Intelligent scheduling method and system for welding assembly

Abstract

The present invention provides an intelligent scheduling method and system for welding assembly, including: implementing nested process coding, determining processing information such as the processing equipment and buffer capacity required for welding; collecting historical production data, extracting key production data and interference data from the welding process, fitting data features, and determining a data distribution function; establishing a dynamic mathematical description model for the welding process and designing a target expression formula; implementing a model-solving algorithm system, including an improved genetic algorithm, a construction heuristic algorithm, and a lower bound estimation method based on problem model theory; solving the established mathematical problem model, generating random cases, comparing them with the estimated lower bound, and revising the optimal parameters input to the algorithm system based on production scheduling indicators such as completion time and equipment utilization; inputting verification production data, and using the optimal parameter system to solve the problem. The present invention solves the technical problems of dynamic problem representation defects, poor simulation results, poor expression accuracy, and poor adaptability.

Description

Technical Field

[0001] The present invention relates to the field of production scheduling simulation optimization, and in particular to a welding assembly intelligent scheduling method and system. Background Art

[0002] At present, the gradual improvement of mathematical modeling technology and intelligent algorithms has greatly improved the accuracy and efficiency of solving scheduling problems. It is especially suitable for production enterprises that are extremely sensitive to fixed costs. The assembly scheduling problem model and improved algorithm used in the assembly manufacturing system solve the scheduling problem of products composed of many parts that have undergone the same or similar processing. It greatly reduces the production costs of manufacturing enterprises and improves the competitiveness of products.

[0003] Although mathematical models for various scheduling problems have emerged, for example, the existing invention patent application document "A Transparent Monitoring Method and System for Intelligent Workshops" with publication number CN107870600A includes the following steps: Step A: Building an intelligent workshop transparent monitoring platform, which includes: Step A1: Establishing a mechanism for interconnecting virtual models and physical objects; Step A2: Static modeling of the workshop; Step A3: Dynamic modeling of the workshop; Step A4: Model and equipment integration; Step B: Implementing a transparent monitoring method for intelligent workshops: Step B1: Specific implementation of three-dimensional simulation of the intelligent workshop; Step B2: Associating the virtual model with the physical model; Step B3: Command transmission and data collection and feedback; Step B4: Data visualization. A transparent monitoring system for intelligent workshops includes an MES module, a unit control module, a SCADA module, and a bus control network module. The essence of the above-mentioned prior art is still a general static model, which fails to break through the defect of static models in representing dynamic problems. There is a lack of analysis for workpiece processing processes similar to multi-layer nested processes, and the intelligent algorithms, heuristic algorithms and lower bound estimation methods for related requirements are also different from the existing designs. In view of the differences in the scheduling process, there are generally two ways to deal with it: one is to idealize the problem, ignore the re-entry of the process and simply add it up, and the treatment of time is also biased towards hypothetical estimation, failing to consider the law of time change; for example, the existing invention patent application document "A large-scale flexible job shop scheduling optimization method" with publication number CN107862411A has the following steps: (1) First, cluster and batch the workpieces with similar processing technology, workpiece sizes in the same range, and the same blank material, so as to reduce the scale of problem solving; (2) Set the initial parameters of the algorithm, adopt three-layer gene encoding technology, OBX crossover method and certain mutation strategy, combine simulation experiments to select the crossover length, and use adaptive improved genetic algorithm to optimize the solution. The aforementioned existing technologies add additional constraints to similar scheduling problems, limit processing conditions, and adopt an average value approach to time processing, which also lacks simulation reflection of the production process. The nested processes existing in the tank welding process significantly affect the accuracy of the production scheduling model expression and the adaptability of the solution method design. The lack of targeted mathematical models and intelligent algorithms fails to fully improve the potential of the production process.

[0004] In summary, the existing technology has technical problems such as defects in dynamic problem representation, poor simulation effect, poor expression accuracy and poor adaptability. Summary of the Invention

[0005] The technical problem to be solved by the present invention is how to solve the technical problems of dynamic problem representation defects, poor simulation effects, poor expression accuracy and poor adaptability.

[0006] The present invention solves the above technical problems by adopting the following technical solutions: A welding assembly intelligent scheduling method includes:

[0007] S1. Implement preset nested process coding to determine the processing equipment information and buffer capacity required for welding;

[0008] S2. Collect production history data, extract production data and interference data of the welding process, fit the data features of the production data and interference data, and determine the data distribution function accordingly;

[0009] S3. Analyze the processing equipment information and buffer capacity based on the data distribution function and establish a dynamic mathematical description model for the welding process;

[0010] S4. Utilize preset algorithms and nested process codes to construct and implement a model-solving algorithm system, wherein the preset algorithms include: an improved genetic algorithm, a construction heuristic algorithm, and a lower bound estimation method based on problem model theory. By utilizing the model-solving algorithm system to solve the problem, a mathematical problem model is constructed and random cases are generated to estimate the lower bound of the bottleneck in stages according to a dynamic mathematical description model of the welding process.

[0011] S5. Compare the model solving algorithm system with the bottleneck lower bound, and modify the applicable input parameters of the model solving algorithm system according to the preset production scheduling index, thereby obtaining an applicable parameter system;

[0012] S6. Input the verified production data into the applicable parameter system to obtain a scheduling solution applicable to the intelligent scheduling of welding assembly.

[0013] This system integrates data analysis, mathematical modeling of nested scheduling problems, and algorithm design to meet the processing needs of complex production workshops. It reconstructs genetic algorithms, constructs heuristic algorithms, and lower bounds for scheduling problems to address the complex problem-solving of welding processes. Utilizing this system to develop optimal production plans enables high-efficiency, low-cost production, shortening overall processing completion time and average throughput time for such enterprises, effectively improving equipment utilization and reducing inventory. This system and method are suitable for production lines and enterprises with high production costs and customization, and are also valuable for optimizing the operation of similar production processes.

[0014] In a more specific technical solution, step S1 includes:

[0015] S11. According to the nested process code, determine the workpiece and its repetition number, and compile it into an outer code;

[0016] S12, determining the process number of each workpiece and determining it as an inner layer code;

[0017] S13. According to the preset processing position, the outer layer code and the outer layer code are allocated to obtain the processing equipment information and the buffer area capacity that meet the processing sequence constraint.

[0018] The present invention can intuitively and accurately describe the dynamics and randomness of the product processing process and equipment utilization by performing multi-layer encoding on the workpiece processing process with nested steps, and has the characteristics of simulation based on empirical data analysis and improves the accuracy of simulation.

[0019] In a more specific technical solution, step S3 includes:

[0020] S31. Calculate the processing time data and the disturbance time variation function based on the processing equipment information, and establish a dynamic mathematical description model based on the welding process to solve the completion time and average throughput time using the following logic:

[0021] f1=minC max and f2 = minP ave ;

[0022] Among them, f i (i=1, 2) represents the i-th scheduling target. max represents the maximum completion time of production (maxcompletion time); P ave The average passing time of each bin is the time between the earliest production start time of its parts and the completion time of the bin.

[0023] S32. Fitting a probability function of random reentry based on the processing process in the production history data, calculating the expected number of reentries, and adding the expected number of reentries to the real-time scheduling model to express a processing model related to the number of reentries;

[0024] S33. Determine the decision variable constraints using the following logic:

[0025] B aklb (t)≥max(D akl-lb (t), B a′k′l′b (t′))

[0026] B aklb (t+1)=D akl′b (t)

[0027] Among them, suppose the lth process of the kth component of the ath tank is processed on the bth machine, B aklb Indicates the start time of the process in the tth processing, D aklbrepresents the end time of the process during the tth processing. The first formula indicates that the start time of the next process must not be less than the end time of the previous process of the same component and the end time of other processes after the same processing equipment is used. The second formula indicates that the machine has rework operations, and the start time of rework is equal to the end time of the previous processing. The decision variable constraints are added to the scheduling formula for solution;

[0028] S34. According to the production demand in the production history data, the decision variables, solution objectives and algorithm fitness are processed accordingly.

[0029] This invention analyzes processing equipment information and buffer capacity to determine decision variable constraints, which are then incorporated into the scheduling formula to facilitate algorithmic solution, improving the model's simulation of the intelligent scheduling process for welding assembly. By establishing a mathematical model with cyclical characteristics, the invention reflects the reentrant nature of the production process and optimizes batch processing and nested process operations in tank welding.

[0030] In a more specific technical solution, in step S32, the probability function of random reentry is fitted using the following logic to calculate the expected number of reentry times T:

[0031]

[0032] Where t is the number of reentry times.

[0033] The present invention has the advantages of selective feasible solution updating and iterative process acceleration to improve the quality of feasible solutions, and is also applicable to other complex scheduling problems, continuously reducing the space complexity and solution complexity of the problem.

[0034] In a more specific technical solution, step S4 includes:

[0035] S41. Acquire and construct heuristic rules based on the characteristics of the tank welding assembly scheduling problem, apply the heuristic rules to the dynamic mathematical description model of the welding process in stages based on the preset processing process, and obtain an initial scheduling solution;

[0036] S42. Designing an adaptive crossover mutation operator of an improved genetic algorithm based on the initial scheduling solution and the nested process coding;

[0037] S43. Construct and use the logistic model to control the iterative process of the improved genetic algorithm and the balance mechanism between local optimum and global optimum;

[0038] S44, taking different stages as bottleneck stages, calculating the lower bound of each bottleneck stage for scheduling, and estimating the overall lower bound accordingly.

[0039] This paper uses a constructed heuristic solution as the initial solution. The improved adaptive genetic algorithm accelerates algorithm search speed, improves the efficiency of exploring feasible solutions, and optimizes solution quality. The paper also designs an adaptive crossover and mutation operator based on nested coding, and adds a logistic control iterative process and a balance mechanism between local and global optima to increase the iteration speed of the genetic algorithm.

[0040] In a more specific technical solution, in step S43, an improved genetic algorithm is used to set a new iteration start method and update method, and a logistic model is constructed using the following logic to update the number of iterations and the population level:

[0041] and r(t)=r0·(1-A(t) / K).

[0042] Where A(t) represents the population size during the t-th iteration, K represents the species limit, A0 represents the initial population size of the species, r0 represents the initial natural growth rate of the population, e represents the natural exponent (=2.71828…), and r(t) represents the population growth rate during the t-th iteration.

[0043] In a more specific technical solution, in step S43, the following search logic is used to adaptively search for a new feasible solution based on the initial scheduling solution:

[0044] p c =A0·p c0 / A(t) and p m =p m0 ·r(t) / r0.

[0045] Among them, p c represents the probability of chromosome crossover, P c0 represents the initial crossover probability of chromosomes, p m represents the probability of chromosome mutation, p m0 represents the initial mutation probability of chromosome.

[0046] In a more specific technical solution, step S44 includes:

[0047] S441. Assume the first stage using the following logic to obtain the lower bound indicator of the bottleneck in the first stage:

[0048]

[0049] Among them, L1 represents the lower bound of the optimization problem assumed in the first stage, O akl represents the lth process of the kth component of the ath tank, δ aklb Is a binary value, if O akl Processing on the bth machine, δ aklbis 1, otherwise 0; Indicates the number of all machines.

[0050] S442. Assume the second stage using the following logic to obtain the lower bound indicator of the bottleneck in the second stage:

[0051]

[0052] S443. Determine the bottleneck lower bound using the first-stage bottleneck lower bound indicator and the second-stage bottleneck lower bound indicator.

[0053] In a more specific technical solution, in step S53, the following logic is used to determine the bottleneck lower bound based on the first-stage bottleneck lower bound indicator and the second-stage bottleneck lower bound indicator:

[0054] L=max{L1, L2}.

[0055] In a more specific technical solution, a welding assembly intelligent scheduling system includes:

[0056] Nested coding implementation module, used to implement preset nested process coding to determine the processing equipment information and buffer capacity required for welding;

[0057] The production history data fitting processing module is used to collect production history data, extract production data and interference data of the welding process, fit the data features of the production data and interference data, and determine the data distribution function accordingly;

[0058] A welding process description model construction module is used to analyze processing equipment information and buffer area capacity according to the data distribution function and establish a dynamic mathematical description model of the welding process. The welding process description model construction module is connected to the nested coding implementation module and the production history data fitting processing module;

[0059] A solution system construction and lower bound estimation module is used to construct and implement a model solution algorithm system using preset algorithms and nested process coding, wherein the preset algorithms include: an improved genetic algorithm, a construction heuristic algorithm, and a lower bound estimation method based on problem model theory. By using the model solution algorithm system to solve, a mathematical problem model is constructed based on it, and random cases are generated to estimate the bottleneck lower bound in stages according to the dynamic mathematical description model of the welding process. The solution system construction and lower bound estimation module is connected to the welding process description model construction module;

[0060] A solution system correction module is used to compare the model solution algorithm system with the bottleneck lower bound, so as to correct the applicable input parameters of the model solution algorithm system according to the preset production scheduling indicators, thereby obtaining an applicable parameter system. The solution system correction module is connected to the solution system construction and lower bound estimation module;

[0061] The scheduling solution module inputs the verified production data into the applicable parameter system to obtain a scheduling solution applicable to the welding assembly intelligent scheduling. The scheduling solution module is connected to the solution system correction module.

[0062] Compared to existing technologies, this invention offers the following advantages: It systematically integrates data analysis, mathematical modeling of nested scheduling problems, and algorithm design to meet the processing needs of complex production workshops. It reconstructs genetic algorithms, constructs heuristic algorithms, and lower bounds for scheduling problems to address the complex problem-solving of welding processes. Utilizing this invention to develop optimal production plans enables high-efficiency, low-cost production, shortening overall processing completion time and average throughput time for such enterprises, effectively improving equipment utilization and reducing inventory. This system and method is suitable for production lines and enterprises with high production costs and customization, and is also instructive for optimizing the operation of similar production processes.

[0063] The present invention can intuitively and accurately describe the dynamics and randomness of the product processing process and equipment utilization by performing multi-layer encoding on the workpiece processing process with nested steps, and has the characteristics of simulation based on empirical data analysis and improves the accuracy of simulation.

[0064] This invention analyzes processing equipment information and buffer capacity to determine decision variable constraints, which are then incorporated into the scheduling formula to facilitate algorithmic solution, improving the model's simulation of the intelligent scheduling process for welding assembly. By establishing a mathematical model with cyclical characteristics, the invention reflects the reentrant nature of the production process and optimizes batch processing and nested process operations in tank welding.

[0065] The present invention has the advantages of selective feasible solution updating and iterative process acceleration to improve the quality of feasible solutions, and is also applicable to other complex scheduling problems, continuously reducing the space complexity and solution complexity of the problem.

[0066] This invention uses a constructed heuristic solution as the initial solution. The improved adaptive genetic algorithm accelerates algorithm search speed, improves the efficiency of exploring feasible solutions, and optimizes solution quality. The invention also designs an adaptive crossover and mutation operator based on nested coding, and adds a logistic control iteration process and a balance mechanism between local and global optima to increase the iteration speed of the genetic algorithm. This invention addresses the technical problems of existing technologies, such as flawed dynamic problem representation, poor simulation results, and poor expression accuracy and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a data flow processing diagram of a welding assembly intelligent scheduling system according to embodiment 1 of the present invention;

[0068] Figure 2This is a schematic diagram of the basic steps of a welding assembly intelligent scheduling method according to embodiment 1 of the present invention;

[0069] Figure 3 Schematic diagram of specific steps for determining processing data in Example 1 of the present invention;

[0070] Figure 4 This is a schematic diagram of specific steps for modeling and expressing Example 1 of the present invention;

[0071] Figure 5 Schematic diagram of specific steps of implementing the model solving algorithm system of embodiment 1 of the present invention;

[0072] Figure 6 This is a schematic diagram of a specific implementation flow of a welding assembly intelligent scheduling method according to embodiment 2 of the present invention;

[0073] Figure 7 This is a schematic diagram of a welding assembly product processing process with nested steps according to Example 2 of the present invention;

[0074] Figure 8 This is a schematic diagram of the improved genetic algorithm flow in Example 2 of the present invention;

[0075] Figure 9 This is a diagram illustrating the coding cross-section of Example 2 of the present invention;

[0076] Figure 10 This is a diagram illustrating coding variations according to Example 2 of the present invention;

[0077] Figure 11(a) to Figure 11(o) A box plot of the results of solving the nested-process workpiece scheduling problem using the heuristic rule, (improved) genetic algorithm, greedy algorithm, and neighborhood search algorithm according to Example 2 of the present invention;

[0078] Figure 12 This is a bi-objective iteration curve diagram of solving the nested process scheduling problem using the improved genetic algorithm according to Example 2 of the present invention;

[0079] Figure 13 This is a production scheduling Gantt chart for solving the nested scheduling problem using the improved genetic algorithm according to embodiment 2 of the present invention. DETAILED DESCRIPTION

[0080] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0081] Example 1

[0082] like Figure 1 As shown, the present invention provides an intelligent scheduling method and system integration for welding assembly, which embeds tools such as data analysis, mathematical modeling simulation, and algorithm adaptive solution. It mainly carries out three core parts including disturbance data analysis and fitting, mathematical modeling of re-entrant nested process production process, and algorithm adaptive solution optimization. The data analysis module performs probability testing and fitting on the historical dynamic processing information, dynamic machine damage probability, and dynamic re-entry process of the welding process to obtain an empirical function that simulates the actual production situation. The mathematical modeling module of the welding nested process takes into account the fitting data and the re-processing characteristics of the nested process, and realizes dynamic time processing and dynamic re-processing on the basis of the classic scheduling model. process, establishes a dual dynamic change mathematical model; the adaptive algorithm test solution optimization module includes test design, automatic data generation, improved algorithm and other intelligent algorithm application solution comparison and adaptive parameter correction parts; the test design part realizes the design of main parameters of the model, algorithm program application, and data application program interface; the automatic data generation realizes production simulation and simulates multi-scenario production; the algorithm application comparison realizes the comparison of improved genetic algorithm, constructed heuristic algorithm and lower bound estimation deviation, and adjusts the optimal algorithm parameters according to the deviation trend; the verification data adjustment realizes the substitution of actual data, instantly optimizes the production plan, and outputs the optimal workpiece processing plan and scheduling plan.

[0083] like Figure 2 As shown, the welding assembly intelligent scheduling method provided by the present invention includes the following basic steps:

[0084] Step S1: Implement nested process coding, determine processing information such as processing equipment required for welding, buffer area capacity, etc.;

[0085] like Figure 3 As shown, in this embodiment, step S1 also includes the following specific steps:

[0086] Step S11, determine the workpiece and its repetition times, and compile them into the outermost code;

[0087] Step S12: determine the process number of each workpiece and determine it as an inner layer code;

[0088] Step S13: distribute the inner and outer layer codes according to the processing positions so as to satisfy the processing sequence constraints.

[0089] Step S2: Collect production history data, extract main production data and interference data of the welding process, fit data features, and determine data distribution function;

[0090] Step S3: establishing a dynamic mathematical description model of the welding process based on data analysis and scheduling theory, and designing an expression formula for solving the target;

[0091] like Figure 4 As shown, in this embodiment, step S3 also includes the following specific steps:

[0092] Step S31: Calculate the processing time data and the disturbance time variation function, and use the sum of the two as the process time information of the scheduling model;

[0093] Step S32: Fitting a random reentry probability function based on the provided processing process, calculating the expected number of reentries, and adding it to the real-time scheduling model to represent a processing model related to the number of reentries;

[0094] Step S33: determine the decision variable constraints and add them to the scheduling formula to facilitate the algorithm to solve the problem;

[0095] Step S34: According to production requirements, the decision variables, solution objectives and algorithm fitness are matched.

[0096] Step S4, implementing a model solving algorithm system, including improving a genetic algorithm, constructing a heuristic algorithm, and a lower bound estimation method of a problem model theory;

[0097] like Figure 5 As shown, in this embodiment, step S4 also includes the following specific steps:

[0098] Step S41: construct heuristic rules according to the characteristics of the problem, apply the scheduling rules to the scheduling model in stages based on the provided processing process, and obtain an initial scheduling solution;

[0099] Step S42: Using the constructed heuristic solution as the initial solution, the genetic algorithm is improved, and an adaptive crossover mutation operator is designed based on nested coding. A logistic control iterative process and a balance mechanism between local optimum and global optimum are added to improve the iteration speed of the genetic algorithm.

[0100] In step S43 , different stages are regarded as bottleneck stages, and the theoretical lower bound of each stage of scheduling is calculated, thereby estimating the overall lower bound.

[0101] Step S5: Solve the established mathematical problem model through algorithm experimentation, automatically generate random cases, apply the designed algorithm to compare with the estimated lower bound, and modify the optimal parameters input to the algorithm system based on indicators such as completion time and equipment utilization rate of production scheduling;

[0102] Step S6: Input and verify production data, use the optimal parameter system to solve the problem, output the optimal scheduling plan and manually check and adjust the actual plan.

[0103] Example 2

[0104] In this embodiment of a rocket tank welding production line, the number of three types of tanks, A, B, and C, is set to, for example, 13, 8, and 5, respectively. In this embodiment, the tanks include, but are not limited to, a single bottom, a short shell, and a barrel segment, with three types of components. The number of the three components of tank A is 3, 2, and 1, respectively; the number of the three components of tank B is 2, 2, and 5, respectively; and the number of the three components of tank C is 4, 2, and 3, respectively. The processing time has both truncated normal and uniform distribution characteristics, and includes a probabilistic rework and reentry process.

[0105] In this embodiment, a welding assembly intelligent scheduling system integrates embedded tools including but not limited to: a data analysis module 1, a welding nesting process mathematical modeling module 2, and an adaptive algorithm test solution optimization module 3, which performs core parts including: disturbance data analysis and fitting, re-entrant nesting process mathematical modeling, algorithm adaptive solution optimization, etc.

[0106] In this embodiment, the data analysis module 1 performs probability testing and fitting on the historical dynamic processing information, dynamic machine damage probability, and dynamic re-entry process of the welding process to obtain an empirical function that simulates the actual production conditions;

[0107] In this embodiment, the welding nesting process mathematical modeling module 2 considers the fitting data and the reprocessing characteristics of the nested process, realizes the dynamic time processing and dynamic reprocessing process on the basis of the classic scheduling model, and establishes a dual dynamic change mathematical model;

[0108] In this embodiment, the adaptive algorithm experiment solution optimization module 3 includes: experiment design, automatic data generation, improved algorithm and other intelligent algorithm application solution comparison and adaptive parameter correction part; the experiment design part realizes the design of the main parameters of the model, the application of the algorithm program, and the data application program interface; the automatic data generation realizes production simulation and simulates multi-scenario production; the algorithm application comparison realizes the comparison of the improved genetic algorithm, the constructed heuristic algorithm and the lower bound estimation deviation, and adjusts the optimal algorithm parameters according to the deviation trend; the verification data adjustment realizes the substitution of actual data, instantly optimizes the production plan, and outputs the optimal workpiece processing plan and scheduling plan.

[0109] Step S1', implement nested process coding, determine processing equipment required for welding, buffer area capacity and other processing information;

[0110] like Figure 6 and Figure 7As shown, in this embodiment, multiple workpieces are processed in a re-entry process. Workpieces of the same type are processed in an assembly line, and workpieces of different types are processed in a job shop. Therefore, the first stage of the processing process is a welding assembly problem in which assembly line processes and job shop processes are nested. The total number of workpieces is 26, and the total number of the three components is 75, 52, and 68, respectively. According to the basic process numbers of 5, 5, and 4, the initial coding forms of the three products are [00…11…22…]. The internal coding number is the number of processes. Considering the number of products, the second-level coding will repeat the internal coding number of times as the number of workpieces.

[0111] Step S2', analyzing historical production data;

[0112] In this embodiment, the process hours follow a truncated normal distribution:

[0113] ψ(X,X / 10,XX / 5,X+X / 5;o)

[0114] In this embodiment, the setting time follows a normal distribution:

[0115] U(xx / 10,x+x / 10)

[0116] In this embodiment, the non-destructive testing results show that about 5% of the parts need to be reworked, and the rework time is the same as the original time.

[0117] Step S3': Establish a dynamic mathematical description model based on data analysis and scheduling theory, and solve the target of completion time and average transit time, that is:

[0118] f1=minC max and f2 = minP ave ;

[0119] In this embodiment, the time consumed by a process in the model is the sum of the processing time and the setup time, that is, p + s; the probability function of random reentry is fitted, and the expected number of reentries is calculated, that is:

[0120]

[0121] Where t is the number of reentry times;

[0122] The core constraint expression of the dynamic scheduling model is:

[0123] B aklb (t)≥max(D akl-1b (t), B a′k′l′b (t′))

[0124] Indicates that the start time of an operation should be after the end time of the previous operation in the same cycle, or after the end time of a different cycle operation on the same equipment, and:

[0125] B aklb (t+1)=D akl′b (t)

[0126] Indicates that the start time of the first cycle operation should be after the end time of the last operation of the previous cycle.

[0127] Step S4', the improved genetic algorithm first sets a new iteration start mode and update mode. In this embodiment, the logistic model uses:

[0128] and r(t)=r0·(1-A(t) / K)

[0129] Update the number of iterations and population level.

[0130] like Figure 8 As shown, in this embodiment, the improved genetic algorithm involved in step S4' includes the following specific steps:

[0131] S41', let A = A0, r = r0;

[0132] S42', set k = 1;

[0133] S43', Retention probability calculation formula. P tk Indicates the chromosome X in the k-th optimization process at the t-th iteration tk The retention probability, e is the natural index, f tk Represents chromosome X tk The fitness value, A and r represent the number of species and growth rate respectively, f tb represents the best individual X tb The fitness value of

[0134] S44', determine whether: P tk >max(sr, A / K), sr represents the selection probability, which is a constant less than 1; A / K represents the ratio of the number of species to its limit value;

[0135] S45', if not, then individual X tk Mutations;

[0136] S46', individual X tk With the best individual X tb cross;

[0137] S47', if so, then X tb =X tk ,f tb =f tk ;

[0138] S48', determine whether: f tk >f gb ;

[0139] S49', if so, then X gb =X tk ,f gb =f tk ; If not, jump to step S410 ';

[0140] S410', random update X tk ;

[0141] S411', determine whether: k≧L t ;

[0142] S412', if yes, then update the number A and growth rate r of invasive species according to the logistic model;

[0143] S413', if not, then k=k+1, jump to step S43', and loop through steps S43' to S412';

[0144] S414′, determining whether the condition A≥η×K is satisfied; in this embodiment, the value range of η may be, for example, (1, A0 / K), where η represents the maximum ratio of the actual number of species to its limit number, and A0 / K represents the ratio of the initial population number to the limit number;

[0145] S415', if not, set t=t+1;

[0146] S416'、L t =[ln(K / A)], L t Indicates the number of optimization attempts at the tth iteration;

[0147] S417', if yes, then keep the L of the previous population t Jump to step S42';

[0148] S418', output the current optimal solution X gb .

[0149] like Figure 9 and Figure 10 As shown, in this embodiment, P is set at the same time c =A0·p c0 / A(t) and p m =p m0 ·r(t) / r0 adaptively searches for new feasible solutions, and its crossover mutation method refers to Figure 3 and Figure 4 .

[0150] In this embodiment, according to the characteristics of the problem, the first stage is a nested processing process and the second stage is an assembly welding process. Therefore, the heuristic indicator is to determine the processing sequence in the first stage to minimize the workpiece completion time target, and then determine the assembly product sequence in the second stage to optimize the overall indicator. The sequence of the two consecutive stages is used as the initial solution;

[0151] Step S5', based on the established mathematical problem model, estimate the lower bound of the bottleneck in stages. In this embodiment, first assume that in the first stage:

[0152]

[0153] Assume that the second stage:

[0154]

[0155] In this embodiment, the lower bound index is used to determine the optimization effect of the algorithm, and the overall lower bound is:

[0156] L = max{L1, L2}

[0157] As shown in FIG11 , in this embodiment, the optimal parameters of the algorithm system input are modified according to indicators such as the completion time of production scheduling and equipment utilization rate; the box diagrams of the modified heuristic algorithm and the improved genetic algorithm are shown in FIG11 . Figure 5 ;

[0158] Description of each sub-figure in Figure 11: Figure 11(a) to Figure 11(c) A box plot showing the algorithm comparison when the number of products, i.e. the number of storage boxes, is 10, 20, and 30 respectively; Figure 11(d) to Figure 11(f) Box plots showing algorithm comparisons when the number of machines is 5, 10, and 15 respectively; Figure 11(g) to Figure 11(i) These are box plots comparing algorithms when the rework probability is 0.05, 0.1, and 0.15 respectively; Figure 11(j) to Figure 11(o) Box plot showing algorithm comparison of six time generation schemes.

[0159] like Figure 12 As shown, in this embodiment, in step S6', the production data for verification is input, the problem is solved using the optimal parameter system, the optimal scheduling plan is output, and the actual plan is manually checked and adjusted;

[0160] like Figure 13 As shown in this embodiment, the optimization curve of the improved genetic algorithm on the two indicators of completion time and average passing time is referenced Figure 12 , refer to the final production scheduling Gantt chart Figure 13 , the optimization effects are 19.92% and 35.11% respectively.

[0161] In summary, the present invention systematically integrates data analysis, mathematical modeling of nested scheduling problems, and algorithm design applications to meet the processing needs of complex production workshops. It reconstructs genetic algorithms, constructs heuristic algorithms and lower bounds for scheduling problems, and solves the complex problem-solving of welding processes. Utilizing the present invention to formulate optimal production plans enables high-efficiency, low-cost production, shortening the overall processing completion time and average throughput time for such enterprises, effectively improving equipment utilization and reducing inventory. This system and method is suitable for production lines and enterprises with high production costs and customization, and is also instructive for optimizing the operation of similar production processes.

[0162] The present invention can intuitively and accurately describe the dynamics and randomness of the product processing process and equipment utilization by performing multi-layer encoding on the workpiece processing process with nested steps, and has the characteristics of simulation based on empirical data analysis and improves the accuracy of simulation.

[0163] This invention analyzes processing equipment information and buffer capacity to determine decision variable constraints, which are then incorporated into the scheduling formula to facilitate algorithmic solution, improving the model's simulation of the intelligent scheduling process for welding assembly. By establishing a mathematical model with cyclical characteristics, the invention reflects the reentrant nature of the production process and optimizes batch processing and nested process operations in tank welding.

[0164] The present invention has the advantages of selective feasible solution updating and iterative process acceleration to improve the quality of feasible solutions, and is also applicable to other complex scheduling problems, continuously reducing the space complexity and solution complexity of the problem.

[0165] This invention uses a constructed heuristic solution as the initial solution. The improved adaptive genetic algorithm accelerates algorithm search speed, improves the efficiency of exploring feasible solutions, and optimizes solution quality. The invention also designs an adaptive crossover and mutation operator based on nested coding, and adds a logistic control iteration process and a balance mechanism between local and global optima to increase the iteration speed of the genetic algorithm. This invention addresses the technical problems of existing technologies, such as flawed dynamic problem representation, poor simulation results, and poor expression accuracy and adaptability.

[0166] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A welding assembly intelligent scheduling method, characterized in that: The method comprises: S1. Implement preset nested process coding to determine the processing equipment information and buffer capacity required for welding; S2. Collecting historical production data, extracting production data and interference data of the welding process, fitting data features of the production data and the interference data, and determining a data distribution function accordingly; S3. Analyze the processing equipment information and the buffer capacity according to the data distribution function, and establish a dynamic mathematical description model of the welding process; S4. Constructing and implementing a model-solving algorithm system using a preset algorithm and the nested process code, wherein the preset algorithm includes: an improved genetic algorithm, a construction heuristic algorithm, and a lower bound estimation method based on problem model theory; solving the problem using the model-solving algorithm system, constructing a mathematical problem model based on the solution, generating random cases, and estimating the bottleneck lower bound in stages according to the dynamic mathematical description model of the welding process; S5. Comparing the model solving algorithm system with the bottleneck lower bound, and modifying applicable input parameters of the model solving algorithm system according to preset production scheduling indicators, thereby obtaining an applicable parameter system; S6. Inputting the verified production data into the applicable parameter system to obtain an applicable scheduling solution for intelligent scheduling of welding assembly; The step S3 comprises: S31. Calculate the processing time data and the disturbance time variation function based on the processing equipment information, and establish a dynamic mathematical description model based on the welding process to solve the completion time and average throughput time using the following logic: and Where, (i=1, 2) represents the i-th scheduling target, where represents the maximum completion time of production, The average transit time of each tank is the span between the earliest production start time of its parts and the completion time of the tank; S32. Fitting a probability function of random reentry based on the processing in the production history data, calculating the expected number of reentries, and adding the expected number of reentries to the real-time scheduling model to express a processing model related to the number of reentries; S33. Determine the decision variable constraints using the following logic: In the formula, suppose the lth process of the kth component of the ath tank is processed on the bth machine, Indicates the start time of the process in the tth processing. represents the end time of the process during the t-th processing. The first formula indicates that the start time of the next process must not be less than the end time of the previous process of the same component and the end time of other processes after the same processing equipment is used. The second formula indicates that the machine has rework operations, and the start time of rework is equal to the end time of the previous processing. The above decision variable constraints are added to the scheduling formula for solution. S34. Processing decision variables, solving objectives, and algorithm fitness accordingly based on the production demand in the production history data; In step S32, the probability function of random reentry is fitted using the following logic to calculate the expected number of reentry times: in, represents the probability of reentry of the decision variable, and t is the number of reentries.

2. The intelligent scheduling method for welding assembly according to claim 1, characterized in that: The step S1 comprises: S11. Determine the workpiece and its repetition number according to the nested process code, and compile it into an outer layer code; S12, determining the process number of each workpiece and determining it as an inner layer code; S13. Allocate the outer layer code and the inner layer code according to the preset processing position to obtain the processing equipment information and the buffer area capacity that meet the processing sequence constraint.

3. The intelligent scheduling method for welding assembly according to claim 1, characterized in that: The step S4 comprises: S41. Acquire and construct a heuristic rule based on the characteristics of the tank welding assembly scheduling problem, and apply the heuristic rule to the dynamic mathematical description model of the welding process in stages based on a preset processing process to obtain an initial scheduling solution; S42. Designing an adaptive crossover and mutation operator of the improved genetic algorithm based on the initial scheduling solution and the nested process code; S43, constructing and utilizing a logistic model to control the iterative process of the improved genetic algorithm and the balance mechanism between local optimum and global optimum; S44, taking different stages as bottleneck stages, calculating the lower bound of each bottleneck stage for scheduling, and estimating the overall lower bound accordingly.

4. The intelligent scheduling method for welding assembly according to claim 3, characterized in that: In step S43, the improved genetic algorithm is used to set a new iteration start mode and update mode, and the following logic is used to construct a logistic model to update the number of iterations and the population level: and In the formula, A(t) represents the population size in the t-th iteration process, K represents the limit size of the population, represents the initial population size of the species, represents the initial natural growth rate of the population, e represents the natural exponent (=2.71828…), and r(t) represents the population growth rate during the t-th iteration.

5. The intelligent scheduling method for welding assembly according to claim 4, characterized in that: In step S43, a new feasible solution is adaptively searched based on the initial scheduling solution using the following search logic: as well as Where, represents the probability of chromosome crossover, represents the initial crossover probability of chromosomes, represents the probability of chromosome mutation, represents the initial mutation probability of chromosome.

6. The intelligent scheduling method for welding assembly according to claim 3, characterized in that: The S44 includes: S441. Assume the first stage using the following logic to obtain the lower bound indicator of the bottleneck in the first stage: Where, represents the maximum completion time of production, represents the lower bound of the optimization problem with assumptions made on the first stage, represents the lth process of the kth component of the ath tank, is a binary value, if Processed on the bth machine, is 1, otherwise 0; Indicates the number of all machines; represents the time consumption of the lth process of the kth component of the ath tank, represents the a-th tank, represents the passing time of the a-th tank; S442. Assume the second stage using the following logic to obtain the lower bound indicator of the bottleneck in the second stage: S443: Determine the bottleneck lower bound using the first-stage bottleneck lower bound indicator and the second-stage bottleneck lower bound indicator.

7. The intelligent scheduling method for welding assembly according to claim 6, characterized in that: In step S443, the bottleneck lower bound is determined according to the bottleneck lower bound indicator of the first stage and the bottleneck lower bound indicator of the second stage using the following logic: 。 8. A welding assembly intelligent scheduling system, characterized in that: The system comprises: Nested coding implementation module, used to implement preset nested process coding to determine the processing equipment information and buffer capacity required for welding; A production history data fitting processing module is used to collect production history data, extract production data and interference data of the welding process, fit the data features of the production data and the interference data, and determine the data distribution function accordingly; a welding process description model construction module, configured to analyze the processing equipment information and the buffer area capacity according to the data distribution function and establish a dynamic mathematical description model of the welding process, wherein the welding process description model construction module is connected to the nested coding implementation module and the production history data fitting processing module; A solution system construction and lower bound estimation module is used to construct and implement a model solution algorithm system using a preset algorithm and the nested process code, wherein the preset algorithm includes: a lower bound estimation method including an improved genetic algorithm, a construction heuristic algorithm, and a problem model theory, and solves by using the model solution algorithm system to construct a mathematical problem model and generate random cases to estimate the bottleneck lower bound in stages according to the dynamic mathematical description model of the welding process. The solution system construction and lower bound estimation module is connected to the welding process description model construction module; a solution system correction module, configured to compare the model solution algorithm system with the bottleneck lower bound, and to correct applicable input parameters of the model solution algorithm system according to preset production scheduling indicators, thereby obtaining an applicable parameter system. The solution system correction module is connected to the solution system construction and lower bound estimation module; A scheduling solution module inputs the verified production data into the applicable parameter system to solve and obtain a scheduling solution applicable to welding assembly intelligent scheduling, and the scheduling solution module is connected to the solution system correction module; Analyzing the processing equipment information and the buffer capacity according to the data distribution function to establish a dynamic mathematical description model of the welding process includes: S31. Calculate the processing time data and the disturbance time variation function based on the processing equipment information, and establish a dynamic mathematical description model based on the welding process to solve the completion time and average throughput time using the following logic: and Where, (i=1, 2) represents the i-th scheduling target, where represents the maximum completion time of production, The average transit time of each tank is the span between the earliest production start time of its parts and the completion time of the tank; S32. Fitting a probability function of random reentry based on the processing in the production history data, calculating the expected number of reentries, and adding the expected number of reentries to the real-time scheduling model to express a processing model related to the number of reentries; S33. Determine the decision variable constraints using the following logic: In the formula, suppose the lth process of the kth component of the ath tank is processed on the bth machine, Indicates the start time of the process in the tth processing. represents the end time of the process during the t-th processing. The first formula indicates that the start time of the next process must not be less than the end time of the previous process of the same component and the end time of other processes after the same processing equipment is used. The second formula indicates that the machine has rework operations, and the start time of rework is equal to the end time of the previous processing. The above decision variable constraints are added to the scheduling formula for solution. S34. Processing decision variables, solving objectives, and algorithm fitness accordingly based on the production demand in the production history data; In step S32, the probability function of random reentry is fitted using the following logic to calculate the expected number of reentry times: in, represents the probability of reentry of the decision variable, and t is the number of reentries.

Images