A Deep Learning-Driven Mobile Interaction System for Industrial Process Design

An Automotive Industry Case Study

Abstract

To address the demands of intelligent transformation in the automotive industry under the paradigm of Industry 4.0 and to overcome limitations associated with experience-dependent process design, prolonged iteration cycles, and insufficient mobile interaction capabilities, a deep learning–driven mobile interaction system based on cloud–edge–end collaboration was proposed for automotive process design scenarios. A hierarchical collaborative architecture is established, in which the cloud layer provides knowledge support and computational assurance through a process knowledge graph and data augmentation based on artificial intelligence generated content (AIGC), the edge layer achieves a balance between responsiveness and efficiency via dynamic model scheduling and multi-terminal collaboration, and the mobile layer enhances interaction quality through adaptive rendering and multimodal fusion techniques. To address key technical bottlenecks, a two-stage lightweight model generation framework, an on-device multimodal scene understanding method, and a realtime simulation optimization scheme driven by Physics-informed neural networks (PINNs) were designed, forming a closed-loop mobile design interaction workflow. The effectiveness of the proposed system was validated using the joining process of automotive door outer panels as a representative case, supported by multi-dimensional comparative experiments and ablation studies. Through the deep integration of deep learning and industrial mobile interaction, the proposed approach enables a paradigm shift in automotive process design from tool-assisted operation toward intelligent co-creation, providing both technical support and practical guidance for mobile intelligent design in industrial applications.