Page 233 - Mechanical Engineers' Handbook (Volume 2)

222 Data Acquisition and Display Systems

There are many techniques for compression of data. As mentioned above, many rely on
assumptions about the underlying nature of the data, such as being continuous data. Where
data are directly related to events, more traditional compression techniques which look for
repeating patterns in the data may be used. These are typically performed by operating
systems and database systems and can therefore be taken advantage of with little or no work
on the part of the engineer.

5 DATA STORAGE
In whatever ways data are sampled, collected, filtered, smoothed, and/or compressed, at
some point the data must be stored on some media if long-term data are to be analyzed
(covered later in this chapter). There are several approaches to data storage that will be
discussed in brief here.

5.1 In-Memory Storage
There are normally limitations on how much data can be stored, particularly when low-
frequency events have high-frequency data surrounding them that are of interest. For ex-
ample, if scientists are monitoring Mt. St. Helens for seismic data, it would be prohibitive
to capture millisecond data for years while waiting for an eruption. It would be of interest
to capture data at high density just before, during, and after each eruption but not in the
quiet times in the intervening years. Collecting the millisecond data on many sensors would
overflow the storage capability of most systems. There are techniques to store subsets of the
data that allow high-density data from constrained time intervals to be stored.
An approach for collecting and later reporting high-density data around an event of
interest is to collect the data continuously using the triggered snapshot method. High-speed
data are temporarily retained for a fixed time interval or memory capacity, with the start and
end time of the data moving forward with time. Older data are discarded as the time range
moves past it. This moving window is useful for creating trends and summary data for that
interval. The user can be shown dynamic displays that update over time and reflect char-
acteristics of the moving window of the process. Periodically, a set of the data can be
extracted to mass storage, especially triggered by some event of interest. An event is rec-
ognized by some means (automatic or user generated) that the engineer has preconfigured
to cause the transfer of the current instance of the moving window to permanent storage.
The relationship between the trigger and the moving window can be configured several
ways, as depicted in Fig. 5. The handling of the data involves moving values through a data
array, adding more recent values at the end and pushing the rest toward the beginning—a
queue.
The triggered snapshot is particularly useful when knowledge about the sequence of
events just before the event of interest can help discover problems. As an example in man-
ufacturing, in sawmills there are often very high speed sequences of events, such as where
a board may come out of one conveyor and is transferred to another conveyor and some
event such as the board leaving the conveyor occurs. High-speed video can be always in
progress, and the detection of the board leaving the system can be used to trigger the transfer
to permanent storage of the video. Events of concern can be safety issues and the triggered
snapshot method can be used to help eliminate potential life-threatening situations. The
triggered snapshot method is particularly useful for discovering the causes of unusual events.