Group Method of Data Handling Neural Networks (GMDH-NN) in predicting the unconfined compressive strength of two mine tailings matrices for potential use in sustainable road base construction in cold regions

ABSTRACT: Mine tailings are the waste materials of the mining industry; they contain all the leftovers from the ore extraction process and the chemical additives that were added in its midst. Mine tailings are traditionally stored in large impoundments (dams) that keep the tailings in a liquid or semi-liquid form. Several environmental catastrophes resulting from such storage method have prompted engineers and tailings practitioners into adopting sustainable tailings management practices. As part of a broad program of study into the reuse of mine tailings matrices, Portland cement has been mixed as a binder with two types of tailings to create new stabilized tailings matrices with potential use in road base layers. The tailings have been taken from two different mines in Eastern Canada. Cement products are known to suffer from large range of variability and hence the need for a powerful numerical technique to predict their values without going through an extensive testing program. This study is an attempt to investigate the validity of using these matrices as road base courses in cold regions and to assess the application of the Group Method of Data Handling Neural Networks (GMDH-NN) in predicting the unconfined compressive strength of these mine tailings matrices. Conclusions and recommendations are drawn to help practitioners in the decision process of tailings use in current and future road base applications.