View metadata, citation and similar papers at core.ac.

uk brought to you by CORE

provided by Sheffield Hallam University Research Archive

Performances Comparison between Real-Time Auto-

Tuning PID and Conventional PID Controller for a Dairy
Industrial Evaporation Process Control
MENG, Qingbo, ZHANG, Hongwei <http://orcid.org/0000-0002-7718-021X>
and HOWARTH, Martin
Available from Sheffield Hallam University Research Archive (SHURA) at:
http://shura.shu.ac.uk/24785/

This document is the author deposited version. You are advised to consult the
publisher's version if you wish to cite from it.
Published version
MENG, Qingbo, ZHANG, Hongwei and HOWARTH, Martin (2019). Performances
Comparison between Real-Time Auto-Tuning PID and Conventional PID Controller
for a Dairy Industrial Evaporation Process Control. In: 2019 International Conference
on Artificial Intelligence, Control and Automation Engineering (AICAE 2019).
DEStech Publications, Inc..

Copyright and re-use policy

See http://shura.shu.ac.uk/information.html

Sheffield Hallam University Research Archive

http://shura.shu.ac.uk
 Performances Comparison between Real-Time Auto-Tuning PID and
Conventional PID Controller for a Dairy Industrial Evaporation Process
Control.
Qingbo MENG1, Hongwei ZHANG1,* and Martin HOWARTH1
1
National Centre of Excellence for Food Engineering, Sheffield Hallam University, Sheffield, S1
1WB, United Kingdom
Keywords: Three-effect falling film evaporator, Modeling, Conventional PID control, Auto-tuning
PID control, Simulation.

Abstract. In this study, an industrial milk evaporation process is introduced and a mathematical
model of a multi-effect falling film evaporator is developed using MATLAB/Simulink to evaluate
the performance of the controller. A real-time closed-loop auto-tuning PID controller is presented as
a candidate control strategy for the evaporation process. The simulation results controlled by auto-
tuning and conventional PID are compared and discussed and the performance improvement by the
auto-tuning PID controller is illustrated.

Introduction
Evaporation in dairy industry is an essential process for concentrating milk, by evaporating excess
water to achieve the desired concentration and to prepare for the subsequent spray drying operation.
Additionally, evaporation, as an energy intensive process, can also help to reduce the energy cost in
milk powder production/drying [1]. During the industrial evaporation process, the milk
concentration (total solid content) is required to be evaporated from approximately 5-10% to 48-52%
before feeding into the spray dryer. However, food production processes present great challenges to
the engineers because it contains many characteristics which make the process particularly difficult
to model and control. These characteristics include dynamic behavior that is non-linear and highly
complex. In addition there are many key variables which are difficult to measure. Many advanced
control theories and methods have been developed and applied to the food industrial manufacturing.
For example, adaptive control, proportional–integral–derivative control (PID), robust control,
predictive control, Multivariate Statistical Process Control (MSPC) and intelligent control. In this
study, a conventional PID controller and an auto-tuning PID are developed and compared for a
three-effect falling film evaporation process.

Figure 1. Three effects falling film evaporator

Falling film evaporators (Fig. 1) are widely used for the concentration of solutions containing solid
particles in the dairy industry. It is normally applied as a part of the powder production process to
produce a very wide range of products, such as whole milk, skim milk and whey powder. However,
research studies on the modelling of multiple effects in falling film evaporators for milk powder
production is limited. A first-principles model of an evaporator was described by Winchester and
Marsh [2], based on thermodynamic physical laws. Runyon et al [3] developed a double effect
evaporator to check the product output concentration consistency with multivariable controllers.
Miranda and Simpson developed a dynamic multiple effect evaporation system model, determining
that the most important parameters of the evaporation process are the global heat transfer coefficient
and the latent heat of vaporization [4]. More recently, two types of dynamic models named
‘lumped’ and ‘distributed’ are developed for an industrial multi-effect milk evaporation process [5].
In this study, the three-effect falling film evaporator simulation model is developed by using
MATLAB/Simulink and based on the Newell and Lee’s evaporation model [6].
This study is organized into four sections: section 2 describes the mathematical modelling details
and the development of the Simulink model. Conventional PID and auto-tuning PID control
strategies are presented in section 3. In the final section, 4, simulation results from both controllers
are compared and discussed.

Mathematical Modelling of the Three Effect Falling Film Evaporator

The three-effect falling film evaporator includes:
(i) A separator, which separates the vapor and liquid phases when the evaporation process is taking
place. The concentrated product forms a level at the base of the separator in order to avoid flow
disturbances caused by the pump running dry.
The Separator Level (L) is calculated by equation (1) as described by Newell & Lee [6]:
𝑑𝐿 1
= (𝐹1 − 𝐹2 − 𝐹4) (1)
𝑑𝑡 𝜌𝐴

Where ρ is the liquid density, A is the cross-sectional area of the separator, and F1, F2, & F4 are the
feed, product and vapor flowrates.
(ii) Heat exchanger, which is where the main evaporation process takes place. The water is heated
and evaporated by contacting the outside tubes heated by the steam. Concentrated liquid drops
down to the bottom of the effect.
The concentration of the liquid is determined by equation (2) in [6]:
𝑑𝑋2 𝐹1 ∗ 𝑋1 − 𝐹2 ∗ 𝑋2
= (2)
𝑑𝑡 20

Where X1 and X2 are the feed concentration and product concentration respectively.
(iii) A condenser, which removes the condensed vapor from the system.
(iv) A steam ejector, which provides hot steam to heat the liquid in the heat exchanger.
The steam temperature (𝑇𝑠𝑡𝑒𝑎𝑚 ) is described by equation (3) in [6]:
𝑇𝑠𝑡𝑒𝑎𝑚 = 0.1538 ∗ 𝑃𝑠𝑡𝑒𝑎𝑚 + 90 (3)
Where, 𝑃𝑠𝑡𝑒𝑎𝑚 is the steam pressure.
There are also four equations which determine the mass and energy balances on both feed and
steam ejector process as follows [6]:
𝑑𝑀
= 𝑀𝑖𝑛 − 𝑀𝑜𝑢𝑡 − 𝑀𝑣 (4)
𝑑𝑡
𝑑(𝜌𝑠 𝑉𝑠 ) 𝜌𝑠 ∗𝑀𝑠𝑖𝑛 𝜌𝑠 ∗𝑀𝑠𝑜𝑢𝑡
= − (5)
𝑑𝑡 𝑣 𝑣
𝑑𝑇
= 𝑀𝑓 ∗ (𝑇𝑖𝑛 − 𝑇𝑠 ) + 𝐶𝑝 ∆𝑇 ∗ (𝑇𝑜𝑢𝑡 − 𝑇𝑠 ) (6)
𝑑𝑡
𝑑𝐻 𝑀𝑠 𝑀𝑠
= ∗ 𝐻1 + 𝑈 ∗ 𝐴 ∗ ∆𝑇 ∗ (𝑇𝑠 − 𝑇) − ∗ 𝐻0 ∗ 𝑉 ∗ 𝑄 (7)
𝑑𝑡 𝑉𝑠 𝑉𝑠

The key parameters are listed in the following Table 1.

Table 1. Main parameters of the falling film evaporator simulation
Parameters Description Value
F1 Feed Flowrate 10[kg/min]
F2 Product Flowrate 2[kg/min]
X1 Feed concentration 5[%]
T1 Feed Temperature 40[°C]

In this study, the milk feed concentration is 5% and the desired final product concentration is 52%.
Normally, the feed liquid product passes the pre-heater to reach the feed temperature before going
into the first effect, but the pre-heater simulation is not considered in this study and the feed
temperature is set as a constant with disturbances. Similarly, assuming that F1 and X1, are input
parameters which have up to 20% disturbances because of the time delay and heat loss during the
product flows from effect to effect. The three effects falling film evaporator simulation model,
which is based on the mass and energy balance equations, is developed using MATLAB/ Simulink
(Fig. 2). In each effect, the separator, heat exchanger, condenser and steam ejector are developed as
a sub-systems (Fig. 3) according to equations Eq. 1 to Eq. 3 above.

Figure 2. Three effects falling film evaporator MATLAB/simulink model

 Figure 3. MATLAB/Simulink model of one effect

Conventional PID controller and Auto-Tuning PID controller

During the last two decades, Proportional-Integral-Derivative (PID) controllers have been improved
and are now used in most industrial process control applications, as a control strategy, to regulate
many industrial variables, including flowrate, pressure, temperature and level. The two main
reasons that PID is commonly used are because of its robustness and ease of operation [8]. The
conventional PID controllers are firstly designed to control the three-effect falling film evaporator.
The set point is the target milk concentration, which is set to 52% in this study, after evaporating.
The product concentration from third effect is treated as a closed-loop feedback signal to calculate
the difference error. The PID controller is for product flow rate control in order to minimize the
error difference between set point and concentration feedback. A PID controller is designed and
tuned for each effect of the evaporator.
The auto-tuning PID works by injecting a test signal to collect the system input and output data for
estimating the frequency response (Fig. 4). After self-tuning, the controller calculates the updated
PID gains according to the data and then replaces the initial PID values.

Figure 4. Schematic of Auto-tuning PID controller

Simulation Results and Controller Performances Comparison
The three effect falling film evaporator for the industrial milk evaporation process mathematical
model is developed by using MATLAB/Simulink. The desired milk solid content is evaporated
from 5% to 52% in the three effects, where the first effect is from 5% to 30%, the second is from
30% to 38% and the third is from 38% to 52%. Running time for the evaporation process is 40
minutes. The simulation results and controllers performances are shown and compared below. The
conventional PID values are listed in table 2.
Table 2. Conventional PID values
P I
Effect 1 0.303 0.018
Effect 2 0.299 0.020
Effect 3 0.281 0.024

The simulation results controlled by conventional PID and auto-tuning PID are showing in Fig. 5 in
each effect. The blue signal, which presents the auto-tuning PID control performance, from 0 to 100
seconds is a self-tuning process. This is why there are large overshoots at the beginning. However,
once the auto-tuning process found the appropriate PID values and updated it to the controller, it is
obvious that once the milk concentration is settled the variations of outputs for each effect are much
smaller than when controlled by the conventional PID controllers. From table 3. The much smaller
variance values indicated better performance of the auto-tuning PID controllers and more accurate
control of the evaporating process.

Figure 5. Simulation results of each effect

Meanwhile, auto-tuning PID not only has better accuracy, but also better stability performance than
the conventional PID. The increasing variance values controlled by conventional PID means that
the more complex the plant is or the more effects the plant contains, the more difficult it is for the
conventional PID to maintain its performances. On the contrary, the variance values controlled by
auto-tuning PID are changing within a very limited range, which means that the auto-tuning PID is
more suitable than a convention PID controller for a complex system.
 Table 3. Variance values of each effect concentration output with the PID and Auto-tuning
controller
PID Auto-tuning
Effect 1 0.057 0.021
Effect 2 0.185 0.015
Effect 3 0.330 0.025

Conclusions
In this study, a three effect falling film evaporator for the milk industry is introduced. A
mathematical model is developed and simulated using MATLAB/Simulink. Two different control
strategies are applied in the model in order to achieve the control targets. The simulation results
shows that the auto-tuning PID controllers work better on more complex system, and are also more
accurate and stable than the conventional PID controllers. Further works need to be focued on other
advanced control strategies, such as Model predictive control, Fuzzy logic control, or a combination
of different control strategies, such as cascade PID, Fuzzy logic PID.

References:
[1] Q. Ruan, H. Jiang, M. Nian, Z. Yan, Mathematical modeling and simulation of countercurrent
multiple effect evaporation for fruit juice concentration, J. Food Eng. 146 (2015) 243-251.
[2] J.A. Winchester, C. Marsh, Dynamics and control of falling film evaporators with mechanical
vapor recompression, Chemical Engineering. 77 (1999) 357-371.
[3] C.H. Runyon, T.R. Rumsey, K.L. McCarthy, Dynamic simulation of a non-linear model of a
double effect evaporator, J. Food Eng. 14 (1991) 185-201.
[4] V. Miranda, R. Simpson, Modelling and simulation of an industrial multiple effect evaporator:
tomato concentrate, J. of Food Eng. 66 (2005) 203-210.
[5] B. Medhat, M.A. Fanaei, H. Zohreie, Mathematical modelling and dynamic simulation of multi-
effect falling-film evaporator for milk powder production, Mathematical and Computer Modelling
of Dynamical Systems. 21 (2014) 1-23.
[6] R.B. Newell, P.L. Lee, Applied Process Control, A Case Study, New York: Prentice-Hall, 1989.
[7] M. Karimi, A. Jahanmiri, Nonlinear Modeling and Cascade Control Design for Multi Effect
Falling Film Evaporators, Iranian J. of Chemical Eng. 3 (2006) 52-63.
[8] F.J. Kong, A. Van Cauwenberghe, Comparison between Conventional and Self-Tuning PID
Controllers for Mould Level Control, IFAC Proceedings Volumes. 28 (1995) 185-191.