Micro-Vision Based High-Precision Space Assembly Approach for Trans-Scale Micro-Device: The CFTA Example

Juan Zhang, Xi Dai, Wenrong Wu, Kai DuSensors 2023 · DOI

TL;DR

This paper presents a novel micro-vision-based approach for high-precision assembly of "trans-scale" micro-devices, specifically focusing on the Capsule Fill Tube Assembly (CFTA) used in inertial confinement fusion (ICF) targets. The method tackles the challenge of assembling components with vastly different sizes (microns to hundreds of microns) by integrating advanced posture alignment, insertion depth control, and quantitative glue mass control techniques.

What Problem Does This Paper Address?

Assembling micro-devices, especially those with components of significantly different scales (trans-scale), presents a unique set of challenges. Traditional micro-assembly often struggles with the trade-off between high precision and a wide field of view (FOV). For trans-scale assembly, this means that while precise manipulation is needed for micron-sized parts, observing larger components and maintaining them within the FOV during adjustments is difficult. Furthermore, specific issues arise in the CFTA assembly:

  • Posture Alignment: Aligning a micro-tube (tens of microns) with a micro-sphere (hundreds of microns) is complicated by the size disparity and the need for precise 6-DOF control, which is less explored than 3-DOF.
  • Insertion Depth Control: During insertion, the micro-tube can occlude the micro-pore on the micro-sphere, making real-time depth detection and control challenging, especially with transparent materials.
  • Glue Mass Control: Precisely dispensing the correct amount of glue (in the picoliter range) for bonding is critical but difficult to achieve quantitatively due to the complex interplay of dispensing parameters.

Methods

The proposed approach is a three-module system:

  1. Posture Alignment Module: This module utilizes a multi-vision monitoring model with four microscopic vision systems. Two systems provide a lower magnification, wider FOV for global posture detection (horizontal alignment), while two others offer higher magnification for detailed local alignment (60-degree view). This allows for 5-DOF posture alignment (attitude and position) by combining coarse and fine adjustment strategies.
  2. Insertion Depth Control Module: This module employs a novel "local deformation detection" (LDD) method. When visual occlusion occurs, it infers insertion depth by monitoring subtle grayscale variations at the micro-tube's contact point with the micro-pore. This allows for precise depth control even when the micro-tube end is not directly visible.
  3. Glue Mass Control Module: This module relies on simulation (using ANSYS finite element simulation with a VOF model) to analyze the influence of various dispensing parameters (pressure, needle diameter) on glue mass. Based on these simulations, a time-pressure dispensing mode is used, with the dispensing needle diameter being a key controllable parameter for achieving quantitative glue mass.

Datasets

Not discussed in the paper. The experiments were conducted using a custom-built micro-assembly robot and the CFTA components.

Key Insights and Ideas

  • Multi-Vision for Trans-Scale: The strategic use of multiple cameras with varying magnifications and angles is crucial for effectively monitoring and aligning components of vastly different sizes. This overcomes the inherent FOV vs. precision trade-off.
  • LDD for Occlusion: The Local Deformation Detection (LDD) method offers an ingenious solution for depth control in visually occluded scenarios, a common problem in micro-assembly. It leverages subtle physical cues rather than direct visual observation.
  • Simulation-Driven Glue Control: Using finite element simulation to understand and predict glue dispensing behavior is a powerful approach to achieve quantitative control over a notoriously difficult parameter like glue mass at the picoliter scale.
  • Integrated Approach: The success hinges on the synergistic integration of posture, depth, and glue control, demonstrating that a holistic approach is necessary for complex micro-assembly tasks.

Key Results

The paper reports impressive experimental results:

  • Posture Alignment: Attitude alignment error is less than ±0.3°, and position alignment control error is less than ±5 µm.
  • Insertion Depth Control: The insertion depth control error is less than ±5 µm.
  • Glue Mass Control: For a 10-µm needle at 0 hPa pressure, the mean glue spot diameter is 62 µm with a deviation of less than 15 µm, indicating good uniformity. The study also shows that different needle diameters can achieve different glue spot sizes.
  • CFTA Assembly: A CFTA was successfully assembled with a position error in 3D space (measured by CT) of less than 5 µm and a glue spot diameter of 56 µm. The estimated glue mass was approximately 23 picoliters.

Limitations and Future Work

  • Limitations:
    • The paper focuses on a specific assembly task (CFTA) and may require adaptation for other trans-scale micro-devices.
  • While the simulation provides insights into glue mass control, real-world manufacturing variations in micro-parts (micro-tube, micro-pore, dispensing needle) can still affect diameter consistency.
  • The assembly process is described as semi-automatic, implying some level of human intervention or supervision.
  • Future Work:
    • The authors suggest that the proposed methods will be packaged to apply to other trans-scale micro-device assemblies, indicating a broader applicability.
  • Further improvements in manufacturing accuracy of micro-parts could further reduce diameter deviation in glue dispensing.
  • Exploring fully automated assembly processes could be a future direction.

Summary updated

BibTeX
@article{zhang2023micro,
  author = {Juan Zhang and Xi Dai and Wenrong Wu and Kai Du},
  title = {{Micro-Vision Based High-Precision Space Assembly Approach for Trans-Scale Micro-Device: The CFTA Example}},
  journal = {Sensors},
  year = {2023},
  doi = {10.3390/s23010450},
}
Figure