Appraisal of Takagi-Sugeno-Kang type of adaptive neuro-fuzzy inference system for draft force prediction of chisel plow implement

Abstract

Required draft force of chisel plow implement during tillage operations was comprehensively apprised. Field experiments were carried out at three levels of plowing depth (PD) (10, 20 and 30 (cm)) and three levels of forward speed (FS) (2, 4 and 6 (km/h)) in a clay loam soil. An intelligent model based on soft computing technique, adaptive neuro-fuzzy inference system (ANFIS), was used to integrally predict draft force. The FS and PD were chosen as input variables and the draft force was considered as output parameter in the first order Takagi-Sugeno-Kang type of ANFIS model. A comparison was also performed between results of the best developed ANFIS model and those of the well-known mathematical model suggested by American Society of Agricultural and Biological Engineers (ASABE). To select the best model with the highest predictive ability, some statistical performance criteria (SPC) (coefficient of determination (R2), root mean square error (RMSE), mean relative deviation modulus (MRDM), mean of absolute values of prediction residual errors (MAVPRE) and prediction error mean (PEM)) were used. The results demonstrated that the best ANFIS model with acceptable SPC values of R2=0.994, RMSE=0.722 (kN), MRDM=3.172%, MAVPRE=0.561 (kN) and PEM=−0.071% was more accurate than the ASABE model. The ANFIS modeling results also showed that the simultaneous or individual increment of FS and PD resulted in nonlinear increment of draft force. Additionally, the interaction of FS and PD on draft force was congruent. Application of physical perception obtained from developed ANFIS model results led to exposition of a new scientific window towards deep root understanding of draft force behavior. Thus, it is practically proposed to employ the ANFIS model for proper selection of tractor type for pulling chisel plow implement in the most efficient manner.