Cable and duct joint layout method, device and medium based on deep reinforcement learning
By employing deep reinforcement learning, the problems of high safety risks and low space utilization in aircraft cable and pipe layout were solved, achieving an efficient and safe integrated layout that adapts to dynamic environmental changes and improves the airworthiness and reliability of aircraft.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2026-04-13
- Publication Date
- 2026-07-31
AI Technical Summary
Existing technologies lack joint optimization in aircraft cable and pipeline layout, resulting in high safety risks and low space utilization. At the same time, traditional algorithms have low computational efficiency in dynamic environments and cannot meet engineering design requirements.
By employing a deep reinforcement learning-based approach, the agent is trained to avoid common-cause failures and coupling failures during the layout process by acquiring the coupling influence range of cables and pipes. Combining safety, assembly, and economic constraints, the optimal layout path is generated.
It effectively prevents fault propagation, improves the airworthiness and reliability of complex electromechanical systems in aircraft, and enables efficient layout path generation in dynamic environments.
Smart Images
Figure CN122490747A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of equipment layout design, and in particular relates to a method, equipment, and medium for the joint layout of cable pipelines based on deep reinforcement learning. Background Technology
[0002] With the development of the aviation industry, cables and conduits, as key media for energy supply and signal transmission, directly determine the airworthiness of aircraft through their rational layout. However, when faced with the complex automated layout design of electromechanical products, existing technologies still have the following significant limitations: 1. Safety Risks and Space Waste Due to Independent Layout: Due to the complexity of multidisciplinary coupling, most existing studies optimize the layout of cables and conduits independently. While this separate approach reduces the difficulty of initial modeling, it makes it difficult to achieve overall optimal space utilization. More seriously, this separate layout ignores the mutual influence and coupling constraints between adjacent cables and conduits (such as the minimum safe distance that must be maintained between them), often hiding high safety risks and making it impractical for engineering applications.
[0003] 2. Poor Adaptability to Dynamic Environments: In actual aircraft layout design and implementation, designers need to frequently adjust the position and quantity of components within the layout area to adapt to the overall aircraft performance requirements. This adjustment leads to dynamic changes in the layout space environment, necessitating iterative redesign of cable and piping layout schemes. Traditional evolutionary algorithms (such as genetic algorithms) are mostly only applicable to deterministic environments. When faced with scenarios involving frequent changes in the layout environment, traditional algorithms need to completely iterate to complete the entire path search. Especially when the scale of the layout object increases, the solution time grows exponentially, resulting in extremely poor computational efficiency, which cannot meet the demand for efficient response in actual engineering design.
[0004] It is evident that existing automatic layout design methods still have significant shortcomings: on the one hand, they lack a systematic consideration of the joint layout of cables and conduits, ignoring safety risks and global space utilization; on the other hand, current technologies are only applicable to deterministic regional environments, and when faced with frequent changes in the layout environment, they need to be repeatedly searched, resulting in poor computational efficiency. Summary of the Invention
[0005] To address the shortcomings of existing technologies, embodiments of the present invention provide a method, device, and medium for joint layout of cable conduits based on deep reinforcement learning.
[0006] In a first aspect, embodiments of the present invention provide a cable conduit joint layout method based on deep reinforcement learning, the method comprising: Obtain the design model of the target layout area of the aircraft; Obtain component data, cable data, pipeline data, and coupling influence range between cables and pipelines in the target layout area, and use this data to obtain the layout space of the target layout area. The layout space is processed using the bounding box method to construct a three-dimensional grid space model, and then the three-dimensional grid space model is converted into a two-dimensional grid map. The layout path of each cable and each pipe is solved by a trained agent based on the two-dimensional grid map. The training process of the agent includes: defining the environment, actions, and rewards, and training the agent in combination with joint constraints, including avoiding common cause failures and avoiding cable and pipeline coupling failures.
[0007] In a second aspect, embodiments of the present invention provide an electronic device, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores one or more computer programs that can be executed by the at least one processor, and the one or more computer programs are executed by the at least one processor to enable the at least one processor to perform the above-described deep reinforcement learning-based cable conduit joint layout method.
[0008] Thirdly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described cable conduit joint layout method based on deep reinforcement learning.
[0009] Fourthly, embodiments of the present invention provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the aforementioned deep reinforcement learning-based cable conduit joint layout method.
[0010] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method for the joint layout of cables and conduits. By extracting the coupling influence range between cables and conduits when acquiring data of the target layout area, and considering constraints to avoid common-cause failures and cable / conduit coupling failures during the training process of the agent, the method guides environment construction, action exploration, and reward feedback. This ensures that the layout scheme generated by the agent naturally meets the minimum safe distance requirements between cables and conduits, and between conduits and high-risk failure components. This effectively breaks the fault propagation chain, avoiding the serious consequences of simultaneous failure of conduits within a region due to a single fault source (such as conduit leakage or component rupture), and greatly improving the airworthiness and reliability of complex electromechanical systems in aircraft. Attached Figure Description
[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0012] Figure 1 A flowchart illustrating the cable conduit joint layout method based on deep reinforcement learning provided in an embodiment of the present invention; Figure 2 A schematic diagram of an aircraft landing gear door design model provided in an embodiment of the present invention; Figure 3 A schematic diagram of a three-dimensional grid space model provided in an embodiment of the present invention; Figure 4 A schematic diagram of the joint constraint conditions provided in the embodiments of the present invention; Figure 5 A schematic diagram of the security constraints provided in this embodiment of the invention; Figure 6 A schematic diagram of assembly constraints provided in an embodiment of the present invention; Figure 7 A schematic diagram of reinforcement learning provided in an embodiment of the present invention; Figure 8 A schematic diagram illustrating the optimal layout path provided in an embodiment of the present invention; Figure 9 A schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0013] 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.
[0014] It should be noted that, unless otherwise specified, the features in the following embodiments and implementation methods can be combined with each other.
[0015] like Figure 1 As shown, this embodiment of the invention provides a cable conduit joint layout method based on deep reinforcement learning, the method comprising: Step S1: Obtain the design model of the target layout area of the aircraft; obtain the component data, cable data to be laid, pipeline data to be laid, and coupling influence range between cables and pipelines in the target layout area, and obtain the layout space of the target layout area accordingly.
[0016] Furthermore, the cables in the aircraft are primarily responsible for power distribution and signal transmission, including power supply cables, control and signal cables, etc. The pipelines in the aircraft are responsible for the transport of fluids (liquids and gases), including hydraulic pipelines, fuel pipelines, pneumatic pipelines, etc.
[0017] In some embodiments, the component data of the target layout area includes: component name, component location, quantity, failure mode and its impact; the cable data to be laid includes: cable type, diameter, endpoint coordinates, and whether there is an assembly requirement; the conduit data to be laid includes: conduit type, diameter, endpoint coordinates, and whether there is an assembly requirement.
[0018] The component failure modes and their impacts in the target layout area include: the service conditions of the components in the target layout area, the possible failure modes of each component under the service conditions, the severity, frequency, and detectability of each failure mode, and whether the failure mode causes common failures of cables and pipelines in adjacent areas and the distance of impact.
[0019] In some embodiments, the coupling effect range between cables and conduits is defined as the minimum safe distance between cables, between conduits, and between cables and conduits as specified in the airworthiness guidelines and the Zone Safety Analysis (ZSA) checklist.
[0020] For example, such as Figure 2 As shown in the figure, in this embodiment, the aircraft target layout area design model is taken as an example of an aircraft landing gear bay design model. Components in the landing gear bay area include landing gear struts and wheels, strut assemblies, upper locks, lower locks, and doors. Typical failure modes in the aircraft landing gear bay area include component failures caused by vibration, chemical contamination, and external environmental conditions. The severity, frequency, and detectability of component failures in this area are measured using a scale of 1 to 10, with higher scores indicating higher significance. As shown in Table 1 below, failure modes such as corrosion crack propagation in struts, landing gear tire bursting, and rupture and leakage of fuel and hydraulic lines have relatively high common-cause failure risks. The impact distance of these component failures should be considered when laying cables and / or pipelines.
[0021] Table 1: Failure Modes and Effects of Aircraft Landing Gear Bay Components Step S2: The layout space is processed using the bounding box method to construct a three-dimensional grid space model, and then the three-dimensional grid space model is converted into a two-dimensional grid map.
[0022] Specifically, step S2 includes the following sub-steps: Step S201: Obtain the layout space of the target layout region, simplify the layout space using the bounding box method, and perform dilation processing; uniformly decompose the dilated bounding box according to a preset granularity to construct a three-dimensional grid space model, such as... Figure 3 As shown.
[0023] In this example, the preset granularity is set as the minimum diameter of cables and / or pipes in the target layout area of the aircraft.
[0024] Step S202: The three-dimensional grid space model is projected and reduced to a two-dimensional grid map through coordinate transformation, and the projected component data, cable data to be laid, pipeline data to be laid, and the coupling influence range between the cable and the pipeline are obtained.
[0025] Specifically, the coordinate transformation expression is as follows: in, The original layout space coordinates; The coordinates are those of a two-dimensional raster map. For hierarchical indexes; Preset grid granularity; This represents the floor function.
[0026] Step S203: Based on the projected component data, the cable data to be laid, the pipeline data to be laid, and the coupling influence range between the cable and the pipeline, define the grid type attribute of each grid in the two-dimensional grid diagram; wherein, the grid type attribute includes: prohibited type, priority type, and normal type.
[0027] Furthermore, prohibited types are grids containing components, endpoints of all deployed objects, and paths of deployed objects; priority types are grids adjacent to bulkheads, grids containing component surfaces without common cause failures, etc.; and normal types are feasible grids that are not priority types.
[0028] Furthermore, the logic for determining whether a grid contains a component is that the grid has at least one internal point of a component.
[0029] It should be noted that, since the component model has been expanded, from the perspective of assembly optimization, cable conduits can be assembled on the surface of components that do not have common cause failures. Therefore, the grid where the surface points of the component are located is not defined as prohibited type in this embodiment.
[0030] Step S3: Based on the two-dimensional grid map, the trained agent solves the layout path of each cable and each pipe.
[0031] Specifically, step S3 further includes: Obtain the raster coordinates of key intermediate points in the layout path of each cable and each conduit; The grid coordinates of key intermediate points are converted into raw spatial coordinates and the drawing is saved. The drawing is then processed through the secondary development interface of CAD software to generate a cable and / or pipeline layout model of the target layout area of the aircraft.
[0032] For example, the optimal layout path generated in this instance is as follows: Figure 4 As shown, this diagram, presented in a two-dimensional grid format, visually represents a globally optimal cable and / or conduit layout path generated by the placement agent after comprehensively considering physical, safety, assembly, and economic constraints. The orange grid represents the agent's starting point (top left), and the dark red grid represents the target endpoint (bottom right). Components labeled with index 1 correspond to prohibited attribute grids in the reinforcement learning state space, indicating that these components have an extremely high risk of common-cause failure in the current placement task. The algorithm strictly prunes actions during path solving, ensuring that the agent can never use grids within the common-cause failure range as paths. Components labeled with index 2 correspond to preferred attribute grids in the state space, representing safe areas without common-cause failure risk on their surfaces, and conforming to engineering preferences for cable and / or conduit assembly.
[0033] Figure 4 The light yellow continuous line segment represents the optimal path ultimately planned by the agent. It can be clearly observed that the yellow path perfectly avoids the component with index 1 in the upper left corner as it extends downwards, proving the effectiveness of the safety and physical constraints in the algorithm. Furthermore, the agent did not choose to walk in a straight line suspended in the dark blue space (although short, it does not meet assembly requirements), but instead closely followed the edge of the component with index 2 (for example, the path closely follows the top of the large component on the right and the edge of the left side). This indicates that the agent actively chose to lay out the wires close to the wall in order to obtain the reward of the priority type grid in the reward function. At the same time, while meeting the wall-hugging requirement, the agent strived to maintain the fewest turns and the shortest layout length, ultimately weaving through the complex terrain to the endpoint in the lower right corner. In summary, the agent made a trade-off between the absolutely untouchable danger zone (index 1) and the assembly zone that is best followed (index 2), ultimately outputting an optimal layout scheme that is both safe and compliant, as well as economical and aesthetically pleasing.
[0034] Furthermore, such as Figure 5 As shown, the training process of the agent includes: Step S301: Set joint constraints and define the environment, action, and reward.
[0035] Furthermore, the joint constraints include safety constraints, assembly constraints, economic constraints, sequence constraints, physical constraints, and performance constraints, as shown in Table 2 below.
[0036] Table 2: Joint Constraints In this embodiment, the safety constraints include: avoiding common-cause failures and avoiding cable / conduit coupling failures; the expressions are as follows: In the formula, , , , These represent the sets of components, cables, conduits, and cable / conduit types that have coupling effects, respectively. For components The failure affects the distance. This refers to the minimum safe distance between the outer surfaces of cable conduits. The minimum distance between the outer surfaces of cable conduits is expressed as: In the formula, , Represents a collection object and collection objects The equivalent radius at this point, .
[0037] Furthermore, the state space includes: The first to third dimensions represent the coordinates of the agent's current position, i.e. The expression is: in, For the size of the layout space; Dimensions 4 through 6 represent the distances from the agent's current position to the destination position, respectively. The expression is: in, The raster coordinates of the endpoint; Dimensions 7 through 14 represent the type and distance of the nearest object to the agent's current position in each 90˚ direction, respectively. .
[0038] It should be noted that the object types described in this embodiment include: index 0, the grid at the endpoint position; index 1, the grid with the prohibited type attribute; and index 2, the grid with the preferred type attribute.
[0039] Furthermore, in this example, the action includes four discrete actions: a single-step movement within the grid image along the positive and negative directions of the orthogonal coordinate axes; that is, A = {up, down, left, right}.
[0040] In this example, the expression for the reward is: in, This represents the reward at time step t. Rewards for reaching the finish line For dynamic distance rewards, For length penalty, Penalty for the number of turns, , , and The weight of each reward.
[0041] It should be noted that the dynamic distance reward is: the Euclidean distance between the agent at the current time step and the endpoint is less than that at the previous time step.
[0042] For example, in this embodiment, reaching the finish line is rewarded. The upper limit is preferably +100, with dynamic distance bonus. The preferred value is +0.2, with a length penalty. The preferred value is -0.1, and the penalty for the number of turns. The optimal value is -0.2, where each reward contributes to the total reward through weight control, and also controls large fluctuations in network training.
[0043] Step S302: Initialize the Actor-Critic network, randomly generate component coordinates and the endpoint coordinates of the cable conduit to be laid, update the prohibition type attribute of the grid in the two-dimensional grid diagram according to safety constraints, sequence constraints, and physical constraints, and update the priority type attribute of the grid in the two-dimensional grid diagram according to assembly constraints.
[0044] Specifically, such as Figure 6 As shown, based on the safety constraints of the current object to be deployed, combined with the common cause failure distance of the component and the minimum safe distance of the already deployed object, the prohibition type attribute of the grid in the two-dimensional grid diagram is supplemented to avoid common cause failure and cable pipeline coupling failure.
[0045] The sequence constraint is based on the endpoint coordinates of all objects to be deployed, and the paths of already deployed objects are considered as prohibited type attribute meshes to avoid collisions between cables and conduits.
[0046] Based on the physical constraints and the location of the component, the prohibited type attribute of the grid in the 2D grid diagram is added to avoid passing through the component.
[0047] Based on the assembly constraints and whether the current object to be deployed has wall-hugging requirements, the priority type attribute of the grid in the two-dimensional grid diagram is supplemented to guide the deployment along the aircraft bulkhead.
[0048] For example, such as Figure 7 As shown, Figure 7 This demonstrates the specific mapping and solution process of considering safety constraints and economic constraints (minimum number of turns) in a two-dimensional raster graph. Figure 7 The left image shows the initial environment and attribute definitions, defining the starting position S and the ending position G. The space contains components represented by gray grids (without common-cause failure) and bulkheads represented by dark red grids, which naturally constitute physical barriers that the agent cannot traverse. A blue path S'-G' has already been deployed in the space. Because a coupling failure risk is identified between the deployed object and the object to be deployed, the algorithm not only marks the blue path itself as an obstacle but also extends it outwards according to a preset minimum safe distance, generating a red coupling influence range. During the agent's training state, this red and blue area is uniformly supplemented with grids representing prohibited type attributes. Figure 7 The right-hand diagram illustrates the agent's solution process, starting from point S and searching for a path to point G. Under the strong constraints of physics and safety, the feasible yellow path perfectly avoids the gray areas of the components and the red and blue areas, which carry the risk of coupling failure. Simultaneously, driven by the economic constraint of minimizing the number of turns in the reward function, the agent does not choose a circuitous detour but precisely plans the solution with the fewest turns, containing only the two key intermediate points P1 and P2.
[0049] For example, such as Figure 8 As shown, Figure 8 It demonstrates the combination and solution process of assembly constraints (wall-hugging requirement) and economic constraints (shortest length). Figure 8 The left figure shows the initial environment and attribute definitions, including a starting point S and an ending point G, as well as dark red bulkheads and gray components. The blue path S'-G' in the space is determined to have no coupling failure in this evaluation; therefore, no red danger isolation zone is generated around it, only the blue path itself is classified as prohibited. To meet the assembly requirements of actual engineering where cables and / or conduits need to be laid "along the aircraft bulkhead" or attached to component surfaces, the algorithm generates green grids marked as priority types around the bulkheads and above the components. This is equivalent to setting high-score rewards in the environment, guiding the agent to tend to lay cables in these areas. Figure 8 The right-hand diagram illustrates the agent's solution process. Driven by the economic penalty of minimizing path length in the reward function, the agent, after comprehensively evaluating the rewards from prioritizing assembly and the penalties for excessively long paths, plans a yellow route. This route starts from S, extends directly to the right to node P1, and then descends vertically to G. This route not only perfectly fits the green priority grid on the right bulkhead in its latter half, satisfying the assembly requirements, but also achieves the economic benefits of the shortest path globally.
[0050] In step S303, the agent observes the state in the current environment and obtains a state space expression including its own coordinates, distance from the endpoint, and the type of the nearest object in each direction.
[0051] In step S304, the agent samples actions based on the observed state.
[0052] In this example, physical constraints are applied to dynamically prune the action space before the action is executed, predicting the agent's next action and eliminating actions that enter prohibited attribute grids to prevent cables and / or pipes from colliding with each other or passing through components. This not only avoids training with invalid action values but also eliminates safety risks.
[0053] In step S305, the agent executes the pruned legal action, the environment undergoes a state transition, and the reward function for the current time step is calculated based on economic and performance constraints.
[0054] It should be noted that the agent is penalized for each step it takes or each turn it makes, while it receives a positive reward for getting close to or reaching the destination. This forces the agent to learn to find the optimal path while ensuring safety and compliance.
[0055] Furthermore, the agent training process also includes: calculating the policy entropy value at each time step in real time to monitor the degree of policy exploration; specifically, it includes: When the policy entropy value is continuous Steps below the low entropy threshold When this occurs, it is determined to be in an entropy collapse state; Activating entropy collapse intervention increases the entropy reward coefficient, with an upper limit equal to the maximum entropy coefficient. ; When the policy entropy value rises back to the high entropy threshold At that time, the original entropy reward coefficient is restored.
[0056] Furthermore, entropy collapse intervention involves reducing the learning rate of the policy network to prevent training oscillations.
[0057] For example, low entropy threshold The preferred value is 0.1, a high-entropy threshold. The preferred value is 0.3, the maximum entropy coefficient. The preferred value is 0.3. The preferred value range is 5 to 10.
[0058] Furthermore, in this example, the parameter settings for reinforcement learning are shown in Table 3 below.
[0059] Table 3: Training network parameter settings for reinforcement learning In summary, this invention provides a method for the joint layout of cables and conduits. By accurately extracting the coupling influence range between cables and conduits when acquiring target layout area data, and considering constraints to avoid common-cause failures and cable / conduit coupling failures during the agent's training process, the invention guides environment construction, action exploration, and reward feedback. This ensures that the layout scheme generated by the agent naturally meets the minimum safe distance requirements between cables and conduits, and between conduits and high-risk failure components. This effectively breaks the fault propagation chain, avoiding the serious consequences of simultaneous conduit failures within a region due to a single fault source (such as conduit leakage or component rupture), and significantly improving the airworthiness and reliability of complex aircraft electromechanical systems.
[0060] Meanwhile, this invention introduces a deep reinforcement learning algorithm framework to replace the traditional evolutionary algorithm for solving the joint layout problem of cable conduits. Since the agent has already traversed and learned a massive amount of random environmental states during the training phase, it possesses extremely strong spatial generalization perception capabilities. When the actual layout environment undergoes dynamic changes such as the addition, deletion, or repositioning of components, this invention does not require time-consuming iterative searches again. Instead, it directly inputs the changed new environment as the observation state into the trained agent, instantly outputting a new optimal layout path through millisecond-level network inference. This technique fundamentally reduces the runtime of iteratively generating solutions due to frequent modifications to the layout environment, perfectly breaking through the computational bottleneck of traditional algorithms and enabling highly efficient adaptation to dynamic and complex aerospace engineering design requirements.
[0061] Accordingly, this application also provides an electronic device, including: one or more processors; a memory for storing one or more programs; and when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the deep reinforcement learning-based cable conduit joint layout method as described above. Figure 9 The diagram shown illustrates a hardware structure of any device with data processing capabilities for the cable conduit joint layout method based on deep reinforcement learning provided in this embodiment of the invention, except... Figure 9 In addition to the processor, memory, and network interface shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.
[0062] Accordingly, this application also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the cable conduit joint layout method based on deep reinforcement learning as described above. The computer-readable storage medium can be an internal storage unit of any data processing device as described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units of any data processing device and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.
[0063] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only.
[0064] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.
Claims
1. A cable conduit joint layout method based on deep reinforcement learning, characterized in that, The method includes: Obtain the design model of the target layout area of the aircraft; Obtain component data, cable data, pipeline data, and coupling influence range between cables and pipelines in the target layout area, and use this data to obtain the layout space of the target layout area. The layout space is processed using the bounding box method to construct a three-dimensional grid space model, and then the three-dimensional grid space model is converted into a two-dimensional grid map. The layout path of each cable and each pipe is solved by a trained agent based on the two-dimensional grid map. The training process of the agent includes: defining the environment, actions, and rewards, and training the agent in combination with joint constraints, including avoiding common cause failures and avoiding cable and pipeline coupling failures.
2. The cable conduit joint layout method based on deep reinforcement learning according to claim 1, characterized in that, The component data for the target layout area includes: component name, component location, quantity, failure mode and its impact; among which, failure mode and its impact includes: service conditions in the target layout area, possible failure modes of each component under the conditions, severity, frequency, detectability of each failure mode, and whether the failure mode will cause common failure of cables and pipelines in the area and the affected distance. The data for the cables to be laid includes: cable type, diameter, endpoint coordinates, and whether assembly is required; The pipeline data to be laid includes: pipeline type, diameter, endpoint coordinates, and whether there are assembly requirements.
3. The cable conduit joint layout method based on deep reinforcement learning according to claim 1, characterized in that, The process of constructing a 3D raster space model and converting it into a 2D raster image includes: The bounding box is uniformly decomposed according to the preset granularity to construct a three-dimensional grid space model; The three-dimensional grid space model is projected and reduced to a two-dimensional grid map through coordinate transformation, and the projected component data, cable data to be laid, pipeline data to be laid, and the coupling influence range between the cable and pipeline are obtained. Based on the projected component data, the cable data to be laid, the pipeline data to be laid, and the coupling influence range between the cable and the pipeline, the grid type attribute of each grid in the two-dimensional grid diagram is defined; wherein, the grid type attribute includes: prohibited type, priority type, and normal type.
4. The cable conduit joint layout method based on deep reinforcement learning according to claim 3, characterized in that, The preset granularity is set as the minimum diameter of cables and / or pipes in the target layout area of the aircraft.
5. The cable conduit joint layout method based on deep reinforcement learning according to claim 1, characterized in that, The training process of the agent includes: The status includes: the coordinates of the current position, the distance of the current position from the destination position, and the type and distance of the nearest object to the current position in each 90˚ direction; The actions include: four discrete actions of single-step movement along the positive and negative directions of the orthogonal coordinate axes within the grid diagram; The rewards include: rewards for reaching the finish line, dynamic distance rewards from the finish line, penalties for the length of cables and / or conduits, and penalties for the number of turns in cables and / or conduits. The joint constraints include safety constraints, assembly constraints, economic constraints, sequence constraints, physical constraints, and performance constraints. Safety constraints include: avoiding common-cause failures and avoiding cable / conduit coupling failures. Assembly constraints include: cables and / or conduits should be laid along the aircraft bulkhead. Economic constraints include: minimizing the length of cables and / or conduits and minimizing the number of bends in cables and / or conduits. Sequence constraints include: prioritizing cables and / or conduits between important equipment and prioritizing large-diameter cables and / or conduits. Physical constraints include: cables and / or conduits should avoid colliding with each other or passing through components. Performance constraints include: minimizing cable slack deviation and minimizing conduit friction loss.
6. The cable conduit joint layout method based on deep reinforcement learning according to claim 3 or 5, characterized in that, The training process of the agent includes: Initialize the environment; including: updating the prohibition type attribute of the grid in the two-dimensional grid diagram according to safety constraints, sequence constraints, and physical constraints, and updating the priority type attribute of the grid in the two-dimensional grid diagram according to assembly constraints; Perform actions; including: prior to performing an action, eliminating actions that enter prohibited type attribute grids based on physical constraints to prevent cables and / or pipes from colliding with each other or passing through components; Calculate rewards; including: calculating reward functions based on economic constraints and performance constraints.
7. The cable conduit joint layout method based on deep reinforcement learning according to claim 1, characterized in that, After solving the layout path of each cable and each pipe using a trained agent based on the two-dimensional grid map, the following steps are also included: Obtain the raster coordinates of key intermediate points in the layout path of each cable and each conduit; The raster coordinates of key intermediate points are converted into raw spatial coordinates to generate a cable and / or pipeline layout model of the target layout area of the aircraft.
8. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores one or more computer programs that can be executed by the at least one processor, the one or more computer programs being executed by the at least one processor to enable the at least one processor to perform the cable conduit joint layout method based on deep reinforcement learning as described in any one of claims 1-7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the cable conduit joint layout method based on deep reinforcement learning as described in any one of claims 1-7.
10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the cable conduit joint layout method based on deep reinforcement learning as described in any one of claims 1-7.