Investigation of encoding methods of the genetic algorithm for the load identification of composite structures

Abstract

In this paper, the method combining genetic algorithm and finite element was used in identification of magnitude and location of load for composite structure. Further study on encoding methods of chromosome was conducted. The binary coding method which encodes the node number of load location and magnitude of load is simple and efficient. There is a mapping error in continuous function discretization, and its encoded string length depends on the required identification accuracy. The length of chromosomes should have a moderate scale, otherwise, too long chromosomes will affect the computational efficiency of the algorithm; The hybrid coding method encodes the load location with binary coding method and encodes the magnitude of load with floating-point coding method such that the length of chromosomes are greatly reduced. This method significantly improves the computation efficiency and accuracy, but it is still difficult to achieve the load identification of non-grid node. This paper effectively achieved the identification of load acting on the node and inner points of the element for composite structure by combining the loaded element discriminance method and the floating-point coding method, where the load magnitude and the abscissa, vertical coordinates of load point are encoded in float-coding. Then, the load is allocated to the nodes of element by the equivalent distribution principle.