Engineering & Mining Journal

JAN 2017

Engineering and Mining Journal - Whether the market is copper, gold, nickel, iron ore, lead/zinc, PGM, diamonds or other commodities, E&MJ takes the lead in projecting trends, following development and reporting on the most efficient operating pr

Issue link: https://emj.epubxp.com/i/772779

Contents of this Issue

Navigation

Page 33 of 59

PLANT ENGINEERING 32 E&MJ; • JANUARY 2017 www.e-mj.com Plant information management sys- tems allow the integration of process data, business data and personnel data, which are the foundation to measuring plant performance. Real-time information management systems gather and histor- ically record data from all the different sources, and include the interfaces to connect with other systems. Tools to con- vert data into information are required to calculate the key performance indicators (KPIs) by applying business rules. A well-implemented asset manage- ment system checklist includes: • Preventive, predictive and condition- based maintenance; • Automatic notifications (process, alarms, events, etc.); • Advanced diagnostics (instrumentation, devices, actuators, etc.); • Downtime reporting and tracking; • Performance monitoring (KPIs) and web visualization; • Integration with computerized mainten- ance management system (CMMS); and • Asset/object information. "Progressive asset management sys- tems enable the performance monitoring system to work together with the asset man- agement solution to provide a framework to measure the performance and health of the plant assets," Almond said. "This perfor- mance measurement includes not only in- dividual machines and process areas, but also the process control performance, in- cluding advanced process control systems." Establishing such systems will cre- ate greater transparency on operations performance and identify areas for im- provement, Almond explained. Integrated operating systems should also free peo- ple and resources to focus on operational excellence and productivity. Enhancing Process Performance Process control systems can be integrated with maintenance systems, enabling ac- cess to real-time data for assets, which is key for condition-based maintenance operations. Plant engineers can combine the original equipment manufacturers' (OEM) data on how the equipment should perform with knowledge about the actu- al wear rates to schedule maintenance. Understanding the risk of continuing op- erations, embedded sensors can provide better intelligence on actual wear rates allowing some repair jobs to be postponed to the next relining period. "The fusion with more and more ad - ditional IoT data streams, from process, maintenance and wear parts, enables us to estimate 'remaining useful life' of the machines' subsystems and components more accurately—and thereby improve operation and maintenance practices," said Steen Christian Knudsen, technical manager, R&D;, for FLSmidth. "With sensors becoming smaller and cheaper, we can apply sensors in more places today," Knudsen said. "The rapid development in information technology and network topology allows the machines to send more signals, more frequently. Today, we are getting more high-quality information at lower costs." The experts at FLSmidth see interesting potential for using intelligent collaborative environ- ments employing automation and sensor technologies to exploit opportunities that increase the plant's operational effective- ness, and reduce its operational costs. Remote systems also allow mining companies to measure how the machines are performing not just relative to the ma- chines' history at one mine site but rela- tive to the performance of an entire group of plants. "Better diagnostics and trou- ble-shooting by fault symptoms signature, severity identification through data clas- sification, and pattern recognition based on neural networks all enhance process performance, and provide better access to engineering services, and proactive main- tenance services," Knudsen said. Engineers have studied neural networks for years, but the speed with which mod- ern computer systems can employ algo- rithms to enhance pattern recognition and machine learning is constantly increasing. Microsoft's Machine Learning is the most common neural network. The system's neural network algorithm, according to Microsoft, tests each possible state of the input attribute against each possible state of the predictable attribute, and calculates probabilities for each combination based on the training data. Engineers can use these probabilities for both classification or regression tasks, to predict an outcome based on some input attributes. Simplifying the explanation, Knudsen said the algorithm detects a pattern in the data. "The computer remembers that the last time it saw this data, this event occurred," Knudsen said. "The key is the human interface with the data. The algo- rithms are providing more useful interpre- tations at a much faster pace." Intel believes the scientific community is just beginning to understand the poten- tial of machine learning. Data scientists, developers and researchers are using ma- chine learning to gain insights previously out of reach. The company said that en- gineers can now scale machine learning and deep learning applications quickly— and gain insights more efficiently—with the existing hardware infrastructure. The principles behind these concepts are not new, Knudsen explained. "Now, however, we see the development and the possibility of data storage and faster computers," Knudsen said. "This allows A team in the U.S. monitors daily operation at a nickel concentrator in South Africa.

Articles in this issue

Links on this page

Archives of this issue

view archives of Engineering & Mining Journal - JAN 2017