Scholarlink

<p>In this study, the elastic modulus is obtained from the eigen frequencies of a cantilever beam. The beam’s elastic modulus is considered a random field with an exponential covariance function. Monte Carlo simulations (MCS) with Cholesky decomposition are employed to generate synthetic elastic modulus and eigen frequency data, which is treated as actual data. The data-driven non-intrusive polynomial chaos expansion (NIPCE) is utilized to estimate the elastic modulus from the eigen frequencies. The difference between the elastic modulus data obtained from MCS and NIPCE is used to train a neural network. For new predictions, the elastic modulus data from NIPCE and the difference predicted by the neural network are combined. The final results exhibit a remarkable resemblance to the actual data from MCS. Additionally, the computational time of the proposed method is significantly lower than that of MCS, as the trained neural network requires considerably less time to generate outputs.</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

Integrated Stochastic Modeling and Neural Networks for Inverse Prediction of Elastic Modulus in Beams

  • 期刊论文

  • 发表日期：

    2025年3月22日

查看全文
  • Rakesh Kumar

摘要

In this study, the elastic modulus is obtained from the eigen frequencies of a cantilever beam. The beam’s elastic modulus is considered a random field with an exponential covariance function. Monte Carlo simulations (MCS) with Cholesky decomposition are employed to generate synthetic elastic modulus and eigen frequency data, which is treated as actual data. The data-driven non-intrusive polynomial chaos expansion (NIPCE) is utilized to estimate the elastic modulus from the eigen frequencies. The difference between the elastic modulus data obtained from MCS and NIPCE is used to train a neural network. For new predictions, the elastic modulus data from NIPCE and the difference predicted by the neural network are combined. The final results exhibit a remarkable resemblance to the actual data from MCS. Additionally, the computational time of the proposed method is significantly lower than that of MCS, as the trained neural network requires considerably less time to generate outputs.