In normal uses, most analyzers will require adjustment to their sensitivity to make up for a gradual reduction in sensitivity over time.
The normal amount of adjustment of the signal should be well understood by the maintenance personnel so that problems other than loss of sensitivity are not corrected by the improper use of the "span" adjustment.
Safety standards often require that, the analyzer must produce an alarm whenever the solvent concentration exceeds 60% LFL or some level that leaves a margin of safety. All analyzer systems should have provisions that prevent alarm levels without enough of a safety margin.
All alarms and deviations from expected operation are opportunities to determine if a fault exists in the process or analyzer system.
An alarm which activates when the rate of increase in solvent concentration exceeds the rate alarm level, can give early warning when a severe upset condition exists that is likely to eventually result in a very high solvent concentration.
Because the alarm triggers on a change in reading per second, and not a reading level, the rate alarm can activate at the very beginning of the analyzer's response time rather than several seconds or more later when the reading has finally climbed above the warning or danger alarm levels.
An analyzer system should be reliable. At minimum alarm actions include an annunciator to notify the operator with a loud, unique sounding, horn or buzzer at the "warning" level, and relays to initiate a shut-down at the “danger" level.
For transient conditions the response time of the analyzer is critical.
Upsets can cause the solvent concentration to reach 100% LFL within seconds. In the US, NFPA-86 indicates that five seconds’ response time could be required of the analyzer in order to make an effective alarm. Experience has shown that process upsets can produce solvent increases of 10% LFL per second or more.
The time lag in an analyzer’s response causes “dynamic error,” which is the instantaneous difference between the actual solvent concentration and the analyzer reading.
Common practice is to assume complete accuracy in the LFL values from whatever authority is recognized at the time. But, as we have seen, one can reasonably assign an uncertainty of 10% LFL to the initial LFL values, as indicated by the precision of the published LFL values and the amount of agreement between the various competent authorities. Temperature effects can account for perhaps an additional 10% LFL, if one chooses to use the lesser amount of correction and in a particular case, the greater amount of correction is appropriate.