E1..En
|
Input
|
Value whose trend is determined. The number of inputs
corresponds to the parameter "Number of inputs/outputs".
|
|
|
|
T1..Tn
|
Trend
|
Trend of the associated input: value "rising" when the
value increases, "stable" when it stays within the deviation,
"falling" when it decreases.
|
|
|
|
|
Number of inputs/outputs
|
Number of channels (one trend output per input),
1 to 64. |
|
Trend logic
|
Defines how the trend is derived from the value over time. For all
methods: if the determined change over the time window is
≥ deviation the output is "rising", if it is
≤ −deviation "falling", otherwise "stable".
- Discrete sampling: at the end of each time window the
current value is compared with the reference value (value at the
start of the window) and the result is output; the current value
then becomes the new reference. The output only changes at the rate
of the configured time. Simple and easy to follow, but reacts only
at the end of the window.
Example: Time = 10 min, deviation = 0.5. Reference at the
start 21.0 °C, after 10 min 21.8 °C → difference +0.8 ≥ 0.5
→ output "rising" (1), and 21.8 °C becomes the new reference.
If the value rises only to 21.9 °C over the next 10 min →
difference +0.1 < 0.5 → "stable" (0).
- Gliding reference: the reference continuously follows the
input value slowly (each second it approaches by the fraction
1/time window). The trend is determined every second from the
difference between the current value and the reference. Reacts
continuously – not only at the end of the window. If the value stays
constant for a while, the reference catches up and the output goes
to "stable".
Example: Time = 10 min, deviation = 0.5. If the
temperature rises steadily, the trailing reference stays below the
current value; as soon as the gap exceeds 0.5 °C, "rising" is output
immediately and stays so while it keeps rising. If the temperature
then stays constant, the reference catches up within about 10 min
and the output switches to "stable".
- Linear regression: evenly spaced samples are collected over
the time window (up to 16) and the slope is calculated from them via
a best-fit line (least squares). This slope is projected onto the
window length and compared with the deviation. Averages over many
points and is therefore insensitive to single outliers.
Example: Time = 10 min, deviation = 0.5. The best-fit
line of the last 10 min has a slope of +0.1 °C/min → projected
onto 10 min = +1.0 °C ≥ 0.5 → output "rising". A single
measurement spike barely shifts the line, so the trend stays
stable.
|
|
Time
|
Length of the time window for the trend evaluation
(unit see "Time unit"). |
|
Time unit
|
Unit for "Time": seconds, minutes or hours. |
|
Deviation
|
Threshold (dead band) above which a change is rated as
rising or falling. Smaller changes result in "stable". |
|
Deviation type
|
- Absolute: the deviation is a fixed value in units of the input.
- Percent: the deviation is calculated relative in % of the current
value.
|
|
Value rising
|
Output value for a rising trend (default 1). |
|
Value stable
|
Output value for a stable trend (default 0). |
|
Value falling
|
Output value for a falling trend (default -1). |
|
Smoothing (s)
|
Smoothing of the input against measurement noise (time
constant in seconds). 0 = no smoothing. |
|
Minimum hold time (s)
|
A newly detected trend is held at least for this time
before it may change. Prevents flickering of the output. 0 = off. |
|
Save persistent
|
Stores the calculation state so the trend continues
after a restart of the controller. Saving is performed cyclically at
most once per hour and on shutdown. |
Trend