Development of a heuristic methodology for designing measurement networks for precise metal accounting

This thesis investigates the development of a heuristic based methodology for designing measurement networks with application to the precise accounting of metal flows in mineral beneficiation operations. The term 'measurement network' is used to refer to the 'system of sampling and weight measurement equipment' from which process measurements are routinely collected. Metal accounting is defined as the estimation of saleable metal in the mine and subsequent process streams over a defined time period. One of the greatest challenges facing metal accounting is 'uncertainty' that is caused by random errors, and sometimes gross errors, that obtain in process measurements. While gross errors can be eliminated through correct measurement practices, random errors are an inherent property of measured data and they can only be minimised. Two types of rules for designing measurement networks were considered. The first type of rules referred to as 'expert heuristics' consists of (i) Code of Practice Guidelines from the AMIRA P754 Code, and (ii) prevailing accounting practices from the mineral and metallurgical processing industry which were obtained through a questionnaire survey campaign. It was hypothesised that experts in the industry design measurement networks using rules or guidelines that ensure requisite quality in metal accounting. The second set of rules was derived from the symbolic manipulation of the general steady-state linear data reconciliation solution as well as from an intensive numerical study on the variance reduction response of measurements after data reconciliation conducted in this study. These were referred to as 'mathematical heuristics' and are based on the general principle of variance reduction through data reconciliation. It was hypothesised that data reconciliation can be used to target variance reduction for selected measurements by exploiting characteristics of entire measurement networks as well as individual measurement characteristics.

Grantor dc:publisher.institution

Department of Chemical Engineering

Year dc:date.issued

2016

Author dc:creator

• Bepswa, Paul Aaron

Advisor dc:contributor.advisor

• Deglon, David A

Language dc:language.iso

eng

Handle dc:identifier.uri

http://hdl.handle.net/11427/20420

OAI identifier oai:identifier

oai:open.uct.ac.za:11427/20420