| Summary: | 碩士 === 淡江大學 === 機械工程學系 === 86 === In the actual engineering optimal design problems, it often requires getting global optimal solution and designing for discrete variables. In this paper, we have analyzed the theory of original Simulated Annealing method to deduce the ameliorative designing idea, and using the ameliorative Simulated Annealing method to strengthen the efficiency of the program, showing its quality of that we need.
At first, this paper aim at the original theory for more perfect programming process, strengthen transacting the constraint problems, and have more detail norm for terminal conditions, becoming ameliorative Simulated Annealing algorithm. Then according to Metropolis criteria, making a more exact judgement to the accept probability of optimal point, doesn''t waste time for needless computing, making the setting of initial and final temperature more reasonable, and for speeding the search efficiency. The other points strengthening of the algorithm include observing the search course forecast the presumable position of next point, and reducing the range of search powerfully, in the way of temperature decreasing. We want to simulate the actual drop in temperature more precisely, we adopt the fact speed of temperature decreasing at high temperature and slow at low temperature. The strategy is using exponent to be the function of temperature decreasing in our paper. So in general, this paper has many ways of research for increasing the efficiency of program.
In the multiple optimal problems, ameliorative Simulated Annealing program increasing the efficiency of program powerfully, it use more reasonable initial temperature and more exact search range, reducing the frequency of computing objective function, shortening the time of program execution. When it meets the problems which include mixed integer adn discete variables, our method can also provide the same algorithm, just change the formula of searching variables, then we can get the optimal result. Therefore ameliorative program have the general and simple principle. It is a method which have quite practicability.
|
|---|
