A prediction-optimization method for surface topography-induced assembly accuracy of a precision mirror

TY - JOUR

T1 - A prediction-optimization method for surface topography-induced assembly accuracy of a precision mirror

AU - Shen, Hongda

AU - Xia, Huanxiong

AU - Liu, Long

AU - Liu, Jianhua

N1 - Publisher Copyright: © 2026 Elsevier Inc.

PY - 2026/8

Y1 - 2026/8

N2 - Surface topography significantly influences the accuracy of precision assemblies; however, its complex, multi-scale geometric characteristics make it difficult to quantitatively establish the relationship between surface topography and assembly accuracy, thereby posing significant challenges for the active design and optimization of surface topography. This paper proposes a novel framework for the efficient prediction and optimization of assembly accuracy induced by surface topography. The framework comprises three procedures: surface topography representation, assembly accuracy prediction, and optimization, where a discrete cosine transform (DCT) method is used to decompose measured surface topographies, a genetic algorithm-backpropagation (GABP) network, trained on finite-element simulated data, predicts the assembly accuracy, and a genetic algorithm is finally implemented to optimize the DCT coefficients. Applied to a precision mirror, the framework achieves a maximum prediction error of 1.22% on the test set and 6.69% against experimental data. Moreover, the method that integrates the trained GABP network with the genetic algorithm successfully optimizes surface topography, reducing the RMS value by 49.27%, from 151.03 nm to 76.61 nm. This framework offers a robust tool for predicting and optimizing assembly accuracy, aiding rapid assessment, selective assembly, and surface-processing guidance in precision manufacturing.

AB - Surface topography significantly influences the accuracy of precision assemblies; however, its complex, multi-scale geometric characteristics make it difficult to quantitatively establish the relationship between surface topography and assembly accuracy, thereby posing significant challenges for the active design and optimization of surface topography. This paper proposes a novel framework for the efficient prediction and optimization of assembly accuracy induced by surface topography. The framework comprises three procedures: surface topography representation, assembly accuracy prediction, and optimization, where a discrete cosine transform (DCT) method is used to decompose measured surface topographies, a genetic algorithm-backpropagation (GABP) network, trained on finite-element simulated data, predicts the assembly accuracy, and a genetic algorithm is finally implemented to optimize the DCT coefficients. Applied to a precision mirror, the framework achieves a maximum prediction error of 1.22% on the test set and 6.69% against experimental data. Moreover, the method that integrates the trained GABP network with the genetic algorithm successfully optimizes surface topography, reducing the RMS value by 49.27%, from 151.03 nm to 76.61 nm. This framework offers a robust tool for predicting and optimizing assembly accuracy, aiding rapid assessment, selective assembly, and surface-processing guidance in precision manufacturing.

KW - Assembly accuracy

KW - Discrete cosine transform

KW - Optical mirror

KW - Precision assembly

KW - Surface topography

UR - https://www.scopus.com/pages/publications/105036034909

U2 - 10.1016/j.precisioneng.2026.04.011

DO - 10.1016/j.precisioneng.2026.04.011

M3 - Article

AN - SCOPUS:105036034909

SN - 0141-6359

VL - 101

SP - 16

EP - 26

JO - Precision Engineering

JF - Precision Engineering

ER -