Scholarlink

<p>The importance of water supply in today’s advanced industrial world and the lack of sufficient water sources and the lack of replacement have led to the emergence of various methods to improve water supply, of which water desalination is one of the most common methods. Capacitive Deionization (CDI) is an emerging method to remove salt from water. Due to its low energy and low environmental pollution, it has been very popular in the last few decades. In this research, two methods of Response Surface Methodology (RSM) and Adaptive Neuro-Fuzzy Inference System (ANFIS) have been used to investigate the performance of the CDI method of water. Input variables include cell voltage value, initial concentration value, and flow rate, while output variables include Salt Removal Percentage (SRP). A single-pass (SP) fluid flow method is used by CDI for desalination tests. A Central Composite Design (CCD) was used to plan the Design of Experiments (DoE). The Analysis of Variance (ANOVA) method is used to investigate the effect of variables and the effects of three input variables on the response variable are significant. RMSE, MSE, R, R<sup>2</sup>Adj, and R<sup>2</sup> indices are used to evaluate the performance of the two methods. The Sugeno fuzzy method is used in ANFIS modeling with five layers and three inputs and one output. Membership functions in ANFIS modeling are Gaussian functions. The values of RMSE and MSE for the RSM model are 0.2209 and 0.470 respectively, whereas these values are 0.0033 and 0.0575 for ANFIS modeling. This fact indicates that both methods have an acceptable prediction for the CDI method. In the statistical analysis, it is found that both methods are reliable and suitable for estimating the CDI method in the range of input variables. In the statistical and graphical comparison of the two models, it is found that the ANFIS model is more suitable for estimating and predicting the RSP of water. Moreover, in the analysis of the variance of the RSM model, the interaction effect of the cell voltage variable and the initial concentration value and the interaction effect of the fluid flow rate and the initial concentration value on the RSP are significant. The results show that presented models can predict the outcome of CDI reliably. Therefore, this can save time and cost by avoiding performing physical experiments to achieve similar results.</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

Comparison Study of ANFIS and RSM Models for Predicting of Water Production by Capacitive Deionization Method

  • 期刊论文

  • 发表日期：

    2026年4月21日

查看全文
  • H. Etezadi,
  • P. Zarafshan,
  • M. Khashehchi,
  • E. Pipelzadeh,
  • G. Fang,
  • K. Salari

摘要

The importance of water supply in today’s advanced industrial world and the lack of sufficient water sources and the lack of replacement have led to the emergence of various methods to improve water supply, of which water desalination is one of the most common methods. Capacitive Deionization (CDI) is an emerging method to remove salt from water. Due to its low energy and low environmental pollution, it has been very popular in the last few decades. In this research, two methods of Response Surface Methodology (RSM) and Adaptive Neuro-Fuzzy Inference System (ANFIS) have been used to investigate the performance of the CDI method of water. Input variables include cell voltage value, initial concentration value, and flow rate, while output variables include Salt Removal Percentage (SRP). A single-pass (SP) fluid flow method is used by CDI for desalination tests. A Central Composite Design (CCD) was used to plan the Design of Experiments (DoE). The Analysis of Variance (ANOVA) method is used to investigate the effect of variables and the effects of three input variables on the response variable are significant. RMSE, MSE, R, R2Adj, and R2 indices are used to evaluate the performance of the two methods. The Sugeno fuzzy method is used in ANFIS modeling with five layers and three inputs and one output. Membership functions in ANFIS modeling are Gaussian functions. The values of RMSE and MSE for the RSM model are 0.2209 and 0.470 respectively, whereas these values are 0.0033 and 0.0575 for ANFIS modeling. This fact indicates that both methods have an acceptable prediction for the CDI method. In the statistical analysis, it is found that both methods are reliable and suitable for estimating the CDI method in the range of input variables. In the statistical and graphical comparison of the two models, it is found that the ANFIS model is more suitable for estimating and predicting the RSP of water. Moreover, in the analysis of the variance of the RSM model, the interaction effect of the cell voltage variable and the initial concentration value and the interaction effect of the fluid flow rate and the initial concentration value on the RSP are significant. The results show that presented models can predict the outcome of CDI reliably. Therefore, this can save time and cost by avoiding performing physical experiments to achieve similar results.