Applying heat treatment to metal (heating to a specific temperature for a set duration) can alter its properties making it suitable for use in specific applications. Due to the costs of heating then testing the metal, it is desirable to minimise the number of such tests. The problem of finding a small set of heat treatments optimised to allow accurate interpolation of the remaining points is investigated in this paper. We propose an evolutionary algorithm (HTEA) to find a set of treatments that maximise the accuracy of the inference. We conclude that HTEA is effective in recommending sets of heat treatments the results of which illuminate the relationship between heat treatments and the desired mechanical property.