This paper describes a novel refinement to a Tabu search algorithm that has been implemented in an attempt to improve the robustness of the search when applied to particularly complex problems. In this approach, two Tabu searches are carried out in parallel. Each search thread is characterised by it's own short term memory which forces that point out of local optima. However, the two search threads share an intermediate term memory so allowing a degree of information to be passed between them. Results are presented for both unconstrained and constrained numerical functions as well as a problem in the field of hydraulic circuit optimization. Simulation of hydraulic circuit performance is achieved by linking the optimization algorithm to the commercial simulation package Bathfp.