Abstract:

[Objective] To address the problems of time-consuming computation and low optimization iteration efficiency inherent in traditional Finite Element Analysis in the design of automotive air conditioning brackets, this paper proposes a new rapid simulation method for mechanical performance. [Methods] A high-fidelity dataset is constructed via parametric modeling and design of experiments. Based on the Deep Operator Network model, this method incorporates Fourier positional encoding, a Snake activation function, and an attention mechanism to establish a unified mapping framework for the ‘Parameter-Space-Full-Field-Indicator’ chain. Parameters are used to characterize design schemes and operating conditions, while spatial coordinates describe structural geometries and boundary conditions. Cross-modal alignment and full-field prediction are implemented within the same network, and key indicators including maximum stress and modal frequency are automatically extracted accordingly. [Results] Ablation experiments validate that the proposed model possesses high accuracy, high robustness, and high computational efficiency. Case studies show that this method can reduce the evaluation time of random vibration and modal analysis for automotive air conditioning brackets from hours to tens of seconds, while ensuring engineering-level precision. The relative prediction errors of the maximum stress and the first-order modal frequency both meet the enterprise’s overall precision requirement for predictive errors, with the allowable margin set at less than 5%. [Conclusions] The proposed method significantly improves the iteration efficiency of product structural design and enterprises’ optimal design capabilities

Key words: automotive air conditioning bracket, DeepONet, mechanical performance prediction, rapid simulation, optimal design