Special Issue "Fuzzy/Neural Network Applications to Dynamics and Control of Mechanical Systems". Optimization of Pipe-Support Allocation by Neural Network.
Fumio HARA · TRANSACTIONS OF THE JAPAN SOCIETY OF MECHANICAL ENGINEERS Series C · 1992
This paper deals with optimization of pipe-support allocation using the neural network algorithm of the Boltzmann Machine, and shows the feasibility of the optimization method to actual design problems and also the convergence characteristics of optimization calculation with respect to two parameters such as the Lagrange multiplier and network temperature used. The pipe in question was modeled as a mass-spring mulidegree-of-freedom system and the support was as a spring-damper. The support allocation problem was formulated for the Boltzmann Machine to be utilized to minimize the response of the piping system to earthquake-like random excitation. We obtained the within 5% best solution among about 10 000 cases of support allocation by using the Boltzann Machine algorithm for each case of the piping systems with 5 to 10 degrees-of-freedom.