AI Case Study
Roy Hill iron ore project optimises mine operations planning based on predicting supply chain outcomes
Roy Hill iron ore project is using Schneider Electric’s Demand Chain Planning and Scheduling (DCPS) system to optimise decision-making at the mining operations planning stage based on predicting its impacts further down at each point in the supply chain.
Mining And Metals
Roy Hill Iron Ore selected Schneider Electric’s Demand Chain Planning and Scheduling (DCPS) system for its iron ore project in Western Australia. According to Schneider Electric, it "leverages production data, the software models the supply chain — including mines, processing plants, stockpiles, transport routes, and port movements — to predict how each entity in the value chain will operate. It lets companies view their supply chains both as discrete components and as a single system that operates according to unified objectives and metrics. Dynamically updated information shared among all supply chain functions arms key personnel with the intelligence to make better decisions faster. These new tools still allow for each unique function to be modelled accurately and used as the basis for optimized overall business performance. Decisions made at the mine planning stage may now be tracked all the way through the supply chain to determine the impact of the fulfilment of shipments or demand. For example, a shipment impacted by a change in the mine extraction sequence may be 10 days away, but the decision on what block to extract needs to be made within the next 24 hours.
With an integrated system, both teams are looking at the same information in a single application. As soon as a variable changes in the model, downstream effects are mirrored in the system. Optimization algorithms map the different time horizons so that planners can see the impact of certain decisions on other processes. Different scenarios can be modelled and analyzed to determine the best decision(s) to make for each supply chain function. A single suite of integrated software tools using a consolidated, unified set of data also enables advanced process modelling. Using real-time data and a common interface, simulators create dynamic models of different processes (including alternative “what if” scenarios) for planning, validating, testing, managing, and optimizing operations."
Schneider Electric claims its "sophisticated modelling approach helps us predict how each entity in the demand chain will operate (mine, plant, rail and port), not as discrete components but as a single system that is operating with unified objectives and associated KPIs. This has allowed us to gain visibility, before any CAPEX [capital expenditure] has been spent, on where we expect system constraints. We can continually test operating scenarios and see potential impact on both throughput and grade variability across the entire demand chain."
"Optimization algorithms map the different time horizons so
that planners can see the impact of certain decisions on other processes. Different scenarios can be modelled and analyzed to determine the best decision(s) to make for each supply chain function" (Schneider Electric). However, it is not clear whether the models are learned based on data or built using prior knowledge, and thus whether it constitutes machine learning or traditional AI planning.
General Supply Chain
From Schneider Electric: "In today’s highly volatile commodities market, mining companies are looking to increase their profits not by ramping up production volume but by improving productivity and efficiencies. Over the past few years, operational costs have skyrocketed. Roy Hill Iron Ore project in Western Australia has more than 2.4 billion tonnes of proven iron ore resource. At full capacity Roy Hill will be mining almost 350Mtpa total movement requiring 14 mining fleets. The high throughput and linear nature of the demand chain (a single mine, process plant, rail track, car dumper and shiploader operating out of a tidal and capacity-constrained port) require Roy Hill to be able to optimize from resource to market. Previously, these planning and scheduling horizons would have been managed by two separate teams in a vacuum — possibly leading to a situation where the wrong product arrives at the right place at the wrong time. Even getting two out of three right would still result in waste and inefficiency."
Using production data to model the supply chain which involves mines, processing plants, stockpiles, transportation routes and ports