Modeling Deep Beam Strengths with a Genetic Programming System

Abstract

This paper proposes a genetic programming system (GPS) to both predict and program deep beam strength. The soft prediction is to provide estimated results for problems and the soft programming is to represent problems with meaningful formulas. In proposed GPS, two models of genetic programming (GP) and weighted genetic programming (WGP) are involved. GP is structured of a tree topology, inputs, and operators. WGP improves GP on introducing weights to balance tree blanches. Both GP and WGP results are provided in the GPS. Besides, this paper further designed a WGP to provide polynomial-like formulas, which seemed pretty helpful to parameter studies. Deep beam strengths were used as the case study. Results reveal that GPS has two distinct advantages over other black-box approaches in programming formulas and parameter studies. Furthermore, GPS does not surrender in prediction accuracy at all. Summarily, uses of proposed GPS are to predict, program, and model engineering problems.