Year: 2026
Type: article
Abstract: Post-weld shift (PWS) significantly degrades the coupling efficiency and consistency of butterfly semiconductor lasers during automated packaging. This study proposes a novel pre-compensation method based on a Particle Swarm Optimization–Backpropagation (PSO-BP) neural network to mitigate these effects. Finite element simulations and experimental analyses reveal that PWS consistently presented negative Y-direction displacement, driven by localized thermal contraction and residual stress. Results indicate that PWS magnitude is highly sensitive to laser energy and welding joint layout. By integrating PSO, the neural network achieved global convergence in just 25 iterations—15 fewer than the conventional BP model—with superior prediction accuracy. Implementation of this pre-compensation framework into an automated packaging system reduced the welding offset range from 0.5–4 μm to 0.5–2 μm, yielding a 26% average increase in optical coupling efficiency. This research provided a high-performance predictive framework suitable for the high-precision mass production of optoelectronic devices.
Source: Photonics
Authors: Jinwen Peng, Lian Duan, Ji-An Duan
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,896
Awards: G7485366771, G8377400998
OpenAlex ID: W7166724730