(404c) Online Fault Diagnosis During Startup of Distillation Process Using Artificial Immune System

Abstract

The startup is one of the most
important operations in a chemical process. It lasts a long period of time,
leads to off-spec products, and costs much energy. Due to its nature of phase
transition, large time delay and strong interaction between variables, it is
prone to fault. In the last decade, various process models and algorithms have
been developed for the fault diagnosis during startup of chemical processes^{[1, 2, 3, 4]}.

However, the establishment of most
traditional methods needs a certain number of training samples. However,
actually in the industrial processes, the samples of faults are quite limited.
When there are only small amount of training samples, the traditional methods
might not work very well. Since the data during startup of chemical process may
be much different between each time, the fault diagnosis models must have
strong self-adaptive ability to retrain themselves with minimal human
intervention.

In the recent years, many
researchers have begun to use the concepts from immunology to solve engineering
problems. A dynamic time warping (DTW) – based artificial immune system
(AIS) was proposed by the authors for the fault diagnosis^[5], in
which antibodies are generated by historical data and antigens are generated by
online real-time data. The system will detect and diagnose faults by
calculating the affinity of the antigen-antibody binding. Both antigens and
antibodies are represented by matrices of time-sampled data as figure 1 shows.
Also a cloning algorithm was proposed to generate the antibodies with variation
by the historical data. So the antibodies may cover different operational
status, and the system can diagnose the fault of chemical processes.

Fig.1. An antibody of the artificial
immune system^[5]

In allusion to the fault diagnosis
during startup of chemical process, we improved DTW-based AIS algorithm,
including dynamic selection of the size of antibodies and antigens, the cloning
algorithm of antibodies, etc. A scheme of the improved DTW-based AIS is shown
in figure 2.

Fig.2.  Flowchart
of improved DTW-based AIS

To develop an online fault
diagnosis system, a distillation process have been designed and installed. The environment
of online fault diagnosis system of distillation process is shown in figure 3.
A DCS system is used to ensure the stable operation of the distillation
process, and get the process data. A real-time database is combined with DCS
system to get and store the online data. And the antibody libraries are also
stored in the real-time database. The online fault diagnosis platform can
detect and diagnose the fault by the information from real-time database, and
show the real time diagnosis result to the operator.

Fig.3.  Environment of online fault diagnosis system of
distillation process

Four different faults were
introduced to the startup of distillation process to validate the online fault
diagnosis system. Such as: condenser leak, heating control failure, reflux
ratio increasing, and feed stopped.

Result shows that the proposed
system will reduce the false detection rate and improve calculation efficiency
during startup of distillation process. Compared with other methods, the
proposed method has better capability when the number of historical fault
samples is limited. When the system works, the antibody library is
automatically updated in background when the startup was finished or a fault
was diagnosed. So the system can learn online with an adaptive nature, and
become more and more intelligent.

References

1. N.J. Scenna (2000) Some aspects of
fault diagnosis in batch processes. Reliability Engineering and System Safety.
70.
pp. 95–110.

2. R. Srinivasan and M.S. Qian (2006) Online fault diagnosis and state
identification during process transitions using dynamic locus analysis. Chemical Engineering Science. 61.
pp.6109-6132.

3. M.J Olanrewajua., B. Huang, and A. Afacana (2010)
Onlinecomposition estimation and experiment validation of distillation processes with switching
dynamics. Chemical Engineering Science. 65. pp. 1597-1608.

4. K. Ghosh and R. Srinivasan
(2011). Immune-System-Inspired Approach to
Process Monitoring and Fault Diagnosis. Industrial &
Engineering Chemistry Research. 50.
pp. 1637-1651.

5. Y. Dai and J. Zhao (2011). Fault Diagnosis of Batch Chemical
Processes Using a Dynamic Time Warping (DTW)-Based Artificial Immune System. Industrial
& Engineering Chemistry Research. 50. pp. 4534-4544.