Development of an algorithm for optimal process control based on artificial intelligence and a digital twin

1. Development of an algorithm for optimal process control based on artificial intelligence and a digital twin

Prepared by:
Stepanenko Ivan
Education program: Industrial and Systems Engineering
Supervisor: Samigulin T.I.
23MD0497 – 2 y.s.

2. Abstract

• In today’s rapidly changing industrial environment, improving
process control is crucial to improve productivity, reliability and
efficiency. One of the products that is widely used in the
technological processes of oil refining, methanol andammonia
production is Hydrogen. The heat of combustion of hydrogen is three
to four times higher than that of hydro carbon fuel. From this it can
be concluded that hydrogen can and should be used effectively as a
secondary energy carrier. Therefore, in order to improve and
optimize the process of obtaining hydrogen by the electrolytic
method, it is proposed to consider a possible algorithm that uses
digital twin technologies and artificial intelligence approaches. In
order to achieve optimal control of technological processes, this study
proposes to develop an algorithm using digital twin technology and
artificial intelligence (AI) approaches. Topic has been selected by the
subject of installing electrolysis systems for hydrogen production as
the focus of my research.

3. Problem statement

• The relevance of this topic lies in the fact that digitalization and robotization
are an integral part of our world at the present time. No factory operates
without automation of control systems, which ensure and guarantee quick
setup of operations in real-time mode. The digital twin is widespread in
many industries, for example, in production. Suppose they want to install a
new temperature sensor to the existing system in production to be able to
see the heating of the substance in the tank, but it is impossible to predict
the behavior of this system with a new sensor. Since in order to implement
and test this sensor, it will be necessary to spend a lot of time and stop
production. And accordingly, the company’s plant will be idle and will not
bring the desired result to the management. There are also systems with
embedded AI that will adjust the technological parameters, depending on
the current state of the system. For example, thermistors with an integrated
AI board can be used to regulate the temperature in boilers in production.
Thus, when the boiler temperature is high in the room and the room
temperature is high, the AI itself will decide whether to turn off the boiler
heating or continue to heat the boiler.

4. LITERATURE REVIEW

Given the dynamically developing digitalization and robotization in production, the use of a digital
twin and AI to optimize production processes at the enterprise has become quite common. There
are two main types of manufacturing helping tools: Digital twin and AI driven tool. After
conducting the analysis, the strengths and weaknesses of the digital twin and AI were identified,
based on this, it can be concluded that using a digital twin to optimize production processes will be
more effective than using AI.
One of the main advantages of the digital twin is the simulation of a real system in real time. This
helps analysts to produce a clearer and more constructive analysis, which in turn leads to the
most effective strategies for optimizing and managing the production system. In turn, AI training
requires extensive and high-quality data, which is not very convenient, since in many production
environments this data is limited. In turn, this can lead to inefficient operation of AI algorithms.
Also, do not forget that using a digital double allows you to perform analysis that is safe for the
physical system and personnel. This significantly reduces the likelihood of errors in the system
and unexpected shutdowns of this system. This, in turn, leads to savings in time and resources at
the enterprise. The use of AI can increase the risk of data security and confidentiality, which is
associated with possible cyber attacks and information leaks. This can create a significant threat
to business. One of the important advantages of the digital twin is that it allows you to create
hybrid systems that combine the advantages of digital modeling and the real world.
Unfortunately, many AI models at the moment cannot boast of their flexibility. They may be too
harsh or insufficiently effective, given new requests and requirements. Also, the principles of AI
algorithms can be difficult for humans to understand, which will make it difficult to interpret the
decisions made. This is especially noticeable when mandatory compliance with certain regulatory
standards is required. Based on all of the above, it can be concluded that using a digital twin is a
more convenient and optimal option for optimizing production processes.

5. Methodology

• In this section, we will consider the adaptive control algorithm for
optimizing the process of obtaining hydrogen by the electrolytic method in
more detail. This digital twin algorithm is an adaptive control algorithm. It
is a continuous analysis of the state of the system. To do this, it uses various
sensors that are installed on the electrolyzers and, if any deviations are
detected, this algorithm adjusts the system parameters in real time to
troubleshoot problems. There are several key steps in this algorithm, if
followed, this algorithm will work optimally. Firstly, it is necessary to collect
data from sensors in order to establish certain system parameters for this
algorithm. Next, data is analyzed using AI to detect various deviations from
the set optimal values. In the future, with the help of a digital twin, possible
scenarios for the development of events are being simulated, which in turn
will allow adjustments to the parameters of the system. After analyzing this
algorithm, it can be argued that its main advantages are reduced energy
consumption, since this algorithm selects the most optimal sensor values
(temperature, voltage, current, electrolyte concentration), it also prolongs
the service life of the equipment and increases the efficiency of the hydrogen
production process by electrolysis.

6.

7. RESULTS

• From the figure below, you can see 1st order model with the current iel as
input and the voltage vel as output can be described by the RC circuit given,
and corresponds to the following state-space model:

8. RESULTS

• A continuous-time transfer function for electrolyzer Fh(s) is considered as
model in the following general form below. Fh(s) = Vel(s) Iel(s) Where Vel(s)
and Iel(s) are the Laplace transforms of vel(t) and iel(t), respectively. The
block diagram in Figure 1 allows to show the Transfer function of the
electrolyzer F(s).
• Controlled variables: F(water), X. Possible uncontrolled indignation: D, Kt.
Ability control efforts: L, P, T. This control plant shows the controlled
variables, possible uncontrolled indignations and ability control efforts
which are used to perform Hydrogen extraction. For better understanding of
the stage of water jet cutting there is a Table 1 which contains information
about the main technical parameters of the control object.

9.

10. Main parameters of the control plant

11.

12. PID regulation tuning

• As we can see, the temporal behavior of the output data of the general
system, namely the transfer function and the step response with a value of
0, have a lag of 5 seconds in time. Which are normal conditions.

13. CONCLUSION

• The results from Matlab simulation helped to construct the PID regulator
for the system with optimal parameters. The obtained value of PID are
showed below. With this parameters system is controlled more smoothly and
accurately by 80%. Furthermore, the savings of the energy increased by
70%.

14. Reference

15. THANK YOU FOR YOU ATTENTION!