applied

sciences
Article
CFD-Rotordynamics Sequential Coupling Simulation
Approach for the Flow-Induced Vibration of Rotor
System in Centrifugal Pump
Hehui Zhang 1, *, Haolin You 1 , Haishan Lu 1 , Kun Li 1 , Zhiyong Zhang 2 and Liangxing Jiang 3, *
1 School of Mechanical Engineering, Xiangtan University, Xiangtan 411105, China;
201921001928@smail.xtu.edu.cn (H.Y.); luhs@xtu.edu.cn (H.L.); likun981120@163.com (K.L.)
2 Hunan M&W Energy Saving Co., Ltd., Xiangtan 411100, China; pump118@163.com
3 School of Metallurgy and Environment, Central South University, Changsha 410083, China
* Correspondence: hhzhang@xtu.edu.cn (H.Z.); lxjiang@csu.edu.cn (L.J.); Tel.: +86-731-5829-2209 (H.Z.)




Received: 30 December 2019; Accepted: 6 February 2020; Published: 10 February 2020


Featured Application: The numerical approach in this article can provide abundant information
about the flow induced vibration of the pump rotor and may offer theoretical guidance to the
design of reducing vibration and safety monitoring of centrifugal pumps.

Abstract: Vibration of the rotor system is closely related with the operation stability of centrifugal
pump, and it is inevitably induced by the unsteady inner flow. An unsteady computational fluid
dynamics model coupling with a rotordynamics model was presented, and the corresponding
numerical calculation program, including a self-designed rotordynamics code, was developed on the
commercial package ANSYS Workbench. The validity of the numerical calculation model was verified
by a hydraulic performance and vibration test based on an industrial centrifugal pump. The hydraulic
radial forces on impeller, pressure pulsation, deformation, and vibration of the main shaft under
nine different flow rates were systematically investigated and explained preliminarily from the view
of inner unstable flow. Results show that the blade passing frequency is the dominant frequency
in the fluctuation of the above dynamical behaviors, which is closely related to the rotor–stator
interaction between the impeller and volute casing. This study built the connection between internal
fluid flow in the centrifugal pump and the vibration of its external rotor structure, and may provide
theoretical references for the design of vibration-reducing and safety monitoring strategies design of
centrifugal pumps.

Keywords: centrifugal pump; unsteady flow; rotor–stator interaction; rotordynamics; vibration

1. Introduction
As an indispensable important energy conversion device, centrifugal pump is extensively applied
in various agricultural and industrial fields. For the reduction of equipment investment and energy
consumption, modern centrifugal pumps have developed towards the direction of higher flow capacity,
rotation speed, head, and efficiency [1–5]. This means that the vibration problem of centrifugal pump
tends to become more prominent due to the increased energy density, and it will inevitably bring about
unfavorable effects on the operational safety, stability, and life. The hazard of the vibration problem is
specially reflected in the rotor assembly of pump, and it may lead to the serious damage of high-speed
running parts such as the shaft, impeller, and bearings [6].
There are mainly two kinds of factors influencing the vibration of centrifugal pump rotor
assembly—the mechanical factors and the hydraulic factors. The mechanical factors generally relate
to the unbalance mass of the rotating parts, misalignment of shaft system and radial rub between

Appl. Sci. 2020, 10, 1186; doi:10.3390/app10031186 www.mdpi.com/journal/applsci

Appl. Sci. 2020, 10, 1186 2 of 33

different parts, so those factors can be effectively weakened by improving the machining accuracy
and assembly quality [7,8]. While for the hydraulic factors, they are predominated by the complicated
fluid–structure interaction (FSI) due to inner wall pressure fluctuation caused by the unsteady flow
in centrifugal pump. As a result, it is very necessary to obtain the induction mechanism of the rotor
assembly vibration from the view of internal unsteady flow, so the vibration level can be restrained at
the source via inner flow field optimization.
The unsteady flow behaviors in centrifugal pump has been systematically investigated in the way
of numerical calculation and experiment test. The intense rotor–stator interaction due to the relative
movement between impeller and volute forms the primary character of internal unsteady flow, so the
dominant frequencies of unsteady pressure fluctuation are usually the blade passing frequency (f BPF )
and multiple shaft frequency (f n ) [9]. Similar pulsation tendency is also fund in the hydraulic radial
force on the impeller, because the hydraulic force mainly depends on the integration of pressure on the
impeller surfaces. In addition, there are also other obvious frequency components rather than f BPF or
multiple f n , which are stimulated by a serious complex unsteady flow behaviors, such as the jet-wake
pattern along the blade extension direction, the flow impact on the tongue, and the flow separation on
the blade suction surface [10–12]. The above behaviors usually become more evident under off-design
conditions and would take much greater risks for the operation of centrifugal pump [13,14].
Based on the research of internal unsteady flow in centrifugal pump, the structural dynamic
response analysis has been introduced to perform the FSI study. Because of the complex structure
and twisted shape of pump, currently the typical application for the FSI problem of pump is the
loosely coupled (explicit) simulation between computational fluid dynamics (CFD) and computational
structural dynamics (CSD) [15]. Considering the fluid-induced vibration of the pump brings very
limited displacement of the fluid–solid interface, and their interaction acts as a sustained effect in
the whole operation process, the FSI problem can be solved with the partitioned method, in which
the equations governing the internal flow and structural deformation can be solved separately and
expediently with independent field solvers. The principal advantage of the partitioned method is that
the whole modeling and solving procedure for the FSI problem can be built on existing and mature
CFD and CSD codes with the widely used arbitrary Eulerian–Lagrangian description for the materials,
where the Eulerian (spatial) method is used to descript the fluid domain and the Largangian method
for the solid component. In this way, not only the required computational resources for both CFD
and CSD solvers can be largely saved, but the information exchange (interface mesh deformation)
and loads transfer (pressure from fluid domain and displacement on the solid surfaces) representing
the FSI effect can be applied in an efficiency and stable way. In recent years, coupling simulation
based on the partitioned method has been widely applied in the FSI problems of pump and other
turbomachineries, either with the strong two-way coupling strategy or the one-way sequential coupling
strategy, and most of the application cases achieved satisfactory prediction results in the deformation,
oscillation, and orbit curves of the impeller [16–20].
However, present reported numerical methods are not able to meet the requirements for the
prediction of rotor vibration in centrifugal pump. The main reason is that the solving of CSD in current
work is built on the general dynamic equation, thus neither the rotary inertia nor gyroscopic effects
can be considered, and the stiffness and damping characteristic of bearing has been ignored. In such
a case, the dynamics strain and stress distribution of the impeller can be obtained for use of strength
checking and safety evaluation, but the motion features about the shaft system cannot be acquired
with acceptable accuracy. In fact, the flow-induced vibration behaviors of the shaft system contain
a lot of useful information, and they can be easily measured in a cost-effective manner [21]. Therefore,
the monitoring and analyzing of shaft vibration data has become the mainstream approach for the
safety assessment and fault diagnosis of fluid machinery [22,23].
Based on the previous work by Zhang et al. [1,24,25], this research presents a new one-way coupling
simulation approach for the flow-induced vibration of rotor assembly in centrifugal pump, with special
attention on the association features of unsteady internal flow and shaft vibration. Compared with
Appl. Sci. 2020, 10, 1186 3 of 33

former works, the first and foremost improvement of this work is the application of rotordynamics in
the CSD, where the bearing damping matrix and gyroscopic matrix have been added to the general
dynamic equation. With this advantage, the above new approach is able to provide a wealth shaft
oscillation information including vibration velocity and shaft centerline orbit, with both the unsteady
hydraulic force on the impeller and the rotor dynamic properties in consideration.
In this article, an industrial split-case double suction centrifugal pump was chosen for study,
and the unsteady flow as well as rotor assembly vibration behaviors under different operation
conditions were investigated. The remaining of this article was organized as follows. In Section 2,
the mathematical model and its overall frame are presented. Section 3 describes the computation
details and experiment setup for a research case. Sections 4 and 5 are the analysis and discussion for
the numerical and experimental results, followed by the conclusions in Section 6.

2. Sequential Coupling Model

2.1. Hypothesis and Simplification

This study constructs the CFD-rotordynamics sequential coupling numerical model of centrifugal
pump mainly based on following hypotheses and simplifications:
(1) The coupling between flow field and rotordynamics belongs to one-way coupling. It means
that only influences of flow induced exciting force on the vibration behavior of the rotor system
were considered, but interferences of the deformation and whirl orbits of impeller to the flow field
are ignored. According to the calculation and experimental results of a centrifugal pump made
by Pei et al. [14,20], compared with the two-way coupling method with consideration of impeller
deformation, the calculated radial force based on one-way coupling shows little differences in numerical
values. In addition, the vibration displacements of impeller predicted by both one-way coupling and
two-way coupling show very small errors with experimental results. However, compared with the
two-way coupling method, the one-way coupling way can save as much as 75% computing time and
achieves stronger convergence ability during numerical solving.
(2) In flow field analysis, internal leakage of fluid media through sealing rings was neglected.
(3) In rotordynamics analysis, mass distribution was hypothesized uniform on the impeller,
namely the unbalanced mass force of the rotating impeller was treated to be zero. In addition, hole and
slot structures of the main shaft are ignored. Therefore, the rotor system is considered as a rotationally
symmetric structure composed of some mass points of impellers, a main shaft, and its bearings.

2.2. Overall Model Framework

The overall framework of the unsteady flow field-rotordynamics coupling model of centrifugal
pump is shown in Figure 1. The coupling model is mainly composed of two parts. One is the CFD
model based on finite volume method and the other is the rotordynamics model based on finite element
method. The CFD model implements the steady-state CFD solving by using a multiple reference frame
(RMF) model in the rotating impeller region, and its computation results are used as initial conditions
for the following unsteady CFD solving, where the time varying hydraulic excitation forces acting
on the impeller can be obtained. Then the exciting forces on three orthogonal spatial directions are
transported to the rotordynamics model though load time series; thus, the sequential coupling process
between those two models can be completed.
Appl. Sci. 2020, 10, 1186 4 of 33
Appl. Sci. 2020, 10, x FOR PEER REVIEW 4 of 35

CFD model Rotordynamics model

Rotordynamics
Momentum equation Continuity equation equation

Turbulence model
Mesh of solid zones
Mesh of fluid zones

MRF model

Steady CFD model APDL calculation

code

Model solution

Sliding meshes Unsteady solution

Unsteady CFD model Load transfer

of F(t, x, y )

Sequential coupling
Vibration and
Model solution deformation of rotor

Time-variable hydraulic
excitation force

Figure 1.
Figure Overall framework
1. Overall framework of
of the
the coupling
coupling model.
model.

2.3. CFD Model

2.3. CFD Model
2.3.1. Governing Equations
2.3.1. Governing Equations
A governing equation group of fluid mechanics in a centrifugal pump is set up according to
A governing
Navier–Stokes equation
equation andgroup of fluid
continuity mechanics in a centrifugal pump is set up according to
equation
Navier–Stokes equation and continuity equation
∂ρ∂ρ →
∇·(ρ u⃑ )=0
++∇∙(ρu )=0 (1)
(1)
∂t ∂t
∂ρ → ∂ρ (ρu ⃑ )+∇∙(ρu → ⃑ )T )]+ρg
→ T⃑ +f →
(ρ u u )⃑ =
⃑ )=-∇P+∇∙[μ(∇u ⃑ +(∇u
→→ u → (2)
(ρ u )∂t+ ∇· −∇P + ∇·[µ(∇ u + (∇ u ) )] + ρ g + f (2)
∂t
where t represents time (s), → 𝑢 ⃑ is the velocity vector−1(m·s−1), p is static pressure (Pa), ρ is density
where−3t represents time (s), u is the velocity vector (m·s ), p is static pressure (Pa), ρ is−1density (kg·m−3 ),
→ m ), 𝑓 is external body force term (N·m ), µ is the effective viscosity (kg·m ·s ) and g is the
(kg· −3 −1

s−2). term (N·m ), µ is the effective viscosity (kg·m ·s ) and g is the gravity
f is external body force −3 −1 −1
gravity constant (m·
constant (m·s−2 ).
2.3.2. Turbulence Model
2.3.2. Turbulence Model
The governing equations are closed by introducing the SST k-ω turbulence model [26]. The SST
The governing
k-ω model integratesequations
advantagesare closed
of bothbyk-ε introducing
turbulencethe SST k-ω
model andturbulence modelmodel,
k-ω turbulence [26]. The SST
and it
k-ω model integrates advantages of both k-ε turbulence model and k-ω turbulence
controls value of dissipation source term in a specific dissipation equation with the blending function model, and it
controls
F value of dissipation source term in a specific dissipation equation with the blending function
1. As shown in Equation (3), coefficient φ can be changed with the variation of F1 function
F1 . As shown in Equation (3), coefficient ϕ can be changed with the variation of F1 function
φ=F1 φ1 +(1-F1 )φ2 (3)
ϕ = F1 ϕ1 + (1 − F1 )ϕ2 (3)
where φ, φ1 , and φ2 are coefficients of the SST k-ω turbulence model, k-ω turbulence model and k-
ε turbulence
where model,
ϕ, ϕ1 , and respectively.
ϕ2 are coefficientsWhen
of theFSST
1 = 0, turbulence in the region far away from the wall is
k-ω turbulence model, k-ω turbulence model and k-ε
solved by the k-ε turbulence model. On contrary,
turbulence model, respectively. When F1 = 0, turbulence the in
k-ωthe
turbulence
region far model under
away from thelow
wallReynolds
is solved
number is applied in the region close to the wall when F1 = 1. As a result, the SST k-ω turbulence
by the k-ε turbulence model. On contrary, the k-ω turbulence model under low Reynolds number is
model are believed to be able to provide a relatively high prediction accuracy of the performance
curves of pump with acceptable computational cost [27,28].
Appl. Sci. 2020, 10, 1186 5 of 33

applied in the region close to the wall when F1 = 1. As a result, the SST k-ω turbulence model are
believed to be able to provide a relatively high prediction accuracy of the performance curves of pump
with acceptable computational cost [27,28].

2.3.3. Treatment of the Gird of Impeller Zone

Appl. Sci. 2020, 10, x FOR PEER REVIEW 5 of 35
A grid technique is needed to treat the transition and connection issues between the rotating
2.3.3. Treatment
impeller and other ofnon-rotating
the Gird of Impeller
flow zones Zone
of centrifugal pump. There are two different grid techniques
in this work. In the steady CFD model,
A grid technique is needed to treat the the MRF model The
transition andMRF model was
connection issuesintroduced
between inthethe steady
rotating
CFD model,
impeller and which applies a stationary
other non-rotating flow zones reference frame in pump.
of centrifugal the suction
Thereand aredischarge chambers,
two different grid
and a rotating reference frame for the impeller zone. In this way, the governing
techniques in this work. In the steady CFD model, the MRF model The MRF model was introduced equations for fluid zones
with
in thedifferent
steady CFDreference
model,fames are applies
which solved independently, and local
a stationary reference velocity
frame andsuction
in the pressure information
and discharge
on their interfaces
chambers, is exchanged
and a rotating by coordinate
reference frame for the transformation
impeller zone. processing.
In this way,While in the unsteady
the governing CFD
equations
model, the sliding meshes technique is used in the rotating impeller zone,
for fluid zones with different reference fames are solved independently, and local velocity and and a uniform stationary
referenceinformation
pressure frame is setonfor theinterfaces
their whole fluid domain, by
is exchanged which can actually
coordinate reflect more
transformation detailedWhile
processing. flow
behaviors
in caused
the unsteady by the
CFD rotor–stator
model, the sliding interaction (RSI) in pump.
meshes technique is used in the rotating impeller zone, and
a uniform stationary reference frame is set for the wholethe
The sliding meshes technique is realized by modifying velocity
fluid termwhich
domain, of the convection
can actuallydiffusion
reflect
reaction
more (CDR)flow
detailed equation for any
behaviors fluidbycontrol
caused volume within
the rotor–stator the computation
interaction (RSI) in pump. domain based on its
movement featuremeshes
The sliding relativetechnique
to the stationary
is realizedreference frame
by modifying the velocity term of the convection
diffusion reaction (CDR) Z
equation forZany fluid control →
volume
Z within
→
the
Z computation domain based
d → →
ρφdVto
on its movement feature relative + the ρφ ( u − u g )reference
stationary ·dA = ψ∇φ·d
frame A + Sφ dV (4)
dt
d
∫ ρϕdV + ∫ ρϕ(u ⃑ -u ⃑⃑ = ∫ ψ∇ϕ∙dA
⃑⃑⃑g ) ∙dA ⃑⃑ + ∫ Sϕ dV → (4)
where Φ represents a general flow scalar for an arbitrary control volume V (m3 ), A is the area vector of
dt
→
2 ), u is the mesh velocity vector of the moving mesh (m·s −1 ), ψ and S are the
the control
where volume (m
Φ represents a general g flow scalar for an arbitrary control volume V (m3), A ⃑⃑ is the area
ψ vector
diffusion
of coefficient
the control volume and(msource
2), ⃑⃑⃑⃑
𝑢𝑔 isterm
theof scalar
mesh Φ respectively.
velocity vector of the moving mesh (m·s-1), ψ and Sψ are
For a centrifugal
the diffusion coefficientpump, no volume
and source termoverlapping
of scalar Φ occurs between dynamic and stationary meshing
respectively.
regions
Forduring the rotating
a centrifugal of theno
pump, impeller
volumezone, which is attributed
overlapping to the cylinder
occurs between interface
dynamic and of dynamic
stationary
and stationary meshing regions. Therefore, volume of the control volume in Equation
meshing regions during the rotating of the impeller zone, which is attributed to the cylinder interface (4) keeps
constant. Hence, it can be simplified as
of dynamic and stationary meshing regions. Therefore, volume of the control volume in Equation (4)
keeps constant. Hence, it can be simplified as
(ρφV )i+1 − (ρφV )i
Z
d
ρφdV = (ρϕV) -(ρϕV) (5)
dt d i+1∆t i (5)
∫ ρϕdV =
dt Δt
where i represents the respective value at the ith time level, and i+1 denotes its next one.
where i represents the respective value at the ith time level, and i+1 denotes its next one.
2.4. Rotordynamics Model
2.4. Rotordynamics Model
Based on the Jeffcott rotor model [29], a rotordynamics model of centrifugal pump can be
Based on
constructed. the be
It can Jeffcott rotorFigure
seen from model [29],a typical
2 that a rotordynamics model
rotordynamics of centrifugal
model is composedpump can be
of a principal
constructed. It can be seen from
axis, an impeller and two bearings. Figure 2 that a typical rotordynamics model is composed of a
principal axis, an impeller and two bearings.
Main shaft Impeller

Bearing
Fixed Constraint

Figure 2.
Figure Schematic diagram
2. Schematic diagram of
of the
the rotordynamics
rotordynamics model.
model.

The rotordynamics model is built based on the rotordynamics equation, which is transformed
from the universal dynamics equation by adding the gyroscopic matrix and rotating damping matrix.
It is expressed as
[M]{𝑈̈}+([C]+[G]){𝑈̇}+([K]+[B]){U}={F} (6)
where is {U} is the displacement matrix; [M], [C] and [K] are the mass matrix, damping matrix and
Appl. Sci. 2020, 10, 1186 6 of 33

The rotordynamics model is built based on the rotordynamics equation, which is transformed
from the universal dynamics equation by adding the gyroscopic matrix and rotating damping matrix.
It is expressed as
.. .
[M]{U} + ([C] + [G]){U} + ([K] + [B]){U} = {F} (6)

where is {U} is the displacement matrix; [M], [C] and [K] are the mass matrix, damping matrix and
stiffness matrix, respectively; [G] is the gyroscopic matrix for describing the rotating inertia of the
rotor system, and its value is determined by rotating speed; [B] is the rotating damping matrix closely
connected to the rotor stability, and its value is also related with rotating speed; {F} is the external force
vector, which mainly includes the hydraulic force caused by the inner unsteady flow and the gravity of
the impeller itself in this study. The Equation (6) is applicable to calculate rotor movement in the same
stationary reference coordinate system.

3. Numerical Calculation Application and Experimental Setup

3.1. Case Examined

In this study, commercial computational codes Fluent and ANSYS Mechanical were applied as
the CFD and rotordynamics analysis tools, respectively. The whole process including pre-processing,
post-processing, physical field coupling, and numerical solving were implemented on the ANSYS
Workbench co-simulation platform [30].
A horizontal split centrifugal pump with a six-blade double suction impeller was chosen as the
research object. The pump was manufactured by Hunan M&W pump Co., Ltd., and it was designed
for the clean water transportation. Its whole major structure is completely symmetric, so the axial
hydraulic force on the impeller is fully balanced in theory due to the double suction structures of the
suction chamber and impeller. The main structural and performance parameters of the centrifugal
pump are listed in Table 1. The three-dimensional structure is shown in Figure 3 and the middle
section of hydraulic model is represented in Figure 4, where the suction chamber, the impeller with six
blade passages and the discharge chamber were exhibited with different colors. In addition, an X-Y
orthogonal coordinate system with its origin located at the impeller center was established, and the
rotating of impeller was set as counterclockwise.
Figure 5 shows the three-dimensional structure and parameters of the main shaft, where the
installation positions of bearing were marked with upward arrows. The main shaft is made of structural
steel and can be divided into five sections with different diameters. Specifically, the middle section with
a diameter of 0.049 m and a length of 0.664 m is designed for carrying the impeller and its accessories.
Two sections next to the impeller (diameter = 0.045 m and length = 0.194 m) are designed for the
mounting of ball bearings and the end sections (diameter = 0.042 m and length = 0.135 m) are used for
the connecting of coupler.

Table 1. Basic parameters of the research object

Parameter Symbol Value

Rated flow Qdes 620 m3 /h
Rated head Hdes 96 m
Rated speed ndes 1480 rev min−1
Specific speed ns 52
NPSH required NPSHr 3.6 m
Impeller inlet diameter D1 0.18 m
Impeller outlet diameter D2 0.56 m
Number of blades Z 6
Impeller outlet width b2 0.029 m
 Rated speed ndes 1480 rev min−1
Specific speed ns 52
NPSH required NPSHr 3.6 m
Impeller inlet diameter D1 0.18 m
Impeller outlet diameter D2 0.56 m
Appl. Sci. 2020, 10, 1186 Number of blades Z 6 7 of 33
Impeller outlet width b2 0.029 m

Appl. Sci. 2020, 10, x FOR PEER REVIEW 7 of 35

Figure 3. Three-dimensional
Three-dimensionalstructure
structureofofthe research
the researchobject: 1—Pump
object: 1—Pump body; 2—Outlet;
body; 3—Base;
2—Outlet; 4—
3—Base;
Coupler; 5—Rolling
4—Coupler; bearing;
5—Rolling 6—Mechanical
bearing; 6—Mechanical seal;seal;
7—Main
7—Mainshaft;shaft;
8—Pump cover;cover;
8—Pump 9—Sealing ring;
9—Sealing
Appl. Sci. 2020, 10, x FOR PEER REVIEW 7 of 35
10—Impeller.
ring; 10—Impeller.

Figure 4. Middle section plane view of the hydraulic model.

Figure 5 shows the three-dimensional structure and parameters of the main shaft, where the
installation positions of bearing were marked with upward arrows. The main shaft is made of
structural steel and can be divided into five sections with different diameters. Specifically, the middle
section with a diameter of 0.049 m and a length of 0.664 m is designed for carrying the impeller and
its accessories. Two sections next to the impeller (diameter = 0.045 m and length = 0.194 m) are
designed for the mounting of ball bearings and the end sections (diameter = 0.042 m and length =
0.135 m) are used for the connecting
Figure of section
4. Middle coupler.plane view of the hydraulic model.
Figure 4. Middle section plane view of the hydraulic model.

Figure 5 shows the three-dimensional structure and parameters of the main shaft, where the
installation positions of bearing were marked with upward arrows. The main shaft is made of
structural steel and can be divided into five sections with different diameters. Specifically, the middle
section with a diameter of 0.049 m and a length of 0.664 mparameters
is designed for carrying the impeller and
of main shaft.
Figure 5. Three-dimensional structure and
its accessories. Two sections next to the impeller (diameter = 0.045 m and length = 0.194 m) are
3.2. Steadyfor
designed State
theCFD Model Setup
mounting of ball bearings and the end sections (diameter = 0.042 m and length =
0.135 m) are used for the connecting of coupler.
3.2.1. Mesh Generation and Turbulence Model
Hexahedral dominated grids for CFD analysis were generated by the meshing module of ANSYS
Workbench. Meanwhile, transition regions like casing tongue were processed by mesh refinement.
For near-wall grids, the allowable maximum y+ was set as 50 by with the meshing adaptive tool of
Fluent after the first CFD analysis (trial computation) [30]. Therefore, the mesh refinement function
Figure
was initiated and grids in 5.
theThree-dimensional
near-wall regionstructure and parameters
were re-meshed basedofonmain
the shaft.
flow distribution results.
In this study, the SST k-ω turbulence model was chosen, whose parameters were listed in Table 2.
Appl. Sci. 2020, 10, 1186 8 of 33

Table 2. Parameter settings of the SST k-ω turbulence model.

Parameter α1 β β0 σ k1 σ ω1 α2 β2 σ k2 σ ω2
Value 0.556 0.075 0.09 0.85 0.5 0.44 0.828 1 0.856

3.2.2. MRF Model Setup and Boundary Condition

The whole computational domain was divided into three sub-domains, namely, the suction
chamber, the impeller and the discharge chamber. These sub-domains exchange the solution information
through their interfaces. The MRF model was applied in the rotating impeller region, in which the
reference frame was set to be clockwise rotating with a speed of 1480 rev min−1 , which equals to the
rated rotating speed of the centrifugal pump.
Boundary conditions were set as follows: the entrance boundary was set as a velocity inlet
perpendicular to the entrance plane, where the turbulent intensity and viscosity ratio were 5% and
10, respectively. An outflow condition with a 100% backflow ratio was adopted for the outlet to
reflect the fully developed flow pattern. The wall condition was used for other surfaces. Specifically,
the roughness was set as 2 × 10−5 m for walls of the impeller and 5 × 10−5 m for walls in the suction
and discharge chambers, which match the actual processing parameters of the pump manufacturer.

3.2.3. Solution Setup

Nine flow rates from Q0.6 to Q1.4 at an interval of 0.1Qdes were investigated in this work. The fluid
media in pump was set as 20 ◦ C clean water, and its density and viscosity are 998.2 kg m−3 and
1.003 × 10−3 Pa s, respectively. Second order upwind scheme was applied for discretion of the
momentum equation and SIMPLE algorithm was used for the iterative computation. The convergence
criterion was less than 10−4 for the maximum residual values for the mass and momentums in
three directions.
The case with Qdes flow rate was chosen to run the grid independence test. Calculated results
including head and efficiency under six different mesh densities were presented in Figure 6, whose
calculation formulas were expressed as Equations (7) and (8).

P0 − Pi
H= +∆ (7)
ρg

where H is pump head, Pi and Po are the total pressure of the inlet and outlet of the pump respectively,
and ∆ represents the height difference between the center points of inlet and outlet faces.

955ρgQH
η= (8)
nT

where Q is volume flow (m3 /h), n is the shaft’s rotational speed (rev·min−1 ) and T denotes the torque
on the main shaft (N m).
Figure 6 shows that both head and efficiency of the pump began to keep stable when the number
of girds exceeds 3.70 × 106 . And the fluctuation ranges of head and efficiency can be controlled within
0.7% and 0.5%, respectively. Therefore, the grid with number of 3.70 × 106 was chosen for further
computation. Numbers of element and node of each sub-domain were listed in Table 3, and the element
quality, skewness factor, and orthogonal quality of the entire gird were 0.79, 0.35, and 0.86, respectively.
Appl. Sci. 2020, 10, 1186 9 of 33
Appl. Sci. 2020, 10, x FOR PEER REVIEW 9 of 35

120 100
Head 95
110 Efficiency
90

100 85

80

Efficiency/ %
Head/ m 90
75
80
70

70 65

60
60
55

50 50
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
6
Grid number/10

Figure 6. Predicted
Predicted head
head and
and efficiency
efficiency values with different grid numbers.

Figure 6 shows that both headTable 3. Mesh

and information
efficiency for calculation.
of the pump began to keep stable when the number
of girds exceeds 3.70 × 10 . Fluid
6 And the
Zonefluctuation ranges
Element of head and
Number efficiency
Node Number can be controlled within
0.7% and 0.5%, respectively. Therefore, the grid with number of 3.70 × 106 was chosen for further
Suction chamber 974,659 801,071
computation. NumbersDischarge
of element and node of
chamber each sub-domain1,261,920
1,261,920 were listed in Table 3, and the
element quality, skewness factor,zone
Impeller and orthogonal quality of the entire
1,463,735 gird were 0.79, 0.35, and 0.86,
1,463,738
respectively. Total 3,700,314 3,526,729

Table 3. Mesh information for calculation

3.3. Unsteady State CFD Model Setup
The unsteady CFD model Fluid
wasZone Element
set basically Number
consistent with Node Number CFD model except for
the steady-state
solving settings and the processing Suction chamber
of the impeller 974,659
rotating regions.801,071
Discharge chamber 1,261,920 1,261,920
In the unsteady CFD model, the frame of the impeller region was changed into a stationary
Impeller zone 1,463,735 1,463,738
reference one, and the sliding meshes were applied to grids in the impeller region. In other words,
Total 3,700,314 3,526,729
all grids in the impeller region make counterclockwise rotation around the impeller center at the rate
of −1
3.3.1480 rev min
Unsteady State. CFD Model Setup
With respect to solving, the calculated results of steady-state CFD model were used as the initial
The unsteady
conditions for unsteadyCFD model
CFD modelwas set basically
under consistent
the same withtothe
flow rate steady-state
assure CFD model
convergence except
performance.
forthis
In solving settings
study, the time and thefor
step processing of the
solution was impeller
1.1261 × 10 rotating
−4 s, which regions.
equals to the time for the impeller to
rotate by 1 . And the whole solving time is 0.2432 s, corresponding towas
In the◦ unsteady CFD model, the frame of the impeller region changed into
six revolutions of athe
stationary
impeller.
reference one, and the sliding meshes were applied to grids in the impeller
The convergence standard adopted by each solving step was that the maximum residual values reached region. In other words,
all grids
10 −4 in the
of the impeller
initial valuesregion
of mass make
andcounterclockwise
momentum, but the rotation
maximumaround the impeller
iteration center
number for at thetime
each rate
of 1480 rev min −1.
step was controlled to be less than 20. Besides, various monitoring settings were added during the
WithCFD
unsteady respect to solving,
solving process,theaiming
calculated results
to get time of steady-state
series CFD model
data of calculated were usedsuch
parameters as the
as initial
mean
conditions for unsteady CFD model under the same flow rate to assure convergence
pressure and flow rate at entrance and exit of the pump, torque on impeller and hydraulic forces. The performance. In
this study, the time step for solution was 1.1261 × 10 −4 s, which equals to the time for the impeller to
computation was carried out on the HP workstation (CPU: E5-2695V2, Memory: 48GB), and it took
rotate 6bydays
about 1°. And the whole
to complete onesolving
case. time is 0.2432 s, corresponding to six revolutions of the impeller.
The convergence standard adopted by each solving step was that the maximum residual values
3.4. Rotordynamics
reached Model values
10−4 of the initial Setup of mass and momentum, but the maximum iteration number for each
time step was controlled to be less than 20. Besides, various monitoring settings were added during
3.4.1. Resonance
the unsteady CFD Check
solving process, aiming to get time series data of calculated parameters such as
meanBefore
pressure and
rotordynamicsflow rate at entrance
analysis, modal and exit ofon
analysis thethe
pump,
rotor torque
systemon of impeller andcentrifugal
the studied hydraulic
forces. was
pump The carried
computationout to was carried
eliminate out on the
possibility of HP workstation
structural (CPU:
resonance E5-2695V2,
under Memory:
the rated rotating48GB),
speed.
and it took about 6 days to complete one case.
The modal analysis shows that the first six orders of inherent frequencies of the rotor system were
3.4. Rotordynamics Model Setup

3.4.1. Resonance Check

Before rotordynamics analysis, modal analysis on the rotor system of the studied centrifugal
Appl. Sci. 2020, 10, 1186 10 of 33
pump was carried out to eliminate possibility of structural resonance under the rated rotating speed.
The modal analysis shows that the first six orders of inherent frequencies of the rotor system were
67.04 Hz,
67.04 Hz, 118.61
118.61 Hz,Hz, 138.60
138.60 Hz,
Hz, 235.26
235.26 Hz,
Hz, 253.01
253.01 Hz,
Hz, and
and 661.44
661.44 Hz,
Hz, respectively.
respectively. Since
Since none
none of
of these
these
were close
were close to
to the
the rotating
rotating frequency
frequencyofofmain
mainshaft
shaftor
orits
itsharmonics,
harmonics,there
thereisisno
norisk
riskof
ofresonance.
resonance.

3.4.2.
3.4.2. Element Type
Type Selection
Selection
Finite
Finite element
element modeling
modeling and calculation
calculation of rotordynamics
rotordynamics model were completed
completed by ANSYS
ANSYS
Parametric
Parametric Design
Design Language
Language (APDL)
(APDL) language
language programming [30]. [30]. The codes of the whole program
were
were provided
provided inin the
the Appendix
Appendix A A of
of this
this article.
article. During
During the programming, different types of finite
elements
elements were
were applied
applied forfor different
differentstructural
structuralparts.
parts.
Beam
Beam 188188 element
element with
with a three-dimensional
three-dimensional linearlinear beam
beam and
and two
two nodes
nodes was
was chosen
chosen for
for the main
shaft, which is suitable for the analysis the problem with
shaft, with large
large rotation
rotation angle
angle and
and large
large structure
structure
deformation. Mass 21 element was assigned to the impeller,
deformation. information of
impeller, and information of the mass and rotating
inertia of
inertia of the
the mass
mass point
point can
can be
be added
added through
through real
real constants.
constants. COMBI 214 element was used for the
bearings, which
bearings, which isis aa kind
kind of
of spring-damper
spring-damper element
element with
with two
two nodes.
nodes.

3.4.3.
3.4.3. Mesh
Mesh and
and Material
Material Attribution
Attribution
As shown in
As shown inFigure
Figure7,7,the
themain
main shaft
shaft meshed
meshed intointo high-quality
high-quality structural
structural hexahedral
hexahedral grids.grids.
The
The minimum grid size was set 0.002 m and each radial section was assured to have at least
minimum grid size was set 0.002 m and each radial section was assured to have at least 20 layers of 20 layers
of elements
elements in in order
order totoassure
assuresolving
solvingaccuracy.
accuracy.Local
Localgrids
gridsatatends
endsofof shaft
shaft were
were zoomed
zoomed up up and
and
positions
positions of
of the
the bearing
bearing were
were marked
marked with
with ‘K0’.
‘K0’.

Figure 7.
Figure 7. Finite element mesh of the rotordynamics
rotordynamics model.
model.

The
The main shaft was made of structural steel, whose physical
physical properties
properties were
were listed
listed in
in Table
Table 4.
According
According to the data
data provided
providedbybythe
themanufacturer,
manufacturer,the
therigidity
rigidity
of of
thethe rolling
rolling bearings
bearings waswas
set set
5×
9
510× N/m,
10 N/m,
9 and and the damping
the damping can can be ignored.
be ignored.

Table4.4.Material
Table Materialproperties
propertiesof
ofthe
themain
mainshaft.
shaft

Density
Density (kg(kg
m−3m
) −3) Elasticity
ElasticityModulus
Modulus(Pa)(Pa) Poisson’s
Poisson’sRatio
Ratio
7890 2.09 × 101111 0.269
7890 2.09 × 10 0.269

3.4.4. Constraint and Load Condition

3.4.4. Constraint and Load Condition
Nodes on the bearings which are far away from the main shaft (representing the outer rings of
Nodes on the bearings which are far away from the main shaft (representing the outer rings of the
the bearings) were fixed. Loading conditions of the rotordynamics model include two parts: (1) forces
bearings) were fixed. Loading conditions of the rotordynamics model include two parts: (1) forces
changing with time was applied on the mass point of impeller, including hydraulic radial forces on
changing with time was applied on the mass point of impeller, including hydraulic radial forces on the
the impeller along the X and Y directions (coming from the calculated results of unsteady CFD model)
impeller along the X and Y directions (coming from the calculated results of unsteady CFD model) and
and the gravity of impeller along the Y direction. (2) Rotational angular velocity was applied onto the
the gravity of impeller along the Y direction. (2) Rotational angular velocity was applied onto the main
main shaft and impeller to calculate the Coriolis force. The Coriolis force comes from the rotating
shaft and impeller to calculate the Coriolis force. The Coriolis force comes from the rotating inertia
inertia of the rotor system, and it is used to describe the migration effect of rectilinear motion mass
of the rotor system, and it is used to describe the migration effect of rectilinear motion mass in the
in the rotating system. The calculation formula of the Coriolis force is shown as
rotating system. The calculation formula of the Coriolis force is shown as

→ →
Fk = −2mωr × v0 (9)
Appl. Sci. 2020, 10, x FOR PEER REVIEW 11 of 35
Appl. Sci. 2020, 10, 1186 11 of 33

⃑⃑⃑′
⃑⃑⃑⃑𝑟 ×𝑣
Fk =-2m𝜔 (9)
→
where Fk isisthe
where thegeostrophic
geostrophicforce
force(N),
(N),mmdenotes
denotesthe mass
the mass(kg), ωr ⃑⃑⃑⃑
(kg), is𝑟the
𝜔 is angular velocity
the angular vector
velocity of rotor
vector of
→
rotor (rad/s),
(rad/s), and vand ⃑⃑⃑
0 is the ′
𝑣 velocity
is the velocity
vector vector relative
relative to the reference
to the reference rotating rotating frame (m/s).
frame (m/s).

3.4.5.Solution
3.4.5. SolutionSetup
Setupand
andDamping
DampingCoefficient
CoefficientSelection
Selection
Unsteadystate
Unsteady statesolving
solvingwas wasalsoalsoapplied
appliedininthe therotordynamics
rotordynamicsfinite finiteelement
elementanalysis
analysismodel,
model,
whosetime
whose timestep
stepand
andsolving
solvingtimetimekeep
keepconsistent
consistentwith
withthose
thoseininthe
theunsteady
unsteadyCFD CFDanalysis.
analysis.Transient
Transient
completemethod
complete methodwas wasactivated
activatedininthe thecomputation
computationprocess,
process,namely
namelythe theunsteady
unsteadyresponses
responsesof ofthe
the
structural to
structural to external loads
loads were
werecalculated
calculatedby bycomplete
completemass, mass, damping,
damping, and stiffness
and matrixes.
stiffness matrixes.In this
In
way,way,
this various nonlinear
various features
nonlinear such as
features elasticity,
such large strain,
as elasticity, large and large
strain, and deformation of main shaft
large deformation of main can
be obtained
shaft more accurately.
can be obtained more accurately.
ToToassure
assuresolving
solvingstability
stabilityduring
duringdynamics
dynamicsanalysis
analysisof ofstructures,
structures,aadamping
dampingmatrixmatrixisisusually
usually
needed to consider
needed considerinfluences
influences of damping
of dampingon structural movements
on structural in practical
movements engineering.
in practical According
engineering.
to the viscous
According to thedamping theory, the
viscous damping rotor the
theory, damping is proportional
rotor damping to the vibration
is proportional velocityvelocity
to the vibration and its
acting
and its direction is opposite
acting direction to the velocity.
is opposite Therefore,
to the velocity. a proportional
Therefore, damping
a proportional was applied
damping onto the
was applied
entire
onto thefinite element
entire analysis analysis
finite element model. This
model.implies
This that
impliesthe system
that thedamping matrix [C]
system damping in Equation
matrix [C] in
(6) is approximately
Equation treated astreated
(6) is approximately the linear combination
as the of mass matrix
linear combination of mass[M]matrix
and the[M] stiffness
and thematrix [K],
stiffness
as expressed
matrix [K], as in the Equation
expressed in the(10)
Equation (10)
[C] = α·[M] + β·[K] (10)
[C]=α∙[M]+β∙[K] (10)
where α and β denote the mass and stiffness damping coefficients respectively.
where α and β denote the mass and stiffness damping coefficients respectively.
With references to research results of Pei et al. [14,20], a stiffness damping coefficient was added
With references to research results of Pei et al. [14,20], a stiffness damping coefficient was added
into the rotordynamics model and its value was determined by trial-and-error method and engineering
into the rotordynamics model and its value was determined by trial-and-error method and
experiences. A trial operation was implemented under the flow rate of Q , in which a relatively small
engineering experiences. A trial operation was implemented under the des flow rate of Qdes, in which a
stiffness damping coefficient (β = 0.001) was set firstly. Figure 8 shows that the velocity curves of mass
relatively small stiffness damping coefficient (β = 0.001) was set firstly. Figure 8 shows that the
point of impeller was very unstable with β = 0.001, and no valid information can be obtained from
velocity curves of mass point of impeller was very unstable with β = 0.001, and no valid information
them. Then, when the value of stiffness damping coefficient was increased to β = 0.002, the velocity
can be obtained from them. Then, when the value of stiffness damping coefficient was increased to β
curves began to be stable and appear periodic fluctuation feature after oscillation in the first revolution
= 0.002, the velocity curves began to be stable and appear periodic fluctuation feature after oscillation
of impellers (corresponding to 0.041 s). Thirdly, when the stiffness damping coefficient β reached
in the first revolution of impellers (corresponding to 0.041 s). Thirdly, when the stiffness damping
0.003, it takes a shorter time for regulation compared with that under β = 0.002. However, it takes
coefficient β reached 0.003, it takes a shorter time for regulation compared with that under β = 0.002.
more system input disturbance under β = 0.003. Therefore, the stiffness damping coefficient in the
However, it takes more system input disturbance under β = 0.003. Therefore, the stiffness damping
computation was finally determined as β = 0.002.
coefficient in the computation was finally determined as β = 0.002.

0.03  0.03


 
0.02
 
0.02
0.01

0.01 0.00
-1

-1
Velocity/ms

Velocity/ms

-0.01
0.00
-0.02

-0.01 -0.03

-0.04
-0.02
-0.05

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
Time/s Time/s

(a) (b)
Figure
Figure8.8.Impeller
Impellermass
masspoint
pointvelocity
velocityversus
versustime
timewith
withdifferent
differentstiffness
stiffnessdamping
dampingcoefficient:
coefficient:(a)
(a)XX
direction; (b) Y direction.
direction; (b) Y direction.
Appl. Sci.
Appl. Sci. 2020,
2020, 10,
10, 1186
x FOR PEER REVIEW 12 of
12 of 33
35

3.5. Experiment Setup

3.5. Experiment Setup
3.5.1. Hydraulic Performance Test
3.5.1. Hydraulic Performance Test
To verify the numerical results, a hydraulic performance test of the industrial centrifugal pump
To verifyout
was carried theinnumerical results,
an industrial testarig,
hydraulic
duringperformance test ofsignals
which vibration the industrial centrifugal
were collected pump
at the was
bearing
carried out in an industrial test rig, during which vibration signals were collected
end cover. The test system was shown in Figure 9, which agree with the level-1 standards in ISO at the bearing end cover.
The test system
9906-2012 [31]. was
The shown
test riginmainly
Figureincludes
9, which two
agreevalves,
with the
two level-1 standards
connected water in tanks,
ISO 9906-2012
a pump,[31].an
The test rig mainly
electromotor and someincludes two valves,
measuring two connected
instruments. Duringwater
test,tanks, a pump,
the water an electromotor
temperature and some
was controlled
measuring instruments. During test, the waterintemperature was controlled within ◦ C ± 0.3 ◦ C.
within 19.8°C ± 0.3 °C. The random uncertainty the process of data acquisition was 19.8
within ± 0.15%.
The random
flow rateuncertainty
was measured in thebyprocess of data acquisition
electromagnetic flowmeter was withwithin ±0.15%.
a measuring Theofflow
error rate Two
± 0.2%. was
measured by electromagnetic
pressure gauges were utilizedflowmeter
to obtain with a measuring
the pump error
head, and ±0.2%. Two
theofmeasuring pressure
error gauges
of each one waswere±
utilized
0.1%. A to obtaintransducer
torque the pump head,
and a and
laserthe measuringwere
tachometer errorapplied
of each one was ±0.1%.
to measure A torque
the torque transducer
and rotation
and
speed a of
laser
thetachometer were appliedand
main shaft respectively, to their
measure the torqueerrors
measurement and rotation
were bothspeed of the
± 0.1%. Thus,main shaft
based on
respectively, and theirthe
uncertainty analysis, measurement
total measurementerrors were both ±0.1%.
uncertainties for the Thus,
headbased on uncertainty
and efficiency analysis,
are 0.21% and
the
0.38%total measurement uncertainties for the head and efficiency are 0.21% and 0.38% respectively.
respectively.

9. Hydraulic
Figure 9. Hydraulic performance
performance test
test rig: 1—Inlet
1—Inlet valve;
valve; 2—Inlet gauge; 3—Laser
2—Inlet pressure gauge; 3—Laser
tachometer; 4—Electromotor;
4—Electromotor;5—Torque
5—Torque transducer; 6—Centrifugal pump; 7—Outlet
transducer; 6—Centrifugal pump; 7—Outlet pressure pressure
gauge;
gauge; 8—Electromagnetic
8—Electromagnetic flowmeter;
flowmeter; 9—Outlet
9—Outlet valve; 10—Water
valve; 10—Water drainage
drainage tank; 11—Base
tank; 11—Base line; 12—
line; 12—Water
Water supply
supply tank. tank.

3.5.2. Vibration Measurement

3.5.2. Vibration Measurement
Vibration
Vibration behaviors
behaviors ofof the
the bearing
bearing end
end cover
cover of
of the
the centrifugal
centrifugal pump
pump waswas collected
collected during
during the
the
test, which can be used to approximately represent the vibration level of the rotor system
test, which can be used to approximately represent the vibration level of the rotor system according according to
Chinese
to Chinese standards
standardsof GB/T
of GB/T29531-2013
29531-2013[32][32]
andand
GB/T 6075.1-2012
GB/T [33].
6075.1-2012 [33].
It
It can be seen from Figure 10 three sensors with a common base were
can be seen from Figure 10 three sensors with a common base were mounted
mounted atat the
the center
center of
of
the bearing end cove. With that test instrument, vibration accelerations on three orthogonal
the bearing end cove. With that test instrument, vibration accelerations on three orthogonal spatial spatial
directions
directions(two(tworadial
radialdirections
directionsandand
oneoneaxialaxial
direction) were measured
direction) with PTZ
were measured piezoelectric
with ceramic
PTZ piezoelectric
sensors, and the sample frequency was set as 10.24 kHz. The frequency response
ceramic sensors, and the sample frequency was set as 10.24 kHz. The frequency response and and measurement
ranges of the sensors
measurement ranges ofare
the0.5 Hz–5 are
sensors kHz0.5 and 0–50
Hz–5 kHzg, and
respectively. The INV3018A
0–50 g, respectively. data collection
The INV3018A data
instrument with 24-bit resolution and 4 simultaneous-capture analog input channels
collection instrument with 24-bit resolution and 4 simultaneous-capture analog input channels was utilizedwas
for
analog acquisition, digital conversion, signal filtering, and data storage, and the
utilized for analog acquisition, digital conversion, signal filtering, and data storage, and theacquisition system
accuracy
acquisition was 0.02%.accuracy was 0.02%.
system
Appl. Sci. 2020, 10, 1186 13 of 33
Appl. Sci. 2020, 10, x FOR PEER REVIEW 13 of 35

Figure 10. Picture of the mounted vibration acceleration sensor.

4. Results Analysis

4.1. Numerical Model Validation

4.1.1. Comparison of Performance Curves

The tested and calculated performance curves of the centrifugal pump were shown in Figure 11,
Figure 10. Picture of the mounted vibration acceleration sensor.
where the unsteady calculation results were the mean values gathered from the unsteady CFD
4. Results Analysis
calculation
4. Results Analysis
for the last impeller revolution.
In general, the predicted performance curves agree well with test results, especially in the nearby
4.1.
rangeNumerical Model
of Qdes. For Validation
the Validation
flow of Qdes, the maximum errors between calculated head coefficients Ch and its
4.1. Numerical Model
test value are less than 0.036, with a relative deviation of about 3.3%, and the absolute difference
4.1.1. Comparison of Performance Curves
between
4.1.1. calculatedofefficiency
Comparison Performance η andCurves
its test value is controlled within 0.5%. When the flow moves far
away from Qdes, the error between calculatedcurves
The tested and calculated performance value of thetest
and centrifugal pump were
value increases shown
with the in Figure
increase 11,
of flow
where The tested
the and calculated
unsteady calculation performance
results curves
were the of the centrifugal
mean values pump were
gathered from shown
the in Figure
unsteady CFD11,
offset, especially in the part-load conditions. However, the maximum relative errors of Ch and η are
where the for
calculation unsteady
the calculation
last3.9%
impeller results were the mean values gathered from the unsteady CFD
revolution.
only about 4.6% and (around Q0.6), respectively.
calculation for the last impeller revolution.
In general, the predicted
1.20 performance curves agree well with test results, 85 especially in the nearby
range of Qdes. For the flow of Qdes, the maximum errors between calculated head 84
coefficients Ch and its
1.15
test value are less than 0.036, with a relative deviation of about 3.3%, and the absolute difference
83
1.10
between calculated efficiency η and its test value is controlled within 0.5%. When the flow moves far
Qdes 82
away from Qdes, the error 1.05 between calculated value and test value increases with the increase of flow
Head coefficient, Ch

Efficiency, /%

81
offset, especially in the1.00 part-load conditions. However, the maximum relative errors of Ch and η are
80
only about 4.6% and 3.9% 0.95
(around Q0.6), respectively.
79
0.90
1.20 Steady numerical head coefficient 85
Unsteady numerical head coefficient 78

0.85 Experimental head coefficient 84

1.15 77
Steady numerical efficiency
Unsteady numerical efficiency 83
0.80
1.10 76
Experimental efficiency
Qdes 82
0.75
1.05 75
0.04 0.05 0.06 0.07 0.08 0.09 0.10 0.11
Head coefficient, Ch

Efficiency, /%

81
1.00 Flow coefficient, Cq
80
0.95
Figure 11. Comparisons of the performance curves.
Figure 11. 79
0.90 Steady numerical head coefficient
In
Thegeneral, the predicted
computational performance curves agree well with test results,78especially in the nearby
Unsteady numerical head coefficient
0.85 errors are mainly attributed to the following two aspects. Firstly, both
Experimental head coefficient
77
range
leakage Qdes . mechanical
of and For the flowloss of Qwere
des , the maximum
ignored in theerrors
CFDbetween
model. calculated
Steady numerical efficiency
Then, errorshead coefficients
may come fromCh and
the
Unsteady numerical efficiency
0.80
its test value
discretion and aresolving
less thanprocesses
0.036, with duea relative
to the deviation of about
flow 3.3%,
Experimental efficiency
complicated and the
behaviors
76
in absolute difference
centrifugal pump.
between calculated
Specifically, efficiency
the solving
0.75 η andofits
accuracy thetest value
k-ω is controlledmodel withindeclines
0.5%. When
75 the flowconditions,
moves far
0.04 0.05 0.06 SST0.07 turbulence
0.08 0.09 0.10 0.11 in part-load
away from Qdes ,inthe
thus resulting error
the between
higher solvingcalculated value
error. Flow and Ctest itvalue
Additionally,
coefficient, can increases
be seen fromwith the increase
Figure of flow
11 that the
q
offset, especially in the part-load conditions. However, the maximum relative errors of Ch and η are
only about 4.6% and 3.9% (around Figure 11. Comparisons
Q0.6 of the performance curves.
), respectively.
The computational errors are mainly attributed to the following two aspects. Firstly, both leakage
The computational
and mechanical loss wereerrors
ignored arein mainly
the CFDattributed
model. Then, to the following
errors may come twofrom
aspects. Firstly, both
the discretion and
leakage processes
solving and mechanical
due to the loss were ignored
complicated flowin the CFDinmodel.
behaviors Then,
centrifugal errors
pump. may comethe
Specifically, from the
solving
discretion and solving processes due to the complicated flow behaviors
accuracy of the SST k-ω turbulence model declines in part-load conditions, thus resulting in the higher in centrifugal pump.
Specifically, the solving accuracy of the SST k-ω turbulence model declines in part-load conditions,
thus resulting in the higher solving error. Additionally, it can be seen from Figure 11 that the
4.1.2. Comparison of Vibration Characteristics of Main Shaft
The variation curves of vibration accelerations along X and Y directions versus time were gained
from numerical calculation and vibration measurement. In the numerical model, a node of the main
shaft which
Appl. Sci. connects
2020, 10, 1186 with the bearing was chosen for study and comparison. It was found14that of 33
spectral distribution characters of vibration acceleration under different flow rates are similar in some
aspects, so the Qdes condition was chosen for the comparisons of the spectral distribution of the
solving error.
vibration Additionally,
accelerations, it caninbeFigure
as shown seen from
12. Figure 11 that the calculation curves of steady-state
CFDIn analysis
Figure 12, distribution laws can be seentobetween
and unsteady CFD analysis differ some extent, and their and
computation differences are much
measurement more
results.
significant under
Specifically, part-load acceleration
the vibration conditions. Generally,
magnitude thereaches
unsteady
theCFD
peakresults
at fBPFprovide higher
(148 Hz), prediction
which is the
accuracy. A possible explanation is that the inner flow is more unstable under part-load
dominant frequency. Meanwhile, the magnitudes at fn (24.7 Hz), fBPF and their harmonics. Moreover, condition,
andacceleration
the the unsteady CFD model
magnitude withXsliding
along meshes
direction technique
is generally can reflect
higher the along
than that RSI better than
the Y the steady
direction. To
CFDup,
sum model with MRF setting
the numerical for the
calculation impeller grids.
of vibration information of the rotor system is reliable.
However, the calculated results are different from measurement results to some extent with
4.1.2. Comparison of Vibration Characteristics of Main Shaft
respect to specific vibration acceleration magnitude and distribution features. On one hand, the
calculated acceleration
The variation curvesmagnitudes
of vibrationunder dominant
accelerations frequency
along is higher versus
X and Y directions than thetimemeasurement
were gained
results. For example,
from numerical the calculated
calculation vibration
and vibration acceleration magnitudes
measurement. alongmodel,
In the numerical X and Y directions
a node of theunder
main
the
shaftfBPF are connects
which 0.88 m with
s−2 and 0.26 m was
the bearing s−2 respectively,
chosen for studywhich are higher than
and comparison. It wasthe
foundcorresponding
that spectral
measurement results (0.46
distribution characters m s−2 and
of vibration 0.19 m under
acceleration s−2). On the other
different hand,are
flow rates the measured
similar in some vibration
aspects,
acceleration magnitudes
so the Qdes condition wasat chosen
some other multiple
for the shaft frequencies
comparisons (e.g.,
of the spectral 98.7 Hz and
distribution 296vibration
of the Hz) are
significantly
accelerations,higher thanin
as shown the calculated
Figure 12. results.

0.9 (a) 148Hz

0.6

0.3 98.7Hz
296Hz

0.0
0.3 148Hz
(b)
0.2
24.7Hz
0.1
-2
Magnitude/ms

0.0
0.45 (c) 148Hz

0.30

296Hz
0.15 98.7Hz

0.00
0.20 (d)
148Hz
0.15

0.10
98.7Hz 296Hz
0.05

0.00
0 100 200 300 400 500 600 700 800 900 1000
Frequency/Hz

Figure
Figure 12. Comparisonsof of
12. Comparisons thethe frequency
frequency domains
domains of of
thethe vibration
vibration acceleration
acceleration under QdesQ
under des :
: (a)
(a) Calculated
Calculated values
values in Xindirection;
X direction;
(b)(b) Calculated
Calculated values
values in in
YY direction;
direction; (c)(c)Test
Testvalues
valuesininXXdirection;
direction;
(d) Test
(d) Test values in Y direction.

In Figure 12, distribution laws can be seen between computation and measurement results.
Specifically, the vibration acceleration magnitude reaches the peak at f BPF (148 Hz), which is the
dominant frequency. Meanwhile, the magnitudes at f n (24.7 Hz), f BPF and their harmonics. Moreover,
the acceleration magnitude along X direction is generally higher than that along the Y direction. To sum
up, the numerical calculation of vibration information of the rotor system is reliable.
However, the calculated results are different from measurement results to some extent with
respect to specific vibration acceleration magnitude and distribution features. On one hand,
the calculated acceleration magnitudes under dominant frequency is higher than the measurement
results. For example, the calculated vibration acceleration magnitudes along X and Y directions
under the f BPF are 0.88 m s−2 and 0.26 m s−2 respectively, which are higher than the corresponding
Appl. Sci. 2020, 10, 1186 15 of 33

measurement −2 −2 ).
Appl. Sci. 2020, 10,results (0.46REVIEW
x FOR PEER m s and 0.19 m s On the other hand, the measured vibration acceleration 15 of 35
magnitudes at some other multiple shaft frequencies (e.g., 98.7 Hz and 296 Hz) are significantly higher
thanDifferences
the calculated between
results.calculated results and measurements are mainly caused by the following
threeDifferences
aspects. Firstly,
betweenthe calculated
unsteady CFD model
results applies Reynolds-average
and measurements are mainly caused methodbytothe average
followingthe
instantaneous
three aspects. values
Firstly,ofthe
turbulence
unsteady phenomenon,
CFD model thus ignoring
applies the effects of high-frequency
Reynolds-average method to average random the
disturbance.
instantaneousSecondly,
values of vibration
turbulenceofphenomenon,
the rotor system thus in the numerical
ignoring the effects model is only attributed
of high-frequency random to
hydraulic
disturbance. exciting. However,
Secondly, due of
vibration to the
mechanical
rotor systemprocessing and assembly,
in the numerical model it isis inevitable to have
only attributed to
some otherexciting.
hydraulic vibrationHowever,
sources in due
practical operation,processing
to mechanical such as external forces caused
and assembly, it isby mass imbalance
inevitable to have
of
someimpeller, rotor asymmetry,
other vibration sources inand electrical
practical motor. such
operation, Thirdly, the computation
as external forces caused model cannot
by mass provide
imbalance
vibration
of impeller,information of bearing
rotor asymmetry, anddirectly duemotor.
electrical to the limitation
Thirdly, the of computation
element typesmodel in the cannot
rotordynamics
provide
finite element
vibration model. of
information Instead,
bearing bearing
directlyvibration
due to the is limitation
approximately reflected
of element types byininformation
the rotordynamicson the
shaft
finite node connected
element model. with the bearing.
Instead, bearing Considering the high stiffness
vibration is approximately of the by
reflected bearing, the vibration
information on the
acceleration measurements
shaft node connected are bearing.
with the smaller than the calculated
Considering values,
the high and there
stiffness of theare multiple
bearing, the harmonic
vibration
components
acceleration in measurementare
measurements results.
smaller than the calculated values, and there are multiple harmonic
In this study,
components a quantitative
in measurement evaluation on vibration energy of a centrifugal pump was carried
results.
out by In using the root-mean-square
this study, (RMS) on
a quantitative evaluation value of vibration
vibration energy ofvelocity. At thepump
a centrifugal first step, curvesout
was carried of
vibration
by using the velocity versus time (RMS)
root-mean-square were gained
value ofthrough
vibration time integral
velocity. At ofthevibration
first step,acceleration series.
curves of vibration
Then,
velocityRMS of vibration
versus time were velocities under different
gained through flow rates
time integral were calculated
of vibration acceleration andseries.
shownThen, in Figure
RMS13. of
It is easy to
vibration see thatunder
velocities the RMS velocity
different flow distribution
rates were feature
calculatedof computation
and shown inresults Figureis13. highly similar
It is easy to
to see
measurements. Generally
that the RMS velocity speaking,feature
distribution vibration velocity level results
of computation is relatively low around
is highly similar to Qdesmeasurements.
, and the RMS
ranges
Generallyare speaking,
between 0.4 m s−1–1.0
vibration m s−1. And
velocity level RMS velocitylow
is relatively along the XQdirection
around des , and the is generally
RMS ranges higher
are
than
betweenthat 0.4
along −1
m sthe–1.0 −1
Y direction.
m s . And Nevertheless,
RMS velocity RMS velocity
along the X increases
direction significantly
is generally higher under thanpar-load
that
and
alonglarge
the Yflow conditions,
direction. and it increases
Nevertheless, RMS velocityto the peak under
increases Q0.6. Moreover,
significantly under par-load the calculated
and largeRMS flow
velocity
conditions,along
andtwo directions
it increases arepeak
to the slightly
under higher than measurements,
Q0.6 . Moreover, the calculated which RMSisvelocity
similar along
with two
the
patterns
directions inare
Figure 12. higher than measurements, which is similar with the patterns in Figure 12.
slightly

2.5

Calculated values in X direction

2.0 Calculated values in Y direction
Test values in X direction
-1
Velocity RMS value/mms

Test values in Y direction

1.5

1.0

0.5

0.0
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
Q/Qdes

Figure 13.
Figure Comparisonsof
13. Comparisons ofthe
thevibration
vibrationvelocity
velocityRMS
RMSvalues
valuesunder
under different
different flow
flow conditions.
conditions.

4.2. Fluctuation of the Hydraulic Radial Force on the Impeller

4.2. Fluctuation of the Hydraulic Radial Force on the Impeller
In unsteady CFD analysis, variations of hydraulic radial forces along the X and Y directions of
In unsteady CFD analysis, variations of hydraulic radial forces along the X and Y directions of
impeller of the centrifugal pump with time under different flow rates in the last revolution of impeller
impeller of the centrifugal pump with time under different flow rates in the last revolution of impeller
were shown in Figure 14a,b, respectively. Radial forces along X and Y directions present sine fluctuating
were shown in Figure 14a,b, respectively. Radial forces along X and Y directions present sine
characteristics with f BPF as the dominant frequency. Under Q0.6 , the hydraulic radial force along the X
fluctuating characteristics with fBPF as the dominant frequency. Under Q0.6, the hydraulic radial force
direction has a high value and it points to the negative direction of the X-axis. When the flow rate
along the X direction has a high value and it points to the negative direction of the X-axis. When the
increases to Q0.7 , the radial force along the X direction decreases to some extent, but it still points to the
flow rate increases to Q0.7, the radial force along the X direction decreases to some extent, but it still
negative direction of X-axis. While under Q0.8 , the absolute value of radial force along the X direction
points to the negative direction of X-axis. While under Q0.8, the absolute value of radial force along
the X direction further declines and its positive and negative direction alter with the rotation of
impeller. Then under large flow conditions, the hydraulic radial force along the X direction increases
gradually with the continuous increase of flow rate and it keeps positive in X-axis. To sum up, the
Appl. Sci. 2020, 10, 1186 16 of 33

further declines and its positive and negative direction alter with the rotation of impeller. Then under
Appl. Sci.
large 2020, 10, x FOR PEER
flow REVIEW 16 of 35
Appl. Sci. 2020,conditions, theREVIEW
10, x FOR PEER hydraulic radial force along the X direction increases gradually with 16 ofthe
35
continuous increase of flow rate and it keeps positive in X-axis. To sum up, the hydraulic radial force
hydraulic
along
hydraulic radial force
Y direction
radial force along
points YYnegative
to the
along directiondirection
direction points to
points to the negative
under
the negative direction
differentdirection under
flow ratesunder different
and its flow
value isflow
different rates
positively
rates
and its
its value
correlated
and value
withis positively
is positively correlated with
flow rate. correlated with flow
flow rate.
rate.

1600 0
1600 0
1400 Q 0.6Q
1400 Q 0.6 0.6Q des
0.6 des
1200 Q -200 0.7Q
1200 Q 0.7 -200 0.7Q des
0.7 des
1000 Q
0.8Q
0.8Q des
1000 Q 0.8 -400 des
800 0.8 -400 0.9Q
800
Q 0.9Q des
Q 0.9 des
600 0.9 -600 1.0Q
600 Qdes 1.0Q des
Qdes -600 des
400 Q1.1
1.1Q
1.1Q des
400 des
Force/N

-800

Force/N
200 Q1.1 1.2Q
Force/N

-800

Force/N
200 Q1.2 1.2Q des
des
0 Q1.2 1.3Q
0 Q -1000 1.3Q des
des
-200 Q 1.3 -1000 1.4Q
-200 1.3 1.4Q des
Q des
-400 Q 1.4 -1200
-400 1.4 -1200
-600
-600 -1400
-800 -1400
-800
-1000 -1600
-1000 -1600
-1200
-1200 -1800
0 0.25 0.5 0.75 1 -1800 0 0.25 0.5 0.75 1
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
Period of the impeller Period of the impeller
Period of the impeller Period of the impeller
(a)
(a) (b)
(b)
Figure 14. Hydraulic
Figure Hydraulic radial
radial force
force versus
versus time
time under
under different
different flow
flow conditions:
conditions: (a)
(a) XX direction;
direction; (b)
(b) YY
Figure 14.
14. Hydraulic radial force versus time under different flow conditions: (a) X direction;
direction.
direction.
(b) Y direction.

4.3. Pressure
4.3. Pressure Pulsation
Pulsation Characteristics
Characteristics
Fluctuation of
Fluctuation of hydraulic
hydraulic radial
radial force
force on
on the
the impeller
impeller isis closely
closely related
related with
with pressure
pressure fluctuation
fluctuation
of the
of the internal
internal unsteady
unsteady flowflow field. On the
field. On the mid-section
mid-section of of discharge
discharge chamber,
chamber, fivefive monitoring
monitoring nodes
nodes
(P1–P5)
(P1–P5) close
close to the volute wall were chosen to investigate the pressure fluctuation
fluctuation
(P1–P5) close to the volute wall were chosen to investigate the pressure fluctuation laws. The distance laws.
laws. The distance
between each
between each node
node andand thethe volute
volute wall
wall is
is 0.02
0.02 m.m. As
m. As seen
seen inin Figure
Figure 15,
15, P1
P1 locates
locates near
near the
the volute
volute
tongue and
tongue and the
the rest
rest four
four points
points are
are on
on the X-axis and
the X-axis
X-axis and Y-axis,
and Y-axis, respectively.
Y-axis, respectively.Flow
respectively. Flowrate
Flow rateQ
rate Qdes
Q des was taken as
des was taken as
an example,
example, under
under which
which the the pressure
pressure fluctuations
fluctuations at five at five
monitoring
an example, under which the pressure fluctuations at five monitoring points in one monitoring
points inpoints
one in
impeller onerevolution
impeller
impeller
revolution
were
revolution were
shownwere shown
in Figure
shown inFigure
16.in Figure
Figure 16.
1616. Figure
shows
Figure 16
the16 shows the
dominate
shows the dominate
pressure
dominate pressure
fluctuation
pressure fluctuation
period period
is 1/6 ofperiod
fluctuation is 1/6
1/6
the rotating
is
of the
period rotating
of the period
impeller, of
whichthe impeller,
agrees which
with the agrees with
experimental the experimental
regularity obtained
of the rotating period of the impeller, which agrees with the experimental regularity obtained by regularity
by Wang obtained
et al. by
[12].
Wang
However,
Wang et al. [12].
et al.pressure However,
values and
[12]. However, pressure values
their amplitude
pressure and their
values andvariations amplitude variations
are different
their amplitude for each
variations are different
monitoring
are different for each
for point.
each
monitoring
Similar point.
pressure Similar
fluctuation pressure
features fluctuation
were features
observed were
under observed
other flow under
rates.
monitoring point. Similar pressure fluctuation features were observed under other flow rates. other flow rates.

Figure 15.
Figure 15. Position setting
15. Position
Position setting of
of the
the pressure
pressure monitoring nodes.
monitoring nodes.
nodes.
Figure setting of the pressure monitoring
Appl. Sci. 2020, 10, 1186 17 of 33
Appl. Sci. 2020, 10, x FOR PEER REVIEW 17 of 35

8.5

8.0

Pressure/10 Pa
7.5
5

7.0
P1
P2
P3
6.5 P4
P5

6.0
0 0.25 0.5 0.75 1
Period of the impeller

Figure 16.
Figure Pressure fluctuation
16. Pressure fluctuation of
of different
different monitoring
monitoring points under Q
points under des..
Qdes

To further
To furthercompare
comparepressure
pressurefluctuation characteristics
fluctuation under
characteristics different
under flow rates,
different flow the time-varying
rates, the time-
curves P ( node, t ) of pressure at different monitoring nodes are decomposed into
varying curves 𝑃(node,t) of pressure at different monitoring nodes are decomposed into the fixed component
the fixed
P(node) and the fluctuating component P e(node, t), as shown in Equation (11).
̃
component P(node) and the fluctuating component P(node,t), as shown in Equation (11).
P (𝑃(node,t) (node) +̃P
node, t) ==PP(node)+P (node, t)
e(node,t) (11)
(11)

where thenumerical
where the numericalvalue
value of(node
of P P(node) is defined
) is defined the arithmetic
the arithmetic meanofvalue
mean value of in
pressure pressure in one
one revolution
revolution
N−1
N-1
11 X
P (node ) = P(node, t0+i∆t)
P(node)=N ∑ P(node,t + i∆t) (12)
(12)
0
N i=0
i=0

where N is the number of time steps, t00 is is the

the initial
initial moment and ∆t
moment and ∆𝑡 isisthe
thetime
timestep
steplength.
length.
ThusThus the value
the value ̃of
of P e(node,ist)the
P
(node,t) is the difference
difference between
between instantaneous
instantaneous pressurevalue
pressure P(node,t
value𝑃( node, )t) and
and
P ( node
P(node) )
e(node, t) = P(node, t) − P(node)
P (13)
P̃(node,t)=P(node,t)-P(node) (13)
e(node, t) is worth in-depth study as it indicates the instability of the inner flow of centrifugal
P
𝑃̃ (𝑛𝑜𝑑𝑒, 𝑡) is worth in-depth study as it indicates the instability of the inner flow of centrifugal
pump. Hence, the dimensionless quantity namely the pressure fluctuation intensity coefficient C' 0P is
pump. Hence, the dimensionless quantity e(node,namely the pressure fluctuation intensity coefficient CP is
defined as the root-mean-square of P ̃ t) divided by reference pressure Pref
defined as the root-mean-square of P (node,t) divided by reference pressure Pref
̃P
P rms(node)
erms (node)
0 '=
CC
P P= (14)
(14)
PPref
re f

̃rms (node) and Pref are shown in Equations (15) and (16), respectively.
where eP
where Prms (node) and Pref are shown in Equations (15) and (16), respectively.
N-1
t 1 N−1
v
u
̃rms (node)=√ 1 ∑
P XP ̃(node,t0 +i∇t)2 (15)
erms (node) = N
P e(node, t0 + i∇t)
P (15)
N i=0
i=0
1
Pref = 1ρu22 2 (16)
Pre f =2 ρu2 (16)
2
where u2 is the circumferential speed at the impeller outlet.
where u2 is the circumferential speed at the impeller outlet.
Figure 17 shows C0'P of the five monitoring nodes under different flow conditions. C'P0 at P1 is
Figure 17 shows CP of the five monitoring nodes under different flow conditions. CP at P1 is
higher than that at other nodes, and it presents a V-shaped variation trend with the increase of flow
higher than that at other nodes, and it presents a V-shaped variation trend with the increase of flow rate,
rate, where it reaches the maximum value (about 0.02) under Q0.6. C'P decreases to the minimum
(slightly higher than 0.01) under Q0.9. Subsequently, it increases continuously with the continuous
Appl. Sci. 2020, 10, 1186 18 of 33

Appl.
where Sci. 2020, 10, x FOR
thePEER REVIEWvalue (about 0.02) under Q0.6 . C0P decreases to the minimum (slightly18 of 35
Appl. Sci.it2020,
reaches
10, x FOR maximum
PEER REVIEW 18 of 35
higher than 0.01) under Q0.9 . Subsequently, it increases continuously with the' continuous increase of
increase
flow rate,of flow
but it israte, but than
still less it is still in
less than that in part-load conditions.
C0P values CP' are
values at P2–P5lower
are
increase of flow rate, but it is that part-load
still less conditions.
than that in part-load at P2–P5
conditions. significantly
CP values at P2–P5 are
significantly
than lower
that at P1, than
andthan that
theythat at P1, and
are generally they are generally higher under part-load conditions.
significantly lower at P1, andhigher under
they are part-load
generally conditions.
higher under part-load conditions.
0.020
0.020
P1

pulsation
P1
P2
pulsation
P2
P3
P3
P4
0.015 P4
P5
0.015 P5
pressure
pressure

0.010
of of

0.010
coefficient
coefficient

0.005
0.005
Strength
Strength

0.000
0.000 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
0.6 0.7 0.8 0.9 Q/Qdes1.1
1.0 1.2 1.3 1.4 1.5
Q/Qdes

Figure
Figure17.
17.Comparisons
Comparisonsof
ofpressure
pressurefluctuation
fluctuationintensity
intensitycoefficient
coefficientunder
underdifferent
differentflow
flow conditions.
Figure 17. Comparisons of pressure fluctuation intensity coefficient under different flow conditions.
conditions.
4.4.Deformation
4.4. Deformationand
andVibration
VibrationBehaviors
Behaviorsofofthe
theMain
MainShaft
Shaft
4.4. Deformation and Vibration Behaviors of the Main Shaft
4.4.1.Deformation
4.4.1. Deformationand andStress
StressDistribution
Distribution
4.4.1. Deformation and Stress Distribution
Thiswork
This workfocuses
focuseson
onthe
themechanics
mechanicsbehaviors
behaviorsof ofthe
therotor
rotorsystem
systemcaused
causedbybythe
theinner
innerunsteady
unsteady
This
flow. At work
Atfirst,
first,thefocuses on the
thedisplacement mechanics
displacementand behaviors
andequivalent
equivalentstress of the rotor
stressdistribution system
distribution of
ofthe caused
themain by
mainshaft the
shaftat inner
atthe
theend unsteady
endmoment
moment
flow.
flow.
of At first,rotordynamics
unsteady the displacement and equivalent
analysis under Q stress
des were distribution
shown in of the 18
Figures main shaft
and 19, at the end moment
respectively.
of unsteady rotordynamics analysis under Qdes were shown in Figures 18 and 19, respectively.
of unsteady rotordynamics analysis under Qdes were shown in Figures 18 and 19, respectively.

(a)
(a)

(b)
(b)

(c)
(c)
Figure 18. Cont.
 (d)
Figure 18. Displacement contour of the main shaft under Qdes (Unite: m, magnification factor: 250): (a)
Appl. Sci. 2020, 10, 1186 19 of 33
X direction; (b) Y direction; (c) Z direction; (d) Total.
Appl. Sci.
Appl. Sci. 2020,
2020, 10,
10, x
x FOR
FOR PEER
PEER REVIEW
REVIEW 19 of
19 of 35
35

(d)
(d)
FigureFigure 19. Equivalent
18. Displacement
Displacement contour stress
ofofthe
the contour
main shaftofunder
the main
QQ shaftm,under
(Unite: Qdes (Unit:
magnification Pa).
factor: 250): (a)
Figure 18. Displacementcontour
contourof themain
mainshaft
shaftunder
underQdes (Unite: m, m,
des (Unite:
des magnification factor:
magnification 250):
factor: (a)
250):
X
X direction;
(a)direction; (b) (b)
X direction;
(b) Y direction;
Y direction; (c) Z
Y direction;
(c) Z direction;
(c)direction; (d) (d)
Z direction;
(d) Total.
Total.
Total.
Figure 18 shows that the deformation occurs mainly in radial directions. The maximum
displacements in X and Y directions reach 0.127 mm and 0.194 mm, while it is only about 0.02 mm in
Z direction (axial direction). Furthermore, deformation of the main shaft is symmetric around the XY
plane located on the origin, where the deformation reaches maximum. On the contrary, the minimum
deformation is at the bearing node due to its constraint function.
Figure 19 the equivalent stress reaches maximum (22.3 MPa) is at the center of the main shaft,
and the strength safety coefficient
Figure
Figure 19. is about
19. Equivalent
Equivalent stress 15.2. The
stress contour
contour of theequivalent
of the main shaft
main stress
shaft under
shaft under
under des decreases
Qdes
Q (Unit: Pa). gradually from the
Figure 19. Q des (Unit:
(Unit: Pa).
Pa).
center of the main shaft to its end.
Figure 18 18 shows
The time Figure
series ofshows that the
that
equivalent the deformation
the deformation occurs
deformation
stress occurs mainly
occurs
at the center mainly in
mainly
of main in radial
in radial directions.
radial
shaft directions. The
directions.
under The
different maximum
The maximum
flow rates were
displacements
displacements in
displacements in X
inXXandand Y
andY Y directions
directions
directions reach
reach
reach 0.127 mm
0.127
0.127 mmmm and
andand0.194 mm,
0.1940.194
mm, mm, while it
whilewhile is only about
it isabout
it is only 0.02
only 0.02
about mm
mm in
0.02
in
gained andZ
analyzed.
direction
Similarly
(axial direction).
direction).
to the
Furthermore,
data processing
deformation
of
of the
pressure
the main
main shaft
fluctuation
is shaft
symmetric
curves in
around around
Section
the XY
XY
4.3,
mm
Z in Z direction
direction (axial (axial direction).
Furthermore,Furthermore, deformation
deformation of of the main
shaft is is symmetric
symmetric around the
the time-varying
plane
the XY
plane
equivalent
located
plane
located on
located
on the onstress
the origin,
origin, where
the was thedecomposed
origin,
where the deformation into mean
reaches
where the deformation
deformation reaches
and fluctuation
maximum.
reaches
maximum. On the contrary,
maximum.
On the
components,
contrary,Onthethe minimum
theminimum
contrary, as shown
in Figuredeformation
20.minimumis
deformation
the isdeformation
at the
at the bearing
bearing node
is node due
at thedue to its
bearing
to itsnode
constraint function.
due tofunction.
constraint its constraint function.
Figure
Figure 19
In Figure 20, the mean
19 the
the equivalent
equivalent
equivalent stress
and fluctuation
stress reaches
reaches maximum
stress reachescomponents
maximum
maximum (22.3(22.3
(22.3 MPa)
ofMPa)
MPa) is
equivalent at
is at
is the
at the center
the stress
center of
center of
of the
reach main
their
the main
the main shaft,
minimums
shaft,
shaft,
and
and the strength safety coefficient is about
about 15.2.
15.2. The equivalent
equivalent stress
stress decreases
decreases gradually
gradually from
from the
at Q0.8 and Q0.9, respectively. Both of them increase significantly with flow rate under the
and the strength safety coefficient is The equivalent stress decreases gradually from large flow
center of
center of the
the main
main shaft
shaft toto its
its end.
end.
conditions. Stress
The time
and
time series
its fluctuation
series of of equivalent
equivalent
level
stress
aretheimportant
at center of
factors
main shaft
for thedifferent
under
fatigueflow failure ofwere
rates were
main shaft
The equivalent stress
stress atat the
the center
center of
of main
main shaft
shaft under
under different
different flow rates
[34], which
gained
gained
gained
is and
one
and of the most
analyzed.
and analyzed.
analyzed. common
Similarly
Similarly
Similarly to the
to the
to
failure
data
the data
data modes
processing
processing
processing ofof
of
of
pumps.
pressure
pressure
pressure
Therefore,
fluctuation
fluctuation
fluctuation curves
curves
curves
the
in equivalent
Section
in Section
in Section 4.3, stress
4.3,
4.3,
information
the predicted equivalent
the time-varying
time-varying
time-varying by the proposed
equivalent stress was
stress model in this
was decomposed
decomposed into workand
into mean
mean and
andis fluctuation
able to provide
fluctuation
fluctuation components,
components,
components, useful quantitative
as shown
as shown
shown
in Figure
referenceininformation
Figure 20.20. for the fatigue strength analysis of centrifugal pumps.
In Figure
In Figure 20,
20, the
the mean
mean andand fluctuation
fluctuation components
components of of equivalent
equivalent stress
stress reach
reach their
their minimums
minimums
at Q and Q , respectively. Both of them increase significantly with flow
at Q0.8 and Q0.9, respectively. Both of them increase significantly with flow rate under large
0.8 0.9 45 rate
0.5 under large flow
flow
conditions. Stress and its fluctuation level are important
Mean value factors for the fatigue failure
conditions. Stress and its fluctuation level are important factors for the fatigue failure of main shaft of main shaft
[34], which
which isis one
one of of 40
the most
most common
common failure Fluctuating
failure modescomponent
of pumps.
pumps.value Therefore, the equivalent stress
[34], the modes of Therefore, the equivalent stress
0.4
Fluctuating component value/MPa

information predicted by the proposed model in this

information predicted by the proposed model in this work is able to work is able to provide
provide useful
useful quantitative
quantitative
35
reference information for the fatigue strength analysis of centrifugal
reference information for the fatigue strength analysis of centrifugal pumps. pumps.
Mean value/MPa

0.3
30 45 0.5
45 0.5
Mean value
Mean value
25 40 Fluctuating component
Fluctuating component value
value
40
0.4 0.2
value/MPa

0.4
componentvalue/MPa

35
35
20
value/MPa

0.3
Meanvalue/MPa

30 0.3 0.1
30
Fluctuatingcomponent

15
25
25 0.2
0.2
Mean

10 0.0
Fluctuating

0.6
20
20 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
Q/Qdes 0.1
0.1
15
15

Figure 20. Comparisons

Figure of of
20. Comparisons thethe equivalent stress
equivalent stress in thein theofcenter
center the shaftof thedifferent
under shaft under different flow
flow conditions.
10
10 0.0
0.0
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
conditions. 0.6 0.7 0.8 0.9 1.0
Q/Q
1.1 1.2 1.3 1.4
Q/Qdes
des

Figure 20.
Figure 20. Comparisons
Comparisons of
of the
the equivalent
equivalent stress
stress in
in the
the center
center of
of the
the shaft
shaft under
under different
different flow
flow
conditions.
conditions.
Appl. Sci. 2020, 10, 1186 20 of 33

In Figure 20, the mean and fluctuation components of equivalent stress reach their minimums
at Q0.8 and Q0.9 , respectively. Both of them increase significantly with flow rate under large flow
conditions. Stress and its fluctuation level are important factors for the fatigue failure of main
shaft [34], which is one of the most common failure modes of pumps. Therefore, the equivalent stress
information predicted by the proposed model in this work is able to provide useful quantitative
reference
Appl. information
Sci. 2020, for REVIEW
10, x FOR PEER the fatigue strength analysis of centrifugal pumps. 20 of 35

4.4.2. Vibration
4.4.2. Vibration Behaviors
Behaviors
The central
The central point
point ofof main
main shaft
shaft waswas chosen
chosen forfor the
the vibration
vibration behavior
behavior analysis
analysis in
in this
this work,
work,
where the impeller is mounted. Its time-varying acceleration and velocity curves
where the impeller is mounted. Its time-varying acceleration and velocity curves under different flowunder different flow
conditions are shown in Figures 21 and 22, respectively. In addition, the corresponding
conditions are shown in Figures 21 and 22, respectively. In addition, the corresponding frequency frequency
distributions were
distributions were shown
shown in in Figures
Figures 23 23 and
and 24.
24.
It can be seen from Figures 23
It can be seen from Figures 23 and 24 that and 24 that for
for both
both acceleration
acceleration and
and velocity,
velocity, the
the dominant
dominant
frequency of the shaft vibration
frequency of the shaft vibration is fBPF is f (148 Hz), and its harmonics (296 Hz, 444 Hz,
BPF (148 Hz), and its harmonics (296 Hz, 444 Hz, …) are also
. . . ) are also
obvious. Besides,
obvious. Besides, harmonic
harmoniccomponent
component of both the vibration
of both acceleration
the vibration and velocity
acceleration is more is
and velocity obvious
more
under part-load conditions, so there are some distinct small peaks for the corresponding
obvious under part-load conditions, so there are some distinct small peaks for the corresponding time domain
time
curves in Figures 21 and 22. Nevertheless, harmonic component is weakened
domain curves in Figures 21 and 22. Nevertheless, harmonic component is weakened gradually with gradually with the
increase of flow rate and there are little harmonic components under large flow
the increase of flow rate and there are little harmonic components under large flow rates, where the rates, where the
time-domain sine
time-domain sine curves
curves are
are relatively
relatively smooth.
smooth.

5 3
0 Q0.6 0 Q0.6
-5
-10 -3
4 4
0 Q 0 Q0.7
0.7
-4 -4
3 2
0 Q0.8 0 Q0.8
-3 -2
-2
-2

Accelerated velocity/ms
Accelerated velocity/ms

3 1
0 Q0.9 0 Q0.9
-3 -1
2 1
0 Qdes 0 Qdes
-2 -1
2 1
0 Q1.1 0 Q1.1
-2 -1
3 1
0 Q1.2 0 Q1.2
-3 -1
3 2
0 Q 0 Q1.3
1.3
-3
-2
5 2
0 Q 0 Q1.4
1.4
-5
-2
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
Period of the impeller Period of the impeller
(a) (b)
Figure
Figure 21.
21. Time-varying
Time-varying acceleration
acceleration under
under different
different flow
flow conditions:
conditions: (a)
(a) X
X direction;
direction; (b)
(b) Y
Y direction.
direction.
Appl. Sci. 2020, 10, 1186 21 of 33

Appl.Sci.
Appl. Sci.2020,
2020,10,
10,xxFOR
FORPEER
PEERREVIEW
REVIEW 21ofof35
21 35

5 3
5 3
0 Q0.6 0 Q0.6
0 Q0.6 0 Q0.6
-5 -3
-54 -34
4 4
0 Q0.7 0 Q0.7
0 Q0.7 0 Q0.7
-4 -4
-4 -42
3 2
3
0 Q0.8 0 Q0.8
0 Q0.8 0 Q0.8
-3 -2
-3 -2
3 1
3 1
0 Q0.9 0 Q0.9

-1
-1

0 Q0.9 0 Q0.9

-1
-1

/mmmss
Velocity/mmss

-3 -1
-3 -1
Velocity/mm

2 1

Velocity/m
2 1
0 Qdes 0 Qdes

Velocity
0 Qdes 0 Qdes
-2 -1
-2 -12
2 2
2
0 Q1.1 0 Q1.1
0 Q1.1 0 Q1.1
-2 -2
-2 -21
3
3 1
0 Q1.1 0 Q1.2
0 Q1.1 0 Q1.2
-3 -1
-3 -1
3 2
3 2
0 Q1.1 0 Q1.3
0 Q1.1 0 Q1.3
-3 -2
-3 -22
5 2
5
0 Q1.1 0 Q1.4
0 Q1.1 0 Q1.4
-5 -2
-5 -2
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
Periodofofthe
theimpeller
impeller Periodofofthe
Period theimpeller
impeller
Period
(a)
(a) (b)
(b)
Figure 22.
Figure22.
Figure Time-varying velocity
Time-varyingvelocity
22.Time-varying under
velocityunder different
underdifferent flowconditions:
differentflow
flow conditions:(a)
conditions: (a) Xdirection;
(a)XX direction;(b)
direction; (b)YY
(b) Ydirection.
direction.
direction.
2.0
5 2.0
5

4 1.5
4
Magnitude/ms -2

1.5
Magnitude/ms
-2

Magnitude/ms -2
Magnitude/ms
-2
3
148Hz 3
148Hz 1.0
1.0
2
2

296Hz 0.5
296Hz 444Hz 1 0.5
1
592Hz
444Hz 592Hz
0 0.0
1.40 1.40.0
1.4
1.3 1.4
1.3
1.2 1.3
1.2 1.3
1.2
1.1 1.2
1.1 Q/Q
1.0
1.1
1.0 Q/Qdes des 1.1
1.0
0.9 1.0 Q/Q
0.8 Q/Qdes
0.9 0.9 des
0.8 0.9
0.8
0.7 0.8
0.7 0.7
0.6 0.7
0.6
0.6 0.6
0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
Frequency/Hz Frequency/Hz
Frequency/Hz Frequency/Hz
(a)
(a) (b)
(b)
Figure23.
Figure 23. Frequencydistribution
distribution of theacceleration
acceleration underdifferent
different flowconditions:
conditions: (a)XXdirection;
direction;
Figure 23.Frequency
Frequency distributionofofthe
the accelerationunder
under differentflow
flow conditions:(a)
(a) X direction;
(b)
(b) Y direction.
(b)YYdirection.
direction.
2.0
5 2.0
In the engineering practice, the peak and 5 RMS values in a certain period of time are counted

for the vibration acceleration and velocity signals, 4

4 respectively. The acceleration peak indicates the
1.5
Magnitude/ms -1

1.5
Magnitude/ms -1
Magnitude/ms
-1

Magnitude/ms
-1

external force induced the vibration, and the3RMS 3 velocity reflects the strength of vibration energy.
1.0
Figure 25 compares the above two vibration indicators in the center of the main shaft under different 1.0
2
2
flow conditions. 0.5
1 0.5
1
It can be seen from Figure 25 that acceleration peak and velocity RMS are relatively low around
0 0.0
Qdes . Given large flow (Q1.3 or higher) or part-load 1.40
1.4
1.3
conditions (Q0.7 or lower), both acceleration peak
1.40.0
1.4
1.3
1.2 1.3
1.3
and velocity RMS increase significantly with 1.2
1.1
1.1
1.0
the increase of deviation from Qdes , and this phenomena 1.2
1.2
1.1
1.0 Q/Q 1.1
1.0
1.0 Q/Q
0.9
Q/Q des
is more obvious under part-load conditions.
0.8
0.7
0.9
0.8 des 0.8 Q/Q
0.9
0.9
0.8
des
des
0.7 0.7
0.6 0.7
0.6 0.6
0 100 200 300 400 500 600 700 800 900 1000 0.6
0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
Frequency/Hz 0 100 200 300 400 500 600 700 800 900 1000
Frequency/Hz Frequency/Hz
Frequency/Hz
(a)
(a) (b)
(b)
Figure24.
Figure 24.Frequency
Frequencydistribution
distributionofofthe
thevelocity
velocityunder
underdifferent
differentflow
flowconditions:
conditions:(a)
(a)XXdirection;
direction;(b)
(b)
YYdirection.
direction.
 des 1.0
0.9 0.9 Q/Qdes
0.8 0.8
0.7 0.7
0.6 0.6
0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
Frequency/Hz Frequency/Hz
(a) (b)
Appl. Figure
Sci. 2020,
23.10,Frequency
1186 distribution of the acceleration under different flow conditions: (a) X direction;22 of 33
(b) Y direction.

2.0
5

4 1.5

-1

-1
Magnitude/ms

Magnitude/ms
3
1.0
2

0.5
1

0 0.0
1.4 1.4
1.3 1.3
1.2 1.2
1.1 1.1
1.0
Appl. Sci. 2020, 10, x FOR PEER REVIEW Q/Qdes 1.0
22 of 35
0.9
0.8 0.9 Q/Qdes
0.8
0.7
0.7
0.6
0.6
0 100
In 200
the300 400 500 600 700 800 900 1000
engineering
Frequency/Hz
practice, the peak and RMS values0 100 200in300
a certain period
400 500 600 of time
700 800 900 1000are counted for
Frequency/Hz
the vibration acceleration (a)
and velocity signals, respectively. The acceleration (b)
peak indicates the
external force induced the vibration, and the RMS velocity reflects the strength of vibration energy.
Figure
Figure 24.
24.Frequency
25 compares
Figure Frequency distribution
the above of of
distribution thethe
two vibration velocity under
velocity under
indicators different
in flow
different
the center conditions:
flow
of main(a)
conditions:
the X(a)
shaft direction; (b)
X direction;
under different
Y direction.
(b) Y direction.
flow conditions.

10 5.0 4.5 3.0

Acceleration peak in X direction Acceleration peak in X direction

4.5
Acceleration peak in Y direction 4.0 Acceleration peak in Y direction
-1

-1
-2

2.5
Acceleration peak in X direction/mms

Acceleration peak in Y direction/mms

-2
Acceleration peak in X direction/ms

Acceleration peak in Y direction/ms

8
4.0
3.5
2.0
3.5
6 3.0
3.0 1.5
2.5
2.5
4 1.0
2.0
2.0

0.5
1.5 1.5
2

1.0 1.0 0.0

0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
Q/Qdes Q/Qdes

(a) (b)
Figure
Figure25.
25.Comparisons
Comparisonsofofthe
thevibration
vibrationindicators
indicatorsunder
underdifferent
differentflow
flowconditions:
conditions:(a)
(a)Acceleration
Acceleration
peak;
peak;(b)
(b)Velocity
VelocityRMS.
RMS.

4.5. It
Evolution Characters
can be seen of the Unsteady
from Figure Flow Field peak and velocity RMS are relatively low around
25 that acceleration
Qdes. Given large flow (Q1.3 or higher) or part-load conditions (Q0.7 or lower), both acceleration peak
4.5.1. Flow Fields under Qdes
and velocity RMS increase significantly with the increase of deviation from Qdes, and this phenomena
is more It is the unstable
obvious under flow in theconditions.
part-load centrifugal pump that induces the various radial force fluctuations,
pressure pulsations, and vibration behaviors presented in the preceding sections. Inner flow field
4.5. Evolution Characters
observation reveals there of the
areUnsteady
obviousFlow flowField
distribution changes in the discharge chamber under
different flow rates and different impeller phases, while it keeps fairly uniform and stable in the suction
4.5.1.
chamberFlowand Fields under zone.
impeller Qdes Thus, the internal flow analysis mainly concentrates on the discharge
chamber in this
It is the work.flow in the centrifugal pump that induces the various radial force fluctuations,
unstable
Firstly, the inner
pressure pulsations, and flowvibration
distribution under Qpresented
behaviors des was analyzed
in the and used assections.
preceding the basisInner
for theflow
follow-up
field
comparative study of flow fields under different flow rates. In this study,
observation reveals there are obvious flow distribution changes in the discharge chamber under the rotation phase of impeller
was defined
different flowzero
rateswhen
and the impeller
different center,phases,
impeller tip of aspecific
while it blade,
keeps and
fairlytipuniform
of volute
andtongue
stablemeet
in theon
a straight line, as illustrated in Figure 4.
suction chamber and impeller zone. Thus, the internal flow analysis mainly concentrates on the
The velocity
discharge chambervectorin thisdistributions
work. on middle section of the discharge chamber with impeller phase
of 0◦Firstly,
and 30the ◦ were shown in Figure 26a,b, respectively. Obviously, fluid media sheds from outlet of
inner flow distribution under Qdes was analyzed and used as the basis for the follow-
blade
up passages along
comparative study the tangential
of flow fields direction and then
under different flowenters
rates.the
Indischarge
this study, chamber with phase
the rotation relatively
of
impeller was defined zero when the impeller center, tip of aspecific blade, and tip of volute tonguein
regular distribution excepet for gentle impact near the tongue and the consequent secondary flow
the diffuser
meet (marked
on a straight line,with dotted frame).
as illustrated in Figure 4.
The velocity vector distributions on middle section of the discharge chamber with impeller
phase of 0° and 30° were shown in Figure 26a,b, respectively. Obviously, fluid media sheds from
outlet of blade passages along the tangential direction and then enters the discharge chamber with
relatively regular distribution excepet for gentle impact near the tongue and the consequent
secondary flow in the diffuser (marked with dotted frame).
Appl. Sci. 2020, 10, 1186 23 of 33

Figure 27 shows the turbulence kinematic energy distribution on the middle plane of the impeller
zone and discharge chamber. It is easy to find that there are not any obvious differences between the
0◦ and 30◦ impeller phases for both velocity and turbulence energy distribution. In Figure 28, high
turbulence energy areas mainly locate at the blade wake region and the diffuser region nearby the
tongue, where there are typical unstable flow structures such as jet-wake, cutting, and impact. Under
Qdes , the maximum values of turbulence energy under 0◦ and 30◦ of impeller phases are close, and they
are 5.31 J kg−1 and 6.54 J kg−1 , respectively.

4.5.2. Flow Fields under Q0.6 and Q1.4

Due to space limitations, the minimum and maximum flow rates (Q0.6 and Q1.4 ) in the calculation
range were chosen as the representative to analyze the evolution laws of flow field. The velocity vector
and turbulence kinematic energy distributions under these two operation conditions were shown from
Figures 28–31.
Figure 28 indicates strong flow impingement and cutting near the tongue, followed by highly
disordered secondary flow and stall flow phenomenons (marked by dotted frame) in the diffuser. In
response, Figure 30 shows the turbulence kinematic energy is high in large scale, especially for the
clearance between blade tips and volute wall, the blade passages near pressure surfaces, and laryngeal.
Furthermore, the flow field differences between the 0◦ and 30◦ impeller phases are significant both for
the velocity and turbulence kinematic energy distributions.
While under Q1.4 , as shown in Figures 29 and 31, there is no visible differences between those
two blade phases. The velocity vector distributions are stable and uniform in the impeller zone and
tip clearance. Although there is large scale stall flow with high turbulence kinematic energy in the
diffuser, but it is in the downstream and far away from impeller zone, so its impact on the impeller is
very limited.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 23 of 35

(a)

(b)
Figure 26. Velocity
Figure vector
26. Velocity distribution
vector distributionon
onthe
the middle sectionofofthe
middle section the discharge
discharge chamber
chamber plane
plane underunder
Qdes :Q(a)
des: Impeller
(a) Impeller phase
phase atat0◦0°; (b)Impeller
; (b) Impeller phase 30◦ .
phaseatat30°.

Figure 27 shows the turbulence kinematic energy distribution on the middle plane of the
impeller zone and discharge chamber. It is easy to find that there are not any obvious differences
between the 0° and 30° impeller phases for both velocity and turbulence energy distribution. In Figure
28, high turbulence energy areas mainly locate at the blade wake region and the diffuser region
nearby the tongue, where there are typical unstable flow structures such as jet-wake, cutting, and
 Figure 27 shows the turbulence kinematic energy distribution on the middle plane of the
impeller zone and discharge chamber. It is easy to find that there are not any obvious differences
between the 0° and 30° impeller phases for both velocity and turbulence energy distribution. In Figure
28, high turbulence energy areas mainly locate at the blade wake region and the diffuser region
nearby the tongue, where there are typical unstable flow structures such as jet-wake, cutting, and
Appl. Sci. 2020, 10,
impact. 1186 Qdes, the maximum values of turbulence energy under 0° and 30° of impeller phases
Under 24 of 33
are close, and they are 5.31 J kg−1 and 6.54 J kg−1, respectively.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 24 of 35

Appl. Sci. 2020, 10, x FOR PEER REVIEW (a) 24 of 35

(b)
Figure 27. Turbulence kinematic energy contour distribution on the middle section of the discharge
chamber under Qdes: (a) Impeller phase at 0°; (b) Impeller phase at 30°.

4.5.2. Flow Fields under Q0.6 and Q1.4

Due to space limitations, the minimum and maximum flow rates (Q0.6 and Q1.4) in the calculation
range were chosen as the representative to analyze the evolution laws of flow field. The velocity
vector and turbulence kinematic energy distributions under these two operation conditions were
shown from Figures 28–31.
Figure 28 indicates strong flow impingement and cutting near the tongue, followed by highly
disordered secondary flow and stall flow phenomenons (marked by dotted frame) in the diffuser. In
response, Figure 30 shows the turbulence kinematic energy is high in large scale, especially for the
(b)
clearance between blade tips and volute wall, the blade passages near pressure surfaces, and
Figure 27. Turbulence
laryngeal.
Figure Furthermore,
27. Turbulence kinematic
the flowenergy
kinematic field contour
energy differencesdistribution
between
contour distribution onon
thethe
the 0°middle section
andsection
middle 30° of discharge
impeller
of the thephases
discharge
are
chamber
chamber under
significant under
both Q
for Q :
the(a)
des : (a) Impeller
Impeller
velocity and phase
phase at 0 ◦ ;(b)
0°;
turbulence (b)Impeller
Impeller
kinematic phase at
phase
energy 30°.
at 30◦.
distributions.
des

4.5.2. Flow Fields under Q0.6 and Q1.4

Due to space limitations, the minimum and maximum flow rates (Q0.6 and Q1.4) in the calculation
range were chosen as the representative to analyze the evolution laws of flow field. The velocity
vector and turbulence kinematic energy distributions under these two operation conditions were
shown from Figures 28–31.
Figure 28 indicates strong flow impingement and cutting near the tongue, followed by highly
disordered secondary flow and stall flow phenomenons (marked by dotted frame) in the diffuser. In
response, Figure 30 shows the turbulence kinematic energy is high in large scale, especially for the
clearance between blade tips and volute wall, the blade passages near pressure surfaces, and
laryngeal. Furthermore, the flow field differences between the 0° and 30° impeller phases are
significant both for the velocity and turbulence kinematic
(a) energy distributions.

(b)
(a)
Figure 28. Velocity
Figure vector
28. Velocity vector distribution
distribution on themiddle
on the middle section
section of the
of the discharge
discharge chamber
chamber 0.6: (a) Q0.6 :
under
under Q
Impellerphase
(a) Impeller phaseat ◦ ; (b)
at 00°; (b)Impeller
Impellerphase
phaseat at 30◦ .
30°.
 Appl. Sci. 2020, 10, x FOR PEER REVIEW 25 of 35

While under Q1.4, as shown in Figures 29 and 31, there is no visible differences between those
two blade phases. The velocity vector distributions are stable and uniform in the impeller zone and
tip clearance. Although there is large scale stall flow with high turbulence kinematic energy in the
Appl. Sci. 2020, 10, 1186 25 of 33
diffuser, but it is in the downstream and far away from impeller zone, so its impact on the impeller
is very limited.

(a)

(b)
Figure 29. Velocity
Figure vector
29. Velocity vectordistribution on the
distribution on themiddle
middle section
section of discharge
of the the discharge chamber
chamber 1.4: (a) Q1.4 :
under
under Q
(a) Impeller ◦ ; (b) Impeller phase at 30◦ .
Impellerphase
phase at
at 00°; (b) Impeller phase at 30°.

Appl. Sci. 2020, 10, x FOR PEER REVIEW (a) 26 of 35

(b)
Figure 30. Turbulence
Figure 30. Turbulencekinematic energy
kinematic energycontour
contour distribution onthe
distribution on themiddle
middle section
section of the
of the discharge
discharge
chamber
chamber underunder
Q0.6Q 0.6: (a)
: (a) Impellerphase
Impeller phaseat ◦ ; (b) Impeller
at00°; Impellerphase
phaseatat 30◦ .
30°.
 (b)
Appl. Sci. 2020, 10, 1186
Figure 30. Turbulence kinematic energy contour distribution on the middle section of the discharge 26 of 33
chamber under Q0.6: (a) Impeller phase at 0°; (b) Impeller phase at 30°.

(a)

(b)
Figure 31. Turbulence
Figure 31. Turbulence kinematic
kinematicenergy
energycontour distributionononthe
contour distribution the middle
middle section
section of discharge
of the the discharge
chamber
chamber under Q1.4Q: 1.4(a)
under : (a) Impellerphase
Impeller phase at
at 0 ◦ ; (b)
0°; (b)Impeller
Impellerphase
phase at 30◦ .
at 30°.

5. Discussion
5. Discussion
The above
The above series
series analysis
analysis provesthat
proves thathydraulic-induced
hydraulic-induced vibration
vibrationand deformation
and deformationof the
ofmain
the main
shaftshaft are closely
are closely relatedwith
related withthe
the unsteady
unsteady flow in in
flow centrifugal pump.
centrifugal Around
pump. the rated
Around theflow rate,
rated the rate,
flow
overall flow field is uniform and the secondary flow phenomenon is controlled well. Consequently,
the overall flow field is uniform and the secondary flow phenomenon is controlled well. Consequently,
the pressure pulsation and turbulence kinematic energy are both weak, which ensures the main shaft
the pressure pulsation and turbulence kinematic energy are both weak, which ensures the main shaft
work in stable operation with relatively low vibration and stress levels. Given the part-load
work in stable operation with relatively low vibration and stress levels. Given the part-load conditions,
conditions, the RSI is strong when the impeller blade sweeps over the volute tongue, and it brings a
the RSI is strong when the impeller blade sweeps over the volute tongue, and it brings a series of
unstable flow phenomena, such as the flow cutting and impingement in the nearby region of tongue,
which has been observed by PIV measuring [10]. The inner unstable flow finally causes high fluctuation
of radial forces of the impeller and significant vibration of the main shaft. While in the large flow
conditions, although the radial forces on the impeller is high, their fluctuation degree is controlled
effectively and the main shaft vibration level is lower than that in part-load conditions. One possible
explanation is that the high flow suppresses the separation degree of fluid media, and it is the flow
rate factor rather than pump structure that dominant the entire flow process. However, illustrating
the flow induced mechanism of the rotor vibration is a very complex task, and a more accurate CFD
computation, such as detached eddy simulation [10] and some new comparative analysis methods,
may be needed in the further research.
Compared with the two-way FSI study of the centrifugal pump made by Benra [16] and Pei [14,20],
this work pays more attention on the vibration behaviors of the main shaft rather than the impeller,
considering the commonly occurring shaft fracture failures. Furthermore, the Coriolis force and
bearing characteristics have been included in the model of this work, which were ignored in previous
research [14,16,20].
Appl. Sci. 2020, 10, 1186 27 of 33

6. Conclusions
The present work emphasizes on the development of a sequential coupling mathematical model
for unsteady flow field-rotordynamics of the centrifugal pump as well as the corresponding numerical
calculation program. A horizontal split centrifugal pump with double-entry impeller was chosen for
research, and the validity and accuracy of the computation were verified through its performance
test and vibration measurement. The coupling model presented in this work provided a series of
interconnected computation information including the inner flow field and deformation and vibration
behaviors of the external rotor system in a flexible high efficiency way, and some conclusions can be
drawn as follows:
(1) f BPF is the dominant frequency in the fluctuation of hydraulic radial force on the impeller,
pressure pulsation in the blade tip clearance, and vibration of main shaft, which are closely related to
the RSI of the centrifugal pump.
(2) Operating stability indexes like fluctuation of hydraulic radial forces on the impeller, vibration,
and deformation levels of the main shaft are predominated by the unsteady flow process in the
centrifugal pump. Those indexes are at the optimal level around the rated flow rate, but get deteriorated
when it deviates from the best work condition, which is more serious in part-load conditions.
(3) Under large flow rates, although the radial forces on the impeller are larger than other
conditions, the vibration levels can be controlled better than those under part-load conditions due to
their relatively stable inner flow field.
In sum, this work mainly discusses the rotor vibration behaviors induces by inner flow process of
centrifugal pump. A more precise unsteady CFD model should be conducted in future work to obtain
more accurate time-varying laws of the hydraulic exciting forces. Moreover, more measuring points,
including other non-intrusive sensors, needed to be supplemented in the experimental validation work.

7. Patents
There has been a Chinese invention patent resulted from this work: Zhang, H.H.; Zhou, T.H.
Vibration Simulation Analysis Method and Device of Vane Pump Rotor System. China Patent
ZL2016110994972, 2016.

Author Contributions: Conceptualization, H.Z. and H.Y.; methodology, H.Z.; software, H.Y. and K.L.; validation,
H.L. and Z.Z.; formal analysis, H.Z.; investigation, H.Z.; resources, H.Z.; data curation, H.Y. and K.L.;
writing—original draft preparation, H.Z.; writing—review and editing, H.Z. and H.Y.; visualization, H.Y.;
supervision, L.J.; project administration, H.Z.; funding acquisition, L.J. All authors have read and agreed to the
published version of the manuscript.
Funding: This research was funded by the National Natural Science Foundation of China, grant number 51674298.
Acknowledgments: The authors would like to acknowledge the constructive reviews given by anonymous reviewers.
Conflicts of Interest: The authors declare no conflict of interest.

Nomenclature
A [m2 ] Area
b2 [m] Impeller outlet width
[C] Damping matrix
gH
Ch Head coefficient (= u22 /2
)
Q
Cq Flow coefficient (= 3600πD2 b2 u2 )
C0p Pressure fluctuation intensity coefficient
D1 [m] Impeller inlet diameter
D2 [m] Impeller outlet diameter
F [N] Force
F1 Function value of the SST k-ω turbulence model
Appl. Sci. 2020, 10, 1186 28 of 33

Fk [N] Geostrophic force

f [N·m−3 ] Body force
f BPF [Hz] Blade passing frequency
fn [Hz] Shaft rotating frequency
[G] Gyroscopic matrix
g [m·s−2 ] Gravity constant
H [m] Head
Hdes [m] Rated head
[K] Stiffness matrix
[M] Mass matrix
m [kg] Mass
NPSHr [m] Required net positive suction head
n [rev min−1 ] Rotating speed
ndes [rev min−1 ] Rated rotating speed
p
ns Specific speed (= 3.65ndes Qdes /Hdes 0.75 )

p [Pa] Pressure
Pi [Pa] Total pressure of the inlet
Po [Pa] Total pressure of the outlet
Pref [Pa] Reference pressure
Q [m3 /h] Volume flow
Qdes [m3 /h] Rated flow
Qζ [m3 /h] Flow rate (subscript ζ represents the flow ratio with Qdes )
SΦ Source term of flow scalar
T [N m] Torque
t [s] Time
{U} Displacement matrix
u [m·s−1 ] Fluid velocity
Ug [m·s−1 ] Mesh velocity
u2 [m·s−1 ] Circumferential speed at the impeller outlet (=πD2 n/60)
V [m3 ] Volume
v [m·s−1 ] Solid velocity
v0 [m·s−1 ] Solid velocity relative to the reference rotating frame
Z Number of blades
α Mass damping coefficient
β Stiffness damping coefficient
∆t [s] Time step
∆Z [m] Height difference
η Efficiency
µ [kg·m−1 ·s−1 ] Effective viscosity
ρ kg·m−3 Density
Φ General flow scalar for an arbitrary control volume
ϕ Coefficients of the SST k-ω turbulence model
ϕ1 Coefficients of the SST k-ω turbulence model
ϕ2 Coefficients of the SST k-ω turbulence model
ψ Diffusion coefficient of flow scalar
ωr [rad/s] Angular velocity

Appendix A
APDL code of the rotordynamics numerical calculation program:
FINISH
*DIM,FFX,TABLE,2160,1,1,TIME
*DIM,FFY,TABLE,2160,1,1,TIME
*TREAD,FFX,’1.0-FX’,’TXT’,’ ’
*TREAD,FFY,’1.0-FY’,’TXT’,’ ’
/PREP7
L1 = 0.135
Appl. Sci. 2020, 10, 1186 29 of 33

L2 = 0.1025
L3 = 0.0915
L4 = 0.332
RZ1 = 0.021
RZ2 = 0.0225
RZ3 = 0.025
KX = 5E9
KY = 5E9
M_IMPELLER = 54.3
R_IMPELLER = 0.278
JZ = 1/2*M_IMPELLER*R_IMPELLER**2
JY = 1/2*JZ
JX = 1/2*JZ
RPM = 1480
ANGULAR_V = 2*3.14*RPM/60
T_PERIOD = 2*3.14/ANGULAR_V
C_N = 360
CYCLE = 6
ET,1,MASS21
ET,2,BEAM188
ET,3,214
R,1,M_IMPELLER,M_IMPELLER,M_IMPELLER,JX,JY,JZ
R,2,KX,KY
MP,EX,1,2.1E11
MP,PRXY,1,0.269
MP,DENS,1,7890
SECTYPE,1,BEAM,CSOLID„0
SECOFFSET,CENT
SECDATA,RZ1,15,10,0,0,0,0,0,0,0,0,0
SECTYPE,2,BEAM,CSOLID„0
SECOFFSET,CENT
SECDATA,RZ2,15,10,0,0,0,0,0,0,0,0,0
SECTYPE,3,BEAM,CSOLID„0
SECOFFSET,CENT
SECDATA,RZ3,15,10,0,0,0,0,0,0,0,0,0
K,1,0,0,-(L1+L2+L3+L4)
K,2,0,0,-(L2+L3+L4)
K,3,0,0,-(L3+L4)
K,4,0,0,-L4
K,5,0,0,0
K,6,0,0,L4
K,7,0,0,L3+L4
K,8,0,0,L2+L3+L4
K,9,0,0,L1+L2+L3+L4
L,1,2
L,2,3
L,3,4
L,4,5
L,5,6
L,6,7
L,7,8
L,8,9
ESIZE,0.002
TYPE,2
MAT,1
SECNUM,1
LMESH,1
SECNUM,2
LMESH,2
SECNUM,2
LMESH,3
SECNUM,3
LMESH,4
SECNUM,3
LMESH,5
SECNUM,2
LMESH,6
Appl. Sci. 2020, 10, 1186 30 of 33

SECNUM,2
LMESH,7
SECNUM,1
LMESH,8
NODEP = NODE(0,0,0)
TYPE,1
REAL,1
E,NODEP
NODEZ1 = NODE(0,0,-(L3+L4))
NODEZ2 = NODE(0,0,L3+L4)
N,10000,0.08,0,-(L3+L4)
N,10001,0.08,0,L3+L4
TYPE,3
REAL,2
E,10000,NODEZ1
E,10001,NODEZ2
/SOL
ANTYPE,4
TRNOPT,FULL
BETAD,0.002
D,10000,ALL
D,10001,ALL
OMEGA,0,0,ANGULAR_V
CORIOLIS,1„,1,
FLST,2,1,1,ORDE,1
FITEM,2,37
!*
!*
/GO
F,P51X,FX, %FFX%
FLST,2,1,1,ORDE,1
FITEM,2,37
!*
!*
/GO
F,P51X,FY, %FFY%
ALLS
OUTRES,ALL,ALL
TIME,T_PERIOD*CYCLE
AUTOTS,1
NSUBST,C_N*CYCLE,C_N*CYCLE,C_N*CYCLE,1
KBC,0
SOLVE
SAVE
/POST26
ALLS
NUMVAR,200
FILLDATA,191„„1,1
REALVAR,191,191
!*
NSOL,2,126,U,X,UX_126,
STORE,MERGE
!*
NSOL,3,126,U,Y,UY_126,
STORE,MERGE
!*
NSOL,4,126,V,X,VX_126,
STORE,MERGE
!*
NSOL,5,126,V,Y,VY_126,
STORE,MERGE
!*
NSOL,6,126,A,X,AX_126,
STORE,MERGE
!*
NSOL,7,126,A,Y, AY_126,
STORE,MERGE
Appl. Sci. 2020, 10, 1186 31 of 33

NSOL,8,37,U,X,UX_37,
STORE,MERGE
!*
NSOL,9,37,U,Y,UY_37,
STORE,MERGE
!*
NSOL,10,37,V,X,VX_37,
STORE,MERGE
!*
NSOL,11,37,V,Y,VY_37,
STORE,MERGE
!* NSOL,12,37,A,X,AX_37,
STORE,MERGE
!*
NSOL,13,37,A,Y, AY_37,
STORE,MERGE
NUMVAR,200
FILLDATA,191„„1,1
REALVAR,191,191
FILLDATA,191„„1,1
REALVAR,191,191
*CREATE,SCRATCH,GUI
*DEL,_P26_EXPORT
*DIM,_P26_EXPORT,TABLE,720,6
VGET,_P26_EXPORT(1,0),1
VGET,_P26_EXPORT(1,1),2
VGET,_P26_EXPORT(1,2),3
VGET,_P26_EXPORT(1,3),4
VGET,_P26_EXPORT(1,4),5
VGET,_P26_EXPORT(1,5),6
VGET,_P26_EXPORT(1,6),7
/OUTPUT,’126’,”,’.’
*VWRITE,’TIME’,’UX_126’,’UY_126’,’VX_126’,’VY_126’,’AX_126’,’AY_126’
%14C %14C %14C %14C %14C %14C %14C
*VWRITE,_P26_EXPORT(1,0),_P26_EXPORT(1,1),_P26_EXPORT(1,2),_P26_EXP
ORT(1,3),_P26_EXPORT(1,4),_P26_EXPORT(1,5),_P26_EXPORT(1,6)
%14.5G %14.5G %14.5G %14.5G %14.5G %14.5G %14.5G
/OUTPUT,TERM
*END
/INPUT,SCRATCH,GUI
NUMVAR,200
FILLDATA,191„„1,1
REALVAR,191,191
*CREATE,SCRATCH,GUI
*DEL,_P26_EXPORT
*DIM,_P26_EXPORT,TABLE,C_N*CYCLE,6
VGET,_P26_EXPORT(1,0),1
VGET,_P26_EXPORT(1,1),8
VGET,_P26_EXPORT(1,2),9
VGET,_P26_EXPORT(1,3),10
VGET,_P26_EXPORT(1,4),11
VGET,_P26_EXPORT(1,5),12
VGET,_P26_EXPORT(1,6),13
/OUTPUT,’37’,”,’.’
*VWRITE,’TIME’,’UX_37’,’UY_37’,’VX_37’,’VY_37’,’AX_37’,’AY_37’
%14C %14C %14C %14C %14C %14C %14C
*VWRITE,_P26_EXPORT(1,0),_P26_EXPORT(1,1),_P26_EXPORT(1,2),_P26_EXPORT(1,3),_P26_EXP
ORT(1,4),_P26_EXPORT(1,5),_P26_EXPORT(1,6)
%14.5G %14.5G %14.5G %14.5G %14.5G %14.5G %14.5G
/OUTPUT,TERM
*END
/INPUT,SCRATCH,GUI
! END OF TIME HISTORY SAVE
FINISH
ALLS
ESEL,S,TYPE„2
ALLSEL,BELOW,ELEM
Appl. Sci. 2020, 10, 1186 32 of 33

!*
/SHRINK,0
/ESHAPE,1
/EFACET,1
/RATIO,1,1,1
/CFORMAT,32,0
/REPLOT
!*
NSEL,R,LOC,Z,-0.06,0.06
ESLN,R
FINISH
*DIM,MAXSEQV_NODE,TABLE,C_N*CYCLE,1,1
*DO,ISET,1,C_N*CYCLE,1
/POST1 SET,1„1„„ISET,
PLNSOL,S,EQV,0
*GET,MAXSEQV_NODEI,PLNSOL,0,MAX
*SET,MAXSEQV_NODE(ISET-1,0,1),(ISET-1)*T_PERIOD/C_N
*SET,MAXSEQV_NODE(ISET-1,1,1),MAXSEQV_NODEI
*ENDDO *MWRITE,MAXSEQV_NODE,MAX_VONS,TXT„KIJ,1,C_N*CYCLE,1
(F15.6)

References
1. Zhang, H.H.; Deng, S.X.; Qu, Y.J. High working efficiency of rapid custom design. World Pumps 2016, 3,
34–36. [CrossRef]
2. Shankar, A.; Kalaiselvan, V.; Umashankar; Subramaniam; Paramasivam; Shanmugamb; Hanigovszkic, N.
A comprehensive review on energy efficiency enhancement initiatives in centrifugal pumping system.
Appl. Energy 2016, 181, 495–513. [CrossRef]
3. Wang, Z.Y.; Qian, Z.D. Effects of concentration and size of silt particles on the performance of a double-suction
centrifugal pump. Energy 2017, 123, 36–46. [CrossRef]
4. Goman, V.; Oshurbekov, S.; Kazakbaev, V.; Prakht, V.; Dmitrievskii, V. Energy efficiency analysis of fixed-speed
pump drives with various types of motors. Appl. Sci. 2019, 9, 5295. [CrossRef]
5. Liu, Y.B.; Tan, L.; Wang, B.B. A review of tip clearance in propeller, pump and turbine. Energies 2018, 11,
2202. [CrossRef]
6. Adamkowski, A.; Henke, A.; Lewandowski, M. Resonance of torsional vibrations of centrifugal pump shafts
due to cavitation erosion of pump impellers. Eng. Fail. Anal. 2016, 70, 56–72. [CrossRef]
7. Fakhari, V.; Shokrollahi, S. A theoretical and experimental disturbance analysis in a product of inertia
measurement system. Measurement 2017, 107, 143–152. [CrossRef]
8. Tresser, S.; Dolev, A.; Bucher, I. Dynamic balancing of super-critical rotating structures using slow-speed
data via parametric excitation. J. Sound Vib. 2018, 415, 59–77. [CrossRef]
9. Wang, K.; Liu, H.L.; Zhou, X.H.; Wang, W.B. Experimental research on pressure fluctuation and vibration in
a mixed flow pump. J. Mech. Sci. Technol. 2016, 30, 179–184. [CrossRef]
10. Zhang, N.; Liu, X.K.; Gao, B.; Xia, B. DDES analysis of the unsteady wake flow and its evolution of
a centrifugal pump. Renew. Energy 2019, 141, 570–582. [CrossRef]
11. Zhang, N.; Liu, X.K.; Gao, B.; Wang, X.J.; Xia, B. Effects of modifying the blade trailing edge profile on
unsteady pressure pulsations and flow structures in a centrifugal pump. Int. J. Heat Fluid Flow 2019, 75,
227–238. [CrossRef]
12. Wang, L.K.; Lu, J.L.; Liao, W.L.; Zhao, Y.P.; Wang, W. Numerical simulation of the tip leakage vortex
characteristics in a semi-open centrifugal pump. Appl. Sci. 2019, 9, 5244. [CrossRef]
13. Kye, B.; Park, K.; Choi, H.; Lee, M.; Kim, J.H. Flow characteristics in a volute-type centrifugal pump using
large eddy simulation. Int. J. Heat Fluid Flow 2018, 72, 52–60. [CrossRef]
14. Pei, J.; Dohmen, H.J.; Yuan, S.Q.; Benra, F.-K. Investigation of unsteady flow-induced impeller oscillations of
a single-blade pump under off-design conditions. J. Fluids Struct. 2012, 35, 89–104. [CrossRef]
15. Degroote, J. Partitioned simulation of fluid-structure interaction. Arch. Comput. Methods Eng. 2013, 20,
185–238. [CrossRef]
16. Benra, F.-K. Numerical and experimental investigation on the flow induced oscillations of a single-blade
pump impeller. ASME J. Fluids Eng. 2006, 128, 783–793. [CrossRef]
Appl. Sci. 2020, 10, 1186 33 of 33

17. Campbell, R.L.; Paterson, E.G. Fluid–structure interaction analysis of flexible turbomachinery. J. Fluids Struct.
2011, 27, 1376–1391. [CrossRef]
18. Kan, K.; Zheng, Y.; Zhang, X.; Yang, C.X.; Zhang, Y.Q. Numerical study on unidirectional fluid-solid coupling
of Francis turbine runner. Adv. Mech. Eng. 2015, 7. [CrossRef]
19. Yin, T.Y.; Pei, J.; Yuan, S.Q.; Osman, M.K.; Wang, J.B.; Wang, W.J. Fluid-structure interaction analysis of
an impeller for a high-pressure booster pump for seawater desalination. J. Mech. Sci. Technol. 2017, 31,
5319–5328. [CrossRef]
20. Pei, J.; Meng, F.; Li, Y.J.; Yuan, S.Q.; Chen, J. Fluid–structure coupling analysis of deformation and stress in
impeller of an axial-flow pump with two-way passage. Adv. Mech. Eng. 2016, 8. [CrossRef]
21. Lu, J.X.; Liu, X.B.; Zeng, Y.Z.; Zhu, B.S.; Hu, B.; Yuan, S.Q.; Hua, H. Detection of the flow state for a centrifugal
pump based on vibration. Energies 2019, 12, 3066. [CrossRef]
22. Sakthivel, N.R.; Nair, B.B.; Elangovan, M.; Sugumaran, V.; Saravanmurugan, S. Comparison of dimensionality
reduction techniques for the fault diagnosis of mono block centrifugal pump using vibration signals. Eng. Sci.
Technol. Int. J. 2017, 17, 30–38. [CrossRef]
23. Kumar, A.; Kumar, R. Time-frequency analysis and support vector machine in automatic detection of defect
from vibration signal of centrifugal pump. Measurement 2017, 108, 119–133. [CrossRef]
24. Zhang, H.H.; Deng, S.X.; Qu, Y.J. Differential amplification method for flow structures analysis of centrifugal
pump between design and off-design points. J. Cent. South Univ. 2017, 24, 1443–1449. [CrossRef]
25. Zhang, H.H.; Deng, S.X.; Qu, Y.J. Numerical investigation of periodic fluctuations in energy efficiency in
centrifugal pumps at different working points. Energies 2017, 10, 342. [CrossRef]
26. Menter, F.R. Two-Equation eddy-visocity turbulence models for engineering applications. AIAA J. 1994, 32,
1598–1605. [CrossRef]
27. Liu, Y.B.; Tan, L. Spatial-temporal evolution of tip leakage vortex in a mixed flow pump with tip clearance.
ASME J. Fluids Eng. 2019, 141, 081302. [CrossRef]
28. Jiang, L.L.; Liu, H.L.; Wang, K.; Shao, C. Effects of radial diffuser outlet pattern on pulsation of sprayer pump.
J. Drain. Irrig. Mach. Eng. 2016, 34, 289–293.
29. Jeffcott, H.H. The lateral vibration of loaded shafts in the neighborhood of a whirling speed. Philos. Mag.
1919, 219, 304–314. [CrossRef]
30. ANSYS, Inc. ANSYS Help System; ANSYS, Inc.: Canonsburg, PA, USA, 2019.
31. The International Standards Organization for Standardization. Rotodynamic Pumps—Hydraulic Performance
Acceptance Tests—Grades 1, 2 and 3, 2nd ed.; ISO 9906-2012; The International Standards Organization for
Standardization: Geneva, Switzerland, 2015.
32. Standardization Administration of China. Methods of Measuring and Evaluating Vibration of Pumps;
GB/T 29531-2013; Standardization Administration of China: Beijing, China, 2013.
33. Standardization Administration of China. Mechanical Vibration3/4Evaluation of Machine Vibration by
Measurements on Non-Rotating Parts3/4Part 1: General Guidelines; GB/T 6075.1-2012; Standardization
Administration of China: Beijing, China, 2012.
34. Luo, Y.Y.; Wang, Z.W.; Liang, Q.W. Stress of Francis turbine runners under fluctuant work conditions.
J. Tsinghua Univ. (Sci. Technol.) 2005, 45, 235–237.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).