Grade Estimation in Deposits with Locally Varying Anisotropy using Ant Colony Algorithm: Case Study, Miduk Porphyry Copper Deposit

Summary:

Anisotropy of a deposit is due to its directional variations of grade or structure. Locally varying anisotropy (LVA) is the specific case of anisotropy in some structurally-controlled deposits. In this research, using ant colony application in geochemical anomaly detection and LVA field of the study area, an algorithm (ACLVA) has been developed to smartly direct the ants into the more continuous paths and ants, meanwhile act as moving average agents over their routes. Ordinary kriging (OK), OKLVA, and ACLVA were applied on borehole samples of Miduk copper deposit as the case study, and estimations were validated with blast hole samples. The estimations were improved with ACLVA. A newly-developed hybrid ant colony with an LVA algorithm (ACLVA) is presented that can modify an initial estimation of the data according to the LVA field. ACLVA is compared with recently-developed OKLVA and OK on borehole samples of a copper deposit. The estimations were validated with blast hole samples.

Neural networks (ANN) have recently been used to estimate grade. They were able to present acceptable models of the resources. Many attempts have been made to incorporate the LVA feature of deposits into geostatistical models. In this research, a new hybrid AC-LVA algorithm has been developed that can produce a more representative map of the continuities.

Methodology and Approaches:

Artificial ants are randomly put in the grid cells and while searching for high values according to the LVA field, act as moving average agents on their routes. To decrease the randomness effect of AC, the program is repeated. The ants’ stability termination condition is. Better initializing would lead to a better Jensen-Shannon (JS) value.

The outputs of OK, OKLVA, and ACLVA were validated with blast hole samples. The results showed that ACLVA performed 4% better than OKLVA and 3% better than OK. The initial number of ants can be set optimally. Other parameters should be changed based on the best JS value. The results would be significant if the deposit has more complex LVA.