Computational intelligence-based identification of maximally sustainable materials: The case of liquid containers

A novel computational intelligence methodology is proposed for identifying the properties of maximally sustainable materials for any given application; identification is based on the known properties of existing candidate materials. Once the correlation surface between material properties and environmental impact values of the candidate materials is created via a general regression artificial neural network, genetic algorithms are used for accurately as well as efficiently determining the minimum/a of the correlation surface, thus uncovering the properties of the maximally sustainable material(s). A demonstration is given for a simplified type of material selection concerning maximally sustainable liquid containers. The proposed methodology is compared to and found more accurate than classic interpolation techniques; additionally, sensitivity and multi-criteria analyses confirm stability of the methodology under variations in both the material property values of the known materials and in the relative importance of the input properties. It is important that this methodology can be directly applied to a variety of material selection tasks by choosing the input properties of interest as well as the desired material selection criteria.