Automatic design and optimization system for hydraulic speed control loop

Abstract

The invention discloses an automatic design and optimization system for a hydraulic speed control loop, which takes a large language model intelligent body as a core, realizes automatic design and optimization of the hydraulic speed control loop through semantic comprehension, knowledge retrieval and parameter optimization, and specifically comprises an LLM intelligent body, a data processing module and a data processing module, the analysis module is used for analyzing natural language input containing design requirements and performance targets of a user, generating a loop structure in combination with a local knowledge base and calling an optimization algorithm; the local knowledge base comprises a basic hydraulic loop, a key empirical relationship and a typical structure template which are artificially extracted; and the simulation and optimization module is used for executing a swarm intelligence optimization algorithm to optimize and adjust parameters of the loop structure. The invention constructs a hydraulic speed control loop automatic design and optimization system with semantic understanding, automatic generation and adaptive optimization capabilities, and aims to realize full-process intelligent design from user demand input to loop structure generation and performance optimization.

Description

Technical Field

[0001] This invention relates to the field of hydraulic circuit design, and in particular to an automatic design and optimization system for hydraulic speed control circuits. Background Technology

[0002] Hydraulic systems, as an important form of power and control, are widely used in aerospace, engineering machinery, industrial automation, and intelligent equipment. Their core function lies in achieving energy conversion and precise control through hydraulic components (such as valves, pumps, and actuators). In traditional design processes, the structure and parameter settings of hydraulic circuits typically rely on rules of thumb and manual adjustments, resulting in long design cycles, high iteration costs, and difficulty in achieving optimal performance matching under multiple operating conditions. To improve design efficiency and reliability, research institutions both domestically and internationally have attempted to introduce numerical simulation and optimization algorithms into hydraulic system design. For example, companies such as Caterpillar and Bosch Rexroth have developed parametric simulation platforms based on AMESim and SimHydraulics, which can achieve local performance optimization through parameter scanning; RWTH Aachen University in Germany proposed a predictive design framework for hydraulic systems based on digital twins, realizing multi-objective performance prediction and parameter tuning under a fixed topology. In addition, some academic institutions have introduced intelligent optimization methods such as genetic algorithms (GA) and particle swarm optimization (PSO) to search and optimize the control parameters of hydraulic servo valves, load sensing systems, etc., in order to reduce energy loss and improve response speed. Although these solutions improve design efficiency to some extent, they still rely on manually defined topologies and static optimization models, and cannot achieve truly end-to-end automatic design and performance adaptive optimization.

[0003] Currently, the design and optimization of hydraulic circuits still mainly rely on manual experience and general simulation platforms, resulting in a low overall level of automation and intelligence, and exhibiting the following main shortcomings:

[0004] 1. The design process relies on expert experience and has a low degree of automation.

[0005] Traditional hydraulic system design typically relies on engineers selecting components based on experience and manually drawing circuit diagrams. Although existing tools (such as AMESim and Simscape Fluids) support parametric modeling and numerical simulation, the system structure still needs to be manually constructed, lacking the ability to automatically generate circuits based on performance targets, resulting in poor design efficiency and consistency.

[0006] 2. Lack of intelligent mapping mechanism between performance targets and loop structure

[0007] The existing system cannot automatically derive the corresponding hydraulic component combination and connection method based on the performance requirements set by the user (such as pressure regulation accuracy, flow range, response time, etc.). The design process lacks semantic understanding and adaptive capabilities, resulting in performance verification relying on repeated trial calculations and experience adjustments.

[0008] 4. The design, simulation, and optimization processes are disconnected, and data interfaces are inconsistent.

[0009] In traditional workflows, loop design, simulation analysis, and parameter optimization are often completed on different software platforms. Data transfer requires manual conversion, which reduces computational efficiency and makes it difficult to ensure parameter consistency, thus failing to form a closed-loop optimization mechanism. Summary of the Invention

[0010] The purpose of this invention is to address the shortcomings of existing technologies. The goal is to construct an automatic design and optimization system for hydraulic speed control loops with semantic understanding, automatic generation, and adaptive optimization capabilities. This system aims to achieve intelligent design throughout the entire process, from user requirement input to loop structure generation and performance optimization. It focuses on solving problems such as reliance on experience in the design process, the inability to automatically generate structures, and the disconnect between the design, simulation, and optimization processes. Specifically:

[0011] By introducing the semantic parsing capabilities of a large language model, users can describe design requirements in natural language. The system can then automatically identify performance targets and call the knowledge base to generate an initial loop structure, achieving a direct conversion from language description to hydraulic speed control loop, significantly reducing the design threshold.

[0012] A hydraulic component template library and structural combination rules were established. LLM reasoning capabilities were used to match component functions with connection logic, enabling the automatic generation of hydraulic speed control loops that meet target requirements based on performance constraints. The system can autonomously select structural modules such as overflow, pressure reduction, throttling, and pilot flow control under different design objectives.

[0013] An integrated design framework is constructed, which combines loop modeling, simulation calculation and swarm intelligence optimization algorithm (such as artificial bee colony ABC) to realize automatic optimization of loop parameters and performance iteration, forming a continuous closed-loop design-simulation-optimization process.

[0014] The objective of this invention is achieved through the following technical solution: an automatic design and optimization system for hydraulic speed control loops. This system, with a large language model intelligent agent as its core, achieves automatic design and optimization of hydraulic speed control loops through semantic understanding, knowledge retrieval, and parameter optimization. Specifically, it includes:

[0015] The intelligent agent is used to parse the user's natural language input containing design requirements and performance goals, combine it with the local knowledge base to generate loop structures and call optimization algorithms;

[0016] A local knowledge base containing manually refined basic hydraulic circuits, key empirical relationships, and typical structural templates;

[0017] The simulation and optimization module is used to execute swarm intelligence optimization algorithms to optimize and adjust the parameters of the loop structure.

[0018] Furthermore, the performance targets input by the user include target output pressure, target output flow rate, and flow response speed.

[0019] Furthermore, the intelligent agent transforms performance goals into executable design tasks, identifies design goals, constraints, and key physical quantities through semantic decomposition, and automatically generates design parameter tables and task structure trees.

[0020] Furthermore, the intelligent agent performs parameter calculations and geometric modeling to form an initial loop structure, automatically establishes a flow and pressure relationship model based on the flow and pressure balance equation, calculates the effective area range of the throttling orifice, pressure drop distribution, and leakage correction coefficient according to the pressure and flow indicators set by the user, and forms a preliminary geometric model.

[0021] Furthermore, the simulation and optimization module takes the preliminary geometric model as input and optimizes the circuit structure parameters under the premise of meeting the constraints of voltage regulation accuracy, dynamic response and energy efficiency through the search and local update mechanism of the bee colony.

[0022] Furthermore, the optimized circuit structure parameters include valve port size, spring stiffness, and damping coefficient.

[0023] Furthermore, based on the circuit structure parameters adjusted by the simulation and optimization module, the circuit is uniformly evaluated using the comprehensive performance index of hydraulic small signals to obtain the pressure stabilization performance, dynamic response capability, and energy utilization efficiency of the hydraulic circuit.

[0024] Furthermore, the pressure stabilization performance, dynamic response capability, and energy utilization efficiency of the hydraulic circuit are calculated simultaneously using candidate solutions with different circuit structure parameters. An automatic cycle of design-simulation-optimization is achieved through LLM scheduling. Based on the evaluation results, each candidate scheme is dynamically scored and ranked, and an optimized circuit, key component parameters, and performance prediction report that meet the requirements are output.

[0025] Furthermore, large-scale language models employ LLM, Code Llama, or Mistral Engineer.

[0026] Furthermore, the swarm intelligence optimization algorithm adopts artificial bee colony algorithm (ABC), genetic algorithm (GA), particle swarm optimization algorithm (PSO), gray wolf optimization algorithm (GWO), or whale optimization algorithm (WOA); or it adopts reinforcement learning (RL) strategy to use the hydraulic circuit performance as a reward signal to achieve dynamic self-learning optimization.

[0027] The beneficial effects of this invention: This invention proposes an automatic design method and system for hydraulic speed control loops, achieving significant breakthroughs compared to existing technologies in terms of design automation, intelligent autonomy, and system consistency. A specific comparison is as follows:

[0028] 1. Semantic-driven automated design and task understanding

[0029] Traditional hydraulic system design relies on human experience and manual modeling, resulting in a lengthy design process, significant human intervention, and a reliance on extensive engineering experience, leading to low design efficiency and high subjectivity. This invention introduces Large Language Modeling (LLM) into hydraulic circuit design for the first time. Through natural language understanding and semantic parsing, it achieves automatic generation from user requirements to design parameters. Users only need to input a single command (e.g., "Design a throttling speed control circuit that can achieve an adjustable flow rate of 20-50 L / min in a 25 MPa system, maintain stable speed under load changes, and requires fast response, no low-speed creep, and low energy depletion"), and the system can automatically infer the appropriate structural type and initial parameters, significantly improving the intelligence of design task transformation.

[0030] 2. Design-Simulation-Optimization Integrated Closed-Loop Architecture

[0031] Existing hydraulic design processes are mostly step-by-step, requiring manual data exchange between design, simulation, and optimization, resulting in low efficiency and poor consistency. This invention establishes an integrated closed-loop optimization system driven by LLM (Liquid Dynamics Management). After design completion, it automatically invokes the simulation module for performance analysis and updates parameters based on the results, achieving a closed-loop design. This significantly shortens the design iteration cycle, improves optimization convergence speed, and realizes truly intelligent closed-loop design.

[0032] This invention automates the entire hydraulic system design process—from performance requirement input to loop structure generation, parameter optimization, and performance evaluation—by introducing semantic understanding from language models and an adaptive search mechanism based on swarm intelligence optimization. This breakthrough overcomes the limitations of traditional hydraulic loop design, which heavily relies on expert experience and manual parameter tuning. It achieves a highly efficient and precise intelligent hydraulic system design process, enabling automatic mapping from natural language descriptions to hydraulic loop structure models. This significantly improves the efficiency and intelligence of the preliminary design stage of hydraulic systems, providing an innovative solution with industrial potential for automated modeling and performance optimization of complex hydraulic equipment. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a schematic diagram of the LADHC process.

[0035] Figure 2 This is a schematic diagram of a throttling speed control circuit.

[0036] Figure 3 This is a schematic diagram of the pressure reduction control loop. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are merely illustrative and not intended to limit the invention.

[0038] like Figure 1 As shown, this invention provides an automatic design and optimization system for hydraulic speed control circuits based on a large language model intelligent agent (LADHC). This system uses a large language model (LLM) intelligent agent based on DeepSeek as its core. Through semantic understanding, knowledge retrieval, and parameter optimization, it achieves automatic design and optimization of hydraulic speed control circuits, automatically generating a hydraulic circuit framework including components such as hydraulic pumps, relief valves, throttling elements, and hydraulic cylinders. This overcomes the limitations of traditional hydraulic design, which relies on human experience and lacks an integrated framework. As an alternative, language models specifically optimized for engineering knowledge, such as Code Llama and Mistral Engineer, can be used to improve the accuracy of parameter calculation and equation generation. These models can also parse natural language input and output the hydraulic system structure.

[0039] The LADHC system consists of three parts: an LLM agent, a local knowledge base, and a simulation and optimization module. The LLM agent is responsible for parsing user input, generating loop structures, and invoking optimization algorithms. The local knowledge base contains manually refined basic hydraulic loops (pressure reduction, throttling, synchronization, etc.), key empirical relationships, and typical structural templates. The simulation and optimization module is responsible for parameter adjustments and swarm intelligence optimization. When designing a hydraulic loop, the user only needs to input design requirements and performance goals in natural language, such as "Design a throttling speed control loop that can achieve an adjustable flow rate of 20-50 L / min in a 25 MPa system, maintain stable speed under load changes, and requires fast response, no low-speed creep, and low energy consumption." LADHC can retrieve loop and structural templates (such as basic knowledge of common hydraulic loops, pilot-operated hydraulic control loops, proportional servo valve design, and other technical literature) based on user input and the local knowledge base. Figure 2 The diagram shows a typical throttling speed control circuit. Besides throttling speed control circuits, this invention is also applicable to the design of typical hydraulic systems such as pressure reduction control circuits. Figure 3 The diagram shows a hydraulic cylinder pressure reduction control circuit. When a user inputs "Design a pressure reduction control circuit that can stabilize the branch pressure at 7 MPa and output a flow rate of 40 L / min in a 25 MPa system, requiring fast response and low energy consumption," the LLM agent will automatically parse the requirement and, combined with relevant pressure reduction control content in the local knowledge base, automatically generate a circuit framework including a hydraulic pump, relief valve, pressure reducing valve, and actuator.

[0040] The automatic design process of this invention can be summarized as follows: First, the system uses an LLM agent to parse the user's input natural language requirements, transforming performance objectives (such as speed ratio, load flow rate, energy efficiency requirements, etc.) into executable design tasks. The LLM identifies design objectives, constraints, and key physical quantities through semantic decomposition, automatically generating a design parameter table and task structure tree. Then, it retrieves the most relevant loop templates and key design experiences from the local knowledge base. By comparing historical samples with theoretical equations, the system determines the most suitable loop type and uses typical hydraulic equations and empirical formulas (such as the relationship between throttle valve flow rate and orifice area) for parameter calculation and geometric modeling, forming the initial loop structure. Next, the system automatically establishes a flow-pressure relationship model based on the flow-pressure balance equation.

[0041]

[0042]

[0043] in It is the pump's output flow rate. It refers to the load flow, that is, the effective flow entering the actuator. It's a data leak. It is the overflow flow, the excess flow that returns to the oil tank through the overflow valve; It is the flow rate of the throttling branch. It is the flow coefficient, used to correct for energy loss in ideal flow. The effective flow area is determined by the valve opening. The intelligent agent calculates the effective area range of the orifice, pressure drop distribution, and leakage correction coefficient based on the pressure and flow indicators set by the user, and forms a preliminary geometric model.

[0044] The main problem with large-scale language models is the excessive time consumption of each iteration, which hinders their optimization of hydraulic circuits that already meet or are close to meeting performance requirements. To further improve performance, LADHC introduces the Artificial Bee Colony (ABC) swarm intelligence algorithm for parallel optimization. ABC takes the initial circuit structure as input and optimizes system parameters, such as valve size, spring stiffness, and damping coefficient, through a bee colony search and local update mechanism, while meeting constraints on pressure regulation accuracy, dynamic response, and energy efficiency, thereby achieving multi-objective performance improvement of the circuit. Besides the ABC algorithm, other swarm intelligence algorithms can be used as alternative paths. For example, the Genetic Algorithm (GA) can optimize valve size and spring stiffness through selection, crossover, and mutation operations; the Particle Swarm Optimization (PSO) algorithm can achieve fast convergence using velocity and position update rules; the Grey Wolf Optimization (GWO) and Whale Optimization (WOA) algorithms are suitable for global search under multi-objective constraints; furthermore, reinforcement learning (RL) strategies can also use hydraulic circuit performance as a reward signal to achieve dynamic self-learning optimization. Although these algorithms differ in their search mechanisms, they can all achieve global optimization of key performance indicators of hydraulic systems and achieve the same goal under different computing architectures.

[0045] Finally, the system was evaluated using the Hydraulic Figure of Merit for Small Signal (HFOM) comprehensive performance index. S A unified evaluation of the circuits was conducted, HFOM S The definition is as follows:

[0046]

[0047] in, For flow gain, For system bandwidth, For energy efficiency, For steady-state error, This refers to pressure fluctuation. This indicator comprehensively reflects the pressure stabilization performance, dynamic response capability, and energy utilization efficiency of the hydraulic circuit, ensuring that the final output hydraulic circuit achieves an optimal balance between response speed, energy utilization, and stability. (Based on HFOM)S Based on the evaluation results, the system outputs optimized circuits, key component parameters, and performance prediction reports, realizing fully automated design from user requirement input to optimal hydraulic circuit output.

[0048] This invention uses a classic throttling speed control circuit as an example, detailing its design process and results. Guided by a local knowledge base, LADHC successfully designed a throttling speed control circuit that met the expected requirements by selecting a suitable circuit template and comparing appropriate component parameters. To evaluate the performance of the throttling speed control circuit, we compared the performance indicators of the expected values, LADHC's initial design values, and LADHC's final design values ​​after further optimization using ABC. The results are shown in Table 1, demonstrating that LADHC possesses the capability to design and optimize throttling speed control circuits.

[0049] Table 1

[0050]

[0051] According to the table, the throttling speed control circuit after applying the ABC algorithm not only met the expected performance but also exhibited higher performance indicators. This indicates that the additional optimization of ABC not only achieved the target performance but also improved the overall performance of the hydraulic circuit.

[0052] This invention is not only applicable to throttling speed control circuits, but can also be extended to typical hydraulic systems such as pressure reduction control, synchronization control, overflow regulation, and pilot control. By combining LLM semantic reasoning with ABC optimization, the system can achieve adaptive circuit generation and parameter optimization under different operating conditions and performance requirements, significantly improving the automation level and performance consistency of hydraulic system design.

[0053] The above embodiments are used to explain and illustrate the present invention, but not to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. An automatic design and optimization system for hydraulic speed control loops, characterized in that, This system, centered on a large language model intelligent agent, achieves automatic design and optimization of hydraulic speed control loops through semantic understanding, knowledge retrieval, and parameter optimization. Specifically, it includes: The intelligent agent is used to parse the user's natural language input containing design requirements and performance goals, combine it with the local knowledge base to generate loop structures and call optimization algorithms; A local knowledge base, including manually refined basic hydraulic circuits, key empirical relationships, and typical structural templates; The simulation and optimization module is used to execute swarm intelligence optimization algorithms to optimize and adjust the parameters of the loop structure.

2. The automatic design and optimization system for hydraulic speed control circuits according to claim 1, characterized in that, User-inputted performance targets include target output pressure, target output flow rate, and flow response speed.

3. The automatic design and optimization system for hydraulic speed control circuits according to claim 1, characterized in that, The intelligent agent transforms performance goals into executable design tasks, identifies design goals, constraints, and key physical quantities through semantic decomposition, and automatically generates design parameter tables and task structure trees.

4. The automatic design and optimization system for hydraulic speed control circuit according to claim 1, characterized in that, The intelligent agent performs parameter calculations and geometric modeling to form an initial loop structure. Based on the flow and pressure balance equation, it automatically establishes a flow and pressure relationship model. According to the pressure and flow indicators set by the user, it calculates the effective area range of the throttling orifice, pressure drop distribution, and leakage correction coefficient, and forms a preliminary geometric model.

5. The automatic design and optimization system for hydraulic speed control loop according to claim 4, characterized in that, The simulation and optimization module takes the preliminary geometric model as input and optimizes the circuit structure parameters under the premise of meeting the constraints of voltage regulation accuracy, dynamic response and energy efficiency through the search and local update mechanism of bee colony.

6. The automatic design and optimization system for hydraulic speed control loop according to claim 5, characterized in that, The optimized circuit structure parameters include valve port size, spring stiffness, and damping coefficient.

7. The automatic design and optimization system for hydraulic speed control circuit according to claim 1, characterized in that, Based on the circuit structure parameters adjusted by the simulation and optimization module, the circuit is uniformly evaluated through the comprehensive performance index of hydraulic small signals to ensure that the final output hydraulic circuit achieves the optimal balance between response speed, energy utilization rate and stability.

8. The automatic design and optimization system for hydraulic speed control loop according to claim 7, characterized in that, The pressure stabilization performance, dynamic response capability, and energy utilization efficiency of the hydraulic speed control circuit are calculated simultaneously using candidate solutions with different circuit structure parameters. The design-simulation-optimization cycle is automatically realized through LLM scheduling. Based on the evaluation results, each candidate scheme is dynamically scored and ranked, and the optimized circuit, key component parameters, and performance prediction report that meet the requirements are output.

9. The automatic design and optimization system for hydraulic speed control circuit according to claim 1, characterized in that, Large language models employ LLM, Code Llama, or Mistral Engineer.

10. The automatic design and optimization system for hydraulic speed control circuits according to claim 1, characterized in that, The swarm intelligence optimization algorithm adopts artificial bee colony algorithm (ABC), genetic algorithm (GA), particle swarm optimization algorithm (PSO), gray wolf optimization algorithm (GWO), or whale optimization algorithm (WOA); or it adopts reinforcement learning (RL) strategy to use the performance of hydraulic speed control loop as reward signal to achieve dynamic self-learning optimization.

Images