WO2019200626A1 - 人工智能体训练系统及无源电路优化设计系统及方法 - Google Patents

人工智能体训练系统及无源电路优化设计系统及方法 Download PDF

Info

Publication number
WO2019200626A1
WO2019200626A1 PCT/CN2018/085640 CN2018085640W WO2019200626A1 WO 2019200626 A1 WO2019200626 A1 WO 2019200626A1 CN 2018085640 W CN2018085640 W CN 2018085640W WO 2019200626 A1 WO2019200626 A1 WO 2019200626A1
Authority
WO
WIPO (PCT)
Prior art keywords
circuit
design
model
optimization
artificial intelligence
Prior art date
Application number
PCT/CN2018/085640
Other languages
English (en)
French (fr)
Inventor
刘洁
陈志熙
石佳
Original Assignee
石家庄创天电子科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 石家庄创天电子科技有限公司 filed Critical 石家庄创天电子科技有限公司
Publication of WO2019200626A1 publication Critical patent/WO2019200626A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G06F30/39—Circuit design at the physical level
    • G06F30/398—Design verification or optimisation, e.g. using design rule check [DRC], layout versus schematics [LVS] or finite element methods [FEM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G06F30/36—Circuit design at the analogue level
    • G06F30/367—Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods

Definitions

  • the embodiments of the present application relate to a design and development technology of a passive circuit, and in particular, to an artificial intelligence training system and a passive circuit optimization design system and a method thereof based on artificial intelligence.
  • passive circuits are widely used in the field of electronics, and passive circuits are circuits composed of basic components such as RCL (ie, resistor R, inductor L, and capacitor C).
  • RCL resistor resistor R, inductor L, and capacitor C.
  • the existing passive circuit design has the following problems:
  • the design of passive circuits requires experienced designers.
  • the entire design process consists of building simulation models, generating calculation results, debugging and optimizing circuits in computer simulation design tools.
  • the current design flow of passive circuits is fragmented and is achieved manually.
  • the main object of the present invention is to provide an artificial intelligence training system and a passive circuit optimization design system and method, which can enable an artificial intelligence body to have an automatic design function of a passive circuit to utilize depth. Reinforce learning to achieve automatic optimization design of passive circuits.
  • Another object of the present invention is to provide an artificial intelligence training and passive circuit optimization design system and method, which have the advantages of high design efficiency, high reusability, and low design cost.
  • the first embodiment of the present application provides an artificial intelligence training system, which is applied to an artificial intelligence body for training passive circuit optimization design performance of the artificial intelligence, and its characteristics.
  • the invention comprises: a receiving module, configured to receive a circuit model; a setting module, configured to provide a setting circuit model design index; and a training optimization design module, configured to perform circuit optimization on the circuit model according to the circuit model design index Designing processing and generating a circuit optimization model; and calculating a analysis module for calculating whether the simulation operation result of the circuit optimization model conforms to the circuit model design index and outputting the analysis result for the training optimization design module
  • the analysis results train to update the circuit optimization design process that it performs.
  • the circuit model design indicator includes a media material type parameter, a circuit component type and a size parameter, a conductor type and a size parameter, an input interface parameter, and an output interface parameter.
  • the artificial intelligence training system further includes a database that stores initialization parameters of circuit elements constituting the circuit model.
  • the training optimization design module performs at least one time for the circuit model according to an initialization parameter of the circuit component in the database and the circuit model design indicator.
  • the circuit is optimized for design processing.
  • the calculating and analyzing module further includes: simulating running the circuit optimization model generated by the training optimization design module to perform the circuit optimization design process each time to generate at least Performing the simulation operation result once, and determining whether the simulation operation result of the circuit optimization model currently generated by the training optimization design module is more than the simulation operation result of the circuit optimization model generated by the previous time. Close to the circuit model design index, if the analysis result is yes, the forward feedback signal is output, and vice versa, the negative feedback signal is output.
  • the training optimization design module further includes: recording each of the circuit optimization design processing according to the forward feedback signal or the negative feedback signal output by the calculation analysis module.
  • the training optimization design module further includes: the forward feedback signal or the negative feedback signal output according to the calculation analysis module, so that the calculation analysis module The probability of outputting the forward feedback signal is continuously increased to an optimized design basis, and the operation of the circuit optimization adjustment process performed for the circuit model next time is determined.
  • the second embodiment of the present application further provides a passive circuit optimization design system, which is implemented based on artificial intelligence, and includes: a receiving unit, configured to receive a circuit model to be optimized; and a setting unit, configured to provide Setting a circuit model design index; and an artificial intelligence body trained by the artificial intelligence body training system according to the first embodiment; wherein the artificial intelligence body is based on the set circuit model design index
  • the received circuit model to be optimized is subjected to at least one optimization design process to generate a circuit optimization model that conforms to the circuit model design specification.
  • a second embodiment of the present application further provides a passive circuit optimization design method, wherein an artificial intelligence body performs an optimized design of a passive circuit, wherein the method includes: inputting, to the artificial intelligence body, a to-be-optimized a circuit model, and an input circuit model design index; wherein the artificial intelligence body performs circuit optimization design processing on the circuit model according to the circuit model design index to generate a circuit optimization model; and simulates running the artificial intelligence body to generate The circuit optimization model to generate a simulation operation result, analyze whether the simulation operation result conforms to the circuit model design indicator, and output an analysis result; and cause the artificial intelligence body to receive the analysis, and when the analysis If the result of the simulation operation does not meet the circuit model design index, the circuit optimization design process is repeatedly performed to generate a new circuit optimization model until the analysis result is consistent with the simulation operation result. When the circuit model design index is completed, the circuit optimization design processing is ended.
  • the circuit model design indicator includes a media material type parameter, a circuit component type and a size parameter, a conductor type and a size parameter, an input interface parameter, and an output interface parameter.
  • the artificial intelligence further stores initialization parameters of circuit elements constituting the circuit model, and the artificial intelligence is based on the stored circuit components.
  • the initialization parameter and the circuit model design indicator perform at least one circuit optimization design process for the circuit model.
  • the method further includes: simulating running the currently generated circuit optimization model to generate the simulation running result; and causing the artificial intelligence to determine the currently generated Whether the simulated operation result of the circuit optimization model is closer to the circuit model design index than the simulation operation result of the circuit optimization model generated in the previous time, and if the analysis result is yes, the output is positive a feedback signal, and vice versa, outputting a negative feedback signal; and causing the artificial intelligence to rely on the output of the forward feedback signal or the negative feedback signal to increase the probability of the outputted forward feedback signal
  • the next operation plan for the circuit optimization design process to be executed for the circuit model is judged, and the next circuit optimization design process is executed for the circuit model to generate a new circuit optimization model.
  • the method further includes: causing the artificial intelligence to record an influence of adjustment of each of the circuit components in each circuit optimization design process on the circuit model design index The historical data and the real-time data of the circuit optimization model affecting the design of the circuit model during the simulation run.
  • the artificial intelligence training system trains artificial intelligence by collecting the design experience of the passive circuit designer, the historical data of the circuit optimization process, and the real-time data of the passive circuit during the simulation operation.
  • the body has the function of automatically optimizing the passive circuit design.
  • the present application also provides a passive circuit optimization design system and method, which can realize automatic optimization design of the passive circuit based on artificial intelligence, and has the advantages of high design efficiency, high reusability and low design cost.
  • FIG. 1 is a schematic diagram of a basic architecture of a human body training system according to an embodiment of the present application
  • FIG. 2 is a schematic view showing different embodiments of a human body training system of the present application.
  • FIG. 3 is a schematic diagram of a basic architecture of a passive circuit optimization design system according to another embodiment of the present application.
  • FIG. 4 is a schematic diagram showing the basic flow of a passive circuit optimization design method according to another embodiment of the present application.
  • FIG. 5 is a flow chart showing a specific embodiment of the passive circuit optimization design method of FIG. 4.
  • FIG. 1 is a schematic diagram of a basic architecture of a human body training system according to an embodiment of the present application.
  • the artificial intelligence training system 10 of the present application is applied to the artificial intelligence body 1 and is mainly used for training the passive circuit optimization design function of the artificial intelligence body 1, which mainly comprises a receiving module 11, a setting module 12, and training.
  • the optimization design module 13 and the calculation analysis module 14 are provided.
  • the receiving module 11 is configured to receive a circuit model.
  • the received circuit model refers to a circuit model that is not designed to meet the circuit model design specifications and needs to be optimized.
  • the setting module 12 is used to provide circuit designers with the desired circuit model design specifications.
  • the circuit model design indicators input by the circuit designer include a dielectric material type parameter, a circuit component type and a size parameter, a conductor type and a size parameter, an input interface parameter, and an output interface parameter.
  • the training optimization design module 13 is configured to perform circuit optimization design processing on the input circuit model according to the circuit model design index to generate a corresponding circuit optimization model.
  • the artificial intelligence training system 10 further includes a data library 15 storing initialization parameters of various types of circuit components for constructing a circuit model
  • the training optimization design module 13 can be At least one circuit optimization design process is performed for the circuit model according to the initialization parameters of the various types of circuit elements stored in the database 15, and the input circuit model design specifications.
  • the calculation analysis module 14 is configured to analyze whether the simulation operation result of the circuit optimization model generated by the training optimization design module 13 meets the circuit model design index received by the receiving module 11, and outputs the analysis result according to the training optimization design module 13 The analysis results train to update the circuit optimization design process performed by it, and perform the next circuit optimization design process accordingly.
  • the calculation and analysis module 14 is configured to simulate the circuit optimization model generated by the operation training optimization design module 13 to generate a corresponding simulation operation result, and analyze whether the simulation operation result satisfies the circuit received by the receiving module 11.
  • the model design index when the analysis result is that the simulation operation result does not satisfy the circuit model design index, the training optimization design module 13 repeats the operation of the circuit optimization design process until the calculation analysis module 14 determines the simulation operation result of the generated circuit optimization model. Can meet the circuit model design indicators.
  • the calculation analysis module 14 is configured to simulate the circuit optimization model generated by the running training optimization design module 13 each time performing the circuit optimization design process to generate at least one simulation operation result, and compare the current generated by the training optimization design module 13. Whether the simulation operation result of the circuit optimization model is closer to the circuit model design index set by the circuit designer than the simulation operation result of the circuit optimization model generated in the previous time, and if the analysis result is yes, it represents the training optimization design module 13
  • the operation scheme of the currently implemented circuit optimization design processing is effective, and a forward feedback signal (for example, a reward signal) is output, and if the analysis result is no, it represents an operation of the circuit optimization design processing currently performed by the training optimization design module 13.
  • a negative feedback signal (such as a penalty signal) is output.
  • the standard for outputting the reward and punishment signal by the calculation and analysis module 14 is not limited to the above technical solution.
  • the application may also use the experience of the circuit designer as a standard for judging the size of the reward and punishment signal during learning. 1. For example, in the process of circuit optimization debugging, when based on the experience of the circuit designer, it is judged that the parameters of a certain circuit component in the circuit model should be down-regulated, and the optimization design process performed by the training optimization design module 13 is also the circuit. When the parameters of the component are down-regulated, the calculation analysis module 14 outputs a bonus signal, otherwise the penalty signal is output.
  • the training optimization design module 13 can record the historical data and the circuit of the influence of the adjustment of each circuit component in each circuit optimization design process on the circuit model design index according to the forward feedback signal or the negative feedback signal output by the calculation analysis module 14. Real-time data that optimizes the impact of various circuit component adjustments on circuit model design specifications during simulation runs. At the same time, the training optimization design module 13 can also calculate the forward feedback signal or the negative feedback signal output by the analysis module 14 to increase the probability that the calculation analysis module 14 outputs the forward feedback signal to an optimized design basis. Determine the operation of the next circuit optimization adjustment process performed for the circuit model.
  • the training optimization design module 13 analyzes the variation law between the simulation operation result (experimental data) and the circuit model design index (theoretical data) according to the output forward feedback signal or the negative feedback signal, thereby modifying the relevant circuit components.
  • the design parameters are executed to perform circuit optimization design processing operations.
  • the simulation operation results simulation results
  • the calculation analysis module 14 may be based on the results of the simulation results.
  • the training optimization design module 13 will select the next circuit optimization design processing operation to be executed according to the feedback reward and punishment signal and the operating environment, and select the next time
  • the principle of the circuit optimization processing operation to be performed is to increase the probability of the reward signal given by the computational analysis module 14. If a circuit optimization design processing operation performed by the training optimization design module 13 results in a positive reward, then the trend of performing the circuit optimization design processing operation later will be strengthened; otherwise, the trend of performing the circuit optimization design processing operation later will be enhanced. Will be weakened.
  • the reinforcement learning is performed for the optimized design function of the training optimization design module 13.
  • the present application enables the training optimization design module 13 to acquire the mapping features of the passive circuit design based on the deep reinforcement learning through the repeated interaction between the training optimization design module 13 and the computational analysis module 14, and utilizes the reinforcement learning technique to learn
  • the optimal decision-making for circuit design-related decision-making is to use migration learning to achieve the efficiency of building different types of passive circuit design models.
  • FIG. 3 is a schematic diagram of a basic architecture of a passive circuit optimization design system according to another embodiment of the present application.
  • the passive circuit optimization design system 20 provided by the present application mainly includes a receiving unit 21, a setting unit 22, and an artificial intelligence body 23.
  • the receiving unit 21 is configured to receive a circuit model to be optimized
  • the setting unit 22 is configured to provide a circuit designer to set a circuit model design index
  • the artificial intelligence body 23 is an artificial intelligence body training system according to FIG. 1 or FIG.
  • the training of 10 is generated, and has the function of passive circuit optimization design, which can perform at least one optimization design on the circuit model to be optimized received by the receiving unit 21 according to the circuit model design index set by the setting unit 22. Processing to automatically generate a circuit optimization model that simulates the running results in accordance with the circuit model design specifications.
  • FIG. 4 is a schematic diagram showing the basic flow of a passive circuit optimization design method according to another embodiment of the present application.
  • the passive circuit optimization design method of the present application implements the optimization design of the passive circuit by the artificial intelligence body, and mainly includes the following processing steps:
  • Step S41 inputting the circuit model to be optimized to the artificial intelligence body, and inputting the circuit model design index, and then performing step S42.
  • the input circuit model design specifications include a dielectric material type parameter, a circuit component type and a size parameter, a conductor type and a size parameter, an input interface parameter, and an output interface parameter.
  • Step S42 the worker agent performs circuit optimization design processing for the circuit model according to the circuit model design index to generate a circuit optimization model, and then performs step S43.
  • the artificial intelligence body further stores initialization parameters of the circuit components for constructing the circuit model, and the artificial intelligence body is executed according to the initialization parameters of the circuit components stored therein and the circuit model design indicators for the circuit model. At least one circuit optimization design process.
  • Step S43 simulating a circuit optimization model generated by running the artificial intelligence body to generate a simulation operation result, and analyzing whether the simulation operation result of the circuit optimization model conforms to the circuit model design index, and outputting the analysis result according to the result (please refer to FIG. 5 for details) Said), then step S44 is performed.
  • Step S44 the artificial intelligence body is configured to receive the analysis result, and when the analysis result is that the simulation operation result does not conform to the circuit model design index, the foregoing circuit optimization design process is repeatedly performed to generate a new circuit optimization model. Until the analysis result is that the simulation operation result meets the circuit model design index, the execution circuit optimization design processing is ended.
  • FIG. 5 is a flow chart showing a specific embodiment of the passive circuit optimization design method of FIG. 4 .
  • step S51 is first executed to simulate the circuit optimization model generated by running the artificial intelligence to generate a simulation operation result, and then step S52 is performed.
  • Step S52 determining whether the simulated operation result of the generated circuit optimization model meets the circuit model design index, and when the judgment result is consistent, the artificial intelligence body outputs the generated circuit optimization model, and ends the circuit model for the to-be-optimized Optimized design processing.
  • step S53 is performed.
  • Step S53 the worker-agent determines whether the simulation operation result of the currently generated circuit optimization model is closer to the circuit model design index than the simulation operation result of the circuit optimization model generated in the previous time, and if the determination result is no, the step is performed. In S641, if the determination result is YES, then step S642 is performed.
  • Step S541 when the determination result is no, the circuit optimization model generated by the artificial intelligence agent is unsuccessful, or the current circuit optimization design processing is not satisfactory, and a negative feedback signal (such as a penalty signal) is output, and then the steps are performed.
  • a negative feedback signal such as a penalty signal
  • Step S542 when the determination result is yes, the circuit optimization model currently generated by the artificial intelligence body achieves the optimization effect, that is, the current circuit optimization design processing performed is successful, and the forward feedback signal (for example, the reward signal is output). Then, step S55 is performed.
  • Step S55 the worker agent according to the received forward feedback signal or the negative feedback signal, so that the probability of the output forward feedback signal is continuously increased to an optimized design reference, and the next circuit to be executed for the circuit model is determined.
  • the operation scheme of the design process is optimized, and the next circuit optimization design process is performed for the circuit model to generate a new circuit optimization model, and the process returns to step S51.
  • step S55 further includes historical data of the influence of the adjustment of each circuit component on the circuit model design index during each circuit optimization design process, and the circuit of the circuit optimization model during the simulation operation.
  • the artificial intelligence analyzes the change between the simulation operation result (experimental data) and the circuit model design index (theoretical data) according to the output forward feedback signal or the negative feedback signal, thereby modifying the design parameters of the relevant circuit component, That is, the principle of performing the circuit optimization design processing operation and selecting the next circuit optimization processing operation to be performed is to increase the probability of the given reward signal. If a certain circuit optimization design processing operation performed by the artificial agent leads to a positive reward, then the trend of executing the circuit optimization design processing operation later will be strengthened; otherwise, the trend of executing the circuit optimization design processing operation later will be weakened. .
  • the artificial intelligence training system digitizes the design experience of the passive circuit designer, and the historical data of the components of the circuit optimization process are adjusted to affect the design index requirements and the passive circuit is online. During the simulation, each component adjusts the real-time data and other information that affect the design index requirements to train the passive circuit optimization design function of artificial intelligence.
  • the present application also discloses a passive circuit optimization design system and method, adopting artificial intelligence and enhanced learning technology to realize automatic optimization design of passive circuits, and has the advantages of high design efficiency, high reusability and low design cost. It also reduces the risk and uncertainty of the design and implementation process and makes the circuit easy to customize.
  • the device embodiments described above are merely illustrative, wherein the modules described as separate components may or may not be physically separate, and the components displayed as modules may or may not be physical modules, ie may be located A place, or it can be distributed to multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the embodiment. Those of ordinary skill in the art can understand and implement without deliberate labor.
  • a machine-readable medium includes read only memory (ROM), random access memory (RAM), magnetic disk storage media, optical storage media, flash storage media, electrical, optical, acoustic, or other forms of propagation signals (eg, carrier waves) , an infrared signal, a digital signal, etc., etc., the computer software product comprising instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to perform the various embodiments or portions of the embodiments described Methods.
  • ROM read only memory
  • RAM random access memory
  • magnetic disk storage media e.g., magnetic disks, magnetic disk storage media, optical storage media, flash storage media, electrical, optical, acoustic, or other forms of propagation signals (eg, carrier waves) , an infrared signal, a digital signal, etc., etc.
  • the computer software product comprising instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to perform the various embodiments or portions of the embodiment
  • embodiments of the embodiments of the invention may be provided as a method, apparatus (device), or computer program product.
  • embodiments of the invention may be in the form of an entirely hardware embodiment, a full software embodiment, or a combination of software and hardware.
  • embodiments of the invention may take the form of a computer program product embodied on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) including computer usable program code.
  • Embodiments of the invention are described with reference to flowchart illustrations and/or block diagrams of methods, apparatus, and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or FIG.
  • These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing device to produce a machine for the execution of instructions for execution by a processor of a computer or other programmable data processing device.
  • the computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture comprising the instruction device.
  • the apparatus implements the functions specified in one or more blocks of a flow or a flow and/or block diagram of the flowchart.
  • These computer program instructions can also be loaded onto a computer or other programmable data processing device such that a series of operational steps are performed on a computer or other programmable device to produce computer-implemented processing for execution on a computer or other programmable device.
  • the instructions provide steps for implementing the functions specified in one or more of the flow or in a block or blocks of a flow diagram.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Design And Manufacture Of Integrated Circuits (AREA)

Abstract

本申请实施例提供了一种人工智能体训练系统及无源电路优化设计系统及方法，通过输入待优化的电路模型及所需的电路模型设计指标，使得人工智能体根据所述电路模型设计指标而针对所述电路模型执行电路优化设计处理，而自动生成符合电路模型设计指标的电路优化模型，可实现不同类型的无源电路设计经验的共享，提高电路设计的可复用性能，并具有设计效率高，设计成本低廉的优点。

Description

人工智能体训练系统及无源电路优化设计系统及方法
本申请要求在2018年4月18日提交中国专利局、申请号为201810350393.7、发明名称为“基人工智能体训练系统及无源电路优化设计系统及方法”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请实施例涉及无源电路的设计开发技术，尤其涉及一种人工智能体训练系统以及基于人工智能而实现的无源电路优化设计系统及其方法。
背景技术
当前，无源电路在电子领域中的使用非常广泛，所谓无源电路，是指仅由RCL(即电阻R，电感L，和电容C)等基础元件组成的电路。现有的无源电路设计存在以下问题：
首先，无源电路的设计需要经验丰富的设计师，整个设计过程主要包括在计算机仿真设计工具中建立仿真模型、生成计算结果、调试和优化电路。然而，当前的无源电路的设计流程是零散的，并且均是通过人工手动来实现。
其次，由于无源电路的种类繁多，故具有较高的设计门槛，对于设计经验较少或者几乎没有设计经验的设计师而言，容易发生因受限于设计经验不足而很难展开电路设计工作的困扰。
再者，无源电路在设计、优化的过程中会生成高度相关的数据，但是目前没有一个系统和方法能够自动学习电路设计中所产生的数据的规律，导致当前电路设计的工作效率低、开发周期长。
综上所述，由于现有技术中不同类型的无源电路具有共有的设计、调试特性，但是没有一个系统和方法能指明如何进行不同类型电路的迁移调试，导致现有电路设计方法的可复用程度低，而如何改善上述问题，即为本申请的待解决的技术课题。
发明内容
为了解决现有技术存在的缺陷，本发明的主要目的在于提供一种人工智能体训练系统及无源电路优化设计系统及方法，可使人工智能体具有无源电路的自动化设计功能，以利用深度强化学习实现无源电路的自动优化设计。
本发明的另一目的在于提供一种人工智能体训练及无源电路优化设计系统及方法，具有设计效率高，可复用程度高以及设计成本低廉的优点。
为达上述目的及其他相关目的，本申请的第一实施方案提供了一种人工智能体训练系统，应用于人工智能体，用于训练所述人工智能体的无源电路优化设计性能，其特征在于，包括：接收模块，用于接收电路模型；设定模块，用于提供设定电路模型设计指标；训练优化设计模块，用于依据所述电路模型设计指标而针对所述电路模型执行电路优化设计处理，并生成电路优化模型；以及计算分析模块，用于计算分析所述电路优化模型的模拟运行结果是否符合所述电路模型设计指标据以输出分析结果，以供所述训练优化设计模块依据所述分析结果训练更新其所执行的所述电路优化设计处理。
可选地，在本申请的任一实施例中，所述电路模型设计指标包括介质材料类型参数、电路元件类型及尺寸参数、导体类型及尺寸参数、输入接口参数及输出接口参数。
可选地，在本申请的任一实施例中，所述人工智能体训练系统还包括资料库，其储存用于构成所述电路模型的电路元件的初始化参数。
可选地，在本申请的任一实施例中，所述训练优化设计模块依据所述资料库中的所述电路元件的初始化参数以及所述电路模型设计指标，针对所述电路模型执行至少一次的电路优化设计处理。
可选地，在本申请的任一实施例中，所述计算分析模块还包括模拟运行所述训练优化设计模块每一次执行所述电路优化设计处理所生成的所述电路优化模型，以生成至少一次所述模拟运行结果，并判断所述训练优化设计模块当前所生成的所述电路优化模型的所述模拟运行结果是否较前一次所生成的所述电路优化模型的所述模拟运行结果更为接近所述电路模型设计指标，若分析结果为是，则输出正向反馈信号，反之，则输出负向反馈信号。
可选地，在本申请的任一实施例中，所述训练优化设计模块还包括依据所述计算分析模块所输出的所述正向反馈信号或负向反馈信号，记录每一次电路优化设计处理中的各所述电路元件的调整对所述电路模型设计指标的影响的历史数据以及所述电路优化模型在模拟运行时各电路元件的调整对所述电路模型 设计指标的影响的实时数据。
可选地，在本申请的任一实施例中，所述训练优化设计模块还包括根据所述计算分析模块所输出的所述正向反馈信号或负向反馈信号，以使所述计算分析模块输出所述正向反馈信号的概率不断增大为优化设计基准，确定下一次针对所述电路模型所执行的电路优化调整处理的操作。
本申请的第二实施方案还提供一种无源电路优化设计系统，为基于人工智能而实现，其特征在于，包括：接收单元，用于接收待优化的电路模型；设定单元，用于提供设定电路模型设计指标；以及由第一实施方案所述的人工智能体训练系统所训练生成的人工智能体；其中，所述人工智能体依据所设定的所述电路模型设计指标，对所接收的所述待优化的电路模型进行至少一次的优化设计处理，以生成符合所述电路模型设计指标的电路优化模型。
本申请的第二实施方案还提供一种无源电路优化设计方法，借由人工智能体执行无源电路的优化设计，其特征在于，所述方法包括：向所述人工智能体输入待优化的电路模型，以及输入电路模型设计指标；令所述人工智能体依据所述电路模型设计指标，针对所述电路模型执行电路优化设计处理，以生成电路优化模型；模拟运行所述人工智能体所生成的所述电路优化模型以生成模拟运行结果，分析所述模拟运行结果是否符合所述电路模型设计指标，据以输出分析结果；以及令所述人工智能体接收所述分析，并当所述分析结果为所述模拟运行结果不符合所述电路模型设计指标时，则重复执行所述电路优化设计处理，以生成新的所述电路优化模型，直至当所述分析结果为所述模拟运行结果符合所述电路模型设计指标时，则结束执行所述电路优化设计处理。
可选地，在本申请的任一实施例中，所述电路模型设计指标包括介质材料类型参数、电路元件类型及尺寸参数、导体类型及尺寸参数、输入接口参数及输出接口参数。
可选地，在本申请的任一实施例中，所述人工智能体还储存有构成所述电路模型的电路元件的初始化参数，且所述人工智能体是依据所储存的所述电路元件的初始化参数以及所述电路模型设计指标，针对所述电路模型执行至少一次的电路优化设计处理。
可选地，在本申请的任一实施例中，所述方法还包括：模拟运行当前所生成的所述电路优化模型，以生成所述模拟运行结果；令所述人工智能体判断当前所生成的所述电路优化模型的所述模拟运行结果是否较前一次所生成的所述电路优化模型的所述模拟运行结果更为接近所述电路模型设计指标，若分析结 果为是，则输出正向反馈信号，反之，则输出负向反馈信号；以及令所述人工智能体依据所输出的所述正向反馈信号或负向反馈信号，以使所输出的所述正向反馈信号的概率不断增大为优化设计基准，判断下一次针对所述电路模型待执行的电路优化设计处理的操作方案，据以针对所述电路模型执行下一次的电路优化设计处理，以生成新的电路优化模型。
可选地，在本申请的任一实施例中，所述方法还包括令所述人工智能体记录每一次电路优化设计处理过程中各所述电路元件的调整对所述电路模型设计指标的影响的历史数据以及所述电路优化模型在模拟运行时各电路元件的调整对所述电路模型设计指标的影响的实时数据。
由上可知，本申请所提供的人工智能体训练系统通过采集无源电路设计人员的设计经验，电路优化过程的历史数据及无源电路在模拟运行过程中的实时数据等信息，以训练人工智能体具有自动优化设计无源电路的功能。
再者，本申请还提供无源电路优化设计系统及方法，可基于人工智能而实现无源电路的自动优化设计，并具有设计效率高，可复用程度高以及设计成本低廉的优点。
附图说明
后文将参照附图以示例性而非限制性的方式详细描述本申请实施例的一些具体实施例。附图中相同的附图标记标示了相同或类似的部件或部分。本领域技术人员应该理解，这些附图未必是按比例绘制的。附图中：
图1为根据本申请的一实施例所示的人体智能体训练系统的基本架构示意图；
图2为显示本申请的人体智能体训练系统的不同实施例示意图；
图3为根据本申请的另一实施例所示的无源电路优化设计系统的基本架构示意图；
图4为根据本申请的另一实施例所示的无源电路优化设计方法的基本流程示意图；以及
图5为显示图4的无源电路优化设计方法的具体实施例流程图。
具体实施方式
实施本发明实施例的任一技术方案必不一定需要同时达到以上的所有优点。
为了使本领域的人员更好地理解本发明实施例中的技术方案，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅是本发明实施例一部分实施例，而不是全部的实施例。基于本发明实施例中的实施例，本领域普通技术人员所获得的所有其他实施例，都应当属于本发明实施例保护的范围。
下面结合本发明实施例附图进一步说明本发明实施例具体实现。
图1为根据本申请的一实施例所示的人体智能体训练系统的基本架构示意图。如图所示，本申请的人工智能体训练系统10应用于人工智能体1，主要用于训练人工智能体1的无源电路优化设计功能，其主要包括接收模块11、设定模块12、训练优化设计模块13、以及计算分析模块14。
接收模块11用于接收电路模型。于本申请的实施例中，所接收的电路模型是指不符合电路模型设计指标的有待进行优化设计的电路模型。
设定模块12用于提供电路设计人员设定所期望的电路模型设计指标。于本申请的实施例中，电路设计人员所输入的电路模型设计指标包括介质材料类型参数、电路元件类型及尺寸参数、导体类型及尺寸参数、输入接口参数及输出接口参数。
训练优化设计模块13用于依据所述电路模型设计指标而针对所输入的所述电路模型执行电路优化设计处理，以生成相应的电路优化模型。
请配合参阅图2，于本申请的实施例中，人工智能体训练系统10中还包括有资料库15，其储存有用于构成电路模型的各类型电路元件的初始化参数，训练优化设计模块13可依据资料库15中所储存的各类型电路元件的初始化参数，以及所输入的电路模型设计指标，针对所述电路模型执行至少一次的电路优化设计处理。
计算分析模块14用于分析训练优化设计模块13所生成的电路优化模型的模拟运行结果是否符合接收模块11所接收的电路模型设计指标，并据以输出分析结果，以供训练优化设计模块13依据分析结果训练更新其所执行的电路优化设计处理，并据以执行下一次的电路优化设计处理。
于本实施例中，计算分析模块14用于模拟运行训练优化设计模块13所生成的电路优化模型，以生成相应的模拟运行结果，并分析所述模拟运行结果是否满足接收模块11所接收的电路模型设计指标，当分析结果为模拟运行结果不满足电路模型设计指标时，训练优化设计模块13将重复执行电路优化设计处理 的操作，直至计算分析模块14判断所生成的电路优化模型的模拟运行结果可满足电路模型设计指标。
具体而言，计算分析模块14用于模拟运行训练优化设计模块13每一次执行电路优化设计处理所生成的电路优化模型，以生成至少一次模拟运行结果，并比较训练优化设计模块13当前所生成的电路优化模型的模拟运行结果是否较前一次所生成的电路优化模型的模拟运行结果进更为接近电路设计人员所设定的电路模型设计指标，若分析结果为是，则代表训练优化设计模块13当前所执行的电路优化设计处理的操作方案是有效的，则输出正向反馈信号(例如奖励信号)，若分析结果为否，则代表训练优化设计模块13当前所执行的电路优化设计处理的操作方案存在问题，则输出负向反馈信号(例如惩罚信号)。需说明的是，计算分析模块14输出奖惩信号的标准并不以上述技术方案为限，于其他实施例中，本申请亦可将电路设计人员的经验作为强化学习时判断奖惩信号大小的标准之一，例如：在电路优化调试的过程中，当基于电路设计人员的经验而判断电路模型中的某一电路元件的参数应该下调，且训练优化设计模块13所执行的优化设计处理也是将该电路元件的参数进行下调时，计算分析模块14就会输出奖励信号，否则输出惩罚信号。
训练优化设计模块13可依据计算分析模块14所输出的正向反馈信号或负向反馈信号，记录每一次电路优化设计处理中的各电路元件的调整对电路模型设计指标的影响的历史数据以及电路优化模型在模拟运行时各电路元件的调整对电路模型设计指标的影响的实时数据。同时，训练优化设计模块13还可根据计算分析模块14所输出的正向反馈信号或负向反馈信号，以使计算分析模块14输出所述正向反馈信号的概率不断增大为优化设计基准，确定其下一次针对电路模型所执行的电路优化调整处理的操作。
例如，训练优化设计模块13根据所输出的正向反馈信号或负向反馈信号，分析模拟运行结果(实验数据)与电路模型设计指标(理论数据)之间的变化规律，据以修改相关电路元件的设计参数，以执行电路优化设计处理操作，电路模型在接受到该电路优化设计处理后，模拟运行结果(仿真结果)亦将发生相应的变化，计算分析模块14可基于仿真结果的优劣结果而产生一个强化信号(奖励或者惩罚)反馈给训练优化设计模块13，训练优化设计模块13将根据所反馈的奖惩信号和运行环境再选择下一次待执行的电路优化设计处理操作，而选择下一次待执行的电路优化处理操作的原则是使计算分析模块14所给出的奖励信号的概率不断增大。如果训练优化设计模块13执行的某个电路优化设计处 理操作导致正向的奖赏，那么其以后执行这个电路优化设计处理操作的趋势便会加强；反之，则以后执行这个电路优化设计处理操作的趋势将减弱。借由上述技术手段，以针对训练优化设计模块13的优化设计功能进行强化学习。因此，本申请通过训练优化设计模块13与计算分析模块14之间的反复交互作用，使训练优化设计模块13可基于深度强化学习来获取无源电路设计的映射特征，并利用强化学习技术，学习电路设计相关决策的最优决策，以采用迁移学习来达到加快搭建不同类型的无源电路设计模型的效率。
图3为根据本申请的另一实施例所示的无源电路优化设计系统的基本架构示意图。如图所示，本申请所提供的无源电路优化设计系统20主要包括接收单元21、设定单元22以及人工智能体23。
接收单元21用于接收待优化的电路模型，设定单元22用于提供电路设计人员设定电路模型设计指标，而人工智能体23为由根据图1或图2所示的人工智能体训练系统10的训练而生成，具有无源电路优化设计的功能，其可依据设定单元22所设定的电路模型设计指标，对接收单元21所接收的述待优化的电路模型进行至少一次的优化设计处理，以自动生成模拟运行结果符合电路模型设计指标的电路优化模型。
图4为根据本申请的另一实施例所示的无源电路优化设计方法的基本流程示意图。如图所示，本申请的无源电路优化设计方法，借由人工智能体执行无源电路的优化设计，其主要包括以下处理步骤：
步骤S41，向人工智能体输入待优化的电路模型，以及输入电路模型设计指标，接着执行步骤S42。于本申请的实施例中，所输入的电路模型设计指标包括介质材料类型参数、电路元件类型及尺寸参数、导体类型及尺寸参数、输入接口参数及输出接口参数。
步骤S42，令人工智能体依据所述电路模型设计指标，针对电路模型执行电路优化设计处理，以生成电路优化模型，接着执行步骤S43。于本申请的实施例中，人工智能体还储存有用于构成电路模型的电路元件的初始化参数，且人工智能体是依据其所储存的电路元件的初始化参数以及电路模型设计指标，针对电路模型执行至少一次的电路优化设计处理。
步骤S43，模拟运行人工智能体所生成的电路优化模型以生成模拟运行结果，分析电路优化模型的模拟运行结果是否符合电路模型设计指标，据以输出分析结果(请容后在图5中予以详述)，接着执行步骤S44。
步骤S44，令所述人工智能体接收分析结果，并当分析结果为模拟运行结 果不符合电路模型设计指标时，则重复执行的前述的电路优化设计处理，以生成新的所述电路优化模型，直至当分析结果为模拟运行结果符合电路模型设计指标时，则结束执行电路优化设计处理。
请配合参阅图5，其为显示图4的无源电路优化设计方法的具体实施例流程图。
如图所示，首先执行步骤S51，模拟运行人工智能体所生成的电路优化模型，以生成模拟运行结果，接着执行步骤S52。
步骤S52，判断所生成的电路优化模型的模拟运行结果是否符合电路模型设计指标，当判断结果为符合时，则人工智能体输出所生成的电路优化模型，并结束针对所述待优化的电路模型的优化设计处理。当判断结果为不符合时，则进行步骤S53。
步骤S53，令人工智能体判断当前所生成的电路优化模型的模拟运行结果是否较前一次所生成的电路优化模型的模拟运行结果更为接近电路模型设计指标，若判断结果为否，则进行步骤S641，若判断结果为是，则进行步骤S642。
步骤S541，当判断结果为否，则代表人工智能体当前所生成电路优化模型不成功，或当前所执行的电路优化设计处理不理想，则输出负向反馈信号(例如惩罚信号)，接着执行步骤S55。
步骤S542，当判断结果为是，则代表人工智能体当前所生成电路优化模型达到了优化的效果，也就是当前所执行的电路优化设计处理是成功的，则输出正向反馈信号(例如奖励信号)，接着执行步骤S55。
步骤S55，令人工智能体依据所接收的正向反馈信号或负向反馈信号，以使所输出的正向反馈信号的概率不断增大为优化设计基准，判断下一次针对电路模型待执行的电路优化设计处理的操作方案，据以针对电路模型执行下一次的电路优化设计处理，以生成新的电路优化模型，并返回步骤S51。
于本申请的实施例中，步骤S55还包括令人工智能体记录每一次电路优化设计处理过程中各电路元件的调整对电路模型设计指标的影响的历史数据以及电路优化模型在模拟运行时各电路元件的调整对电路模型设计指标的影响的实时数据。例如，人工智能体根据所输出的正向反馈信号或负向反馈信号，分析模拟运行结果(实验数据)与电路模型设计指标(理论数据)间的变化，据以修改相关电路元件的设计参数，即执行电路优化设计处理操作，选择执行的下一次的电路优化处理操作的原则是使所给出的奖励信号的概率增大。如果人工智能体执行的某个电路优化设计处理操作导致正向的奖赏，那么其以后执行这 个电路优化设计处理操作的趋势便会加强；反之，则以后执行这个电路优化设计处理操作的趋势将减弱。
综上所述，本申请所提供的人工智能体训练系统通过将无源电路设计人员的设计经验数据化，采集电路优化过程的各个元器件调整对设计指标要求影响的历史数据及无源电路在线仿真时各个元器件调整对设计指标要求影响的实时数据等信息，以训练人工智能的无源电路优化设计功能。
同时，本申请还揭露了无源电路优化设计系统及方法，采用人工智能和增强学习技术，实现无源电路的自动优化设计，具有设计效率高，可复用程度高以及设计成本低廉的优点，还能够减少设计和实现过程的风险和不确定性，并使电路易于定制。
以上所描述的装置实施例仅仅是示意性的，其中所述作为分离部件说明的模块可以是或者也可以不是物理上分开的，作为模块显示的部件可以是或者也可以不是物理模块，即可以位于一个地方，或者也可以分布到多个网络模块上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。本领域普通技术人员在不付出创造性的劳动的情况下，即可以理解并实施。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到各实施方式可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件。基于这样的理解，上述技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品可以存储在计算机可读存储介质中，所述计算机可读记录介质包括用于以计算机(例如计算机)可读的形式存储或传送信息的任何机制。例如，机器可读介质包括只读存储器(ROM)、随机存取存储器(RAM)、磁盘存储介质、光存储介质、闪速存储介质、电、光、声或其他形式的传播信号(例如，载波、红外信号、数字信号等)等，该计算机软件产品包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行各个实施例或者实施例的某些部分所述的方法。
最后应说明的是：以上实施例仅用以说明本申请实施例的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围。
本领域的技术人员应明白，本发明实施例的实施例可提供为方法、装置(设备)、或计算机程序产品。因此，本发明实施例可采用完全硬件实施例、完全软 件实施例、或结合软件和硬件方面的实施例的形式。而且，本发明实施例可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本发明实施例是参照根据本发明实施例的方法、装置(设备)和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。

Claims (13)

  1. 一种人工智能体训练系统，应用于人工智能体，用于训练所述人工智能体的无源电路优化设计性能，其特征在于，包括：
    接收模块，用于接收电路模型；
    设定模块，用于提供设定电路模型设计指标；
    训练优化设计模块，用于依据所述电路模型设计指标而针对所述电路模型执行电路优化设计处理，并生成电路优化模型；以及
    计算分析模块，用于计算分析所述电路优化模型的模拟运行结果是否符合所述电路模型设计指标，并据以输出分析结果，以供所述训练优化设计模块依据所述分析结果训练更新其所执行的所述电路优化设计处理。
  2. 根据权利要求1所述的人工智能体训练系统，其特征在于，所述电路模型设计指标包括介质材料类型参数、电路元件类型及尺寸参数、导体类型及尺寸参数、输入接口参数及输出接口参数。
  3. 根据权利要求2所述人工智能体训练系统，其特征在于，所述人工智能体训练系统还包括资料库，用于储存构成所述电路模型的电路元件的初始化参数。
  4. 根据权利要求3所述的人工智能体训练系统，其特征在于，所述训练优化设计模块依据所述资料库中的所述电路元件的初始化参数以及所述电路模型设计指标，针对所述电路模型执行至少一次的电路优化设计处理。
  5. 根据权利要求4所述的人工智能体训练系统，其特征在于，所述计算分析模块还包括模拟运行所述训练优化设计模块每一次执行所述电路优化设计处理所生成的所述电路优化模型，以生成至少一次所述模拟运行结果，并判断所述训练优化设计模块当前所生成的所述电路优化模型的所述模拟运行结果是否较前一次所生成的所述电路优化模型的所述模拟运行结果更为接近所述电路模型设计指标，若分析结果为是，则输出正向反馈信号，反之，则输出负向反馈信号。
  6. 根据权利要求5所述的人工智能体训练系统，其特征在于，所述训练优化设计模块还包括依据所述计算分析模块所输出的所述正向反馈信号或负向反馈信号，记录每一次电路优化设计处理中的各所述电路元件的调整对所述电路模型设计指标的影响的历史数据以及所述电路优化模型在模拟运行时各电路元件的调整对所述电路模型设计指标的影响的实时数据。
  7. 根据权利要求6所述的人工智能体训练系统，其特征在于，所述训练优化设计模块还包括根据所述计算分析模块所输出的所述正向反馈信号或负向反馈信号，以使所述计算分析模块输出所述正向反馈信号的概率不断增大为优化设计基准，确定下一次针对所述电路模型所执行的电路优化调整处理的操作。
  8. 一种无源电路优化设计系统，为基于人工智能而实现，其特征在于，包括：
    接收单元，用于接收待优化的电路模型；
    设定单元，用于提供设定电路模型设计指标；以及
    由根据权利要求1至7中任一项所述的人工智能体训练系统训练生成的人工智能体；其中，
    所述人工智能体依据所设定的所述电路模型设计指标，对所接收的所述待优化的电路模型进行至少一次的优化设计处理，以生成符合所述电路模型设计指标的电路优化模型。
  9. 一种无源电路优化设计方法，借由人工智能体执行无源电路的优化设计，其特征在于，所述方法包括：
    向所述人工智能体输入待优化的电路模型，以及输入电路模型设计指标；
    令所述人工智能体依据所述电路模型设计指标，针对所述电路模型执行电路优化设计处理，以生成电路优化模型；
    模拟运行所述人工智能体所生成的所述电路优化模型以生成模拟运行结果，分析所述模拟运行结果是否符合所述电路模型设计指标，据以输出分析结果；以及
    令所述人工智能体接收所述分析，并当所述分析结果为所述模拟运行结果不符合所述电路模型设计指标时，则重复执行所述电路优化设计处理，以生成新的所述电路优化模型，直至当所述分析结果为所述模拟运行结果符合所述电 路模型设计指标时，则结束执行所述电路优化设计处理。
  10. 根据权利要求9所述的无源电路优化设计方法，其特征为，所述电路模型设计指标包括介质材料类型参数、电路元件类型及尺寸参数、导体类型及尺寸参数、输入接口参数及输出接口参数。
  11. 根据权利要求10所述的无源电路优化设计方法，其特征为，所述人工智能体还用于储存构成所述电路模型的电路元件的初始化参数，且所述人工智能体是依据所储存的所述电路元件的初始化参数以及所述电路模型设计指标，针对所述电路模型执行至少一次的电路优化设计处理。
  12. 根据权利要求11所述的无源电路优化设计方法，其特征为，所述方法还包括：
    模拟运行当前所生成的所述电路优化模型，以生成所述模拟运行结果；
    令所述人工智能体判断当前所生成的所述电路优化模型的所述模拟运行结果是否较前一次所生成的所述电路优化模型的所述模拟运行结果更为接近所述电路模型设计指标，若分析结果为是，则输出正向反馈信号，反之，则输出负向反馈信号；以及
    令所述人工智能体依据所输出的所述正向反馈信号或负向反馈信号，以使所输出的所述正向反馈信号的概率不断增大为优化设计基准，判断下一次针对所述电路模型待执行的电路优化设计处理的操作方案，据以针对所述电路模型执行下一次的电路优化设计处理，以生成新的电路优化模型。
  13. 根据权利要求12所述的无源电路优化设计方法，其特征为，所述方法还包括：令所述人工智能体记录每一次电路优化设计处理过程中各所述电路元件的调整对所述电路模型设计指标的影响的历史数据以及所述电路优化模型在模拟运行时各电路元件的调整对所述电路模型设计指标的影响的实时数据。
PCT/CN2018/085640 2018-04-18 2018-05-04 人工智能体训练系统及无源电路优化设计系统及方法 WO2019200626A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810350393.7A CN108595815B (zh) 2018-04-18 2018-04-18 人工智能体训练系统及无源电路优化设计系统及方法
CN201810350393.7 2018-04-18

Publications (1)

Publication Number Publication Date
WO2019200626A1 true WO2019200626A1 (zh) 2019-10-24

Family

ID=63611194

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/085640 WO2019200626A1 (zh) 2018-04-18 2018-05-04 人工智能体训练系统及无源电路优化设计系统及方法

Country Status (2)

Country Link
CN (1) CN108595815B (zh)
WO (1) WO2019200626A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113553794A (zh) * 2021-07-12 2021-10-26 苏州贝克微电子有限公司 一种用于电路设计的人工智能实现系统及方法
US11836602B2 (en) 2021-07-12 2023-12-05 Batelab Co., Ltd Method for AI-based circuit design and implementation system thereof

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109447245A (zh) * 2018-10-29 2019-03-08 石家庄创天电子科技有限公司 基于神经网络的等效模型生成方法以及建模方法
CN111368497A (zh) * 2020-03-13 2020-07-03 浪潮商用机器有限公司 一种电路板原理图设计方法、装置、电子设备及存储介质
CN115048885B (zh) * 2022-08-12 2022-11-15 阿里巴巴（中国）有限公司 电路设计参数调整方法、装置、电子设备和存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102024067A (zh) * 2009-09-09 2011-04-20 中国科学院微电子研究所 一种模拟电路工艺移植的方法
CN106022523A (zh) * 2016-05-23 2016-10-12 广东工业大学 一种基于集成仿真的自动化生产线优化设计方法
CN107679266A (zh) * 2017-08-22 2018-02-09 珠海泓芯科技有限公司 闪存电路的仿真方法及仿真装置
CN107679282A (zh) * 2017-09-05 2018-02-09 珠海博雅科技有限公司 电荷泵的仿真方法及仿真装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6876961B1 (en) * 1999-08-27 2005-04-05 Cisco Technology, Inc. Electronic system modeling using actual and approximated system properties
CN101714176B (zh) * 2009-12-21 2011-09-21 宁波大学 一种模拟运算放大器集成电路优化方法
CN103279824A (zh) * 2013-05-22 2013-09-04 国家电网公司 一种继电保护整定计算系统的建模方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102024067A (zh) * 2009-09-09 2011-04-20 中国科学院微电子研究所 一种模拟电路工艺移植的方法
CN106022523A (zh) * 2016-05-23 2016-10-12 广东工业大学 一种基于集成仿真的自动化生产线优化设计方法
CN107679266A (zh) * 2017-08-22 2018-02-09 珠海泓芯科技有限公司 闪存电路的仿真方法及仿真装置
CN107679282A (zh) * 2017-09-05 2018-02-09 珠海博雅科技有限公司 电荷泵的仿真方法及仿真装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113553794A (zh) * 2021-07-12 2021-10-26 苏州贝克微电子有限公司 一种用于电路设计的人工智能实现系统及方法
US11836602B2 (en) 2021-07-12 2023-12-05 Batelab Co., Ltd Method for AI-based circuit design and implementation system thereof

Also Published As

Publication number Publication date
CN108595815B (zh) 2021-05-28
CN108595815A (zh) 2018-09-28

Similar Documents

Publication Publication Date Title
WO2019200626A1 (zh) 人工智能体训练系统及无源电路优化设计系统及方法
JP7233807B2 (ja) 人工ニューラル・ネットワークにおける不確実性をシミュレートするためのコンピュータ実施方法、コンピュータ・システム、およびコンピュータ・プログラム
US12014258B2 (en) Method and device for optimizing simulation data, and computer-readable storage medium
KR102399535B1 (ko) 음성 인식을 위한 학습 방법 및 장치
US20190004920A1 (en) Technologies for processor simulation modeling with machine learning
US11501153B2 (en) Methods and apparatus for training a neural network
US11341034B2 (en) Analysis of verification parameters for training reduction
CN105900116A (zh) 分层型神经网络装置、判别器学习方法以及判别方法
US7707528B1 (en) System and method for performing verification based upon both rules and models
CN105528652A (zh) 一种预测模型的建立方法及终端
CN113343630A (zh) 建模方法及建模装置、电子设备及存储介质
US10537801B2 (en) System and method for decision making in strategic environments
WO2007050622A2 (en) Weighted pattern learning for neural networks
US20180114281A1 (en) System and tool to configure well settings for hydrocarbon production in mature oil fields
CN110941934A (zh) 一种fpga原型验证开发板分割仿真系统、方法、介质及终端
CN116258097A (zh) 一种更新水文模型方法、装置、电子设备和介质
CN114547917A (zh) 仿真预测方法、装置、设备及存储介质
KR20220117123A (ko) Ai트윈 기반 경제 선순환 시뮬레이션 시스템 및 방법
KR102334532B1 (ko) 단백질 상호 작용 네트워크에서 신호 전파 간섭을 통한 약력학적 약물 상호 작용 예측 장치 및 방법
US11409928B2 (en) Configurable digital twin
US11966851B2 (en) Construction of a machine learning model
CN112288032B (zh) 一种基于生成对抗网络的量化模型训练的方法及装置
CN113762061A (zh) 神经网络的量化感知训练方法、装置及电子设备
CN114238106A (zh) 测试时间预测方法及装置、电子设备、存储介质
WO2020134011A1 (zh) 展示信息组合确定方法、装置、存储介质及电子设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18915086

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18915086

Country of ref document: EP

Kind code of ref document: A1