Doctoral thesis is dedicated to meta-modeling application for dynamic system analysis and optimization. There has been developed methodology for a dynamic system meta-modeling formation on numeric and natural experiment basis and to make use of meta-modeling in multiobjective action. Method is a dynamic system of transition or stationary process simplified model (metamodel), describing it in a small number of parameters. Afterwards a mathematical model was designed with a help of combination of numeric and natural experiment and the approximation of non-modern parametrical methods; it describes the transition process parameter dependence on the system variable entry factors. The model hereinafter is used for process analysis and in multiobjective optimization. The particular case as a dynamic system examines model for concrete moulding vibropressing process. The first analysis phase includes creation by the response surface method of a simplified vibropressing transition process model where the inputs are three variables vibropressing process parameters: f0-pressing force, fA-power amplitude, ω-vibration frequency, but outputs are three factors which characterize compact process as a function of time. Metamodel is used to create the type of experimental design by Latin hypercube, varying variables that created several series of samples using the Instron testing machine Zwick 8802. Part of the samples were tested in the laboratory of Ltd. "TMB Elements". It is designed vibropressing process describing metamodels using polynomial approximation and kriging methods. The paper sets out a dynamic system optimization multiobjective purpose functions and limitations. The parameters of samples of vibropressing molding process are used in multiobjective optimization using the Pareto boundary method and software EDAOpt. Numerical experiments were validated with natural. There is a good correlation between the natural and the numerical experiments.