CN-121980812-A - Intelligent product design system based on deep learning

CN121980812ACN 121980812 ACN121980812 ACN 121980812ACN-121980812-A

Abstract

The invention relates to the technical field of product design, in particular to an intelligent product design system based on deep learning, which comprises an information analysis module, a product design module, a generation judgment module and a training adjustment module, wherein the information analysis module comprises an acquisition unit used for acquiring product design information and an analysis unit connected with the acquisition unit and used for analyzing the product design information to obtain design element characteristics, the product design module comprises a conversion unit used for converting a design target input by a user into constraint conditions, the generation judgment module is used for judging whether the generation accuracy of a product design scheme meets the requirements according to the constraint condition satisfaction rate, the sampling judgment module is used for judging whether the sampling temperature of a deep learning model meets the requirements according to the verification failure rate of the product design scheme, and the training adjustment module is used for determining the disturbance resistance intensity in the model training process according to the minimum characteristic disturbance amplitude judged by boundary conditions. The invention improves the generating accuracy of the product design scheme.

Inventors

CHEN HANFENG
ZHU QINGQIN
DONG YU

Assignees

义乌工商职业技术学院

Dates

Publication Date: 20260505
Application Date: 20260206

Claims (10)

1. An intelligent product design system based on deep learning, comprising: the information analysis module comprises an acquisition unit used for acquiring product design information, an analysis unit connected with the acquisition unit and used for analyzing the product design information to obtain design element characteristics, and a mapping unit connected with the analysis unit and used for constructing association constraint relations among the design element characteristics; the product design module is connected with the information analysis module and comprises a conversion unit for converting a design target input by a user into a constraint condition, a model training unit for training an initial model according to the design element characteristics and the association constraint relation to obtain a deep learning model, and a design unit connected with the model training unit for generating a product design scheme according to the constraint condition and the deep learning model; The generation judging module is connected with the product design module and is used for judging whether the generation accuracy of the product design scheme meets the requirement according to the constraint condition satisfaction rate; The sampling judgment module is respectively connected with the product design module and the generation judgment module and is used for judging whether the sampling temperature of the deep learning model meets the requirement according to the verification failure rate of the product design scheme; and the training adjustment module is respectively connected with the product design module and the sampling judgment module and is used for determining the disturbance resistance intensity in the model training process according to the minimum characteristic disturbance amplitude judged by the boundary condition.
2. The intelligent product design system based on deep learning of claim 1, wherein the generation determination module determines that the accuracy of the generation of the product design is satisfactory in response to the constraint satisfaction rate being greater than a preset satisfaction rate.
3. The deep learning based intelligent product design system of claim 2, wherein the generation determination module determines that the accuracy of the generation of the product design is not satisfactory in response to the constraint satisfaction rate being less than or equal to the preset satisfaction rate.
4. The intelligent product design system based on deep learning of claim 3, wherein the sampling decision module decides whether the sampling temperature of the deep learning model meets the requirement according to the verification failure rate of the product design in response to the condition that the generation accuracy of the product design is not met.
5. The deep learning based intelligent product design system of claim 4, wherein the sampling determination module determines that the sampling temperature of the deep learning model meets the requirement in response to the verification failure rate of the product design being less than or equal to a preset failure rate.
6. The deep learning based intelligent product design system of claim 5, wherein the sampling determination module determines that the sampling temperature of the deep learning model is unsatisfactory in response to the verification failure rate of the product design being greater than the preset failure rate.
7. The intelligent product design system based on deep learning of claim 6, wherein the training adjustment module determines whether the accuracy of the determination of the constraint condition boundary is satisfactory according to the minimum characteristic disturbance amplitude determined by the boundary condition in response to the condition that the sampling temperature of the deep learning model is not satisfactory.
8. The intelligent product design system based on deep learning of claim 7, wherein the training adjustment module determines that accuracy of determination of constraint boundaries is satisfactory in response to the minimum characteristic disturbance magnitude determined by the boundary conditions being greater than a preset disturbance magnitude.
9. The intelligent product design system based on deep learning of claim 8, wherein the training adjustment module determines that accuracy of discrimination of constraint boundaries is not satisfactory and increases the strength of the opposing disturbance during model training in response to the minimum characteristic disturbance magnitude determined by the boundary conditions being less than or equal to the preset disturbance magnitude.
10. The deep learning based intelligent product design system of claim 9, wherein the magnitude of the increase in the resistance to disturbance intensity during the model training process is determined by the difference between the preset disturbance magnitude and the minimum characteristic disturbance magnitude determined by the boundary condition.

Description

Intelligent product design system based on deep learning Technical Field The invention relates to the technical field of product design, in particular to an intelligent product design system based on deep learning. Background In the fields of intelligent manufacturing, consumer electronics and the like, an intelligent product design system based on deep learning has become a key technology for accelerating innovation period, optimizing product performance and realizing personalized customization. Because the product design involves the cooperation of multi-dimensional complex elements such as morphology, structure, materials, user experience and the like, the scheme quality is directly related to market competitiveness and production cost. When relying on traditional manual experience or parameterized CAD tools, creative divergence is insufficient due to cognition limitation or rule solidification of designers, or design schemes with better performance are omitted due to the fact that massive parameter combinations are difficult to exhaust, and design innovation and engineering feasibility are weakened. Although the existing intelligent design tool has realized basic modeling generation, a single model or a simple optimization algorithm is mostly adopted, and the dynamic balance capability among design diversity, engineering constraint and user preference is insufficient, so that the high-efficiency, collaborative and high-quality automatic design is difficult to realize in real product development. Therefore, the intelligent product design system based on deep learning needs to be capable of integrating a generated type countermeasure network, constraint satisfaction reasoning and a multi-objective optimization technology, and realizing intelligent exploration of creative space, autonomous satisfaction of cross-domain constraint and accurate mapping of user intention. The invention discloses a Chinese patent publication No. CN119378146A, which discloses an electromechanical product design method and system based on deep learning of a part database, wherein the method comprises the steps of S1, establishing the part database, wherein the part database comprises a plurality of part model index data, a plurality of part assembly schemes and a plurality of electromechanical product 3D models, when the types and the numbers of parts are determined, all the part model index data and any part assembly scheme are combined to correspond to a unique electromechanical product 3D model, S2, establishing an initial hybrid design optimization model, training and testing the initial hybrid design optimization model through the part database to obtain a hybrid design optimization model, S3, acquiring electromechanical product design requirements, acquiring a plurality of required part model index data corresponding to the electromechanical product design requirements based on the electromechanical product design requirements and a part assembly requirement analysis model, inputting the plurality of required part model index data into the hybrid design optimization model, outputting a plurality of electromechanical product 3D models corresponding to the electromechanical product design requirements, and S4, and producing the corresponding electromechanical product according to the electromechanical product 3D models. Therefore, the electromechanical product design method and system based on deep learning of the part database have the problem that the design requirement mapping has limitation to cause insufficient accuracy of the generation of the product design scheme because the hybrid design optimization model training depends on the fixed database. Disclosure of Invention Therefore, the invention provides an intelligent product design system based on deep learning, which is used for solving the problem that the generation accuracy of a product design scheme is insufficient due to the limitation of design demand mapping because the training of a hybrid design optimization model depends on a fixed database in the prior art. In order to achieve the above object, the present invention provides an intelligent product design system based on deep learning, comprising: the information analysis module comprises an acquisition unit used for acquiring product design information, an analysis unit connected with the acquisition unit and used for analyzing the product design information to obtain design element characteristics, and a mapping unit connected with the analysis unit and used for constructing association constraint relations among the design element characteristics; the product design module is connected with the information analysis module and comprises a conversion unit for converting a design target input by a user into a constraint condition, a model training unit for training an initial model according to the design element characteristics and the association constraint relation to obtain a deep learning model, and a