A Neuro-Fuzzy-Analysis of Variance Approach for Improved Estimation and Prediction of Tire Reliability

This paper presents a novel adaptive approach consisting of artificial neural network (ANN) and fuzzy linear regression (FLR) for improved estimation and analysis of tire reliability. This approach undertakes to improve estimating the actual reliability of tire in the field using tire life data, which are obtained from a laboratory test data. In this study, a particular failure mode known as tread and belt separation (TBS) is considered and viewed as the resultant of some tire geometry and physical properties. According to the proposed approach, ANN and FLR approaches were applied to the provided data in order to capture the potential complexity, uncertainty, and non-linear relation between reliability function and its determinants. Next, the preferred ANN was selected based on sensitivity analysis results of mean absolute percentage error (MAPE) function while the preferred FLR was identified regarding to the analysis of variance (ANOVA), MAPE and index of confidence (IC) results. The results obtained from performance comparison of both models demonstrated the considerable superiority of the preferred ANN over preferred FLR considering the nonlinear and complex nature of reliability function. This is the first study that presents an ANN-FLR approach for accurate estimation and analysis of tire reliability capable of handling complexity and non-linearity, uncertainty, pre-processing and post-processing.

Tire Reliability Estimation; Artificial Neural Network; Fuzzy Linear Regression; Complexity; Nonlinearity; Limited and Inconsistent Data; Analysis of Variance