Scholarlink

<p>To address the difficulty of obtaining Guide to the Expression of Uncertainty in Measurement (GUM)-based measurement uncertainty for sensors employing artificial neural networks (ANNs) in the field of mechanical measurement, this paper proposes and demonstrates a GUM-based uncertainty evaluation method for a nonlinear optical angle sensor that uses a multilayer perceptron artificial neural network (MLP-ANN). The proposed sensor decodes the angular information of a target by inputting a measured second harmonic generation spectrum into a trained MLP-ANN to enable wide-range angle measurement. However, the measurement uncertainty of this method has not yet been evaluated, and such an analysis is essential to assess the accuracy and reliability of the sensor. This paper demonstrates the measurement uncertainty evaluation of the nonlinear optical angle sensor using a specifically trained MLP-ANN and thereby confirms the feasibility of the proposed GUM-based method. Mathematical derivations are provided to obtain the sensitivity coefficients required for uncertainty estimation. Notably, this derivation is one of the most challenging aspects when applying GUM to ANN-based sensors. The uncertainty of the reference instrument used to generate training labels is also considered by incorporating the effects of the ANN training. The proposed method reveals that the measurement uncertainty of the MLP-ANN estimation ranges from 0.555 arcseconds to 30.264 arcseconds across different measurement regions. A detailed discussion of the results is also provided.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GUM-Based Measurement Uncertainty Analysis of a Nonlinear Optical Angle Sensor Using Artificial Neural Network

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月7日

查看全文
  • Jiahui Lin,
  • Zhiyang Zhang,
  • Ryo Sato,
  • Kuangyi Li,
  • Yasuhiro Mizutani,
  • Hiraku Matsukuma,
  • Wei Gao,
  • Robert X. Gao

摘要

To address the difficulty of obtaining Guide to the Expression of Uncertainty in Measurement (GUM)-based measurement uncertainty for sensors employing artificial neural networks (ANNs) in the field of mechanical measurement, this paper proposes and demonstrates a GUM-based uncertainty evaluation method for a nonlinear optical angle sensor that uses a multilayer perceptron artificial neural network (MLP-ANN). The proposed sensor decodes the angular information of a target by inputting a measured second harmonic generation spectrum into a trained MLP-ANN to enable wide-range angle measurement. However, the measurement uncertainty of this method has not yet been evaluated, and such an analysis is essential to assess the accuracy and reliability of the sensor. This paper demonstrates the measurement uncertainty evaluation of the nonlinear optical angle sensor using a specifically trained MLP-ANN and thereby confirms the feasibility of the proposed GUM-based method. Mathematical derivations are provided to obtain the sensitivity coefficients required for uncertainty estimation. Notably, this derivation is one of the most challenging aspects when applying GUM to ANN-based sensors. The uncertainty of the reference instrument used to generate training labels is also considered by incorporating the effects of the ANN training. The proposed method reveals that the measurement uncertainty of the MLP-ANN estimation ranges from 0.555 arcseconds to 30.264 arcseconds across different measurement regions. A detailed discussion of the results is also provided.