Artificial intelligence creative design agent for high-performance multi-axis machine tool structure and implementation method thereof

Abstract

The application provides an AI creative design intelligent agent for a high-performance multi-axis machine tool structure and an implementation method thereof, and comprises the following steps: representing a design space as a constraint vector and analyzing upper and lower bounds of the value of a design parameter; inputting a plurality of preset performance indicators and the constraint vector into a structure creative design model to generate a plurality of groups of dimensionless candidate design parameters represented in a normalized scale in a dimensionless design space constructed by the upper and lower bounds; converting the plurality of groups of dimensionless candidate parameters into actual candidate design parameters; performing performance evaluation on the plurality of groups of actual candidate design parameters through a performance prediction model to obtain a plurality of groups of performance prediction results; comparing the plurality of groups of performance prediction results with the corresponding preset performance indicators to obtain a plurality of groups of performance deviations; training the structure creative design model by using the plurality of groups of performance deviations; inputting a target performance indicator and a target constraint vector into the trained structure creative design model to obtain corresponding dimensionless candidate parameters and finally convert the dimensionless candidate parameters into target actual design parameters.

Description

Technical Field

[0001] This invention relates to the field of multi-axis machine tool structural parameter design technology, and more specifically, to an AI generative design agent for high-performance multi-axis machine tool structures and its implementation method. Background Technology

[0002] Machine tool design is at the forefront of multi-axis machine tool development, directly determining its machining accuracy, dynamic response, and long-term service stability. Domestic high-end multi-axis machine tool design primarily relies on reverse engineering, typically using mature foreign models as references. This involves analyzing and replicating their overall configuration, key component layout, transmission chain configuration, and support connection methods through external mapping. However, the core of high-performance machine tool design lies not only in its structural shape but also in the deep design knowledge accumulated over time, such as load-bearing paths, static and dynamic thermal performance coordination, and assembly precision control. These deep physical laws and design knowledge largely exist as tacit experience of experts and are difficult to acquire through geometric reverse engineering. Therefore, traditional reverse design often results in a superficial resemblance, failing to directly meet the high-performance requirements of machine tools. Due to the complex structure and multi-level coupling of components in multi-axis machine tools, meeting the design requirements of high-performance machine tools usually requires repeated trial and error verification through local modifications, modeling analysis, and simulation verification, leading to long development cycles and high costs.

[0003] Secondly, while existing forward design techniques such as topology optimization can generate high-performance structural solutions to some extent, they still have limitations in the overall design of high-performance multi-axis machine tools. On the one hand, traditional topology optimization is usually focused on a single component or limited operating conditions, resulting in long solution cycles and difficulties in multi-objective, multi-constraint collaborative design. It is difficult to simultaneously consider the comprehensive requirements of the machine tool's static stiffness, dynamic response, thermal stability, accuracy retention, and assembly feasibility. On the other hand, its optimization results often have complex boundaries, with problems such as difficulty in parametric reconstruction and insufficient manufacturing and assembly adaptability, making it difficult to directly convert them into usable machine tool CAD assembly models. In addition, although existing generative AI models have shown some ability to generate solutions in engineering structural design, due to the lack of physical mechanism constraints and performance interpretability, they are prone to producing structural forms that are "unmanufacturable, unassembleable, and unverifiable," such as geometric discontinuities, unreasonable load-bearing paths, and local weaknesses, making it difficult to meet the reliability and engineering feasibility requirements of high-performance machine tool design.

[0004] Therefore, there is an urgent need to develop an AI generative design agent for high-performance multi-axis machine tool structures, so as to realize an intelligent design method that can quickly generate CAD assembly models of multi-axis machine tool structures from performance requirements. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide an AI generative design agent for high-performance multi-axis machine tool structures and its implementation method.

[0006] The present invention provides a method for implementing an AI generative design agent for high-performance multi-axis machine tool structures, comprising: Step S1: Represent the geometric dimension range, spatial boundary, machining stroke and assembly constraints in the design space as a constraint vector, and parse the constraint vector into the upper and lower bounds of the corresponding design parameters; Step S2: The multiple preset performance indicators and the constraint vector are used together as input to the structural generative design model. The structural generative design model generates multiple sets of dimensionless candidate design parameters expressed in normalized proportions within the dimensionless design space constructed by the upper and lower bounds of the values. Step S3: Convert the multiple sets of dimensionless candidate parameters into actual candidate design parameters through mapping relationships; Step S4: Evaluate the performance of multiple sets of actual candidate design parameters using the performance prediction model to obtain multiple sets of performance prediction results; compare the multiple sets of performance prediction results with the corresponding preset performance indicators to obtain multiple sets of performance deviations; Step S5: Use multiple sets of performance deviations to perform feedback training and network parameter updates on the structural generative design model so that the design parameters can meet the corresponding design constraints and performance index requirements. Step S6: Input the target performance index and target constraint vector into the trained structure creation design model to obtain the corresponding dimensionless candidate parameters, and convert the obtained dimensionless candidate parameters into the final target actual design parameters.

[0007] Preferably, the performance prediction model includes: Step S4.1: Perform parametric modeling for various types of multi-axis machine tool structures, generate multiple sets of machine tool structure instances through parameter sampling, and obtain corresponding performance index data for each structure instance through simulation, thereby forming a mapping dataset between design parameters and performance indexes; Step S4.2: Preprocess the mapping dataset between design parameters and performance indicators to obtain the preprocessed mapping dataset between design parameters and performance indicators; Step S4.3: Construct a performance prediction model. Use the preprocessed mapping dataset between design parameters and performance indicators to train the performance prediction model and obtain the trained performance prediction model.

[0008] Preferably, step S4.1 includes: Step S4.1.1: Perform layered modeling of the overall structure of multi-axis machine tools of various types, including: parametrically expressing the geometric dimensions, spatial positions and connection relationships of key components of each type of multi-axis machine tool that meet the preset requirements; Step S4.1.2: Establish the constraint relationships between parameters, including geometric constraints and assembly constraints, to ensure that the structure remains effective throughout the parameter changes; Step S4.1.3: Within the preset parameter range, generate multiple sets of machine tool structure instances through parameter sampling, and obtain corresponding performance index data for each structure instance through simulation to form a mapping dataset between design parameters and performance indexes; The simulation includes: performing simulation using any one or more methods, including finite element analysis, geometric tolerance modeling, and Monte Carlo simulation. The performance indicators include any one or more of the following: structural mass, static stiffness, modal characteristics, thermal stability, and assembly accuracy.

[0009] Preferably, step S4.2 includes: Step S4.2.1: Identify and remove outlier data in the mapping dataset between design parameters and performance indicators; Step S4.2.2: Perform unified scale mapping on different physical quantities in the mapped dataset after anomaly processing to transform them into dimensionless or unified scale representations; Step S4.2.3: Standardize or normalize the mapped dataset after uniform scale mapping to eliminate the impact of dimensional differences.

[0010] Preferably, the performance prediction model includes: a surrogate model based on neural networks or a regression model based on statistical learning; The performance prediction model is based on any set of design parameters to obtain the corresponding performance prediction results.

[0011] An AI generative design agent for high-performance multi-axis machine tool structures, provided by the present invention, includes: Module M1: Unifies the geometric dimension range, spatial boundary, machining stroke and assembly constraints in the design space into constraint vectors, and resolves the constraint vectors into the upper and lower bounds of the corresponding design parameters; Module M2: The structural generative design model takes multiple preset performance indicators and the constraint vector as inputs, and generates multiple sets of dimensionless candidate design parameters expressed in normalized proportions within the dimensionless design space constructed by the upper and lower bounds of the values. Module M3: Converts multiple sets of dimensionless candidate parameters into actual candidate design parameters through mapping relationships; Module M4: The performance prediction model is used to evaluate the performance of multiple sets of actual candidate design parameters to obtain multiple sets of performance prediction results; the multiple sets of performance prediction results are compared with the corresponding preset performance indicators to obtain multiple sets of performance deviations; Module M5: Uses multiple sets of performance deviations to perform feedback training and network parameter updates on the structural generative design model, so that the design parameters can meet the corresponding design constraints and performance index requirements; Module M6: Input the target performance index and target constraint vector into the trained structure creation design model to obtain the corresponding dimensionless candidate parameters, and convert the obtained dimensionless candidate parameters into the final target actual design parameters.

[0012] Preferably, the performance prediction model includes: Module M4.1: Performs parametric modeling for various types of multi-axis machine tool structures. It generates multiple sets of machine tool structure instances through parameter sampling, and obtains corresponding performance index data for each structure instance through simulation, thereby forming a mapping dataset between design parameters and performance indicators. Module M4.2: Preprocesses the mapping dataset between design parameters and performance indicators to obtain a preprocessed mapping dataset between design parameters and performance indicators; Module M4.3: Construct a performance prediction model. Use the preprocessed mapping dataset between design parameters and performance metrics to train the performance prediction model and obtain the trained performance prediction model.

[0013] Preferably, module M4.1 includes: Module M4.1.1: Performs layered modeling of the overall structure of multi-axis machine tools of various types, including: parametrically expressing the geometric dimensions, spatial positions and connection relationships of key components of each type of multi-axis machine tool that meet preset requirements; Module M4.1.2: Establishes constraints between parameters, including geometric constraints and assembly constraints, to ensure that the structure remains effective during parameter changes; Module M4.1.3: Within the preset parameter range, multiple sets of machine tool structure instances are generated through parameter sampling, and the corresponding performance index data are obtained through simulation for each structure instance, forming a mapping dataset between design parameters and performance indexes; The simulation includes: performing simulation using any one or more methods, including finite element analysis, geometric tolerance modeling, and Monte Carlo simulation. The performance indicators include any one or more of the following: structural mass, static stiffness, modal characteristics, thermal stability, and assembly accuracy.

[0014] Preferably, the module M4.2 includes: Module M4.2.1: Identifies and removes outlier data in the mapping dataset between design parameters and performance indicators; Module M4.2.2: Performs unified scale mapping on different physical quantities in the mapped dataset after anomaly handling, transforming them into dimensionless or unified scale representations; Module M4.2.3: Standardizes or normalizes the mapped dataset after uniform scale mapping to eliminate the impact of dimensional differences.

[0015] Preferably, the performance prediction model includes: a surrogate model based on neural networks or a regression model based on statistical learning; The performance prediction model is based on any set of design parameters to obtain the corresponding performance prediction results.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. Transform the traditional machine tool structure design mode that relies on experience and trial and error into an intelligent generative design mode that is directly driven by performance requirements, significantly shortening the design cycle and reducing R&D costs; 2. This invention overcomes the limitations of traditional topology optimization results, which are difficult to parametrically reconstruct and generate manufacturable and assemblable machine tool design models, and enables the design results to have the advantages of parametric expression and CAD reconstruction. 3. This invention improves the physical rationality, performance reliability, and engineering verifiability of AI-generated structures, reduces the risks of geometric failure, abnormal load-bearing paths, and local weaknesses, and makes the generated results reliable and engineering-applicable. 4. This invention solves the problem that the traditional machine tool structure "reverse copying" design mode, which relies on experience accumulation and iterative trial and error, cannot meet the needs of high-performance and high-efficiency research and development. Attached Figure Description

[0017] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 Flowchart of the AI ​​generative design intelligent agent implementation method for high-performance multi-axis machine tool structures.

[0018] Figure 2 This is a schematic diagram of the parametric model of a five-axis vertical machining center.

[0019] Figure 3 A schematic diagram of a structure creation design model based on supervised learning.

[0020] Figure 4 This document provides the design requirements input and structural CAD assembly model output schematic diagram for a five-axis vertical machining center. Detailed Implementation

[0021] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0022] Example 1 According to the present invention, an AI generative design agent implementation method for high-performance multi-axis machine tool structures is provided, such as... Figure 1 As shown, it includes: Step S1: Represent the geometric dimension range, spatial boundary, machining stroke and assembly constraints in the design space as a constraint vector, and parse the constraint vector into the upper and lower bounds of the corresponding design parameters; Step S2: The multiple preset performance indicators and the constraint vector are used together as input to the structural generative design model. The structural generative design model generates multiple sets of dimensionless candidate design parameters expressed in normalized proportions within the dimensionless design space constructed by the upper and lower bounds of the values. Step S3: Convert the multiple sets of dimensionless candidate parameters into actual candidate design parameters through mapping relationships; More specifically, the geometric dimensional range, spatial boundaries, machining strokes, and assembly constraints in the design space are uniformly represented as constraint vectors and used as inputs in the calculation process of the structural generative design model. The structural generative design model outputs a relative proportional representation of the design parameters within the corresponding constraint range. This proportional representation is in dimensionless form and describes the position of the design parameters within their value intervals. Subsequently, based on the upper and lower bounds of the design parameters, the dimensionless proportions are converted into actual design parameters through a proportional mapping relationship. Through this mechanism, the structural generation process is decoupled from specific geometric dimensions. Thus, when the design space changes, only the constraint vectors need to be adjusted to achieve adaptive structural generation, without the need to retrain the model.

[0023] Step S4: Evaluate the performance of multiple sets of actual candidate design parameters using the performance prediction model to obtain multiple sets of performance prediction results; compare the multiple sets of performance prediction results with the corresponding preset performance indicators to obtain multiple sets of performance deviations; Step S5: Use multiple sets of performance deviations to perform feedback training and network parameter updates on the structural generative design model so that the design parameters can meet the corresponding design constraints and performance index requirements. Step S6: Input the target performance index and target constraint vector into the trained structure creation design model to obtain the corresponding dimensionless candidate parameters, and convert the obtained dimensionless candidate parameters into the final target actual design parameters.

[0024] The performance prediction model includes: Step S4.1: Perform parametric modeling for various types of multi-axis machine tool structures, generate multiple sets of machine tool structure instances through parameter sampling, and obtain corresponding performance index data for each structure instance through simulation, thereby forming a mapping dataset between design parameters and performance indexes; More specifically, a parametric 3D model of a multi-axis machine tool structure is established, and layered modeling is performed on the overall structure of various types of machine tools. The geometric dimensions, spatial positions, and connection relationships of key components of each type of multi-axis machine tool are parametrically expressed. Based on this, the machine tool structure parameters are divided into design parameters and associated parameters. Design parameters directly control the structural morphology and topological features, while associated parameters are derived from the design parameters through preset geometric relationships or engineering rules to ensure structural integrity and assembly consistency. Simultaneously, constraint relationships between parameters are established, including geometric constraints and assembly constraints, to ensure the structure remains effective throughout parameter changes. Within the preset parameter range, multiple machine tool structure instances are generated through parameter sampling, and corresponding performance index data are obtained through refined simulation for each structural instance, forming a mapping dataset between design parameters and performance indices. The simulation methods include, but are not limited to, finite element analysis, geometric tolerance modeling, and Monte Carlo methods. The performance indices include, but are not limited to, structural mass, static stiffness, modal characteristics, thermal stability, and assembly accuracy.

[0025] Step S4.2: Preprocess the mapping dataset between design parameters and performance indicators to obtain the preprocessed mapping dataset between design parameters and performance indicators; More specifically, the dataset undergoes preprocessing to ensure the stability and consistency of subsequent modeling processes. This includes: identifying and removing outlier data; performing unified scale mapping on different physical quantities to transform them into dimensionless or unified scale representations; and standardizing or normalizing the data to eliminate the impact of dimensional differences. Furthermore, it includes: addressing parameter differences in different machine tool configurations by using configuration identification and parameter filtering methods to mask or constrain parameters unsuitable for the current configuration, enabling different types of machine tool structures to be expressed and modeled in a unified parameter space, thereby achieving unified processing of multi-configuration structures.

[0026] Step S4.3: Construct a performance prediction model. Use the preprocessed mapping dataset between design parameters and performance indicators to train the performance prediction model and obtain the trained performance prediction model.

[0027] More specifically, the performance prediction model takes design parameters as input and corresponding performance indicators as output. By training on sample data, it establishes a mapping relationship between design parameters and performance indicators, enabling the model to quickly output corresponding performance prediction results when given any set of design parameters. During the construction of the performance prediction model, the data is partitioned for training, validation, and testing to improve the model's generalization ability and prediction stability. In subsequent steps, the performance prediction model serves as a performance evaluation unit, coupled with the structure creation design model, to constrain and adjust the performance of the generated structure. The performance prediction model can be implemented in various ways, including but not limited to surrogate models based on neural networks and regression models based on statistical learning.

[0028] Step S4 further includes: based on the established performance prediction model, constructing a structural generative design model to generate corresponding design parameters according to the target performance index and design constraints. The structural generative design model takes the target performance index and design constraints as input and the design parameters as output, and is constructed through learning or mapping data samples to generate structural parameters from performance-geometric requirements. In this embodiment, the structural generative design model is coupled with the performance prediction model, so that the output of the structural generative design model can be input into the performance prediction model for performance evaluation, thereby forming a closed-loop structure. The process includes: the structural generative design model generating candidate design parameters according to the input target performance index; the performance prediction model evaluating the candidate design parameters to obtain corresponding performance prediction results; comparing the performance prediction results with the target performance index to obtain performance deviations; and performing feedback training and network parameter updates on the structural generative design model based on the performance deviations. The structural creation design model and its coupling with the performance prediction model can be implemented in various ways, including but not limited to: supervised learning, which establishes a mapping relationship between performance targets and design parameters and uses the performance prediction model for error feedback and constraints; reinforcement learning, which uses performance prediction results to construct evaluation indicators or reward signals to guide the structural creation design process; and hybrid mechanisms, which combine the above methods.

[0029] The present invention also provides an AI generative design agent for high-performance multi-axis machine tool structures. The AI ​​generative design agent for high-performance multi-axis machine tool structures can be implemented by executing the process steps of the implementation method of the AI ​​generative design agent for high-performance multi-axis machine tool structures. That is, those skilled in the art can understand the implementation method of the AI ​​generative design agent for high-performance multi-axis machine tool structures as a preferred embodiment of the AI ​​generative design agent for high-performance multi-axis machine tool structures.

[0030] This invention provides an AI generative design agent and its implementation method for high-performance multi-axis machine tool structures. Based on user-input performance indicators and design constraints, it can automatically and rapidly generate CAD models of multi-axis machine tool structure assemblies that meet performance requirements, enabling intelligent, systematic, and efficient generative design of function-driven multi-axis machine tool structures. It can be applied to scenarios such as reverse engineering optimization design of high-end CNC machine tools and rapid product modification design. Addressing the problem of difficulty in meeting multiple performance indicators in multi-axis machine tool reverse engineering design, this invention can quickly improve machine tool structure design to meet static, dynamic, thermal performance, and accuracy requirements. For different processing scenarios, including different part materials, geometric dimensions, and site limitations, where existing machine tools cannot directly meet engineering requirements, this invention can rapidly generate structural modification solutions based on existing machine tool product R&D experience, demonstrating significant application value and industrialization prospects.

[0031] Example 2 Example 2 is a preferred example of Example 1. According to the present invention, an AI generative design agent implementation method for high-performance multi-axis machine tool structures is provided, such as... Figures 2 to 4 As shown, it includes: This embodiment focuses on the structural design of a certain type of five-axis vertical machining center. The machine is required to simultaneously meet the requirements of high static stiffness, high first-order natural frequency, low thermal deformation, and high assembly accuracy within limited installation space and assembly boundaries. The overall structure is divided into a bed module, a column module, a crossbeam module, a ram mounting module, and local reinforcement modules. Each module is parametrically expressed, with design parameters including: total length of the bed. Total width of bed Column base width Column top width Column height There are 50 parameters, including the position of the guide rail mounting surface, the center position of the lead screw mounting, the connection position between the crossbeam and the column, the coordinates of the mounting hole system, and the relative position of the reference plane. These associated parameters are derived from the design parameters based on preset geometric relationships and engineering rules, and are used to ensure the geometric integrity and assembly consistency of the model during parameter changes. Figure 2As shown. Within a preset parameter range, design parameters are sampled, and Latin hypercube sampling is used to generate 3000 machine tool structure variants with different parameter combinations. For each structural instance, a CAD modeling program is automatically invoked to generate a corresponding 3D geometric model, and a CAE simulation program is further invoked to obtain performance indicators. The simulations include: static simulation, used to obtain the static displacement and static stiffness of key points; natural frequency simulation based on modal analysis, used to obtain the first and lower-order natural frequencies of the whole machine or key load-bearing structures; thermal deformation simulation based on steady-state thermal analysis, used to obtain the thermal drift of key locations under thermal load; and assembly accuracy evaluation based on geometric tolerance modeling and Monte Carlo analysis, used to obtain the pose error and error distribution of key assembly interfaces. This forms a mapping dataset between design parameters and performance indicators, where performance indicators include structural mass, static stiffness, first-order natural frequency, thermal deformation, and assembly accuracy.

[0032] First, invalid samples are identified in the obtained mapping dataset, including samples with failed geometric modeling, samples with entity self-intersection or interference, simulation divergence samples, and abnormal samples that significantly deviate from the reasonable engineering range, and these are removed. Then, different physical quantities are processed using a unified scale. Since mass, stiffness, frequency, thermal deformation, and assembly error vary greatly in magnitude, each performance index is standardized or normalized to eliminate the influence of dimensions. Furthermore, to achieve unified processing of machine tool data for different configurations, parameters are uniformly expressed. A unified parameter space template is defined for different machine types such as vertical machining centers, gantry machining centers, and horizontal coordinate boring machines. Parameters not involved in the current configuration are processed using masking tags, zero-filling, or constraint masks. For example, when a gantry machine tool has a beam span parameter that a vertical machining center does not, it can be distinguished by configuration labels and mask matrices, allowing different machine types to be expressed within the same data framework.

[0033] In this embodiment, the performance prediction model employs a feedforward neural network. The input layer contains all normalized design parameters; the intermediate hidden layers are set to three layers, with 128 neurons in each layer; the output layer corresponds to the five metrics from step 1. To improve generalization ability, batch normalization layers, residual connections, and Dropout mechanisms are introduced. The sample data is divided into a 70% training set and a 30% test set. During training, a weighted mean squared error is used as the loss function, with higher weights assigned to key metrics such as stiffness, frequency, and thermal deformation to ensure that the model focuses on learning high-performance-related features. Training ends after verifying on the test set whether the model's prediction errors for each performance metric are within an acceptable range.

[0034] In this embodiment, the input to the structural generative design model includes: target performance indicators, geometric boundary conditions, and aircraft type labels. The output is a set of candidate design parameters containing all design variables. This embodiment employs supervised learning coupled with a performance prediction model. The process involves the user inputting the target performance and design constraints; the structural generative design model generating a set of candidate parameters; the performance prediction model rapidly predicting the performance corresponding to these candidate parameters; comparing the prediction results with the target performance to obtain a performance deviation vector; and updating the structural generative design model parameters or iteratively correcting the candidate parameters based on the performance deviation until a design result that meets the requirements is obtained.

[0035] In this embodiment, the geometric dimensional range, spatial boundaries, mounting base, restricted areas, and assembly limitations in the current design task are uniformly encoded into constraint vectors. These constraint vectors may include: upper and lower limits for bed length, upper and lower limits for bed width, column installation positions, and the range of areas where crossbeams can be arranged. Then, the structural creation design model does not directly output actual dimensions, but rather outputs the relative proportions of each design parameter within its value range. For example, for the bed length parameter... The model outputs its normalized proportion. Then, based on the upper and lower bounds given by the current design constraints. and Mapped to actual size:

[0036] Other parameters are mapped in the same way.

[0037] In this embodiment, the structural generative design model is trained to gradually learn the mapping relationship between design parameters and performance targets under the combined effect of a performance prediction model and engineering constraints. After the model training or optimization is completed, the structural generative design model is applied to the actual machine tool structure design. Based on the target performance indicators and design constraints input by the user, the corresponding design parameters are directly generated, and a three-dimensional structural model is generated through a parametric machine tool structure model. The generated results can be further used for structural analysis, optimization design, or engineering manufacturing, realizing the automated conversion from performance requirements to structural design results.

[0038] This embodiment trains the structural generative design model through the combined action of a performance prediction model, a structural generative design model, and a dimensionless parameter mapping based on constraint fields. This allows the model to gradually master the generation rules from performance requirements and design constraints to structural parameters. After training, the model is applied to actual machine tool design tasks. In a specific design task, the user inputs the following requirement: the total mass of the key load-bearing structure should not exceed a preset value; The static stiffness in the main cutting direction is not lower than a preset value; the first natural frequency is not lower than a preset value; the thermal deformation is not higher than a preset value; and the given installation space boundary is satisfied. The system first generates constraint vectors and target performance vectors based on user input; then, candidate parameters are output from the structural creation design model; these are then converted into actual structural dimensions via dimensionless mapping; subsequently, performance is evaluated by the performance prediction model; finally, structural design parameters that meet performance and engineering constraints are obtained. Finally, this set of parametric CAD models is used to automatically generate a three-dimensional assembly model of the machine tool structure.

[0039] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0040] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for implementing an AI generative design intelligent agent for high-performance multi-axis machine tool structures, characterized in that, include: Step S1: Represent the geometric dimension range, spatial boundary, machining stroke and assembly constraints in the design space as a constraint vector, and parse the constraint vector into the upper and lower bounds of the corresponding design parameters; Step S2: The multiple preset performance indicators and the constraint vector are used together as input to the structural generative design model. The structural generative design model generates multiple sets of dimensionless candidate design parameters expressed in normalized proportions within the dimensionless design space constructed by the upper and lower bounds of the values. Step S3: Convert the multiple sets of dimensionless candidate parameters into actual candidate design parameters through mapping relationships; Step S4: Use the performance prediction model to evaluate the performance of multiple sets of actual candidate design parameters and obtain multiple sets of performance prediction results; Multiple sets of performance prediction results are compared with the corresponding preset performance indicators to obtain multiple sets of performance deviations; Step S5: Use multiple sets of performance deviations to perform feedback training and network parameter updates on the structural generative design model so that the design parameters can meet the corresponding design constraints and performance index requirements. Step S6: Input the target performance index and target constraint vector into the trained structure creation design model to obtain the corresponding dimensionless candidate parameters, and convert the obtained dimensionless candidate parameters into the final target actual design parameters.

2. The AI ​​generative design intelligent agent implementation method for high-performance multi-axis machine tool structures according to claim 1, characterized in that, The performance prediction model includes: Step S4.1: Perform parametric modeling for various types of multi-axis machine tool structures, generate multiple sets of machine tool structure instances through parameter sampling, and obtain corresponding performance index data for each structure instance through simulation, thereby forming a mapping dataset between design parameters and performance indexes; Step S4.2: Preprocess the mapping dataset between design parameters and performance indicators to obtain the preprocessed mapping dataset between design parameters and performance indicators; Step S4.3: Construct a performance prediction model. Use the preprocessed mapping dataset between design parameters and performance indicators to train the performance prediction model and obtain the trained performance prediction model.

3. The AI ​​generative design intelligent agent implementation method for high-performance multi-axis machine tool structures according to claim 2, characterized in that, Step S4.1 includes: Step S4.1.1: Perform layered modeling of the overall structure of multi-axis machine tools of various types, including: parametrically expressing the geometric dimensions, spatial positions and connection relationships of key components of each type of multi-axis machine tool that meet the preset requirements; Step S4.1.2: Establish the constraint relationships between parameters, including geometric constraints and assembly constraints, to ensure that the structure remains effective throughout the parameter changes; Step S4.1.3: Within the preset parameter range, generate multiple sets of machine tool structure instances through parameter sampling, and obtain corresponding performance index data for each structure instance through simulation to form a mapping dataset between design parameters and performance indexes; The simulation includes: performing simulation using any one or more methods, including finite element analysis, geometric tolerance modeling, and Monte Carlo simulation. The performance indicators include any one or more of the following: structural mass, static stiffness, modal characteristics, thermal stability, and assembly accuracy.

4. The AI ​​generative design intelligent agent implementation method for high-performance multi-axis machine tool structures according to claim 2, characterized in that, Step S4.2 includes: Step S4.2.1: Identify and remove outlier data in the mapping dataset between design parameters and performance indicators; Step S4.2.2: Perform unified scale mapping on different physical quantities in the mapped dataset after anomaly processing to transform them into dimensionless or unified scale representations; Step S4.2.3: Standardize or normalize the mapped dataset after uniform scale mapping to eliminate the impact of dimensional differences.

5. The AI ​​generative design intelligent agent implementation method for high-performance multi-axis machine tool structures according to claim 2, characterized in that, The performance prediction model includes: a surrogate model based on neural networks or a regression model based on statistical learning; The performance prediction model is based on any set of design parameters to obtain the corresponding performance prediction results.

6. An AI generative design agent for high-performance multi-axis machine tool structures, characterized in that, include: Module M1: Unifies the geometric dimension range, spatial boundary, machining stroke and assembly constraints in the design space into constraint vectors, and resolves the constraint vectors into the upper and lower bounds of the corresponding design parameters; Module M2: The structural generative design model takes multiple preset performance indicators and the constraint vector as inputs, and generates multiple sets of dimensionless candidate design parameters expressed in normalized proportions within the dimensionless design space constructed by the upper and lower bounds of the values. Module M3: Converts multiple sets of dimensionless candidate parameters into actual candidate design parameters through mapping relationships; Module M4: The performance prediction model is used to evaluate the performance of multiple sets of actual candidate design parameters and obtain multiple sets of performance prediction results. Multiple sets of performance prediction results are compared with the corresponding preset performance indicators to obtain multiple sets of performance deviations; Module M5: Uses multiple sets of performance deviations to perform feedback training and network parameter updates on the structural generative design model, so that the design parameters can meet the corresponding design constraints and performance index requirements; Module M6: Input the target performance index and target constraint vector into the trained structure creation design model to obtain the corresponding dimensionless candidate parameters, and convert the obtained dimensionless candidate parameters into the final target actual design parameters.

7. The AI ​​generative design agent for high-performance multi-axis machine tool structures according to claim 6, characterized in that, The performance prediction model includes: Module M4.1: Performs parametric modeling for various types of multi-axis machine tool structures. It generates multiple sets of machine tool structure instances through parameter sampling, and obtains corresponding performance index data for each structure instance through simulation, thereby forming a mapping dataset between design parameters and performance indicators. Module M4.2: Preprocesses the mapping dataset between design parameters and performance indicators to obtain a preprocessed mapping dataset between design parameters and performance indicators; Module M4.3: Construct a performance prediction model. Use the preprocessed mapping dataset between design parameters and performance metrics to train the performance prediction model and obtain the trained performance prediction model.

8. The AI ​​generative design agent for high-performance multi-axis machine tool structures according to claim 7, characterized in that, The module M4.1 includes: Module M4.1.1: Performs layered modeling of the overall structure of multi-axis machine tools of various types, including: parametrically expressing the geometric dimensions, spatial positions and connection relationships of key components of each type of multi-axis machine tool that meet preset requirements; Module M4.1.2: Establishes constraints between parameters, including geometric constraints and assembly constraints, to ensure that the structure remains effective during parameter changes; Module M4.1.3: Within the preset parameter range, multiple sets of machine tool structure instances are generated through parameter sampling, and the corresponding performance index data are obtained through simulation for each structure instance, forming a mapping dataset between design parameters and performance indexes; The simulation includes: performing simulation using any one or more methods, including finite element analysis, geometric tolerance modeling, and Monte Carlo simulation. The performance indicators include any one or more of the following: structural mass, static stiffness, modal characteristics, thermal stability, and assembly accuracy.

9. The AI ​​generative design agent for high-performance multi-axis machine tool structures according to claim 7, characterized in that, The module M4.2 includes: Module M4.2.1: Identifies and removes outlier data in the mapping dataset between design parameters and performance indicators; Module M4.2.2: Performs unified scale mapping on different physical quantities in the mapped dataset after anomaly handling, transforming them into dimensionless or unified scale representations; Module M4.2.3: Standardizes or normalizes the mapped dataset after uniform scale mapping to eliminate the impact of dimensional differences.

10. The AI ​​generative design agent for high-performance multi-axis machine tool structures according to claim 7, characterized in that, The performance prediction model includes: a surrogate model based on neural networks or a regression model based on statistical learning; The performance prediction model is based on any set of design parameters to obtain the corresponding performance prediction results.

Images