Hydrocarbon Processing - April 2022 - 87

Process Controls, Instrumentation and Automation
Use Case 4: Diagnosing cause and
effect to find the needle in the haystack.
Poor performance is sometimes
isolated to single controllers, and the
need for retuning or valve maintenance
can be quickly determined. In more
complex situations, poor performance is
widespread, with a single problem propagating
across many controllers. Before
the problem can be addressed, the controller-which
is the root cause of the
variability-must be identified.
Root cause identification can be tedious,
particularly in processes with recycle
streams, heat integration and features
that mask cause and effect relationships.
Causality algorithms mathematically
quantify cause and effect relationships
among signals, revealing relationships
obscured on time series trends. Without
analytics to guide troubleshooting and
narrow the potential root causes, control
engineers may spend many hours
performing invasive process tests and inspecting
data trends.
In this use case, a plant-wide oscillation
lasting for about 2 hr persists through the
measurements and controllers designated
with blue and red circles (FIG. 6). The
dataset of oscillating controller process
variables (FIG. 7) is input to the causality
analysis. The directional arrows in the
generated causal map (FIG. 8) guide the
engineer toward controller LC2 causing
the related oscillations in TC1 and TC2.
Combining the causality results with process
dynamics and connectivity knowledge,
the engineer agrees that LC2 is a
suitable candidate for the root cause and
confirms this with additional plant testing.
Complementing analytics tools for
monitoring, diagnostic tools like causality
can streamline tedious investigations, enabling
the full range of control workflows.
Good things come in threes: Process
data + control engineers +
analytics. Exemplified by the four use
cases spanning single controllers to
many and analytics from monitoring to
diagnostics, advanced analytics is an invaluable
component for control and automation
success. By integrating control
engineer analytics within a comprehensive
and extensible analytics platform,
the collaborative value with other operations
personnel is greatly enhanced,
enabling a superior approach vs. nichefocused
control analytics packages. FeaT13
21.
Condenser
Condenser
22.
6.
FI1
4.
1.
PC1
2.
FC3
Steam
Condenser
7 & 8.
FC1
Feed
FIG. 6. Chemical process exhibiting plant-wide oscillations. Red circles designate cycling
measurements and blue circles designate cycling controllers.
P11 9.
T11
3.
LC1
5.
17.
LC3
T12
10.
FC4
12.
T14
Decanter
LC2
19.
FC5
18. F12
T15
11.
13.
TC1
T16
15.
FC6 14.
Steam
Condenser
Condenser
F15 20.
Product
FC8
23.
PC2 16.
Separator
Decanter
27.
26.
T18
T17
Condenser
F14 24.
F13 29.
TC2
25.
28.
PI2
30.
FC7
Steam
FIG. 7. Oscillating controller data over a 2-d period. With all controllers cycling
at a similar frequency, the oscillation source is not readily apparent.
ture-laden advanced analytics, applied
from a process control perspective and
utilizing the engineer's expertise and
process understanding, can be used to
quickly uncover process value hidden in
historian data and to accelerate process
control benefits.
JOHN COX is a Principal Analytics
Engineer with Seeq Corp.
He earned BS and MS degrees
in chemical engineering and
has an industrial background
in chemical process control.
FIG. 8. Causal map generated for the
dataset in FIG. 7 accurately guides the
user towards controller LC2 as the
root cause of plantwide oscillations.
Hydrocarbon Processing | APRIL 2022 87
Column 3
Reboiler
Column 2
Column 1
Condenser
Reboiler

Hydrocarbon Processing - April 2022

Table of Contents for the Digital Edition of Hydrocarbon Processing - April 2022