Scholarlink

<p>A data-driven artificial intelligence framework is developed to predict and optimize the performance of nanophotonics-enabled solar membrane distillation reactor for water and wastewater treatment. An artificial neural network (ANN) is constructed to model the nonlinear relationships between operating conditions and system performance. The ANN model considers five input parameters: feed temperature, coolant temperature, solar irradiance, feed flow rate, and sweeping air humidity, and predicts three key performance indicators: permeate flux, gain-to-output ratio (GOR), and heat loss. To address the challenge of ANN architecture selection, a genetic algorithm (GA) is employed to systematically optimize the network architecture: hidden layers and neurons. Unlike traditional ANN-based membrane distillation models that rely on manual tuning, the proposed GA-optimized framework provides a computationally efficient and globally optimal approach for ANN architecture optimization. The constructed ANN model using GA demonstrates high accuracy without overfitting, achieving coefficients of determination values approaching 0.99, mean squared error below <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(1\times {10}^{-3}\)</EquationSource> </InlineEquation>, and average relative error under 6%. Single-objective GA optimization is applied to determine the optimal operating conditions for maximizing flux, minimizing heat loss, and maximizing GOR. To account for trade-offs among these performance indicators, multi-objective optimization is conducted using the nondominated sorting genetic algorithm II. The optimization results identify practical operating ranges, with a maximum flux of <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(1.38-1.54 ( \text{kg}/{\text{m}}^{2}/\text{h})\)</EquationSource> </InlineEquation>, heat losses ranging from <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(32.47\)</EquationSource> </InlineEquation> to <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(33.38\%\)</EquationSource> </InlineEquation>, and a GOR reaching 0.944–1.24. The algorithm is validated against published experimental data and demonstrates superior predictive accuracy over trial-and-error ANN models, confirming its robustness and applicability for membrane distillation optimization.</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

Data-driven modeling and optimization of nanophotonics-enabled solar membrane distillation (NESMD) for water and wastewater treatment

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月6日

查看全文
  • Mayar Elrakhawi,
  • Amr Abdelrazek,
  • Ibrahim A. Said,
  • Ahmed F. Tayel

摘要

A data-driven artificial intelligence framework is developed to predict and optimize the performance of nanophotonics-enabled solar membrane distillation reactor for water and wastewater treatment. An artificial neural network (ANN) is constructed to model the nonlinear relationships between operating conditions and system performance. The ANN model considers five input parameters: feed temperature, coolant temperature, solar irradiance, feed flow rate, and sweeping air humidity, and predicts three key performance indicators: permeate flux, gain-to-output ratio (GOR), and heat loss. To address the challenge of ANN architecture selection, a genetic algorithm (GA) is employed to systematically optimize the network architecture: hidden layers and neurons. Unlike traditional ANN-based membrane distillation models that rely on manual tuning, the proposed GA-optimized framework provides a computationally efficient and globally optimal approach for ANN architecture optimization. The constructed ANN model using GA demonstrates high accuracy without overfitting, achieving coefficients of determination values approaching 0.99, mean squared error below \(1\times {10}^{-3}\) , and average relative error under 6%. Single-objective GA optimization is applied to determine the optimal operating conditions for maximizing flux, minimizing heat loss, and maximizing GOR. To account for trade-offs among these performance indicators, multi-objective optimization is conducted using the nondominated sorting genetic algorithm II. The optimization results identify practical operating ranges, with a maximum flux of \(1.38-1.54 ( \text{kg}/{\text{m}}^{2}/\text{h})\) , heat losses ranging from \(32.47\) to \(33.38\%\) , and a GOR reaching 0.944–1.24. The algorithm is validated against published experimental data and demonstrates superior predictive accuracy over trial-and-error ANN models, confirming its robustness and applicability for membrane distillation optimization.