ALEXANDRE, G. B.; http://lattes.cnpq.br/7510601624292508; ALEXANDRE, Gerônimo Barbosa.
Abstract:
This work presents solutions for detecting, diagnosing and mitigating of failures in
industrial processes, the solutions were implemented in MATLAB-Simulink®.
The first solution is an approach for detection and identification of fault in
industrial processes type white-box, making use of techniques based models integrated
for anomaly detection, a decision logic with fixed thresholds to classify and identify
damaged component and mitigation failure uses a module fault tolerant controller which
the severity of the fault and its logical signal for reprogramming of the control law. The
proposed approach is applied to a level and temperature plant.
The second strategy intended to processes type black box, using the process history
for the monitoring of abnormal events. The detection, identification and reconstruction
process failures are computed by Principal Component Analysis (PCA); the detection is
done through τ2 e Q statistics; the identification is made by checking the contribution of
process variables in the model PCA at the time of occurrence of failure and the
reconstruction is carried out by the correction vector of samples, using the magnitude
value of the fault and the direction of the fault. The proposed approach is applied to
Fractionator column crude oil (Shell Process).
The third strategy describes a system of monitoring of abnormal events considering
the process energy (energy variables), in other words, the energy information contained
in measured signals. Its implementation is need to the energy signature of the process,
computing input, output and consume energy intake by the process, the energy
efficiency of unit and actions to optimize energy involved in procedural unit. This
system diagnose interact with statistical monitoring system (PCA) to reduce the
occurrence of false alarm or loss alarm monitoring system of the unit, giving more
reliability to operator of the process. This strategy was evaluated for an atmospheric
fractionator tower. It also proposes a methodology designer adaptive thresholds
operational changes, organized a strategic policy management alarms.
By the end of work will be presented the technical specifications of software fault
monitoring and alarms management. The software will gather all the strategies
developed in one product. The software should be integrated into the plant supervisory,
assisting in operation and routine maintenance of the plant supporting the operator
while minimizing unscheduled shutdowns.
The main contributions are: methodological plan thresholds for characterizing
adaptive, plan methodology for characterizing the initial model PCA, PCA as a tool for
project evaluation instrumentation, strategy for energy signature of process unit.