University of Zagreb, Faculty of Chemical Engineering and Technology, Laboratory for Automation and Measurement

Transcription

1 MEASUREMENT NOISE SIGNAL FILTERING

2 MEASUREMENT & PROCESS NOISE There is a certain level of random deviations in the around the actual value; These are called measuring noise or interference; In addition to measuring noise in some processes physical conditions varies due to turbulence, mixing, agitation, evaporation and the like. These are called process noise; Both phenomena cause random variations around. These variations are rapidly changing and random in nature, so they can not be reduced or eliminated.

3 NOISE AND SIGNAL FILTERING Measured signal consists of four components: 1. Controllable disturbances (controller can supress them) 2. Uncontrollable disturbances (too fast for controller) 3. Measurement noise 4. Stray electrical pickup Bandwidth of the controller is the highest frequency to which the controller can effectively respond; Disturbances with lower frequencies are controllable; Disturbance with higher frequencies are uncontrollable.

4 MEASUREMENT NOISE PROBLEM The noise is especially prevalent in measurements such as flow, level and weight gauges; Every analog signal also contains stray electrical pickup, primarily at either 50 or 60 Hz; When noise is present in the measured variable the controller transfers this noise to the controller output (): K C = 1 K C = 3 SP SP K c >1 amplify the noise

5

6 THE PROBLEM OF MEASURING NOISE MV Valve PROCESS CV Transducer CONTROLLER e SP Consequences of the measurement noise excessive movement of the valve; frequent changes in direction of the valve movement; valve wearing and damage.

7 Filter FILTER AND FILTER CONSTANT Low-pass filter attenuate t the components above a specified frequency (cut-off frequency) which controller can not supress; The most common filter in use is exponential filter (smoothing); In some application arithmetic moving average filter is used. It is referred as a moving average filter. Adjustable controller parameter is called filter constant t (time) - τ f Filtered

8 SIGNAL FILTERING How to reduce the noise? MV PROCESS CV Valve Transducer CONTROLLER e Filter SP Using adequate filtering we will end up with smoother signal and less valve movement. CAUTION! Excessive filtering can add delay in the loop and thus affect the dynamics of the system.

9 THE EXAMPLE OF SIGNAL FILTERING Raw signal Filtered signal

10 FILTERING - FLOW CONTROL No filtering τ f = 1,8 s K c = 0,3 τ i = 35,4 s τ d = 0 s Filter time of 1,8 seconds gives smoother signal and somewhat less movement of the valve.

11 FILTERING - FLOW CONTROL No filtering τ f = 6,0 s K c = 0,3 τ i = 35,4 s τ d = 0 s Filter time of od 6,0 second gives smoother signal and rather less movement of the valve.

12 FILTERING - FLOW CONTROL No filtering τ f = 18,0 s K c = 0,3 τ i = 35,4 s τ d = 0 s Time e( (min) Filter time of 18,0 seconds is too large and dampens signal dynamics. Consequently, the opening of the valve is practically constant!

13 FILTER ACTION WITIHIN FLOW LO IN A PLANT Filter time = 0 s Filter time = 3 s has measurement noise of approx. ±1 m 3 /h After introduction filter time of 3 sec (0,05 min) : - noise is reduced to ± 0,5 05m 3 /h, - the movement of the valve is reduced.