Research On Alloy Addition Optimization Based On Improved Multi-objective Particle Swarm Optimization

Steel is the basic material of industry and determines the country's industrial base.With the rapid development of the steel industry and the increasingly fierce market competition,the concept of reducing costs and increasing benefits promote steel companies to improve steel quality,reduce costs and improve profitability.Basic oxygen furnace is currently the most widely used steelmaking method.So that improving the automation technology of basic oxygen steelmaking is the key to improve the competitiveness of steel enterprises.However,basic oxygen steelmaking is a complex and rapid process of reaction,which is difficulted to produce automatically.Most calculation method of material input are only depended on artificial experience,which are easy to waste materials and increase the costs.Therefore,a reasonable material calculation model for basic oxygen steelmaking is an urgent need to produce the cost-benefit steel.The minimization of the alloy addition cost and the maximization of the steel quality can be regarded as a multi-objective optimization problem.Multi-objective optimization is different from single-objective optimization,the result obtained in multi-objective optimization is a set of solutions which can't be compared with each other.As a swarm intelligence technique algorithm,the particle swarm optimization(PSO)could obtain a number of available optimal solution,it is suitable for solving multi-objective optimization problem.In PSO,How to balance the search ability is a key problem,which affects PSO's optimization performance.How to improve the performance of multi-objective particle swarm optimization(MOPSO)algorithm can still be discussed and studied deeply.The issues on the following were mainly studied in this master degree thesis:(1)To solve the problem of global search and local search in multi-objective particle swarm optimization,a bi-group MOPSO algorithm based on diversity metric is proposed.First,a diversity metric is introduced to MOPSO algorithm and improved based on the characters of MOPSO.Second,we divide the whole swarm to bi-group with different searching tasks.One of the groups keeps population's diversity during evolution to search better in the whole search space.And the other group keeps its convergence to local search nearby the Pareto front.Further more,we adjust the searching behavior of the groups based on the diversity metric to balance the diversity and convergence.The simulations on several standard test functions show that our proposal is effective.(2)Since the adding amount of alloy was difficult to be calculated in basic oxygen steelmaking process,we proposed a multi-objective optimization model to calculate the optimum adding amount of alloy.Our method could reduce the costs of alloy and the error of elements.On the basis of analysing the alloy adding problem,we designed a soft measurement model based on the extreme learning machine(ELM).And then,an improved MOPSO algorithm is proposed and applied to solve the alloy adding problem.At last,we choose the best decision solution to get the adding amount of alloy.The simulation results showed our mothod is effective to find optimum alloy adding amount.