It is well known that the mining industry today has a wealth of data from operations. This includes data related to equipment performance, orebody (resources, reserves), quality (grade), plant operations (recovery), drill & blast, energy/fuel consumption, and maintenance, to name a few elements. Most mines have implemented some form of daily and monthly reporting and wonder whether this is the right time to dive into proof of concepts on machine learning. Many do, only to later realize that these experimentations are unable to scale, had an ill-defined problem statement, are too complex to manage in the long run, lack data quality or do not provide the desired benefits. In essence, a structured approach is necessary and application of data analytics technology should progress as the business matures, to derive insights from data for decision making.
Insights from data
The following section covers some key approaches in adopting data driven strategies in the mining industry and can be applied by miners large or small depending on their maturity.
Design thinking led KPI and data visualization: Most miners have a good level of reporting in place and have invested in data-warehouses as a central data repository and use business intelligence tools to generate reports. The problem typically is that these reports by themselves do not tell a story (with lack of trends, or relevant KPIs) and are often just a historical statement of figures that are not necessarily useful for management. Many times these reports are not holistic and have a poor visualization that doesn’t address real world problems – managing the mine to mill interface, re-handling tons, variations to plans and what is needed to get back to plan. Good visualization of data can trigger management and supervisors to ask the right questions for business improvement and address ill-defined problem statements.
Model driven business improvement approach: Most miners have used value driver trees at some point as a tool but these are mostly Excel based and not dynamic. They usually provide only a point-in-time snapshot into the cost or productivity hierarchy but don’t drill down into the underlying technical levers or allow for what-if analysis to plan for short term improvements. Ideally a model driven approach can be adopted to identify bottlenecks which can be improved upon before iterating to the next bottleneck. The levers that are controllable such as breaks, operator availability, fill factors can be acted upon either using additional technology in the field or operational work practices to guide improvement and achieve scale and sustainability in the long run.
Real time data analytics: This is an approach much talked about with the advent of IoT and data platforms. They are really a complement to the operational data sources and are good for decision making at a shift level or a day and could provide additional insight to the information that underlying mining systems provide today (e.g. an inventory view of a silo in real time along with the rate of mining from the pit etc.). They could also help with short interval control approaches for decision making.
Machine learning based approach: Machine learning can provide tremendous business value where there is a well-defined problem statement, with enough granularity from data exploration on domain variables and possible outcomes and a strategy on how the outcomes can be acted upon. Often the last bit is ignored—a good correlation chart tells the story but either cannot be acted upon or the business process does not include it in the workflow.
There are good examples where predictive maintenance has been successful in identifying issues such as survival analysis of haul trucks, prediction of conveyor failures using decision tree algorithms, etc. However they have lacked a process or approach to make these initiatives sustainable and embed them into the work process.
Implementation journey
Below table indicates a data driven journey and can be considered as an important guideline to in a mining organization:
Some key considerations
The following points are important guidelines while adopting a data driven strategy.
Conclusion
The benefits of a structured program in mining is well documented in media. Productivity benefits of 20% or higher is not unheard of, but that just applies to a segment of the value chain. Mature analytics approaches can solve very high profile problems such as productivity, which is again well documented in the media. In the end, for a successful program, a mining company’s ability to implement would depend on:
Sudip Chaudhuri,
Global Practice Head - Mining, Wipro Limited.
Sudip Chaudhuri heads the Mining Practice for the Energy, Natural Resources and Utilities business unit at Wipro. With over 20 years of varied information technology experience in mining and mineral processing, Sudip has worked with numerous clients in the mining and minerals industry on transformational and advisory assignments and designing end to end programs . With deep domain experience across the mining supply chain and execution he has been able to effectively apply new technologies to bring about improvement in productivity and safety in the mining context. He can be reached at sudip.chaudhuri@wipro.com