An improved constrained quasi-Newton method for the solution of inverse electromagnetic problems

This paper presents the formulation of an improved direct technique called modified constrained quasi-Newton method (PLBA-CR) achieved with constraint correction and objective reduction algorithms which may be used for the solution of inverse electromagnetic problems. Two problems are discussed and the results are compared with the quadratic extended penalty function (QUAP) and the augmented Lagrangian multiplier (ALM) method in terms of accuracy and calculations required. The first problem consists in the minimization of the weight of an EI core inductor. The second problem consists of the shape optimization of an electromagnet in order to maintain the magnetic flux density constant at a prescribed value in its air gap. The results show that the PLBA-CR technique is substantially faster in terms of computation time and would appear to have certain important advantages over the other methods.