Scholarlink

<p>This study examined the accuracy of linearized and nonlinearized forms of kinetics and isotherm models in fitting methylene blue (MB) adsorption by waste-derived biochar. The biochars were effective at MB removal, achieving adsorption capacities of 4.15–34.39&#xa0;mg/g. The best fitting model was assessed using determination coefficient (R<sup>2</sup>) and four error functions. Nonlinearized models provided a better data fit, showing higher determination coefficients (R<sup>2</sup>) of 0.86–0.999 compared to linearized models (0.229–0.988) and lower errors (9.57–36% versus 15.75–48.5%). The use of linearized forms should be avoided since modern common software readily supports nonlinear fitting. Additionally, a regression tree model was developed using machine learning to identify key factors influencing MB adsorption and offer accurate estimations of MB adsorption. Regression tree modelling exhibited excellent predictive capability (R<sup>2</sup> = 0.99). Using feature importance analysis, the strongest predictors of adsorption capacity were initial concentration &gt; carbon and nitrogen contents &gt; adsorber pH &gt; contact time. Regression tree modelling can capture process parameters and adsorbent characteristics into an easy-to-use model which can be used in process operations and optimization. The study revealed that treating of 1 m<sup>3</sup>&#xa0;of dye-contaminated wastewater cost was estimated at AUD $27–230. Biochars reusability for 3 cycles was evaluated, noting a significant reduction in effectiveness (<i>p</i> &lt;  &lt; 0.001). Despite the observed decrease in adsorption capacity, waste-derived biochars continue to offer a cost-effective, environmentally sustainable solution aligning with the concept of "treating waste with waste”. The study highlights potential of using non-conventional materials to reduce the environmental impacts and cost of wastewater treatment, alongside the benefits of machine learning for process 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

Nonlinear modeling and machine learning techniques are needed for accurate prediction of contaminant sorption

  • 期刊论文

  • 发表日期：

    2025年1月5日

查看全文
  • Z. Mahdi,
  • A. E. Hanandeh,
  • C. Pratt

摘要

This study examined the accuracy of linearized and nonlinearized forms of kinetics and isotherm models in fitting methylene blue (MB) adsorption by waste-derived biochar. The biochars were effective at MB removal, achieving adsorption capacities of 4.15–34.39 mg/g. The best fitting model was assessed using determination coefficient (R2) and four error functions. Nonlinearized models provided a better data fit, showing higher determination coefficients (R2) of 0.86–0.999 compared to linearized models (0.229–0.988) and lower errors (9.57–36% versus 15.75–48.5%). The use of linearized forms should be avoided since modern common software readily supports nonlinear fitting. Additionally, a regression tree model was developed using machine learning to identify key factors influencing MB adsorption and offer accurate estimations of MB adsorption. Regression tree modelling exhibited excellent predictive capability (R2 = 0.99). Using feature importance analysis, the strongest predictors of adsorption capacity were initial concentration > carbon and nitrogen contents > adsorber pH > contact time. Regression tree modelling can capture process parameters and adsorbent characteristics into an easy-to-use model which can be used in process operations and optimization. The study revealed that treating of 1 m3 of dye-contaminated wastewater cost was estimated at AUD $27–230. Biochars reusability for 3 cycles was evaluated, noting a significant reduction in effectiveness (p <  < 0.001). Despite the observed decrease in adsorption capacity, waste-derived biochars continue to offer a cost-effective, environmentally sustainable solution aligning with the concept of "treating waste with waste”. The study highlights potential of using non-conventional materials to reduce the environmental impacts and cost of wastewater treatment, alongside the benefits of machine learning for process optimization.