Modification of the Levenberg – Marquardt Algorithm for Solving Complex Computational Construction Problems

Systems of nonlinear algebraic equations have a special place in solving modeling and calculation problems in various areas of construction Sciences. These are tasks for calculating building structures, calculating distributed loads, mass and heat transfer tasks, calculating engineering networks, and many other computational and optimization tasks related to the construction industry. As is known, there are a sufficient number of fundamentally different methods for solving systems of nonlinear algebraic equations, which is related to the specifics of the systems, their degree of nonlinearity and conditionality. Non-linear ill-conditioned systems are particularly difficult; an erroneous choice of the method for solving such a system can lead to significantly distorted or simply incorrect results, which is absolutely unacceptable when performing the calculated stages of solving construction problems. In this paper, we propose to use a modification of the well – known Levenberg – Marquardt method developed by the authors, based on the regularization of the Jacobi matrix used in the classical Newton method, to solve complex, ill-conditioned systems of nonlinear equations. The modified method allows solving poorly defined systems and can significantly reduce the amount of calculations needed to ensure the required accuracy by reducing the number of iterative procedures. The paper presents a detailed description of the algorithm, given the solution of the model problem and, as an example the task of modeling and optimization of the port berthing facilities – mooring wall of massive masonry with various add-ons. When composing mathematical equations, the method of limit States was used, which is generally accepted for calculating such port hydraulic structures. The choice of the optimum scheme was carried out by minimizing the cost of structure during limit conditions the reliability of structures when using existing construction materials. The most significant geometric dimensions of the structure were selected using independent parameters to be optimized.