The rigorous method for retrofitting flexible or multi-period heat exchanger networks (HENs) was developed. Following flexibility concept of KOTJABASAKIS and LINNHOFF [1] no major structural topology changes of HEN were allowed though some limited modifications as e.g. addition of new heat exchangers in series to existing ones are possible. The method presented in the paper differs also from the approach of authors [1] in solution procedure. Stochastic optimisation approach was applied instead of tedious and unreliable simulation based calculations. The paper presents the formulation of optimisation problem including goal function, mathematical model of a HEN and constraints. Because of disjunctive cost function and conditional relations, the problem is of MINLP type. The direct search solution procedure developed in this paper solves this as NLP optimisation problem with non-linear goal function, simple inequality constraints and conditional statements. Moreover, non-linear equality constraints can be solved as linear set of equations. The application of the suggested approach is presented for industrial task from [2]. The approach is much easier to use than the simulation based method or equation oriented mathematical programming one. Simple direct search optimisation algorithm from [3] applied in this work appeared quite efficient and reliable tool.