Predictive capability evaluation of RSM and ANN models in adsorptive treatment of crystal violet dye simulated wastewater using activated carbon prepared from Raphia hookeri seeds

Abstract

A comparative evaluation of the predictive capability of response surface methodology (RSM) and artificial neural network (ANN) in adsorptive treatment of dye simulated wastewater using acid activated Raphia hookeri seeds (AARHS) as adsorbent was investigated. FTIR spectrum and SEM displayed the functional groups and surface morphology of AARHS, respectively. A 2⁴ central composite experimental design (CCD) and a three-layered (4:n:1) feed-forward architecture of artificial neural network trained by the Levenberg–Marquard back-propagation algorithm were developed to model and predict the process parameters. The best performance for the ANN architecture was obtained at n = 10 neurons. The response (% crystal violet dye adsorbed) was a function of four controllable input variables: adsorbent dosage, time, solution temperature and pH. Sensitivity analyses of the input variables on the response evaluated by Pareto analysis (P) and Relative Contribution (RC) for RSM and ANN, respectively, reveal that pH has the highest effect on the response. The predictive ability of the ANN and RSM models were evaluated in terms of RMSE (0.912; 1.186), χ² (0.267; 0.472), MPE (0.549; 1.085) and R² (0.9950; 0.9867), respectively. The result showed the superiority of ANN in capturing the nonlinear behaviour of the adsorptive system. An optimal response of 98.05% was recorded at 3.812 g adsorbent dosage, 65.58 min, 47.83 °C and pH of 7.928.

Keywords:

• Response surface methodology
• artificial neural network
• optimisation
• AARHS
• crystal violet dye

Disclosure statement

No potential conflict of interest was reported by the authors.