Layout design and optimization of industrial robot automated production line based on genetic algorithm

Abstract

With the development of the world consumption economy, the demand of the society on the production efficiency of the workshop is increasing. In order to improve the production efficiency of the industrial robot automatic production line, the working mathematical model of the industrial robot automatic production line is built, and the improved genetic algorithm is used to optimize the layout of the industrial robot automatic production line. The simulation results show that, compared with the selected design method based on a standardized production line and the optimization method based on genetic algorithm, the layout production scheme output by the improved genetic algorithm has higher space utilization efficiency, shorter material handling distance in the workshop, and better overall layout effect. When the experimental time is 14 units, the space utilization rate of the output solution of the improved genetic algorithm is 94%, and the index of the other methods is lower than 70%. Moreover, the material handling distance of the former is the closest to the straight line, and the zoning planning of production equipment and materials is the most reasonable. The experimental data show that the layout optimization method of industrial robot automatic production line based on improved genetic algorithm can effectively improve the planning quality of the production line and improve the production efficiency.

Keywords

Genetic algorithm industrial robot production line layout availability model fitness model task assignment

1. Introduction

With the increasingly fierce competition in the global market, how to improve the competitiveness of enterprises, reduce production costs, and improve profit margins has become an urgent problem for enterprises to solve [1, 2]. However, the impact of an enterprise’s facility layout on its production efficiency and profit margin is more important than any other factor. How to adjust the original facility layout is a very important issue. The industrial robot automation production line, as a place that integrates many high-techs in the field of machinery manufacturing automation – CNC machine tools, industrial robots, and intelligent control, caters to the needs of today’s social development for the industry [3, 4]. However, the traditional flow production line layout design method has the problem of low production efficiency. Therefore, it has certain practical significance to optimize the layout of the industrial robot production line [5].

Kowalski et al. [6] proposed a production line layout optimization method based on system layout design-genetic algorithm, analyzes the process flow and logistics state of the production line, establishes an objective function with the constraints of the workshop logistics intensity and workshop area, and improves the system layout design. (SLP) analysis process and initial population of genetic algorithm (GA), and use the combined method of GA and SLP to plan and study the layout of the production workshop. Through the Plant-Simulation platform, the layout of the production line is simulated and solved, and the dynamic planning of the production line layout and the calculation of the logistics intensity are realized. Analysis of the experimental results shows that the layout scheme of this method is more in line with the actual production and can provide a reasonable layout scheme for the production line, but this method cannot maximize the space utilization of the production workshop. Zhang et al. [7] proposed a design method for a standardized processing production line. According to the processing process flow and the principle of quality formation, it is equipped with key processing equipment, and a standardized production line is designed. The process parameters of each module are optimized, and the production line test experiment is carried out. The experimental results show that the production line has compact equipment connection, scientific layout, simple operation and stable operation, but the material handling distance in the workshop is long. Zhang et al. [8] proposed a Witness simulation analysis and optimization method of workshop production line. This method uses the Witness simulation platform to obtain the production efficiency of each process, analyzes the bottleneck process and key process, and uses the production line balance rate index and genetic algorithm to evaluate the equipment. Parameters such as layout and equipment resources are optimized, and finally, the Witness simulation platform is used for verification. The experimental results show that the output of the optimized model is increased, and the equipment utilization rate is significantly improved, but the layout of the work area is not reasonable.

Aiming at the problems of low space utilization, long material handling distances, and unreasonable layout in the existing methods, a layout design and optimization method of industrial robot automatic production line based on genetic algorithm is proposed. Genetic algorithm is used to solve the equipment layout optimization, which has the advantages of excellent solution set performance and fast calculation speed. Therefore, this paper studies the layout optimization of industrial robot automatic production line, sets the objective function, establishes the mathematical model of production line layout, obtains the optimal layout scheme by using genetic algorithm, and verifies the effectiveness of the optimized production line layout through simulation experiments.

2. Model design of industrial robot automated production line

Coordinate the various processes of the production line through steps such as robotic assembly line balancing and line task assignment to achieve the goal of optimizing the production line layout.

2.1 Production line availability model

At present, decomposition or approximate modeling methods are usually used to solve the buffer capacity allocation problem of production line, but the existing methods do not consider the buffer capacity constraint problem, and there is the problem of uncoordinated intervals of production line. Therefore, an approximate modeling method of production line availability and buffer capacity considering equipment reliability and buffer capacity constraint is proposed. The basic idea is to decompose the industrial robot production line composed of multiple equipment and logistics system into several buffer equivalent models, and then iteratively approximate the functional relationship between the availability of the production line and the buffer capacity.

The analytical modeling method of availability of industrial robot production line with buffer is divided into two steps: approximation and decomposition. The model is divided into continuous model and discrete model. Specifically, the continuous production model is also called process production. The process of this production is continuous. The raw materials in the process will generally change in a chemical sense. The chemical industry such as plastics, drugs, soap and fertilizers, as well as oil refining, metallurgy, food and paper making, are all continuous production. The products produced in the discrete production model are often processed and assembled by various parts provided by different suppliers, and the process continuity is not strong. For example, most machinery, automobile, household appliances and electronic manufacturing industries are typical discrete production industries. Among them, the continuous modeling method has important reference significance for the industrial robot production line discussed in this paper. Therefore, this paper will study the continuous availability modeling method. Figure 1 shows the buffer equivalent model of a continuous industrial robot automatic production line. In order to ensure its continuity, the following assumptions are made.

Figure 1.

The buffer equivalent model of a continuous industrial robot automatic production line.

(1)

The first stage of the production line is not hungry, that is, there are enough raw materials; the output of the last stage machine is not blocked, that is, there is a large enough finished product warehouse.

(2)

Any machine will not fail during the stand-by period (regardless of blocking or starvation).

(3)

There is no fault in the process of transferring the workpiece from the buffer library, and the transmission time of the workpiece in the buffer library is not counted.

(4)

The production line is continuously produced, and there is no single product.

(5)

The production line has been balanced, and all equipment is produced at the same frequency.

Assume that the production cycle of the production line is $B$ , the failure rate of equipment $M_{i}$ is $f_{i}$ , the repair rate is $r_{i}$ , and the capacity of buffer $D_{i}$ is $C_{i}$ . Equipment failure and repair time are exponentially distributed events and are independent of each other. Decompose an industrial robot production line into several buffer equivalent models:

$\displaystyle S_{i}=\sum\limits_{i=1}^{n}{\mu_{i}({\alpha_{1}-\alpha_{2}})^{2}}$ (1) $\displaystyle S_{t}=\sum\limits_{t=1}^{n}{\mu_{t}\sqrt{\beta_{1}+\beta_{2}}}$ (2)

In the formula, $S_{i}$ is the buffer capacity; $\mu_{i}$ is the production line equipment; $n$ is the number of equipment; $i$ is the equipment; $\alpha_{1}$ and $\alpha_{2}$ are the equipment productivity; $S_{t}$ is the buffer surplus level; $t$ is the equipment production capacity; $\mu_{t}$ is the production line equipment failure rate; $\beta_{1}$ and $\beta_{2}$ are the equipment work efficiency.

The steps of the industrial robot automated production line in Fig. 1 are decomposed, iterated, and divided into the following steps:

(1) Equivalent $M_{1}-D_{1}-M_{2}$ to equipment $M_{a1}$ , Eqs (3)–(5) give its equivalent availability, repair rate and failure rate:

$\displaystyle S_{a1}=\frac{\exp({\alpha_{1}-\alpha_{2}})S_{1}S_{2}}{\alpha_{1}% S_{1}-\alpha_{2}S_{2}}$ (3) $\displaystyle\mu_{a1}=\frac{1-({S_{a1}\times S_{1}})}{1-({S_{a1}\times S_{2}})}$ (4) $\displaystyle\alpha_{a1}=\frac{\mu_{a1}({1-S_{a1}})}{S_{a1}}$ (5)

In the formula, $S_{1}$ and $S_{2}$ represent the failure rate and maintenance rate of the production line equipment, respectively.

(2) Referring to step (1), the equivalent availability, repair rate, and failure rate of $M_{ai}-D_{i+1}-M_{i+2}$ are obtained as $S_{ai+1}$ , $\mu_{ai+1}$ , and $\alpha_{ai+1}$ , respectively, where $i=1,2,\ldots,N$ .

(3) The availability of the industrial robot production line is finally obtained as:

$\displaystyle S_{ak}=S_{a({n-1})}=\frac{\exp({-k({n-2})})S_{n}}{S_{a({n-1})}-% \alpha_{a({n-1})}}$ (6)

In the formula, $S_{ak}$ represents the output of the production line; $k$ represents the number of iterations.

Buffer optimization problems can be divided into two categories, that is, the goal of minimizing buffer space or maximizing production line output. This paper focuses on the layout of production lines and discusses the maximization of production line output (availability) under certain space constraints. Therefore, the objective function for this problem is $\max({S_{ak}})$ .

Due to the limited area occupied by each buffer, the total buffer is set to be less than $z$ . In addition, in order to ensure normal work, each station must have a work in process, so the constraints are:

$\displaystyle\sum\limits_{i=1}^{n}{U_{i}}\leqslant z\quad U_{i}\geqslant 1$ (7)

In the formula, $U_{i}$ represents the equipment blocking rate.

2.2 Balancing algorithm of robot production assembly line

2.2.1 Problem description

The industrial robot assembly line is composed of several workstations. Each workstation is generally equipped with a robot to complete the assembly process. Different assembly tasks with determined priority are executed according to the given order, and finally the finished products are produced. In this paper, considering the energy consumption problem, should not only reasonably allocate these assembly tasks to each workstation under the condition of given workstation beat time and task priority relationship, but also allocate the most suitable robot to each workstation because different types of robots have different time and energy consumption to perform the same task, so as to minimize the number of workstations and total energy consumption of the production assembly line. In this paper, the energy consumption of the assembly line is divided into the following four parts: processing energy consumption, mold change energy consumption, transmission energy consumption and standby energy consumption.

In order to describe the problem more effectively, the following assumptions are made: (1) Each assembly task can not be divided, and the switching relationship between the priority graph of assembly products and assembly tasks is known in advance; (2) The process time and energy consumption of each task depend on the robot of the station, and the time and energy consumption of the robot performing each task are determined; (3) Each station is assigned a robot, but there is no limit on the number of different types of robots; (4) There are no restrictions on the allocation of tasks and robots; (5) Material handling, loading and unloading time and robot tool or fixture switching time and energy can not be ignored; (6) For the same task, the switching time between different tools is equal; (7) Assembly line is a traditional assembly line that produces only one single product.

2.2.2 Mathematical model

According to the above description, the following definitions and properties are first given:

Definition 1:

Processing energy consumption $E_{i}$ , that is, the energy consumption required by the robot to complete the assembly task, excluding the energy consumption of fixture or tool replacement. It is mainly the electric energy consumed by the robot in the assembly work, which is divided into the electric energy consumed by the robot in the execution and the electric energy consumed in the workpiece transmission process. Since the assembly tasks that the robot needs to complete in the work are limited and discrete, the processing energy consumption is displayed in the form of accumulation, as shown in Eq. (8).

$\displaystyle E_{i}=\sum\limits_{i=1}^{n}{\partial_{is}}\phi_{is}$ (8)

In the formula, $\partial_{is}$ represents the execution energy consumption of the robot; $\phi_{is}$ represents the workpiece transmission energy consumption.

Property 1:

The tasks assigned to workstation $s$ are excluded according to the principle of limited relationship and the minimum number of tool switching times, and the task number in the workstation is recorded as $i_{n}$ , where $n=1,2,\ldots,N$ , the minimum number of tool switching times in workstation $s$ is:

$\displaystyle E_{s}=\int_{i}^{n}{v_{n-1},v_{n}}+v_{i_{N}}$ (9)

In the formula, $v_{n}$ represents the decision variable; $v_{i_{N}}$ represents the energy consumption of the robot.

Prove:

Since the robot in the workstation executes tasks cyclically, the previous execution task of task $t_{1}$ is task $t_{N}$ . If it is $v_{i_{N}}=1$ , tool switching is required before task $t_{1}$ is executed; otherwise, it is not required.

Definition 2:

Die change energy consumption $E_{l}$ , mainly including fixture adjustment energy consumption and tool replacement energy consumption. Because of the attitude of the object to be processed, the constituent materials, the processing procedures and other reasons, it is impossible for the processing robot arm to complete all the fixture adjustment without changing the fixture, and the fixture adjustment will inevitably consume additional energy. Moreover, the energy consumption of the robot arm will be different with different blades, so changing the tool will also bring additional energy consumption. At the same time, there is no interaction between fixture adjustment energy consumption and tool change energy consumption. Therefore, the final robot model reduction energy consumption can be obtained by direct addition. See Eq. (10) for the calculation method.

$\displaystyle E_{l}=\sum\limits_{l=1}^{n}{p_{l}}+x_{l}$ (10)

In the formula, $p_{l}$ represents the energy consumption of fixture adjustment; $x_{l}$ represents the energy consumption of tool replacement.

Definition 3:

Transmission energy consumption $E_{w}$ refers to the energy consumed by the transfer of workpieces from one workstation to the next. This part of energy consumption mainly includes the energy consumption required for handling when loading and unloading workpieces and transferring workpieces between workstations, such as the energy consumption of conveyor belt and handling manipulator. Since the transmission energy consumption of the workpiece in the transfer process is directly related to the weight of the workpiece itself, the friction coefficient between the construction and the transmission belt, the transportation distance and other factors, it is impossible to accumulate the energy consumption of each workpiece transmission in an accumulation manner. However, for the vast majority of production plants with stable operation, the total transmission workload in adjacent long periods will not fluctuate significantly. Therefore, the remaining maintenance time in each period can be multiplied by the average equipment transmission energy consumption, and then the transmission energy consumption in each period can be superposed to obtain the final total transmission energy consumption of the robot project, as shown in Eq. (11).

$\displaystyle E_{w}=\sum\limits_{i=1}^{n}{L_{ti}\times\overline{e_{w}}}$ (11)

In the formula, $L_{ti}$ represents the remaining maintenance time; $\overline{e_{w}}$ represents the average transmission energy consumption of the equipment.

Definition 4:

Standby energy consumption $E_{y}$ refers to the energy consumed by the robot in an idle state because the workpiece has not arrived. This part of energy consumption mainly includes the energy consumption of the robot during machine tool idling and mold changing, tool switching, and loading and unloading parts. For the assembly line production workshop, even if there is no product transmission on the production line, in most cases, the machine tools in the assembly line cannot stop running immediately. Instead, they need to keep idling and replace the templates, blades and extension auxiliary parts required for subsequent production. These steps need to consume more energy, and there is no correlation between the three. Therefore, the total standby energy consumption is calculated by direct superposition, as shown in Eq. (12).

$\displaystyle E_{y}=e_{1}+e_{2}+e_{3}$ (12)

In the formula, $e_{1}$ represents the energy consumption of the robot during machine tool idling and mold change; $e_{2}$ represents the energy consumption of the robot during tool switching; $e_{3}$ represents the energy consumption of the robot during loading and unloading.

According to the above problem description, assumptions and properties, the balance problem of industrial robot assembly line considering energy consumption is modeled as follows:

Objective function:

$\displaystyle\textit{MinP}=({p_{1},p_{2}})$ (13) $\displaystyle p^{1}(x)=\sum\limits_{i=1}^{n}{\sqrt{\max x_{ik}}}$ (14) $\displaystyle\omega_{p,p_{j}}=\frac{1}{Z(p)}\exp\left({-\frac{d({\bar{M}\psi_{% p},})\bar{M}\psi_{p_{j}}}{\sigma^{2}}}\right)p^{2}(x)=E_{i}+E_{l}+E_{w}+E_{y}$ (15)

In the formula, $P$ represents the total optimization objective; $p_{1}$ represents the optimization objective of the number of workstations; $p_{2}$ represents the optimization objective of the total energy consumption of the assembly line; $x_{ik}$ represents the decision variable.

The constraints are as follows:

$\displaystyle ux_{ik}-ux_{is}\leqslant 0$ (16) $\displaystyle\sum\limits_{i=1}^{n}{x_{is}}=1$ (17) $\displaystyle\sum\limits_{i=1}^{n}{y_{is}}=1$ (18)

Among them, Eq. (16) guarantees the priority relationship between tasks; Eq. (17) indicates that each task can only be assigned to one workstation; Eq. (18) indicates that each workstation can only be assigned one robot.

2.3 Task assignment of industrial robot assembly line

While there is an increasing focus on robotic systems, robotic collaboration has also received a lot of attention. In particular, robot task assignment has become an important research topic. When researchers design robotic systems, there is always a fundamental question: which robot should perform which task to achieve the goal? A task refers to a sub-goal that is necessary to achieve an overall goal and can be achieved independently of other sub-goals. The target task can be discrete, for example, a robotic arm picks up parts and delivers them to another robotic arm for assembly, or it can be continuous. Task assignment is a consideration even for relatively simple robotic systems, and the more complex the study system, the greater the importance of the task assignment problem.

In this paper, a fitness-based task allocation strategy for industrial robot assembly line work is proposed. The fitness-based task allocation strategy mainly involves two aspects: the fitness model and the subtask allocation algorithm based on the model.

2.3.1 Fitness model

In order to realize the rational allocation of the work tasks of the industrial robot assembly line, that is, to solve the problem of “which robot should perform which task to achieve the goal”, the concept of fitness is introduced in the model, and the fitness is divided into internal fitness and external fitness, which are respectively used to describe the requirements of the task itself and the ability of the executing robot. Let $\theta_{ij}$ be the fitness of subtask $t_{i}$ to robot $r_{i}$ , which consists of the following two components:

(1)
Internal fitness: the internal fitness is the fitness that is completely considered from the factors of the subtask itself. It reflects the appropriateness of executing the subtask under the current situation. The larger the value, the more appropriate the selection of subtask $t_{i}$ . That is to say, when other conditions are the same, the subtask with the larger internal fitness value can be regarded as the higher priority of processing. When the robot’s processing or production capacity is limited and cannot meet all production requirements, if other considerations are known, the subtask with the lowest internal fitness should be abandoned.
(2)
External fitness: external fitness is considered from the two aspects of subtask and robot. It reflects whether the capability requirements of subtask adapt to the capability configuration of robot. The larger its value is, the better the subtask adapts to the robot in terms of capability requirements. From the perspective of external fitness index, the goal of industrial robot pipeline task planning is to make as many subtasks as possible obtain greater external fitness. Therefore, only allowing some special subtasks, such as those with high internal fitness, to obtain higher external fitness, can not well complete the pipeline task allocation of industrial robots. In fact, this allocation method is not conducive to the pipeline work.

The fitness of all subtasks shall meet:

$\displaystyle\prod\limits_{i=1}^{n}{\theta_{ij}}=1$ (19)

Before calculating the fitness, 3 state variables are calculated first: the subtask end flag, the subtask function requirement flag, and the upper limit of the required number of robots reached flag.

Let $\psi_{i}(t)$ be the end mark of subtask $t_{i}$ , and its expression is:

$\displaystyle\psi_{i}(t)=\left\{{\begin{array}[]{ll}1&i\geqslant 0\\ 0&i<0\\ \end{array}}\right.$ (20)

In the formula, 1 means that subtask $t_{i}$ at time $t$ has been completed; 0 means that subtask $t_{i}$ at time $t$ has not been completed.

Let $\zeta_{i}(t)$ be the sign that robot $r_{i}$ meets the functional requirements of subtask $t_{i}$ , and its expression is:

$\displaystyle\zeta_{i}(t)=\left\{{\begin{array}[]{ll}1&V_{n}\in V_{p}\\ 0&V_{n}\notin V_{p}\\ \end{array}}\right.$ (21)

In the formula, $V_{n}$ represents the allowed number of robots; $V_{p}$ represents the upper limit of the number of robots.

According to the above formula, if subtask $t_{i}$ is still not completed at time $t$ , robot $r_{i}$ meets the functional requirements of subtask $t_{i}$ , and the upper limit of the number of robots in subtask $t_{i}$ has not been reached, then $c_{i}(t)=1$ , the switch is turned on, and robot $r_{i}$ can select the subtask; otherwise, the switch is closed and robot $r_{i}$ has no right to select subtasks.

According to the above conclusions, the expressions of the internal fitness model and the external fitness model are given:

$\displaystyle\theta_{ij}({in})=\frac{1}{Z_{1}}\exp\left({\frac{|{p_{i}-p_{j}}|% }{\lambda}}\right)$ (22) $\displaystyle\theta_{ij}(\textit{out})=\frac{1}{Z_{2}}\exp\left({\frac{|{p_{i}% -p_{j}}|}{\lambda}}\right)$ (23)

In the formula, $Z_{1}$ represents the internal fitness coefficient; $Z_{2}$ represents the external fitness coefficient.
2.3.2 Subtask assignment algorithm

According to the constructed fitness model, a robot subtask selection algorithm is proposed, which is simple, fast and effective. Assuming that robot $r_{i}$ selects a subtask at time $t=hT$ , the specific steps of the selection process are as follows:

(1)
Update the robot’s own status and check its own functions;
(2)
Interact with other robots to obtain the status and functional characteristics of other robots and task consumption time;
(3)
Update the status of all subtasks;
(4)
Calculate the fitness value $\theta_{ij}$ of all subtasks;
(5)
Select subtask $t_{i}=\max\theta_{ij}$ .

When each task selection period $T$ arrives, the robot subtask selection module determines whether the robot is currently performing a task. If the robot is in an idle state, the above task selection algorithm is activated; otherwise, it monitors the robot’s task execution status. In the process of monitoring the robot’s execution of tasks, if the monitoring program finds that the robot has a hardware failure or the execution times out, the current task of the robot is interrupted, so that the robot is in an idle state.
3. Layout design and optimization of industrial robot automated production line

The genetic algorithm used to optimize the industrial robot production line has poor robustness, slow convergence speed and the global optimality of the solution needs to be improved due to its own characteristics. Therefore, this research improves the genetic algorithm from the steps of chromosome coding scheme, fitness function and cross operation.

3.1 Principle of genetic algorithm

(1) Determination of chromosome coding scheme

In the genetic algorithm, the method of transforming the feasible solution of a problem from its solution space to the search space that the genetic algorithm can identify is called decoding. In the process of production line layout design, fundamentally, the solution is to solve the order of equipment. Therefore, the order of machine tools can be directly encoded, and the variables of the problem can be directly encoded as chromosome genes [9, 10]. However, due to the long and lack of regularity of the binary code string, the binary code is not suitable for direct application to the production planning problem. Therefore, it is improved here that the sequence number is used to represent the assignment of job elements, and each gene is represented by a corresponding sequence number.

(2) Generation of the initial population

$n$ data string is randomly generated, each data string is called an individual, and the generated $n$ individuals are called a population. During the operation, the randomly generated individual is used as the initial value to iterate. When running the genetic algorithm, it is necessary to comprehensively weigh the settings of the group parameters. Generally speaking, the group size is set within 10–200 [11, 12].

(3) Calculate chromosome fitness

In the early stage of calculation of genetic algorithm, the fitness of individuals in the population is greatly different, and individuals with large fitness may appear. These individuals are more likely to enter the reference population, thus reducing the diversity of the population and making the algorithm fall into local optimization. Therefore, in the early stage of the algorithm calculation, we should reduce the population diversity caused by extreme individuals by reducing the fitness. At the end of the algorithm calculation, the fitness difference between individuals in the algorithm is often too small, which easily leads to insufficient individual competitiveness. Therefore, it is necessary to increase the fitness appropriately to improve the competitiveness of individuals in the population. To sum up, the fitness function $f(J)$ of the genetic algorithm is set as Eq. (24):

$\displaystyle f(J)=1-\frac{J}{\mu}$ (24)

In Eq. (24), $J$ represents fitness and $\mu$ represents a constant not less than $J$ .

(4) Designing genetic operators

Design a selection genetic operator, which is used to select individuals with higher adaptability for inheritance; design a crossover genetic operator, which is used for chromosomal crossover to generate new individuals. Due to the constraint of the sequence of factory operations, the routine cross operation is easy to lead to the invalidation of the individual. Therefore, the priority relationship method is adopted to generate the offspring. Specifically, in this method, the cross point gene segments of the first generation of parents are sorted with reference to the gene sequence of the corresponding other parents, and the gene sequence of other parts remains unchanged. Thus, after the generation of offspring chromosomes, the subsequent offspring chromosomes are also formed in the same way. If the crossover probability is too low, the parameter individuals will be directly copied to the next generation parameter individuals, and the search of the algorithm is in a stopped state; therefore, the suggested crossover probability ranges from 0.4 to 0.99. The mutation genetic operator is designed to generate new individuals by chromosomal mutation. When performing genetic algorithm, the range of mutation probability is generally recommended to be 0.0001–0.1. Termination condition: if $t<T$ , go to step (2); if $t>T$ , take the individual with the largest fitness in the evolution process as the optimal solution output, and terminate the calculation. The basic flow of the improved genetic algorithm designed in this study is shown in Fig. 2.

Figure 2.

Basic flow chart of improved genetic algorithm.

According to the above design, the complete design of the improved genetic algorithm program is completed.

3.2 Mathematical model of production line layout optimization

According to the actual situation of the workshop, in order to increase the breadth and practicability of the mathematical model for the optimization of the production line layout, it is necessary to make some assumptions about the model. The assumptions are as follows:

(1)
It is assumed that the areas that need to be laid out in the workshop are all rectangular block structures, and the length and width are known.
(2)
Each work unit is arranged in rows and parallel to the X-axis.
(3)
Assume that the input and output cargo points are located at the midpoint of the work unit.
(4)
Multi-line layout, that is, the arrangement of work units should be arranged in at least 2 lines.
(5)
Materials, semi-finished products and finished products all run in the horizontal or vertical direction.

The production line layout studied in this paper is to minimize the total material handling distance between each operation unit, and its objective function can be expressed as:

$\displaystyle C_{f}=\sum\limits_{i,j=1}^{n}{G_{ij}}Y_{ij}$ (25)

In the formula, $G_{ij}$ represents the state of the material; $Y_{ij}$ represents the logistics route.

The distance expression between work units is as follows:

$\displaystyle D_{ij}=\frac{|{a_{i}-a_{j}}|}{|{b_{i}-b_{j}}|^{2}}$ (26)

In the formula, $a_{i}$ and $a_{j}$ represent the work unit of the x-axis; $b_{i}$ and $b_{j}$ represent the work unit of the y-axis.

The formula for solving the abscissa of each work unit is:

$\displaystyle x_{ij}=\frac{1}{a_{i}+a_{j}}$ (27)

The formula for solving the ordinate of each work unit is:

$\displaystyle y_{ij}=\frac{1}{b_{i}+b_{j}}$ (28)

The abscissa and ordinate need to satisfy the constraint that the operation units of the same line do not overlap. Under this constraint, the production line layout optimization model is constructed:

$\displaystyle\vartheta_{k}=\sum\limits_{i,j=1}^{n}{C_{f}}\times({x_{ij}+y_{ij}})$ (29)

Considering the constraints in the actual production line, this paper designs the production line layout optimization scheme through genetic algorithm.
4. Simulation experiments

In order to verify the effectiveness of the layout design and optimization method of industrial robot automatic production line based on improved genetic algorithm (hereinafter referred to as “method 1”), experiments were conducted. The design method of standardized processing production line (hereinafter referred to as “method 2”) and production line layout optimization design method based on classical genetic algorithm (hereinafter referred to as “method 3”) are used as comparative methods to carry out parallel experiments.

4.1 Experimental environment and parameter settings

Simulation environment: Win10, 64-bit system, clocked at 2.4GHz, i7-2430, dual-core Intel second-generation Core processor. This paper uses the Plant Simulation software simulation platform to simulate the working conditions of the production line before and after the optimization design.

Company A was selected as the experimental object. The company mainly produces escalators. The main work of the workshop is to assemble escalators. There is an area inside the workshop for the production of steps, mainly including standard steps and non-standard steps. At present, the company’s annual production capacity is about 6000 units. Since the number of steps used on each escalator is different, the specific number is not counted. The total area of the company’s cascade production line is 65 m $\times$ 50 m, with a total of 16 areas, 1–16 are: raw material area, temporary storage area, stamping line 1, pedal production line 1, kick plate production line 1, pedal production line 2, kick plate production line 1, board production line 2, stamping line 2, material area 1, electrophoresis area, assembly area 1, assembly area 2, finished product area 1, material area 2, material area 3, finished product area 2. The product transportation of this production line is mainly carried by forklifts and trailers, and does not need to be carried by a hanging crane. Therefore, the costs incurred by the crane can be ignored. Table 1 shows the specific conditions of each working area of the production line.

Table 1
Basic situation of each working area of the production line

Workspace number	Workspace name	Area/m ${}^{2}$
1	Raw material area	120
2	Staging area	180
3	Stamping 1 line	80
4	Pedal production line 1	70
5	Kick plate production line 1	50
6	Pedal production line 2	80
7	Kick plate production line 2	80
8	Stamping 2 lines	70
9	Material Zone 1	110
10	Electrophoresis area	150
11	Assembly Zone 1	130
12	Assembly Zone 2	150
13	Finished Zone 1	200
14	Material Zone 2	230
15	Material Zone 3	220
16	Finished Zone 2	240

4.2 Analysis of results

(1) Space utilization

The greater the space utilization rate, the more rationally the space in the production line can be used. Therefore, using the space utilization rate as the experimental index, the layout optimization effects of the three methods are compared, and the results are shown in Fig. 3.

Figure 3.

Comparison results of space utilization of different methods.

It can be seen from Fig. 3 that with the increase of the number of experiments, the space utilization rate of the three methods is on the rise. Among them, the difference between the space utilization ratio of method 2 and method 3 is small, and the maximum value of both methods is not more than 70%. In contrast, the space utilization rate of method 1 proposed in this study is the highest at any experimental time, and its own maximum value is 94%. It shows that the robot production line layout planning method based on improved genetic algorithm proposed in this study has the best effect in space utilization, and the automatic production line layout is the most reasonable.

(2) Material handling distance in workshop

The smaller the material handling distance in the workshop, the more reasonable the route setting, which can save a lot of handling time and improve production efficiency. Therefore, taking the material handling distance in the workshop as the experimental index, compare the layout optimization effects of the three methods, and the results are shown in Fig. 4.

Figure 4.

Comparison results of material handling distances in workshops by different methods.

Figure 4 shows the workshop material handling paths of the three methods, where the closer the line is to a straight line, the shorter the path, that is, the shorter the workshop material handling distance. It can be seen from Fig. 4 that the material handling path in the workshop of method 3 is the closest to a straight line. Compared with method 1 and method 2, it has obvious advantages, indicating that the proposed method effectively reduces the material handling distance in the workshop and is conducive to improving the production efficiency of the production line.

(3) Layout effect analysis

The simulation model of the production line layout before and after the improvement is established in the Plant Simulation software, as shown in Fig. 5.

Figure 5.

Layout effect.

It can be seen from Fig. 5 that there are similar problems between various areas in the production line before optimization, such as short intervals and long distances between different areas. For example, the material area is not closely connected with the assembly area and finished product area. However, the optimized layout makes the intervals closer, which can shorten the material transmission distance. Therefore, it can be seen that the layout optimization effect of the proposed method is better.

5. Discussion

In this study, the layout design and optimization application effect of the industrial robot automatic production line in the workshop based on the improved genetic algorithm, the standardized method and the classical genetic algorithm are simulated and tested from the three angles of space utilization, material handling distance in the workshop and layout effect. The test results are analyzed and compared with similar studies in the industry.

From the perspective of space utilization, method 1, that is, in the experimental scheme of using the improved genetic algorithm proposed in this study for robot pipeline task and layout planning, the space utilization rate at any experimental time is significantly higher than the other two commonly used planning methods in the industry. Specifically, when the experimental time is 14, the space utilization rates of method 1, method 2 and method 3 are 94%, 65% and 59%, respectively. This is because the coding strategy, genetic and cross calculation rules and chromosome stress moderation calculation function of the genetic algorithm used to optimize the task in method 1 have been adjusted in consideration of the production reality of the robot factory. The improved genetic algorithm can calculate the global optimality better than the classical genetic algorithm. The research results of Huang Q and others in the aircraft maintenance industry also show the same law, that is, improving the genetic algorithm according to the actual application situation can improve the global optimality of its output solution [13].

From the perspective of vehicle material handling distance, since the ideal moving path should be a straight line parallel to the horizontal axis, the closer the planned transport path is to the straight line, the better the planned effect. As is evident in Fig. 4. The first method based on the improved genetic algorithm is the shortest transportation distance, followed by the standardized planning method. The planning method based on the genetic algorithm outputs the longest planning path. The reason why the results of method 1 are more effective is still that the optimized algorithm can generate a better population, thus improving the global optimality of the output results. The research results of Elhefnawy also prove this point [14].

From the perspective of layout effect, the layout of each production robot in the escalator production and assembly workshop after the optimization of method 1 is more reasonable. For example, the finished product area is close to each other, and the material area is adjacent to the assembly area. This planning method that integrates and partitions the processing procedures makes the production efficiency of the workshop and the transportation of semi-finished products and raw materials more convenient and efficient. While the output results of the other two methods have been similarly integrated in some sub regions, but the degree of division is significantly worse than that of the first method. The limited results of the production lines of the leather industry by Jiang Dy and others are inconsistent with that of this study, and the optimization effect is poor [15]. This is because the solution ability of the optimization method they choose is worse.

6. Conclusion

In order to improve the space utilization rate of the production line, reduce the material handling distance in the workshop, and improve the rationality of the layout, a layout design and optimization method of the industrial robot automated production line based on genetic algorithm is proposed. The main research contents of this method are as follows:

(1)

By establishing a production line availability model to ensure the continuity of the production line, by designing a robot production assembly line balancing algorithm and an industrial robot assembly line work task allocation method, improve the adaptability of task capability requirements and robot capability configuration;

(2)

Use genetic algorithm to establish a mathematical model of production line layout optimization, establish an objective function, and realize the layout optimization of industrial robot automated production line through this model;

(3)

The experimental results show that the space utilization rate of the proposed method has been greatly improved, the material handling distance in the workshop is effectively reduced, and the production line layout is more reasonable.

References

Leiber

Eickholt

Vuong

Reinhart

. Simulation-based layout optimization for multi-station assembly lines. J Intell Manufact. 2021; 33(2): 537-554.

Saeedi

Feizi

. Modeling and optimization of batch production based on layout and cutting problems under uncertainty. RAIRO-Oper Res. 2021; 55(1): 61-81.

Souza

Lima

Baioco

. Process plant layout optimization of an FPSO. IEEE Latin Am Trans. 2019; 17(10): 1724-1732.

Padhee

Nandan

. Design of automated visual inspection system for beverage industry production line. Traitement du Signal. 2021; 38(2): 461-466.

Kaltenbrunner

Huka

Gronalt

. Heuristic based approach for short term production planning in highly automated customer oriented pallet production. J Intell Manufact. 2022; 33(4): 1087-1098.

Kowalski

Izdebski

Goda

Manerowski

. Planning and management of aircraft maintenance using a genetic algorithm. Eksploatacja i Niezawodnosc-Maint Reliab. 2021; 23(1): 143-153.

Zhang

Fan

Wang

Tang

. Design and application of flat green tea standardized processing line. Mod Food Sci Technol. 2021; 37(6): 217-221.

Zhang

Gui

Deng

Zhou

. Optimization design of modular support structure for fish egg machine based on response surface methodology. Comput Simul. 2020; 37(10): 217-220+262.

Xiao

Qian

. Research on simulation technology of FMS flexible manufacturing line. Comput Simul. 2019; 36(8): 204-209+270.

10.

Daves

Ervin

Zeng

. Experimental data interpretation using genetic algorithm for global health assessment of reinforced concrete slabs subjected to cracking. Adv Struct Eng. 2021; 24(3): 411-421.

11.

Sochacki

Garus

Szmidla

Nabiaek

Garus

. Designing two-band mechanical wave filters using genetic algorithm. Acta Phys Polonica A. 2021; 139(5): 479-482.

12.

Panda

Nayak

. Realization of VLSI circuit partitioning using advanced genetic algorithm. Solid State Technol. 2021; 63(5): 9129-9145.

13.

Huang

Zhou

Zhang

Wang

Fan

Sun

. Layout optimization of dip dyeing workshop based on system layout planning-genetic algorithm. J Textile Res. 2020; 41(3): 84-90.

14.

Elhefnawy

Ragheb

Sharaf

Hegazy

. Optimised coordinated chassis stability control using fuzzy genetic algorithm. Int J Heavy Veh Syst. 2021; 27(6): 800-816.

15.

Jiang

Zhou

Dan

. Simulation analysis and optimization of tannery production line by Witness. J Comput Appl. 2020; 40(2): 117-120.

Layout design and optimization of industrial robot automated production line based on genetic algorithm

Abstract

Keywords

1. Introduction

2. Model design of industrial robot automated production line

2.1 Production line availability model

2.2.1 Problem description

2.2.2 Mathematical model

2.3.1 Fitness model

3.1 Principle of genetic algorithm

(1) Determination of chromosome coding scheme

(2) Generation of the initial population

(3) Calculate chromosome fitness

(4) Designing genetic operators

4.1 Experimental environment and parameter settings

Table 1 Basic situation of each working area of the production line

(1) Space utilization

(2) Material handling distance in workshop

(3) Layout effect analysis

6. Conclusion

References

Table 1
Basic situation of each working area of the production line