Scholarlink

Abstract <p>Clean-in-place (CIP) is a typical industrial process used to clean equipment and pipes to maintain hygiene, prevent contamination, and ensure that equipment operates optimally. However, this process uses significant amounts of water, increasing resource usage and environmental impact. This study aims to enhance CIP sustainability by minimizing water consumption through experimental data, machine learning, and optimization techniques. Experimental data was collected using a laboratory-scale clean-in-place system, which circulated a NaOH (sodium hydroxide) alkaline detergent solution during the rinsing step, utilizing pulsed and variable setpoints. A Long Short-Term Memory Network (LSTM) was employed to model the process, and the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) was used to determine the optimal flow rates and time intervals for reducing water usage. The results demonstrate that the LSTM-NSGA-II approach effectively minimizes water consumption in response to turbulence generated by the flow rate setpoint variations. The obtained optimal solution achieved a Root Mean Square Error (RMSE) of 0.07004 (normalized conductivity value) and water consumption of 14.20 L, with a 4.5% deviation from algorithm estimates, and was validated by experimental results. Compared to the traditional constant flow rate method, our approach reduced water consumption by 77.38% and rinsing time by 43.36%. Implementing this method could lead to substantial improvements in water resource usage and sustainability in the industry.</p> Graphical abstract <p></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

Enhancing clean-in-place system sustainability with recurrent neural network and multi-objective optimization techniques

  • 期刊论文

  • 发表日期：

    2025年11月29日

查看全文
  • R. Sislian,
  • P. R. P. Allegro,
  • R. Gedraite,
  • G. O. Meloni,
  • P. R. S. Oliveira,
  • F. V. Silva,
  • D. L. Toufen

摘要

Abstract

Clean-in-place (CIP) is a typical industrial process used to clean equipment and pipes to maintain hygiene, prevent contamination, and ensure that equipment operates optimally. However, this process uses significant amounts of water, increasing resource usage and environmental impact. This study aims to enhance CIP sustainability by minimizing water consumption through experimental data, machine learning, and optimization techniques. Experimental data was collected using a laboratory-scale clean-in-place system, which circulated a NaOH (sodium hydroxide) alkaline detergent solution during the rinsing step, utilizing pulsed and variable setpoints. A Long Short-Term Memory Network (LSTM) was employed to model the process, and the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) was used to determine the optimal flow rates and time intervals for reducing water usage. The results demonstrate that the LSTM-NSGA-II approach effectively minimizes water consumption in response to turbulence generated by the flow rate setpoint variations. The obtained optimal solution achieved a Root Mean Square Error (RMSE) of 0.07004 (normalized conductivity value) and water consumption of 14.20 L, with a 4.5% deviation from algorithm estimates, and was validated by experimental results. Compared to the traditional constant flow rate method, our approach reduced water consumption by 77.38% and rinsing time by 43.36%. Implementing this method could lead to substantial improvements in water resource usage and sustainability in the industry.

Graphical abstract