This paper builds on the companion paper (Part I) and presents an improved method to introduce manufacturing rules into the evolutionary algorithm for shape optimisation of cold-formed steel (CFS) sections. Unlike the companion paper, where the cross-sections were arbitrarily drawn with short elements and the manufacturing rules were introduced as constraints into the algorithm, the manufacturing rules are introduced herein as an intrinsic part of the cross-sectional shapes, therefore avoiding the complexity of dealing with constraints in the Genetic Algorithm. Specifically, the crosssections in the first generation are arbitrarily drawn with a fixed number of consecutive flat segments (greater than 10 mm long) directly resulting in manufacturable cross-sections. The improved algorithm is applied to optimise the simply-supported, singly-symmetric and free-to-warp columns analysed in the companion paper. The obtained cross-sectional shapes are compared to those found in the companion paper. The convergence rate and computation time are also compared between the two methods. Results show that the improved method is more efficient in terms of computation time than using the Hough transform in the companion paper.