Scholarlink

<p>Hydraulic systems are increasingly used in machinery and industrial applications owing to their inherent advantages. In general, monitoring the dynamic responses of hydraulic pumps is an effective approach to maintaining the technical condition of hydraulic systems during operation. In the present study, we investigate dynamic responses of an axial piston pump with uncertainties in geometrical and operational parameters using Karhunen–Loève (KL) expansion and neural-network (NN) surrogates. Based on the established dynamic model, the Runge–Kutta (RK) numerical method is used to determine the pump characteristics, serving as input data for constructing surrogate models. Then, the dynamic characteristics of the pump system are analyzed considering the uncertainties of the system parameters via multiscale-informed approaches. The findings indicate that the reconstructed fifth-order KL decomposition and four-layer neural network exhibit excellent predictive accuracy in capturing the dynamic behavior of the pump with a relative error threshold of <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(2\times 10^{-3}\)</EquationSource> </InlineEquation>, including the non-smooth characteristics observed during transition periods. The reconstructed NN surrogate for predicting pump characteristics demonstrates stability in the number of computational iterations and the accuracy of its predictive capability. Furthermore, in terms of uncertainty quantification with two survey scenarios, the statistical behavior of the pump performance characteristics tends to follow a normal distribution, regardless of whether the system input parameters are normally or uniformly distributed.</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

A combined reduced-order and neural-network approach to uncertainty propagation in dynamic response of axial piston hydraulic pump

  • 期刊论文

  • 发表日期：

    2026年2月11日

查看全文
  • Van-Hai Trinh

摘要

Hydraulic systems are increasingly used in machinery and industrial applications owing to their inherent advantages. In general, monitoring the dynamic responses of hydraulic pumps is an effective approach to maintaining the technical condition of hydraulic systems during operation. In the present study, we investigate dynamic responses of an axial piston pump with uncertainties in geometrical and operational parameters using Karhunen–Loève (KL) expansion and neural-network (NN) surrogates. Based on the established dynamic model, the Runge–Kutta (RK) numerical method is used to determine the pump characteristics, serving as input data for constructing surrogate models. Then, the dynamic characteristics of the pump system are analyzed considering the uncertainties of the system parameters via multiscale-informed approaches. The findings indicate that the reconstructed fifth-order KL decomposition and four-layer neural network exhibit excellent predictive accuracy in capturing the dynamic behavior of the pump with a relative error threshold of \(2\times 10^{-3}\) , including the non-smooth characteristics observed during transition periods. The reconstructed NN surrogate for predicting pump characteristics demonstrates stability in the number of computational iterations and the accuracy of its predictive capability. Furthermore, in terms of uncertainty quantification with two survey scenarios, the statistical behavior of the pump performance characteristics tends to follow a normal distribution, regardless of whether the system input parameters are normally or uniformly distributed.