The promises for these near-miss analyses have been discussed in an On the Horizon article for CEP ( Oktem et al., 2013). With compacted data, new approaches for statistical Bayesian analyses to determine the failure probabilities of the safety systems, and the probabilities of plant trips (emergency shut-downs) and accidents, have been developed ( Meel and Seider, 2006 Pariyani et al., 2012b). Initial successes have been achieved in data mining, beginning with the storage of data in event trees and the compaction of their large databases ( Pariyani et al., 2012a). But, recently, there have been moves to analyze all of these data in thorough dynamic risk analyses that is, to analyze the wealth of near-miss data available during normal operation. Until the past decade, most of these data were disregarded, as plant operators and managers focused their attention on those alarms leading to plant shut-downs and accidents.
Plant data historians can record thousands of alarm system events (e.g., activation, deactivation, acknowledgement) daily. Oktem, in Computer Aided Chemical Engineering, 2014 2.2 Near-MissesĪlarms and existing safety systems provide excellent barriers for abnormal events – generally bringing the process to a safe condition and often back to normal operation (with alarmed variables returned to within their alarm thresholds). If you see the level rising over 10 percent, it likely indicates a problem, and you should investigate the source of the broadcast. ▪īroadcasts When a network is fully populated and traffic loads are within normal utilization percentages, overall percentage of broadcasts versus real network traffic should be low. If errors occur at a high rate on a regular basis, they can indicate symptoms that could soon lead to a more serious event such as a failed segment or a bad port on a router or a switch. These errors should rarely occur, but they do manifest themselves from time to time during normal operations. ▪Įrrors Errors consist of jabbers, FCS, short frames, late collisions, and more. ▪Ĭollisions A high level of collisions (a constant level between 5 percent and 10 percent of overall load) indicates that there is a possible problem with the media or that there are too many stations on a given segment. (There are many reported thresholds listed in various places, but this is the general guideline to follow.) If there are long stretches in which the average load is more than 60 percent, additional network segmentation or equipment such as routers and switches should be considered. Utilization From a network-planning perspective, typical constant network loads should hover between 40 percent and 50 percent.
In general, the following scenarios are applicable to typical networks operating under normal conditions: ▪ You might want to reconfigure the alarm thresholds to better meet the characteristics of your specific network. You should look at a couple of standard events when you're troubleshooting Ethernet networks.
0 kommentar(er)
