Finite Element Neural Network Computing And Its Circuit Realization

It is known that there is an unsolved contradiction between the computationand memory of the Von Neumann computer. So many problems of engineeringmechanics that have large computation scale haven't been solved. And there stillhaven't breakthrough in some new or pendent problems because of such reasonas computation, etc. Therefore, new theory and method of computationalmechanics need to be developed to solve above problems.Solving method of finite element and neural network design were studied inthis paper. The neural network of finite element solving was obtained on thebasis of Hopfield neural network that was reformed. And the no-error solving offinite element neural net computation was realized in theory. And a designmethod to construct an artificial neuron by using electronic devices such asoperational amplifier, digital controlled potentiometer and so on was presented.A programmable hardware neural network of finite element can be build up byusing analog switches to interconnect inputs/outputs of hardware neurons. Theweights, biases and connection in the hardware neural network of finite elementcan be adjusted automatically by microprocessor according to the results of trainto controlling system, This programmable hardware neural network of finiteelement has some more adaptability for different systems.