Scholarlink

<p>High-pressure sectors like mining and construction require multi-stage screw compressors that can operate reliably at pressures over 16 bar. Single-stage compressors frequently encounter constraints such as elevated temperatures, rotor bending deformation, imperfect cooling effect of the injected oil, condensate, and diminished bearing longevity, rendering them inadequate for these specifications. This paper introduces a comprehensive modelling and optimisation approach for multi-stage screw compressors, integrating a physics-based chamber model with machine learning via Gaussian process regression. The framework employs Bayesian optimisation to methodically refine stage-specific parameters, enhancing performance and dependability while ensuring computing economy. The innovation is in its capacity to precisely forecast the performance of both individual and final stages, experimentally validated with a two-stage air screw compressor for water-well applications, attaining an error margin below 5%. A case study illustrated the framework’s efficacy by decreasing specific power usage by 2% via the optimisation of fluid injection parameters. This approach represents a significant advancement in compressor technology, providing a scalable and efficient solution for designing and optimising multi-stage screw compressors in high-pressure applications.</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 modelling and optimisation framework for multi-stage screw compressors utilising Gaussian process regression and Bayesian methods

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月2日

查看全文
  • Abhishek Kumar,
  • Ahmed Kovacevic,
  • Sathiskumar Anusuya Ponnusami

摘要

High-pressure sectors like mining and construction require multi-stage screw compressors that can operate reliably at pressures over 16 bar. Single-stage compressors frequently encounter constraints such as elevated temperatures, rotor bending deformation, imperfect cooling effect of the injected oil, condensate, and diminished bearing longevity, rendering them inadequate for these specifications. This paper introduces a comprehensive modelling and optimisation approach for multi-stage screw compressors, integrating a physics-based chamber model with machine learning via Gaussian process regression. The framework employs Bayesian optimisation to methodically refine stage-specific parameters, enhancing performance and dependability while ensuring computing economy. The innovation is in its capacity to precisely forecast the performance of both individual and final stages, experimentally validated with a two-stage air screw compressor for water-well applications, attaining an error margin below 5%. A case study illustrated the framework’s efficacy by decreasing specific power usage by 2% via the optimisation of fluid injection parameters. This approach represents a significant advancement in compressor technology, providing a scalable and efficient solution for designing and optimising multi-stage screw compressors in high-pressure applications.