Greg Stanley and Associates
 Home   About Us   Products   Services   Success Stories   Tech Resources   Contact Us 
HomeTech ResourcesFault DiagnosisFiltering > Spike Filter >

Spike Filtering

Filtering topics:


This page describes spike filtering. This page is part of the section on
Filtering that is part of
A Guide to Fault Detection and Diagnosis.

Overview of spike filtering

The purpose of a spike filter is to suppress extreme changes in measured variable values, since they probably don’t reflect actual changes in the monitored process. Small input changes are passed through without modification. For example, given existing pump capacity, it may be physically impossible for a drum level to increase or decrease more than a few percent in one time interval, unless there is a rupture in the drum or piping. Or even if it is physically possible, a flow might not normally change more than 10% within one data sampling interval. However, in case the extreme change remains in place n times in a row, then the change must be accepted as real. For example, there may have been a large change in setpoint, or loss of flow due to pump failure. Or even if the change is physically impossible, there may have been an instrument recalibration or sudden change in sensor bias that persists.

When using a spike filter, it should be placed first in the signal path, ahead of any other filter. That way, it can be tuned independently of the other filters. Other filters would dampen the spike to an extent dependent on their tuning. In normal conditions, operation of the other filters remains unaffected because the spike filter passes through routine small variations without modification.

A simple spike filter has two parameters, M and n. M is a maximum change parameter.  M is set so that in normal operations, the input does not change by as much as M from one time step to the next. The parameter n is the number of extreme changes in a row that will be rejected before finally accepting a large change, with possible values n=0,1,2,…

In this form, the filter stores its previous output and a counter c for the number of times in a row an input rate of change has been violated. The filter would be initialized to c = 0 and the input matching the output. 


An example implementation of a spike filter

For a “pseudocode” implementation example, we use the filter notation already introduced:

 IF ( | (x(k) – y(k-1) ) > M |  AND c < n  ) {      // recent big change: hold previous safe value
     c = c+1
     y(k) = y(k-1) 
} ELSE {               // normal operation, or else the recent big change must be real after all
  c = 0
  y(k) = x(k)
x(k) is the raw input at time step k
y(k) is the spike-filtered output at time step k
c counts how many times in a row the input values have been outside the legal range

Since extreme changes may result from failures, or result from sudden setpoint changes made during plant upsets (and sample intervals may be fairly slow), n will not typically be greater than 1 or 2, or diagnosis could be significantly delayed. The n= 0 case is included so that spike filtering may be easily turned off.

Copyright 2010 - 2013, Greg Stanley

Return to Filtering   Next: Filtering with Hysteresis: Dead Bands, Change Bands, and Latching

Return to A Guide to Fault Detection and Diagnosis


Share this page: Share this page via LinkedIn... Bookmark or share this page on Delicious... Share this page by e-mailing link...    

 Home   About Us   Products   Services   Success Stories   Tech Resources   Contact Us 
Enter keywords to search this site:
Now operated by Performity LLC