Multi-constraint task flow optimization method for multi-physics simulation of automobiles and the like

Abstract

The present application belongs to the technical field of complex equipment simulation, and particularly relates to a multi-constraint task flow optimization method for automobile multi-physical field simulation, which comprises the following steps: constructing an automobile multi-physical field simulation task optimization model, generating an initial simulation flow scheme set, generating an initial simulation flow scheme set, evaluating the technical feasibility and efficiency of the flow scheme, iteratively optimizing and repairing constraints based on a swarm intelligence algorithm, and outputting an optimal automobile multi-physical field simulation flow scheme. The present application constructs a mathematical programming model of the multi-specialty collaborative flow optimization problem in automobile multi-physical field simulation, and solves the optimization model by using an improved genetic algorithm, so as to minimize the total time for completing all tasks under the constraints of the sequence dependency relationship between different professional simulation tasks, the partial task deadline constraint, and the available resource constraint. The method of the present application can also be directly or adaptively adjusted and applied to the simulation flow optimization of other complex equipment, such as aerospace vehicles, ships, heavy machinery, etc.

Description

Technical Field

[0001] This invention belongs to the field of complex equipment simulation technology, specifically involving a multi-constraint task process optimization method for automotive multiphysics simulation. Background Technology

[0002] In the design and analysis of complex equipment (such as automobiles, aerospace vehicles, and ships), iterative simulation analyses involving multiple disciplines such as fluid dynamics, solid-state dynamics, thermal dynamics, electrical dynamics, and magnetism are often required to optimize the design performance across multiple disciplines. Taking the development of modern automobiles as an example, its forward design process heavily relies on simulation analyses covering multiple physical fields, including structural mechanics, fluid dynamics, thermal management, and electromagnetic compatibility, to predict and optimize the vehicle's safety, economy, comfort, and reliability. This iterative process frequently involves different physical laws governing equations, geometric modeling of different simulation objects, model discretization, and numerical solutions. These multi-disciplinary simulation tasks have vastly different computational loads and complex cross-disciplinary data dependencies (for example, automotive crash safety analysis depends on the results of vehicle body structure analysis, while battery thermal management analysis needs to be coupled with the overall vehicle thermal flow field and battery electrochemical characteristics). This makes it difficult to rationally plan and effectively coordinate the multi-disciplinary simulation task process, severely impacting the efficiency of product (especially highly integrated and complex equipment like automobiles) design and analysis iterations. The complexity stems from: (1) Pre-task constraints: There is a strict pre-execution order between tasks in the workflow.

[0003] (2) Task deadline constraints: Some tasks must be completed before the set time.

[0004] (3) Resource availability constraints: Some resources (workgroups, computer resources) may not be able to complete certain tasks.

[0005] Traditional process design methods typically rely on human experience or heuristic rules for task sequencing and resource allocation. While these methods are simple to implement and have low computational overhead, their applicability and stability are significantly limited when dealing with process design problems involving multiple tasks, multiple resources, and complex pre-dependencies. Especially when there are constraints on task sequence, resource availability, and multi-objective optimization requirements, traditional methods often can only be adjusted locally, making it difficult to achieve an effective balance between solution quality and constraint satisfaction.

[0006] To overcome the aforementioned problems, metaheuristic optimization methods such as genetic algorithms have been gradually introduced into the field of process design and scheduling optimization, utilizing swarm search and stochastic evolution mechanisms to improve the ability to explore the solution space. However, existing related solutions mostly focus on objective function optimization, and constraint handling usually adopts penalty functions or simple repair strategies. They lack a systematic mechanism that can simultaneously and collaboratively repair the task execution order and resource allocation results, thus making it difficult to strictly guarantee the complete satisfaction of hard constraints such as pre-dependent relationships and resource availability in complex scenarios. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention proposes a multi-constraint task process optimization method for multiphysics simulations such as automotive. This method models a process optimization problem with the objective of minimizing task completion time under multiple complex constraints, including task pre-dependency constraints, resource availability constraints, and multi-task deadline constraints. It also proposes a multi-task process optimization method based on an improved genetic algorithm to solve the mathematical programming model. This invention constructs a mathematical programming model for the multi-disciplinary collaborative process optimization problem in automotive multiphysics simulation and uses an improved genetic algorithm to solve the optimization model. This minimizes the total completion time of all tasks under the constraints of sequence dependency relationships, partial task deadline constraints, and available resource constraints (workgroups, computer resources) among different professional simulation tasks. This invention's method can also be directly or adaptively applied to the simulation process optimization of other complex equipment, such as aerospace vehicles, ships, heavy machinery, and other complex equipment fields with multi-disciplinary collaborative simulation requirements.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The optimization method for multi-constraint task flow in automotive multiphysics simulation comprises the following steps: Step (1) Construct an optimization model for the multiphysics simulation task of automobiles Based on the simulation requirements of automotive R&D, a simulation process to be optimized is defined. This process includes multiple simulation tasks covering different automotive physics fields, specifically two or more of the following: automotive body structural mechanics, fluid dynamics, thermal management, and electromagnetic compatibility. A mathematical model of this simulation process is then developed: (1.1) Design variables Assume there is Each simulation task needs to be assigned to This can be accomplished by a workgroup (e.g., computer resources): (1) Design variables It is the execution sequence of each task within the working group. This indicates a task. Assigned to work group : (2) by =5, Taking =2 as an example, [1, 3, 4, 5, 2] and [1, 2, 2, 1, 2] represent the execution order of simulation tasks and the work group allocation scheme, respectively. It means that work group 1 executes simulation task 1 first, and then executes simulation task 5; work group 2 executes simulation task 3 first, then executes simulation task 4, and finally executes simulation task 2.

[0009] (1.2) Objective function: The objective function is to minimize the total time spent completing all tasks, with the optimization objective being to minimize the total simulation cycle. Penalties are also considered for tasks that violate constraints. (6) in, This represents the maximum completion time for all tasks. These are all penalty coefficients, used to control the severity of penalties for violating constraints; This indicates the total timeout for violating pre-requirement constraints. Represents the set of preconditions. Indicates task Completion time, Indicates task Completion time; This indicates the total timeout for violating the task deadline constraint. Indicates task Completion time, Indicates task The deadline; This indicates the total timeout for violating resource availability constraints. This indicates whether the corresponding task can be allocated feasible resources. Time indicates resources Unable to process task ,at this time It must be 0; when This indicates that the resources are feasible.

[0010] (1.3) Constraints (1.3.1) Pre-task constraints: Based on the data dependencies between various automotive physics simulation tasks (e.g., the electromagnetic analysis simulation task depends on the data information output by the thermal management system analysis simulation task after completion), the pre-task execution order is defined to ensure that downstream simulation tasks can only begin after the upstream tasks they depend on have been completed. (3) Equation (3) represents the task Only in its preceding tasks It can only begin after completion. For the set of prerequisite constraints, if... , indicating task Dependency Task Finish.

[0011] (1.3.2) Resource Availability Constraints: Based on the simulation software and computing hardware requirements of each automotive physics simulation task, define the set of available simulation workgroups, i.e., resource availability constraints: (5) (1.3.3) Project Deadline Constraints: Based on the completion milestones of the automotive R&D project, mandatory deadlines are set for key simulation tasks. (4) Equation (4) represents the task set. Each task must be completed before its designated deadline.

[0012] Step (2) Generate an initial simulation process scheme set (2.1) The simulation process scheme is represented by a unified code, which includes: (a) Task execution sequence: The execution order that characterizes the data dependencies between various automotive physics simulation tasks; (b) Resource allocation sequence: Characterized by the specific simulation workgroup (computer resources) allocated to each simulation task; (2.2) The initialization process includes: first, generating a random task sequence; then, performing topological sorting based on the technical data dependencies between tasks to obtain a basic task sequence that satisfies all task preconditions; and then, for each task in the sequence, determining a resource allocation scheme by assigning the task with the shortest execution time to the work group from the set of feasible resources that meet the requirements.

[0013] Step (3) Technical feasibility and efficiency assessment of the process plan The process schemes obtained in step (2) are simulated and evaluated, specifically including: (3.1) Simulate the process corresponding to the code described in step (2) and calculate the total simulation cycle required to complete all automotive multiphysics simulation tasks; (3.2) Detect the violation of task pre-constraints, resource availability constraints, and project deadline constraints in the detection process plan, and calculate the corresponding penalty value, i.e. , and ; (3.3) Combining the total simulation cycle time and the total penalty value, the fitness evaluation value of the process scheme is obtained, i.e. the objective function, which is used to characterize its technical feasibility and execution efficiency.

[0014] Step (4) Iterative optimization and constraint repair based on swarm intelligence algorithm By deeply integrating genetic manipulation with constraint repair mechanisms, we ensure that new individuals rapidly converge to feasible solutions during evolution. The main operations are as follows: (4.1) Crossover operation: (4.1.1) Task execution sequence crossover: Perform two-point crossover on the task execution sequence of two parent individuals with a preset probability (e.g., 80%). After exchanging the gene segments between the two crossover points, for the missing task genes in the offspring, fill in the blank positions according to their original order in the parent sequence to generate a new sequence.

[0015] (4.1.2) Workgroup (resource) allocation sequence crossover: Generate a random number for each task. If the random number is greater than a preset threshold (e.g., 0.5), then exchange the resource allocation information of the two parent generations on the task; otherwise, the child generation inherits the resource allocation information of the current parent generation.

[0016] (4.1.3) Post-crossover repair: Constraint detection is performed on the new individuals generated by the crossover operation. If the sequence of task execution violates the technical data dependency constraint, the sequence repair mechanism is initiated. For individuals that violate the constraint, first consider an arrangement of N tasks and name it set Z. Consider the predecessor task of each task and generate an empty set of the same size as Z and name it set C. The specific process is as follows: First, traverse the elements in set Z. If the predecessor task of a certain task is a member of set C, add it to set C and continue traversing from the beginning until set Z is empty, thus obtaining a feasible task sequence that does not violate the constraint. Then, following the same steps as initializing the chromosome, first obtain its feasible resource list, then select the resource with the shortest execution time in the list with a 70% probability and randomly select it with a 30% probability, thereby synchronously repairing the resource allocation sequence to ensure that the individual satisfies all hard constraints, and finally obtain a feasible offspring crossover individual.

[0017] (4.2) Mutation operation: Perform mutation on the selected individuals with a preset probability (e.g., 30%): (4.2.1) Task execution sequence mutation: Randomly swap two gene positions in the task execution sequence. If the swap results in a violation of technical data dependency constraints, the above sequence repair mechanism is immediately invoked to repair it.

[0018] (4.2.2) Workgroup allocation sequence mutation: For two tasks whose positions are swapped, their resource allocation information is inherited from the resource information before mutation with a certain probability (e.g., 50%). Otherwise, a different resource is randomly selected from their feasible resource set to replace it. After mutation, constraint detection and repair are performed on the individual again to ensure its feasibility.

[0019] (4.3) Elite preservation strategy: In each generation of population evolution, the individual with the best current fitness value is directly preserved in the next generation of population to prevent the loss of excellent genes and ensure the convergence of the algorithm.

[0020] (4.4) Repeat steps (4.1) to (4.3) to perform iterative optimization for a preset number of generations. In each generation, a new population is generated through selection, crossover, and mutation, and the feasibility of the population is ensured by using a constraint repair mechanism. At the same time, the population is guided to evolve towards a better solution by elite retention.

[0021] Step (5) Output the optimal multiphysics simulation process for automobiles. When the iteration reaches the preset number of times or the convergence condition is met, the optimization is terminated, and the automotive multiphysics simulation process scheme with the best fitness evaluation value among all the process schemes is output. The optimal automotive multiphysics simulation process scheme with the best fitness evaluation value clarifies the execution order of each automotive multiphysics simulation task, the allocated heterogeneous simulation computing resources, and the start and end times of each task, which is used to guide and control the actual execution of the simulation platform.

[0022] Beneficial effects of this invention: Compared with existing technologies, this invention addresses the problems of complex task dependencies, strict resource constraints, and the tendency of traditional optimization methods to generate numerous infeasible solutions in multi-disciplinary collaborative processes for automotive multiphysics simulation. It proposes a process optimization method that introduces constraint detection and joint repair during the iterative optimization of the process solution. By treating task dependency constraints and resource availability constraints as hard constraints that must be satisfied, the process solution remains feasible throughout the entire optimization process, guiding the optimization process to obtain a high-quality solution. This method achieves collaborative optimization of task execution order and resource allocation, improving the stability and optimization efficiency of the process solution. It is applicable to multi-disciplinary collaborative simulation process optimization scenarios under complex constraints and has good versatility and engineering application value. Attached Figure Description

[0023] Figure 1 This is an overall flowchart of the multi-constraint task process optimization method for automotive multiphysics simulation of the present invention; Figure 2 This is a schematic diagram illustrating the calculation principle of the fitness function with penalty; Figure 3 This is a diagram illustrating the first sub-vector intersection; Figure 4 This is a diagram illustrating the intersection of the second sub-vector; Figure 5 This is the result obtained by the genetic algorithm (shown as a Gantt chart); Figure 6 This is the result obtained by the particle swarm optimization algorithm (shown as a Gantt chart). Detailed Implementation

[0024] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and technical solutions.

[0025] The overall flowchart of the multi-constraint task process optimization method for automotive multiphysics simulation of the present invention is as follows: Figure 1 As shown.

[0026] This embodiment optimizes the simulation process for multiphysics coupling in the automotive field. It decomposes the automotive multiphysics simulation process into 30 interrelated simulation tasks, covering key areas such as vehicle body structure, fluid dynamics, thermal management, and electromagnetic compatibility. Specifically, the process comprises 30 simulation tasks, executed simultaneously by 5 working groups (M1, M2, M3, M4, M5). Task execution time is measured in days. The execution schedule, deadline constraints, pre-task constraints, and unavailable resource constraints for all tasks are shown in Tables 1, 2, 3, and 4, respectively.

[0027] Table 1 is the task execution schedule. For example, the second row of data indicates that the completion time for simulation task 1 on workgroups M1, M2, M3, M4, and M5 is 8 days, 7 days, 6 days, 5 days, and 4 days, respectively. Table 2 is the task deadline schedule. For example, the second row indicates that the completion time for task 10 must be within 40 days. Table 3 is the task prerequisite constraint table. For example, the second row of simulation task J3 and prerequisite task 1 indicates that simulation task 3 can only be performed after simulation task 1 is completed. The seventh row of simulation task J8 and prerequisite tasks 1 and 6 indicates that simulation task 8 can only be performed after both simulation task 1 and simulation task 6 are completed. Table 4 is the unavailable resource constraint table. For example, the second row of task 1 and resource 3 indicates that resource (workgroup) M3 cannot execute simulation task 1, and so on.

[0028] Table 1 Task Execution Schedule (in days)

[0029] Table 2 Task Deadline Schedule

[0030] Table 3 Task Prerequisite Constraints

[0031] Table 4 Unavailable Resource Constraints

[0032] Below are 30 simulation tasks in the automotive multiphysics simulation workflow: 1. Static analysis of vehicle body structure: Analysis of stress and strain distribution of the vehicle body under static loads. 2. Aerodynamic flow field analysis: Simulating the airflow distribution and drag coefficient of the vehicle at different speeds. 3. Fluid-thermal coupling analysis of the thermal management system: assessing the impact of the cooling system on the overall vehicle thermal management. 4. Electromagnetic field analysis of the motor: Calculate the electromagnetic field distribution and electromagnetic force of the motor. 5. Electromagnetic compatibility analysis of electronic systems: Evaluate the electromagnetic interference and immunity capabilities of onboard electronic systems. 6. Full-vehicle frontal collision structural analysis: Simulating structural deformation and energy absorption during a frontal collision. 7. Hydraulic pressure analysis of the braking system: Evaluate the pressure distribution of the braking system under different operating conditions. 8. Gear meshing analysis of the transmission system: Analyze the meshing characteristics of gearbox gears under different loads. 9. Electrochemical-thermal coupling analysis of power batteries: evaluating the thermal behavior of batteries during charge and discharge processes. 10. Dynamic Vibration and Noise Analysis of Vehicle Doors: Analyze the vibration and noise characteristics of vehicle doors at different speeds. 11. Signal Integrity Analysis of Vehicle Electronic Systems: Assessing Signal Transmission Quality and Interference. 12. Airflow field analysis of air conditioning system ducts: Optimizing air conditioning system duct design to improve comfort. 13. Dynamics Analysis of Wheel Suspension System: Analyzing the impact of the suspension system on vehicle handling. 14. Vehicle Body Modal Analysis: Identifying Key Modal Frequencies and Mode Shapes of the Vehicle Body 15. Battery Pack Thermal Management Optimization Analysis: Optimize the battery pack cooling system design to reduce temperature. 16. Electromagnetic Characteristics Analysis of Motor Stator Slot Type: Analyzing the Influence of Different Stator Slot Types on Motor Performance 17. Motor winding temperature field analysis: Evaluate the temperature distribution of the motor windings under different operating conditions. 18. Motor NVH Performance Analysis: Evaluate the noise and vibration characteristics of the motor at different speeds. 19. Battery Pack Structural Strength Analysis: Analyze the structural strength of the battery pack under impact and vibration conditions. 20. Simulation of Battery Thermal Runaway Propagation: Simulating the propagation process of battery thermal runaway within a battery pack. 21. Power Integrity Analysis of Electronic Systems: Evaluating the stability of power systems under different loads. 22. Electromagnetic Radiation Analysis of Electronic Systems: Evaluating the electromagnetic radiation emitted by electronic systems to the surrounding environment. 23. Fluid-thermal coupling analysis of motor cooling system: Optimizing motor cooling system design 24. Temperature Field Distribution Analysis of Vehicle Body Materials: Analyze the temperature distribution of vehicle body materials under different environmental conditions. 25. Fatigue Life Analysis of Vehicle Body Structure: Evaluating the fatigue life of the vehicle body structure under long-term use. 26. Strength Analysis of Motor Rotor Structure: Analyze the structural strength of the motor rotor under high-speed rotation. 27. Battery Pack Vibration Characteristics Analysis: Analyze the vibration characteristics of the battery pack under different road surface conditions. 28. Battery Pack Thermal Management Solution Optimization: Optimize the design of the battery pack thermal management system. 29. Electromagnetic-Vibration-Noise Coupling Analysis of Motors: Analyzing the coupling relationship between the electromagnetic field of a motor, structural vibration, and noise. 30. Multiphysics Collaborative Optimization of the Whole Vehicle: Integrating simulation results from various subsystems to optimize the overall vehicle performance.

[0033] Step 1: Chromosome Encoding Method In this embodiment, to simultaneously describe the execution order and resource allocation of simulation tasks, each process scheme is encoded using a chromosome. The chromosome consists of two sequences of equal length: the first sequence represents the task execution order, and the second sequence represents the work group to which the corresponding task is assigned. Taking 10 tasks and 3 work groups as an example (as shown in Table 5), Task 1, the static analysis of the vehicle body structure, will be assigned to work group 2; Task 2, the aerodynamic flow field analysis, will be assigned to work group 1, and so on. Tasks appearing earlier in the sequence are those that are executed first. In the resource allocation sequence, tasks at the same position in the task order corresponding to each gene locus are executed by the designated work group. This chromosome can completely represent the simulation process scheme, achieving unified encoding of task execution order and resource allocation, providing a data foundation for subsequent constraint detection, repair, and optimization.

[0034] Table 5 Chromosome data structure (N=10, M=3 as an example)

[0035] Step 2: Initialize the process scheme set First, a population is generated. This step begins by randomly generating a sequence of tasks. This sequence is completely random, for example, […30, 4, 3, 8, 1…]. This chromosome-based approach is not feasible because vehicle structure analysis (Task 1) is fundamental and must precede collision safety analysis (Task 8). Electromagnetic analysis (Task 4) relies on thermal management system analysis (J3) to perform topological sorting on the randomly generated task sequence to obtain the basic sequence. Optimization and comprehensive analysis (J30) is the final stage and depends on all preceding critical tasks. Therefore, a topological sorting is needed on the initially randomly generated task sequence. Topological sorting means arranging all tasks in a row, placing each task after its dependent task. After topological sorting, the sequence becomes […1, 3, 4, 8, 30]. This task sequence satisfies the following conditions: vehicle structure analysis (Task 1) is fundamental and must precede collision safety analysis (Task 8); electromagnetic analysis (Task 4) relies on thermal management system analysis (J3) to perform topological sorting on the randomly generated task sequence to obtain the basic sequence; and optimization and comprehensive analysis (J30) is the final stage and depends on all preceding critical tasks. Then, resources are allocated to each task. First, a list of feasible resources is obtained. For example, feasible work teams are selected for task 1 (static analysis of vehicle body structure). However, due to resource limitations, work team 3 cannot complete the static analysis of the vehicle body structure. Therefore, feasible work teams are selected from the remaining four work teams: [1,2,4,5]. Since the resources of each work team are different and the experience and manpower of the engineers are different, the time for each work team to complete the task is also different. The time for the static analysis of the vehicle body structure to be completed by work teams [1,2,4,5] is [8,7,5,4], respectively. Then, the resource with the shortest execution time among the work teams is selected with a 70% probability, and a random selection is made with a 30% probability. The above operation is repeated to obtain an initial population containing 2000 process scheme chromosomes.

[0036] Step 3: Individual Assessment For each individual in the population, the fitness evaluation function is called to calculate its fitness, such as... Figure 2 As shown. This function simulates a multi-task workflow and calculates the total completion time. Simultaneously, it iterates through all tasks with deadlines and tasks that violate unavailable resource constraints and preconditions, accumulating penalty values ​​(timeout amounts) for tasks that violate constraints. 100). Total fitness = Total completion time + Total penalty value. For example, an individual in the population obtained after initialization satisfies all task preconditions and resource availability constraints. The total completion time of all tasks is 82, that is, the completion time of task 30 optimization and synthesis analysis is also 82. However, the deadline for task 30 is 80. In this case, the total fitness of this individual = total completion time + total penalty value = 82 + 2. 100 = 282, and 282 is the total fitness of the individual.

[0037] Step 4: Perform genetic operations on the selected chromosomes. (1) When using an improved genetic algorithm to solve the process optimization problem, perform genetic operations and constraint repair: (1.1) Intersection: such as Figure 3 As shown (taking N=8, M=3 as an example), the first sub-vector crossover is performed. The crossover process is carried out on the two selected parent individuals with an 80% probability. First, two non-repeating crossover points are selected for the two parents, and the gene sequences in the middle part of the two crossover points are swapped (task 5 of parent one: electromagnetic compatibility analysis of electronic system, task 8: gear meshing analysis of transmission system, task 2: aerodynamic flow field analysis, and task 1 of parent two: static analysis of vehicle body structure, task 3: fluid-thermal coupling analysis of thermal management system, and task 6: frontal collision structure analysis of whole vehicle). Then, for genes that are not present in the offspring, they are directly moved from the corresponding positions of the parents. Finally, for missing gene sequences, they are directly filled into the blank task positions of the offspring according to the task order of the parents, and finally the crossover result of the first sub-vector is obtained.

[0038] like Figure 4 As shown (using N=8, M=3 as an example), the second sub-vector crossover is performed. First, a random number array [0,1] is generated. If the random number for this task is greater than 0.5, the parent machine information is swapped; otherwise, the machine information of the current parent is retained. For example, if the random number for the electromagnetic compatibility analysis of the electronic system in Task 5 is 0.22, which is less than 0.5, then the parent work team information is retained, meaning that work team 2 still handles the electromagnetic compatibility analysis of the electronic system in Task 5.

[0039] The offspring obtained through the above crossover process may violate constraints. Therefore, after crossover, the tasks that violate the preconditions and resource availability constraints should be sequenced to obtain a feasible task sequence.

[0040] For individuals that violate the constraints, first consider an arrangement of N tasks and name it set Z. Consider the prerequisite tasks for each task and generate an empty set of the same size as Z and name it set C. The specific operation is as follows: First, traverse the elements in set Z. If a prerequisite task of a task is a member of set C, add it to set C and continue traversing from the beginning until set Z is empty, thus obtaining a feasible sequence of tasks that do not violate the constraints. Then, following the same steps as initializing the chromosome, first obtain its feasible resource list, then select the resource with the shortest execution time in the list with a 70% probability, and randomly select the feasible offspring crossover individuals with a 30% probability.

[0041] (1.2) Mutation: First, a chromosome is randomly selected from the population, and mutation is performed on the individual with a 30% probability. First, two genes in the task sequence are randomly swapped. If the swap violates the predecessor constraint, the above-mentioned correction mechanism is used for sequence repair. Suppose an individual is mutated, and two genes in the task sequence are randomly selected. These two genes are Task 4 Electromagnetic Analysis and Task 3 Thermal Management System Analysis. At this time, Task 3 is analyzed before Task 4 and does not violate the predecessor constraint. However, after the swap, the predecessor constraint is violated, so sequence repair is required. For the resource allocation sequence, except for the swapped Task 3 and Task 4, the rest are directly inherited from the parent. For the swapped Task 3 Thermal Management System Analysis and Task 4 Electromagnetic Analysis, there is a 50% probability that the parent's work team is inherited. Otherwise, one is randomly selected from the set of feasible work teams (which needs to be different from the original). After the above operation, the mutated chromosome is obtained. After mutation, the above-mentioned sequence repair needs to be performed.

[0042] (1.3) Elite preservation: The best individuals in each generation are directly preserved to the next generation to prevent the loss of excellent genes.

[0043] (2) When using particle swarm optimization to solve process optimization problems, particle swarm operations and constraint repair can also be performed: (2.1) Compare the fitness value of each particle with the best position it has ever experienced, and if it is better, take it as the current best position; (2.2) Compare the fitness values ​​of all particles with the global optimum. If a particle has a better fitness, update the position and fitness value of the global optimum. (2.3) Update the velocity and position of the particles, decode the continuous position vector into a discrete process scheme, and finally call the sequence repair function to repair the pre-constraint violation and unavailable resource problem in the sequence.

[0044] Step 5: Iteration and Output Repeat steps 3 and 4 for a total of 100 generations. After the evolution is complete, output the individual with the best fitness across all generations. Perform a final evaluation and verification of the optimal solution, confirming that it satisfies all prerequisites and resource constraints, and output its final total completion time, the start and end times of each task, and resource allocation details. Finally, output the optimal process file to a CSV file, such as... Figure 5 and 6The results are presented for the genetic algorithm and the particle swarm optimization algorithm, respectively. Under the conditions of the above examples, the obtained process schemes all satisfy the task pre-dependency constraints and resource availability constraints. Taking the genetic algorithm results as an example, the deadline for Task 10, door dynamic vibration and noise analysis, is 30 seconds; the deadline for Task 20, battery thermal runaway propagation simulation, is 50 seconds; and the deadline for Task 30, whole vehicle multiphysics collaborative optimization, is 80 seconds. The completion times of the three tasks are 23, 43, and 66 seconds, respectively, which meet their deadline requirements. The results show that both algorithms can obtain feasible and stable process schemes within a finite number of iterations, effectively achieving the collaborative optimization of task execution order and resource allocation, and verifying the feasibility and effectiveness of the method of this invention in complex multi-constraint simulation process optimization problems.

[0045] This invention addresses the characteristics of multi-disciplinary collaborative process optimization problems in complex equipment simulation. It selects a swarm intelligence algorithm—Genetic Algorithm (GA) and Particle Swarm Optimization (PSO)—as the core solution framework for multi-task process optimization under multiple resource constraints. This problem is a typical combinatorial optimization problem, aiming to minimize the overall task completion time and penalty term for constraint violations while satisfying pre-task constraints, resource availability constraints, and deadline constraints, thereby obtaining a near-globally optimal process solution. In summary, this invention constructs an efficient and scalable collaborative simulation optimization solution framework by applying an improved genetic algorithm and particle swarm optimization algorithm to multi-constraint task process optimization problems. This framework ensures solution feasibility while obtaining a near-globally optimal process solution within a reasonable computation time, achieving efficient optimization and rational resource allocation for multi-disciplinary collaborative processes in complex equipment simulation. It provides important theoretical and technical support for intelligent process optimization in complex system design and simulation cloud platforms.

Claims

1. A multi-constraint task flow optimization method for automotive multiphysics simulation, characterized in that, The steps are as follows: Step (1) Construct an optimization model for the multiphysics simulation task of automobiles Based on the simulation requirements of automotive R&D, a simulation process to be optimized is defined. This process includes multiple simulation tasks covering different automotive physical fields, specifically two or more of the following: automotive body structural mechanics, fluid dynamics, thermal management, and electromagnetic compatibility. A mathematical model of this simulation process is then developed. (1.1) Design variables It has Each simulation task needs to be assigned to A working group will be established to complete this task. ; Design variables Indicates task Assigned to work group ; Define the task-resource execution time matrix , indicating task In the working group Execution duration; (1.2) Objective function: Minimize the total simulation cycle, while also considering penalties for tasks that violate constraints: in, The maximum completion time for all tasks; , , This is the penalty coefficient; First penalty item middle, For the set of pre-constraints, For the task The start time, For the task Completion time; Second penalty item middle, For the task Completion time, For the task The deadline; Third penalty item middle, Indicates task By working group implement, Indicates working group Do not execute; (1.3) Constraint Definition Prerequisites for the task: Resource availability constraints: Project deadline constraints: Among them, the task Completion time ; Step (2) Generate an initial simulation process scheme set (2.1) The simulation process scheme is represented by a unified code, which includes: (a) a task execution sequence, representing the dependent execution order between tasks; and (b) a resource allocation sequence, representing the specific work group to which each task is allocated. (2.2) First, a random task sequence is generated, and topological sorting is performed based on the data dependencies between tasks to obtain a basic task sequence that satisfies the preconditions; then, for each task in the sequence, according to the matrix... Resource allocation is determined primarily based on the shortest execution time from its set of feasible resources; Step (3) Technical feasibility and efficiency assessment of the process plan (3.1) Based on the work groups assigned to each task and the matrix described above Determine the execution time of each task and calculate the total simulation cycle required to complete all tasks; (3.2) Detect the violation of task pre-constraints, resource availability constraints and project deadline constraints by the detection process scheme, and calculate the corresponding penalty value; (3.3) Combining the total simulation cycle and the total penalty value, the fitness evaluation value of the process scheme is obtained; Step (4) Iterative optimization and constraint repair based on genetic algorithm Perform selection, crossover, and mutation operations on the current population, and perform constraint repair on the generated individuals, retaining elite individuals, and repeat the iteration until the termination condition is met; Step (5) Output the optimal simulation process scheme After terminating the optimization, the simulation process scheme with the best fitness is output.

2. The method according to claim 1, characterized in that, The cross operation and constraint repair described in step (4) are as follows: The crossover operation includes: performing two-point crossover on the task execution sequence with a preset probability, exchanging gene segments between the two crossover points, and filling the missing task genes in the offspring into the blank positions according to the original order of the parents; performing uniform crossover on the work group allocation sequence with a preset threshold, and determining whether to exchange resource allocation information based on a random number greater than the threshold. The constraint repair process includes: Let set Z be a permutation of N tasks, and set C be an empty set of the same size as Z; traverse the elements in Z, and if all the predecessor tasks of a certain task belong to C, then move the task into C and remove it from Z, repeating the traversal until Z is empty, thus obtaining a sequence of tasks that satisfy the predecessor constraints; then obtain a list of feasible resources, and select the resource with the shortest execution time with a 70% probability and a 30% probability of random selection, simultaneously repairing the resource allocation sequence.

3. The method according to claim 1, characterized in that, The genetic algorithm described in step (4) is replaced by the particle swarm optimization algorithm, specifically: The fitness value of each particle is compared with its historical best position and updated. The fitness values ​​of all particles are compared with the global best value and updated. The velocity and position of the particles are updated. The continuous position vector is segmented and decoded into a task sequence and resource allocation scheme. Then, the sequence repair function is called to repair the problems of prior constraint violation and unavailable resources.

Images