Welding process parameter inference method and device based on saint-mlp fusion model and differential evolution, and medium

By combining the SAINT-MLP fusion model with the differential evolution algorithm, the problems of high cost and limited generalization ability of traditional welding process parameter optimization are solved, realizing efficient and accurate welding process parameter inference, and improving welding quality and efficiency.

CN121502243BActive Publication Date: 2026-06-09HEBEI UNIV OF SCI & TECH
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HEBEI UNIV OF SCI & TECH
Filing Date
2026-01-14
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Traditional welding process parameter optimization relies on a large number of experiments or finite element simulations, which are computationally expensive and difficult to perform reverse calculations for specific weld morphologies. Conventional models are not sensitive to small sample data and have limited generalization ability.

Method used

By employing the SAINT-MLP fusion model and differential evolution algorithm, welding feature information is extracted from the BIM model of the welded parts. The welding parameters are then inferred using the SAINT-MLP cascade model and differential evolution algorithm, and a weld morphology prediction model is established to achieve efficient reasoning of process parameters.

Benefits of technology

It enables high-precision welding process parameter inference under small sample data conditions, improving the reliability and efficiency of welding quality and reducing computational costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121502243B_ABST
    Figure CN121502243B_ABST
Patent Text Reader

Abstract

The application provides a welding process parameter inference method and device based on a SAINT-MLP fusion model and differential evolution, and a medium, first, a BIM model of a to-be-welded piece is acquired; then, welding feature information in the BIM model is extracted; finally, the welding feature information is input into a pre-established welding parameter inference model to obtain recommended process parameters; wherein the welding parameter inference model is obtained by backstepping according to a weld appearance prediction model; and the weld appearance prediction model is a SAINT-MLP series connection model. The application connects the SAINT model and the MLP in series, simultaneously uses a differential evolution algorithm to search for parameters, thereby efficiently learning the mapping relationship between the welding parameters and the weld appearance, and realizing the backstepping of the process parameters under the condition of a given target appearance.
Need to check novelty before this filing date? Find Prior Art