Modelling total weighted vibration of a trailer seat pulled by a two-wheel tractor consumed diesel–biodiesel fuel blends using ANFIS methodology

Abstract

The application of adaptive neural-fuzzy inference system (ANFIS) to predict the total weighted vibration acceleration of a trailer seat pulled by a two-wheel tractor is set out in the present paper. The vibration acceleration signals were obtained in a field experiment using a 9.5 kW two-wheel tractor. Three accelerometers were installed at three orthogonal directions according to ISO 2631-1 standard on the trailer seat for measuring and recording the vibration acceleration signals. The two-wheel tractor consumed diesel–biodiesel fuel blends, and the engine speed and gear ratio were varied to cover normal range of operation in transportation conditions on asphalt rural road. The digital recorded vibration acceleration signals in time domain were converted to the frequency domain using Fast Fourier Transform algorithm, and the one-third octave frequency bands were obtained. The one-third octave frequencies were weighted according to the ISO standard. Altogether, 100 patterns were generated for training and evaluation of ANFIS. The input parameters of ANFIS were the tractor engine speed, transmission gear ratio of the tractor and fuel blend type, and the output parameter was the trailer seat total weighted vibration acceleration. The ANFIS structure was designed based on the grid partition, fuzzy c-means clustering and subtractive clustering. The results revealed that the ANFIS with subtractive clustering method was the best for accurate prediction of the total weighted vibration acceleration of the trailer seat. The number of rules, root-mean-square error (RMSE) of train, RMSE of test, correlation coefficient (R) of train and R of test values for the optimized structure were 5, 0.1479, 0.2141, 0.9175 and 0.8804, respectively.