A Predictive Approach for Aero-engine Support Based on Multi-Agent Modeling

Abstract

This invention proposes a predictive method for aero-engine support based on multi-agent modeling. This method comprehensively considers the main factors influencing the number of backup engines available for use. Addressing the variable states of aero-engines during operation and maintenance, it proposes an agent state machine model for aero-engines. Using multi-agent simulation, it recreates various mission scenarios for aero-engines, including ordering, use, overhaul, and return to the factory after failure. This enables dynamic simulation of the entire fleet's aero-engine usage process, predicting the optimal reserve number of aero-engines for a certain period in the future—that is, how many backup engines should be equipped to meet a given operational readiness level. Simulation results show that this invention can effectively simulate the support process of a fleet's aero-engines and accurately determine the number of backup engines required under a given operational readiness level.

Description

Technical Field

[0001] This invention relates to the field of aero-engine maintenance and support technology, specifically to an aero-engine support prediction method based on multi-agent modeling. Background Technology

[0002] Aircraft engines are essential for ensuring an aircraft can perform its missions normally. If an aircraft engine malfunctions, the aircraft will be unable to complete its mission. Because aircraft engines are high-value products, the time from ordering to delivery is considerable; therefore, temporary orders cannot solve the engine shortage problem.

[0003] Engine shortages fall into two categories: First, foreseeable shortages. For example, some engines are expected to reach the end of their service life within the year, meaning they will need major overhauls at some point during the year; others are expected to reach their total service life within the year, meaning they will need to be phased out within the year. The second category is unforeseen shortages. For instance, engines may experience major malfunctions (which are random) and require early return for repairs. The number of these two types of engines together determines the number of spare engines that need to be planned in advance. Since aircraft lifespans are typically long, while engine lifespans are relatively short, the demand for and repair volume of engines is significantly higher than that of aircraft. Therefore, sufficient spare aircraft engines should be provided to cope with unexpected malfunctions and replenish engines reaching their service life. However, due to the complexity and variability of engine shortages, influencing factors include flight workload, aircraft usage and maintenance, and random engine malfunctions. This often results in some years having too many spare engines in reserve, leading to higher storage costs, reduced spare engine lifespan, and higher maintenance costs; conversely, some years have too few spare engines in reserve, leading to shortages that affect flight operations. Therefore, adopting scientific methods to macroscopically analyze and calculate the number of engines that need to be prepared for a certain period in the future is of significant practical importance.

[0004] Analyzing and calculating the number of engines that need to be prepared in a certain period of the future falls under the category of spare parts demand forecasting. Numerous methods have been studied for spare parts demand forecasting, but aero engines are high-value products with long ordering cycles. Their usage, consumption, and maintenance processes differ significantly from those of repairable spare parts; therefore, traditional spare parts calculation methods cannot be simply applied.

[0005] Current research on the number of spare aircraft engines mainly focuses on the following aspects:

[0006] The analysis utilizes simulation methods: treating aero-engines as important spare parts with a limited lifespan, and employing discrete event modeling to describe the characteristics of the aero-engine operating system, the study investigates the impact of factors such as different scrap rates, failure rates, maintenance cycles, timely replacement probabilities, and average delay times on backup requirements. However, the discrete event modeling simulation is relatively small in scale and does not provide a global analysis. Furthermore, it does not consider the impact of important indicators such as flight workload and aero-engine failure rates on the number of backups.

[0007] From an empirical and statistical perspective, a formula for calculating the number of spare aircraft engines is provided: the main factors affecting the number of aircraft engine backups are analyzed, and models such as the number of normal backups, the number of initial backups, and the number of engines ordered throughout the aircraft's life cycle are proposed; however, the statistical formula is too rough, and the results may have large errors.

[0008] From the perspective of optimization and decision theory, a computational model is constructed: taking the minimum maintenance and support cost as the objective function, a decision-making method for backup engines based on scheduling plans is proposed; however, the model construction is relatively coarse, and there are problems with data acquisition, making it difficult to apply in practice. Summary of the Invention

[0009] To address the problems existing in the prior art, this invention proposes a predictive method for aero-engine support based on multi-agent modeling. This method comprehensively considers the main factors affecting the number of backup engines for aero-engines. Addressing the variable states of aero-engines during use and maintenance, it proposes an agent state machine model for aero-engines. Utilizing multi-agent simulation, it recreates various mission scenarios for aero-engines, including ordering, use, overhaul, and return to the factory after failure. This enables dynamic simulation of the entire fleet's aero-engine usage process, predicting the optimal reserve number of aero-engines for a certain period in the future—that is, how many spare engines should be equipped to meet a given operational readiness level for the aircraft.

[0010] The technical solution of this invention is as follows:

[0011] The aforementioned aero-engine support prediction method based on multi-agent modeling includes the following steps:

[0012] Step 1: Design the state machine model of the aero-engine intelligent agent;

[0013] Step 2: Design simulation driving events for aero-engines;

[0014] Step 3: Based on the aero-engine intelligent agent state machine model designed in Step 1 and the aero-engine simulation driving events designed in Step 2, perform a multi-agent-based aero-engine support process simulation, and obtain the average satisfaction rate of the aero-engine through simulation.

[0015] Step 4: Adjust the number of spare engines, repeat step 3 to simulate the aircraft engine support process, and finally obtain the number of spare engines that meet the average availability requirements of aircraft engines.

[0016] In a further preferred embodiment, in step 1, the aero-engine intelligent agent state machine model has 5 states: working, overhaul, fault return for repair, backup, and scrap.

[0017] In a further preferred embodiment, in step 1, the transition models for each state of the aero-engine intelligent agent state machine model include:

[0018] (1) The transition of an aero-engine from working state to overhaul state can be divided into two situations: early return to the factory for overhaul and the engine reaching the stage of its lifespan.

[0019] When an aero-engine experiences a major malfunction and needs to be returned to the factory for overhaul ahead of schedule, it enters the overhaul state, the remaining lifespan of the stage becomes 0, and the stage lifespan consumed ahead of schedule is included. During simulation, this transition event occurs based on probability, and the probability value is determined with reference to the historical statistical value of the aero-engine early return-to-factory rate.

[0020] When an aircraft engine reaches the end of its service life, it enters a major overhaul phase.

[0021] (2) Transition of an aircraft engine from operational status to failure and return to factory for repair:

[0022] When an engine malfunctions during operation and requires factory repair, it will be returned to the factory for repair. Once repaired, the engine will be sent back to the factory, and its service life will not be reduced after the engine is returned.

[0023] (3) The transition of an aero-engine from working state to scrap state:

[0024] When an engine reaches the end of its service life, it shall be scrapped.

[0025] (4) Transition of aircraft engines from overhaul status to backup status:

[0026] After a certain overhaul cycle, the repaired engine is sent back as a backup engine and enters backup status.

[0027] (5) Transition of aircraft engines from faulty return-to-factory repair status to backup status:

[0028] After a certain period of fault repair, the repaired engine is sent back as a backup engine and enters backup status.

[0029] (6) Transition of aircraft engines from backup state to operational state:

[0030] The transition is triggered by a message; when an event of missing engines occurs, the system broadcasts a missing engine message and randomly selects an engine from the backup engines to receive the message, so as to replace the missing engine and enter the working state.

[0031] In a further preferred embodiment, the aero-engine simulation driving events designed in step 2 are flight plan generation event, mission acceptance event, and mission allocation event.

[0032] In a further preferred embodiment, step 2, the flight plan generation event is implemented through the following process:

[0033] First, based on the predicted total annual flight hours of the aircraft and statistical patterns, a monthly flight plan for the aircraft is randomly generated, and the annual flight time is allocated to each month. Then, within each month, the total flight time is randomly allocated to each day to form a daily flight hour plan.

[0034] In a further preferred embodiment, in step 2, the task acceptance event is achieved through the following process: according to the daily flight hour plan, the total daily flight time is allocated to each aircraft according to statistical rules.

[0035] In a further preferred embodiment, in step 2, the task allocation event is implemented through the following process: on a flight day, the engine operating hours are represented by a triangular distribution function; on a maintenance day, the engine operating time is the set time consumption for start-up checks; when the aircraft has no flight or maintenance tasks, or is in a malfunction, the engine operating time is 0.

[0036] In a further optimized approach, step 3, the simulation flow of the aero-engine support process based on multi-agent technology is as follows:

[0037] Step 3.1: Generate the aero-engine intelligent agent; initialize the parameters according to the set known conditions, set the initial state of the aero-engine, and randomly generate the daily flight hours of each aero-engine intelligent agent using the aero-engine simulation driving events designed in Step 2.

[0038] Step 3.2: The simulation is executed in days, with each aero-engine agent undergoing state changes; and the number of days with missing engines and grounded aircraft is counted until the simulation ends.

[0039] Step 3.3: Record and display the simulation results; after the simulation, calculate the average satisfaction rate of the aero-engine according to the formula:

[0040]

[0041] In a further optimization scheme, step 3 involves obtaining the correlation between various parameters and the average engine fulfillment rate during the aero-engine support process through multiple simulations, thereby determining the support parameters that affect the average engine fulfillment rate.

[0042] Beneficial effects

[0043] This invention proposes a predictive method for aero-engine support based on multi-agent modeling. Addressing the dynamic state changes inherent in aero-engines during use and maintenance, it proposes an agent state machine model for aero-engines. Utilizing multi-agent simulation, it recreates various mission scenarios for aero-engines, including ordering, use, overhaul, and return to the factory after failure. This enables dynamic simulation of the entire fleet's aero-engine usage process, predicting the optimal reserve number of aero-engines for a future period—that is, how many spare engines should be equipped to meet a given operational readiness level. Simulation results show that this invention can effectively simulate the support process of a fleet's aero-engines and accurately determine the number of spare engines required under a given operational readiness level.

[0044] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0045] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0046] Figure 1 Maintenance procedures for aircraft engines;

[0047] Figure 2 State machine model of aero-engine intelligent agent;

[0048] Figure 3 Simulation flowchart;

[0049] Figure 4 Simulation results of engine satisfaction rate;

[0050] Figure 5 The relationship between the number of vehicles returned for major repairs ahead of schedule and the average satisfaction rate; Detailed Implementation

[0051] The embodiments of the present invention are described in detail below. These embodiments are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0052] From the moment of delivery, an aircraft engine enters the operational support phase, and its support process is as follows: Figure 1As shown. First, aircraft engines operate on board. If a malfunction occurs or reaches the end of its service life, it needs to be sent to a factory for repair. After repair or overhaul, it is returned to the factory as a spare engine for replacement when needed. If the engine reaches the end of its service life, it is scrapped.

[0053] Clearly, a certain number of aircraft engines should be kept in reserve during peacetime to cope with potential engine shortages. Too few engines in reserve will result in no engines available during shortages, affecting the aircraft's operational readiness; too many engines in reserve will lead to resource waste. This invention proposes an aircraft engine support prediction method based on multi-agent modeling. This method can determine how many spare engines should be equipped to meet a given level of aircraft operational readiness.

[0054] The method specifically includes the following steps:

[0055] Step 1: Design the state machine model of the aero-engine intelligent agent.

[0056] The aforementioned aero-engine intelligent agent state machine model has five states: operating, overhaul, fault return for repair, backup, and scrap. Transitions between these states can occur under certain circumstances, specifically as follows: Figure 2 As shown.

[0057] (1) The transition of an aero-engine from operational status to overhaul status. This transition may involve two scenarios:

[0058] One scenario is premature return for major overhaul. This occurs when an engine experiences a significant malfunction, requiring it to be returned to the factory for major overhaul ahead of schedule. During the overhaul, the remaining stage life becomes zero, and this prematurely consumed stage life is included in the calculation. In the simulation, this transition event occurs based on probability, with the specific probability value referring to historical statistics on the premature return rate of aero-engines.

[0059] Second, the engine reaches its stage life, which is the life corresponding to a major overhaul, such as the stage life T for each stage. S =1000 hours, the total lifespan is 3000 hours, then after the first stage is reached, a major overhaul is required. After the overhaul is completed, the lifespan of the second stage is reset to 1000 hours. After working for t hours, the total lifespan is recorded as 1000+t.

[0060] (2) The transition of an aircraft engine from its working state to its faulty state requiring return to the factory for repair.

[0061] When an engine malfunctions during operation and requires factory repair, it will be returned for repair and then sent back. The stage life will not decrease after returning. For example, if the stage life was T1 before repair, it will remain T1 after returning.

[0062] (3) The transition of an aero-engine from working state to scrapped state.

[0063] When an engine reaches the end of its service life, it is scrapped.

[0064] (4) The aircraft engine transitions from overhaul status to backup status.

[0065] After a certain overhaul cycle (the value is based on historical overhaul cycle statistics), the repaired engine is returned as a backup engine and enters backup status. At this time, the stage life is reassigned to T. S .

[0066] (5) The aircraft engine transitions from a faulty state to a backup state.

[0067] After a certain fault repair cycle (the value is based on historical fault repair cycle statistics), the repaired engine is returned as a backup engine and enters backup status.

[0068] (6) The transition of an aircraft engine from backup state to working state.

[0069] The transition is triggered by a message. When an event occurs where an engine is missing, the system broadcasts a missing engine message and randomly selects one of the backup engines to receive the message, thereby replacing the missing engine and putting it into operation.

[0070] Step 2: Design simulation driving events for aero-engines.

[0071] In order to accurately simulate the aero-engine support process, this invention designs three types of event-driven simulation models that are continuously advanced: flight plan generation events, mission acceptance events, and mission allocation events.

[0072] 1. Flight plan generation event:

[0073] Typically, an aircraft's total annual flight hours are predictable, while each aircraft's daily flight time is random, though statistical patterns exist. Furthermore, to avoid situations where tasks cannot be fully allocated during task assignment, this embodiment employs a two-step allocation strategy. First, based on the predicted total annual flight hours and statistical patterns, a monthly flight plan is randomly generated: the annual flight time is allocated to each month. Since monthly flight time is influenced by seasonal factors, certain statistical patterns also exist. Then, within each month, the total flight time is randomly allocated to each day, forming a daily flight hour plan.

[0074] 2. Mission Acceptance Event

[0075] For engines that are in normal working order and on board, to describe their operating time, a random assignment method is used, randomly assigning them to aircraft in the normal aircraft pool that are capable of performing flight missions. Based on the daily flight hour schedule, the total daily flight time is allocated to each aircraft according to statistical patterns. When the total daily flight time is low, some operational aircraft may not receive assignments, indicating that their engines are not being used and they are in a standby state. This is consistent with objective reality, as long as the daily flight time is allocated.

[0076] 3. Task assignment event

[0077] Extensive statistical data reveals that engine operating time is related to the daily usage of the aircraft: on flight days, the engine's operating hours can be represented by a trigonometric distribution function; on maintenance days, the engine's operating time mainly consists of the time consumed for start-up checks, which is relatively short and can be directly set during simulation; when the aircraft has no flight or maintenance tasks, or when there is a malfunction, the engine's operating time is 0.

[0078] Step 3: Based on the aero-engine intelligent agent state machine model designed in Step 1 and the aero-engine simulation driving events designed in Step 2, perform a multi-agent-based aero-engine support process simulation, and obtain the average satisfaction rate of the aero-engine through simulation.

[0079] The simulation process of aero-engine support based on multi-agent technology is as follows: Figure 3 As shown.

[0080] Step 3.1: Generate an aero-engine agent; initialize parameters according to the set known conditions to set the initial state of the aero-engine; and use the aero-engine simulation driving events designed in Step 2 to randomly generate the daily flight hours of each aero-engine agent.

[0081] Step 3.2: The simulation is executed in days, with each aero-engine intelligent agent following the... Figure 2 The state machine undergoes a state change. It also counts the number of days with missing engines and grounded aircraft each day until the simulation ends. Here, the simulation ends when the usage time reaches one year.

[0082] Step 3.3: Record and display the simulation results. After the simulation, calculate the average satisfaction rate of the aero-engine according to the formula:

[0083]

[0084] Step 4: Adjust the number of spare engines, repeat step 3 to simulate the aircraft engine support process, and finally obtain the number of spare engines that meet the average availability requirements of aircraft engines.

[0085] Here is a specific simulation example. Assume there are 20 engines in total. At the beginning of the year, 2 engines are undergoing major overhauls (repaired for 20 days), 1 engine has been returned to the factory for repair due to a malfunction (repaired for 10 days), 15 engines are in use (assuming there are 15 single-engine aircraft), and 2 engines are reserved as spares. The initial state data for each engine at the beginning of the year are shown in Table 1, and the simulation parameters are shown in Table 2. A one-year simulation cycle is used. Assume the total monthly flight hours are 500 hours.

[0086] Table 1 Simulation parameter settings

[0087]

[0088]

[0089] For the above initial state, the result of a single simulation using the multi-agent-based aero-engine support process simulation flow proposed in this invention is as follows: Figure 4 As shown, Figure 4 The engine satisfaction rate for each day is given. It can be seen that the engine satisfaction rate fluctuates over time, but overall, the average satisfaction rate is 76.25%.

[0090] The results obtained after 10 simulations are shown in Table 2:

[0091] Table 2 Results of 10 simulations

[0092]

[0093] The results of 10 simulations show that the average engine satisfaction rate ranges from 68.57% to 84.51%, with an average of 75.86%, which does not meet the requirement that the average engine satisfaction rate should not be less than 90%. Therefore, simulations were conducted with the addition of 1, 2, and 3 new spare engines, and the results are shown in Tables 3, 4, and 5.

[0094] Table 3 Simulation results when one new standby engine is added.

[0095]

[0096]

[0097] Table 4 Simulation results when two new backup engines are added.

[0098]

[0099] Table 5 Simulation results when 3 new standby engines are added.

[0100]

[0101]

[0102] As shown in the table, when one additional backup engine is added, the average engine fulfillment rate increases to 87.12%; when two additional backup engines are added, the average fulfillment rate increases to 89.47%; and when three additional backup engines are added, the average fulfillment rate increases to 93.16%. This demonstrates that the average engine fulfillment rate increases significantly with the increase in the number of backup engines. It also indicates that if an average engine fulfillment rate of no less than 90% is required, at least three additional backup engines are needed.

[0103] The simulation results above show that the aero-engine support simulation method based on multi-agent simulation can effectively simulate the support process of a fleet of aero-engines and accurately obtain the number of spare engines that should be equipped under a given level of combat readiness.

[0104] Furthermore, the simulation method proposed in this invention can also be used to obtain the correlation between various parameters and the average satisfaction rate of the engine during the aero-engine support process. As shown in Table 6, these are the correlation coefficients between various parameters and the average satisfaction rate of the engine after 10 simulations in the initial state.

[0105] Table 6. Correlation coefficients between various parameters and average engine satisfaction rate.

[0106]

[0107] The number of engine overhauls has remained relatively stable at 17-19, with a correlation of only 0.3547 with engine fulfillment rate; the number of engines reaching their service life stage is approximately 15-18, with a correlation of 0.4737 with engine fulfillment rate; however, the number of engines returned to the factory for overhaul ahead of schedule has a significant impact on the average engine fulfillment rate. Figure 5 As shown in Table 2, for example, when the number of early overhauls is 4, the average engine fulfillment rate is the lowest at 68.57%. The number of engine returns due to malfunction has little impact on the average engine fulfillment rate. The number of retired engines remains relatively stable at 2-3 units, having a very small impact on the average engine fulfillment rate. The cumulative grounding days due to engine shortages are generally inversely proportional to the average engine fulfillment rate. From the above analysis, it is clear that there is a significant correlation between the average engine fulfillment rate and the number of early engine returns and the cumulative grounding days due to engine shortages, as shown in Table 6.

[0108] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention without departing from the principles and spirit of the present invention.

Claims

1. A predictive method for aero-engine support based on multi-agent modeling, characterized in that: Includes the following steps: Step 1: Design the state machine model of the aero-engine intelligent agent; The aero-engine intelligent agent state machine model has 5 states: working, overhaul, fault return for repair, backup, and scrap. The transition models for each state of the aero-engine intelligent agent state machine model include: (1) The transition of an aero-engine from working state to overhaul state can be divided into two situations: early return to the factory for overhaul and engine reaching the stage of life. When an aero-engine experiences a major malfunction and needs to be returned to the factory for overhaul ahead of schedule, it enters the overhaul state, the remaining life of the stage becomes 0, and the stage life consumed ahead of schedule is included. In the simulation, this situation occurs based on probability, and the probability value is determined with reference to the historical statistical value of the early return rate of aero-engines. When an aircraft engine reaches the end of its service life, it enters a major overhaul phase. (2) Transition of an aero-engine from operational status to failure and return to factory for repair: When an engine malfunctions during operation and requires factory repair, it will be returned to the factory for repair. Once repaired, the engine will be sent back to the factory, and its service life will not be reduced after the engine is returned. (3) The transition of an aero-engine from working state to scrap state: When an engine reaches the end of its service life, it shall be scrapped. (4) Transition of aircraft engines from overhaul status to backup status: After a certain overhaul cycle, the repaired engine is sent back as a backup engine and enters backup status. (5) Transition of aircraft engines from faulty return-to-factory repair status to backup status: After a certain period of fault repair, the repaired engine is sent back as a backup engine and enters backup status. (6) Transition of aircraft engines from backup state to operational state: The transition is triggered by a message; when an event of missing engines occurs, the system broadcasts a missing engine message and randomly selects an engine from the backup engines to receive the message, so as to replace the missing engine and enter the working state. Step 2: Design simulation driving events for aero-engines; Step 3: Based on the aero-engine intelligent agent state machine model designed in Step 1 and the aero-engine simulation driving events designed in Step 2, perform a multi-agent-based aero-engine support process simulation, and obtain the average satisfaction rate of the aero-engine through simulation. Step 4: Adjust the number of spare engines, repeat step 3 to simulate the aircraft engine support process, and finally obtain the number of spare engines that meet the average availability requirements of aircraft engines.

2. The aero-engine support prediction method based on multi-agent modeling according to claim 1, characterized in that: The aero-engine simulation driving events designed in step 2 are flight plan generation event, mission acceptance event, and mission allocation event.

3. The aero-engine support prediction method based on multi-agent modeling according to claim 2, characterized in that: In step 2, the flight plan generation event is achieved through the following process: First, based on the predicted total annual flight hours of the aircraft and statistical patterns, a monthly flight plan for the aircraft is randomly generated, and the annual flight time is allocated to each month. Then, within each month, the total flight time is randomly allocated to each day to form a daily flight hour plan.

4. The aero-engine support prediction method based on multi-agent modeling according to claim 2, characterized in that: In step 2, the task acceptance event is achieved through the following process: according to the daily flight hour plan, the total daily flight time is allocated to each aircraft according to statistical rules.

5. The aero-engine support prediction method based on multi-agent modeling according to claim 2, characterized in that: In step 2, the task allocation event is implemented through the following process: on flight days, the engine operating hours are represented by a triangular distribution function; on maintenance days, the engine operating time is the set time consumption for start-up checks; when the aircraft has no flight or maintenance tasks, or when there is a malfunction, the engine operating time is 0.

6. The aero-engine support prediction method based on multi-agent modeling according to claim 1, characterized in that: In step 3, the simulation process of the aero-engine support system based on multi-agent technology is as follows: Step 3.1: Generate the aero-engine intelligent agent; initialize the parameters according to the set known conditions, set the initial state of the aero-engine, and randomly generate the daily flight hours of each aero-engine intelligent agent using the aero-engine simulation driving events designed in Step 2. Step 3.2: The simulation is executed in days, with each aero-engine agent undergoing state changes; and the number of days with missing engines and grounded aircraft is counted until the simulation ends. Step 3.3: Record and display the simulation results; after the simulation, calculate the average satisfaction rate of the aero-engine according to the formula: 。 7. The aero-engine support prediction method based on multi-agent modeling according to claim 1, characterized in that: In step 3, through multiple simulations, the correlation between each parameter and the engine's average fulfillment rate during the aero-engine support process is obtained, and the support parameters that affect the engine's average fulfillment rate are determined.

Images