Abstract
Micro-force measurement devices are widely used in diverse fields such as pharmaceuticals, biomedicine, and precision manufacturing. However, existing systems suffer from limitations including low accuracy, poor linearity, structural complexity, and limited working range. This paper proposes a novel laser-based micro-force measurement system integrating multi-stage optical amplification with Transformer-based residual learning. The design employs a compact three-mirror structure that amplifies angular displacement through consecutive reflections, enabling micron-scale deformation detection under forces as low as 55 μN while maintaining a wide measurement range of 0-1 N. A hybrid algorithm combining grayscale centroid and Gaussian fitting achieves sub-pixel laser spot localization at 0.05 px precision. Further enhanced by a Transformer-based residual learning framework that processes temporal grayscale sequences, the system reduces steady-state noise by 19.9%. The system also exhibits excellent linearity, achieving a linear fit coefficient (R2) of 0.9997. Experimental validation through continuous water droplet experiments confirms the system’s real-time micro-force variation measurement capacity, validating its dynamic performance. The proposed system provides a compact, high-precision solution for applications such as pharmaceutical dosage control and nanomaterial mechanical testing.
Funding Information
  • Beijing Natural Science Foundation (F2024201068)
  • Beijing Natural Science Foundation (L243014)
  • Beijing Natural Science Foundation (L222138)
  • National Natural Science Foundation of China (62333023)
  • National Natural Science Foundation of China (62025307)
  • National Natural Science Foundation of China (U25A20475)
  • Aeronautical Science Foundation of China (2022Z0500M0001)
  • Fundamental Research Funds for the Central Universities

This publication has 18 references indexed in Scilit: