IM 2017 November 17 | Page 20

BIG DATA AND DIGITALISATION utilise a maintenance management solution that approaches data management and analysis from a problem-solving perspective. While collecting ‘all data, all the time’ is impressive, collecting the right data at the right time is necessary to analyse these datasets in a way that quickly identifies results-driven actions.” Acquiring telemetry data from assets pertaining to the engine, chassis, drive system, tyres, and even related oil analysis, in real time, arms maintenance teams with plenty of raw information to evaluate component health. From there, they can project the component’s estimated remaining useful life, or even take immediate action (repair or replace) to mitigate a catastrophic failure. Modular argues that when considering the health of equipment components, maintenance departments must identify high-impact failure modes, which includes: n The failing component n The root cause of the component’s failure n Component health degradation trends for the failure type (is the failure detectable earlier on the P-F curve?) n Actions to take to prevent future failures, or a plan to replace the component before it fails. can send the data you need now (the ‘right’ data) as a higher priority to better prevent catastrophic failures, while still collecting and transmitting the rest of the data (the Big Data, not required now) in the background for longer-term processing and analysis. “For example, if a haul truck’s engine oil levels are critically low, a maintenance technician will receive immediate notification. This real-time information drives technician action to stop the unit and address lubrication levels before major engine damage occurs.” While intervening to avoid this catastrophic failure will deliver direct and often-significant component cost savings, other indirect yet substantial cost savings can also be realised: n Reduction in required personnel and material resources, since it takes far less labour time to address the lubrication levels than it does to replace major engine components n Potential avoidance of interruptions to operations, as a down truck may block route access if it fails in a critical path area of the mine n Avoidance of interruption to planned work, since units with planned maintenance now fall to a lower priority or delay, increasing Van Wegen adds: “An effective maintenance management system provides plenty of data to confidently identify the above, and some of the recently released systems can often process this data in real time by two different technological means: edge computing and cloud computing. When utilised the right way, both of these computing means will help increase equipment availability, minimise troubleshooting efforts, mitigate catastrophic failures, and provide analyses to identify component health degradation and predict when potential failure points will occur. These technologies also facilitate necessary changes to procedures or maintenance plans in a timely manner, helping to improve the visibility of operational data across the mine.” So-called edge computing, the best option for processing real-time data information streams, occurs at the equipment level to reduce latency and process algorithms locally without the need for high-speed internet. “This serves as a mechanism to quickly notify operators and maintenance teams about critical, time-sensitive conditions while also enabling rapid data transmission for central processing. The logic that would otherwise require time to send data to the cloud, run analysis, and push notifications back to an operator can now live in the equipment’s on-board mobile system, essentially learning from big data analytics, and immediately feeding this logic back to the edge to prioritise critical issues and accelerate proper actions where and when they’re needed.” Modular states that edge computing devices their potential for failure n Reduction in technician hazard risks and associated liabilities, as remote analysis reduces on-premises, physical analysis of components. 18 International Mining | NOVEMBER 2017 Cloud computing and Big Data Since Big Data generally requires some amount of analysis, an ‘all data, all the time’ streaming capability is less suited for edge computing than it is for cloud computing; the Big Data priority focuses more on analysis than it does immediate message transmission. Furthermore, stream ing large volumes of data requires massive storage and advanced analytics capabilities, unfavourable to edge computing. Some maintenance management systems employ cloud computing, in addition to edge computing, to manage the large quantities of real-time and batch data that could have any influence over a component’s health, spanning as many different informational points as possible. Van Wegen from Modular adds: “Cloud computing’s scalable mass collection and data storage facilitates long-term analysis. This enables mines to improve their maintenance plan by reviewing predictions of a component’s remaining life, and provides a feedback loop to the edge configuration as new trends are discovered or old trends are disproved. The cloud’s vast resource capabilities process heavier computing algorithms to predict potential issues related to component health, helping to improve root cause analysis of component failures.” This Big Data management concept is not new, but mining companies have only recently started focusing on the information and required actions behind the data they gather. “The more data we can collect, and the more we can learn about how components operate and fail, the more opportunity to replace components based on their condition, rather than a fixed time interval. Many maintenance teams will replace components even though they are still in healthy condition and could have run for many more hours simply because the planned maintenance schedule dictates their replacement. While OEM guidelines specify replacement at certain intervals, actual operating conditions and duty cycle can often extend equipment lifetime well beyond those specifications. One company proved this when they ran an engine until its condition required replacement, rather than replacing it at the OEM-specified 18,000 hours. The company monitored critical parameters such as blow-by pressure to ensure optimum engine life extension without running the component to failure, and extended the engine life by an additional 22,000 hours.” Cloud computing’s massive storage capability also facilitates easier machine learning and analytics. Gone are the days that require an engineer to sift through huge raw data files; cloud computing feeds large datasets into intensive machine-learning algorithms that can identify trends, calculate predictions about component failures, identify components that are operating differently than other components with similar operating hours, and much more. “In the end, these complex algorithms and constantly- evolving data models provide maintenance departments with trusted information about component health, facilitating more planned maintenance instead of costly unplanned efforts.” An example of this cited by Modular comes from a large mine in Australia, which reduced its unplanned events by about 25% in roughly a year’s time after implementing an intelligent maintenance management system that utilises these complex algorithms. “Considering that unplanned maintenance is typically three to 10 times more costly than planned maintenance, the shift to a more planned schedule saved the mine significantly in equipment downtime, as well as in component and labour costs.” Cloud computing technology also facilitates remote maintenance management; by monitoring equipment and components remotely, maintenance technicians no longer have to inspect equipment on site. This remote capability allows maintenance personnel to diagnose a component issue before they go to repair it, arming them with all necessary tools