Diagnostic model of electric multiple units brake control system based on particle swarm optimization

Abstract

In order to explore the application of particle swarm optimization (PSO) in the control of electric multiple unit (EMU) brake control system, the principle of EMU braking system, block partition and Quantum swarm optimization (QSO) is understood. PSO is applied to the diagnosis model of EMU braking process, and then the data are analyzed through computer simulation experiments. The results show that PSO algorithm has obvious effect on the diagnosis model of EMU braking process under railway block partition. When using PSO, the number of iterations of EMU braking system model to achieve convergence equilibrium is much less than that of ordinary model, which has obvious rapidity and ease of implementation and has great significance especially in safety. In summary, PSO is very suitable for being applied the diagnosis model of EMU brake control system.

Keywords

Brake system blocking zone model diagnosis PSO

1. Introduction

Railway is not only an indispensable infrastructure of a country, but also a preferred means of travel. It is the backbone of the comprehensive transportation system. Railway plays an irreplaceable role in promoting sound and rapid economic and social development. In January 2004, the State Council adopted the first “Medium- and Long-Term Railway Network Planning” in Chinese history. The plan proposes that by 2020, the total scale of the railway network will exceed 100,000 kilometres and the passenger dedicated line will exceed 120,000 kilometres. In 2008, due to the new situation and demand of China’s national economic development, the State Council made timely adjustments to the plan: the total scale of the national railway network rose from 100,000 km to 120,000 km, and the passenger dedicated line was also adjusted from 12,000 km to 16,000 km. Nowadays, Automatic Train Control (ATC) has been widely used in the railway system of most countries in the world. This system not only includes microelectronics and communication technology, but also includes automatic control and railway signal system [1, 2]. It integrates them well and provides more powerful guarantee for the safety, reliability and comfort of the train in operation [3]. Based on the research on the train control system of high-speed railway in developed countries, China has developed Chinese Train Control System (CTCS) which is mainly used to detect the running speed of the train at the current time [4]. When the train runs over speed, the brake system will automatically brake, so as to avoid safety accidents caused by unnatural conditions such as artificial mis-operation [5, 6]. Because the whole running process of high-speed train has many uncertainties, it is easy to be affected by many factors from outside (climate environment, line condition, etc.) [7]. It is difficult to realize the automatic adjustment of traction/braking force under different operating conditions. Therefore, in order to ensure that the train can operate safely under changeable conditions, the train braking system must have good safety and strong adaptability [8].

At present, common braking modes are usually adopted in the operation of electric multiple units (EMUs). According to the characteristics of regenerative braking and air braking, the main common braking mode is combining regenerative braking and air braking. Particle swarm optimization (PSO) is applied to the braking system model of EMU to calculate the data iteratively, which proves that PSO is efficient, economical and fast for braking force in the process of EMU braking. The advantages of PSO are proved by the case results of specific simulation cases.

2. Method

2.1 Braking characteristic of motor vehicle and principle of PSO

The braking of high-speed train has two characteristics: one is the coordination of various braking modes and the installation of anti-skid devices; the other is that the braking control of train generally adopts more sensitive and rapid systems such as electronic control or computer-controlled electrical instructions, and the system control chart is shown in Fig. 1.

Figure 1.

Control chart of brake control system for EMU.

Braking force and braking distance. In the process of train operation, the braking force that hinders the train operation is produced by manual operation, and the braking force is used to reduce the train running speed or stop, and this is the train braking [9]. Now the braking equipment of EMU has been perfected. When the braking behavior of train is needed, the braking equipment will produce corresponding resistance according to the actual situation, which is called the train braking force. Through the function of train braking force, the purpose of reducing the running speed of EMU is achieved. The natural basic resistance and additional resistance can be used but cannot be controlled, while the train braking force is generated by the braking equipment, which can be manually controlled to achieve the control of the braking force [10]. It is precisely because of the controllability of the braking force that the most direct and effective train braking behavior can be produced through artificial operation. The braking force of high-speed EMU is generally given by the manufacturer by the braking characteristic curve or braking deceleration. From this, the braking distance formula of high-speed train is obtained as follows:

$\displaystyle S_{b}=\frac{v_{o}}{t_{k}}+\frac{1}{25.92}\times\frac{\left({v_{1% }^{2}-v_{2}^{2}}\right)}{\left({a_{b}+\frac{\left(i_{j}\times g\right)}{\left(% {1+r}\right)\times 1000}}\right)}$ (1)

In Eq. (1), $t_{k}$ refers to braking time; $V_{o}$ indicates initial braking speed; $V_{l}$ and $V_{2}$ are the initial velocity and final velocity of the interval, which should not exceed 10 km/h, and the smaller the interval is, the more accurate it is.

$r$ suggests rotating mass coefficient; $a_{b}$ is raking deceleration (including deceleration caused by various braking forces and basic resistance); $i_{j}$ means braking section added slope; $g$ is gravity acceleration, about 9.80 m/s ${}^{2}$ .

Similar to other evolutionary algorithms, PSO uses the concepts of “population” and “evolution” and operates according to the fitness of individuals (particles). The difference is that, unlike other evolutionary algorithms, PSO does not use evolutionary operators for individuals, but regards each individual as a particle without weight and volume in the $n$ -dimensional search space, and flies at a certain speed in the search space. The individual flight experience and group flight experience of the flight speed are adjusted dynamically.

2.2 Automatic block

Automatic block is a method of automatically changing signal display according to train operation and the state of closed block zones. While the blocking method that the driver drives by signal is characterized by: dividing the stations into several blocking zones, occupying inspection equipment, driving by signal display, locomotive signal or on-board signal controlled by train operation; realizing train tracking between stations; automatically handling blocking procedures and automatically changing signal display during departure route.

At present, the passenger dedicated lines constructed in China are divided into two speed classes: one is the high-speed passenger dedicated lines with speeds of 300–350 kilometres per hour, such as the Wuhan-Guangzhou passenger dedicated line and the Beijing-Shanghai high-speed railway; the other is the passenger dedicated lines with speeds of 200–250 kilometres per hour. The latter can be subdivided into two types: one is pure passenger dedicated line (such as Guangzhou-Zhuhai Intercity Railway) and the other is passenger dedicated line (such as Shanghai-Hanrong Railway along the Yangtze River, Eastern Coastal Railway). For pure passenger dedicated lines, according to Quasi-mobile blocking mode, blocking zones can be designed with equal length. However, for passenger dedicated lines with 200 km/h or more EMUs, 160 km/h passenger trains and 120 km/h or less freight trains, the train control technology is complex and totally different from the actual situation at home and abroad. It is necessary to consider different train speed and control modes to design blocking zones so as to give full play to the system compatibility and transport capacity.

This paper mainly aims at this kind of mixed passenger and cargo passenger dedicated line, and carries on the analysis of the related factors in the design of block zoning. The 200 km/h and above EMUs adopt the target distance mode curve control mode of on-board train control equipment, and the train control adopts the locomotive signal. Passenger trains below 140 km/h and freight trains below 140 km/h adopt on-board equipment graded speed control and speed difference automatic blocking system. In the design of blocking zones, the blocking zones are divided according to the speed difference automatic blocking system, and the layout of the signal passing through is carried out. Then, the braking distance and tracking interval of 200 km/h and above EMUs are checked.

2.3 Algorithm solving

First, determine the number of block partitions. According to Eqs (2) and (3), the number range of block partitions that can be divided is obtained, and then the dimension of each signal point particle can be determined. At the same time, it is also prepared for the coding of solution and the generation of initial population.

$\displaystyle n=N+1$ (2) $\displaystyle\frac{l_{\text{section}}}{l_{\max}}\leqslant n\leqslant\frac{l_{% \text{section}}}{l_{\min}}$ (3)

Second, use $X^{(k)}=(X_{1}^{(k)},X_{2}^{k},\ldots,X_{N}^{k})$ to represent a solution, namely a division scheme. Iterative algebra is $k$ , $K_{N}^{(k)}$ refers to the location of Nth signal on the line. The location information of $K_{N}^{(k)}$ is represented by real number. Because the results obtained by real coding are of high precision and easy to search, it is better to solve the optimization problem. The dimension of the solution, i.e. the number of signal points arranged, is related to the length of the block partition.

Third, generate the initial population. In the preparation stage of the algorithm, it is necessary to produce an optimal solution and set the corresponding constraints of the optimal solution. If the optimal initial population is random in the solution space, some infeasible solutions will inevitably appear. In order to make the particles tend to search in the feasible solution space, and then accelerate the convergence, an initial population generation algorithm is developed. The calculation flow is shown in Fig. 2.

Figure 2.

Flow chart solution of PSO.

The initial population is generated as follows:

Step 1:

Define the number of initial population as NP. Initialize from the first population, make $i=$ 1, and divide the number of blocked zones into $n$ . In order to avoid too wide range of signal position change, set $l=$ lsection/n, and set the number of signal machines to $j$ , so that $j=$ 1;

Step 2:

Generate $X_{ij}$ randomly in [ $l\times j-R,L\times j+R$ ], and the range of $R$ is 100–300;

Step 3:

Arrange the next signal point position, $j=j+1$ ;

Step 4:

If $j\leqslant n+1$ , then turn to step 2; otherwise $i=i+1,j=1$ ;

Step 5:

If $j\leqslant Np$ , turn to step 2;

Step 6:

End and output $X_{ij}$ . Each row of the matrix $X_{ij}$ represents a layout scheme.

2.4 Simulation experiment

The length of the interval to be divided between station A and station B is 27.4 km. The maximum common braking deceleration of CRH ${}_{5}$ train type is selected as N8 value, and the parking deceleration of incoming station is selected as N5 value. According to Quasi-mobile block, block partition is divided into $l_{\min}=$ 1100 m, and $l_{\max}=$ 3500 m. In order to facilitate the analysis of the division scheme, the average ramp of the line is 4 degrees per thousand.

3. Results and discussion

The simulation experimental results are shown in Figs 3 and 4.

Figure 3.

Ordinary iteration data.

Figure 4.

Iterative data of PSO.

From the simulation experiment data in Figs 3 and 4, it is seen that when $N=$ 20, the data of the model using the ion swarm algorithm has tended to converge smoothly, while when $N=$ 150, the data of the general data model tends to converge smoothly. The model chosen in this experiment is CRH ${}_{5}$ , the maximum speed of the type train is 200 km/h, the turnout of 14# is selected in the station, and the lateral allowable speed of the turnout is 90 km/h. The block partition model based on PSO is used to re-divide the block partition of this section of line, aiming at the number of iterations. The number of signal points is reduced to 5, the number of block partitions is reduced to 6, and the maximum tracking interval is 2 minutes. The above figures also suggest that the use of PSO algorithm, on the one hand, reduces the investment and maintenance workload of maintenance personnel and has a good economy; the most important is that it greatly improves the efficiency and safety of the work of the motor vehicle brake system.

4. Conclusion

Through the understanding of the principle of EMU braking system and the principle of model and PSO, through the experiment and analysis of the experimental data of system model, we can draw a conclusion that the application of PSO in the diagnosis model of EMU braking system is obvious. The number of convergent iterations is reduced from 150 to 20, which greatly improves the working efficiency of EMU braking system, reduces the work time, improves the security, and partitions the blocked partitions by using PSO, which greatly reduces the number of blocked partitions. There are also some shortcomings: the parameters of the algorithm need to be optimized continuously to ensure the accuracy and rapidity of the algorithm. There is no precise reference data for the original set of values in the PSO algorithm. It is necessary to further verify the initial values of the PSO algorithm and find a more accurate algorithm to describe the internal relationship between the initial values.

References

Mahmood

H.A.

Adam

N.M.

and Sahari

B.B.

, Development of a particle swarm optimisation model for estimating the homogeneity of a mixture inside a newly designed CNG-H 2-AIR mixer for a dual fuel engine: An experimental and theoretic study, Fuel217 (2018), 131–150.

Wang

Lin

and Jin

, Optimizing the shunting schedule of electric multiple units depot using an enhanced particle swarm optimization algorithm, Computational Intelligence and Neuroscience2016(1) (2016), 1–11.

Wang

Tian

and Wang

T.Y.

, Fault diagnosis of the foundation brake rigging system based on fault tree and bayesian network, Key Engineering Materials693 (2016), 1734–1740.

Wang

G.G.

Gandomi

A.H.

and Alavi

A.H.

, A hybrid method based on krill herd and quantum-behaved particle swarm optimization, Neural Computing & Applications27(4) (2016), 989–1006.

Gordan

Armaghani

D.J.

and Hajihassani

, Prediction of seismic slope stability through combination of particle swarm optimization and neural network, Engineering with Computers32(1) (2016), 85–97.

Yao

and Ping

, An improved particle swarm optimization for carton heterogeneous vehicle routing problem with a collection depot, Annals of Operations Research242(2) (2016), 303–320.

Singh

R.P.

Mukherjee

and Ghoshal

S.P.

, Particle swarm optimization with an aging leader and challengers algorithm for the solution of optimal power flow problem, Applied Soft Computing40 (2016), 161–177.

Peng

and Zhou

, Stochastic optimal operation of microgrid based on chaotic binary particle swarm optimization, IEEE Transactions on Smart Grid7(1) (2016), 66–73.

Bonyadi

M.R.

and Michalewicz

, Particle swarm optimization for single objective continuous space problems: a review, Evolutionary Computation25(1) (2017), 1–54.

10.

Gong

Y.J.

J.J.

and Zhou

, Genetic learning particle swarm optimization, IEEE Transactions on Cybernetics46(10) (2017), 2277–2290.