Predicting the rheological properties of waste vegetable oil biodiesel-modified water-based mud using artificial neural network

Abstract

ABSTRACTOil-based drilling muds have the greatest preference for drilling operations. However, utilization of environmentally friendly components in drilling mud is fast becoming a requirement prompting production of different types of drilling mud. While there is abundance of prediction models for the rheological properties of oil-based drilling mud, there is scarcity of the same for drilling mud with environmentally friendly additives. In this work, an artificial neural network (ANN) and a multiple nonlinear regression (MNLR) model were developed aimed at predicting the apparent viscosity, plastic viscosity and yield point of waste vegetable oil biodiesel-modified water-based mud. The mean squared errors and correlation coefficient were the key parameters to evaluate and compare the performance of both models. The results indicate that prediction of the ANN perfectly matched the experimental values better than those of MNLR, reflecting its superior performance.Abbreviations: WVO: Waste vegetable oil; WVB: Waste vegetable oil biodiesel; WBMM: Waste vegetable oil modified mud; CMC: Carboxylmethyl cellulose sodium salt; SNPH: Sulfomethyl humate and phenolic resin; SMP-3: Sulfonated methyl phenol; WBM: Water-based mud; Θ600: Dial reading at 600 rpm; Θ300: Dial reading at 300 rpm; PV: Plastic viscosity; AV: Apparent viscosity; YP: Yield point; ANN: Artificial neural network; LM-BP: Levenberg–Marquardt back propagation; FFBPN: Feed-forward backprop network; W1i: Weight in the hidden layer; W2i: Weight in the output layer; b1: Bias of the hidden layer; b2: Bias of the output layer; MSE: Mean squared error; YExp,m: Experimental value; YPred,m: Predicted value; YExp,m: Average of the experimental value; R2:Coefficient of determination; AAPE: Mean absolute percent error; APE: Average percentage error