Siemens Wind Competition 2017
Investigate the codes leading up to the visit for patterns in seeing either particular codes or a specific pattern of code sequence. Ideas: look at between/among different sites, different groups of turbine types, seasonality indications, time of day the visit occurs (within vs outside expected working day hours).
Generate common sequences or paths among the all alarm sequences. Can visits be segmented according to these common paths?
Investigate commonalities by various factors: time of day of visit start, quantities of codes associated with the visit, what percent of total turbines at the same Park were visited on the same day, similarity of type or pattern of associated codes, length of the visit (very short, <5 minutes vs 25 minutes), and/or other various provided Factors.
What variables are significant indicators of visit duration or visit segmentation (from task 1 above)?
Consider not just patterns and occurrence of codes, but also the timing of the codes relative to each other. Your analysis should distinguish between a code sequence spread out over many hours vs. just a few minutes. Remember, some codes occur on every turbine on a periodic basis as a matter of normal operation; so the occurrence of a code may, or may not, be relevant to the visit.
Consider breaking the analysis into codes and patterns occurring prior to the Visit code, separately from those occurring on or after the Visit code starts. Classifying visits in groups or clusters with similar “before visit” behavior vs. “after visit” behavior can provide valuable insight. There may be codes during the visit which indicate similar work was performed, even if/when no pattern is apparent in the codes leading up to the visit, and vice versa. Be able to capture a similarity, even if it only exists in the pre-visit codes or in the during-visit (or after-visit) codes.
Take special note of many turbines at a Park being visited on the same day. It may be useful to specifically seek out visits occurring on the same day at a site, as this could be an indication of a common task or cause. On the other hand, there may be a pattern or oddity noted in investigating only “one-off” visits (where only one turbine experienced a visit at a Park in a single day) vs. the occurrence of multiple visits to various turbines at a Park in a day.
Not every instance of a local/remote switch signal input might actually be caused by a physical turning of the switch at the turbine. It is possible that some sort of electrical glitch could cause the switch input signal to latch in, when a technician is not actually activating the input. Absence of additional turbine codes in the hours proceeding or during the visit could indicate this. A second clue to this situation would be observing many visit starts occurring simultaneously (or nearsimultaneously) across a significant portion of turbines at any specific Park (it’s unlikely that a limited # of teams would be at every turbine at once and/or putting the turbine into local mode simultaneously). A third clue to this situation could be observing abnormal Visit Start Times outside of what is normally seen at a Park -- generally before 6AM or after 8PM. Siemens is particularly interested in identifying any code “signature” which indicates visits could fall into this category of “potential electrical glitches”.