A multi-mode simulation and adaptive decision method for aircraft mixed-line assembly

Abstract

The application discloses a kind of multi-mode simulation and adaptive decision method for aircraft mixed-line assembly, belong to system assembly simulation analysis field.The application constructs unified simulation model including entity set, activity set and future event list, establishes model initialization mapping function, and according to order simulation mode, online simulation mode or plan simulation mode, external input data is mapped to initial state or initial future event list of entity;When entity state changes trigger decision, the score value of each decision rule is calculated based on observation set, and the rule with the highest score value is selected for execution;According to the reward function, the decision reward value is calculated, the decision sample is extracted and stored in the sample library, and the sample is extracted at fixed time intervals to update the score parameters.The application can realize order evaluation, risk prediction and plan verification under unified model architecture, and improve the prediction accuracy and decision availability in mixed-line assembly scenario.

Description

Technical Field

[0001] This invention relates to the field of assembly system simulation analysis, and specifically to a multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft. Background Technology

[0002] Aircraft mixed-line assembly processes are characterized by parallel production of multiple configurations, significant process differences, complex constraints on production factors (such as professionals, tooling, equipment, and materials), and frequent on-site disturbances. Therefore, accurately assessing the capacity of multi-configuration orders, timely predicting potential on-site risks, and effectively verifying the feasibility of production plans are the core challenges of assembly line management. Assembly process simulation methods, which can effectively describe the operational behaviors and logical constraints of the assembly line, have become an important technical means to address these core challenges.

[0003] However, existing assembly process simulation methods mostly rely on pre-defined, fixed rules to handle decision-making issues such as parallel operation execution, resource allocation, and material delivery. These rules cannot adaptively adjust to dynamic changes in the physical states of entities such as fleet size, resources, and materials during simulation, leading to discrepancies between simulation results and complex and ever-changing on-site conditions. Furthermore, for different business scenarios such as order evaluation, risk prediction, and plan verification, existing methods often require the establishment of different analytical models, forcing on-site personnel to frequently switch models and making it difficult to form consistent decision-making judgments within a unified data framework.

[0004] For example, Chinese patent CN118348918A discloses a method and system for multi-source data perception, fusion, and predictive analysis of aircraft assembly lines. It achieves capacity prediction based on production line conditions using a long short-term memory network model. However, due to the inherent black-box nature of deep learning models, it is difficult to reveal the reasoning processes between core assembly activities such as job execution, resource utilization, and material matching, resulting in insufficient interpretability of the prediction results and thus failing to directly support on-site decision optimization. Similarly, Chinese patent CN117055374A proposes a method for rapidly constructing production line simulation models for various aircraft models. However, this method is currently mainly used for verifying process division and resource allocation schemes in the early planning stage of aircraft production lines. Its application focuses on static, long-term planning analysis, lacking simulation reasoning and adaptive decision-making capabilities for multi-business needs on-site, and cannot cope with dynamic disturbances and plan adjustments during production execution.

[0005] Therefore, there is an urgent need for an assembly process simulation method that can take into account both unified modeling and adaptive decision control across multiple business scenarios, in order to improve the prediction accuracy and decision availability in mixed-line assembly scenarios. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft. By constructing a unified simulation model that includes differentiated features of multiple configurations and business decision-making rules, the method characterizes the behavior of production factors in the mixed-line assembly process. The method uses assembly events to drive the evolution of the model state, adaptive decision-making rules to control the event scheduling process, and combines input data and inference constraints from different business scenarios to achieve multi-mode simulation and adaptive decision-making for order evaluation, risk prediction, and plan verification.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] This invention provides a multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft, including:

[0009] S1. Construct a simulation model for mixed-line assembly. The simulation model includes an entity set, an activity set, and a future event list. The entity set includes flight entities, resource entities, and material entities. The activity set includes job execution, resource allocation, and material supply. The future event list is used to drive the state update of each entity in the entity set.

[0010] S2. Establish a model initialization mapping function, which is used to map external input data to the initial state of the entity or the initial list of future events according to the order simulation mode, online simulation mode or planned simulation mode respectively.

[0011] S3. Event-driven execution of multi-mode simulation reasoning and adaptive decision-making, including: selecting the earliest occurring event from the future event list as the current event, advancing the simulation clock, and updating the state of the corresponding entity in the entity set; when an entity's state change triggers an execution decision, calculating the score of each candidate decision rule based on the current entity's observation set, and selecting the candidate decision rule with the highest score as the current decision rule; determining the execution order of parallel operations, the response order of resource requests, or the response order of material requests according to the current decision rule; generating corresponding new events in the determined order, and inserting the generated new events into the future event list;

[0012] S4. Based on the update result of the entity state, calculate the reward value of the selected decision rule according to the preset reward function, and combine the entity state before the decision, the selected decision rule and the reward value into a decision sample and store it in the sample library.

[0013] S5. Repeat steps S3 to S4, and extract multiple decision samples from the sample library at fixed time intervals, update the scoring parameters of each candidate decision rule, until the preset simulation termination condition is met.

[0014] S6. Output simulation results according to the simulation mode. The simulation results include the completion time of each flight, resource waiting time, material waiting time, or the deviation between the planned start time and the predicted start time.

[0015] Further, in S1, the flight entity is represented as:

[0016]

[0017] in, For the first entity in the set Each flight entity, For the first The unique identifier for each flight entity For the first The product configuration of each flight entity. For the first The status of each flight entity;

[0018] The resource entity is represented as follows:

[0019]

[0020] in, For the first entity in the set One resource entity, For the first The type name of each resource entity. For the first The product configuration that a resource entity is suited for. For the first The status of each resource entity;

[0021] The material entity is represented as:

[0022]

[0023] in, For the first entity in the set A material entity, For the first The type name of a material entity. For the first The product configuration that each material entity is adapted to. For the first The inventory status of each material entity.

[0024] Furthermore, in S2, under the order simulation mode, the product configuration and production time of each flight in the production order are mapped to flight production events, which serve as the initial list of future events;

[0025] In the online simulation mode, the available quantity of various resource entities, the inventory quantity of various material entities, and the work execution progress under the on-site working conditions are mapped to the initial state of the corresponding entities; if there is a material delay under the current working conditions, the adjusted material supply time and supply quantity are mapped to a new material supply event, which serves as the initial list of future events.

[0026] In the planned simulation mode, the planned start time and planned completion time of the operation are mapped to planned start events of the operation, which serve as the initial list of future events.

[0027] Furthermore, in S3, in online simulation mode, for the completion event of a job being performed, the occurrence time is calculated based on the time corresponding to the input working condition and the remaining working hours, and the completion event is stored in the future event list;

[0028] In the planned simulation mode, when resources or materials cannot meet the requirements for the operation to start at the planned time node, the occurrence time of the operation start event is corrected to the maximum value of the occurrence time of the resource matching event and the occurrence time of the material matching event, and the corrected operation start event is stored in the future event list.

[0029] Furthermore, in S3, the observation set of the sortie entity includes the operation time, the remaining total time, the maximum value, minimum value, average value, and variance of the operation request satisfaction; the candidate decision rules of the sortie entity include: shortest operation time priority, longest operation time priority, shortest remaining total time priority, longest remaining total time priority, and earliest operation request satisfaction priority.

[0030] The observation set of the resource entity includes the resource request time, resource occupancy time, maximum value, minimum value, average value and variance of resource request satisfaction; the candidate decision rules of the resource entity include: earliest resource request priority, latest resource request priority, shortest occupancy time priority, longest occupancy time priority, and earliest satisfied resource request priority.

[0031] The observation set of the material entity includes the material request time, material delivery distance, maximum value, minimum value, average value and variance of material request satisfaction; the candidate decision rules of the material entity include: earliest material request priority, latest material request priority, shortest delivery distance priority, longest delivery distance priority, and earliest satisfied material request priority.

[0032] Furthermore, in S3, random disturbance events are generated using a probability distribution model and stored in a future event list. These random disturbance events include material delay events, tooling failure events, or personnel leave events. When a tooling failure event occurs, if the faulty tooling is in an occupied state, the completion event occurrence time of the corresponding operation in the future event list is updated, and the tooling failure duration is increased accordingly.

[0033] Furthermore, in S3, the reward function of the decision rule for the flight entity is calculated based on the deviation between the operation waiting time and the completion time; the reward function of the decision rule for the resource entity is calculated based on the resource utilization rate; and the reward function of the decision rule for the material entity is calculated based on the material residence time.

[0034] Furthermore, in S3, the state change of an entity triggers the execution decision, including: when the job completion event is triggered, the job execution order decision for the flight entity is entered;

[0035] When a resource release event, resource repair event, or personnel leave event is triggered, the resource request response order decision is made for the resource entity.

[0036] When a material supply event is triggered, the material request response order is determined for the material entity.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0038] (1) This invention constructs a unified simulation model comprising an entity set, an activity set, and a future event list. During the entity definition process, the operational network, resource requirements, and material requirements corresponding to different product configurations are represented as entity attributes and activity constraints, enabling the simulation model to describe the differentiated assembly requirements of multiple configuration frames under mixed-line assembly conditions. Based on this, a model initialization mapping function is established, mapping external input data to the initial state of entities or the initial future event list for order simulation mode, online simulation mode, and planned simulation mode, respectively. By uniformly abstracting the input data of different business scenarios into entity states and event types recognizable by the simulation model, it eliminates the need to establish independent simulation models for different scenarios, achieving decision-making and scheme adjustment under unified data support, and improving the coherence and consistency of simulation analysis across multiple business scenarios.

[0039] (2) In the event reasoning process, this invention matches the corresponding job set according to the product configuration of the flight entity, and further matches the resource type and material category required for job execution to ensure that job requests, resource responses, and material responses meet the process differentiation requirements of different configurations. On this basis, an adaptive decision-making mechanism based on observation sets and scoring parameters is also introduced. When a change in entity state triggers a decision, the system calculates the score value of each candidate decision rule based on the current entity's observation set through a scoring function, selects the rule with the highest score value for execution, and determines the execution order of parallel jobs, the response order of resource requests, or the response order of material requests. The observation set describes the trend of entity state changes, and the scoring parameters quantify the applicability of each decision rule in different states, enabling the decision rules to be dynamically adjusted according to the real-time state, thereby improving the matching degree between simulation and actual on-site working conditions.

[0040] (3) After each decision is made, the present invention calculates the reward value of the decision according to the updated entity state and a preset reward function. The entity state before the decision, the selected decision rule, and the reward value are combined into a decision sample and stored in the sample library. The sample is extracted at fixed time intervals to update the scoring parameters of each candidate decision rule. The reward function of the flight entity rule is based on the job waiting time and completion time deviation, the reward function of the resource entity rule is based on the resource utilization rate, and the reward function of the material entity rule is based on the material residence time. The scoring parameters are updated incrementally according to the reward value. The scoring parameters of rules that obtain positive rewards increase and the tendency to be selected increases, while the scoring parameters of rules that obtain negative rewards decrease and the tendency to be selected decreases. This enables the system to continuously learn and optimize decision rules during simulation operation and continuously improve the quality of decision-making.

[0041] (4) This invention predicts the completion time of each flight through an order simulation mode, providing a basis for capacity assessment; it integrates on-site working condition data through an online simulation mode, calculates completion events based on remaining working hours, and realizes simulation deduction starting from the intermediate state of the working condition, timely predicting potential bottlenecks and risks; it verifies the feasibility of planned time nodes through a planning simulation mode, and identifies redundant and conflicting nodes. At the same time, it generates random disturbance events such as material delays, tooling failures, and personnel leave using a probability distribution model, making the simulation deduction closer to the actual on-site situation. The rules adopted in each decision-making link have clear business meanings, and the decision-making process is transparent and explainable, enabling on-site management personnel to understand the reasons for the formation of simulation results and make more evidence-based decision adjustments. Attached Figure Description

[0042] Figure 1 A flowchart of a multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft provided in an embodiment of the present invention;

[0043] Figure 2 This is a graph showing the flight completion prediction results in the order simulation mode of this invention.

[0044] Figure 3 This is a comparison chart of the predicted cumulative working hours and the actual cumulative working hours in the online simulation mode of this invention embodiment;

[0045] Figure 4 This is a diagram showing the resource and material identification results in the online simulation mode of this invention.

[0046] Figure 5 This is a diagram showing the redundancy / conflict identification results of the planned simulation mode in an embodiment of the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Example

[0049] This embodiment provides a multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft, including the following steps:

[0050] S1. Construct a simulation model for mixed-line assembly, wherein the simulation model includes a set of entities, a set of activities, and a list of future events.

[0051] The "flight" entity is used to represent each flight entity to be assembled on the aircraft assembly line, and is represented as:

[0052]

[0053] in, For the first entity in the set One flight entity; For the first A unique identifier for each sortie entity, used to distinguish different sorties and track their status during the reasoning process; For the first The product configuration to which each flight belongs. For the first In this embodiment, the product configuration includes basic type B, derivative type L, and derivative type S; the status of the flight entity includes states such as operation not started, waiting for resources, waiting for materials, in progress, and completed.

[0054] Resource entities are used to represent the various production resources required in the assembly process, and are represented as follows:

[0055]

[0056] in, For the first entity in the set One resource entity, For the first The type name of each resource entity. For the first The product configuration that a resource entity is suited for. For the first The status of each resource entity; in this embodiment, resource entities include four categories: professional personnel, tooling, equipment and handling devices, and the status of resource entities includes resource occupancy, equipment maintenance, equipment failure, personnel leave, etc.

[0057] Material entities are used to represent various materials required in the assembly process, and are represented as follows:

[0058]

[0059] in, For the first entity in the set A material entity, For the first The type name of a material entity. For the first The product configuration that each material entity is adapted to. For the first The inventory status of each material entity; in this embodiment, the material entities include three categories: standard parts, structural parts, and system parts.

[0060] The activity set includes job execution, resource allocation, and material supply, used to characterize assembly behaviors that change the state of model entities during aircraft production. In mixed-line assembly scenarios, job execution activities match the corresponding job content based on the product configuration of the aircraft, while resource allocation and material supply activities match the corresponding resource types and material categories based on the job requirements under that configuration. Resource allocation refers to the process where resources are allocated in place and the job can obtain the required resources when the available quantity of resource entities meets the job request; material supply refers to the process where materials are delivered in complete sets and the job can obtain the required materials when the inventory quantity of material entities meets the job request. In the aircraft final assembly process described in this embodiment, each workstation follows a serial logic. Jobs with clear process dependencies within a workstation must be executed serially, while multiple jobs without spatial interference or strict process dependencies, and whose resources and materials meet the execution conditions, are allowed to be executed simultaneously.

[0061] The constraints for job execution in an activity set are represented as follows:

[0062]

[0063] in, Indicates the first Constraints on a sequence of tasks Indicates the first A sequence of tasks, , express The first in One assignment, This represents the total number of jobs in the job sequence. express Suitable product configuration. Represented as:

[0064]

[0065] in express The name, express Working hours, express resource requirements, express Material requirements express The set of immediate tasks, ,in express The Middle A pressing task ahead, express The total number of pre-contraction operations.

[0066] The future event list will be used to drive the state updates of each entity in the entity set, and will be represented as follows:

[0067]

[0068] in, Indicates the first A future event, express The time of occurrence, for The occurrence corresponds to the entity, express The parameters required for the execution process are as follows: for example, the flight commissioning event requires the flight identifier and commissioning time as parameters; the operation start-up event requires the operation identifier and start-up time as parameters; the resource matching event requires the resource type and matching quantity as parameters; and the material supply event requires the material type and supply quantity as parameters.

[0069] S2. Instantiate the rack entity, resource entity, and material entity based on the multi-configuration orders, resource configurations, and material configurations at the assembly site; establish a model initialization mapping function, which is used to map external input data to the initial state of the entity or the initial list of future events according to the order simulation mode, online simulation mode, or planned simulation mode, respectively.

[0070] In order simulation mode, with all flights not yet in production as the initial state, the product configuration and production time of each flight in the production order are mapped to flight production events, which serve as the initial list of future events.

[0071] In online simulation mode, the initial state is the work execution progress of the in-process rack, which includes the remaining working hours of the started work and the list of the unstarted work. The available quantity of various resource entities, the inventory quantity of various material entities, and the work execution progress under the on-site working conditions are mapped to the initial state of the corresponding entities.

[0072] If there is a material delay under the current operating conditions, the adjusted material supply time and quantity will be mapped to a new material supply event, and the new material supply event will be used as the initial list of future events.

[0073] In the planning simulation mode, the planned start time and planned completion time of the operation are mapped to the planned start event of the operation, which serves as the initial list of future events.

[0074] S3. Execute multi-modal simulation reasoning and adaptive decision-making based on event-driven execution, specifically including the following sub-steps:

[0075] S301. Select the event with the earliest occurrence time from the list of future events as the current event, advance the simulation clock, and update the state of the corresponding entity in the entity set.

[0076] In online simulation mode, the time of occurrence of the completion event for a job in progress. Based on the time corresponding to the input working condition and remaining working hours calculate:

[0077]

[0078] Save the completion event to the list of future events.

[0079] In the planned simulation mode, when resources or materials cannot meet the requirements for the operation to start at the planned time node, the occurrence time of the operation start event is corrected to the maximum value of the occurrence time of the resource matching event and the occurrence time of the material matching event (i.e., the later one of the occurrence time of the resource matching event and the occurrence time of the material matching event).

[0080] Save the revised job start event to the future event list.

[0081] Furthermore, a probability distribution model is used to control the generation of random disturbance events, including material delay events, tooling failure events, and personnel leave events. The generated random disturbance events are stored in a future event list. When a tooling failure event occurs, if the faulty tooling is in use, the completion event time of the corresponding operation in the future event list is updated, and the tooling failure duration is increased accordingly.

[0082] S302. When a job completion event is triggered, the job execution order decision is entered into the flight entity.

[0083] When a resource release event, resource repair event, or personnel leave event is triggered, the resource request response order decision is made for the resource entity.

[0084] When a material supply event is triggered, the material request response order is determined for the material entity.

[0085] When executing each type of decision, the system calculates the score of each rule in the candidate decision rule set corresponding to the entity based on the observation set of the current entity through a scoring function, and executes the rule with the highest score.

[0086] The physical flight at the moment observation set Represented as:

[0087]

[0088] in, , , and They represent the times respectively. The maximum, minimum, average, and variance of all operating hours in the entity that meet the start-up conditions. , , and These represent the maximum, minimum, average, and variance of the remaining total working hours in the current flight entity, respectively. , , and These represent the maximum, minimum, average, and variance of the job request satisfaction rate in the current flight entity, respectively.

[0089] The set of candidate decision rules for each sortie entity Represented as:

[0090]

[0091] in, This indicates that the shortest operation time takes priority. This indicates that the longest working time takes priority. This indicates that the shortest remaining total working time takes precedence. This indicates that the longest remaining total working hours takes precedence. This indicates that the earliest job request will be fulfilled first.

[0092] Candidate Decision Rule Set The formulas for calculating the score of each rule are as follows:

[0093]

[0094] in, express The Middle Each candidate decision rule at time... The rating value, for The Middle Each candidate decision rule at time... The scoring parameters for The Middle The fit value of each candidate decision rule.

[0095] Regarding the rules and , The calculation is shown in the following formula:

[0096]

[0097] Regarding the rules , and , The calculation is shown in the following formula:

[0098]

[0099] in, express The Middle The evaluation index values ​​corresponding to each candidate decision rule are work hours, remaining total work hours, or job request satisfaction rate. For each rule... and Take the work hours; for the rules and Take the remaining total working hours; for the rule , and determine the job request satisfaction rate. Indicates the current time The maximum value of the indicator corresponding to all candidate decision rules. Indicates the current time The minimum value of the index corresponding to all candidate decision rules. To prevent extremely small positive numbers with a denominator of zero, this embodiment... All are taken as .

[0100] Candidate Decision Rule Set The rule with the highest score for:

[0101]

[0102] According to the rule of the highest score The set of parallel jobs for the current simulation clock-down entity is sorted as follows:

[0103]

[0104] in, Indicates at time The order of job execution The sorting function for sorting flight entities. Indicates at time A set of parallel jobs that meet the execution conditions. Resource request events and material request events are generated sequentially according to the order of parallel jobs and stored in the future event list.

[0105] Resource entities at any time observation set Represented as:

[0106]

[0107] in, , , and These represent the maximum, minimum, average, and variance of resource request times, respectively. , , and These represent the maximum, minimum, average, and variance of resource usage time, respectively. , , and These represent the maximum, minimum, average, and variance of the resource request satisfaction rate, respectively.

[0108] Set of candidate decision rules for resource entities Represented as:

[0109]

[0110] in, This indicates that the earliest resource request takes precedence. This indicates that the latest resource request takes precedence. This indicates that the shortest time takes priority. This indicates that the longest occupied time takes precedence. This indicates that the earliest resource request will be satisfied first.

[0111] Candidate Decision Rule Set The formulas for calculating the score of each rule are as follows:

[0112]

[0113] in, express The Middle Each candidate decision rule at time... The rating value, express The Middle One candidate decision rule, for The Middle Each candidate decision rule at time... The scoring parameters for The Middle The fit value of each candidate decision rule.

[0114] Regarding the rules and , The calculation formula is as follows:

[0115]

[0116] Regarding the rules , and , The calculation formula is as follows:

[0117]

[0118] in, express The Middle The evaluation index values ​​corresponding to each candidate decision rule are resource request time, resource occupancy time, or resource request fulfillment time, for each rule. and Take the resource request time; for rules and Take the time the resource is occupied; for rules , and obtain the resource request satisfaction rate. Indicates the current time The maximum value of the indicator corresponding to all candidate decision rules. Indicates the current time The minimum value of the indicator corresponding to all candidate decision rules.

[0119] Candidate Decision Rule Set The rule with the highest score for:

[0120]

[0121] According to the rule of the highest score Sort the request set of resource entities under the current simulation clock:

[0122]

[0123] in, Indicates at time The order of resource request and response. The sorting function represents the resource entity. Indicates at time The resource request queue. After determining the resource request response order, it is checked whether the resource requests of the job can be satisfied in turn. If satisfied, a resource-related event is generated and stored in the future event list.

[0124] Material entity at time observation set Represented as:

[0125]

[0126] in, , , and These represent the maximum, minimum, average, and variance of the material request time, respectively. , , and These represent the maximum, minimum, average, and variance of the material delivery distance, respectively. , , and These represent the maximum, minimum, average, and variance of the material request satisfaction rate, respectively.

[0127] The set of candidate decision rules for material entities Represented as:

[0128]

[0129] in, This indicates that the earliest material request takes priority. This indicates that the latest material request takes priority. This indicates that the shortest delivery distance is prioritized. This indicates that the longest delivery distance will be prioritized. This indicates that the earliest material request will be fulfilled first.

[0130] Candidate Decision Rule Set The formulas for calculating the score of each rule are as follows:

[0131]

[0132] in, express The Middle Each candidate decision rule at time... The rating value, express The Middle One candidate decision rule, for The Middle Each candidate decision rule at time... The scoring parameters for The Middle The fit value of each candidate decision rule.

[0133] Regarding the rules and , The calculation formula is:

[0134]

[0135] Regarding the rules , and , The calculation formula is:

[0136]

[0137] in, express The Middle The evaluation index values ​​corresponding to each candidate decision rule are material request time, material delivery distance, or material request fulfillment time. and Material request time; for rules and Take the material delivery distance; for rules Take the material request satisfaction rate. Indicates the current time The maximum value of the indicator corresponding to all candidate decision rules. Indicates the current time The minimum value of the indicator corresponding to all candidate decision rules.

[0138] Candidate Decision Rule Set The rule with the highest score for:

[0139]

[0140] According to the rule of the highest score Sort the request set of material entities under the current simulation clock:

[0141]

[0142] in, Indicates at time The material request response order, The sorting function represents the material entities. Indicates at time The material request queue. After determining the material request response order, it is checked whether the material requests for the operation can be satisfied in turn. If satisfied, a material matching event is generated and stored in the future event list.

[0143] S4. Evaluate and update the parameters of the executed decision rules, which includes the following sub-steps:

[0144] S401. Based on the update result of the entity state, calculate the reward value for this decision according to the preset reward function, specifically:

[0145] The reward function for the flight entity rule is calculated based on the deviation between job waiting time and completion time:

[0146]

[0147] in, Indicates the time of flight entity The reward value, Indicates the time of flight entity The waiting time for the assignment This indicates the allowable job waiting time reference value. Indicates the time of flight entity Completion time deviation, This indicates the allowable deviation from the completion time reference value. and In this embodiment, the weights for job waiting time and completion time deviations are... , .

[0148] The reward function for resource entity rules is calculated based on resource utilization rate:

[0149]

[0150] in, Indicates the resource entity at time. The reward value, This indicates the duration of busy work for a resource entity. This indicates the duration of time required to respond to the corresponding resource request. This indicates the total available time of resources within the simulation statistical period.

[0151] The reward function for material entity rules is calculated based on material residence time:

[0152]

[0153] in, Indicates the material entity at time [time]. The reward value, Indicates the material entity at time [time]. The length of stay This indicates the allowable material residence time reference value.

[0154] S402, Change the entity state before the decision Selected decision rules and the reward value Combined into a single decision sample , and store in the sample library.

[0155]

[0156] Indicates the entity type, take , or These correspond to the flight entity, resource entity, or material entity, respectively.

[0157] S403, at fixed time intervals λ samples are drawn from the sample library (in this embodiment) (Using a 24-hour timeframe and λ = 72), this is used to update the scoring parameters for each candidate decision rule:

[0158]

[0159]

[0160] Indicate entity type The Each candidate decision rule at time... The scoring parameters; in this embodiment, the initial scoring parameters for each candidate decision rule during simulation initialization are: , Indicates the entity type, take , or These correspond to the flight entity, resource entity, or material entity, respectively. Indicates temporary scoring parameters. Indicate entity type The The average reward value of each candidate decision rule in this sampling. Indicate entity type The overall average reward value in this sampling; Indicates the parameter update step size. This represents the lower limit of the scoring parameter. This indicates that the upper limit of the rating parameter is set to prevent the rating parameter from increasing or decreasing excessively due to the accumulation of consecutive rewards; the updated rating parameter is limited to a preset range. In this embodiment , ,

[0161]

[0162]

[0163] in This indicates that the entity type in this sample is And the selected rule is the first A sample set of candidate decision rules. express The number of samples, Indicates sample The reward value, This indicates that the entity type in this sample is The complete set of decision samples express The number of samples.

[0164] The scoring parameters of each candidate decision rule are used to characterize the applicability of each candidate rule under different states. They affect the rule evaluation value by participating in the calculation of the rule evaluation function and are updated according to the reward result after the rule is executed, so that the high reward rule has a higher tendency to be selected in subsequent similar states.

[0165] S5. Repeat steps S3 to S4 until the preset simulation end condition is met. The simulation end condition is: the cumulative number of simulation cycles, and the simulation ends when the number of simulation cycles reaches a preset value; wherein, the simulation cycle is completed when the simulation clock reaches the preset end time or all sorties are completed.

[0166] S6. Output simulation results according to the simulation mode:

[0167] Output the completion time of each flight in the order simulation mode.

[0168] In online simulation mode, the resource waiting time and material waiting time corresponding to each task are output.

[0169] In the planning simulation mode, the deviation between the planned start time and the predicted start time is output.

[0170] To better illustrate the beneficial effects of the present invention, the order simulation mode, online simulation mode, and planning simulation mode in the embodiments of the present invention have been verified.

[0171] The assembly line involves basic model B, derivative model L, and derivative model S, comprising 5 stations, over 800 operations, 15 resource entities, and 12 material entities. The 15 resource entities are divided into four categories: professional personnel, tooling, equipment, and handling devices. Among them, equipment includes 5 types: lifting platforms, docking tooling, engine brackets, large component lifting tools, and measuring tools; handling devices include two types: AGV-1 and AGV-2; professional personnel include 8 types: electrical engineering, electrical engineering, electrical engineering, hydraulics, fuel, high-altitude operations, control, and oxygen; the 12 material entities are divided into three categories: standard parts, structural parts, and system parts. Among them, standard parts include 4 types: duct laying accessories, interior accessories, avionics system cable assemblies, and portable fuel components; structural parts include 5 types: vertical tail docking accessories, extended fuselage connecting parts, standard fuselage connecting parts, landing gear mounting assemblies, and fuselage docking fairings; and system parts include 3 types: engine accessories, L-type engine accessories, and system insulation assemblies.

[0172] The order contains 12 flights, numbered CXS011, CXL012, CXB013, CXB014, CXS015, CXB016, CXB017, CXB018, CXS019, CXB020, CXB021, and CXL022.

[0173] (1) Validation of order simulation mode

[0174] The production start time of 12 flights in the order is used as input.

[0175] Figure 2 The invention demonstrates the predicted assembly time for each flight under ideal working conditions and without random disturbances. The predicted assembly time is consistent with the theoretical cycle time of the assembly line design, proving that the invention can accurately predict the assembly time of each flight in multi-configuration orders under order simulation mode, providing a reliable basis for capacity assessment.

[0176] (2) Online simulation mode verification

[0177] Selecting segments of the CXB013 assembly process where resource usage and material supply are relatively stable, and using the operational data at the completion time of OP10 operation in CXB013 as input, compare the actual cumulative operating hours and the predicted cumulative operating hours for each operation. For example... Figure 3 As shown, the cumulative working hours predicted by the method provided by this invention are in high agreement with the actual cumulative working hours, proving that the online simulation mode of this invention can accurately predict the completion time of subsequent operations based on on-site working condition data.

[0178] During the on-site assembly of aircraft CXB018, a tooling failure occurred. The online simulation mode of this invention was used for prediction. (Reference) Figure 4 As can be seen, the simulation results identify the risk of delays in subsequent operations, clarify the resources and materials causing the delays, and demonstrate the waiting time for resources and materials for each operation. This indicates that the present invention can predict the potential risks brought about by on-site disturbances in a timely manner, providing early warning and decision support for on-site management personnel.

[0179] (3) Verification of the planning simulation mode

[0180] During on-site production, management personnel rearranged the disrupted production plan and adjusted the resource allocation scheme. Using the planning simulation mode of this invention, the planned start and finish times for each operation after rearrangement, as well as the resource allocation scheme, are input.

[0181] refer to Figure 5 As can be seen, the simulation results predicted potential redundant and conflicting nodes in the rescheduling plan. Redundant nodes indicate that the planned start time is later than the simulated feasible time, indicating a time buffer; conflicting nodes indicate that the planned start time is earlier than the simulated feasible time, indicating a time conflict. This demonstrates that the present invention can effectively verify the feasibility of planned time nodes, helping management personnel identify problems in the plan and make adjustments.

[0182] (4) Performance comparison between the method of the present invention and the fixed rule simulation method

[0183] To verify the effectiveness of the adaptive decision-making rule of this invention, the performance of the method of this invention was compared with that of a simulation method based on fixed rules under the same simulation conditions. The fixed rules included: shortest operation time priority, longest operation time priority, and earliest request fulfillment priority. Performance evaluation indicators included on-time delivery rate, production cycle time, production line balance rate, and resource utilization rate. The performance comparison results are shown in Table 1.

[0184] Table 1. Performance comparison between the method of this invention and the fixed rule simulation method

[0185]

[0186] As shown in Table 1, the method of this invention outperforms the three fixed-rule methods in four indicators: on-time delivery rate, production cycle time, production line balancing rate, and resource utilization rate. This invention adaptively optimizes the job execution sequence, resource request response sequence, and material request response sequence based on the dynamic changes in the entity state during the reasoning process, achieving capacity optimization and cycle time balancing control in the assembly process and improving the utilization efficiency of key resources. Compared to fixed rules that cannot adjust according to state changes, this invention can adjust the decision rules in real time according to the dynamic changes in the entity state, making the simulation results closer to the actual working conditions on site, significantly improving the rationality of simulation predictions and the usability of decision-making.

[0187] The specific embodiments of the present invention are provided to enable those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention.

[0188] It should be understood that the present invention is not limited to the content already described above, and various modifications and changes can be made without departing from its scope. The scope of the present invention is limited only by the appended claims.

Claims

1. A multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft, characterized in that, include: S1. Construct a simulation model for mixed-line assembly. The simulation model includes an entity set, an activity set, and a future event list. The entity set includes flight entities, resource entities, and material entities. The activity set includes job execution, resource allocation, and material supply. The future event list is used to drive the state update of each entity in the entity set. S2. Establish a model initialization mapping function, which is used to map external input data to the initial state of the entity or the initial list of future events according to the order simulation mode, online simulation mode or planned simulation mode respectively. S3. Event-driven execution of multi-mode simulation reasoning and adaptive decision-making, including: selecting the earliest occurring event from the future event list as the current event, advancing the simulation clock, and updating the state of the corresponding entity in the entity set; when an entity's state change triggers an execution decision, calculating the score of each candidate decision rule based on the current entity's observation set, and selecting the candidate decision rule with the highest score as the current decision rule; determining the execution order of parallel operations, the response order of resource requests, or the response order of material requests according to the current decision rule; generating corresponding new events in the determined order, and inserting the generated new events into the future event list; The observation set of the sortie entity includes the operation time, the remaining total operation time, the maximum value, minimum value, average value and variance of the operation request satisfaction; the candidate decision rules of the sortie entity include: shortest operation time priority, longest operation time priority, shortest remaining total operation time priority, longest remaining total operation time priority, and earliest operation request satisfaction priority. The observation set of the resource entity includes the resource request time, resource occupancy time, maximum value, minimum value, average value and variance of resource request satisfaction; the candidate decision rules of the resource entity include: earliest resource request priority, latest resource request priority, shortest occupancy time priority, longest occupancy time priority, and earliest satisfied resource request priority. The observation set of the material entity includes the material request time, material delivery distance, maximum value, minimum value, average value and variance of the material request satisfaction; the candidate decision rules of the material entity include: earliest material request priority, latest material request priority, shortest delivery distance priority, longest delivery distance priority, and earliest satisfied material request priority. S4. Based on the update result of the entity state, calculate the reward value of the selected decision rule according to the preset reward function, and combine the entity state before the decision, the selected decision rule and the reward value into a decision sample and store it in the sample library. S5. Repeat steps S3 to S4, and extract multiple decision samples from the sample library at fixed time intervals, update the scoring parameters of each candidate decision rule, until the preset simulation termination condition is met. S6. Output simulation results according to the simulation mode. The simulation results include the completion time of each flight, resource waiting time, material waiting time, or the deviation between the planned start time and the predicted start time.

2. The multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft according to claim 1, characterized in that, In S1, the flight entity is represented as: in, For the first entity in the set Each flight entity, For the first The unique identifier for each flight entity For the first The product configuration of each flight entity. For the first The status of each flight entity; The resource entity is represented as follows: in, For the first entity in the set One resource entity, For the first The type name of each resource entity. For the first The product configuration that a resource entity is suited for. For the first The status of each resource entity; The material entity is represented as: in, For the first entity in the set A material entity, For the first The type name of a material entity. The product configuration adapted to the nth material entity. This represents the inventory status of the nth material entity.

3. The multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft according to claim 1, characterized in that, In S2, under the order simulation mode, the product configuration and production time of each flight in the production order are mapped to flight production events, which serve as the initial list of future events. In the online simulation mode, the available quantity of various resource entities, the inventory quantity of various material entities, and the work execution progress under the on-site working conditions are mapped to the initial state of the corresponding entities; if there is a material delay under the current working conditions, the adjusted material supply time and supply quantity are mapped to a new material supply event, which serves as the initial list of future events. In the planned simulation mode, the planned start time and planned completion time of the operation are mapped to planned start events of the operation, which serve as the initial list of future events.

4. The multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft according to claim 1, characterized in that, In S3, in online simulation mode, the occurrence time of a completion event for a job in progress is calculated based on the time corresponding to the input working condition and the remaining working hours, and the completion event is stored in the future event list. In the planned simulation mode, when resources or materials cannot meet the requirements for the operation to start at the planned time node, the occurrence time of the operation start event is corrected to the maximum value of the occurrence time of the resource matching event and the occurrence time of the material matching event, and the corrected operation start event is stored in the future event list.

5. The multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft according to claim 1, characterized in that, In S3, random disturbance events are generated using a probability distribution model and stored in a future event list. These random disturbance events include material delay events, tooling failure events, or personnel leave events. When a tooling failure event occurs, if the faulty tooling is in an occupied state, update the completion event time of the corresponding job in the future event list and add the tooling failure duration to it.

6. The multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft according to claim 1, characterized in that, In S3, the reward function of the decision rule for the flight entity is calculated based on the deviation between the operation waiting time and the completion time; the reward function of the decision rule for the resource entity is calculated based on the resource utilization rate; and the reward function of the decision rule for the material entity is calculated based on the material residence time.

7. The multi-mode simulation and adaptive decision-making method for mixed-line assembly of aircraft according to claim 1, characterized in that, In S3, the state change of an entity triggers the execution decision, including: when the job completion event is triggered, the job execution order decision for the flight entity; When a resource release event, resource repair event, or personnel leave event is triggered, the resource request response order decision is made for the resource entity. When a material supply event is triggered, the material request response order is determined for the material entity.

Images