Research On State Transition Algorithm And Its Application In The Zinc Purification Process

State transition algorithm (STA) is a new intelligent optimization algorithm; it has advantages of simple structure, parallelism good and easy to understand. STA has made good application effects on single-objective optimization problem.In this paper (1) the improved rotation operator and Translation operator is proposed based on the detailed analysis of the basic principles of STA; the time complexity of STA was been discussed and simulation compared with genetic algorithms; the convergence of STA was been studied and proved.(2) For solving the overlapping peaks problem in multi-component detection of zinc hydrometallurgical process, an online analysis method (STAWNN) for polarographic detection signal of multi-metal ion concentrations is proposed based on improved wavelet neural network. We adopted state transition algorithm to optimize the parameters of wavelet neural network in offline training to avoide local optimum. Before detection, a method of extracting feature points is provided to solve information redundancy and computational complexity problem. In STAWNN, firstly, the first derivative of polarographic signal is obtained through the discrete wavelet transform. Then, correspond eigenvalues are obtained as wavelet neural network input based on the original signal and the first derivative of polarographic signal. STAWNN was tested by the actual polarographic overlapping peaks signal of are zinc and cobalt, simulation results of STAWNN was superior to conventional curve fitting and BP neural network algorithm.(3) By analyzing the characteristics of multi-objective optimization problem, multi-objective state transition algorithm (MOSTA) and constrained multi-objective state transfer algorithm (CMOSTA) is proposed to solve unconstrained and constrained multi-objective optimization problems. Both MOSTA and CMOSTA adopted a search strategy with multiple populations, and combined with the sorting tactic of Pareto non-dominated ranking. Furthermore, a mutation operator is proposed to improve the distribution diversity of solution set on the Pareto front. Compared with MOSTA, CMOSTA combined constraint processing technology, in addition, feasible solutions and infeasible solutions in the intermediate populations were sorted according to different methods in CMOSTA. Experimental simulation results show that both MOSTA and CMOSTA could effectively solve multi-objective optimization problem.(4) CMOSTA was applied to solve multi-objective optimization model of amount of zinc and cobalt ion concentration. Based on TOPSIS, the satisfactory solution of the optimal control schemes is obtained to achieve optimal control of the amount of zinc.