JUL-AUG 2017

Issue link:

Contents of this Issue


Page 13 of 53

14 INTECH JULY/AUGUST 2017 WWW.ISA.ORG COVER STORY tors who typically are not as intimate with the metallurgy and geology of the reserve and are more focused on hard- ware, software, and traditional unit con- trol than on product control. This is an opportunity for the industry as a whole. Collaboration with the METS There must be the realization that having METS expertise in improvement teams (for both vendor and operator) will have a huge impact. It is the experienced METS who can point out why issues arise through the (in)consistency of the metal- lurgy, geology, geo-technology, and geo- chemical functions and how data and initial models can relate to functions like planning, scheduling, operations, main- tenance, and potential downtime. An integrated team that talks to tech- nology vendors, but is also intimate with the process, IT/OT technology, and even the softer side of people and change processes will help the best companies adopt change rapidly and truly achieve high performance. Forced by the pro- ductivity agenda, the majors are mak- ing inroads in creating some of these cross-functional teams. To them, it will soon be clear how production variables relate to energy and water use, as well as how inputs relate to safer and more ef- ficient operations. They will soon better understand how these parameters cause downtime directly or indirectly. The teams that are persistent in their journey will ultimately find the physi- cal or natural limits of equipment and its relationships to ore attributes like hardness. As in other industries, the in- novators and early adopters will soon be followed by the early and late major- ity that will include many of the tier 2 players that are also starting to invest in this capability. By then, software solu- tions and the processes that lead to fact finding, interpretation, and modeling for this industry will have matured. Embedding best practice Significant challenges are still ahead before this can be achieved. Recent re- search shows that only 50 percent of analytics projects give a significant and expected upside. Even though some have achieved clustering, decision trees, and nearest neighbor machine learning algorithms (to mention just a couple). The mature teams that have been "playing in the sandpit" for a while start seeing how everything is hanging togeth- er for key bottlenecks in the pit or plant. For mining experts to be efficient in such an environment and achieve productiv- ity results requires some "hands on the tools" and training to use various ma- chine-learning models. It takes time for these teams to under- stand the basic requirements of first hav- ing high-quality data and using one com- mon language to help point (with respect) "creative analytics propeller heads" in the right direction. If they focus only on ana - lytics tools, it can take a while before the process is understood from the data. Context When IT and analytics teams are remote, it is hard for them to put data in the right context. Optimization projects performed remotely from the site take much more time or very significant (and therefore ex- pensive) experience to make significant findings. "Not knowing" the physical pro- cess of the plant and its variability issues is still the top reason why many analytics and im provement projects fail. When good piping and instrumenta- tion diagrams, process control charts, and alarming limits are at hand and the physical design of the asset is available in two- and three-dimensional drawings or models, being remote will likely be much less of an issue. Still, there is a strong ben- efit of being local. Experience shows that engineering artifacts of brownfield sites are not always kept up to date, and plants continuously keep changing to achieve or improve nameplate levels. Engineer- ing design databases and drawings are a great source for analysts, who often need to call upon site experts to understand is sues or physical plant limitations. It is a pity that traditionally engineer- ing, procurement and construction (EPC) and engineering, procurement, and construction management (EPCM) do not have much opportunity to jointly put effort in a longer start-up phase with operations. Teams have to rely on pro- cess control vendors or system integra- Another example of an enterprise model is modeling shared rail capacity with production efficiency in mind. Allo- cating rail capacity dynamically for multi- ple end users based on current status and flexibility of the value chain against con- tract variables and service levels is a great area to explore with this new technology. In the future, these kinds of "super models" could have an even larger overall impact on revenue instead of optimizing product volume, quality, or grade for any one specific site. Even more value can be captured for both producers and end users if real-time slot booking (including penal - ties for any over or under delivery) can be applied over time, as is the case in other in - dustries (e.g., oil terminals and refineries). Challenge Capturing such complex challenges is potentially not far away now that com - bined technology and software solu- tions are starting to scale and are more open in nature, and now that initial good proof of concepts have already captured significant value. There have been some strong potential advantages for the more recent greenfield assets to adopt new technology that has more and better sensor data. Those greenfield sites have typically adopted different and more dynamic reporting systems and architectures. You would expect these sites to be po sitioned for better analytics out- comes. However, from a traditional min- ing perspective, these newer sites often struggle to ramp up to nameplate levels. They often do not have a mature team or do have a predictable canonical and functional/role-based reporting solu - tion to fall back on, as IT budgets have gone to the latest and greatest analyt - ics architectures and minimal money is spent on traditional reporting. Two-level approach There is a strong case, therefore, for start- ing a lightweight hybrid reporting model and a best-practice analytics architec- ture. Mining analytics initiatives also re- quire mining experts to be more than a little computer savvy when working with big data servers and various nontradi- tional client tools to perform regression,

Articles in this issue

Links on this page

Archives of this issue

view archives of InTech - JUL-AUG 2017