Automatic onset detection

The following section starts with the description of automatic onset detection methods, which are extremely useful when dealing with large data sets, but are the most error-prone of the localization algorithms. 

There are also general dilFerences in the type of localization procedure. Different localization techniques will be discussed after the automatic onset detection description, beginning with methods that provide only a rough estimate of the source coordinates, and building up to more complex methods that derive these coordinates with best possible accuracy. 

In the domain of acoustic emissions, Kurz et al. [2005] applied the principle of the autoregressive AIC picker in an automatic onset detection procedure. The results of the autoregressive AIC picker were then compared to manual picks and to an auto picker based on the Hinkley criterion, developed for acoustic emissions by Grosse and Reinhardt [1999]. Details about these approaches and the results can be found in Kurz et al. [2005]. In the following, only a short summary will be given. 

One of the advantages of this automatic system is that the state-of-health data recorded for the numerous subsystems, including blowers, component states, temperatures, and other critical information, allow remote failure detection, diagnosis, and possibly prevention. As an example, variation in detector temperature may show the onset of failure of a mechanical cooler. Remote diagnostics are used to schedule repair trips and minimize down time. 

MIDAS (Motorway Incident Detection and Automatic Signalling System) loops at 100m intervals (compared with 500m on other motorways) for more accurate and rapid response to the onset of congestion or incidents. 

A main advantage of the SEC detection is that it follows the evolution of the MWD, particularly important in living -type reactions and in copolymerization reactions where complex mixtures of reagents make unfractionated spectroscopic resolution of comonomers difficult. On the other hand, continuous detection provides a much more detailed characterization of the reaction and, interestingly, in the case where a bimodal MWD was produced, the continuous method automatically detected the onset of the second mode in a model independent fashion, whereas SEC conld only discern the bimodaUty by applying preconceived models. 

---