have been proposed on the basis of the tabu search algorithm. In addition, researchers also successfully apply algorithms such as a genetic algorithm, an ant colony algorithm, a particle swarm algorithm, a Petri network model and the like to the production scheduling problem and the flexible job scheduling problem solution.

With the development of global integration and the formation of a regional economic landscape, the production of a product is no longer produced by traditional local equipment, but rather is cooperatively produced by a plurality of distributed production units. The traditional manufacturing model, in which local area processing capacity is the dominant, has been gradually replaced by distributed manufacturing centered on logistics transportation, resulting in a shift from modern manufacturing to distributed manufacturing. Therefore, a good scheduling algorithm is needed to coordinate distributed devices, so as to perform global optimization on the production of products. Because the traditional workshop scheduling problem only limits the scheduling algorithm to a relatively small workshop, the traditional workshop scheduling problem takes the transportation and distribution coordination problem of products into less consideration. Thus, in the last 10 years, scholars began conducting scheduling optimization studies on distributed manufacturing.

Most distributed plant scheduling problems have the NP-hard nature. At present, a meta-heuristic algorithm provides an effective and feasible scheme for solving a distributed workshop scheduling problem, wherein the scheme comprises a simulated annealing algorithm, a genetic algorithm, an immune algorithm, a particle swarm algorithm and the like, the algorithms can obtain approximate solutions of the problem within feasible time, but because the algorithms do not utilize the transportation characteristics of distributed equipment, the flexible processing capacity of the equipment and the transportation relation of the equipment are not considered sufficiently, and the global optimal solution of the distributed scheduling cannot be guaranteed.

Yet another key factor in the distributed scheduling process is the choice of output point, i.e. where the task is done. The final completion position of the distributed product is a practical problem to be considered in distributed scheduling, and has strong practical significance. For example, in a distributed schedule, with a specific location as an exit point, the process of distributed production performs task allocation around the exit point. In this regard, the distance of the processing tool from the exit point and the association of the structural characteristics of the processing task with the distributed tool selection need to be considered in the scheduling optimization process.

Disclosure of Invention

The invention aims to provide a fixed-point output-oriented distributed manufacturing scheduling method and device, which are used for reasonably distributing tasks and minimizing processing time.

The technical solution for realizing the purpose of the invention is as follows: a fixed point output oriented distributed manufacturing scheduling method, the method comprising the steps of:

step 1, constructing a task structure measurement p model;

step 2, constructing an equipment distribution state measurement q model;

and 3, based on the p model and the q model, performing reverse distribution scheduling according to the reverse sequence of the process.

Further, the task structure metric p model in step 1 is:

in the formula, p_iIs a process v_iProcessing time weight vector, foreword (v) based on task structure_i) Represents the step v_iSet of preceding steps of (1), dis (v)_i,v_j) Represents the step v_iAnd process v_jMinimum path distance in the task structure diagram, α is a control parameter, A_j,.Indicating the j-th row of the processing matrix A, i.e. process v_jThe time required for processing on each piece of equipment.

Further, the device distribution state metric q model in step 2 is:

in the formula, q_jIs a device m_jMachining time weighting u based on the distribution of the equipment_k,jIs a device m_kAnd m_jThe transport time between, σ is the distance control parameter, R_kAnd R_jRespectively a process in a device m_kAnd device m_jThe latency of the process, | M | represents the number of devices in the distributed environment.

Further, in step 3, based on the p model and the q model, the reverse allocation scheduling is performed according to a reverse order of the processes, and the specific process includes:

step 3-1, distributing a termination procedure to termination equipment;

step 3-2, acquiring an alternative process set of the allocated processes, wherein the alternative process set comprises processes immediately before the allocated processes;

step 3-3, acquiring an alternative equipment set of alternative processes, specifically: if a certain process v has been assigned to equipment m, i.e. e (v) ═ m, the set of alternative equipment of the immediately preceding process r (v) of v comprises e (v) and equipment immediately adjacent to e (v);

3-4, calculating the priority O of each alternative procedure v and the alternative equipment m thereof_v,mThe formula used is:

in the formula, O_i,jIs a process v_iWith its alternative device m_jPriority of (d), min (p)_i) Representing the vector p_iMinimum value of (1);

step 3-5, selecting the minimum priority O_v,mV and m with corresponding values, and distributing the working procedure v to the equipment m for processing;

and repeating the steps 3-2 to 3-5 until all the working procedures are distributed.

A fixed point output oriented distributed manufacturing scheduling apparatus, the apparatus comprising:

the first model building module is used for building a task structure measurement p model;

the second model building module is used for building an equipment distribution state measurement q model;

and the distribution scheduling module is used for carrying out reverse distribution scheduling according to the reverse sequence of the working procedures on the basis of the p model and the q model.

A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:

step 1, constructing a task structure measurement p model;

step 2, constructing an equipment distribution state measurement q model;

and 3, based on the p model and the q model, performing reverse distribution scheduling according to the reverse sequence of the process.

A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:

step 1, constructing a task structure measurement p model;

step 2, constructing an equipment distribution state measurement q model;

and 3, based on the p model and the q model, performing reverse distribution scheduling according to the reverse sequence of the process.

Compared with the prior art, the invention has the following remarkable advantages: 1) the invention designs a reverse allocation scheduling algorithm RSM aiming at the fixed-point output problem of tasks by utilizing the distribution state and the transportation relation of equipment in a distributed manufacturing system, wherein the algorithm adopts a reverse scheduling process and starts from equipment termination and procedure termination to allocate equipment for the procedure reverse. In the scheduling process, task structure measurement and equipment distribution state measurement are automatically established according to the task structure and the distribution state of the equipment network, and a basis is provided for task scheduling. Compared with HEF, CEFT and DCP algorithms, the algorithm simplifies the scheduling process, reduces the overall scheduling time and improves the scheduling efficiency of distributed manufacturing; 2) compared with the similar reverse scheduling algorithm, the RSM algorithm has better equipment balance load, can fully utilize low-efficiency equipment in the scheduling process, reduces the total execution time and improves the workshop operation efficiency; 3) the RSM algorithm is optimized from two aspects of tasks and equipment structures, fixed-point output constraint of the tasks can be better guaranteed, scheduling procedures are reduced, and distributed manufacturing efficiency is improved to the maximum extent.

The present invention is described in further detail below with reference to the attached drawing figures.

Drawings

FIG. 1 is a flow diagram of reverse dispatch scheduling in one embodiment.

FIG. 2 is a diagram of a task structure model in one embodiment.

Fig. 3 is a device distribution diagram in one embodiment.

FIG. 4 is an allocation diagram and a positive scheduling result diagram of each process in one embodiment, where (a) is v₆And v₇The distribution chart of (a), wherein (b) is v₅The distribution chart of (c) is v₂And v₄The distribution diagram of (d) is v₁The distribution chart of (a), (b) is v₃The map (f) is a map of the scheduling result in the positive sequence.

FIG. 5 is a diagram of scheduling results for the tasks shown in FIG. 1 using HEFT in one embodiment.

FIG. 6 is a function diagram of 1/(dis + α) in one embodiment.

FIG. 7 is a diagram illustrating the distribution of σ, F, and F in one embodiment.

FIG. 8 is a graph of the distribution of the maximum α values of V1, V2, V3 and V4 in one embodiment.

FIG. 9 is a graph of the distribution of the maximum σ values of M1, M2, M3, and M4 in one embodiment.

FIG. 10 is a scores distribution plot for each algorithm at Set1-4 in one embodiment.

FIG. 11 is a device occupancy diagram for each algorithm in one embodiment.

FIG. 12 illustrates an execution time scatter plot of 500 tasks in a 6 device set, under an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

Description of distributed manufacturing scheduling problem:

(1) the distributed system task is scheduled in a slicing mode, and the following constraints, namely DAG constraints, are satisfied among the procedures:

1) distributed tasks have a unique termination procedure (DAG egress node);

2) the distributed equipment has only one termination device, namely, the termination process needs to complete processing on the termination device;

3) a certain process is only executed in equipment meeting the requirements of the process, the equipment which can meet the requirements of the same process is not unique, and the execution time of the process in each equipment is known;

4) any equipment can only execute one process at the same time;

5) if v is_jIs v is_iA step immediately after (2), v_jIs required to be at v_iWait v after execution_iTo v_jTransporting; if v is_iAnd v_jWhen the method is executed in the same equipment, the transportation time is 0;

6) each process is executed only when the processes immediately before the process are executed and the raw materials transported by all the previous processes are collected;

7) the equipment is in a discrete distribution state, and the transportation time among the equipment is influenced by the state of the equipment access;

8) the task has a designated termination device, i.e., the termination process must complete the process on the termination device.

(2) The notation of the DAG scheduling model is as follows:

1)V＝{v₁，v₂… … represents a set of steps for a task, where | V | is the number of steps;

2)M＝{m₁，m₂… …, where | M | represents the number of devices in the distributed environment;

3) u denotes the transport time matrix, U_i,jPresentation apparatus m_iTo device m_jIf the equipment m is transported_iTo device m_jOne shortest circuit { m }_i,m_i',m_i”,…,m_j',m_jU, then u_i,j＝u_i,i'+u_i',i”+…+u_j',j；

4) r (v) represents the step immediately before v, if v_i,v_jIs v is_kImmediately before, i.e. r (v)_k)＝{v_i,v_jH, then v_kMust be at v_i,v_jThe processing is finished and the raw materials transported by all the previous processes can be started to be executed when the raw materials are collected;

5) e (v) means assigned to step v;

6) a represents a processing matrix, A_i,jRepresents the step v_iAt the equipment m_jTime required for upper working, A_i,. denotes a process v_iThe processing time on all devices constitutes a vector.

The distributed manufacturing scheduling problem is a type of NP problem that does not have an optimal allocation scheme for a limited time. In contrast, the invention provides a fixed-point output scheduling algorithm for minimizing the processing time, and a reverse allocation adaptation strategy, wherein the scheduling principle takes the subsequent equipment occupation state and the equipment distribution state of the task as guidance when equipment is allocated.

In one embodiment, a fixed point output oriented distributed manufacturing scheduling method is provided, the method comprising the steps of:

step S01, constructing a task structure measurement p model;

step S02, constructing an equipment distribution state measurement q model;

and step S03, based on the p model and the q model, performing reverse distribution scheduling according to the reverse sequence of the process.

The invention provides quantitative indexes for the distribution of the processes and the equipment and comprehensively expresses the rationality of the distribution of the tasks and the equipment at the same time.

The invention utilizes the transportation characteristic of the distributed equipment, considers the transportation relation of the equipment while considering the flexible processing capacity of the equipment. The invention not only considers the structural characteristics of tasks and the processing capacity of equipment, but also considers the transportation relation among the equipment, and in addition, for the problem of specifying an exit node, the invention also considers the constraint of the equipment for terminating the processing. Therefore, the complexity of the problem researched by the invention is higher than that of the traditional algorithm, and the method has better practicability.

Further, in one embodiment, the task structure metric p model in step S01 is:

in the formula, p_iIs a process v_iProcessing time weight vector, foreword (v) based on task structure_i) Represents the step v_iSet of preceding steps of (1), dis (v)_i,v_j) Represents the step v_iAnd process v_jMinimum path distance in the task structure diagram, α is a control parameter, A_j,.Indicating the j-th row of the processing matrix A, i.e. process v_jThe time required for processing on each piece of equipment.

The model can measure the process v_iAnd its subsequent device dependency of the scheduling process.

Here, the principle of constructing the p model is: the task structure measurement aims to estimate the dependency of each process on equipment in the subsequent scheduling process according to the distribution state of the process in the task and the equipment dependency. If the process v is carried out_jIs the process v_iPreceding procedure of (a), in accordance with a reverse allocation adaptation strategy, v_jNeeds to wait for v_iThe dispensing device ends. Therefore, if v_jAnd v_iThe farther away v is_jDevice dependency pair v_iThe less impact the device allocation decision will have.

Illustratively, in conjunction with FIGS. 2 and 3, when v is₈Is distributed to m₇In equipment, step v₆And v₇And the process is changed into a schedulable process. For the process v₆V is a preceding step₂,v₄,v₁,v₃Each preceding step with v₆Are 1,1,2, respectively. Thus, at v₆When making scheduling decisions, v needs to be considered₆,v₂,v₄,v₁,v₃According to the formula of the p model, α is taken to be 1, and the process v is performed₆，p₆＝1×A_6,.+0.5×A_2,.+0.5×A_4,.+0.33×A_1,.+0.33×A_3,.For the step v₇，p₇＝1×A_7,.+0.5×A_5,.+0.33×A_3,.。

Further, in one embodiment, the device distribution state metric q model in step S02 is:

in the formula, q_jIs a device m_jMachining time weighting u based on the distribution of the equipment_k,jIs a device m_kAnd m_jThe transport time between, σ is the distance control parameter, R_kAnd R_jRespectively a process in a device m_kAnd device m_jThe latency of the process, | M | represents the number of devices in the distributed environment.

The model expresses a device m_jThe importance degree of the position relative to the process is known from the q model formula, if m_jThe shorter the transport time with other equipment and m_jThe more devices there are in the vicinity, q_jThe larger the value of (a) is, the higher the weight of the device is.

Here, the principle of construction of the q model is: the equipment distribution state measurement is a measurement of the processing capacity of a local area centered on a specific equipment, which is established based on the processing time of the equipment in the local area to each process and the transportation time between the equipments. When v is shown in FIG. 3₈Is distributed to m₇After the apparatus, step v₆And v₇Becomes a schedulable procedure, at which time v₆And v₇Alternative apparatus is m₆,m₃,m₇Wherein m is₃Is { m }₁,m₅,m₆,m₇}，m₆Is { m }₃,m₄,m₇}，m₇Is { m }₃,m₆}，m₃Than m₆And m₇The transportation environment is good, and richer equipment selection schemes can be provided for subsequent tasks, so that the competition of equipment is reduced.

Further, in one embodiment, with reference to fig. 1, the step S03 is to perform reverse allocation scheduling according to a reverse order of the processes based on the p model and the q model, and the specific process includes:

step S03-1, allocating a termination process to the termination equipment;

step S03-2, acquiring an alternative process set of the allocated processes, wherein the alternative process set comprises processes immediately before the allocated processes;

step S03-3, acquiring an alternative equipment set of alternative processes, specifically: if a certain process v has been assigned to equipment m, i.e. e (v) ═ m, the set of alternative equipment of the immediately preceding process r (v) of v comprises e (v) and equipment immediately adjacent to e (v);

step S03-4, calculating the priority O of each alternative process v and the alternative equipment m thereof_v,mThe formula used is:

in the formula, O_i,jIs a process v_iWith its alternative device m_jPriority of (d), min (p)_i) Representing the vector p_iMinimum value of (1);

step S03-5, selecting the minimum priority O_v,mV and m with corresponding values, and distributing the working procedure v to the equipment m for processing;

and repeating the steps S03-2 to S03-5 until all the procedures are distributed.

In one embodiment, there is provided a fixed point output oriented distributed manufacturing scheduling apparatus, the apparatus comprising:

the first model building module is used for building a task structure measurement p model;

the second model building module is used for building an equipment distribution state measurement q model;

and the distribution scheduling module is used for carrying out reverse distribution scheduling according to the reverse sequence of the working procedures on the basis of the p model and the q model.

Further, in one embodiment, the allocation scheduling module includes:

a first distribution unit for distributing the termination process to the termination device;

an alternative process acquisition unit for acquiring an alternative process set of the assigned processes, the set including a process immediately before the assigned process;

an alternative equipment acquiring unit, configured to acquire an alternative equipment set of an alternative process, specifically: if a certain process v has been assigned to equipment m, i.e. e (v) ═ m, the set of alternative equipment of the immediately preceding process r (v) of v comprises e (v) and equipment immediately adjacent to e (v);

a priority calculating unit for calculating the priority O of each alternative process v and the alternative equipment m_v,mThe formula used is:

in the formula, O_i,jIs a process v_iWith its alternative device m_jPriority of (d), min (p)_i) Representing the vector p_iMinimum value of (1);

a second allocation unit for selecting the minimum priority O_v,mV and m with corresponding values, and distributing the working procedure v to the equipment m for processing;

and continuously operating the alternative process acquisition unit to the second distribution unit until all the processes are distributed.

For specific definition of the fixed-point output-oriented distributed manufacturing scheduling apparatus, reference may be made to the above definition of the fixed-point output-oriented distributed manufacturing scheduling method, and details are not described herein again. The modules in the fixed-point output-oriented distributed manufacturing scheduling apparatus may be implemented in whole or in part by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:

step S01, constructing a task structure measurement p model;

step S02, constructing an equipment distribution state measurement q model;

and step S03, based on the p model and the q model, performing reverse distribution scheduling according to the reverse sequence of the process.

For the specific definition of each step, see the above definition of the fixed point output-oriented distributed manufacturing scheduling method, which is not described herein again.

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:

step S01, constructing a task structure measurement p model;

step S02, constructing an equipment distribution state measurement q model;

and step S03, based on the p model and the q model, performing reverse distribution scheduling according to the reverse sequence of the process.

For the specific definition of each step, see the above definition of the fixed point output-oriented distributed manufacturing scheduling method, which is not described herein again.

As a specific example, in one embodiment, the fixed-point output-oriented distributed manufacturing scheduling method (RSM algorithm) of the present invention is further verified and described:

FIG. 2 is a DAG structure for a task, where a node represents a process, node v₈To terminate a process, directed edges represent the ordered relationship of the process. Table 1 below shows the execution times of the processes in the respective facilities, and lists the execution times of the processes in the respective 8 facilities m_1-8The execution time of. Table 2 below is the inter-equipment transit time matrix U. FIG. 3 shows each apparatus m_1-8The weight value of the distribution state of (2) represents the transportation time of the adjacent equipment, m₇Is a terminating device.

TABLE 1 execution time of the process of FIG. 2 on each plant

TABLE 2 inter-Equipment transport time matrix U

1. Scheduling example of the reverse allocation scheduling algorithm of the invention

The present embodiment takes the task structure of fig. 2 and the device distribution state of fig. 3 as examples, and schedules the tasks shown in fig. 2. The task structure metric p depends only on the structural features of the tasks, and therefore the task structure metric p of the process does not change during the scheduling process. As can be known from the device distribution state metric q model formula, the device distribution state metric q depends on the waiting time R of the device, and when scheduling is performed, the waiting time of the device changes with the change of the scheduling time, so that the q of the device changes in the scheduling process. Table 3 below shows p of each process and q of each time device in the task scheduling process, where the parameter values of p and q are α ═ 1 and σ ═ 5, respectively. The scheduling process shown in table 3 is as follows:

table 3 scheduling process for tasks shown in fig. 2

(1) When t is 0, m is due to the device₇And task v₈The apparatus m is required to be a terminating apparatus and a terminating process, respectively₇Assignment to task v₈；

(2) When t is 8, v₈At the equipment m₇When the upper working is finished, v₈Preceding step v of₆And v₇Becomes an alternative process, m₇And the device m directly adjacent thereto₃,m₆Becoming an alternative device. Therefore, it is necessary to judge v₆,v₇And device m₃,m₆,m₇The priority O. At this time, since there is no processing task on all of the 3 devices, R is₃＝R₆＝R ₇0. In table 3, when t is 8, the priorities are: o is_6,3＝3.79,O_7,3＝1.67,O_6,6＝17.24,O_7,6＝15.41,O_6,7＝7.17,O_7,74.22, O with the smallest O is selected_6,3And O_7,7As a task allocation scheme. v. of₆And v₇The task allocation process of (a) is shown in fig. 4 (a);

(3) when t is 15, v₇At the equipment m₇When the upper working is finished, v₇Preceding step v of₅Becomes an alternative process, m₇And the device m directly adjacent thereto₃,m₆Becoming an alternative device. At this time, the device m, as shown in FIG. 4(b)₃Is under processing, and m is₃T is 16, so R ₃1. Selecting O according to the priority O_5,3As a distribution scheme;

(4) when t is 16, v₆At the equipment m₃When the upper working is finished, v₆Preceding step v of₂,v₄Becomes an alternative process, m₃And the device m directly adjacent thereto₁,m₂,m₅,m₆,m₇Becoming an alternative device. At this time, the device m, as shown in FIG. 4(c)₃Is under processing, and m is₃T 23, so R ₃7. Selecting O according to the priority O_2,1,O_4,2As a distribution scheme;

(5) when t is 19, v₂After finishing the processing on the device m1, v₂Preceding step v of₁Becomes an alternative process, m₁And the device m directly adjacent thereto₃,m₄Becoming an alternative device. At this time, the device m, as shown in FIG. 4(d)₂,m₃Respectively have tasks v₄,v₅Is being processed, and m₂,m₃T 22, t 23, so R₂＝3,R ₃4. Selecting O according to the priority O_1,1As a distribution scheme;

(6) when t is 23, v₅At the equipment m₃Is finished at the upper working, and v₄Has been at device m when t is 22₂At this time, v is finished₃Becomes an alternative process, m₃、m₂And the device m directly adjacent thereto₁,m₅,m₆,m₇Becoming an alternative device. As shown in fig. 4(e), at this time, R of all the devices is 0. Selecting O according to the priority O_3,2As an allocation scheme.

Since the RSM algorithm is a reverse scheduling algorithm that starts from the termination node, performing mirror symmetry on the scheduling result of the RSM algorithm results in the forward scheduling result shown in fig. 4(f), which has a total time of 28 (hours). FIG. 5 is the result of the scheduling of the tasks shown in FIG. 2 using HEFT, with a total time of 35 (hours). Because the HEFT is a forward scheduling algorithm, in order to realize fixed-point output of the HEFT, the invention firstly utilizes rank of the HEFT to carry out reverse scheduling, namely scheduling from a termination node to a starting node. In the rank calculation process, the mean value of the transportation time of the adjacent equipment is taken as an input.

2. The parameter analysis of the invention specifically comprises the following steps:

the invention has 2 parameters alpha and sigma, wherein alpha controls the influence of the subsequent process on the alternative process, and sigma controls the influence of the adjacent processing equipment on the alternative equipment. For this embodiment, the influence of 2 parameters on the scheduling result is analyzed. Fig. 6 is a functional image of 1/(dis + α) in the task structure metric p-model equation, and it can be seen from the α functional image that the larger α is, the smaller the influence of the change in dis on 1/(dis + α), that is, the smaller the influence of the subsequent process on the alternative process.

The following equation is exp (- (f/sigma) in the device distribution state metric q model formula²) And (3) the influence function of the corresponding transportation time on the alternative equipment, wherein F is the transportation time, and F is the distribution proportion of different transportation times F when sigma is fixed. The distribution of F is shown in fig. 7.

As can be seen from the distribution state F in fig. 7, when σ is fixed, the longer the transport time, the smaller its influence on the alternative equipment. When σ is 4, it is close to 0 for the alternative device if the transit time f is greater than 3. However, the transportation time is greatly different for different equipment tasks and equipment transportation networks, such as shown in FIG. 3The mean transit time of the neighboring devices is 4.67, and if σ is 1, the influence of the neighboring devices on the candidate device is close to 0. Therefore, the value of σ should be proportional to the transit time f. In this regard, the present invention takes n × g as a value of σ, where g is a mean value of communication time of neighboring devices, i.e., g — average (U)_i,j|m_iAnd m_jIs a neighboring device) and n is a coefficient.

To verify the value intervals of α and σ, this embodiment randomly generates α and σ test data using a Workflow Generator.

(1) Alpha test data. The test data comprises 4 groups of tasks V1, V2, V3 and V4, each group of tasks comprises 100 tasks, the structure of each task is randomly generated, and the generation parameters of the 4 groups of task data are shown in the following table 4. All 4 groups of tasks are scheduled in the same equipment relation network, and the generation parameters of the equipment relation network are as follows: the number of the devices is 30, the average transport time of the adjacent devices is 10, and the link density between the devices is 0.25;

table 44 set of generation parameters for task data

(2) Sigma test data. The test data comprises 4 groups of tasks M1, M2, M3 and M4, each group of tasks comprises 100 device networks, the structure of each device network is randomly generated, and the generation parameters of the 4 groups of device network data are shown in the following table 5. The 4 groups of equipment networks all schedule the same processing task, and the generation parameters of the processing task are as follows: the number of the processes is 200, the average processing time of each process is 15, and the link density between the processes is 0.2;

table 54 set of process equipment data generating parameters

(3) Alpha test procedure. 4 groups of tasks V1, V2, V3 and V4 are scheduled in the same equipment relation network by using an RSM algorithm, a task G (a task V1 or V2 or V3 or V4) is scheduled for 7 times according to alpha ═ 1,2,3,4,5,6 and 7 respectively when being scheduled, and an alpha value with the optimal scheduling result of the task G is recorded. And counting the alpha value distribution of the optimal scheduling result in 7 times of scheduling of each group of 100 tasks. The result is shown in fig. 8, where each histogram represents the number of times each group of tasks obtains the optimal scheduling result under different values of α. For example, 15 tasks out of 100 tasks of V4 obtain the optimal scheduling result when alpha is less than or equal to 2. Fig. 8 visually shows that the optimal value intervals of α of the 4 groups of tasks are all α ═ 3,4,5 };

(4) sigma test procedure. The same task is scheduled in 4 groups of device networks M1, M2, M3, M4 using RSM algorithm. And scheduling for 30 times according to the sigma of {0.1G,0.2G, … and 3G } during scheduling, and recording the sigma value of the optimal scheduling result of the task G. And carrying out statistics on sigma value distribution of the optimal scheduling result. The result is shown in fig. 9, where each histogram represents the number of times each group of device networks obtains the optimal scheduling result in different σ value intervals. For example, 30 of 100 device networks of M4 obtain the optimal scheduling result within the interval of 1.5g < sigma ≦ 2 g. Fig. 9 visually shows that the optimal value intervals of σ of the 4 groups of tasks are all 0.5g < σ ≦ 2 g.

3. Experimental analysis:

(1) experimental Environment

To verify the effectiveness of the CESM algorithm, the experimental part of this embodiment generates experimental data by using Workflow Generator to simulate the DAG task structure of the distributed system. Hardware: intel i3 processor, 4G memory. In the aspect of comparison algorithm, the present embodiment uses the HEFT, CEFT, DCP algorithm in the scheduling problem as the comparison algorithm. Since each algorithm does not consider the problem of fixed-point output, the following algorithms need to be improved for fixed-point output:

1) the improved method of the HEFT algorithm is to use rank of the HEFT to carry out reverse scheduling, namely scheduling from a termination node to a start node. In the rank calculation process, the average value of the transportation time of adjacent equipment is used as transportation as input, and rank is used as the priority of process selection;

2) the CEFT algorithm is improved by reversely calculating the length of a key path from a termination node and reversely splitting a task result by using the length of the key path;

3) the DCP algorithm selects the priority of the device by using the sum of the latest start time and the earliest end time of the process, and performs reverse scheduling from the end process of the task.

(2) Scheduling total time analysis

This embodiment divides the data into 4 data sets of Set1, Set2, Set3 and Set4, each data Set containing 100 'task-device pairs'. The generation parameters for each data set are as follows:

table 6 task-device generation parameter table

Since the total time difference of each task-device pair' is large, the present embodiment will utilize the score model as follows to normalize the total time of each algorithm:

in the formula, time_i(n) denotes the processing time of algorithm i on the nth 'task-device pair', max { time (n) } denotes the maximum value of the processing time of each algorithm on the nth 'task-device pair', score_iRepresents the cumulative score of algorithm i over 100 'task-device pairs', with lower score indicating that the processing time of the algorithm is relatively smaller. For example, the total scheduling time of the 'task-device pair' by RSM, HEFT, CEFT, DCP is: 20, 25,15, 10, the scores of the algorithms for the 'task-device pair' are respectively: 20/25,25/25,15/25,10/25, i.e. 0.8,1,0.6, 0.2. Fig. 10 is the scores distribution of the algorithm on Set1-4, where the score value of RSM is lowest relative to the other 3 algorithms, indicating that the overall scheduling time of RSM is lowest.

(3) Device dependency analysis

Device dependency is the selection tendency of an algorithm to different devices. In order to analyze the device dependency of each algorithm, the present embodiment uses the device occupancy as a metric index, and the expression of the occupancy (occupancy rate) is as follows:

in the formula, o (m)_j) Presentation apparatus m_jThe time occupied, | M | is the total number of devices. If the occupancy rate of a certain algorithm on a certain type of equipment is higher, the dependency of the algorithm on the equipment is stronger. The better the load balancing performance of the algorithm, the more uniform the occupancy rate of the device is. In this embodiment, 5 kinds of machines 1-5 are designed, each Machine has 6 units, and 30 units in total, and the average time of each equipment processing procedure is as follows: 5,10, 15, 20, 25. With the average transit time of the adjacent devices being 10 and the device network density being 0.25, 100 networks containing the 30 devices were randomly generated. With the task link density of 0.3 and the number of processes of 100 as parameters, 100 tasks are randomly generated. And respectively scheduling 100 tasks on 100 device networks by using RSM, HEFT, CEFT and DCP, and calculating the average occupancy rates of 5 devices. Fig. 11 is a device occupancy histogram of the results of 4 algorithm runs. From the comparison of fig. 11, the occupancy rates of the HEFT, CEFT, and DCP algorithms in the Machine1-3 are significantly higher than those of other devices, and the occupancy rates from the Machine1 to the Machine5 are in a downward trend, which indicates that the HEFT, PCH, and HHDS algorithms all depend on high-efficiency devices. The utilization rate of the RSM algorithm on 5 devices is relatively uniform, which shows that the RSM algorithm has better device load balancing performance, and can fully utilize inefficient devices, thereby reducing the total execution time.

(4) Device balance analysis

The present embodiment analyzes the execution efficiency of the RSM algorithm for different device sets. The experimental process is as follows: 1) generation data D500: 500 tasks, wherein each task comprises 150 processes, the average processing time of the processes is 10, and the task link density is 0.2; 2) the equipment in the equipment group was divided into 3 types (average processing time of high efficiency type equipment was 10, average processing time of general type equipment was 20, average processing time of low efficiency type equipment was 30), density of equipment network was 0.25, and average transportation time of adjacent equipment was 10. The 6 device sets designed for this experiment are shown in table 7 below, where each device set contains 9 devices. For example, the device Set1 includes: 2 high-efficiency devices, 4 general devices and 3 low-efficiency devices; 3) and scheduling each group of tasks by using 4 algorithms, and recording the execution time distribution of the D500 tasks in each group of data in 6 equipment sets. Fig. 12 is a scatter diagram of 500 task execution time distributions of 4 algorithms in 6 device sets, in which the work efficiency gradually decreases from MSet1 to MSet6 device sets, and the task execution time distributions of 4 algorithms gradually increase. The RSM in FIG. 12 performs significantly better in the set of MSet1 and MSet2 devices than the other 3 algorithms, while the RSM performs close to the other 3 algorithms in MSet5 and MSet6, because: the distribution of 3 devices in MSet1 and MSet2 is more even, at which time the advantage of RSM algorithm considering balanced load is obvious, while in MSet5 and Set6 the device types tend to be consistent, at which time the RSM algorithm behaves close to the other 3 algorithms.

TABLE 76 device sets

Therefore, the effectiveness and feasibility of the algorithm are fully verified.

The invention designs a reverse allocation scheduling algorithm RSM aiming at the fixed-point output problem of tasks by utilizing the distribution state and the transportation relation of equipment in a distributed manufacturing system, wherein the algorithm adopts a reverse scheduling process and starts from equipment termination and procedure termination to allocate equipment for the procedure reverse. In the scheduling process, task structure measurement and equipment distribution state measurement are automatically established according to the task structure and the distribution state of the equipment network, and a basis is provided for task scheduling. Compared with HEF, CEFT and DCP algorithms, the algorithm simplifies the scheduling process, reduces the overall scheduling time and improves the scheduling efficiency of distributed manufacturing. Compared with the similar reverse scheduling algorithm, the RSM algorithm has better equipment balance load, can fully utilize low-efficiency equipment in the scheduling process, reduces the total execution time and improves the workshop operation efficiency. In addition, the RSM algorithm is optimized from two aspects of tasks and equipment structures, fixed-point output constraint of the tasks can be better guaranteed, scheduling procedures are reduced, and distributed manufacturing efficiency is improved to the maximum extent.

The foregoing illustrates and describes the principles, general features, and advantages of the present invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.

Claims (6)

1. A fixed point output-oriented distributed manufacturing scheduling method is characterized by comprising the following steps:

step 1, constructing a task structure measurement p model;

step 2, constructing an equipment distribution state measurement q model;

and 3, based on the p model and the q model, performing reverse distribution scheduling according to the reverse sequence of the process.

2. The fixed-point output-oriented distributed manufacturing scheduling method according to claim 1, wherein the task structure metric p model in step 1 is:

in the formula, p_iIs a process v_iProcessing time weight vector, foreword (v) based on task structure_i) Represents the step v_iSet of preceding steps of (1), dis (v)_i,v_j) Represents the step v_iAnd process v_jMinimum path distance in the task structure diagram, α is a control parameter, A_j,.Indicating the j-th row of the processing matrix A, i.e. process v_jThe time required for processing on each piece of equipment.

3. The fixed-point output-oriented distributed manufacturing scheduling method of claim 1, wherein the device distribution state metric q model in step 2 is:

in the formula, q_jIs a device m_jMachining time weighting u based on the distribution of the equipment_k,jIs a device m_kAnd m_jThe transport time between, σ is the distance control parameter, R_kAnd R_jRespectively a process in a device m_kAnd device m_jThe latency of the process, | M | represents the number of devices in the distributed environment.

4. The fixed-point output-oriented distributed manufacturing scheduling method according to claim 2 or 3, wherein the step 3 is to perform reverse distribution scheduling in a reverse order of the process steps based on the p model and the q model, and the specific process includes:

step 3-1, distributing a termination procedure to termination equipment;

step 3-2, acquiring an alternative process set of the allocated processes, wherein the alternative process set comprises processes immediately before the allocated processes;

step 3-3, acquiring an alternative equipment set of alternative processes, specifically: if a certain process v has been assigned to equipment m, i.e. e (v) ═ m, the set of alternative equipment of the immediately preceding process r (v) of v comprises e (v) and equipment immediately adjacent to e (v);

3-4, calculating the priority O of each alternative procedure v and the alternative equipment m thereof_v,mThe formula used is:

in the formula, O_i,jIs a process v_iWith its alternative device m_jPriority of (d), min (p)_i) Representing the vector p_iMinimum value of (1);

step 3-5, selecting the minimum priority O_v,mV and m with corresponding values, and distributing the working procedure v to the equipment m for processing;

and repeating the steps 3-2 to 3-5 until all the working procedures are distributed.

5. The scheduling apparatus of the fixed-point output-oriented distributed manufacturing scheduling method according to any one of claims 1 to 4, wherein the apparatus comprises:

the first model building module is used for building a task structure measurement p model;

the second model building module is used for building an equipment distribution state measurement q model;

and the distribution scheduling module is used for carrying out reverse distribution scheduling according to the reverse sequence of the working procedures on the basis of the p model and the q model.

6. The scheduling apparatus of the fixed-point output-oriented distributed manufacturing scheduling method according to claim 5, wherein the allocation scheduling module comprises sequentially executing:

a first distribution unit for distributing the termination process to the termination device;

an alternative process acquisition unit for acquiring an alternative process set of the assigned processes, the set including a process immediately before the assigned process;

an alternative equipment acquiring unit, configured to acquire an alternative equipment set of an alternative process, specifically: if a certain process v has been assigned to equipment m, i.e. e (v) ═ m, the set of alternative equipment of the immediately preceding process r (v) of v comprises e (v) and equipment immediately adjacent to e (v);

a priority calculating unit for calculating the priority O of each alternative process v and the alternative equipment m_v,mThe formula used is:

in the formula, O_i,jIs a process v_iWith its alternative device m_jPriority of (d), min (p)_i) Representing the vector p_iMinimum value of (1);

a second allocation unit for selecting the minimum priority O_v,mV and m corresponding to the value willThe working procedure v is distributed to equipment m for processing;

and continuously operating the alternative process acquisition unit to the second distribution unit until all the processes are distributed.

Images