Data-driven SIRMs-connected FIS for prediction of external tendon stress

• Main Authors: Tay, K.M, See, Hung Lau, Chee, Khoon Ng
• Format: E-Article
• Language: English
• Published: 2015
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://ir.unimas.my/id/eprint/8144/1/CAC41061X_Lau%20See%20Hung.pdf; http://ir.unimas.my/id/eprint/8144/

This paper presents a novel harmony search (HS)-based data-driven single input rule modules (SIRMs)- connected fuzzy inference system (FIS) for the prediction of stress in externally prestressed tendon. The proposed method attempts to extract causal relationship of a system from an input-output pairs of data even without knowing the complete physical knowledge of the system. The monotonicity property is then exploited as an additional qualitative information to obtain a meaningful SIRMs-connected FIS model. This method is then validated using results from test data from the literature. Several parameters, such as initial tendon depth to beam ratio; deviators spacing to the initial tendon depth ratio; and distance of a concentrated load from the nearest support to the effective beam span are considered. A computer simulation for estimating the bond reduction coefficient   u  is then reported. The contributions of this paper is two folds; (1) it contributes towards a new monotonicity-preserving data-driven FIS model in fuzzy modeling and (2) it provides a novel solution for estimating the u  even without a complete physical knowledge of unbonded tendons.