Mechanism and Machine Theory 146 (2020) 103731

Contents lists available at ScienceDirect

Mechanism and Machine Theory

journal homepage: www.elsevier.com/locate/mechmachtheory

Research paper

Synthesis and analysis method for powertrain configuration of

single motor hybrid electric vehicle
Hang Peng a, Datong Qin a,b,∗, Jianjun Hu a,b, Chunyun Fu a,b
a
School of Automotive Engineering, Chongqing University, No. 174, Shazheng Road, Shapingba District, Chongqing 400044, China
b
State Key Laboratory of Mechanical Transmissions, Chongqing University, No. 174, Shazheng Road, Shapingba District, Chongqing
400044, China

a r t i c l e i n f o a b s t r a c t

Article history: Existing hybrid powertrain configuration analysis methods mainly focus on power-split hy-
Received 26 September 2019 brid electric vehicles with multiple planetary gears. However, these methods cannot be ap-
Revised 16 November 2019
plied directly to single motor hybrid electric vehicles with automatic transmission. Hence,
Accepted 29 November 2019
an analysis method to optimize the powertrain configuration of single motor hybrid vehi-
Available online 20 December 2019
cles is proposed to overcome this limitation. In the proposed method, two different adja-
Keywords: cency matrices are established to generate all powertrain configurations, and a correspond-
Hybrid electric vehicle ing matrix analysis method is proposed to analyze their working modes. Subsequently, a
Powertrain configuration set of working modes are established to improve the efficiency of obtaining mathemat-
Matrix analysis method ical models of powertrain configurations. Finally, an energy management strategy based
Performance analysis on dynamic programming is developed to analyze the performance of the generated con-
figurations. The proposed method is utilized to analyze the single motor hybrid vehicle
with a planetary gear and a continuously variable transmission, and a new configuration
is obtained after the analysis for validation. It is observed that the fuel efficiency increases
by 12.2% and the 0–100 km/h acceleration time is reduced by 8.7%, in comparison to the
existing powertrain configuration.
© 2019 Elsevier Ltd. All rights reserved.

1. Introduction

Hybrid electric vehicles (HEVs) can help in reducing emission of greenhouse gases as well as in mitigating global warm-
ing [1]. Several factors such as management strategies [2,3], system component size [4,5], driving condition prediction [6],
and powertrain configuration [7] can help in enhancing the fuel economy, drivability, and emission performance of HEVs.
Amongst the above mentioned factors, powertrain configuration physically determines the energy coupling mechanism and
the energy flow of a HEV. The energy flow is not affected by the component parameters and the energy management
strategies. Thus, the powertrain configuration is a fundamental factor that affects the performance of a vehicle and requires
a detailed study.
To identify powertrain configuration with good fuel economy and acceleration, various methods have been proposed in
the existing literature. The lever analogy method is the most widely used method, which describes the speed and torque

∗
Corresponding author at: School of Automotive Engineering, Chongqing University, No. 174, Shazheng Road, Shapingba District, Chongqing 40 0 044,
China.
E-mail address: [email protected] (D. Qin).

https://doi.org/10.1016/j.mechmachtheory.2019.103731
0094-114X/© 2019 Elsevier Ltd. All rights reserved.
2 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731

Nomenclature

HEV hybrid electric vehicle

CVT mechanical continuously variable transmission
FG stepped automatic transmission with multiple fixed speed ratios
PG planetary gear
DP dynamic programming
WLTC worldwide light-duty test cycle
SOC battery state of charge
C1, C2, C3 clutch
B1 brake

Subscripts
e engine
m motor
d Wheels
CVTp, CVTs primary pulley node and secondary pulley node of CVT
FGs, FGb driving gear node and driven gear node of FG
PGs, PGr, PGc sun gear node, ring gear node and carrier node of PG
power power source component
coupling power coupling component

Variables
A0 , α 0 , H0 matrix of the fundamental matrix equation
Ix , ω˙ x , Tx rotational inertia, angular acceleration and torque of node
TCVTp input torque of CVT
FFG , FPG internal force of FG and PG
iCVT speed ratio of CVT
ηCVT transmission efficiency of CVT
ZFGs , ZFGb number of teeth of driving gear and driven gear
ZPGs , ZPGr number of teeth of sun gear and ring gear
J0 _x matrix related to inertia of nodes
DT 0 _x , DT 1 _x matrix related to torque relationship between nodes
Dn 0 _x , Dn 1 _x matrix related to speed relationship between nodes
FT 0 _x , FT 1 _x matrix related to transmitted power of components
α n 0 _x matrix related to angular acceleration of nodes
Hn 0, HT 0 vector related to speed and the torque relationships between nodes
nclutch_MaxNum number of possible locations to arrange clutches/brakes
nCVT , nFG , nPG number of CVT, FG and PG
C, E adjacency matrix
A1 , α 1 , H1 matrix of the system dynamic equation
ωn 1 _x matrix related to angular velocity of nodes
M, N, O, Q matrix for obtaining the speed equation of the working mode
rank(G) rank of matrix G
DT 2 , DT 3 , TT 3 matrix for obtaining the torque equation of the working mode
KPGs , KPGr coefficient used to classify the powertrain configuration
Xk eco , Xk dri state variable vector at k step
Uk eco , Uk dri control variable vector at k step
Jk eco cost of fuel economy analysis at k step
Pfuel fuel consumption in form of power (kW)
Pele electricity consumption in form of power (kW)
Preq demand power (kW)
Pext additional power loss (kW)
ωe engine speed (rpm)
Te engine torque (Nm)
࢞ωe , ࢞iCVT , ࢞Mode change of engine speed, CVT speed ratio, and working mode
SOCk battery state of charge at k time step
Pe output power of engine (kW)
be fuel consumption rate (g/(kW·h))
 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731 3

Qfuel combustion value of gasoline (kJ/kg)

Pm output power of motor (kw)
ηe working efficiency of engine
ηm working efficiency of motor
ηtran working efficiency of power coupling system
ηbat working efficiency of battery
Qbat capacity of the battery (Ah)
Rbat internal resistance of battery ()
Jk dri cost of drivability analysis at k step
Tm motor torque (Nm)
Fd (Xk dri , Uk dri ) vehicle traction force (N)
Ff vehicle rolling resistance (N)
Fw (vk ) vehicle air resistance at k step (N)
vk vehicle velocity at k step (km/h)

relationship in a planetary gear (PG) set based on the lever principle [8]. This method can be easily applied to analyze the
power transfer characteristics of input-split and output-spilt hybrid powertrain systems using a single PG [9,10].
In case of powertrain configuration with multiple PGs, such as the compound-split HEV [11] and the dual-mode power-
split HEV [12], a compound lever model is required to analyze the powertrain performance using two lever variables. How-
ever, in the compound lever model, the topological information of the powertrain is missing. Hence, a block matrix is devel-
oped in [13] to study the relationship between the power source and the power coupling mechanism. Based on this matrix
representation method, the impact of clutch arrangement on power coupling form of the powertrain configuration is ana-
lyzed, and a transformation matrix is proposed to generate the powertrain configurations of power-split HEVs with double
PGs [14] and multiple PGs [15,16].
The graph theory [17] is also used to analyze complex powertrain configurations. In this method, points and edges are
used to represent the links and motion pairs, respectively, and the kinematics, statics, and power flow characteristics of
PGs are analyzed using the fundamental circuit method. Using the graph theory, the topological characteristics and the
constraints in the clutch arrangement of hybrid powertrain configurations with double PGs were studied [18]. Even though
this method can be used to express the topological relationship between the system components efficiently [19], it is difficult
to analyze the powertrain configurations with an automatic transmission.
The configuration analysis methods mentioned above mainly focus on power split HEVs with PGs. For example, the
current matrix representation method determines the distribution law of the characteristic matrix elements of feasible con-
figuration on the basis of the characteristics of the PG mechanism [14,15]. These methods cannot be applied directly to
single motor parallel HEVs with automatic transmission due to the difference between the working characteristics of PG
mechanisms and automatic transmissions.
In recent times, single motor parallel HEVs with automatic transmission have gained significant attention from automak-
ers and researchers [2], owing to their advantages such as ease of control, compact structure and good production inheri-
tance [20]. In [21], the performance of six parallel hybrid systems with different power coupling forms and different trans-
mission types is compared, and the impact of the configuration and the component size on fuel economy is investigated. A
single motor HEV with PG and dual-input gear transmission was investigated as mentioned in [22], and a pre-coupled HEV
with a PG and a mechanical continuously variable transmission (CVT) was discussed as seen in [23]. The results show that
these powertrain configurations combine the advantages of the PG and automatic transmission and provide more working
modes and enhanced vehicle performance. However, certain questions have not been addressed properly such as the num-
ber of single motor hybrid powertrain configurations that can be obtained using a PG and an automatic transmission, and
the number of working modes present in these systems.
To address these questions, this paper proposes a synthesis and analysis method for studying the powertrain configu-
ration of single motor HEVs. This study involves two major steps. In the first step, two new types of adjacency matrices
are established to generate all the powertrain configurations of single motor HEVs with an automatic transmission and/or
PGs, and a corresponding matrix analysis method is proposed to analyze the system’s working modes. Hence, there is a
huge scope for identifying powertrain configuration on basis of matrix representation method. In the second step, a set of
working modes are proposed to describe the relationship between the adjacency matrices, system speed and torque equa-
tions, and the power coupling form. This set of working modes is helpful in the analysis of hybrid powertrain configuration,
including working mode feasibility, powertrain mathematical model and vehicle performance.
The remainder of this paper is organized as follows. Section 2 states the rules for establishing the fundamental ma-
trix equation and adjacency matrices and proposes a corresponding matrix analysis method. The classification method of
configurations and the establishment of a set of working modes for the single motor HEV with PG and CVT are presented
in Section 3. Section 4 states the energy management strategy based on dynamic programming to analyze the configura-
tion performance and the working characteristics of the newly obtained configuration. The conclusions of this study are
presented in Section 5.
4 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731

Fig. 1. Schematic diagram of a parallel hybrid vehicle.

Fig. 2. Fundamental matrix equation.

2. Improved matrix analysis methods for powertrain configurations

2.1. Establishment of the fundamental matrix equation

To meet the power demand under different driving conditions and improve energy efficiency, a hybrid powertrain system,
which is capable of distributing the engine power and motor power through a power coupling mechanism, can be utilized.
The schematic diagram of a parallel HEV is shown in Fig. 1.
Based on the dynamic characteristics of the different components, a single PG is equivalent to three nodes, namely, the
sun gear node PGs, the ring gear node PGr, and the carrier node PGc. A stepped automatic transmission with multiple fixed
speed ratios is equivalent to two nodes, namely, the driving gear node FGs and the driven gear node FGb. A CVT is equivalent
to two nodes, namely, the primary pulley node CVTp and the secondary pulley node CVTs. The engine, motor, and wheels
are equivalent to a single node and represented as e, m, and d, respectively. By combining the dynamic equations of various
components, the fundamental matrix equation is established, as shown in Fig. 2.
In Fig. 2, Ix , ω˙ x , and Tx represent the rotational inertia, angular acceleration, and torque of different nodes, respectively.
TCVTp , iCVT , and ηCVT represent the input torque, speed ratio, and transmission efficiency of CVT, respectively. FFG and FPG
represent the internal force of the FG and PG, respectively. ZFGs and ZFGb represent the number of teeth of the driving gear
and the driven gear of FG, respectively. ZPGs and ZPGr represent the number of teeth of the sun gear and the ring gear of PG,
respectively. The value of ZPGr /ZPGs is the characteristic parameter of the PG.
 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731 5

In this study, the system components are divided into power source components and power coupling components. Hence,
the fundamental matrix equation can be simplified and represented in the following form:
⎧
⎪A0 α0 = H0
⎪
⎪  
⎪
⎪ J0_power J0_coupling DT 0_power DT 0_coupling
⎪
⎨A0 =
Dn0_power Dn0_coupling 0 0 (1)
⎪
⎪  T
⎪
⎪α0 = αn0_power αn0_coupling
⎪
⎪
FT 0_power FT 0_ coupling
⎩  T
H0 = HT 0 Hn0

where the subscripts power and coupling represent the nodes of power source components and power coupling components,
respectively. The square matrix A0 includes the matrix J0 _x describing the inertia of nodes, the matrix DT 0 _x describing the
torque relationship between nodes, and the matrix Dn 0 _x describing the speed relationship between nodes. The vector α 0
comprises of the vector α 0 _x which is related to the node angular accelerations, and the vector F0 _x which is related to the
power transmitted by the system components. The vector H0 comprises of the vector Hn 0 which is related to the speed
relationships, and the vector HT 0 which is related to the torque relationships. It should be noted that the dimensions of
each matrix in Eq. (1) are determined by different types and numbers of system components. On the basis of Eq. (1), the
following and equations can be used to describe the speed and torque relationships of the system components, respectively.
  T
Dn0_ power Dn0_coupling αn0_ power αn0_coupling = Hn0 (2)

  T   T
J0_ power J0_coupling αn0_ power αn0_coupling + DT 0_ power DT 0_coupling FT 0_ power FT 0_coupling = HT 0 (3)

To determine the possible locations of the clutches/brakes for a powertrain configuration, the following assumptions
are made [14]: (1) the clutch is able to connect the nodes of any two power coupling components, (2) only one clutch is
required to connect any two nodes of the PG, and (3) the brake can ground any node except the output node. On the basis
of the above assumptions, the maximum number of locations can be obtained using the following equation:

nclutch_MaxNum = C (2nCV T , 1 )C (2nF G , 1 )C (3nPG , 1 ) + nPG + (2nCV T + 2nF G + 3nPG − 1 ) (4)

where, C(nx ,k) determines the number of k-combinations for a given set of nx elements, and nCVT , nFG , and nPG represents
the number of CVT, FG, and PG, respectively.

2.2. Establishment of two adjacency matrices

To build a relationship between the powertrain configuration and the matrix variables, two types of adjacency matrices
C and E have been discussed in this paper. Both the adjacency matrices are represented as unit diagonal matrices in their
initial stage. By changing the values of the elements in these two matrices, matrix C can be used to describe the connection
between different nodes as well as to reflect the working state of the clutch. Matrix E can be used to describe the working
state of the nodes and also to reflect the working state of the brake.
On the basis of the location and working state of the clutches, the connection state between the nodes can be deter-
mined, and the values of elements Ci , j and Cj , i can be calculated using the following equation:

1, if node i and node j are connected.

Ci, j = C j,i = (5)
0, if node i and node j are not connected.

On the basis of the working states of the brakes and components and the state of connection between the nodes, the
value of the diagonal element Ek , k can be calculated using the following equation:
⎧
⎨0, if node l and node k are connected (l < k ).
⎪
0, if node k is braked.
Ek,k = (6)
⎩0, if component is not working.
⎪
1, the remaining diagonal elements.
Connecting of node 8 and 9 and braking of node 8 are employed as an example to illustrate the change in the value of
elements in the two adjacency matrices. For matrix C, the values of the corresponding elements C8,9 and C9,8 change from 0
to 1 owing to the connection between node 8 and 9, as shown in the red parallelogram in Fig. 3(a). At this point, when any
two nodes of the PG are connected, the three nodes rotate at the same speed. Hence, the values of the elements C8,10 , C9,10 ,
C10,8 and C10,9 are to be changed from 0 to 1, as shown in the blue rectangle in Fig. 3(a). For matrix E, the values of the
diagonal elements E9,9 and E10,10 need to be changed from 1 to 0 on the basis of the connection between nodes, as shown
in the red parallelogram in Fig. 3(b). As node 8 is connected to the gearbox, the working states of that node and the other
connected nodes undergo a change, and the values of the corresponding diagonal elements (E8,8 , E9,9 and E10,10 ) change to
6 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731

Fig. 3. Rules for establishing adjacency matrices (a) matrix C, (b) matrix E. (For interpretation of the references to color in this figure, the reader is referred
to the web version of this article.)

0, as shown in the blue rectangle in Fig. 3(b). Furthermore, the PG does not transmit power owing to the working state of
the actuator (clutch and brake), hence, the value of element E16,16 becomes 0, as shown in the green ellipse in Fig. 3(b).
Matrices C and E not only describe the working state of a hybrid vehicle in different working modes, but also explain
the connection between the power source components (engine, motor, and wheels), power coupling components (plane-
tary gear and automatic transmission), and clutches/brakes. Thus, the relationship between the matrix and the powertrain
configuration is mapped, which helps in searching various hybrid powertrain configurations using numerical variables.

2.3. Matrix analysis method of system working mode

When a clutch/brake is in different working condition, different working modes of a powertrain can be obtained. In this
study, the fundamental matrix equation and the two adjacency matrices are employed in Eq. (7) to analyze the dynamic
equations of different working modes of a system.
⎧
⎨A1 α1 = H1 T T
⎪
A1 = E (C A0C )E
(7)
⎩α1 = E α0
⎪
H1 = E (C H0 )
Since matrix E causes all elements to be zero in some rows of matrix α 1 , Eq. (7) can only help in describing the speed
and torque relationship between the power source nodes and the power coupling nodes connected to them. In order to
obtain the speed and torque relationships between the power source components from Eq. (7), a matrix analysis method is
proposed which is based on the difference in the working characteristics of different power coupling components.
By expressing Eq. (7) in the form of Eq. (1), the relationship between the speeds of the connecting nodes can be obtained:
  T
Dn1_ power Dn1_coupling ωn1_ power ωn1_coupling =0 (8)

where the matrices Dn 1 _x and ωn 1 _x are obtained from matrices A1 and α 1 in Eq. (7), and matrix ωn 1 _x denotes the angular
velocities of the nodes.
As the proposed adjacency matrix C can explain the connection between the power source nodes and the power coupling
nodes, Eq. (8) can be extended in combination with matrix C and the speed characteristics of different power coupling
components, as described in the following equation:.
⎧  T
⎪
⎨ 1 −iCV T ωCV T p
⎪ ωCV Ts = 0
  T (9)
⎪ ZF Gs ZF Gb ωF Gs ωF Gb = 0
⎪
⎩  T
−ZPGs −ZPGr ZPGs + ZPGr ωPGs ωPGr ωPGc =0

Therefore, by using the matrix elementary transformation method, the relationship between the speeds of all nodes can
be derived, which is represented by the following equation:.
 T  T
M p×2 ωe ωm + N p×1 ωd = −O p×7 ωCV T p ωCV Ts ωF Gs ωF Gb ωPGs ωPGr ωPGc (10)
where, coefficient matrices M, N and O are obtained from the extended equation of Eq. (8), and the subscript p × 2, p × 1,
and p × 7 denote the dimensions of these matrices.
 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731 7

Then, all equations satisfying Eq. (11) are searched in the Eq. (10) to determine the relationship between the speeds of
power source components.
⎧  T 
⎪
⎪ M ω e ω m = 0
⎨ p×2
 (i,1 )
N ω = 0 , for i ∈ {1, 2, ..., p} (11)
⎪
⎪ p×1  d (i,1) T 
⎩ O p×7 ωCV T p ωCV Ts ωF Gs ωF Gb ωPGs ωPGr ωPGc =0
(i,1 )
where (X)( i ,1) refers to the entry in the i-th row and first column of the matrix X.
Thus, the speed equation of the working mode of a system can be expressed using Eq. (12). In case of a feasible working
mode, a linear relationship exists between the speeds of engine, motor, and wheel. In other words, the working mode
becomes infeasible when the number q in the following equation becomes zero.
 T
ωe ωm ωd = 0, if the working mode is infeasible (q = 0 )
 T (12)
Qq×3 ωe ωm ωd = 0, if the working mode is feasible (q > 0 )
where, coefficient matrix Q is obtained from the equations satisfying Eq. (11), and q is the number of the equations satisfied.
For a feasible system working mode, the relationship between the torques of the connecting nodes can also be obtained
using Eq. (7), which is shown as follows:
DT 1_coupling FT 1_coupling = −DT 1_ power FT 1_ power (13)
where, matrices DT 1 _x and FT 1 _x are obtained from matrices A1 and α 1 in Eq. (7).
While solving the torque equation of the working mode of a system, the above method cannot be used, as it is the
power transfer path and the transmitted power of the hybrid powertrain that determine the torque at each node. Since
Eq. (13) mainly describes the torque relationship between the power source nodes and the power coupling nodes connected
to them, the internal force/torque of the coupling mechanism FT 1 _coupling is treated as an unknown variable, which is repre-
sented by the following equation:
 T
FT 1_coupling = G3×3 TCV T p FF G FPG (14)
where G is a 3 × 3 diagonal matrix, and its rank is represented by rank(G).
All the homogeneous linear equations that satisfy Eq. (15) can be extracted from Eq. (13).
 
DT 1_coupling FT 1_coupling = 0
 (i,1) for i ∈ {1, 2, . . . , 10} (15)
DT 1_ power FT 1_ power =0
(i,1 )

These homogeneous linear equations are represented by the following equation:

DT 2(h×3) FT 1_coupling = 0 (16)
where matrix DT 2 is obtained from matrix DT 1_ coupling , and h is the number of homogeneous linear equations.
When h becomes equivalent to rank(G), the unknown FT 1 _coupling has only a zero solution, and the output torque of each
power source component at this time is 0. When h<rank(G), an infinite number of solutions are found for this unknown
variable. However, for a feasible working mode, FT 1 _coupling must have a unique solution. Therefore, all non-homogeneous
linear equations which do not satisfy Eq. (15) are found within Eq. (13), and a set of non-homogeneous linear equations
which satisfy Eq. (17) is established.
⎧
⎪
⎨DT 3 FT 1_coupling = TT 3
   
DT 2 DT 2 0 (17)
⎪
⎩rank = rank = rank(G )
DT 3 DT 3 TT 3

where matrix DT 3 is obtained from matrix DT 1_ coupling , and matrix TT 3 is obtained from -DT 1_ power FT 1_ power .
The torque equation of working mode of a system can be solved as follows:
⎧ T
⎪
⎨ Te Tm −Td = 0, if the working mode is infeasible (h = rank(G ))
 −1  
DT 2 0 (18)
⎪
⎩DT 1_ power FT 1_ power = −DT 1_coupling , if the working mode is feasible (h < rank(G ))
DT 3 TT 3

On the basis of the matrix transformation theory used in linear algebra, two adjacent matrices and the working charac-
teristics of all power coupling mechanisms have been incorporated in the proposed method. This helps in making changes
to the fundamental matrix equation. This is followed by analysis of the feasibility of each working mode of the system, and
its speed and torque equations are obtained. The method proposed in this paper is applicable to various single motor hybrid
powertrain configurations owing to consideration of the difference in the working characteristics of the PG and automatic
transmission. Moreover, the proposed method also helps in developing a systematic mathematical computation procedure
for computer-aided analysis of the working mode of a system.
8 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731

Fig. 4. Schematic diagram of the original hybrid powertrain configuration.

Fig. 5. (a) Combination of the values of lever coefficients, (b) configuration group s1, (c) configuration group s7.

3. Analysis method for studying the hybrid powertrain configuration

3.1. Classification method of powertrain configuration

The feasibility of the proposed method can be verified by analysis of the configuration of a single motor parallel HEV,
comprising of an engine, a motor, a PG, and a CVT. The schematic diagram of the original powertrain configuration is shown
in Fig. 4.
On the basis of the number of system components and the power transmission characteristics of the PG, 12 types of
connection between the power source components and power coupling mechanisms can be found. In the lever analogy
method, the lever coefficient is used to denote the connection between the power source component and the PG [9, 10].
Therefore, all configurations can be classified into 12 groups on the basis of different combinations of the values of lever
coefficients, as shown in Fig. 5(a).
In the configuration group s1-s6, wheels are connected to the output of CVT, and different connections exist between
the motor, engine, and PG nodes. The relationship between the speed of the motor and engine is represented by the first
equation in the following equation:
  T
−KPGs −KPGr KPGs + KPGr ωe ωm ωo = 0, for group s1 − s6
  T (19)
−KPGs −KPGr KPGs + KPGr ωo ωm ωd = 0, for group s7 − s9
 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731 9

Fig. 6. All 11 clutch locations.

According to the group number, the values of the coefficients KPGs and KPGr can be obtained from Fig. 5(a). For example,
in the configuration group s1, the values of KPGs and KPGr are equal to ZPGs and ZPGr , respectively. Hence, on the basis of
the speed characteristics of the PG, the engine is connected to the sun gear, and the motor is connected to the ring gear. A
schematic diagram of the configuration group s1 is shown in Fig. 5(b).
In the configuration group s7-s12, the engine is connected to the input of CVT, whereas different connections exist be-
tween the motor, wheels, and PG nodes. The relationship between the speed of the motor and wheels is represented by
the second equation in Eq. (19). According to Fig. 5(a), in the configuration group s7, the values of KPGs and KPGr are equal
to ZPGs and ZPGr , respectively. Hence, the motor is connected to the ring gear, and the wheel is connected to the carrier. A
schematic diagram of the configuration group s7 is shown in Fig. 5(c).
According to Eq. (4), the studied hybrid powertrain system theoretically has up to 11 locations where the clutches can be
arranged, as shown in Fig. 6. For each configuration group, different powertrain configurations can be obtained by changing
the number and location of clutches. In view of the space limitation and cost of the mechanism, the number of clutches for
a powertrain configuration is limited to four, and at most two clutches are present between PG and CVT. Besides, at least
one clutch is required as the powertrain system has two power source components and three degrees of freedom at initial
stage.

3.2. Establishment of working modes set

On the basis of the proposed matrix analysis method, the powertrain configuration of a HEV can be described using the
adjacency matrices. Apart from that, the speed and torque equations of various working modes can also be obtained using
the matrix analysis method. Hence, the adjacency matrices and the corresponding speed and torque equations are treated
as an element to establish different sets of working modes for different configuration groups.
The analysis results show that 12 identical working modes exist for two groups of configurations with the same values of
leverage coefficients. Thus, based on the configuration classification method, a set of working modes is established, as shown
in Fig. 7(a). This set comprises of 30 elements and describes the corresponding relationship between the adjacency matrices,
the speed and torque equations of a working mode, and the power source coupling form. All feasible working modes can
be divided into groups namely the engine-only mode (ICE), motor-only mode (EV), speed coupling hybrid mode, and torque
coupling hybrid mode. The classification of the working modes is shown in Fig. 7(b). Since there are different adjacency
matrix pairs describing the same power coupling form, the total number of adjacency matrix pairs corresponding to each
working mode is also shown in Fig. 7(b). In the torque coupling hybrid mode group, few power reflux modes exist, such as
mode 21 and mode 22. The characteristic of the power reflux mode is that part of the power transmitted by PG circulates
to the output shaft of the power source through CVT. For the power reflux mode, the range of the variable speed ratio of
the system is improved, and transmission efficiency of the system at low speed but high torque is also improved. However,
the solution of the dynamic equations of the power reflux mode is complex [24]. The matrix analysis method proposed in
this paper can solve these problems by computer and avoid mistakes, which is another advantage of this method.
As the working modes of an arbitrary configuration belong to a subset of the above mentioned set of working modes,
the speed and torque equations for different working modes and the mathematical models of different configurations can
be obtained using the lever coefficient, adjacency matrices, and the working mode set. This results in reduction of the
computation time. In order to verify this point, the speed and torque equations of the powertrain configurations with two,
three and four clutches are analyzed with the proposed matrix analysis method and the results are shown in Table 1. The
10 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731

Fig. 7. (a) A set of working modes, (b) number of adjacency matrix pairs describing the same working mode.
 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731 11

Table 1
The computation time to obtain the speed and torque equations of powertrain configurations.

Number of Computation time (s)

configurations
Without using the Establishing the By using the working
working mode set working mode set mode set

Configuration with two 660 4206 3128 49

clutches
Configuration with 1980 25,393 388
three clutches
Configuration with 3960 89,481 2012
four clutches
In total 6600 119,080 5577

simulation software which has been used for computation is MATLAB R2014a, and the CPU used is i7-670 [email protected] 0 GHz with
8GB memory.
As seen in Table 1, the speed and torque equations of 660 hybrid powertrain configurations with two clutches can be
obtained in 4206 s, 1980 configurations with three clutches can be obtained in 25393s, and 3960 configurations with four
clutches can be obtained in 89,481 s. However, using the working mode set and a combination of the values of leverage
coefficients, it only took 49 s, 388 s and 2012s, respectively, for the above-mentioned combinations. In addition, only 3128 s
were required to obtain the working mode set, as seen in Fig. 7(a). Hence, the total computation time is reduced by 95.4%.

4. Performance analyses

4.1. Fuel economy and drivability analysis

As the performance of a vehicle can be affected by control strategy, a global optimal control strategy, based on dynamic
programming (DP), is used to obtain optimum performance of each powertrain configuration [25]. Thus, it is possible to
analyze the impact of configuration on the performance of a vehicle.
The fuel economy of the hybrid powertrain configuration is evaluated using the worldwide light-duty test cycle (WLTC)
with fuel consumption for every 100 kms. This driving cycle comprises of the low-, medium-, high- and extra-high speed
phases [26], and it can better represent the real driving conditions in comparison to the previous driving cycles, such as
the New European Driving Cycle. The engine speed ωe , engine torque Te , CVT speed ratio iCVT , and working mode Mode
are chosen as the control variables Ueco . The battery state of charge (SOC) is selected as the state variable Xeco . The state
constraint is that the terminal and initial SOC values are equivalent. The cost function in this problem consists of three
parts: (1) the total energy loss of hybrid powertrain configuration, including the power losses of engine, motor, battery and
power coupling system; (2) the penalty function for sustaining SOC; (3) the penalty function for avoiding frequent changes
in engine speed, CVT speed ratio and working mode. The cost at k step is attained using Eq. (20):
⎧ 
⎪
⎪Jkeco (Xkeco, Ukeco ) = Pf uel (k ) + Pele (k ) − Preq (k ) + β1 |Pext (k )| + (β2 ωe + β3 iCV T + β4 Mode )
⎪
⎪
⎨Pf uel (k ) = Pe (X eco, U eco )be (X eco, U eco )Q f uel /(3600 × 1000 )
k k k k
(20)
⎪
⎪
−sign(Pm ) −sign(Pm )
Pele (k ) = Pm (Xkeco, Ukeco )ηm ηbat
⎪
⎪
⎩
Pext (k ) = [(SOCk − SOCini )Qbat ] Rbat (SOCk )
2

where Pfuel is the fuel consumption in the form of power, Pele is the electricity consumption in the form of power, Preq repre-
sents the power demand, Pext is the amount of power required to restore the current SOCk to the initial SOCini (reflecting the
offset of SOC),. ࢞ωe , ࢞iCVT , and ࢞Mode denote the changes in the engine speed, CVT speed ratio, and mode shift, respec-
tively, β x is the weight coefficient of the corresponding index, Pe denotes the engine output power, be stands for the fuel
consumption rate, Qfuel is the gasoline combustion value (44,0 0 0 kJ/kg), Pm represents the motor output power; ηm denotes
the motor working efficiency, and ηbat , Qbat and Rbat represent the working efficiency, capacity and internal resistance of the
battery, respectively.
The drivability of a hybrid powertrain configuration is evaluated in terms of the time taken to accelerate from rest to
100 km/h on a flat road. The acceleration process is divided into several stages, and the maximum output torque of the
system is calculated at different speeds. Then, the minimum acceleration time of each acceleration stage is accumulated to
obtain the overall acceleration performance. The engine speed ωe is selected as the state variable Xdri , and the engine torque
Te , motor torque Tm , CVT speed ratio iCVT , and working mode Mode are chosen as the control variables Udri . The cost at k
step is attained using Eq. (21):

δm
Jkdri (Xkdri , Ukdri ) = v (21)
Fd (Xkdri , Ukdri ) − Ff − Fw (vk )
12 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731

Fig. 8. Main steps in proposed analysis method.

where Fd , Ff and Fw represents the traction force, rolling resistance and air resistance of the vehicle respectively, vk is the
vehicle velocity at k step, and ࢞v is the velocity variation.
When analyzing the powertrain configuration performance, the following constraints should be satisfied to ensure the
safety of the system components:

⎧SOC ≤ SOC ≤ SOC

⎪ min
⎪ωe min ≤ ωe ≤ ωe max
max
⎪
⎪
⎪
⎪
⎪Te min ≤ Te ≤ Te max
⎨
ωm ≤ |ωm max |
(22)
⎪Tm ≤ |Tm max |
⎪
⎪
⎪
⎪0.442 ≤ iCV T ≤ 2.432
⎪T
⎪
⎩ CV T p ≤ 190
Mode ∈ Modeavail abl e

Table 2 presents the parameters of the vehicle and the main components. The vehicle model is designed on the basis of
the vehicle longitudinal dynamic equation. The fuel consumption of the gasoline engine and the efficiencies of the motor
and the CVT can be tested through a bench test. Based on this test, the quasi-static models of these components can be
established. The ideal torque model is considered for clutch/brake instead of the mathematical model [27] which is found
to be more accurate. The charge/discharge current of the lithium battery is calculated using the battery internal resistance
model, and the battery SOC is estimated using the ampere hour counting approach [28]. On the basis of these models, the
accuracy of the simulation results is validated, and the computational load can be reduced. Fig. 8 shows the main steps
involved in the proposed design and the analysis method.
 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731 13

Fig. 9. Number of hybrid powertrain configurations satisfying constraints and performance requirements.

Table 2
Vehicle parameters.

Parameter Value Unit

Vehicle mass 1590 kg

Aerodynamic drag coefficient 0.321
Frontal area 2.261 m2
Rolling resistance coefficient 0.012
Wheel radius 0.307 m
Final drive ratio 5.24
Engine maximum torque 137/3500 Nm/rpm
Engine maximum power 67/5250 kW/rpm
Motor rated power 25 kW
Motor rated/maximum speed 2700/6000 rpm
Nominal capacity of the battery 30 Ah
Characteristic parameters of the PG 2.6

Table 3
Non-dominated solutions.

Serial number System working modes Fuel economy (L/100 km)/Improvement Drivability (s)/Improvement

Original configuration 1,5,6,9,12 5.62 10.3

C4s5n299 6,8,9,11,12,26 4.97/11.5% 9.8/4.8%
C4s7n269 37,8,13,14,24,26,30 5.19/7.6% 9.3/9.7%
C4s11n104 1,7,8,11,14,17,25,26 4.93/12.2% 9.4/8.7%
C4s12n79 7,8,11,14,22,24,26 5.21/7.3% 8.3/19.4%

4.2. Simulation results

The simulation results show that the original hybrid powertrain configuration, as seen in Fig. 4, has a fuel economy of
5.62 L/100 km under the WLTC cycle, and a 0–100 km/h acceleration time of 10.3 s. On the basis of the configuration analysis
method, a total of 6600 configurations (12 groups) were generated, out of which 4680 configurations were able to fulfill the
components constraint. The performance of all powertrain configurations were analyzed and 310 configurations were found
to have an acceleration time of less than 10.3 s, out of which 274 configurations consumed less than 5.62 L/100 km. Fig. 9
shows the number of powertrain configurations with two, three and four clutches/brakes, which satisfy different constraints
and performance requirements. It can be seen that as the number of clutches/brakes increases, more configurations are
generated, and they are found to have better fuel economy and drivability performance.
Fig. 10(a)–(c) shows the 0–100 km/h acceleration time and fuel consumption per hundred kilometers for different pow-
ertrain configurations with two, three, and four clutches/brakes. The C4s11n104 configuration indicates that this hybrid
powertrain is equipped with four clutches and belongs to the 104th configuration of the C(11,4)∗ 12 = 3960 configurations of
the group s11. The group number is determined by Fig. 5 and the connection between power source components and power
14 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731

Fig. 10. Performances of hybrid powertrain configurations with (a) two clutches, (b) three clutches and (c) four clutches.

Fig. 11. Schematic diagram of C4s11n104 configuration.

 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731 15

Fig. 12. (a)WLTC driving cycle, (b) engine and motor power outputs of the original configuration, (c) engine and motor power outputs of the C4s11n104
configuration.

coupling components. The serial number of the configuration is determined by the combination of the clutch locations in
Fig. 6.
By comparing the performances of the configurations, non-dominated solutions can be obtained. Table 3 lists the working
modes, fuel economy, drivability, and the corresponding improvement of the non-dominated solutions.
The analysis results show that the fuel economy and drivability of C4s11n104 configuration is improved by 12.2% and
8.7%, respectively, in comparison to the original powertrain configuration. The schematic diagram of the new powertrain
configuration is shown in Fig. 11.
To understand the underlying reasons for enhancement of fuel economy, the working conditions of two different con-
figurations were investigated. The speed of the WLTC cycle is shown in Fig. 12(a). The engine and motor power outputs of
the original configuration under WLTC cycle are shown in Fig. 12(b) and those of the C4s11n104 configuration are shown in
Fig. 12(c).
In comparison to the original configuration, C4s11n104 configuration reduces the output power of the engine under
driving conditions and increases the regenerative braking power under braking conditions. This phenomenon arises due
to the difference in the power flow paths of the two powertrain systems. The frequency of usage of the working mode
16 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731

Fig. 13. Working mode usage frequencies of C4s11n104 configuration under (a) driving condition, (b) braking condition, and (c) 0–100 km/h accelerating
condition.

Fig. 14. Comparison between original configuration and C4s11n104 configuration in terms of efficiency distribution of (a) engine, (b) motor and (c) power
coupling system, and (d) power loss of the hybrid powertrain under WLTC.

of C4s11n104 configuration were analyzed for driving, braking, and accelerating conditions and the results are shown in
Fig. 13. As the CVT works only under the working modes 1, 14, 17 and 25, hence, its involvement in both driving and
braking conditions is only 23%. However, the involvement of CVT in the accelerating condition is approximately 83%.
The power losses of the power source components and the power coupling components were studied to analyze the
influence of powertrain configuration on the working conditions of the components of a system. The efficiency distributions
of engine, motor, and transmission are shown in Fig. 14(a)–(c) respectively, and the power losses of the components under
the WLTC driving cycle is shown in Fig. 14(d).
Fig. 14 shows that the average working efficiency of a motor shows a marked improvement in case of C4s11n104 con-
figuration. It can be seen in Fig. 12(b) and (c) that the motor output power under driving condition and the regenerative
power under braking condition is greater than that of the original configuration. Thus, the loss incurred on account of the
 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731 17

conversion of electrical energy increases for C4s11n104 configuration. It is also observed that the working efficiency of an
engine decreases in case of C4s11n104 configuration while the engine output power under driving condition is lesser than
that of the original configuration, hereby reducing the energy conversion loss by 2.5% approximately. It is clearly observed
in Fig. 14 that the efficiency of the power coupling system has significantly improved and the power loss has reduced con-
siderably to 85.6%.
The above analysis indicates that the powertrain configuration has a significant impact on the power distribution, power
flow transmission path, system component efficiency and system working mode usage of a HEV. The proposed configuration
analysis method can also improve the performance of the HEV with respect to the powertrain configuration.

5. Conclusion

In this study, a new synthesis and analysis method is proposed for studying the powertrain configuration of single motor
hybrid electric vehicle, which overcomes the limitations of the configuration analysis method based on matrix representa-
tion. In case of the single motor HEV with a PG and a CVT, a new configuration was determined using the proposed matrix
analysis method where a marked improvement was observed in fuel economy by 12.2% and a reduction in 0–100 km/h
acceleration time by 8.7%. The four main advantages of the new method are: (1) All the powertrain configurations for single
motor HEVs with an automatic transmission and/or PGs can be generated using the proposed adjacency matrices and fun-
damental matrix equation (2) The computer-aided systematic mathematical computation procedure for obtaining the speed
and torque equations of working modes can be determined using the proposed matrix analysis method. (3) The efficiency
of obtaining mathematical models of different hybrid powertrain configurations can be improved using the devised set of
working modes. (4) The proposed configuration analysis method can improve the performance of the single motor HEV from
the perspective of powertrain configuration.
The main objective of this study is to optimize the powertrain configuration of single motor HEVs with different power
coupling components. Apart from the above, the system component parameters also have a significant impact on the per-
formance of a vehicle. In future research, intelligence algorithms such as genetic algorithm and particle swarm optimization
can be employed to optimize the system component parameters of different HEVs. Thus, a hybrid power system optimiza-
tion method can be devised on the basis of the impact of powertrain configuration and component parameters on the
performance of the HEV.

Declarations of Competing Interest

None.

Acknowledgments

This work was supported by the National Key Research and Development Program of China [grant number
2016YFB0101402], the Fundamental Research Funds for the Central Universities [grant number 106112016CDJXZ338825] and
the Chongqing Key Research and Development Program [project number cstc2018jszx-cyztzx0047].

References

[1] V. Krithika, C. Subramani, A comprehensive review on choice of hybrid vehicles and power converters, control strategies for hybrid electric vehicles,
Int. J. Energy Res. 42 (5) (2017) 1789–1812, doi:10.1002/er.3952.
[2] M. Yao, D. Qin, X. Zhou, S. Zhan, Y. Zeng, Integrated optimal control of transmission ratio and power split ratio for a CVT-based plug-in hybrid electric
vehicle, Mech. Mach. Theory 136 (2019) 52–71, doi:10.1016/j.mechmachtheory.2019.02.014.
[3] X. Hu, N. Murgovski, L. Johannesson, L. Johannesson, B. Egardt, Energy efficiency analysis of a series plug-in hybrid electric bus with different energy
management strategies and battery sizes, Appl. Energy 111 (4) (2013) 1001–1009, doi:10.1016/j.apenergy.2013.06.056.
[4] Y. Zeng, D. Qin, L. Su, M. Yao, Cost, fuel consumption and emission multi-objective parameter optimization for the powertrain of a plug-in hev, Autom.
Eng. 38 (4) (2016) 397–402, doi:10.19562/j.chinasae.qcgc.2016.04.001.
[5] X. Wu, B. Cao, X. Li, J. Xu, X. Ren, Component sizing optimization of plug-in hybrid electric vehicles, Appl. Energy 88 (3) (2011) 799–804, doi:10.1016/
j.apenergy.2010.08.018.
[6] Q. Gong, P. Tulpule, V. Marano, S. Midlam-Mohler, G. Rizzoni, The role of ITS in PHEV performance improvement, in: Proceedings of the 2011 American
Control Conference, IEEE, 2011, pp. 2119–2124, doi:10.1109/ACC.2011.5990968.
[7] A.E. Bayrak, Y. Ren, P.Y. Papalambros, Topology generation for hybrid electric vehicle architecture design, ASME J. Mech. Des. 138 (8) (2016) 1–9 no.
081401, doi:10.1115/1.4033656.
[8] J. Liu, L. Yu, Q. Zeng, Q. Li, Synthesis of multi-row and multi-speed planetary gear mechanism for automatic transmission, Mech. Mach. Theory 128
(2018) 616–627, doi:10.1016/j.mechmachtheory.2018.07.007.
[9] H. Yang, S. Cho, N. Kim, W. Lim, S. Cha, Analysis of planetary gear hybrid powertrain system part 1: input split system, Int. J. Autom. Technol. 8 (6)
(2007) 771–780, doi:10.1007/s12239- 009- 0044- y.
[10] H. Yang, B. Kim, Y. Park, W. Lim, S. Cha, Analysis of planetary gear hybrid powertrain system part 2: output split system, Int. J. Autom. Technol. 10 (3)
(2009) 381–390, doi:10.1007/s12239- 009- 0044- y.
[11] W. Wang, R. Song, M. Guo, S. Liu, Analysis on compound-split configuration of power-split hybrid electric vehicle, Mech. Mach. Theory 78 (2014)
272–288, doi:10.1016/j.mechmachtheory.2014.03.019.
[12] J. Kim, J. Kang, Y. Kim, T. Kim, B. Min, H. Kim, Design of power split transmission: design of dual mode power split transmission, Int. J. Autom. Technol.
11 (4) (2010) 565–571, doi:10.1007/s12239- 010- 0068- 3.
[13] J. Liu, H. Peng, A systematic design approach for two planetary gear split hybrid vehicles, Veh. Syst. Dyn. 48 (11) (2010) 1395–1412, doi:10.1080/
00423114.2010.512634.
18 H. Peng, D. Qin and J. Hu et al. / Mechanism and Machine Theory 146 (2020) 103731

[14] X. Zhang, S.E. Li, H. Peng, J. Sun, Efficient exhaustive search of power-split hybrid powertrains with multiple planetary gears and clutches, ASME J.
Dyn. Syst., Meas. Control 137 (12) (2015) 121006, doi:10.1115/1.4031533.
[15] W. Zhuang, X. Zhang, H. Peng, L. Wang, Rapid configuration design of multiple-planetary-gear power-split hybrid powertrain via mode combination,
IEEE/ASME Trans. Mechatron. 21 (6) (2016) 2924–2934, doi:10.1109/TMECH.2016.2575359.
[16] Z. Qin, Y. Luo, K. Li, H. Peng, Optimal design of single-mode power-split hybrid tracked vehicles, J. Dyn. Syst. Meas. Control 140 (10) (2018) 1–14 No.
101002, doi:10.1115/1.4039687.
[17] W. Yang, H. Ding, Automatic detection of degenerate planetary gear trains with different degree of freedoms, Appl. Math. Model 64 (2018) 320–332,
doi:10.1016/j.apm.2018.07.038.
[18] H. Ngo, H. Yan, Configuration synthesis of parallel hybrid transmissions, Mech. Mach. Theory 97 (2016) 51–71, doi:10.1016/j.mechmachtheory.2015.11.
002.
[19] Z. Peng, J. Hu, T. Xie, C. Liu, Design of multiple operating degrees-of-freedom planetary gear trains with variable structure, J. Mech. Des. 137 (9) (2015)
1–11 no. 093301, doi:10.1115/1.4030856.
[20] G. Wu, Z. Dong, Design, analysis and modeling of a novel hybrid powertrain system based on hybridized automated manual transmission, Mech. Syst.
Signal Process. 93 (2017) 688–705, doi:10.1016/j.ymssp.2016.12.029.
[21] T. Hofman, S. Ebbesen, L. Guzzella, Topology optimization for hybrid electric vehicles with automated transmissions, IEEE Trans. Veh. Technol. 61 (6)
(2012) 2242–2450, doi:10.1109/TVT.2012.2196299.
[22] J.K. Sun, K.S. Kim, D. Kum, Feasibility assessment and design optimization of a clutchless multimode parallel hybrid electric powertrain, IEEE/ASME
Trans. Mechatron. 21 (2) (2016) 774–786, doi:10.1109/TMECH.2015.2469592.
[23] Y. Yang, X. Zhao, D. Qin, Z. Duan, H. Guo, Design and operation mode coupling characteristics analysis on the new transmission system of hybrid
electric vehicle, Autom. Eng. 34 (11) (2012) 675–968, doi:10.19562/j.chinasae.qcgc.2012.11.002.
[24] Y. You, D. Sun, D. Qin, Shift strategy of a new continuously variable transmission based wheel loader, Mech. Mach. Theory 130 (2018) 313–329,
doi:10.1016/j.mechmachtheory.2018.08.004.
[25] Y. Yang, X. Hu, H. Pei, Z. Peng, Comparison of power-split and parallel hybrid powertrain architectures with a single electric machine: dynamic pro-
gramming approach, Appl. Energy 168 (2016) 683–690, doi:10.1016/j.apenergy.2016.02.023.
[26] J. Demuynck, D. Bosteels, M.D. Paepe, C. Favre, J. May, S. Verhelst, Recommendations for the new WLTP cycle based on an analysis of vehicle emission
measurements on NEDC and CADC, Energy Policy 49 (1) (2012) 234–242, doi:10.1016/j.enpol.2012.05.081.
[27] T. Ouyang, S. Li, G. Huang, F. Zhou, N. Chen, Mathematical modeling and performance prediction of a clutch actuator for heavy-duty automatic trans-
mission vehicles, Mech. Mach. Theory (2019) 190–205 2019, doi:10.1016/j.mechmachtheory.2019.03.003.
[28] J. Hu, Z. Guo, H. Peng, D. Zheng, Research on regenerative braking control strategy of plug-in hybrid electric vehicle considering cvt ratio rate of
change, proceedings of the institution of mechanical engineers, Part D J. Autom. Eng. 232 (14) (2018) 1931–1943, doi:10.1177/0954407017735681.