PCB intelligent design method and system
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUBEI UNIV OF TECH
- Filing Date
- 2026-05-08
- Publication Date
- 2026-06-02
Smart Images
Figure CN122133595A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of PCB design automation and machine learning technology, and in particular to a PCB intelligent design method and system. Background Technology
[0002] As electronic products develop towards higher integration, higher speed and smaller size, the number of PCB layers, the number of networks and design constraints continue to increase, significantly increasing the complexity of routing tasks. Especially in high-speed signal, differential pair signal and high-density interconnect scenarios, routing not only needs to meet basic indicators such as line length and number of vias, but also needs to strictly follow design rules such as impedance consistency, topology order and electromagnetic compatibility. This makes PCB routing gradually become one of the most time-consuming and most dependent on engineering experience in the design process.
[0003] Most existing PCB automatic routing technologies are based on rule-driven or fixed heuristic strategies. They complete the routing process by pre-setting priorities, cost functions, and search parameters. These methods usually fix the network order and parameter configuration at the beginning of the routing process, making it difficult to dynamically adjust the strategy based on intermediate routing results. When routing space is limited or there are many conflicts between networks, problems such as insufficient routing rate, severe local congestion, or a large number of design rule violations are likely to occur. Often, manual intervention or repeated routing is required to obtain acceptable results. Overall efficiency and global optimization capabilities are limited. Therefore, in complex PCB design scenarios, the PCB routing quality is not high. Summary of the Invention
[0004] In view of this, it is necessary to provide a PCB intelligent design method and system to solve the technical problem of low PCB routing quality in complex PCB design scenarios.
[0005] To address the aforementioned problems, in a first aspect, the present invention provides a PCB intelligent design method, comprising: Step 1: Build a PCB knowledge base based on the acquired PCB board layout information, component placement, netlist, and design rules; Step 2: Construct a routing execution layer scheduled by an intelligent agent based on the PCB knowledge base; Step 3: Determine the set of unrouted networks based on the PCB knowledge base, select the network to be routed at the current time from the set of unrouted networks, and route the network to be routed at the current time based on the key parameters of the network to be routed at the current time and the corresponding routing execution layer to obtain the routing result of the network to be routed at the current time. Step 4: After performing a disconnection operation on the routing result of the network to be routed at the current moment, update the set of unrouted networks. Select the network to be routed at the next moment from the updated set of unrouted networks. Based on the key parameters of the network to be routed at the next moment and the corresponding routing execution layer, route the network to be routed at the next moment to obtain the routing result of the network to be routed at the next moment. Step 5: Take the routing result of the network to be routed at the next moment as the new routing result of the network to be routed at the current moment, and repeat Step 4 until the preset maximum number of dismantling and rerouting iterations is reached or the routing rate of the network to be routed at the current moment and the network to be routed at the next moment reaches a stable balance, then stop the iteration to obtain the PCB routing result.
[0006] In one possible implementation, the construction of the PCB knowledge base based on the acquired PCB layout information, component placement, netlist, and design rules includes: The acquired PCB board layout information, component layout, netlist, and design rules are structured and encoded to build a PCB knowledge base. The netlist includes network types and network connection relationships, and the network types include ordinary networks and differential pair networks.
[0007] In one possible implementation, the routing execution layer includes a 3D path exploration module, a differential pair topology adaptive routing module, a routing cost control module, an exploration rate adjustment module, and a spatial resolution adjustment module; the routing of the network to be routed at the current moment based on the determined key parameters of the network to be routed at the current moment and the corresponding routing execution layer includes: The type of the network to be wired at the current moment is determined. When the type of the network to be wired at the current moment is a normal network, the key parameters are input to the wiring cost control module, the exploration rate adjustment module, and the spatial resolution adjustment module. Then, the network to be wired at the current moment is wired based on the wiring cost control module, the exploration rate adjustment module, the spatial resolution adjustment module, and the three-dimensional path exploration module. When the type of the network to be routed at the current moment is a differential pair network, after inputting the key parameters into the routing cost control module, the exploration rate adjustment module, and the spatial resolution adjustment module, the network to be routed at the current moment is routed based on the routing cost control module, the exploration rate adjustment module, the spatial resolution adjustment module, and the differential pair topology adaptive routing module.
[0008] In one possible implementation, the key parameters include routing cost weight parameters, exploration step size parameters, and spatial resolution parameters.
[0009] In one possible implementation, the routing of the network to be routed at the current moment based on the routing cost control module, the exploration rate adjustment module, the spatial resolution adjustment module, and the three-dimensional path exploration module includes: The weight parameters in the comprehensive cost function are set based on the routing cost control module, wherein the weight parameters include the line length cost weight, the via number cost weight, and the design rule violation number cost weight. The exploration step size parameter of the three-dimensional path exploration module is set based on the exploration rate adjustment module; The spatial resolution parameters of the three-dimensional path exploration module are set based on the spatial resolution adjustment module. Based on the weight parameters, exploration step size parameters, and spatial resolution parameters in the comprehensive cost function, a three-dimensional path exploration module is used to route the network to be routed at the current moment.
[0010] In one possible implementation, the step of using a three-dimensional path exploration module to route the network to be routed at the current moment, based on the weight parameters, exploration step size parameters, and spatial resolution parameters in the comprehensive cost function, includes: Constructing a three-dimensional discrete space based on a PCB knowledge base; Construct the comprehensive cost function based on the weight parameters in the comprehensive cost function; The starting and ending points of the network to be wired at the current moment are determined based on the three-dimensional discrete space. The routing path of the network to be routed at the current moment is explored with the starting point as the initial exploration node, the exploration step size parameter as the exploration step size, and the goal of minimizing the comprehensive cost function. The routing path that satisfies the design rules and minimizes the comprehensive cost is obtained. The spatial resolution of the candidate positions in the routing path is verified based on the spatial resolution parameter.
[0011] In one possible implementation, the comprehensive cost function is: , in, The value of the comprehensive cost function. For cumulative line length, To accumulate the number of vias, To accumulate the number of rule violations, , , These are the line length cost weight, the via quantity cost weight, and the design rule violation frequency cost weight, respectively.
[0012] In one possible implementation, the routing of the network to be routed at the current moment based on the routing cost control module, the exploration rate adjustment module, the spatial resolution adjustment module, and the differential pair topology adaptive routing module includes: Determine the start and end points of the network to be wired at the current moment; Starting from the positive pin on the starting side and the fan-out target corresponding to the positive pin on the starting side as the endpoint, the three-dimensional path exploration module is called to explore the routing path on the starting side. Starting from the negative pin on the starting side and the fan-out target corresponding to the negative pin on the starting side as the endpoint, the three-dimensional path exploration module is called to explore the routing path on the starting side to obtain the candidate routing results on the starting side. Starting from the positive pin on the endpoint side and taking the fan-out target corresponding to the positive pin on the endpoint side as the endpoint, the three-dimensional path exploration module is called to explore the routing path on the starting side. Starting from the negative pin on the endpoint side and taking the fan-out target corresponding to the negative pin on the endpoint side as the endpoint, the three-dimensional path exploration module is called to explore the routing path on the starting side to obtain the candidate routing results on the endpoint side. With the connection between the candidate routing results on the starting side and the candidate routing results on the ending side as the goal, and the center coordinates of the positive and negative fan-out points on the starting side as the starting point of the middle section and the center coordinates of the positive and negative fan-in points on the ending side as the ending point of the middle section, the three-dimensional path exploration module is called to explore the routing path of the middle section and obtain the center trajectory of the middle section. Based on the candidate routing results at the starting point, the candidate routing results at the ending point, and the center trajectory of the middle section, the positive and negative paths of the network to be routed at the current moment are obtained. After verifying the positive and negative paths based on the design rules, the final routing path of the network to be routed at the current moment is obtained.
[0013] In one possible implementation, the mid-segment center trajectory is: , in, This is a 3D path exploration module. The trajectory is the center of the middle section. The center coordinates of the positive and negative fan-out points on the starting side. The coordinates are the center coordinates of the positive and negative fan-in points on the endpoint side.
[0014] Secondly, the present invention also provides a PCB intelligent design system, comprising: The knowledge base construction module is used to build a PCB knowledge base based on the acquired PCB board layout information, component layout, netlist, and design rules. The routing execution layer construction module is used to construct a routing execution layer scheduled by an intelligent agent based on the PCB knowledge base; The routing module is used to determine the set of unrouted networks based on the PCB knowledge base, select the network to be routed at the current time from the set of unrouted networks, and perform routing on the network to be routed at the current time based on the key parameters of the network to be routed at the current time and the corresponding routing execution layer, so as to obtain the routing result of the network to be routed at the current time. The cable teardown and rerouting module is used to tear down the routing results of the network to be routed at the current moment, update the set of unrouted networks, select the network to be routed at the next moment from the updated set of unrouted networks, and route the network to be routed at the next moment based on the key parameters of the network to be routed at the next moment and the corresponding routing execution layer to obtain the routing result of the network to be routed at the next moment. The multiple disconnect and redistribution module is used to take the routing result of the network to be routed at the next moment as the routing result of the network to be routed at the current moment, and disconnect and redistribute the network to be routed at the current moment until the preset maximum number of disconnect and redistribution iterations is reached or the routing success rate of the network to be routed at the current moment and the network to be routed at the next moment reaches a stable balance, and then the iteration stops to obtain the PCB routing result.
[0015] The beneficial effects of this invention are as follows: A routing execution layer scheduled by an intelligent agent is constructed based on a PCB knowledge base; a set of unrouted networks is determined based on the PCB knowledge base; a network to be routed at the current moment is selected from the unrouted network set; and routing is performed on the network to be routed at the current moment based on the determined key parameters of the network to be routed at the current moment and the corresponding routing execution layer, obtaining the routing result of the network to be routed at the current moment; after performing a disconnection operation on the routing result of the network to be routed at the current moment, the set of unrouted networks is updated; a network to be routed at the next moment is selected from the updated unrouted network set; and routing is performed on the network to be routed at the next moment based on the determined key parameters of the network to be routed at the next moment and the corresponding routing execution layer, obtaining the network to be routed at the next moment. The routing results are used as follows: the routing results of the network to be routed at the next moment are used as the routing results of the network to be routed at the new current moment. The new current moment network to be routed is then stripped and rerouted until the preset maximum number of stripping and rerouting iterations is reached or the routing success rate of the network to be routed at the current moment and the network to be routed at the next moment reaches a stable equilibrium. The iteration stops when this happens, thus obtaining the PCB routing results. By performing multiple rounds of stripping and rerouting on the network to be routed at the current moment and the next moment, and by selecting and adjusting the key parameters of the network to be routed at different moments during the multiple rounds of stripping and rerouting, the joint decision-making and routing optimization of the network to be routed at different moments are realized, which improves the overall routing success rate of the PCB and the routing quality of the PCB in complex PCB design scenarios. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.
[0017] Figure 1 A flowchart of an embodiment of the PCB intelligent design method provided by the present invention; Figure 2 A flowchart of a common network routing process for the three-dimensional path exploration module of the PCB intelligent design method provided by the present invention; Figure 3 The flowchart of differential pair network routing in the differential pair topology adaptive routing module of the PCB intelligent design method provided by the present invention; Figure 4 This is a structural schematic diagram of an embodiment of the PCB intelligent design system provided by the present invention. Detailed Implementation
[0018] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.
[0019] In this document, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0020] A specific embodiment of the present invention discloses a PCB intelligent design method, such as... Figure 1 As shown, the PCB intelligent design method includes: S101. Construct a PCB knowledge base based on the acquired PCB layout information, component placement, netlist, and design rules.
[0021] S102. Construct a routing execution layer scheduled by an intelligent agent based on the PCB knowledge base; It should be noted that the routing execution layer includes a 3D path exploration module, a differential pair topology adaptive routing module, a routing cost control module, an exploration rate adjustment module, and a spatial resolution adjustment module. The corresponding 3D path exploration module or differential pair topology adaptive routing module is selected for routing according to the network type.
[0022] S103. Determine the set of unrouted networks based on the PCB knowledge base, select the network to be routed at the current time from the set of unrouted networks, and perform routing on the network to be routed at the current time based on the key parameters of the network to be routed at the current time and the corresponding routing execution layer, so as to obtain the routing result of the network to be routed at the current time.
[0023] S104. After performing a teardown operation on the routing result of the network to be routed at the current moment, update the set of unrouted networks, select the network to be routed at the next moment from the updated set of unrouted networks, and perform routing on the network to be routed at the next moment based on the key parameters of the network to be routed at the next moment and the corresponding routing execution layer to obtain the routing result of the network to be routed at the next moment.
[0024] Step S105: Use the routing result of the network to be routed at the next moment as the new routing result of the network to be routed at the current moment. Repeat step S104 until the preset maximum number of iterations for unrobbing and rerouting is reached or the routing success rate of the network to be routed at the current moment and the network to be routed at the next moment reaches a stable balance. Stop the iteration to obtain the PCB routing result.
[0025] It should be noted that the selection of routing networks and the adjustment of key parameters corresponding to different routing networks during multiple rounds of routing and rerouting improved the overall routing success rate of the PCB and the routing quality of the PCB in complex PCB design scenarios. The specific conditions for the routing success rate to reach a stable equilibrium are that the improvement of the overall routing success rate of the PCB routing tends to be gradual, the difference in routing success rate between two adjacent iterations is less than a certain threshold, and key networks (such as high-speed signals and power supplies) are 100% routed, and the number of DRC (Design Rule) errors tends to be zero.
[0026] In some embodiments, in step S101, a PCB knowledge base is constructed based on the acquired PCB layout information, component placement, netlist, and design rules. The acquired PCB layout information, component placement, netlist, and design rules are structured and encoded to construct the PCB knowledge base. The netlist includes network types and network connection relationships. Network types include ordinary networks and differential pair networks. The PCB layout information, component placement, network connection relationships, and design rules are uniformly structured and modeled to form a PCB knowledge base that can be directly used for reasoning and scheduling. This transforms the original PCB design data into a unified, structured knowledge representation suitable for agent scheduling and reasoning. Determined connection relationships are used as input constraints for subsequent automatic routing. First, the PCB layout data set is determined, consisting of four types of data: layout information, component placement, netlist, and design rules, and described in a unified coordinate system. The layout information is defined as... It is used to describe the geometric boundaries and stack-up structure of a PCB, including the outer border polygon. and layer set The outer polygonal border is used to define the planar area on the PCB where wiring is possible, and each layer... Corresponding to a fixed layer number This allows for the determination of unique height coordinates in three-dimensional space; device layout is defined as... This describes the placement of all components on the PCB, and each component... Includes device reference designation, device package geometry, and pinout. Each pin Based on pin number, pin geometry, and pin center coordinates The only certainty is that, The layer number is consistent with the layer number where the pin is located; the netlist is defined as follows: It consists of multiple networks Each network consists of a set of pins with electrical connections. Based on their electrical properties, networks are classified into two categories: ordinary networks and differential pair networks. For ordinary networks… It contains a single pin set; for differential pair networks It consists of two sub-networks, one for positive terminals and one for negative terminals, corresponding to the pin sets respectively. and Meanwhile, to ensure that subsequent cabling tasks have a definite input, the connection set is directly given as part of the netlist and input along with the netlist. The connection set of a regular network is defined as follows: And the connection relationship of each atom is defined as The set of connection relationships in a differential pair network is defined as follows: And each difference connection is defined as Design rules are defined as follows: It is used to constrain the geometric parameters of conductors and vias and their interrelationships during the routing process, including linewidth parameters corresponding to different network types. Through-hole diameter parameters and the minimum spacing parameter between different network objects The network type parameters serve as inviolable hard constraints during subsequent routing exploration, incorporating board layout information. Component layout Netlist and design rules Perform unified structured coding to form a PCB knowledge base. , Represented in a textual structure, its structure is determined by the following four fields: ,in The field contains both the pin set and the connection relationship set for each network. This allows subsequent wiring modules to directly use the connection relationships as the set of wiring input tasks without relying on additional preprocessing.
[0027] In some embodiments, in step S102, a routing execution layer scheduled by an intelligent agent is constructed based on a PCB knowledge base. This routing execution layer includes a 3D path exploration module, a differential topology adaptive routing module, a routing cost control module, an exploration rate adjustment module, and a spatial resolution adjustment module. These modules collaboratively complete automatic routing operations that satisfy design rule constraints within a unified 3D spatial state. As an execution layer scheduled by the intelligent agent, the routing execution layer consists of multiple cooperative and configurable functional modules. All modules operate in a unified 3D spatial state and receive calling instructions and parameter configurations from the intelligent agent through a standardized interface. The 3D path exploration module is defined as... This module is used to perform path search in a three-dimensional discrete space for a given start and end point while satisfying design rule constraints, and generates a routing path based on the comprehensive cost of line length, via quantity, and design rule violation. The differential pair topology adaptive routing module is defined as follows: This module is used for differential pair networks to perform coordinated routing of the positive and negative paths during the fan-out, parallel routing, and splitting phases at the start and end points. It continuously maintains the topological order of the differential pair paths during routing to ensure that the final routing path matches the differential pin arrangement order. The routing cost control module is defined as follows: This module is used by the agent to directly set the weight parameters in the comprehensive cost function before network routing begins. During path exploration, it sets the cost weights for line length, via quantity, and design rule violations to adjust the emphasis of different networks on routing optimization objectives. The exploration rate adjustment module is defined as follows: This module is used to directly set the exploration step size parameter of the 3D path exploration module during state expansion by the intelligent agent. Based on the spatial region where the candidate exploration position is located, the exploration step size in the 3D path exploration module is regionally adjusted. A smaller exploration step size is used in the obstacle neighborhood and the endpoint neighborhood, while a larger exploration step size is used in ordinary areas, improving overall exploration efficiency while ensuring wiring accuracy. The spatial resolution adjustment module is defined as... It is used to allow the agent to directly set the spatial resolution parameters of the 3D path exploration module during state pruning. Based on the spatial region where the candidate exploration position is located, the minimum repeated exploration distance in the 3D path exploration module is set regionally. A smaller spatial resolution is used in the obstacle neighborhood and the endpoint neighborhood, and a larger spatial resolution is used in the ordinary region, so as to compress the search space and maintain the path expression ability in the key region. Intelligent agents are defined as The intelligent agent, acting as the core of decision-making and scheduling, collaborates with the core cabling module system, which serves as the cabling execution layer, driving the orderly execution of the entire automated cabling process. Built upon a general-purpose large language model foundation, the selected large model foundation is limited to a general-purpose autoregressive language model that supports multi-step complex reasoning, long-context understanding, planning capabilities, and external tool invocation capabilities. This is to ensure... The behavior in cabling tasks is goal-oriented and subject to professional constraints. System-level prompts are configured for this behavior; these prompts are defined as follows: As a fixed context that cannot be overridden by subsequent user prompts, the system input layer injected into the large model base is used to define the agent's role, professional capability boundaries, and global optimization goals. An example of its system prompt is: "A professional PCB autorouting agent possesses complete PCB design knowledge, understands multi-layer PCB board structures, component placement, netlist connections, and design rule constraints. Its task is to complete PCB routing step-by-step according to system scheduling during the autorouting process. The global optimization goal is to maximize the overall routing success rate while minimizing bus length, via count, and DRC violation count, while satisfying all design rule constraints. It needs to follow established functional module call specifications during routing and aims for continuous improvement in global routing quality." This is the structure of the agent. In the system prompt Under strong constraints, it possesses stable role cognition and goal orientation capabilities, providing a unified behavioral constraint basis for subsequent intelligent agents to carry out wiring network planning and parameter scheduling.
[0028] In some embodiments, in step S103, a set of unrouted networks is determined based on the PCB knowledge base. The network to be routed at the current moment is selected from this set. Based on the determined key parameters of the network to be routed at the current moment and the corresponding routing execution layer, routing is performed on the network to be routed at the current moment to obtain the routing result. Before routing each network begins, the agent selects the network to be routed at the current moment from the set of unrouted networks. The set of unrouted networks is as follows: , The type of the network to be routed at the current moment is determined. This network can be either a normal network or a differential pair network, with differential pair networks typically preferred. When the network is a normal network, key parameters are input into the routing cost control module, exploration rate adjustment module, and spatial resolution adjustment module. The routing is then performed based on these modules, along with the 3D path exploration module. When the network is a differential pair network, key parameters are input into these modules as well. The routing is then performed based on these modules, along with the differential pair topology adaptive routing module, to obtain the routing result for the network at the current moment. When the network to be routed is a normal network at the current moment, the weight parameters in the comprehensive cost function are set based on the routing cost control module. These weight parameters include the cost weight for wire length, the cost weight for the number of vias, and the cost weight for the number of design rule violations. The interface of the routing cost control module is as follows: , in, Indicates the cost weight of line length. Indicates the cost weight of the number of vias. The cost weight for the number of DRC violations; The exploration step size parameter of the 3D path exploration module is set based on the exploration rate adjustment module. Explore step size In the 3D path exploration module, the minimum advance distance is defined as the distance when extending a planar traverse. Its value controls the rate of state expansion during path exploration, i.e., the number of candidate states that can be generated per unit time, thus affecting the path search speed and computational resource consumption, without changing the definition of path geometric accuracy. To differentiate the exploration rate in different spatial regions, the 3D space is divided into three types of regions based on the current wiring request: obstacle neighborhood, endpoint neighborhood, and normal region. Different exploration step sizes are set for each region. The obstacle neighborhood is defined as a spatial region that satisfies the following condition: the distance from the candidate exploration position to the obstacle set. The minimum distance is less than the preset threshold The endpoint neighborhood is defined as a spatial region that satisfies the following condition: the Euclidean distance from the candidate exploration location to the target endpoint coordinates is less than a threshold. Spaces other than the two types of regions mentioned above are uniformly defined as ordinary regions. Region determination is performed in real-time based on the spatial coordinates of candidate positions before each state expansion of the 3D path exploration module. Within the obstacle neighborhood and the endpoint neighborhood, the exploration step size... A smaller value is chosen to reduce the speed of state expansion, resulting in a smoother search process in complex or critical regions; in ordinary regions, the exploration step size is adjusted. A larger value is chosen to increase the state expansion rate and accelerate the path search in open space. By setting different exploration step sizes, the overall exploration rate can be adjusted without changing the wiring accuracy constraints. The interface of the exploration rate adjustment module is as follows: , in, The value is the exploration step size corresponding to the neighborhood of the obstacle. The value is set to the exploration step size corresponding to the neighborhood of the endpoint. The step size is the value corresponding to the normal area. The step size parameter is used in the 3D path exploration module. The source of regionalized values; The spatial resolution parameters of the 3D path exploration module are set based on the spatial resolution adjustment module. In the 3D path exploration module, the minimum repeated exploration distance is defined and used to determine whether a new candidate exploration location has been visited at the current resolution, thus deciding whether to prune the candidate location. The spatial resolution adjustment module adopts the same region division method as the exploration rate adjustment module, dividing the 3D space into three types of regions: obstacle neighborhood, endpoint neighborhood, and normal region. For each candidate exploration location, the region to which it belongs is first determined based on the distance relationship between the candidate exploration location and the set of obstacles and the target endpoint coordinates. Then, the spatial resolution parameter of the corresponding region is selected as the repeated exploration determination threshold for that location. During the path exploration process, when the Euclidean distance between the new candidate exploration location and the already explored location is less than the minimum repeated exploration distance threshold corresponding to the region to which the candidate location belongs, the repeated exploration is determined. When a candidate position is deemed a duplicate, it is pruned and the search is not expanded further. By using smaller spatial resolution parameters in the obstacle and endpoint neighborhoods and larger spatial resolution parameters in ordinary regions, the feasible solution space is effectively compressed while maintaining the expressive power of path search within critical regions. The interface of the spatial resolution adjustment module is as follows: , in, The minimum repeated exploration distance corresponding to the neighborhood of the obstacle. The minimum repeated exploration distance corresponding to the neighborhood of the endpoint. The minimum repeated exploration distance for a normal area; the spatial resolution parameter is used as the spatial resolution parameter in the 3D path exploration module. The regionalized value sources; based on the weight parameters, exploration step size parameters, and spatial resolution parameters in the comprehensive cost function, a 3D path exploration module is used to route the network to be routed at the current time. For a flowchart of the 3D path exploration module for routing ordinary networks, please refer to [link to flowchart]. Figure 2 ,include: S1031. Constructing a three-dimensional discrete space based on the PCB knowledge base; Based on the constructed PCB knowledge base, a three-dimensional discrete space is built under a unified coordinate system. The planar wire-accessible area is defined by the outer border of the board diagram. The layer direction is determined by the layer set. Determined, each layer Corresponding unique height coordinates Meanwhile, based on the device layout The device pin geometry, completed wiring paths, and via structures are used to map all impassable or minimum spacing constraints into a unified set of obstacles. , serving as a spatial constraint in the path exploration process.
[0029] S1032. Construct the comprehensive cost function based on the weight parameters in the comprehensive cost function; The overall cost function is: , in, The value of the comprehensive cost function. For cumulative line length, To accumulate the number of vias, To accumulate the number of rule violations, , , These are the line length cost weight, the via quantity cost weight, and the design rule violation frequency cost weight, respectively.
[0030] S1033. Determine the start and end points of the network to be wired at the current moment based on the three-dimensional discrete space. For a cabling request where the network to be cabled is a normal network at the current moment, the 3D path exploration module uses the starting coordinates... coordinates of the endpoint As boundary conditions for path search, the design rule set is satisfied. Execution path exploration under constraints.
[0031] S1034. Using the starting point as the initial exploration node, the exploration step size parameter as the exploration step size, and the minimum comprehensive cost function as the objective, explore the routing path of the network to be routed at the current time to obtain the routing path that satisfies the design rules and minimizes the comprehensive cost. Among them, the spatial resolution of the candidate positions in the routing path is verified based on the spatial resolution parameter. Path exploration employs a gradual expansion strategy guided by comprehensive cost, which will As the initial exploration node, and to construct a set of candidate nodes. With the set of explored nodes Candidate node set Organized into a minimum-cost priority queue sorted by overall cost, each candidate node records its spatial coordinates, cumulative cost, and connections to the previous node; the set of explored nodes... Used to record the explored locations that have been visited at the current spatial resolution, during the exploration loop, from The candidate node with the minimum overall cost is selected as the current node. If the candidate node and the endpoint coordinates are... If the connection criteria are met, the exploration terminates and a complete routing path is generated by backtracking; otherwise, state extension is performed on the current node. State extension includes two types of operations: first, extending the wire along the eight directions (up, down, left, right, and four diagonal directions) within the current layer; second, introducing a via at the current node position to switch to other layers. In planar wire extension, the exploration step size is defined as... This represents the minimum advancing distance of a single traverse extension within the plane; the new candidate position is advanced from the current coordinates along the corresponding direction. During the via operation, the new candidate positions retain their planar coordinates, only changing the layer number to represent inter-layer connections. Before each state expansion, spatial resolution and obstacle constraint checks are sequentially performed on the generated candidate positions. The spatial resolution is determined by parameters... Definition: When the candidate position is... The Euclidean distance at any position in the middle is less than At this point, the location has been explored and pruned at the current resolution. Candidate locations verified by resolution are then further subjected to obstacle spacing detection. When a candidate conductor segment or via is combined with an obstacle... When the minimum distance is less than the minimum spacing allowed by the design rules, the corresponding DRC violation is recorded and the number of violations is accumulated. For all candidate positions that pass the verification, the cumulative path cost is calculated. The path cost is evaluated and calculated using a comprehensive cost function. After the cost calculation is completed, the candidate node is inserted. And record the spatial location of the candidate node to The exploration process continues until the destination is successfully reached or the candidate node set is empty. When multiple candidate paths reach the destination, the path with the lowest overall cost is selected as the final result through a minimum-cost priority queue selection mechanism. A complete wiring path is generated by backtracking the connection relationships between nodes and defined as... Completed It is written into the three-dimensional spatial state, and the spatial region it occupies is incorporated into the obstacle set. This is used to constrain subsequent routing requests, where the routing path is specified. Defined as an ordered list of conductor segments and vias arranged in spatial order: , When element When it is a conductor segment, its elements for: , in, and These are the start and end coordinates on the same layer. For layer numbering, For line width parameters; when element When it is a via, element for: , in, The coordinates of the hole center are, and The layer number is the one being connected. All aperture parameters meet the design rules. Constraints; its 3D path exploration module interface is: , in, and These are the three-dimensional coordinates of the start and end points of the wiring, respectively. The wiring path that meets design rules and minimizes overall cost.
[0032] When the network to be routed is a differential pair network at the current moment, please refer to the flowchart of the differential pair topology adaptive routing module for routing the differential pair network. Figure 3 ,include: S1035. Determine the start and end points of the network to be wired at the current moment; When the network to be cabled is a differential pair network at the current moment, determine the start and end points of the differential pair network, and for any differential connection in the differential pair network... ,in, For the starting side, For the end point side.
[0033] S1036. Starting from the positive pin on the starting side and the fan-out target corresponding to the positive pin on the starting side as the endpoint, the three-dimensional path exploration module is called to explore the routing path on the starting side. Starting from the negative pin on the starting side and the fan-out target corresponding to the negative pin on the starting side as the endpoint, the three-dimensional path exploration module is called to explore the routing path on the starting side to obtain the candidate routing results on the starting side. At the starting point, fan-out routing is performed. During the fan-out phase, starting from the differential pin pair, pairs of fan-out targets are generated to both sides, ensuring that the positive and negative pin routings are parallel in space. Since differential pairs only have two topological orders—either the positive pin is on one side of the negative pin or the opposite side—the differential pair topology adaptive routing module constructs candidate fan-out patterns corresponding to these two topological relationships during this phase. For each candidate fan-out pattern, the 3D path exploration module interface is called, using the positive pin and its corresponding fan-out target as the start and end points. , The 3D path exploration module interface is then invoked using the negative pin and its corresponding fan-out target as the start and end points. , This generates candidate routing results for the starting stage.
[0034] S1037. Starting from the positive pin on the endpoint side and the fan-out target corresponding to the positive pin on the endpoint side, the three-dimensional path exploration module is called to explore the routing path on the starting side. Starting from the negative pin on the endpoint side and the fan-out target corresponding to the negative pin on the endpoint side, the three-dimensional path exploration module is called to explore the routing path on the starting side to obtain the candidate routing results on the endpoint side. On the endpoint side, endpoint routing is performed in a manner corresponding to the fan-out phase of the starting point. The endpoint routing phase also generates candidate routing methods corresponding to the two topology sequences around the differential pin pairs. The 3D path exploration module interface is then called with the routing target and the endpoint pin as the start and end points, respectively. Specifically, the 3D path exploration module is called to explore the routing path on the starting point side, with the positive pin on the endpoint side as the start point and the fan-out target corresponding to the positive pin on the endpoint side as the end point. , And taking the negative pin on the endpoint side as the starting point and the fan-out target corresponding to the negative pin on the endpoint side as the endpoint, the 3D path exploration module is invoked to explore the wiring path on the starting point side: , Obtain the candidate routing results at the endpoint.
[0035] S1038. Taking the connection between the candidate routing results on the starting side and the candidate routing results on the ending side as the goal, and taking the center coordinates of the positive and negative fan-out points on the starting side as the starting point of the middle section and the center coordinates of the positive and negative fan-in points on the ending side as the ending point of the middle section, the three-dimensional path exploration module is called to explore the routing path of the middle section and obtain the center trajectory of the middle section. After completing the starting fan-out and ending branch, the mid-section parallel routing is executed. The mid-section parallel stage aims to connect the starting fan-out results with the ending branch results, treating the positive and negative paths as a set of parallel paths for synchronous extension. This stage first uses the center coordinates of the positive and negative fan-out points on the starting side as the mid-section starting point, denoted as: , And the center coordinates of the positive and negative fan-entry points on the end side are taken as the mid-segment end point, denoted as: , Then, the 3D path exploration module interface is invoked to generate the mid-section center trajectory: , in, This is a 3D path exploration module. The trajectory is the center of the middle section. The center coordinates of the positive and negative fan-out points on the starting side. The coordinates are the center coordinates of the positive and negative fan-in points on the endpoint side.
[0036] S1039. Based on the candidate routing results at the starting point, the candidate routing results at the ending point, and the center trajectory of the middle section, obtain the positive and negative paths of the network to be routed at the current moment. After verifying the positive and negative paths based on the design rules, obtain the final routing path of the network to be routed at the current moment. Based on the candidate routing results at the starting and ending points, and using the mid-section center trajectory as a reference, positive and negative paths are generated synchronously with a fixed differential spacing on both sides of the trajectory. Simultaneously, design rule checks are performed on both paths. The DRC violation status of the differential pair is determined by merging the violation results of the two paths. Throughout the differential pair routing process, the module continuously maintains the topology order of the two differential paths. The topology order is defined as the left-right relationship of the positive path relative to the negative path when viewed in the routing direction. During the starting fan-out stage, the module establishes a corresponding topology marker for each candidate fan-out method, and maintains this marker as routing progresses during the mid-section parallel and ending split stages. When routing continues in the opposite direction after a layer change, the module identifies the topology order switch and updates the topology marker accordingly. During the endpoint connection phase, the module compares the topology markers of the current routing path with the arrangement order of the endpoint differential pin pairs. Only if they match is the final endpoint connection allowed to be completed by calling the 3D path exploration module interface. If they do not match, the current path is abandoned and another candidate topology or intermediate path is selected to continue routing, thus adaptively ensuring the consistency of differential pairs in spatial topology and network connectivity. The interface of the differential pair topology adaptive routing module is as follows: , in, It is a difference connection. and These are the final routing paths for the positive and negative networks, respectively, which represent the final routing paths of the network to be routed at the current moment.
[0037] In some embodiments, in step S104, after performing a teardown operation on the routing result of the network to be routed at the current time, the set of unrouted networks is updated. The next network to be routed is selected from the updated set of unrouted networks. Based on the determined key parameters of the next network to be routed and the corresponding routing execution layer, the next network to be routed is routed to obtain the routing result of the next network to be routed. A teardown operation is performed on the routing result of the network to be routed at the current time. That is, networks with many conflicts, insufficient routing rate, severe local congestion, or a large number of design rule violations in the current network to be routed result are to be torn down. The networks after the teardown operation are merged into the set of unrouted networks. The set of unrouted networks is updated. The agent selects the next network to be routed from the updated set of unrouted networks. The network to be routed is selected based on the list of the unrouted network set, and this network can be a regular network or a differential pair network. Routing cost weight parameters, exploration step size parameters, and spatial resolution parameters are configured for the selected network. As the number of teardown and rerouting rounds increases, the cost weight corresponding to DRC violations is gradually increased. The network to be wired in the next moment is determined, denoted as . It also outputs the cabling cost weight parameters bound to the network. Explore step size parameters and spatial resolution parameters Subsequently, the agent writes parameters into the execution layer module by calling the module in the wiring execution layer, where the wiring cost control module... After completing the cost weight configuration, the exploration rate adjustment module was activated. After completing the exploration step size configuration, the spatial resolution adjustment module was activated. After configuring the spatial resolution, its parameters remain fixed during the next network routing process. Based on the selected network type, the interface of the corresponding module in the routing execution layer is called to complete the routing. When it is a regular network, the 3D path exploration module interface is called for each connection in its connection relationship. Generate routing paths and complete routing; when When it is a differential pair network, the differential pair topology adaptive cabling module interface is called to define its differential connection relationships. Complete the paired wiring. After completing the wiring, from Remove from the middle to obtain the routing result of the network to be routed at the next moment.
[0038] In some embodiments, in step S105, the routing result of the network to be routed at the next time step is used as the new routing result of the network to be routed at the current time step. Step S104 is repeated until the preset maximum number of teardown and rerouting iterations is reached or the routing success rate of the network to be routed at the current time step and the network to be routed at the next time step reaches a stable equilibrium. Then the iteration stops to obtain the PCB routing result. After obtaining the routing result of the network to be routed at the next time step, the routing result of the network to be routed at the current time step is updated. Teardown operations are performed on the updated routing result of the network to be routed at the current time step. The set of unrouted networks is then updated. The network to be routed is selected from the updated set of unrouted networks. The corresponding 3D path exploration module or differential pair topology adaptive routing module is selected according to the type of the network to be routed to perform teardown and rerouting. The teardown and rerouting iteration process is terminated when any of the following conditions are met: the current round reaches the maximum number of teardown and rerouting iterations. If the routing success rate no longer increases after two consecutive rounds of rerouting, the final output will be a PCB routing result that meets the design rules and achieves a stable balance between routing success rate, line length and via number.
[0039] The multi-round teardown and redeployment process consists of an outer-layer teardown and redeployment iterative process and an inner-layer routing process according to network sequence. The outer-layer process controls the global optimization process, while the inner-layer process handles specific network selection, parameter configuration, and routing execution. The outer-layer process is the teardown and redeployment iterative process, and its iteration rounds are denoted as follows: ,in The maximum number of iterations for teardown and re-layout is preset, at the [number]th iteration. At the start of each round of cable teardown and rerouting, a cable teardown operation is performed on the routing results of the previous round. Within each iteration of the cable teardown and rerouting process, the following steps are taken: The network executes wiring operations sequentially. Before each network wiring operation begins, it is controlled by the intelligent agent. Perform joint decision-making, simultaneously determining the current cabling network and its corresponding cabling parameter configuration; the model inputs used to drive the joint decision-making are concatenated in a fixed order as follows: , in, For contextual information, provided by the PCB knowledge base Textual content, current unwired network set List of network names and current suture removal and re-laying round number Composed of various elements; its task-level prompts It is set to: "In the current teardown and redeployment round, please select a network from the list of unrouted networks to be prioritized for routing. Prioritize differential pair networks and high-speed or critical networks. Please also provide the routing cost weight parameters, exploration step size parameters, and spatial resolution parameters for the selected network. As the teardown and redeployment rounds increase, the cost weight corresponding to DRC violations should be gradually increased."
[0040] In summary, the PCB intelligent design method provided by this invention constructs a PCB knowledge base based on acquired PCB board layout information, component placement, netlist, and design rules; it then constructs a routing execution layer scheduled by an intelligent agent based on the PCB knowledge base, wherein the routing execution layer includes a 3D path exploration module, a differential pair topology adaptive routing module, a routing cost control module, an exploration rate adjustment module, and a spatial resolution adjustment module; based on the PCB knowledge base, it determines a set of unrouted networks, selects the network to be routed at the current moment from the set of unrouted networks, and performs routing on the network to be routed at the current moment based on the determined key parameters of the network to be routed at the current moment and the corresponding routing execution layer, obtaining the routing result of the network to be routed at the current moment; and finally, it performs routing on the network to be routed at the current moment. After the routing result of the network to be routed is stripped, the set of unrouted networks is updated. The next network to be routed is selected from the updated set of unrouted networks. Based on the key parameters of the network to be routed in the next time step and the corresponding routing execution layer, the network to be routed in the next time step is routed to obtain the routing result of the network to be routed in the next time step. The routing result of the network to be routed in the next time step is used as the routing result of the network to be routed in the current time step. The new network to be routed in the current time step is stripped and rerouted until the preset maximum number of stripping and rerouting iterations is reached or the routing rate of the network to be routed in the current time step and the network to be routed in the next time step reaches a stable balance. The iteration stops when the PCB routing result is obtained, which improves the routing quality of the PCB.
[0041] To better implement the PCB intelligent design method in the embodiments of the present invention, based on the PCB intelligent design method, correspondingly, as follows: Figure 4 As shown, this embodiment of the invention also provides a PCB intelligent design system, the PCB intelligent design system 400 including: The knowledge base construction module 401 is used to build a PCB knowledge base based on the acquired PCB board layout information, component layout, netlist and design rules; The routing execution layer construction module 402 is used to construct a routing execution layer scheduled by an intelligent agent based on the PCB knowledge base; The routing module 403 is used to determine the set of unrouted networks based on the PCB knowledge base, select the network to be routed at the current time from the set of unrouted networks, and perform routing on the network to be routed at the current time based on the key parameters of the network to be routed at the current time and the corresponding routing execution layer, so as to obtain the routing result of the network to be routed at the current time. The cable teardown and rerouting module 404 is used to tear down the routing result of the network to be routed at the current moment, update the set of unrouted networks, select the network to be routed at the next moment from the updated set of unrouted networks, and route the network to be routed at the next moment based on the key parameters of the network to be routed at the next moment and the corresponding routing execution layer to obtain the routing result of the network to be routed at the next moment. The multiple disconnect and redistribution module 405 is used to take the routing result of the network to be routed at the next moment as the routing result of the network to be routed at the current moment, disconnect and redistribute the network to be routed at the current moment until the preset maximum number of disconnect and redistribution iterations is reached or the routing success rate of the network to be routed at the current moment and the network to be routed at the next moment reaches a stable balance, and then the iteration stops to obtain the PCB routing result.
[0042] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.
Claims
1. A PCB intelligent design method, characterized in that, include: Step 1: Build a PCB knowledge base based on the acquired PCB board layout information, component placement, netlist, and design rules; Step 2: Construct a routing execution layer scheduled by an intelligent agent based on the PCB knowledge base; Step 3: Determine the set of unrouted networks based on the PCB knowledge base, select the network to be routed at the current time from the set of unrouted networks, and route the network to be routed at the current time based on the key parameters of the network to be routed at the current time and the corresponding routing execution layer to obtain the routing result of the network to be routed at the current time. Step 4: After performing a disconnection operation on the routing result of the network to be routed at the current moment, update the set of unrouted networks. Select the network to be routed at the next moment from the updated set of unrouted networks. Based on the key parameters of the network to be routed at the next moment and the corresponding routing execution layer, route the network to be routed at the next moment to obtain the routing result of the network to be routed at the next moment. Step 5: Take the routing result of the network to be routed at the next moment as the new routing result of the network to be routed at the current moment, and repeat Step 4 until the preset maximum number of dismantling and rerouting iterations is reached or the routing rate of the network to be routed at the current moment and the network to be routed at the next moment reaches a stable balance, then stop the iteration to obtain the PCB routing result.
2. The PCB intelligent design method according to claim 1, characterized in that, The PCB knowledge base constructed based on the acquired PCB board layout information, component placement, netlist, and design rules includes: The acquired PCB board layout information, component layout, netlist, and design rules are structured and encoded to build a PCB knowledge base. The netlist includes network types and network connection relationships, and the network types include ordinary networks and differential pair networks.
3. The PCB intelligent design method according to claim 2, characterized in that, The routing execution layer includes a three-dimensional path exploration module, a differential pair topology adaptive routing module, a routing cost control module, an exploration rate adjustment module, and a spatial resolution adjustment module. The process of routing the network based on the determined key parameters of the network to be routed at the current moment and the corresponding routing execution layer includes: The type of the network to be wired at the current moment is determined. When the type of the network to be wired at the current moment is a normal network, the key parameters are input to the wiring cost control module, the exploration rate adjustment module, and the spatial resolution adjustment module. Then, the network to be wired at the current moment is wired based on the wiring cost control module, the exploration rate adjustment module, the spatial resolution adjustment module, and the three-dimensional path exploration module. When the type of the network to be routed at the current moment is a differential pair network, after inputting the key parameters into the routing cost control module, the exploration rate adjustment module, and the spatial resolution adjustment module, the network to be routed at the current moment is routed based on the routing cost control module, the exploration rate adjustment module, the spatial resolution adjustment module, and the differential pair topology adaptive routing module.
4. The PCB intelligent design method according to claim 3, characterized in that, The key parameters include wiring cost weighting parameters, exploration step size parameters, and spatial resolution parameters.
5. The PCB intelligent design method according to claim 3, characterized in that, The process of routing the network to be routed at the current moment based on the routing cost control module, exploration rate adjustment module, spatial resolution adjustment module, and three-dimensional path exploration module includes: The weight parameters in the comprehensive cost function are set based on the routing cost control module, wherein the weight parameters include the line length cost weight, the via number cost weight, and the design rule violation number cost weight. The exploration step size parameter of the three-dimensional path exploration module is set based on the exploration rate adjustment module; The spatial resolution parameters of the three-dimensional path exploration module are set based on the spatial resolution adjustment module. Based on the weight parameters, exploration step size parameters, and spatial resolution parameters in the comprehensive cost function, a three-dimensional path exploration module is used to route the network to be routed at the current moment.
6. The PCB intelligent design method according to claim 5, characterized in that, The process of routing the network to be routed at the current moment using a three-dimensional path exploration module, based on the weight parameters, exploration step size parameters, and spatial resolution parameters in the comprehensive cost function, includes: Constructing a three-dimensional discrete space based on a PCB knowledge base; Construct the comprehensive cost function based on the weight parameters in the comprehensive cost function; The starting and ending points of the network to be wired at the current moment are determined based on the three-dimensional discrete space. The routing path of the network to be routed at the current moment is explored with the starting point as the initial exploration node, the exploration step size parameter as the exploration step size, and the goal of minimizing the comprehensive cost function. The routing path that satisfies the design rules and minimizes the comprehensive cost is obtained. The spatial resolution of the candidate positions in the routing path is verified based on the spatial resolution parameter.
7. The PCB intelligent design method according to claim 6, characterized in that, The comprehensive cost function is: ， in, The value of the comprehensive cost function. For cumulative line length, To accumulate the number of vias, To accumulate the number of rule violations, , , These are the line length cost weight, the via quantity cost weight, and the design rule violation frequency cost weight, respectively.
8. The PCB intelligent design method according to claim 5, characterized in that, The process of routing the network to be routed at the current moment based on the routing cost control module, exploration rate adjustment module, spatial resolution adjustment module, and differential pair topology adaptive routing module includes: Determine the start and end points of the network to be wired at the current moment; Starting from the positive pin on the starting side and the fan-out target corresponding to the positive pin on the starting side as the endpoint, the three-dimensional path exploration module is called to explore the routing path on the starting side. Starting from the negative pin on the starting side and the fan-out target corresponding to the negative pin on the starting side as the endpoint, the three-dimensional path exploration module is called to explore the routing path on the starting side to obtain the candidate routing results on the starting side. Starting from the positive pin on the endpoint side and taking the fan-out target corresponding to the positive pin on the endpoint side as the endpoint, the three-dimensional path exploration module is called to explore the routing path on the starting side. Starting from the negative pin on the endpoint side and taking the fan-out target corresponding to the negative pin on the endpoint side as the endpoint, the three-dimensional path exploration module is called to explore the routing path on the starting side to obtain the candidate routing results on the endpoint side. With the connection between the candidate routing results on the starting side and the candidate routing results on the ending side as the goal, and the center coordinates of the positive and negative fan-out points on the starting side as the starting point of the middle section and the center coordinates of the positive and negative fan-in points on the ending side as the ending point of the middle section, the three-dimensional path exploration module is called to explore the routing path of the middle section and obtain the center trajectory of the middle section. Based on the candidate routing results at the starting point, the candidate routing results at the ending point, and the center trajectory of the middle section, the positive and negative paths of the network to be routed at the current moment are obtained. After verifying the positive and negative paths based on the design rules, the final routing path of the network to be routed at the current moment is obtained.
9. The PCB intelligent design method according to claim 8, characterized in that, The central trajectory of the middle section is: ， in, This is a 3D path exploration module. The trajectory is the center of the middle section. The center coordinates of the positive and negative fan-out points on the starting side. The coordinates are the center coordinates of the positive and negative fan-in points on the endpoint side.
10. A PCB intelligent design system, characterized in that, include: The knowledge base construction module is used to build a PCB knowledge base based on the acquired PCB board layout information, component layout, netlist, and design rules. The routing execution layer construction module is used to construct a routing execution layer scheduled by an intelligent agent based on the PCB knowledge base; The routing module is used to determine the set of unrouted networks based on the PCB knowledge base, select the network to be routed at the current time from the set of unrouted networks, and perform routing on the network to be routed at the current time based on the key parameters of the network to be routed at the current time and the corresponding routing execution layer, so as to obtain the routing result of the network to be routed at the current time. The cable teardown and rerouting module is used to tear down the routing results of the network to be routed at the current moment, update the set of unrouted networks, select the network to be routed at the next moment from the updated set of unrouted networks, and route the network to be routed at the next moment based on the key parameters of the network to be routed at the next moment and the corresponding routing execution layer to obtain the routing result of the network to be routed at the next moment. The multiple disconnect and redistribution module is used to take the routing result of the network to be routed at the next moment as the routing result of the network to be routed at the current moment, and disconnect and redistribute the network to be routed at the current moment until the preset maximum number of disconnect and redistribution iterations is reached or the routing success rate of the network to be routed at the current moment and the network to be routed at the next moment reaches a stable balance, and then the iteration stops to obtain the PCB routing result.