There must be some equivalent between doing this and variational calculus. I wonder if anyone uses similar methods for solving FE models.