Research And System Implementation Of Parallel Global Optimal Algorithm Based On Interval Mathematics

Interval Mathematics is an important branch of computational mathematics，it usesan interval variables instead of points to computer, the purpose is to control the errorestimate of the result to ensure the reliability of the result. However, its inherentdrawback: the super width severely limits its development.The super width of interval mathematics is known as a dependent problem, it isthe width of the result interval in the interval calculation is larger than that of the truefunction in the point. When the width of the interval result is much larger than thewidth of the true function, the interval result would be meaningless. Therefore, thispaper will analyze super width, make use of the branch and bound method to reducethe super width, and cite heuristic strategy to reduce the computation of the intervalglobal optimal algorithm and reduce memory consumption of the computer.Due to the problems of the large amount of calculation and the long calculatedtime, the interval global optimal algorithm has been limited in some degree in thefields of engineering. Particle swarm global optimal algorithm is the global optimalalgorithm which makes use of the swarm intelligent rule to search for globaloptimization in the specified range, converging to the global optimum fast. Therefore,this paper will analyze the particle swarm algorithm and propose a new hybrid globaloptimization algorithm. the algorithm inherits the properties of the particle swarm,searches in the initial interval, accelerates the convergent rate, then finds the globaloptimum interval.The traditional interval global optimization algorithm has the shortcoming of thelong calculated time and possesses the valuable computing resources in the long term.Therefore, this paper researches the global optimal algorithm runs on a cluster in theload balancing way. Parallel computing is the computing mission is divided into smallparallel parts, the small parts will be run in the some processors in the same time, thepurpose is to shorten the calculated time. Therefore, for the hybrid global optimalalgorithm, the paper will first make the algorithm parallel, then use load balancingstrategy to schedule the parallel hybrid global optimization algorithm to run in thelight load node, avoid the phenomena that the cluster appears busy and improve theefficiency of the system.The main work in this paper is as follows:1) Research the Deletion Strategy of Interval Mathematics. According to thesplitting method, this paper proposed a new heuristic selection strategy which canreduce the amount of computation and reduce memory consumption.2) Research the particle swarm global optimal algorithm. According to itsfeatures that contain less parameters and fast convergence, the paper proposes a newhybrid global optimization algorithm to overcome the shortcomings in the latersplitting term. 3) In order to solve the problem that the traditional global optimizationalgorithms often consume large amounts of cluster resources, this paper developed aweb system based on load balancing strategy. This paper will firstly make the newhybrid global optimal algorithm parallel, shortening the calculation time of thealgorithm, then the web system will schedule the algorithm to run in the light loadedserver in the according of load balancing strategy. The load balancing strategy canshorten the response time of the cluster and improve resource utilization rate.