World Patents Pending ©AirSense Technology Ltd. 1999
ISSUE 2.4
Page 16
T E C H N I C A L
.
M A N U A L
n
The peaks indicated on the chart recorder may correspond to levels that should have
triggered an alarm.
n
The two weeks required for setting it up are inconvenient and are hence frequently
omitted or reduced, and the alarm trigger levels set at a ‘safe’ level (i.e., ‘safe’ from
unwanted alarms where the system is comparatively insensitive).
As with all very sensitive equipment, variations in the detector zero point and sensitivity
will occur, and these variations must be taken into account in the installer’s judgement,
when setting the trigger levels. An arbitrary element is contained in this procedure, and
it is unlikely that two different installers will set the same trigger levels.
An alternative method of setting the detector scaling is to determine that zero on the
measurement scale will correspond to the mean (average) of the background level of
pollution, and the full scale point set a given amount above the alarm trigger level. The
alarm trigger level with this method is set by performing a continual statistical analysis
of the background pollution and continuously calculating the level at which a given
probability of a signal occurring is acceptably low. Such measurements and calculations
are well within the capability of the modern micro-processor, and all that is required is
a suitable method. Since the zero and full scale points are set relative to the background
pollution levels, this makes the detector inherently more suited to its specific environ-
ment. It also makes the movements of the detector’s output for zero pollution
immaterial since it is constantly defining its own scale. However, since the scale is set
according to a history of constantly changing background environment, the method
of implementation must place a limit on the duration of the history. For example, the
background level readings taken thirty days ago are likely to be considerably less
significant than the readings taken thirty minutes ago. This method is basically an on-
going ‘learning method’, and to be successful it must ‘learn’ logically
.
There are a
number of minor problems with this method but all of them may be successfully
overcome by using a micro-processor, in fact the micro-processor becomes the centre
of the detector, governing all its procedures and actions.
Relative Scaling
