Stiction
is a well-known villain in industry because of the limit-cycle
imposed on the controller. Several methodologies are reported in the
literature to automatically detect this problem using only normal
operating data. However, this becomes more difficult when the loop
with stiction is affected by disturbances or the sticky valve is inside
a cascade loop. This study proposes two methods to automatically diagnose
valve stiction when the reference signal is variable and centers primarily
on recognizing triangular or sinusoidal patterns. The first method
is based on the slope of the signal peaks and the second on data segmentation.
These techniques were compared to a curve-fitting method, providing
similar results when the reference is fixed. However, for processes
significantly affected by disturbances or when the sticky valve was
inside a cascade loop, stiction detection was better for both methods
proposed. These results are corroborated by simulation and industrial
data.
Oscillation in control loops is a problem of high incidence, stiction in control valves being the most frequent cause. In the last two decades, many automatic stiction detection methods were proposed. These methods require certain signal preprocessing for proper use, where the choice of the appropriate preprocessing technique and parameters depends on the detection method and signal characteristics and, due to the lack of information about its automatic application, are mostly led manually. The need for user interaction turns an automatic into a nonautomatic detection method, which makes the application unfeasible to an entire plant with hundreds or even thousands of control loops. This work proposes techniques for automatic preprocessing according to the demands of each detection method. The techniques cover the main problems related to stiction identification: removal of noise excess, identification of peaks and valleys, and removal of variable mean and zerocrossing identification. Satisfactory results are obtained for both simulated and industrial data.
Oscillatory control loop is a frequent
problem in process industries.
Its incidence highly degrades the plant profitability, which means
oscillation detection and removal is fundamental. For detection, many
automatic techniques have been proposed. These are usually based on
rules compiled into an algorithm. For industrial application, in which
the time series have very distinct properties and are subject to interferences
such as noise and disturbances, the algorithm must include rules covering
all possible time series structures. Since the development of this
algorithm is near impractical, it is reasonable to say that current
rule-based techniques are subject to incorrect detection. This work
presents a machine learning-based approach for automatic oscillation
detection in process industries. Rather than being rule-based, the
technique learns the features of oscillatory and nonoscillatory loops
by examples. A model based on deep feedforward network is trained
with artificial data for oscillation detection. Additionally, two
other models are trained for the quantification of the number of periods
and oscillation amplitude. The evaluation of the technique using industrial
data with different features reveals its robustness.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.