processes

Article
Optimization Design of Centrifugal Pump Flow Control System
Based on Adaptive Control
Yuqin Wang * , Haodong Zhang, Zhibo Han and Xiaoqiang Ni

School of Mechanical Engineering, Chaohu University, Chaohu 238000, China; [email protected] (H.Z.);
[email protected] (Z.H.); [email protected] (X.N.)
* Correspondence: [email protected]; +86-182-5655-6898

Abstract: In this paper, in order to improve the control characteristics of the centrifugal pump flow
control system, a mathematical model of the centrifugal pump flow control system was established
based on an analysis of the basic structures, such as the frequency converter, motor, and centrifugal
pump. Based on the adaptive control theory, the recursive least squares algorithm with a forgetting
factor was used to estimate the real-time parameters of the centrifugal pump control system, and the
self-tuning PID control method was used to optimize the mathematical model of the centrifugal pump
flow control system. The simulation results showed that the adjustment time of the optimized system
was shortened by 16.58%, and the maximum overshoot was reduced by 83.90%, which improved the
rapidity and stability of the transient response of the system. This showed that adaptive control had
a significant effect on improving the robustness and anti-interference ability of the centrifugal pump
control system. In order to further verify the accuracy of the self-tuning PID control method, a flow
adaptive control system test platform was built. The test results showed that under the conditions of
constant frequency and variable frequency, the actual flow rate of the centrifugal pump was always



kept near the set flow rate, the error was small, and it had better real-time followability. The research


results showed that adaptive control could revise the parameters in real-time according to changes
Citation: Wang, Y.; Zhang, H.; Han,
to the centrifugal pump control system, which improved the stability and robustness of the system.
Z.; Ni, X. Optimization Design of
Therefore, adaptive PID control could effectively improve the adaptability of centrifugal pumps to
Centrifugal Pump Flow Control
various complex working conditions and improve the working efficiency of centrifugal pumps.
System Based on Adaptive Control.
Processes 2021, 9, 1538. https://
doi.org/10.3390/pr9091538
Keywords: centrifugal pump; adaptive control; self-tuning PID; simulation analysis

Academic Editor: Arkadiusz Gola

Received: 31 July 2021 1. Introduction

Accepted: 26 August 2021 Pumps are widely used general machinery, mainly used in all aspects of production
Published: 29 August 2021 and life [1]. Centrifugal pumps account for a larger share of the total pump usage. There
has been various research about centrifugal pumps recently. Gülich J. F et al. [2] predicted
Publisher’s Note: MDPI stays neutral hydraulic design information, the turbine performance curve, torsional rotor vibration,
with regard to jurisdictional claims in partial load flow and hydraulic excitation force by studying the interaction between cen-
published maps and institutional affil-
trifugal pump fluid flow and vibration and material wear. Lobanoff VS et al. [3] accurately
iations.
predicted the service life of the pump by studying the vibration and noise characteristics
of the centrifugal pump, and performed rotor dynamics analysis and pump vibration
diagnosis. L. Gevorkov, A et al. [4] proposed to simulate the flow control system of a cen-
trifugal pump based on Simulink software, developed a specific centrifugal pump model,
Copyright: © 2021 by the authors. and carried out a series of simulations. With the advancement of science and technology
Licensee MDPI, Basel, Switzerland. and the development of intelligent control technology, research into automation control is
This article is an open access article also deepening, and the degree of industrial automation is also constantly improving. In
distributed under the terms and
industrial production, how to realize the automatic control of the output flow of centrifugal
conditions of the Creative Commons
pumps is one of the important research directions of scientific and technological workers.
Attribution (CC BY) license (https://
Resolving this problem reasonably and effectively will save a lot of manpower and material
creativecommons.org/licenses/by/
resources and bring huge economic benefits.
4.0/).

Processes 2021, 9, 1538. https://doi.org/10.3390/pr9091538 https://www.mdpi.com/journal/processes

Processes 2021, 9, 1538 2 of 15

Processes 2021, 9, 1538 2 of 14

entific and technological workers. Resolving this problem reasonably and effectively will
save a lot of manpower and material resources and bring huge economic benefits.
Centrifugal pump
Centrifugal pumpflow flowcontrol
control refers
refers to the
to the regulation
regulation of theofoutput
the output
flow offlow of cen-
centrifugal
trifugal pumps in industrial production, so that the output
pumps in industrial production, so that the output flow reaches a certain value set by flow reaches a certain value
the
set by the user. The output changes according to the rules
user. The output changes according to the rules set by the user to meet the normal use set by the user to meet the
normal
in use inproduction.
industrial industrial production.
In practice, the In practice,
externalthe external environments
environments of centrifugal of pumps
centrifugalare
pumps areThe
uncertain. uncertain. The characteristics
characteristics of the transmissionof the transmission
medium willmedium will also be
also be disturbed in dis-
the
turbed in theprocess,
transmission transmissionresultingprocess,
in theresulting
output flow in theof output flowpumps
centrifugal of centrifugal
changing pumps
with
achanging
change in with a change
working in working
conditions. Atconditions. At present,
present, centrifugal centrifugal
pump pumpin
flow control flow control
industrial
in industrialisproduction
production mostly regulated is mostly by regulated
manpower, bywhich
manpower, which is
is inefficient and inefficient and slow
slow in response,
in response,
which which is not
is not conducive conduciveproduction
to reducing to reducing production
costs costs and
and improving improving
economic eco-
benefits.
Therefore, manyTherefore,
nomic benefits. scholars have manystudied
scholarsthe haveproblem
studiedofthe centrifugal
problem of pump flow control.
centrifugal pump
Sang-Ho Suh et
flow control. al. [5] pointed
Sang-Ho Suh et out that
al. [5] the useout
pointed of variable
that the frequency regulation
use of variable technology
frequency regu-
in flowtechnology
lation control could reduce
in flow energy
control couldconsumption
reduce energy to a consumption
certain extenttoand improve
a certain the
extent
service life of centrifugal
and improve the service pumps. Mingxin Chen
life of centrifugal pumps.et al. [6] usedChen
Mingxin fuzzyetPID control
al. [6] used to adjust
fuzzy PIDa
centrifugal pump through frequency conversion, which improved
control to adjust a centrifugal pump through frequency conversion, which improved the the robustness of the
flow controlofsystem,
robustness the flow reduced
controlovershoot,
system, reducedshortened the adjustment
overshoot, shortened time theand enhancedtime
adjustment the
anti-interference capability. Lu Jun et al. [7] used the H∞ method
and enhanced the anti-interference capability. Lu Jun et al. [7] used the H∞ method to to control a centrifugal
pump,
controlso a that the centrifugal
centrifugal pump, so pump
that flow rate could change
the centrifugal pump flow in real-time
rate could withchange
the external
in re-
environment,
al-time with the and the following
external error was
environment, andsmall.
the following error was small.
In
In the
the current
currentindustrial
industrialcontrol
controltheory,
theory,the adaptive
the adaptive control
control theory
theoryhashasthe the
advan-ad-
tages of eliminating the uncertainty of the controlled object,
vantages of eliminating the uncertainty of the controlled object, having good robustnesshaving good robustness to
disturbance and being suitable for frequently changing working
to disturbance and being suitable for frequently changing working conditions. In this conditions. In this paper,
the adaptive
paper, control control
the adaptive theory is proposed
theory to optimize
is proposed the centrifugal
to optimize pump flow
the centrifugal pump control
flow
system. It can make the centrifugal pump flow control system
control system. It can make the centrifugal pump flow control system more adaptable to more adaptable to frequently
changing
frequentlyworking
changing conditions. This will reduce
working conditions. the influence
This will reduce theofinfluence
the environment on the
of the environ-
output flow and improve work efficiency.
ment on the output flow and improve work efficiency.
2. Modeling of Centrifugal Pump Flow Control System
2. Modeling of Centrifugal Pump Flow Control System
2.1. Structure Design
2.1. Structure Design
The centrifugal pump flow control system consists of a centrifugal pump, controller,
The centrifugal
frequency converter,pump
motorflow
and control
flowmeter,system consistsinofFigure
as shown a centrifugal
1. When pump, controller,
the centrifugal
frequency converter, motor and flowmeter, as shown in Figure 1. When
pump works, the flowmeter monitors the centrifugal pump output flow in real-time and the centrifugal
pumpitworks,
feeds back tothe
theflowmeter
controller. monitors the centrifugal
The controller compares itpump output
with the flowvalue.
set flow in real-time and
The motor
feeds current
input it back frequency
to the controller. The controller
can be changed compares
by controlling it with the
the inverter. set the
When flowmotor
value. The
speed
motor input
changes, currentflow
the output frequency can be changed
of the centrifugal pump canby controlling
be adjusted.the inverter.
Through theWhen the
real-time
motor speedofchanges,
monitoring the flow
the output output andflow
theofreal-time
the centrifugal pump
regulation of can be adjusted.
the motor, Through
the automatic
the real-time
speed monitoring
regulation of the output
of the centrifugal pump flow and the
impeller real-timeand
is realized, regulation of the motor,
the automatic the
regulation
automatic speed regulation of the centrifugal pump
control of the centrifugal pump output flow is completed [8,9]. impeller is realized, and the auto-
matic regulation control of the centrifugal pump output flow is completed [8,9].
u(k) Frequency y(k)
Centrifugal
Controller Motor
converter pump

Flowmeter

Figure
Figure 1. 1. Structure
Structure diagram
diagram ofof centrifugal
centrifugal pump
pump flow
flow control
control system.
system.

2.2. Frequency Converter Mathematical Model

The speed regulation of AC
AC motors
motors mainly
mainly depends
depends on
on the
the frequency
frequency converter
converter to
to
change the frequency of the input
input current
current of
of the
the motor.
motor. The control flow isis shown
shown in
in
Figure 2.
Processes 2021, 9, 1538 3 of 14

Figure 2. Flow chart of frequency converter control.

Since the time constant of the frequency converter is much smaller than the time
constant of the asynchronous motor, the dynamic response of the frequency converter can
be ignored, and the mathematical model can be simplified [7]. The relationship between
the input voltage uc and the input current frequency f 1 of the frequency converter is
as follows:
f 1 = kint × uc . (1)
In Equation (1), f 1 is the power frequency, kint is the voltage frequency coefficient, and
uc is the input voltage of the frequency converter.
The value of uc is [0 − 10]V, the value of f 1 is [0 − 50]Hz, and kint = 5 Hz/V [10].
The relationship between the induced electromotive force E1 and stator phase voltage
u1 in a steady state is [11]:

u1 = E1 + I1 ( R1 + j2π f 1 L1 ). (2)

In Equation (2), I1 is the stator current, L1 is leakage inductance of each phase of stator
winding, and R1 is the stator winding resistance.
When the motor works at a low frequency, the torque of the motor decreases greatly,
so the compensation voltage u2 is required, and its relation formula is [12]:

220 − u2
u1 = × kint × uc + u2 . (3)
50
In general, u2 is small and can be ignored. The above formula can be changed into:

u1 = kint × k f × uc . (4)

In Equation (4), k f is the conversion factor.

According to the above formula, the system control model of the frequency converter
is a proportional link, namely:

u1 ( s )
G1 (s) = = kint × k f = 22. (5)
uc (s)

2.3. Motor Mathematical Model

Asynchronous motors generally have two inputs: voltage and frequency. In the
process of variable frequency speed control, the two inputs need to be adjusted at the
same time. The output of the asynchronous motor generally includes the motor speed and
magnetic flux. The magnitude of magnetic flux determines the motor’s torque, which is a
very important parameter in the motor model.
Processes 2021, 9, 1538 4 of 14

The motor model system is relatively complex and integrates high-order, nonlinear
and strong coupling. In order to reduce the calculation difficulty, the influences of magnetic
circuit saturation, core loss, space harmonics and temperature change are ignored. Accord-
ing to relevant knowledge, the motor model is simplified as an electrical link connected in
a series with a mechanical link [13]. The transfer function of the motor system is:

kL
G2 (s) = . (6)
( TL s + 1)( TS s + 1)

In Equation (6), TL is the mechanical time constant of the motor, TS is the electrical
time constant of the motor, and k L is the motor gain factor.
In this paper, TL = 0.012s, TS = 0.002s, k L = 1.

2.4. Centrifugal Pump Mathematical Model

The centrifugal pump is a kind of pump that relies on a centrifugal force generated
by a motor to drive the impeller in order to transfer liquid. Since the working state of
the centrifugal pump is very complicated, the following assumptions are made when
describing the centrifugal pump mathematical model [11]:
(1) The pulsation of the centrifugal pump is ignored;
(2) The leakage state of the centrifugal pump is laminar flow;
(3) The dynamic process of the centrifugal pump is neglected.
Assuming that leakage is neglected, the actual output flow of the centrifugal pump
is equal to the theoretical output flow. The speed formula of the asynchronous motor is
as follows:
n = 60 f 1 (1 − s1 )/Pn (7)
In Equation (7), f 1 is the power frequency of the asynchronous motor, s1 is the slip
rate, and Pn is the pole logarithm of the asynchronous motor.
Since the output flow rate Q1 of the centrifugal pump is proportional to the rotating
speed [14], the transfer function of the centrifugal pump mathematical model is as follows:

Q1 ( s )
G3 (s) = = k. (8)
n1 ( s )

2.5. Control System Mathematical Model

Combining Equations (5), (6) and (8), and setting k = 1 in the centrifugal pump
mathematical model, it can be seen that the mathematical model of the flow control system
is as follows:
k × k L × kint × k f 2, 200, 000
G (s) = = 2
. (9)
( TL s + 1)( TS s + 1) 2.4s + 1400s + 100, 000

Since the system is regulated by unit feedback H (s), the closed-loop transfer function
of the system is:

G (s) 2200, 000

Gk (s) = = . (10)
1 + G (s) H (s) 2.4s2 + 1400s + 2, 300, 000

In Equation (10), H (s) = 1.

3. Flow Adaptive Control Method for Centrifugal Pump

3.1. Control Scheme
Adaptation is an inherent feature of organisms, which means that when the living
environment changes, organisms can change their habits to adapt to environmental changes.
Adaptive control is a theory and method of controller design that applies the adaptive
characteristics of organisms to controller design. By refining this adaptive feature of biology,
 3. Flow Adaptive Control Method for Centrifugal Pump
3.1. Control Scheme
Adaptation is an inherent feature of organisms, which means that when the livin
environment changes, organisms can change their habits to adapt to environmenta
Processes 2021, 9, 1538 5 of 14
changes. Adaptive control is a theory and method of controller design that applies th
adaptive characteristics of organisms to controller design. By refining this adaptive fea
ture of biology, the controller of the conventional feedback system is used as the inne
the controller
loop, and thenof the
theconventional feedback
adaptive outer loop system is usedThe
is designed. as the inner loop,
adaptive and system
control then theis show
adaptive outer loop
in Figure 3 [15]. is designed. The adaptive control system is shown in Figure 3 [15].

Figure3.3.Structure
Figure Structure diagram
diagram of adaptive
of adaptive control
control system.
system.

ItItcan
canbebe seen
seenfrom Figure
from Figure3 that the adaptive
3 that control
the adaptive systemsystem
control is composed of modelof mode
is composed
parameter identification, a controller parameter design, an adjustable parameter
parameter identification, a controller parameter design, an adjustable parameter con controller,
a controlled object and a detection device. The actual controlled object is uncertain, so
troller, a controlled object and a detection device. The actual controlled object is uncer
the adaptive control system uses a controller design model with uncertain parameters
tain, so the adaptive control system uses a controller design model with uncertain pa
to describe the controlled object. An adaptive controller with adjustable parameters is
rametersastothe
designed describe the of
inner loop controlled
the adaptiveobject. An adaptive
controller. controller
The outer loop is with adjustable pa
an adaptive
rameters is designed
identification control loopasforthetheinner loop parameter
controller of the adaptive
design.controller. The outer
The main function of theloop is a
adaptive
outer loop isidentification
to measure many control
kindsloop for the controller
of information, parameter
such as the design. The
external interference, main func
input
ution of the
(t), the outer
output y(t)loop is to
and the measure
tracking many
error kinds ofthe
e(t) between information,
reference inputsuch
w(tas the the
), and external in

terference, of
environment input
() , the
the controlled output
object
()y t
and the and
and its own instability tracking
()
actual output of the controlled object in real time without interruption, so as to master the
u t e t
error Based onbetween
uncertainty. the th

reference
control decision
()
obtained object information, according to a certain adaptive theory design method, the
w t
inputis made ,toand the actual
update output parameters,
the structure, of the controlled object
or control in realoftime
function the withou
inner loop controller. Thus, in a sense, optimal or suboptimal control is realized, and a
certain desired control target is achieved.
In order to reduce the regulation times of parameters, the self-tuning PID controller
is selected to optimize the centrifugal pump flow control system. Its main feature is that
the PID controller is used as the parameter adjustable controller in the control system. The
self-tuning control theory is adopted to perform self-tuning control on the entire system,
reducing the number of adjustments. At the same time, the required parameters can
be modified in real-time according to changes to the object characteristics, thus enhanc-
ing the stability and robustness of the system and realizing the adaptive control of the
control object.
The self-tuning PID controller structure is mainly composed of four parts: the con-
trolled object, least square recursive parameter estimation [16,17], on-line design of con-
troller parameters and an adjustable parameter PID regulator [18,19]. The control structure
is shown in Figure 4.
 can be modified in real-time according to changes to the object characteristics, thus en-
hancing the stability and robustness of the system and realizing the adaptive control of
the control object.
The self-tuning PID controller structure is mainly composed of four parts: the con-
trolled object, least square recursive parameter estimation [16,17], on-line design of con-
Processes 2021, 9, 1538 6 of 14
troller parameters and an adjustable parameter PID regulator [18,19]. The control struc-
ture is shown in Figure 4.

Figure4.4. Structure
Figure Structurediagram
diagramof
ofself-tuning
self-tuningPID
PIDcontrol
controlsystem.
system.

3.2.
3.2. Description
Description of
of Controlled
ControlledObject
ObjectModel
Model
The
Thecontrolled
controlledobject
objectofof
self-tuning
self-tuningcontrol
controlcancan
be described by aby
be described discrete-time stochas-
a discrete-time sto-
tic linear model. The recursive least squares algorithm is used to calculate the model.
chastic linear model. The recursive least squares algorithm is used to calculate the model. The
discrete-time stochastic
The discrete-time linearlinear
stochastic model is [20]:
model is [20]:

AA((zz−−1 )1 y)y((kk)) ==BB((zz−−11)u)u((kk−−dd))++εε((kk)). (11)

(11)
In Equation (11), u(uk()kis) the input variable of the controlled object, y(k) is ythe
(k )output
In Equation (11), is the input variable of the controlled object, is the
variable of the controlled object, ε(k) is white noise, and d is the number of output delays
of the controlled
output variable object.
of the controlled object,
ε (k ) is white noise, and d is the number of
Setting
output theofsampling
delays TS = 0.01s, the centrifugal pump control system Gk (s) is
periodobject.
the controlled
discretized by the backward difference = 0.01sto obtain:
T method
Setting the sampling period S , the centrifugal pump control system
−
Gk ( s ) −1
1
1.007z + 0.05569
k ( zbackward
is discretized byGthe ) = −2 difference−method .
to obtain: (12)
z + 0.1079z 1 + 0.002928
−1
1.007 z + 0.05569
To set the output delay z −1 ) = d−2= 1, the random
Gk (number linear model disturbed by white
(12)
noise is: z + 0.1079 z −1 + 0.002928
To set(1the
+ 36.85z −1
output + 341.53z
delay −2
number )y(dk)==1 ,(19.02 + 343.92z
the random −1
linear )u (k − 1disturbed
model ) + ε(k). by white
(13)
noise is:
In Equation (13), ε(k ) is constant interference.
(1 + 36.85 z −1 + 341.53z −2 ) y(k ) = (19.02 + 343.92 z −1 )u (k − 1) + ε (k ) (13)
3.3. Real-time Parameter Estimation Algorithm
The forgetting
In Equation (13),
ε (krecursive
factor ) is constant
least squares method is used to estimate the real-time
interference.
parameters of the centrifugal pump control system [21]. Equation (11) is equivalently
converted into:

y ( k ) = G ( z −1 ) u ( k ) + H ( z −1 ) ε ( k ). (14)
In Equation (14), G (z−1 ) = B(z−1 )/A(z−1 ) is the system model, and H (z−1 ) = 1/A(z−1 )
is the noise model. It can be expressed as:

y ( k ) = ϕ T ( k ) θ + ε ( k ). (15)
 T
In Equation (15), parameter vector θ = a1 , a2 , · · · ana , b1 , b2 , · · · bnb .
The data information vector ϕ(k ) is:
ϕ(k) = [−y(k − 1), −y(k − 2), · · · − y(k − n a ), u(t − d − 1), u(t − d − 2), · · · u(t − d − nb )] T (16)
The performance index is as follows:

L h i2
J= ∑ λ L−K y(k) − ϕ T (k)θ̂ . (17)
K =1
Processes 2021, 9, 1538 7 of 14

In Equation (17), λ is the forgetting factor (0 < λ ≤ 1).

The algorithm expression of forgetting factor recursive least squares parameter esti-
mation is as follows [22]:

T
 
 θ̂ (k ) = θ̂ (k − 1) + K (k) y(k) − ϕ (k)θ̂ (k − 1)


−1
K (k) = P(k − 1) ϕ(k) λ + ϕ T (k) P(k − 1) ϕ(k)]

. (18)

1
P ( k ) = 1 − K ( k ) ϕ T ( k ) P ( k − 1)

  
λ

PID control is adopted for the system. In order to eliminate constant disturbance,
the controller must have an integral link [23]. The corresponding PID controller can be
expressed as:

F1 (z−1 )u(k) = R(z−1 )yr (k) − G (z−1 )y(k). (19)

In Equation (19), F1 ( z −1 )
= F (z−1 )∆.
Further:
−1 −1 −n f
 F (z ) = 1 + f 1 (z ) + · · · + f n f (z

 )
− 1 − 1
G ( z ) = g0 + g1 ( z ) + g2 ( z )− 2 . (20)

 − 1
R ( z ) = G ( 1 ) = g0 + g1 + g2
Equation (20) is arranged so that the closed-loop characteristic polynomial is the
denominator polynomial of the desired transfer function, thus obtaining:

AF + z−d BG = Am . (21)

4. System Simulation and Analysis

4.1. Parameter Determination
The standard form of the second-order system is:

ωn 2
G (s) = . (22)
s2 + 2ξωn s + ωn 2

The following results can be obtained by calculation: ξ = 0.2979, and the optimal
damping ratio is adjusted to ξ e = 0.707.
After adjustment, the expected transfer function is:

2, 200, 000
Ge (s) = . (23)
2.4s2 + 3322s + 2, 300, 000
The difference equation is:

(1 − 1624.92z−1 + 1026483.27z−2 )y(k) = (−197.29 + 980, 496.82z−1 )u(k) (24)

The initial value P0 = 106 I, θ̂ (0) = 0.001 and forgetting factor λ = 0.97 of the PID
controller are set to optimize the original system.
The parameter optimization process of the self-tuning PID control algorithm is
as follows:
(1) The initial data is inputted, and the initial values θ̂ (0), P(0) and the forgetting factor
λ are set;
(2) The current actual output of the acquisition system is y(k);
(3) The real-time parameter θ̂ of the control target is estimated by the recursive least
squares method with the forgetting factor through model parameter identification;
(4) Solving F and G in Equation (20), taking F1 = ∆F and R = G (1);
(5) Calculate the control quantity u(k) and regulate the system through Equation (19);
(6) Return to step (2) and continue the loop (k → k + 1) .
The object parameter estimation results are shown in Figure 5:
Processes 2021,Processes
9, 1538 2021, 9, 1538 8 of 14 9 of

400

Parameter estimation
a2
300
a1
200

100
0

−100
0 5 10 15 20 25 30 35 40
Number of iterations
(a) Parameter estimation
400
Parameter estimation

b1
300
b0
200

100

0
0 5 10 15 20 25 30 35 40
Number of iterations
(b) Parameter estimation
Figure 5. Object parameter estimation results.
Figure 5. Object parameter estimation results.
As can be seen from Figure 5, the object parameters are estimated to be a1 = 36.85,
a2 = 341.53, b0 = = 343.92.
b1 from a = 36.8
As19.02,
can beand
seen Figure 5, the object parameters are estimated to be 1
− 1 − 1
Then: Am (z ) = 1 − 1624.92z + 1, 026, 483.27z . − 2
a2 = 341.53 b0 = 19.02 b = 343.92
The optimized ,
difference equation is: 1
, and .
−1 −1 −2
Then: Am (−z1 ) = 1 − 1624.92 z + 1，026，
483.27 z .
(1 + 37z + 344.3z−2 )y(k) = (19.02 + 346.7z−1 )u(k). (25)
The optimized difference equation is:
The optimized system transfer function is:
(1 + 37 z + 344.3z −2 ) y(k ) = (19.02 + 346.7 z −1 )u(k )
−1
(2
201.1s + 171, 400
The optimizedGsystem
r ( s ) = transfer
2
function is: . (26)
s + 584.2s + 179, 200
201.1s + 171，400
4.2. System Simulation and Analysis Gr ( s) = 2
(2
s + 584.2s + 179，200
The input signal is set as a unit step signal, and the original system Gk (s) and the
system Gr (s) after adding self-tuning PID control are simulated, respectively. The system
4.2. System
response curves Simulation
are shown and Analysis
in Figures 6 and 7, respectively.
G ( s)
The system transient response indicators before and after optimization are extracted
The 7,
from Figures 6 and input signalin
as shown set as1.a unit step signal, and the original system k
is Table and t
G ( s)
system r after adding self-tuning PID control are simulated, respectively. T
Table 1. Transient response indexes of original system and optimized system.
system response curves are shown in Figure 6 and Figure 7, respectively.
Transient Response Performance Change
Original System Optimized System
Index (%)
Rise time/s 0.00135 0.00425 −214.81
Peak time/s 0.0033 0.0067 −103.03
Adjusting time/s 0.0187 0.0156 16.58
Maximum overshoot 0.3752 0.0604 83.90
 Processes2021,
Processes 2021,9,9,1538
1538 9ofof15
10 of 14
Processes 2021, 9, 1538 10 15

Figure6.
Figure
Figure Stepresponse
6.6.Step
Step responsecurve
response curveof
curve ofoforiginal
originalsystem.
original system.
system.

Figure7.
Figure
Figure Stepresponse
7.7.Step
Step responsecurve
response curveof
curve ofofsystem
systemafter
system afteradding
after addingself-tuning
adding self-tuningPID
self-tuning PIDcontrol.
PID control.
control.

According
The
The to the dataresponse
system transient
system transient analysisindicators
response in Table 1, before
indicators the riseand
before timeafter
and andoptimization
after peak time of the
optimization arecentrifugal
are extracted
extracted
pump flow
from Figures control
Figures 66 and and 7,system
7, as after
as shown
shown in adding
in Table the
Table 1.1. self-tuning PID controller are both large. When
from
parameter identification and PID control are added, the control process of the system
obviously
Table
Table 1. increases.
1. Transient
Transient The indexes
response
response control of
indexes time
of of thesystem
original
original system
system and
and also increases.
optimized
optimized Although the optimized
system.
system.
control time increases, the adjustment time to a steady state after system optimization is
Transient Response
Transient Response Index
Index Original System
System Optimized System
System Performance Change
Change (%)
(%)
shortened by 16.58%, whichOriginalimproves the rapidity Optimizedof the system Performance
transient response. The
Rise time/s
Rise time/s 0.00135
0.00135 0.00425
0.00425 −214.81
−214.81
optimized output
Peak time/s
time/s
response curve of
0.0033
the system is relatively
0.0067
gentle, the maximum
−103.03
overshoot
Peak 0.0033 0.0067 −103.03
is reduced by time/s
Adjusting
Adjusting
83.90%,
time/s and the stability
0.0187 of the system
0.0187
is greatly improved. 16.58
0.0156
0.0156 16.58
Maximum overshoot
Maximum overshoot 0.3752
0.3752 0.0604
0.0604 83.90
83.90
5. Experimental Analysis of Centrifugal Pump Flow Adaptive Control System
5.1. According
Test Platform
According toConstruction
to the data
the data analysis
analysis in
in Table
Table 1,
1, the
the rise
rise time
time and
and peak
peak time
time of
of the
the cen-
cen-
In pump
trifugal
trifugal order to
pump verify
flow
flow the accuracy
control
control system of theadding
system after
after self-tuning
adding PID controlPID
the self-tuning
the self-tuning algorithm,
PID controller
controllertheare
NGL002
are both
both
large. When parameter identification and PID control are added, the control processflow
centrifugal
large. When pump comprehensive
parameter test
identification device
and PID is selected
control to
are build
added, a centrifugal
the control pump
process of
of
the system obviously increases. The control time of the system also increases. Althoughin
adaptive
the system control system
obviously test
increases. platform.
The The
control test
time device
of the and
system test platform
also are
increases. shown
Although
Figures
the
the 8 and control
optimized
optimized 9, respectively.
control time increases,
time increases, the
the adjustment
adjustment time time toto aa steady
steady state
state after
after system
system
optimization is
optimization is shortened
shortened by by 16.58%,
16.58%, which
which improves
improves the the rapidity
rapidity ofof the
the system
system transient
transient
response. The
response. The optimized
optimized output
output response
response curve
curve of of the
the system
system is is relatively
relatively gentle,
gentle, the
the
maximum overshoot is reduced by 83.90%, and the stability of the
maximum overshoot is reduced by 83.90%, and the stability of the system is greatly im-system is greatly im-
proved.
proved.
 5. 5. Experimental
Experimental Analysis
Analysis ofof Centrifugal
Centrifugal Pump
Pump Flow
Flow Adaptive
Adaptive Control
Control System
System
5.1.
5.1. Test
Test Platform
Platform Construction
Construction
InIn order
order to to verify
verify thethe accuracy
accuracy of of
thethe self-tuning
self-tuning PIDPID control
control algorithm,
algorithm, thethe NGL002
NGL002
Processes 2021, 9, 1538 centrifugal pump comprehensive test device is selected to build a centrifugal
centrifugal pump comprehensive test device is selected to build a centrifugal pump 10 pump flow
of
flow 14
adaptive control system test platform. The test device and test platform are shown in in
adaptive control system test platform. The test device and test platform are shown
Figure
Figure 8 and
8 and Figure
Figure 9, 9, respectively.
respectively.

Figure
Figure
Figure 8. Centrifugal
Centrifugal
8.8.Centrifugal pump
pump comprehensive
comprehensive
pump test
test
comprehensive device.
device.
test device.

Figure
Figure 9.9. Test
Test platform
platformof centrifugal pump flow adaptive control system.
Figure 9. Test platform of of centrifugal
centrifugal pump
pump flow
flow adaptive
adaptive control
control system.
system.
It can be seen from Figure 9 that the centrifugal pump draws water from the first water
It It
tank, can
andcan
be be
that seen
seen
the from
from
water Figure
Figure
flows 9 that
9 that
through thethe
the centrifugal
centrifugal
flowmeter topump
pump draws
draws
be injected water
water
into from
from
the thethe
second first
first
water
water
water tank,
tank.tank,
The and and that the
that thefeeds
flowmeter water
water flows
flows
the through
flowthrough the flowmeter
the flowmeter
signal back to be injected
to be injected
to the controller. into the second
into the second
The controller controls
water
water
the motortank.
tank. The
speedThe flowmeter
flowmeter
through the feeds
feeds thethe
frequency flow
flow signal
signal
converter, back
back
and to to
then thethe controller.
controller.
controls The
the outputThe controller
controller
flow of the
centrifugal pump.

5.2. Data Acquisition and Analysis of Constant Frequency Flow

The rated flow rate of the centrifugal pump used in the test is Qe = 5 m3 /h, and the
electrical frequency is f = 50 Hz. The working frequency of the frequency converter is set
to f 1 = 45.8 Hz, and data are collected every 10 s for 10 times in total. The data acquisition
results are shown in Table 2.
 The rated flow rate of the centrifugal pump used in the test is Qe = 5 m 3 /h , and the
electrical frequency is f = 50 Hz . The working frequency of the frequency converter is
set to f1 = 45.8 Hz , and data are collected every 10 s for 10 times in total. The data ac-
Processes 2021, 9, 1538 quisition results are shown in Table 2. 11 of 14

Table 2. Test results collection.

Times/s
Table 2. Test results collection.
Theoretical Flow
A / (m3 /h ) Measured Flow B / (m3 /h ) Error Rate (%)
10
Times/s Theoretical 3.954
Flow A/(m3 /h) Measured Flow3.852
B/(m3 /h) 2.580
Error Rate (%)
20
10 3.951
3.954 3.8523.848 2.607
2.580
30
20 3.958
3.951 3.8483.853 2.653
2.607
30
40 3.958
3.948 3.8533.845 2.653
2.609
40
50 3.948
3.953 3.8453.847 2.609
2.682
50 3.953 3.847 2.682
60
60 3.956
3.956 3.8513.851 2.654
2.654
70
70 3.952
3.952 3.8493.849 2.606
2.606
80
80 3.957
3.957 3.8543.854 2.603
2.603
90
90 3.954
3.954 3.8523.852 2.580
2.580
100 3.949 3.844 2.659
100 3.949 3.844 2.659

Curve fitting
Curve fitting is
is performed according
according to
to the
thedata
datain
inTable
Table2 2[24].
[24].The
Thecurves
curvesofoftheoret-
theo-
ical flow and actual collected flow can be obtained
retical flow and actual collected flow can be obtained at aat a constant frequency, as shown
shown in in
Figure 10.
Figure 10.

Figure10.
Figure Constantfrequency
10.Constant frequencyflow
flowcurve
curvefitting.
fitting.

FromTable
From Table 22 and
and Figure
Figure 10,10, itit can
can be
beseen
seenthat
thatthe
thetheoretical
theoreticalflow
flow rates
rates of
ofthe
thepump
pump
are always greater than the actual flow rates at a constant frequency.
are always greater than the actual flow rates at a constant frequency. The main reason forThe main reason for
this is that the pump has a certain amount of leakage during operation.
this is that the pump has a certain amount of leakage during operation. With the increase With the increase
inacquisition
in acquisitiontime,
time,the
thetheoretical
theoreticalflow flowrate
rate of
of the
the pump
pump is is always
always maintained
maintained at at about
about
3.95 m33/h, and the variation range is less than 0.25%. The actual flow rate of the pump
3.95 m /h , and the variation range is less than 0.25%. The actual flow rate of the pump
is basically kept around 3.95, and the change range is less than 0.26%. Therefore, under
isthe
basically
conditionkept
of around
constant3.95, and thethe
frequency, change
adaptiverangePIDis regulator
less than can
0.26%. Therefore,
effectively under
adjust the
the condition
system of constant
parameters frequency,
and improve thethe adaptive
stability PIDpump.
of the regulatorThecan
erroreffectively
between adjust the
the actual
system
flow curveparameters
and theand improveflow
theoretical the stability
curve of ofthethe pump.
pump Thelarge,
is not error and
between the actual
the accuracy is
relatively high, which is basically controlled within 3%.

5.3. Data Acquisition and Analysis of Variable Frequency Flow

The theoretical flow of the centrifugal pump is Ql = Qe • f 1 / f , and the theoretical flow
of the pump is directly proportional to the operating frequency of the frequency converter.
The output frequency of the frequency converter and the actual flow rate of the pump
can be acquired through experiments. The upper limit value of the frequency converter’s
working frequency is set to 48 Hz, and the lower limit value is set to 30 Hz. The data
acquisition results are shown in Table 3.
 converter. The output frequency of the frequency converter and the actual flow rate of
the pump can be acquired through experiments. The upper limit value of the frequency
converter's working frequency is set to 48 Hz, and the lower limit value is set to 30 Hz.
The data acquisition results are shown in Table 3.
Processes 2021, 9, 1538 12 of 14
Table 3. Test results collection.

Frequency Theoretical Flow Measured Flow Error Rate

f1 Test
Table 3. / Hzresults collection. A / (m /h )
3 3
B / (m /h ) (%)
30 2.458 2.428 1.221
Frequency f1 /Hz Theoretical Flow A/(m3 /h) Measured Flow B/(m3 /h) Error Rate (%)
32 2.603 2.586 0.653
30 2.458 2.428 1.221
34 2.865 2.822 1.501
32 2.603 2.586 0.653
36
34 3.002
2.865 2.954
2.822 1.599
1.501
38
36 3.223
3.002 3.198
2.954 0.776
1.599
38
40 3.223
3.408 3.198
3.357 0.776
1.496
40 3.408 3.357 1.496
42 3.672 3.644 0.763
42 3.672 3.644 0.763
44
44 3.856
3.856 3.797
3.797 1.530
1.530
46
46 4.106
4.106 4.102
4.102 0.097
0.097
48
48 4.603
4.603 4.589
4.589 0.304
0.304

Curve
Curvefitting
fittingisisperformed
performedaccording
accordingtotothe
thedata
datain
inTable
Table3.3.The
Thecurves
curvesof
ofthe
thetheo-
theo-
retical
reticalflow
flowand
andactual
actualcollected
collectedflow
flowcan
canbe
beobtained
obtainedatataavariable
variablefrequency,
frequency,as
asshown
shownin in
Figure
Figure11.11.

Figure
Figure11.
11.Variable
Variablefrequency
frequencyflow
flow curve
curve fitting.
fitting.

Ascan
As can be
be seen
seen from Table
Table 33 and
and Figure
Figure11,11,the
theupper
upperand
andlower
lowerlimits
limitsofofthe working
the work-
frequency of the constant frequency are set to achieve accurate control of the
ing frequency of the constant frequency are set to achieve accurate control of the pump pump flow.
As the frequency increases, the theoretical flow and actual flow of the pump
flow. As the frequency increases, the theoretical flow and actual flow of the pump both both increase
in directinproportion.
increase This shows
direct proportion. This that under
shows thatthe condition
under of frequency
the condition conversion,
of frequency the
conver-
designed
sion, adaptive adaptive
the designed control method
controlcan effectively
method improve the
can effectively working
improve theefficiency
workingofeffi-
the
pump, and the control quality is good. With the change in the working frequency of the
frequency converter, the actual flow rates of the pump are close to the theoretical flow
rates. The actual flow is always kept near the preset flow, and the error is controlled within
2%. Therefore, the optimized centrifugal pump flow control system has better real-time
followability.

6. Conclusions
(1) By analyzing the components of a frequency converter, motor, and centrifugal pump,
the frequency converter and centrifugal pump were simplified as a proportional link.
The motor was simplified as an electrical link connected in a series with a mechanical
link, and the mathematical model of the centrifugal pump adaptive control system
was established.
(2) A self-tuning PID controller based on the adaptive control method was constructed, a
discrete-time stochastic linear model was used to describe the controlled object model,
and the forgetting factor recursive least squares method was used to estimate the pa-
Processes 2021, 9, 1538 13 of 14

rameters in real-time. The simulation results showed that the dynamic characteristics
of the optimized system were significantly improved, and the rapidity and stability
of the transient response of the system were both improved.
(3) The test platform was built based on the NGL002 centrifugal pump comprehensive
test device, and a good control effect was obtained. After optimization, the centrifugal
pump flow control system had strong robustness and had better real-time followability.
The accuracy of the self-tuning PID control algorithm was further verified.
(4) In order to improve the robustness and stability of the centrifugal pump, the math-
ematical model of the control system was obtained on the basis of deducing the
mathematical model of the frequency converter, motor and centrifugal pump. The
adaptive control algorithm was used to optimize the design of the system. The least
squares recursive algorithm was used to estimate the parameters, and the self-tuning
PID controller was added to adjust the parameters of the centrifugal pump control
system. The experimental results showed that the optimized centrifugal pump con-
trol system shortened the adjustment time, reduced the maximum overshoot, and
the system’s rapidity and stability were significantly improved. Based on the test
platform of the centrifugal pump flow adaptive control system, the flow data of a
fixed frequency and variable frequency were collected. The collected data showed
that the error between the actual flow curve and the theoretical flow curve of the
pump was small, which verified the accuracy of the self-tuning PID control algorithm.
It could be seen that based on the adaptive control algorithm, the performance of
each index of the centrifugal pump flow control system could be improved, and
the stability and robustness of the system could also be improved. This has certain
guiding significance for the ability of centrifugal pumps to effectively cope with
various complicated working conditions and to improve work efficiency during the
operation process.

Author Contributions: Conceptualization, Y.W.; methodology, Z.H.; software, H.Z.; validation,

Y.W., H.Z. and X.N.; formal analysis, Z.H.; investigation, X.N.; resources, Z.H.; data curation, X.N.;
writing—original draft preparation, Y.W.; writing—review and editing, Y.W.; visualization, Y.W.;
supervision, Y.W.; project administration, Y.W.; funding acquisition, Y.W. All authors have read and
agreed to the published version of the manuscript.
Funding: This research was funded by the Anhui Province University Discipline (Professional) Top
Talent Academic Funding Project (No. gxbjZD2021076). This project was funded by the National
College Student Innovation and Entrepreneurship Training Program Project (No. 202110380020).
This project was funded by the Key Project of Natural Science Research in Colleges and Universities
of Anhui Province (No. KJ2017A450). This project was funded by the Key Project of Natural Science
Foundation of Chaohu University (No. XLZ-201902).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The study did not report any data.
Conflicts of Interest: The authors declare that there is no conflict of interest regarding the publication
of this paper.

References
1. Wang, X.; Kuang, K.; Wu, Z.; Yang, J. Numerical simulation of axial vortex in a centrifugal pump as turbine with S-blade impeller.
Processes 2020, 8, 1192. [CrossRef]
2. Gülich, J.F. Centrifugal Pumps; Springer: Heidelberg, Germany, 2010.
3. Lobanoff, V.S.; Ross, R.R. Centrifugal Pumps: Design and Application; Gulf Publishing Company: Houston, TX, USA, 1992.
4. Gevorkov, L.; Rassõlkin, A.; Kallaste, A.; Vaimann, T. Simulink based model for flow control of a centrifugal pumping system. In
Proceedings of the International Workshop on Electric Drives: Optimization in Control of Electric Drives, Moscow, Russia, 31
January–2 February 2018; pp. 1–4.
5. Suh, S.H.; Kim, K.W.; Kim, H.H.; Yoon, I.S.; Cho, M.T. A Study on energy saving rate for variable speed condition of multistage
centrifugal pump. J. Therm. Sci. 2015, 24, 566–573. [CrossRef]
Processes 2021, 9, 1538 14 of 14

6. Mingxin, C.; Lingfeng, T.; Jian, H. Research on flow control system of pump based on fuzzy PID. J. Xinxiang Univ. 2014, 31, 36–40.
(in Chinese).
7. Lu, J.; Chen, M. Reach on the H∞ control system of the centrifugal pumps. J. Tongren Univ. 2018, 20, 68–75. (in Chinese).
8. Xie, S.; Chai, T. A kind of fuzzy PID controller with parameter adaption. Inf. Control. 1998, 27, 255–259. (in Chinese).
9. Perissinotto, R.M.; Verde, W.M.; Perles, C.E.; Biazussi, J.L.; de Castro, M.S.; Bannwart, A.C. Experimental analysis on the behavior
of water drops dispersed in oil within a centrifugal pump impeller. Exp. Therm. Fluid Sci. 2019, 112, 1–18. [CrossRef]
10. Hu, J. Research of Wet Desulfurization Dust Control System. Master’s Thesis, Anhui Polytechnic University, Wuhu, China, 2014.
(in Chinese).
11. Gang, C. The Research of The Method on The Frequency Conversion Energy-Conservation of the Hydraulic Apparatus Controls.
Master’s Thesis, Guizhou University, Guiyang, China, 2010. (in Chinese).
12. Zeng, G.; Qian, M. Frequency conversion driving technology of AC asynchronous motor. Electr. Abstr. 2014, 5, 13–17. (in Chinese).
13. Jun, W. Research on low frequency torque insufficiency of frequency converter. China Sci. Technol. Inf. 2015, 2, 183–184.
(in Chinese).
14. Zhao, X.S.; Hu, J.; Shi, P.C. Hydromechanics research of pump flow control system based on BP neural network PID. Appl. Mech.
Mater. 2013, 327, 222–226. [CrossRef]
15. Wen, M. Analysis of flow adjustment method of centrifugal pump. China Pet. Chem. Stand. Qual. 2012, z2, 117. (in Chinese).
16. Jiang, J.; Zhang, Y. A revisit to block and recursive least squares for parameter estimation. Comput. Electr. Eng. 2004, 30, 403–416.
[CrossRef]
17. Ding, F.; Wang, X.; Chen, Q.; Xiao, Y. Recursive least squares parameter estimation for a class of output nonlinear systems based
on the model decomposition. Circuits Syst. Signal Process. 2016, 35, 3323–3338. [CrossRef]
18. Alexandrov, A.; Palenov, M. Self-tuning PID-I controller. Autom. Remote Control 2011, 44, 3635–3640. [CrossRef]
19. Makarem, S.; Delibas, B.; Koc, B. Data-driven tuning of PID controlled piezoelectric ultrasonic motor. Actuators 2021, 10, 148.
[CrossRef]
20. Xie, X.; Ding, F. Adaptive Control System; Tsinghua University Press: Beijing, China, 2002. (in Chinese)
21. Paleologu, C.; Benesty, J.; Ciochina, S. A robust variable forgetting factor recursive least-squares algorithm for system identification.
IEEE Signal Process. Lett. 2008, 15, 597–600. [CrossRef]
22. Pang, Z.; Cui, H. System Identification and Adaptive Control Matlab Simulation; Beihang University Press: Beijing, China, 2013.
(in Chinese)
23. Luo, H.; Mao, Y.; Zhang, J. Design and simulation of a parameter adaptive fuzzy PID controller. Autom. Instrum. 2001, 3, 10–12.
(in Chinese).
24. Hu, M.; Feng, J.; Zheng, J. An additional branch free algebraic B-spline curve fitting method. Vis. Comput. 2010, 26, 801–811.
[CrossRef]