Prognostication of vertical stress transmission in soil profile by adaptive neuro-fuzzy inference system based modeling approach

• Analysis of soil stress propagation by run-off-road vehicles in soil bin testing facility.
• Application of ANFIS technique as modeling implementation tool.
• We used a hybrid method of the gradient descent and the least-squares method to find optimal learning parameters.
• Supremacy of Guassian MF compared to other tested MFs of ANFIS in modeling.

Prediction of vertical stress transmission in real soil profile using adaptive neuro-fuzzy inference system (ANFIS) is documented in this investigation. A soil bin facility holding a single-wheel tester was utilized to arrange controlled condition for exploration of the effects of wheel load, forward velocity, slippage and depth each at three different levels. A profile housing seven load cells was buried at different depths when data were transmitted to a data acquisitioning system for derivation of 81 data points and then to build ANFIS-based model. The Sugeno-type fuzzy rules were constituted with various membership functions in the representations. In the Sugeno-type fuzzy inference approach, the modal was developed according to the four input parameters. Performance evaluation criteria (i.e. MSE, MRE and R2) were incorporated in the study to find the highest quality solution. It was deduced, on the basis of performance criteria, that a Guassian membership function outperformed other tested membership functions. The results could serve as a catalyst to expedite the investigations in the realm of artificial intelligence application in prediction of soil stress transmission created by wheeled vehicle trafficking.