Adjustment scheduling strategy optimization method based on improved genetic algorithm

Abstract

The invention provides a debugging scheduling strategy optimization method based on an improved genetic algorithm, relates to the field of intelligent manufacturing and debugging scheduling, and solves the problem of low coding and execution efficiency of a traditional genetic algorithm caused by huge and complex strategy sets in production debugging scheduling. The method comprises the steps of obtaining a product order information set and a production commissioning and testing strategy set, and performing genetic algorithm parameter initialization; compiling the same kind of mutually exclusive strategies into the same genome position of different individuals, calling a production commissioning and testing simulation algorithm when executing the genetic algorithm, and determining the fitness of the current evolution generation; when the execution process of the genetic algorithm does not meet the preset condition, repeating the evolution process in the genetic algorithm, and determining the fitness again after each evolution is completed; when a preset condition is met, outputting an individual with the maximum fitness in the genetic algorithm as an optimized production commissioning and testing scheme; according to the method, the reasonability judgment on the codes is omitted, and the execution efficiency of the algorithm in the production commissioning scheduling is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing and commissioning and scheduling, and in particular to a commissioning and scheduling strategy optimization method based on an improved genetic algorithm. Background Art

[0002] In the contemporary industrial manufacturing environment, digital transformation has become a key driving force for productivity growth, especially in heavy industries such as military and steel, where the production process involves complex multi-step commissioning processes that require extremely high accuracy and efficiency. With the growing market demand for customization and rapid response, traditional production scheduling methods can no longer meet the dynamic adjustment needs of modern factories. In this context, intelligent optimization algorithms, especially genetic algorithms, have become a powerful tool for optimizing production commissioning strategies due to their powerful global search capabilities and adaptability to large-scale complex problems.

[0003] However, genetic algorithms face multiple challenges when applied to production commissioning. First, the scale of the production strategy set is large and changes with business scenarios, resulting in complex interactions between strategies (such as compatibility and mutual exclusivity). For example, when processing high-priority orders, the "full-amount waiting" strategy may conflict with the "timed waiting" strategy of low-priority orders, which requires the algorithm to effectively identify and avoid these conflicts to achieve optimal resource allocation.

[0004] Traditional genetic algorithms use binary coding to represent the selection of production strategies, but when initializing the population, it is necessary to avoid coding duplication and ensure the feasibility of the coded strategy set, which becomes extremely difficult when the number of strategies is huge. In addition, the rationality verification of the coding, that is, ensuring that mutually exclusive strategies are not activated at the same time, will further increase the computational burden, especially when the strategy set is expanded. This not only increases the difficulty of algorithm design, but also significantly reduces the algorithm's operating efficiency.

[0005] Therefore, the current production debugging system has obvious limitations when facing complex and changing business scenarios, especially when dealing with large-scale production strategy sets. The traditional genetic algorithm encoding method fails to fully consider the dynamic interaction between production strategies and lacks an effective mechanism to deal with the expansion of strategy sets and the increased complexity they bring, thus limiting its performance in actual industrial applications.

[0006] Specifically, the problems of the existing technology are as follows: when dealing with large-scale production strategy sets, the existing algorithms have high coding design complexity and low operating efficiency; there is a lack of effective mechanisms to manage the mutual exclusivity and compatibility between production strategies, resulting in irrational resource allocation; as the production strategy set expands, the coding duplication detection and strategy feasibility verification during population initialization become bottlenecks, affecting algorithm performance; the current method fails to fully utilize the dynamic characteristics of production scheduling in modern industrial scenarios, limiting the improvement of the level of intelligence. Therefore, how to adapt to the ever-changing industrial manufacturing environment and provide a more flexible, efficient, and intelligent production adjustment and scheduling method is the current focus of technical personnel in this field. Summary of the invention

[0007] Based on the current status of the background technology, the purpose of the present invention is to solve the problem of low coding and execution efficiency of traditional genetic algorithms in production test scheduling due to the huge and complex strategy set, and therefore proposes a test scheduling strategy optimization method based on an improved genetic algorithm. The present invention adopts a new coding method, which encodes mutually exclusive strategies of the same type into the same genome and selects them randomly. Such coding omits the judgment of the rationality of the coding and improves the execution efficiency of the algorithm; at the same time, the present invention improves the mutation rules, so that the genetic algorithm can find the optimal population coding scheme more efficiently and accurately, effectively improving the execution efficiency of production test scheduling.

[0008] The present invention adopts the following technical solutions to achieve the purpose:

[0009] A method for optimizing a scheduling strategy based on an improved genetic algorithm comprises the following steps:

[0010] S1. Obtain product order information set and production commissioning strategy set, select a genetic algorithm and perform parameter initialization of the genetic algorithm;

[0011] S2. Based on the production and debugging strategy set, the mutually exclusive strategies of the same type are compiled into the same genome position of different individuals in the genetic algorithm, and then the genetic algorithm is executed;

[0012] S3, when the genetic algorithm is executed, the production commissioning simulation algorithm is called to determine the fitness of the current evolutionary generation of the genetic algorithm based on the product order information set;

[0013] S4. When the execution process of the genetic algorithm does not meet the preset conditions, the evolution process in the genetic algorithm is repeated, and the fitness is determined again after each evolution is completed;

[0014] S5. When the execution process of the genetic algorithm has met the preset conditions, the individual with the maximum fitness in the genetic algorithm is output as the optimized production adjustment plan.

[0015] Specifically, in step S1, the product order information set includes product information and order priority information; the order priority information is used to determine the degree to which an order is prioritized during production line scheduling, and each type of product information has corresponding order priority information.

[0016] Specifically, in step S1, the production commissioning strategy set includes multiple categories of production commissioning strategies; each category of production commissioning strategies includes multiple mutually exclusive strategies of the same category and / or compatible strategies of the same category.

[0017] Furthermore, in step S1, a coding sequence is used to describe an individual in the genetic algorithm, representing a candidate production commissioning scheme; the coding sequence of individual I is recorded as I = {I1, I2, ..., I i ,...,I d}, d represents the total number of categories of production and debugging strategies; at the same time, the number of individuals N in the population of the genetic algorithm and the maximum number of evolution generations g are preset. max 、Number of consecutive minimum fitness changes max , crossover probability P c , mutation probability P m , determine the initialized population and complete the parameter initialization of the genetic algorithm.

[0018] Further, in step S2, for each type of production and testing strategy, a strategy of the same type and mutually exclusive is randomly selected and placed in the corresponding genome position of individual I, that is, a strategy randomly selected from the i-th type of production and testing strategy corresponds to the i-th type of individual I. i The genome position is entered, and then the genetic algorithm is executed to output an individual I from the population as a candidate production adjustment plan.

[0019] Preferably, in step S3, the fitness of the current evolutionary generation of the genetic algorithm is calculated by the fitness function Fitness. For individual I in the genetic algorithm, the fitness function calculation formula is as follows:

[0020] Fitness(I)=T total

[0021] Where, T total Represents the total duration of production commissioning and scheduling; T t0tal The production commissioning scheme represented by individual I is obtained by calculating it through the production commissioning simulation algorithm, T total The smaller the value, the higher the production capacity of the production line per unit time.

[0022] Furthermore, the preset conditions of step S4 and step S5 are: the current evolutionary generation g is less than or equal to the maximum evolutionary generation g max , and the current minimum fitness unchanged times n is greater than or equal to the consecutive minimum fitness unchanged times n max.

[0023] Specifically, in step S4, the evolution process includes a selection operation, a crossover operation, and a mutation operation; wherein the mutation operation is performed under the constraint of the number of mutually exclusive strategies of the same type in each type of production and commissioning strategy.

[0024] Furthermore, in the selection operation, by introducing an exponential scaling factor, the probability of actual selection in the evolutionary process depends on the following formula:

[0025] Fitness SEL (I) = [Fitness (I)] l

[0026] In the formula, Fitness(I) represents the fitness function of individual I; l represents the introduced exponential scaling factor, whose value is greater than 1; the probability of individual I being selected during the evolution process is normalized by Fitness SEL (I) determine;

[0027] The crossover operation is used to exchange the coding segments of the same genomic position between different individuals. For a given pair of individuals, a genomic position is first randomly selected as the crossover point, and all the coding segments before the crossover point are exchanged; the crossover probability P c Decide whether to perform a crossover operation on a pair of individuals;

[0028] In the mutation operation, firstly, a genomic position of individual I is randomly selected, and the number a of mutually exclusive strategies of the same type in a class of production and debugging strategies corresponding to the genomic position is determined. Then, the coding fragment value of the genomic position is obtained and recorded as k, where k represents the strategy ranked as k among multiple mutually exclusive strategies of the same type in this class of production and debugging strategies. The coding fragment value k of the genomic position is changed to any new value except k, and the new value satisfies the constraint that it is less than or equal to the number a of mutually exclusive strategies of the same type.

[0029] The present invention also provides a computer device system, including a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the aforementioned method for optimizing the debugging and scheduling strategy based on the improved genetic algorithm.

[0030] In summary, due to the adoption of this technical solution, the beneficial effects of the present invention are as follows:

[0031] The present invention eliminates the complicated steps of judging the rationality of coding in traditional genetic algorithms by concentrating mutually exclusive strategies in the same genome. This coding method ensures that any generated individual coding automatically satisfies the basic constraints between strategies, greatly reducing the time consumption of the algorithm in the coding stage and significantly improving the execution efficiency of the genetic algorithm.

[0032] The present invention also innovates the mutation mode of the genetic algorithm, so that it can explore the solution space more efficiently while maintaining population diversity, thereby accelerating the process of discovering the optimal generation strategy combination. This improvement is particularly advantageous when dealing with large-scale strategy sets, ensuring that the algorithm can converge to a high-quality solution within a limited time.

[0033] In view of the diversity of industrial production commissioning needs, the method of the present invention can flexibly adapt to different business scenarios, such as the unique processes in the military industry, steel and other industries. By optimizing the coding and mutation methods, the present invention can more effectively handle the compatibility and mutual exclusivity between production strategies, thereby improving the intelligence level and response speed of production commissioning.

[0034] Through the method of the present invention, the algorithm improvement measures can simplify the coding design, reduce the burden of relevant system developers, and make the algorithm implementation easier. This makes it easier for enterprises to integrate production adjustment and optimization technology into existing production management systems without investing a lot of manpower and material resources for complex system transformation.

[0035] Thanks to the optimization of encoding and mutation methods, the present invention significantly improves the performance of genetic algorithms in industrial production commissioning. This not only brings a faster decision cycle, but also reduces the waste of resources caused by improper production commissioning, which ultimately translates into effective control of production costs and increased economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The overall process diagram of the method of the present invention is briefly described;

[0037] Figure 2 Detailed schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0039] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] Example 1

[0041] A method for optimizing the scheduling strategy based on an improved genetic algorithm. The process of this method can be found in Figure 1 The steps are summarized as follows:

[0042] S1. Obtain product order information set and production commissioning strategy set, select a genetic algorithm and perform parameter initialization of the genetic algorithm;

[0043] S2. Based on the production and debugging strategy set, the mutually exclusive strategies of the same type are compiled into the same genome position of different individuals in the genetic algorithm, and then the genetic algorithm is executed;

[0044] S3, when the genetic algorithm is executed, the production commissioning simulation algorithm is called to determine the fitness of the current evolutionary generation of the genetic algorithm based on the product order information set;

[0045] S4. When the execution process of the genetic algorithm does not meet the preset conditions, the evolution process in the genetic algorithm is repeated, and the fitness is determined again after each evolution is completed;

[0046] S5. When the execution process of the genetic algorithm has met the preset conditions, the individual with the maximum fitness in the genetic algorithm is output as the optimized production adjustment plan.

[0047] This embodiment will follow the order of the above steps, combined with Figure 2 The detailed process is given, and the details of each step are specifically introduced. The method of this embodiment is to apply the improved genetic algorithm to the production commissioning process, so as to solve the optimization problem of the commissioning scheduling strategy combination. Since the relationship between strategies exists in two types: mutually exclusive and compatible, rather than the traditional random combination, it is necessary to consider the relevant rules when performing individual mutation and inheritance. This embodiment improves the traditional 0-1 encoding method of the genetic algorithm, and at the same time improves the mutation method, which is introduced in detail below.

[0048] First, it is necessary to give a brief description of the production scheduling strategy as an example scenario for the subsequent application method. The product order information set in step S1 includes product information and order priority information; the order priority information is used to determine the degree to which the order is prioritized during the production line scheduling process, and each type of product information has corresponding order priority information. For production and commissioning lines of products of the same model, the priority of each order is usually the same; when different types of products are mixed, the priorities between different models may be different. This leads to the first example class strategy shown in Table 1 below, namely the order priority determination strategy.

[0049] Table 1 Order priority determination strategy

[0050] Serial number Strategy describe 1 Average method All orders have the same priority 2 Time ladder method Set priorities based on when orders will be completed

[0051] In step S1 of this embodiment, the production debugging strategy set includes, in addition to the above-mentioned order priority determination strategy, other types of strategies, that is, the strategy set contains multiple categories of production debugging strategies; each type of production debugging strategy includes multiple mutually exclusive strategies of the same type and / or compatible strategies of the same type.

[0052] For the processes of the production and commissioning production line, the process is a description of the work stage at a lower and more detailed level than the above-mentioned order. Generally, the higher the order priority to which the process belongs, the higher its own priority; the shorter the process duration, the priority should be processed (thereby reducing long waiting times); the later processes in the process flow should be processed first (thereby completing the current order as soon as possible), and the position of the process in the process flow can be further considered. When there are situations where the order priority is the same and the process duration is the same but the process position is different (in the commissioning of complex equipment, such as the normal temperature performance test needs to be performed multiple times, the above situation may exist), the later process is given priority. This leads to the second example type strategy shown in Table 2 below, namely the process priority determination strategy.

[0053] Table 2 Process priority determination strategy

[0054]

[0055] The process start timing strategy is used to decide how to arrange the process in the "waiting for production" state to start production when production resources are idle; the process start timing strategy is shown in Table 3 below, where "first come, first served" is mutually exclusive with the other two strategies, and the two "predictive waiting" strategies can be compatible, but the three are mutually exclusive when they are of the same category.

[0056] Table 3 Process start timing strategy

[0057]

[0058] In the production scheduling process, when using equipment such as a temperature chamber for environmental testing, considering the long test time, multiple products are usually tested at the same time if the equipment space allows. This is called order combination. This leads to the fourth example type strategy shown in Table 4 below, namely the order combination strategy.

[0059] Table 4: Order-combining strategies

[0060]

[0061] The above four types of policies constitute the production debugging policy set of the scenario in this embodiment. In actual application, the content of the policy set is not limited thereto and can be modified and adjusted according to actual needs.

[0062] After determining the available production adjustment strategies, the next step is to search for the optimal plan and actual execution strategy set for the desired goal, that is, the optimized production adjustment plan, based on the actual production task characteristics through an intelligent combined optimization method, that is, the improved genetic algorithm in this embodiment.

[0063] The traditional genetic algorithm uses a binary sequence to describe a solution to feature selection, that is, an individual in the genetic algorithm. The traditional genetic algorithm individual encoding method and the improved method in this embodiment are the same in parameter notation, both of which are individual I = {I1, I2, ..., I i ,...,I d}, but the meaning is different. In traditional genetic algorithms, d represents the total number of all strategies. When the i-th strategy is selected, I i =1, otherwise I i =0. Therefore, during the initialization process, the traditional genetic algorithm needs to check whether there are multiple individuals with exactly the same code, and then regenerate the relevant individuals to meet the population quantity requirements. At the same time, since there is a mutual exclusion relationship between the strategies, that is, the code values ​​of some strategies cannot be 1 at the same time (such as the full waiting strategy and the timed waiting strategy in Table 4 above), when the traditional genetic algorithm executes the code, it will additionally judge the rationality of the code here, which increases the complexity of the algorithm execution. In particular, when the production and commissioning strategy set is large, a large number of judgment calculations will occur, which reduces the efficiency of the algorithm.

[0064] Therefore, this embodiment improves the encoding-related feature requirements in the genetic algorithm. First, when initializing the parameters, the encoding sequence is used to describe the individuals in the genetic algorithm, representing the candidate production and commissioning solutions; the encoding sequence of individual I is also recorded as I = {I1, I2, ..., I i ,...,I d}, but d represents the total number of categories of production and debugging strategies. At the same time, the number of individuals N in the population of the genetic algorithm and the maximum number of evolution generations g are preset. max 、Number of consecutive minimum fitness changes max , crossover probability P c , mutation probability P m , determine the initialized population.

[0065] The key to improving coding-related features is to make the i-th coding segment I in the genome of individual I i, that is, the fragment that originally determines whether a specific strategy is selected or not, is used to determine which of the mutually exclusive strategies of the same type is selected. The specific method is as follows: in step S2, for each type of production and testing strategy, a strategy of the same type of mutually exclusive strategy is randomly selected and placed in the corresponding genome position of individual I, that is, a strategy randomly selected from the i-th type of production and testing strategy corresponds to the i-th individual I. i The genome position is entered, and then the genetic algorithm is executed to output an individual I from the population as a candidate production adjustment plan.

[0066] For example, an individual I={I1,I2,...,I i ,...,I d}, where I i In the strategy of combining orders in Table 4, I i =1, individual I will adopt the strategy of "waiting for full quota", I i = 2, the "timed waiting" strategy will be adopted, I i =3, the "counting and waiting" strategy will be adopted, I i The value of is no longer limited to the binary sequence of 0 and 1. This encoding omits the rationality judgment of the encoding and improves the efficiency of algorithm execution.

[0067] Since the genetic algorithm is a heuristic search strategy, its core task is to find a fitness function that can accurately evaluate the quality of the solution. The greater the fitness, the better the solution. Therefore, in step S3 of this embodiment, the fitness of the current evolutionary generation of the genetic algorithm is calculated by the fitness function Fitness. For individual I in the genetic algorithm, the fitness function calculation formula is as follows:

[0068] Fitness(I)=T total

[0069] Where, T total Represents the total duration of production commissioning and scheduling; T total The production commissioning scheme represented by individual I is obtained by calculating it through the production commissioning simulation algorithm, T total The smaller the value, the higher the production capacity of the production line per unit time, that is, more products can be produced in the same time range.

[0070] In this embodiment, the preset conditions of step S4 and step S5 are: the current evolutionary generation g is less than or equal to the maximum evolutionary generation g max , and the current minimum fitness unchanged times n is greater than or equal to the consecutive minimum fitness unchanged times n maxThat is, when calculating and determining the fitness after each evolution process, the fitness of the corresponding individual can be stabilized at the minimum number of evolutions without changing the ideal value, which means that the individual that can be selected at this time has met the optimization requirements of the production and debugging plan and can be output as the determined plan.

[0071] The evolution process of step S4 includes selection operation, crossover operation and mutation operation; wherein the mutation operation is performed under the constraint of the number of mutually exclusive strategies of the same type in each type of production and commissioning strategy. Specifically:

[0072] In genetic algorithms, the selection of individuals is a key component of the evolutionary process, and its goal is to simulate the natural selection mechanism in nature, that is, individuals with higher fitness have a greater chance of being selected to participate in subsequent genetic operations. This process is usually implemented through fitness proportionate selection, in which the probability of each individual being selected is directly proportional to its fitness value.

[0073] In the selection operation, by introducing an exponential scaling factor, the probability of actual selection in the evolutionary process depends on the following formula:

[0074] Fitness SEL (I) = [Fitness (I)] l

[0075] In the formula, Fitness(I) represents the fitness function of individual I; l represents the introduced exponential scaling factor, whose value is greater than 1; the probability of individual I being selected during the evolution process is normalized by Fitness SEL (I) Decision.

[0076] The crossover operation is used to exchange the coding segments of the same genomic position between different individuals. For a given pair of individuals, a genomic position is first randomly selected as the crossover point, and all the coding segments before the crossover point are exchanged; the crossover probability P c Determine whether to perform crossover operation on a pair of individuals.

[0077] The mutation operation itself is based on the initialization probability P m To randomly change the coding of individuals, thereby generating new individuals and avoiding the search from falling into the local optimum. The traditional genetic algorithm only needs to change the coding segment value to its opposite value to achieve the binary transformation between 0 and 1, but due to the improvement of the coding features in this embodiment, the traditional mutation method is no longer applicable.

[0078] This embodiment first randomly selects a genome position of individual I, determines the number a of mutually exclusive strategies of the same type in a class of production and debugging strategies corresponding to the genome position, then obtains the coding segment value of the genome position and records it as k, where k represents the strategy ranked as k among multiple mutually exclusive strategies of the same type in the class of production and debugging strategies; changes the coding segment value k of the genome position to any new value other than k, and makes the new value satisfy the constraint of being less than or equal to the number a of mutually exclusive strategies of the same type, so that the genome position will correspond to another mutually exclusive strategy of the same type in the class of production and debugging strategies, thereby realizing the mutation operation.

[0079] The above method of this embodiment can be applied to a computer device system. The system may include a memory, a processor and a computer program stored in the memory. The processor executes the computer program to implement the steps of the debugging scheduling strategy optimization method based on the improved genetic algorithm.

[0080] Example 2

[0081] Based on Example 1, this example applies the method as a specific example, and better illustrates the execution process of the method through simulation. The simulation experiment of this example is implemented on a three-dimensional visualization centralized intelligent management and control platform. The production scheduling subsystem in the platform integrates the testing and control functions of the entire process from module to complete machine, from normal temperature to environmental testing, and can perform full-factor simulation verification of the entire production scheduling process based on the method proposed in Example 1.

[0082] The simulation experiment must first determine the process flow and required resources for production scheduling. The production line resources and process flow set in this embodiment are shown in Tables 5 and 6 below.

[0083] Table 5 Simulation production line resource table

[0084] Workbench Name Working hours Prepare the workbench Monday to Friday, 8h / day Test bench 1 Monday to Friday, 8h / day Test bench 2 Monday to Friday, 8h / day Test bench 3 Monday to Friday, 8h / day Test bench 4 Monday to Friday, 8h / day Test bench 5 Monday to Friday, 8h / day Test bench 6 Monday to Friday, 8h / day Temperature bench 1 24 hours a day, all year round Temperature bench 2 24 hours a day, all year round

[0085] Table 6 Simulation process flow chart

[0086]

[0087]

[0088] The hardware environment of the simulation experiment is: desktop PC model: ThinkStation P340; operating system: Windows 10 Professional Edition; CPU: Intel(R)Core(TM)i7-10700K CPU@3.80GHz3.79GHz; memory: DDR4 16.0GB; hard disk: 2TB.

[0089] This embodiment also sets the product order information set D = {D1, D2, ..., D i,...,D 30}, a total of 30 orders. Set the production commissioning strategy set ST, using all the strategies proposed in Example 1, so:

[0090]

[0091] The simulation experiment process is as follows:

[0092] Initialization parameters, given the above production and commissioning strategy set ST, the number of individuals in the population N = 4, the maximum evolutionary generation g max = 20, the number of consecutive minimum fitness unchanged times n max =5, crossover probability P c =0.8, mutation probability P m =0.1.

[0093] According to the mutually exclusive rules, the average method and the time gradient method are mutually exclusive; the inherited order, order and process time, and process position are mutually exclusive; first-come-first-served, priority prediction waiting, and personnel busy and idle prediction waiting are mutually exclusive; full quota waiting, timed waiting, and counted waiting are mutually exclusive. In this embodiment, individuals are numbered from 0 and recorded as the letter P, and the individuals in the population after random initialization are as follows:

[0094] P0 = {average method, process location, priority prediction waiting, count waiting}

[0095] P1 = {time gradient method, process location, first come first served, full quota waiting}

[0096] P2 = {time gradient method, process location, first come first served, count waiting}

[0097] P3 = {time gradient method, process location, first come first served, timed waiting}

[0098] From the current population, two individuals are selected in sequence to form individual pairs (such as P0 and P1), and the crossover probability P c =0.8 determines whether each group of individuals participates in the crossover. If so, a crossover point is randomly generated, and the fragments before the two are exchanged to update the current population. This embodiment simulates the crossover of P0 and P1. The crossover position is the first fragment, and the corresponding strategies are exchanged. After the crossover, the original P0 and P1 disappear, and new individuals P4 and P5 are generated. The individuals in the updated population are as follows:

[0099] P4 = {time gradient method, process location, priority prediction waiting, count waiting}

[0100] P5 = {average method, process location, first come first served, full quota waiting}

[0101] P2 = {time gradient method, process location, first come first served, count waiting}

[0102] P3 = {time gradient method, process location, first come first served, timed waiting}

[0103] For each individual in the current population (such as P2), according to the mutation probability P m =0.1 to determine whether it participates in mutation. If mutation occurs, a mutation point is randomly selected (the simulation is the second genome position, that is, the corresponding strategy category is the process priority determination strategy), whose value is k (the simulation corresponds to the process position), and then the value is changed to any other value of the same category that is not k (the simulation is changed to inheritance order), which means that P2 has mutated, the original P2 disappears and a new individual P6 is generated. The individuals in the updated population are as follows:

[0104] P4 = {time gradient method, process location, priority prediction waiting, count waiting}

[0105] P5 = {average method, process location, first come first served, full quota waiting}

[0106] P6 = {Time gradient method, inherited order, first come first served, count waiting}

[0107] P3 = {time gradient method, process location, first come first served, timed waiting}

[0108] Increase the evolutionary algebraic index value g, make it increment by 1, and repeat the evolution process until g>g max Or the total duration T of production commissioning scheduling for 5 consecutive population simulations total When the value of is at the minimum and unchanged, the evolution is stopped and the individual with the best fitness is output.

[0109] In this embodiment, after 6 minutes and 8 rounds of iteration, the simulation population iteration results are as follows:

[0110] P i1 ={time gradient method, process location, priority prediction waiting, count waiting}

[0111] P i2 ={Average method, process location, first come first served, full quota waiting}

[0112] P i3 ={Average method, inherited order, first come first served, count and wait}

[0113] P i4 ={Time gradient method, process location, first come first served, timed waiting}

[0114] P i1 The scheduling duration simulation is 50784min; i2 The scheduling duration simulation is 55284min;i3 The scheduling duration simulation is 50995min; i4 The scheduling duration simulation is 50630min.

[0115] Select the individual P with the largest fitness (i.e., the smallest schedule duration) in the final population i4 , as the solution for feature optimization, it is output and used as the optimized production debugging plan.

Claims

1. A method for optimizing a scheduling strategy based on an improved genetic algorithm, characterized in that: The method comprises the following steps: S1. Obtain product order information set and production commissioning strategy set, select a genetic algorithm and perform parameter initialization of the genetic algorithm; S2. Based on the production and debugging strategy set, the mutually exclusive strategies of the same type are compiled into the same genome position of different individuals in the genetic algorithm, and then the genetic algorithm is executed; S3, when the genetic algorithm is executed, the production commissioning simulation algorithm is called to determine the fitness of the current evolutionary generation of the genetic algorithm based on the product order information set; S4. When the execution process of the genetic algorithm does not meet the preset conditions, the evolution process in the genetic algorithm is repeated, and the fitness is determined again after each evolution is completed; S5. When the execution process of the genetic algorithm has met the preset conditions, the individual with the maximum fitness in the genetic algorithm is output as the optimized production adjustment plan.

2. The method for optimizing the debugging and scheduling strategy based on the improved genetic algorithm according to claim 1, characterized in that: In step S1, the product order information set includes product information and order priority information; the order priority information is used to determine the degree to which an order is prioritized during production line scheduling, and each type of product information has corresponding order priority information.

3. The method for optimizing the debugging and scheduling strategy based on the improved genetic algorithm according to claim 1, characterized in that: In step S1, the production commissioning strategy set includes multiple categories of production commissioning strategies; each category of production commissioning strategies includes multiple mutually exclusive strategies of the same category and / or compatible strategies of the same category.

4. The method for optimizing the debugging and scheduling strategy based on the improved genetic algorithm according to claim 1, characterized in that: In step S1, a coding sequence is used to describe an individual in the genetic algorithm, representing a candidate production commissioning scheme; the coding sequence of individual I is recorded as I = {I1, I2, ..., I i ,...,I d }, d represents the total number of categories of production and debugging strategies; at the same time, the number of individuals N in the population of the genetic algorithm and the maximum number of evolution generations g are preset. max 、Number of consecutive minimum fitness changes max , crossover probability P c , mutation probability P m , determine the initialized population and complete the parameter initialization of the genetic algorithm.

5. The method for optimizing the debugging and scheduling strategy based on the improved genetic algorithm according to claim 4 is characterized in that: In step S2, for each type of production and testing strategy, a strategy of the same type and mutually exclusive is randomly selected and placed in the corresponding genome position of individual I, that is, a strategy randomly selected from the i-th type of production and testing strategy corresponds to the i-th type of individual I. i The genome position is entered, and then the genetic algorithm is executed to output an individual I from the population as a candidate production adjustment plan.

6. The method for optimizing the debugging and scheduling strategy based on the improved genetic algorithm according to claim 1, characterized in that: In step S3, the fitness of the current evolutionary generation of the genetic algorithm is calculated by the fitness function Fitness. For individual I in the genetic algorithm, the fitness function calculation formula is as follows: Fitness(I)=T total Where, T total Represents the total duration of production commissioning and scheduling; T total The production commissioning scheme represented by individual I is obtained by calculating it through the production commissioning simulation algorithm, T total The smaller the value, the higher the production capacity of the production line per unit time.

7. The method for optimizing the debugging and scheduling strategy based on an improved genetic algorithm according to claim 4, characterized in that: The preset conditions of step S4 and step S5 are: the current evolutionary generation g is less than or equal to the maximum evolutionary generation g max , and the current minimum fitness unchanged times n is greater than or equal to the consecutive minimum fitness unchanged times n max .

8. The method for optimizing the debugging and scheduling strategy based on the improved genetic algorithm according to claim 4 is characterized in that: In step S4, the evolution process includes selection operation, crossover operation and mutation operation; wherein the mutation operation is performed under the constraint of the number of mutually exclusive strategies of the same type in each type of production and commissioning strategy.

9. The method for optimizing the debugging and scheduling strategy based on the improved genetic algorithm according to claim 8, characterized in that: In the selection operation, by introducing an exponential scaling factor, the probability of actual selection in the evolutionary process depends on the following formula: Fitness SEL (I)=[Fitness(I)] l Where, Fitness(I) represents the fitness function of individual I; l represents the introduced exponential scaling factor, whose value is greater than 1; The probability of individual I being selected during the evolution process is normalized by Fitness SEL (I) determine; The crossover operation is used to exchange the coding segments of the same genomic position between different individuals. For a given pair of individuals, a genomic position is first randomly selected as the crossover point, and all the coding segments before the crossover point are exchanged; the crossover probability P c Decide whether to perform a crossover operation on a pair of individuals; In the mutation operation, firstly, a genomic position of individual I is randomly selected, and the number a of mutually exclusive strategies of the same type in a class of production and debugging strategies corresponding to the genomic position is determined. Then, the coding fragment value of the genomic position is obtained and recorded as k, where k represents the strategy ranked as k among multiple mutually exclusive strategies of the same type in this class of production and debugging strategies. The coding fragment value k of the genomic position is changed to any new value except k, and the new value satisfies the constraint that it is less than or equal to the number a of mutually exclusive strategies of the same type.

10. A computer device system, comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the debugging scheduling strategy optimization method based on the improved genetic algorithm according to any one of claims 1 to 9.

Images