CN113642808B - A dynamic scheduling method for cloud manufacturing resource changes - Google Patents

Abstract

The invention provides a dynamic scheduling method for cloud manufacturing resource changes, and belongs to the technical field of advanced manufacturing technology. It is characterized in that the method firstly analyzes the resource changes in the cloud manufacturing environment by combining the needs of the cloud service platform with the interference of resource attribute changes, new resource access, and resource withdrawal that may occur during the resource scheduling process, and constructs a oriented The dynamic scheduling model of cloud manufacturing resource changes. Finally, the improved particle swarm optimization algorithm is used to solve the dynamic scheduling model of resources, so as to obtain the resource scheduling scheme and real-time resource scheduling update strategy in the cloud manufacturing environment that can meet the conditions. The present invention is widely applied to processing and manufacturing service enterprises under the cloud manufacturing platform, realizes the scheduling and optimization of dynamic resources in the cloud manufacturing environment, and improves the problems of high cost and unbalanced resource load of manufacturing enterprises.

Description

A dynamic scheduling method for cloud manufacturing resource changes

technical field

The present invention relates to a dynamic scheduling method oriented to changes in cloud manufacturing resources. The scheduling method first integrates the needs of the resource requester and the resource provider to upload to the cloud service platform together with the resource attribute changes that may occur in the process of resource dynamic scheduling. The interference of resource access, resource withdrawal and resource maintenance, analyze the changes in resource scheduling in the cloud manufacturing environment, and then construct a dynamic scheduling oriented to cloud manufacturing resource changes based on the attributes and characteristics of manufacturing tasks and resources as well as the actual resource scheduling rules Finally, the improved particle swarm optimization algorithm is used to solve the resource dynamic scheduling model to obtain the resource scheduling scheme and real-time resource scheduling update strategy in the cloud manufacturing environment that can meet the conditions. The invention belongs to the field of advanced manufacturing technology.

Background technique

At present, the production mode of the manufacturing industry in the world is shifting to intelligent manufacturing on a large scale, and with the rapid development of cloud computing, manufacturing grid, Internet and other emerging IT technologies, the production and development of enterprises are increasingly connected with the network. In this context, the concept of cloud manufacturing came into being. Cloud manufacturing is a new service-oriented, high-energy and low-consumption networked manufacturing model. It integrates IT technologies such as cloud computing, Internet of Things, big data, and semantic Web. Features of information sharing and low loss. As a networked manufacturing model, cloud manufacturing can virtualize various scattered manufacturing resources and manufacturing capabilities, form a resource pool after service, and conduct centralized management and operation through the network, which can improve resource utilization and reduce enterprise costs. production cost. In the cloud manufacturing environment, the system has a large number of highly dynamic manufacturing resources. How to dynamically schedule resources with high efficiency and high quality is an urgent problem to be solved at present. The current traditional scheduling method has been difficult to meet the needs of dynamic scheduling of cloud manufacturing resources. At this time, a new more efficient and stable scheduling method is needed. appears extremely important.

Contents of the invention

The purpose of the present invention is to provide a dynamic scheduling method that can solve the above problems and efficiently and stably provide sufficient resources for resource demanders in a cloud manufacturing environment. This method can effectively improve the flexibility of the cloud manufacturing service platform, and can guarantee the lowest service cost, the smallest service completion time, the highest resource reliability and the best resource overall service efficiency under the premise of effectively completing tasks.

Technical scheme of the present invention is:

The present invention first integrates the needs of resource demanders and resource providers to upload to the cloud service platform together with the resource attribute changes, new resource access, resource withdrawal, and resource maintenance that may occur during the dynamic resource scheduling process. The changes in resource scheduling in the manufacturing environment are analyzed, and then a dynamic scheduling model for cloud manufacturing resource changes is constructed according to the attributes and characteristics of manufacturing tasks and resources as well as actual resource scheduling rules. Finally, the improved particle swarm optimization algorithm is used to optimize the dynamic scheduling model of resources Solve to obtain the resource scheduling scheme and real-time resource scheduling update strategy in the cloud manufacturing environment that can meet the conditions. The implementation steps of the method include:

(1) Service composition and dynamic analysis of cloud manufacturing resources

In the cloud manufacturing environment, there are many kinds of manufacturing resources with different functions stored in the cloud platform in the form of data, and their status attributes change dynamically in real time with the service process. The cloud manufacturing system can classify cloud manufacturing resources according to function or granularity through resource classification technology. According to the different resource requirements of different subtasks in the cloud platform, different types of resources can be combined to complete manufacturing tasks through resource matching technology.

(2) Dynamic scheduling model based on changes in cloud manufacturing resources

Based on the distribution, diversity, dynamics, and sampling characteristics of resources in the cloud manufacturing environment, combined with the results of the dynamic attribute analysis of resources in the cloud manufacturing environment, the related concepts of resource dynamic scheduling constraints in the cloud manufacturing environment are given and constructed. A multi-objective optimization model with total manufacturing service time, total manufacturing service cost, average reliability of manufacturing resources, and average service efficiency of manufacturing resources as optimization objectives is established.

(3) Resource dynamic scheduling optimization solution based on improved particle swarm optimization algorithm

On the basis of the established dynamic scheduling model for cloud manufacturing resource changes, the changes of various resources in scheduling are coded to complete the formal description of resource scheduling problems. The AHP is used to calculate the service demand weight in the scheduling scheme, and the resource scheduling preference for different services is determined. Finally, the resource scheduling problem is solved by the particle swarm optimization algorithm.

As a preference, the resources in the above step 1 cloud manufacturing environment can be divided into broad-sense resources and narrow-sense resources from the perspective of definition. The equipment that participates in the entire manufacturing process is called narrow-sense resources. The sum of all kinds of resources involved is called generalized resources. From the perspective of resource granularity, it can be divided into single-function resources and composite-function resources. Single-function resources have a single and efficient capability, and composite-function resources have multiple functions at the same time, with strong comprehensive service capabilities. All resources are stored in the cloud manufacturing service platform in the form of data.

The dynamic scheduling of resources in the cloud manufacturing environment can regard the manufacturing process as a dynamic process. There will be many different disturbance factors in the whole cycle of resource scheduling. This disturbance is called the dynamic nature of resources. The dynamics of resources in the cloud manufacturing environment are mainly reflected in the following perspectives:

(1) Changes in resource attributes. Resources in the cloud manufacturing environment will change their attributes as they are used and evaluated by users. At time t ₁ of the cloud manufacturing service manufacturing cycle, the initial resource set E ¹ ={E ₁ ,E ₂ ...E _i ...E _l } comprehensive service capabilities in the cloud manufacturing system because the initial task set U ₁ ={F ₁ ,F ₂ ...F _i ...F _n } The user's evaluation has changed, E ¹ is updated to E ² , Serving a new set of manufacturing tasks />

(2) New resource access. At time t ₂ of the resource scheduling life cycle, new resources such as E _l+1 , E _l+2 ... E _l+c are connected to the cloud manufacturing service platform, and the resource set Form a new resource set /> A new manufacturing task set formed to serve the unfinished tasks at time _t2 />

(3) Resource maintenance. At time _t3 of the cloud manufacturing service cycle, resources such as E _b , E _b+1 ... E _b+q in the cloud manufacturing platform fail and need to be repaired, and the cloud manufacturing service is withdrawn because the maintenance The required time is different, so the time for maintenance resources to re-connect to the platform is also different. The resource set in the cloud manufacturing platform is updated from E ³ to E ⁴ , and the cloud manufacturing service platform compares the unfinished tasks at time t ₃ according to the updated resource set E ⁴ A new set of manufacturing tasks composed of tasks Continue to serve.

(4) The resource is withdrawn, and the execution process of the manufacturing task advances to time _t4 . At this time, resources such as E _p , E _p+1 , and E _p+w in the cloud manufacturing platform fail and withdraw from the cloud manufacturing service. The cloud manufacturing service The resource set in the platform is updated from E ⁴ to E ⁵ , and the revoked resources will no longer participate in the scheduling service, and the cloud manufacturing service platform will perform the unfinished tasks of the fourth stage task set U ₄ at time t ₄ according to the updated resource E ⁵ A new set of manufacturing tasks Continue to serve.

In the cloud manufacturing environment, due to the existence of various disturbance factors, the scheduling of resources will present certain dynamics. In order to make the process of dynamic scheduling of cloud manufacturing resources go smoothly, the relevant scheduling rules are formulated as follows:

(1) Each manufacturing resource can independently complete one or several secondary manufacturing tasks, and the same secondary manufacturing task can only be processed by the same manufacturing resource at the same time;

(2) The first-level manufacturing task is decomposed to form a second-level manufacturing task, and the second-level manufacturing task is called the smallest unit to be served;

(3) The processing of the secondary manufacturing task and the next secondary manufacturing task between different manufacturing resources will generate certain logistics costs and logistics time;

(4) The service method of the task is to accept the service in an orderly manner according to the order of the secondary manufacturing tasks. The start time of the latter secondary manufacturing task must be greater than the end time of the previous secondary manufacturing task and the interval between the resources selected for the two tasks. The sum of logistics time;

(5) The total cost of logistics between different resources is proportional to the logistics time and logistics distance;

(6) Different types of primary manufacturing tasks have the same execution priority;

(7) All cloud manufacturing resources can be used at zero time;

(8) Once the secondary task starts processing until its manufacturing service is completed, the period cannot be interrupted.

Preferably, in the above step 2, the basic goal of resource dynamic scheduling is determined, the concept of resource dynamic scheduling constraints in the cloud manufacturing environment is given, a multi-objective optimal scheduling model is established, and the resource scheduling process is formally expressed. Cloud manufacturing The resource dynamic scheduling constraints in the environment mainly include the following contents:

(1) The total manufacturing service time constraint, the execution time of each task cannot exceed the maximum completion time specified by the user, that is: T=Max(T _{ij_end} )≤T _max (i=1,2...n), T represents the manufacturing task The total time, T _max represents the latest delivery date specified by the user.

(2) The total manufacturing service cost constraint, the total manufacturing service cost cannot exceed the user's total budget, namely: where C _max represents the total budgeted cost of the user.

(3) resource average reliability constraint,

(4) Constraints on the average service efficiency of manufacturing resources,

(5) The processing time sequence of the secondary manufacturing task is constrained. The service method of the task is to accept the service according to the order of the secondary manufacturing task. The service start time of the subsequent secondary manufacturing task cannot be less than the sum of the end time of the preceding secondary manufacturing task and the logistics time:

in, Indicates the completion time of the secondary manufacturing task F _ij , /> Indicates the logistics time between the secondary manufacturing task F _ij and the resource where F _i(j+1) is located, /> Indicates the task start time of the secondary manufacturing task F _i(j+1) .

(6) The granularity constraints of subtasks. The secondary manufacturing task formed after the task is decomposed is the smallest unit granularity of processing. Each manufacturing resource can independently complete one or several types of secondary manufacturing tasks. The same secondary manufacturing task Tasks cannot be processed by different manufacturing resources at the same time, namely:

This resource scheduling method selects the total manufacturing service time of manufacturing tasks, the total manufacturing service cost, the average service efficiency of manufacturing resources, and the reliability of manufacturing resources as the optimization objectives of the resource dynamic scheduling model. The main contents are:

(1) The total manufacturing service time refers to the total time from the start of the manufacturing service to the completion of the last secondary manufacturing task;

(2) The total manufacturing service cost includes the manufacturing processing cost of manufacturing the secondary manufacturing task and the logistics cost between two adjacent secondary manufacturing tasks between different service resources. The manufacturing resources provided by the cloud manufacturing service platform to users should meet the goal of the least total manufacturing service cost.

(3) In the cloud manufacturing environment, various factors including the comprehensive strength of resource providers, equipment service capabilities, and computing capabilities will affect the service efficiency of resources. At the same time, different resources will have different efficiencies when serving the same task. .

(4) The reliability of resources in the cloud manufacturing environment, as a resource's own attribute, depends on the user's comprehensive evaluation of the service quality and service time reliability of resources. As resources continue to participate in manufacturing, users continue to evaluate the reliability of resources. dynamic updates.

As a preference, the multi-objective function of dynamic scheduling based on resource changes in the cloud manufacturing environment is:

Among them, f ₁ (x) represents the objective function of total manufacturing service time, f ₂ (x) represents the objective function of total manufacturing service cost, f ₃ (x) represents the objective function of average reliability of manufacturing resources, and f ₄ (x) represents the objective function of manufacturing resources Average service efficiency objective function.

Preferably, in the above step 3, the changes of various resources in the scheduling are coded to complete the formalized description of the resource scheduling problem. The AHP is used to calculate the service demand weight in the scheduling scheme, to determine the resource scheduling preference for different services, and finally to solve the resource scheduling problem by improving the particle swarm optimization algorithm.

Preferably, the above-mentioned resource scheduling in the cloud manufacturing environment includes a resource layer, a primary manufacturing task layer and a secondary manufacturing task layer, wherein the resource layer is an optional resource set of the secondary manufacturing task layer. In addition, during the use of cloud manufacturing resources, interference events such as resource attribute changes, new resource access, resource withdrawal, and resource maintenance will also affect the process of resource scheduling. This method uses the improved particle swarm optimization algorithm to solve the scheduling problem of dynamic resources, and mainly includes the following steps:

(1) Coding. This method initializes a group of particles within the range of feasible solutions according to the interference events that may occur in the resource scheduling problem and the specific solutions to the corresponding interference events. Each particle represents the resource dynamic scheduling problem. a potentially optimal solution.

(2) The construction of the fitness function. This method uses linear weighting to transform the multi-objective optimization problem into a single objective function for solution. When constructing the fitness function, it is necessary to satisfy the minimum value of manufacturing time and manufacturing service cost, and the average manufacturing service cost Efficiency and average resource reliability take the maximum value.

(3) Iterative operation, the iterative operation mainly includes calculating the fitness value of each particle in the population, updating the individual optimal particle and the global optimal particle, and updating the speed and position of the particle according to the speed and position update formula.

Compared with the existing methods, the method of the present invention has the advantages of: targeting at the dynamic changes of resources in the cloud manufacturing environment, it can deal with the possible occurrence of cloud manufacturing resources during the scheduling process, such as resource attribute changes, resource damage, maintenance, and resource revocation The problem is analyzed and the scheduling strategy of the service system after the cloud manufacturing resource disturbance event is given to ensure the smooth completion of the cloud manufacturing task. Combined with the main factors in the scheduling problem, such as manufacturing service time, manufacturing service cost, resource average reliability and resource average service efficiency, a multi-objective optimization model was established, and the improved particle swarm optimization algorithm was used to optimize and solve the problem, effectively Improve the flexibility of the cloud manufacturing service platform. Under the premise of effectively completing tasks, it can guarantee the lowest service cost, the shortest service completion time, the highest average resource reliability and the best resource overall service efficiency.

Description of drawings

Fig. 1 is a relationship diagram of resource changes in the present invention

Fig. 2 is the whole-stage flow chart of the resource scheduling of the improved particle swarm optimization algorithm based on resource changes in the present invention

Detailed ways

The present invention will be further described below in conjunction with accompanying drawing.

Embodiment 1: a kind of dynamic scheduling method for cloud manufacturing resource change, as shown in Figure 1, the basic content of resource dynamic scheduling process includes:

(1) The service composition and dynamic analysis of cloud manufacturing resources. In the cloud manufacturing environment, the state attributes of cloud manufacturing resources change dynamically in real time with the service process. The cloud platform needs to adjust the scheduling strategy in real time according to the changes of resource status attributes.

(2) Based on the dynamic scheduling model of cloud manufacturing resource changes, combined with the research results of resource dynamic attributes in cloud manufacturing environment, the related concepts of resource dynamic scheduling constraints in cloud manufacturing environment are given, and a multi-objective optimal scheduling model is constructed.

(3) Resource dynamic scheduling optimization solution based on the improved particle swarm optimization algorithm. On the basis of the above dynamic scheduling model, use the AHP to calculate the service demand weight in the scheduling plan, determine the resource scheduling preference for different services, and finally pass the particle swarm The algorithm implements the solution to the resource scheduling problem.

As shown in Figure 2, it is the whole-stage process of resource scheduling based on the improved particle swarm optimization algorithm. The specific optimization scheduling solution steps are as follows:

(1) Initialize the particle population, and set the initial position, initial velocity, and number of iterations of the population particles.

(2) Calculate the minimum value T _min of the total manufacturing service time when considering the manufacturing time alone, the minimum value C _min of the manufacturing service cost when considering the total manufacturing service cost alone, and the average reliability when considering the average reliability of manufacturing resources separately The maximum value Rel _max of the resource service efficiency is the maximum value E _max of the resource service efficiency when the average service efficiency of the resource is considered separately, and then each individual target is normalized to obtain the normalized fitness function, and finally the particle in the population is calculated fitness value.

(3) Update the individual optimal particle and the global optimal particle.

(4) Update the particle's position and velocity according to the position formula and velocity formula.

(5) Judging whether the maximum number of iterations has been reached, and if the maximum number of iterations has been reached, return the tasks that have been completed at this moment and their corresponding resource numbers. If the maximum number of iterations is not reached, return to step 2.

(6) Decode the optimal individual and convert it into a scheduling result.

(7) Save the manufacturing task set that has not started at the time t=t ₁ in step 6.

(8) E ₁ , E ₂ ···E _i and other resource attributes are changed and updated.

(9) Perform data processing on the new manufacturing task set and the updated resource, and then operate according to step 1, step 2, step 3, step 4, and step 5.

(10) Decode the optimal individual and convert it into a scheduling result;

(11) Save the manufacturing task set that has not started at the time t=t ₂ in step 10.

(12) New resources such as E ₁₊₁ , E ₁₊₂ ···E _1+c are connected to the cloud platform.

(13) Perform data processing on the new manufacturing task set and resources saved in the previous step, and then operate according to step 1, step 2, step 3, step 4, and step 5.

(14) Decode the optimal individual and convert it into a scheduling result;

(15) Save the manufacturing task set that has not started at the time t= _t3 in step 14.

(16) E _b , E _b+1 ··· E _b+q and other resources are maintained.

(17) Perform data processing on the new manufacturing task set and resources saved in the previous step, and then operate according to step 1, step 2, step 3, step 4, and step 5.

(18) Decode the optimal individual and convert it into a scheduling result;

(19) Save the manufacturing task set that has not started at time t=4 in step 18.

(20) Resources such as E _p , E _p+1 ··· E _p+w are withdrawn from the cloud platform.

(21) Perform data processing on the new manufacturing task set and resources saved in the previous step, and then operate according to step 1, step 2, step 3, step 4, and step 5.

(22) Decode the optimal individual and convert it into a scheduling result;

(23) Store the completed manufacturing tasks and their corresponding resource numbers, and output the global scheduling scheme.

In this paper, taking a cloud manufacturing resource scheduling scenario as an example, based on the analytic hierarchy process, the service demand weight in the scheduling scheme is solved by the following steps:

(1) Construct a hierarchical model, based on the principle of solving the weight of the AHP, combined with the research object of this method to construct the optimal resource scheduling plan as the target layer, the total manufacturing service time, total manufacturing service cost, average manufacturing service efficiency, resource The average reliability is a step-level model of the index factor layer.

(2) This method is based on the basic principles and solving steps of the AHP, and uses the AHP to solve the weights of the influencing factors in the scheduling model. In this example it is assumed that total manufacturing service cost is equally important as total manufacturing service time, total manufacturing service time is slightly more important than average manufacturing service efficiency, both are slightly more important than resource average reliability, and average manufacturing service efficiency is equally important as resource average service efficiency , construct the initial judgment matrix A=(a _ij ) _n×n as follows.

(3) Calculate the weight

Normalize the initial judgment matrix A to get the standard matrix as follows,

right The elements in are added by row and normalized to get the final weight vector />

W=[0.333,0.333,0.167,0.167] ^T

Calculate the maximum eigenvalue λ _max of the judgment matrix A,

λ _max =4

Perform a consistency check on the weights sought,

According to the table below, when n=4, the average random consistency index RI=0.90,

CR＝0.0237≤0.1

To sum up, the required weight conforms to the consistency test, and the required weight in this example is:

W=[0.333,0.333,0.167,0.167] ^T

Based on the dynamic scheduling of resources based on the improved particle swarm optimization algorithm, the specific operation steps of the scheduling algorithm are as follows:

(1) Coding

The invention adopts real number encoding and integer encoding for particles, linear weighting mode is adopted for fitness function, and the calculation of each index weight value is solved by analytic hierarchy process, so that tasks and resources to be served constitute the initial solution space to solve the scheduling problem. For example, the list of task types and secondary manufacturing task types of manufacturing task set U is:

That is, when there are n first-level manufacturing tasks, each first-level manufacturing task F _i can be decomposed into k _i second-level manufacturing tasks F _ij . The integers between are assigned, which can be expressed as [1 11 1 1 2 2 2 2 2]. The second layer of particle encoding represents the second-level task execution sequence layer. Initially, it can be assigned a random real number between [0,1] Completed, it can be expressed as [0.3 0.4 0.5 0.2 0.6 0.7 0.42 0.53 0.62 0.27], the third layer of coding indicates the processing resource number layer assigned to each secondary manufacturing task, which can be expressed as [1 3 5 2 4 3 1 2 4 3]; the final particle code is expressed as:

(2) Construction of fitness function

According to the definition of the objective function, that is, when the manufacturing time and manufacturing service cost are minimized, the average manufacturing service efficiency and average resource reliability are maximized. For the improved particle swarm optimization algorithm, a single objective function is used as the fitness function to evaluate and select particles , so it is necessary to transform the multi-objective optimization problem into the form of a single objective function for solution.

Among them, W ₁ represents the total manufacturing service time, W ₂ represents the total manufacturing service cost, W ₃ represents the average service efficiency of manufacturing resources, W ₄ represents the weight value of the average reliability of manufacturing resources; T _min represents the manufacturing The minimum value of service time, C _min represents the minimum value of service cost when only considering manufacturing service cost, E _max represents the maximum value of manufacturing service efficiency when only considering average manufacturing service efficiency, Rel _max represents when only considering average resource reliability The maximum value of the time-average resource reliability.

(3) Particle position and velocity update

Update the particle speed and position by recording the particle individual optimal value and population optimal value, and at the same time complete the update of the task execution sequence layer and resource layer, and generate new task execution by updating the task execution sequence layer and resource layer sequence, and calculate the fitness function value corresponding to the new task execution sequence. The update operation of particle velocity and position is carried out according to the following formula, where ω represents the inertia weight.

(4) Improvement of particle swarm algorithm

The present invention uses linear weights instead of inertial weights to eliminate the problem that a larger inertial weight value leads to excessive particle velocity and directly passes through the optimal solution position after the next particle update and a smaller inertial weight value leads to excessive particle velocity. Computational failures are caused by the problem of small particles getting trapped in a local search space. The inertia weight ω varies linearly according to the formula.

Among them, Max-ITER represents the maximum number of iterations, ω _max and ω _min represent the maximum and minimum values of the inertia weight ω respectively, and f represents the current number of iterations.

Through iterative operation, the scheduling scheme (excerpt) of the manufacturing task set U can be finally obtained.

Claims (3)

1. A dynamic scheduling method for cloud manufacturing resource change is characterized in that: the method comprises the following steps:

step 1, comprehensively analyzing resource scheduling demand conditions in a cloud manufacturing environment by a cloud manufacturing platform according to manufacturing service demands uploaded by a resource demand party and a resource provider;

step 2, constructing a dynamic scheduling model oriented to cloud manufacturing resource change according to the attribute characteristics of manufacturing resources and by combining with an actual resource scheduling rule;

step 3, optimizing and solving a dynamic scheduling process of the resources by applying an improved particle swarm algorithm to obtain a resource scheduling scheme and a real-time scheduling update strategy under the cloud manufacturing environment meeting the conditions;

the application of the improved particle swarm algorithm in the step 3 to dynamically schedule and optimize resources in a cloud manufacturing environment comprises the following steps:

(1) Determining a coding mode of a resource scheduling problem, and realizing formal description of multi-level tasks and resources in resource scheduling;

(2) Calculating the fitness value of each particle in the population according to the particle swarm algorithm flow, and updating the individual optimal particles and the global optimal particles;

(3) Updating the speed and the position of the particles according to a speed and position updating formula;

(4) Executing a loop condition, judging whether the algorithm reaches preset iteration times, if so, ending the algorithm and outputting the global optimal particles and fitness values thereof at the moment, and if not, continuing to execute the loop;

the particle swarm algorithm comprises the following steps:

(1) Setting various parameters in an algorithm, and randomly initializing the initial position and the speed of the particles;

(2) Calculating the fitness value of each particle;

(3) Comparing the fitness value of the current position of the particle with the fitness value of the optimal position of the particle in the searching process, reserving the optimal fitness value as the historical optimal fitness value of the individual particle, and simultaneously updating the historical optimal position by using the current position of the particle;

(4) Comparing the size of the historical optimal fitness value of the individual particles with the fitness value of the particle population at the optimal position, and reserving the larger fitness value as the current global optimal fitness value, wherein the current position is used as the optimal position;

(5) Updating the speed and the position of the particles according to a speed and position updating formula;

(6) And (3) recalculating the particle fitness value after the position and speed updating, and if the particle fitness value fails to reach the termination condition, continuing the operation of the step (2).

2. The cloud manufacturing resource variation oriented dynamic scheduling method as described in claim 1, wherein: the method comprises the following scheduling rules:

(1) Each manufacturing resource can independently complete one or more secondary manufacturing tasks, and the same secondary manufacturing task can only be processed by the same manufacturing resource at the same time;

(2) The first-level manufacturing task is decomposed to form a second-level manufacturing task, and the second-level manufacturing task is called a minimum unit to be serviced;

(3) Processing between different manufacturing resources for a secondary manufacturing task and a next secondary manufacturing task can create logistic costs and logistic time;

(4) The service mode of the task orderly receives service according to the front-to-back sequence of the secondary manufacturing task, and the starting time of the latter secondary manufacturing task is required to be larger than the sum of the ending time of the former secondary manufacturing task and the logistics time between the resources selected by the two tasks;

(5) The total cost of the logistics among different resources is proportional to the logistics time and the logistics distance;

(6) Different types of first-level manufacturing tasks have the same execution priority;

(7) All cloud manufacturing resources can be used at time zero;

(8) The secondary tasks are uninterrupted once they begin processing until their manufacturing service is complete.

3. The cloud manufacturing resource variation oriented dynamic scheduling method as described in claim 1, wherein: the constraint requirements described mainly include total manufacturing service time constraints, total manufacturing service cost constraints, resource average reliability constraints, manufacturing resource average service efficiency constraints, secondary manufacturing task processing timing constraints, and subtask granularity constraints.