Unreliable series-parallel connection production line performance evaluation method considering setting time
Through the Markov chain-based method, considering the setting time and evaluating the performance of unreliable series-parallel production lines, the problem of unconsidered impact of setting time in the prior art is solved, and accurate evaluation of production line performance and control decision support is achieved.
Patent Information
- Application Number
- CN202411095515.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-12-03
AI Technical Summary
The existing technology fails to effectively consider the impact of setting time on the performance evaluation of series and parallel production lines, making it difficult to solve system performance indicators, especially when the production line needs to frequently switch products, the length of setting time has a significant impact on the system state transition and performance.
Based on the Markov chain characterization system state transfer process, an unreliable series-parallel production line performance evaluation method considering the setting time is proposed. By establishing a two-stage and multi-stage production line analytical model, the Markov process is used to solve the system steady-state probability distribution. Calculate performance indicators such as throughput, product-in-process inventory level, hunger rate and blockage rate.
Through numerical experimental verification, the method can accurately evaluate the performance indicators of multi-stage production lines, with small errors and high degree of consistency, providing a more scientific basis for making production control decisions.
Smart Images

Figure CN119066846A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of performance evaluation of unreliable series-parallel production lines, and in particular to a performance evaluation method of unreliable series-parallel production lines taking setup time into consideration. Background Art
[0002] The series-parallel production line is also called a flexible assembly line or a mixed assembly line. It consists of workstations with parallel machines and buffers that are connected in series alternately. It aims to achieve an effective conversion of resource input and output. It is one of the commonly used production organization forms in modern production. Compared with traditional assembly lines, it has stronger adaptability and flexibility.
[0003] In modern manufacturing, in order to meet the diversified needs of customers, production lines are designed to have a certain degree of flexibility so that limited resources can be used to produce a variety of products with different specifications. Therefore, series-parallel production lines are widely used in the manufacturing fields of automobiles, steel, home appliances, and electronics. For example, in automobile production, cars of the same brand have different styles, different powers, and different on-board equipment, and models with different configurations can be assembled in the same production line; in the production of LCD TVs, products of different sizes, appearances and colors, visual effects, and additional functions can be assembled in the same production unit. In addition to being able to meet the personalized needs of customers, in most flexible manufacturing production environments, the actual setup time (Setup Time) when switching product types is very long, making the setup time non-negligible. Therefore, when the same production unit switches from one product to another, the setup time needs to be considered.
[0004] The diversity of products, setup time, machine unreliability and the limited buffer are coupled with each other, making the production system a random, unreliable and complex system, which not only affects the effective output of the single-station machine, but also propagates disturbances along the upstream and downstream of the production line, thereby causing greater output changes in the system, resulting in a nonlinear relationship between the input and output processes of the series-parallel production line, and performance indicators are difficult to solve. The performance evaluation of the production system aims to use mathematical models or simulation methods to characterize the production process, reveal the laws of system operation and evolution, and complete the evaluation of performance indicators by analyzing and quantifying the propagation effects of random disturbance factors in the system. It is not only an important tool for evaluating the effectiveness and measuring the benefits of the production system, but also an important basis for improving the production system's capacity and lean design. It has important guiding significance for scientific control and decision-making of the production process.
[0005] Common methods for evaluating the performance of production systems include simulation methods and analytical methods. The simulation method is suitable for complex production lines. It mainly establishes a simulation model to simulate the operation of the production line and obtain system performance indicators. It is suitable for simulating and analyzing complex manufacturing systems. The analytical method establishes a mathematical model to characterize the relationship between parameters such as equipment reliability, quality characteristics, and buffer capacity and performance indicators such as production line throughput and average number of work-in-progress, which can deeply reveal the operation mechanism of the production system. The analytical method includes precise analytical method and approximate analytical method. The former describes the system state through Markov process, uses mathematical tools such as random process to express the relationship between material flow, workstation state and buffer capacity, establishes the system state transition equation, and obtains the analytical expression of the system performance indicator by solving the state transition equation. The latter obtains an approximate solution that is close to the actual performance by calling the precise model of the two-station production line for multiple iterations. The approximate analytical method also includes aggregation method and decomposition method. Specifically, the aggregation method merges two workstations into a single equivalent workstation, and realizes the recursive iteration of machine parameters through alternating backward and forward aggregation processes until convergence and obtaining approximate solutions of performance indicators such as the system throughput; the decomposition method decomposes the initial production system into multiple two-machine-one-buffer subsystems, and uses an iterative algorithm to solve the system performance indicators based on the relationship between the subsystems. Summary of the invention
[0006] At present, relevant research on production system performance evaluation mainly focuses on the impact of characteristic parameters such as workstation reliability and buffer capacity on system performance, and has not yet considered the impact of setup time on system state transition and performance index characterization. In actual production, the mold change (setup) characteristics of series-parallel production lines that produce multiple products cannot be ignored. In order to effectively cope with the above-mentioned difficulties and challenges, the present invention focuses on actual production needs, targets unreliable series-parallel production lines that take setup time into consideration, characterizes the system state transition process based on Markov chains, explores its operating mechanism, and proposes a performance evaluation method for unreliable series-parallel production lines that take setup time into consideration, providing a performance evaluation tool for production control decisions.
[0007] The technical solution of the present invention is:
[0008] A method for evaluating the performance of an unreliable series-parallel production line considering setup time comprises the following steps:
[0009] Step 1: Establish the unreliable series-parallel production line system model and determine the unreliable series-parallel production line performance indicators:
[0010] The unreliable series-parallel production line consists of m workstations {WS1, WS2, …, WS m} and (m-1) buffers B i Each workstation contains S i Parallel unreliable machines Where i = 1, 2, ..., m; the workpiece to be processed passes through the workstation WS in sequence i Processing and buffer B i Storage and transfer from the last workstation WS m Outflow system;
[0011] The series-parallel production line produces a total of K products, of types 1, 2, 3, ..., K. All products arrive randomly, and the probability of each type of product entering the system is equal;
[0012] When the product currently processed by the machine is of a different type from the previous product, setup operations are required before processing. Assume that the setup time for different products on different machines is the same, and the setup time is TS;
[0013] The performance indicators adopt the throughput rate, work-in-process inventory level, blocking rate and / or starvation rate of the unreliable series-parallel production line;
[0014] Step 2: Merge parallel machines in series and parallel production lines:
[0015] Merge S parallel machines into one equivalent virtual machine Its parameters are in is the average working time of the combined machines, is the combined machine failure working time, For the combined machine Processing cycle;
[0016] Step 3: Two-stage line performance evaluation considering setup time:
[0017] A two-stage production line analytical model is established based on the Markov process, and the system performance evaluation index is calculated by solving its steady-state probability distribution;
[0018] Step 4: Multi-stage production line performance evaluation considering setup time:
[0019] Based on the analytical model of the two-stage production line, the two workstations are aggregated into a single equivalent workstation, and the aggregation process is recursively iterated until convergence to obtain the performance indicators of the multi-stage production line.
[0020] Furthermore, in step 1, the unreliable series-parallel production line system model is assumed as follows:
[0021] In the same processing stage, all types of products share a buffer, assuming that the buffer will not fail; the transfer time of the workpiece between the workstation and the buffer is not considered;
[0022] All machine reliability models follow exponential distribution. For all types of products, the workstation WSi The service rate, failure rate, and repair rate of the machine are μ i,j ,p i,j ,r i,j , i=1,2,…,m,j=1,2,…,S i ;
[0023] At a certain moment, if a machine M i,j The workpiece has been processed and the upstream buffer B i-1 At the previous moment, the state is empty, so machine M i,j Being hungry at the present moment;
[0024] At a certain moment, if a machine M i In working state, downstream buffer B i It was already full at the previous moment, and the machine M i+1 No workpiece will be extracted from the buffer, then machine M i is blocked at the current moment;
[0025] The first station will never be hungry, and the last station will never be blocked;
[0026] Assume that the failure mode of the machine is operation-related failure, that is, the machine will fail only when it is in working state, and will not fail in hungry or blocked state;
[0027] Assume that the blocking mechanism obeys the pre-service blocking, that is, within a period of time, machine M i,j If the machine M is in working state, and its downstream buffer is full at the end of the previous period, and the downstream workstation cannot extract the workpiece from the box for processing at the beginning of the period, then the machine M is considered to be i,j During this period, the machine is blocked and cannot process the workpiece.
[0028] Furthermore, in step 1, the performance indicator is defined as:
[0029] Throughput TP: The average number of workpieces produced per unit time by the last machine or workstation when the system is running in steady state;
[0030] Work-in-process inventory level WIP: the average total occupancy of all buffers when the system is running in a steady state;
[0031] Blocking rate BL i : The probability that machine i is in working state, the downstream buffer is full and the downstream machine cannot extract the workpiece under the steady-state operation of the system;
[0032] Hunger rate ST i : The probability that the upstream buffer is empty when machine i takes a workpiece under steady-state operation of the system.
[0033] Furthermore, in step 2
[0034]
[0035] in machine The service rate, μ i is the service rate of the i-th machine among the S parallel machines, T down,i is the failure time of the i-th machine among the S parallel machines, T up,i is the working time of the i-th machine among the S parallel machines.
[0036] Furthermore, in step 3, the specific process of performing the two-stage production line performance evaluation considering the setup time is as follows:
[0037] Step 3.1: System state modeling:
[0038] The system state is represented by a ternary vector (n, α, β) constructed by the WIP inventory level n, the upstream machine state α, and the downstream machine state β. According to the different WIP inventory levels in the system, the system state is divided into upper boundary state, internal state, and lower boundary state sets.
[0039] Step 3.2: System state transition analysis:
[0040] Based on the system state transition probability analysis, the state transition equation of the two-stage series-parallel production line is established to characterize the transformation relationship between the probabilities of the system being in each state;
[0041] Step 3.3: Steady-state probability solution:
[0042] The system state transfer equation and the normalized equation are combined to construct an equation system, and the equation system is solved using Cramer's rule to obtain the probability of being in each state under steady-state conditions as the steady-state probability distribution of the system;
[0043] Step 3.4: Calculation of performance indicators of the two-stage production system:
[0044] According to the steady-state probability distribution of the system, the performance indicators are solved: throughput rate, work-in-process inventory level, starvation rate and blocking rate.
[0045] Furthermore, in step 3.1, the upstream machine state α is
[0046]
[0047] The downstream machine state β is
[0048]
[0049] Define a judgment function for α and β respectively:
[0050]
[0051]
[0052] Flag is the state identification variable. In the Val(·) function, if the value of the machine state variable (α, β) is the same as the Flag value, it is true and the function return value is 1. Otherwise, the function return value is 0. Under steady-state conditions, any state of the series-parallel production line system corresponds to a non-negative probability number p(n, α, β), which is between 0 and 1, representing the probability that the system is in its corresponding state.
[0053] The entire system state space is represented as Φ = {(n, α, β)}; where Φ can be divided into three sub-state sets, namely the upper boundary state, the internal state and the lower boundary state, denoted as Φ = Φ L ∪Φ I ∪Φ U :
[0054] (1) Lower boundary state set Φ L :When n=0, the possible states of the upstream machine are working, faulty, and set, that is, α=0,1,2; the possible states of the downstream machine are working, faulty, and starving, that is, β=0,1,3. Therefore, there are T L =3·3·1=9 states;
[0055] (2) Internal state set Φ I :When 0<n<B, the possible states of the upstream machine are working, faulty, and set, that is, α=0,1,2; the possible states of the downstream machine are working, faulty, and set, that is, β=0,1,2. Therefore, T I =3·3·(B-1)=9(B-1) states;
[0056] (3) Upper boundary state set Φ U :When n=B, the state of the upstream machine can only be blocked, that is, α=3; the state of the downstream machine may be working, faulty, or set, that is, β=0,1,2. Therefore, T L =1·3·1 states;
[0057] In summary, the total number of states in the system state space Φ is T = T L +T I +T U =9B+3.
[0058] Furthermore, in step 3.2, the system state is expressed as follows:
[0059]
[0060] The system state transition probability is analyzed as follows:
[0061] (1) Leaving the current state due to a failure of the upstream machine, that is, changing from state (n, 1, β) to state (n, 0, β), with a transition rate of p1;
[0062] (2) Leaving the current state due to a failure of the downstream machine, that is, changing from state (n, α, 1) to state (n, α, 0), with a transition rate of p2;
[0063] (3) Leaving the current state due to the completion of repair by the upstream machine, that is, changing from state (n, 0, β) to state (n, 1, β), with a transfer rate of r1;
[0064] (4) The downstream machine completes the repair and leaves the current state, that is, changes from state (n, α, 0) to state (n, α, 1), with a transfer rate of r2;
[0065] (5) Leaving the current state due to the completion of the setup by the upstream machine, that is, the transition rate from state (n, 2, β) to state (n, 1, β) is
[0066] (6) Leaving the current state due to the completion of the setup by the downstream machine, that is, the state (n, α, 2) becomes (n-1, α, 1):
[0067] 1) When n≠B: the transfer rate is
[0068] 2) When n = B:
[0069] ①The upstream machine enters the working state, and the transfer rate is
[0070] ②The upstream machine enters the setup state, and the transfer rate is
[0071] (7) When the upstream machine completes the processing and leaves the current state:
[0072] 1) When 0<n<B-1, the current state is (n,1,β):
[0073] ① The upstream machine enters the working state, that is, enters the state (n+1,1,β), and the transfer rate is
[0074] ② The upstream machine enters the setup state, that is, enters the state (n+1,2,β), and the transfer rate is
[0075] 2) When n = 0, β = 0, 1, 3, that is, the current state is (0, 1, β):
[0076] When the current state is (0,1,3),
[0077] ① The upstream machine enters the working state, and the downstream machine also enters the working state, that is, enters the state (0,1,1), and the transfer rate is
[0078] ② The upstream machine enters the setup state, and the downstream machine enters the working state, that is, enters the state (0,2,1), and the transfer rate is
[0079] ③ The upstream machine enters the working state, and the downstream machine enters the setting state, that is, enters the state (1,1,2), and the transfer rate is
[0080] ④ The upstream machine enters the setup state, and the downstream machine also enters the setup state, that is, enters the state (1,2,2), and the transfer rate is
[0081] The current state is (0,1,β), β=0,1,
[0082] ① The upstream machine starts working, and the downstream machine state remains unchanged, that is, the system enters the state (1,1,β), and the transfer rate is
[0083] ② The upstream machine enters the setting, and the downstream machine state remains unchanged, that is, the system enters the state (1,2,β), and the transfer rate is
[0084] 3) When n = B-1, the current state is (B-1, 1, β):
[0085] ① The upstream machine enters the blocking state, that is, enters the state (B, 3, β), and the transfer rate is μ1;
[0086] (8) The situation where the downstream machine completes the processing and causes the departure:
[0087] 1) When 0<n<B, that is, the current state (n, α, 1):
[0088] ① The downstream machine enters the working state, that is, enters the state (n-1,α,1), and the transfer rate is
[0089] ② The downstream machine enters the setup state, that is, enters the state (n,α,2), and the transfer rate is
[0090] 2) When n = B, and the current state is (B, 3, 1):
[0091] ① The upstream machine enters the working state, and the downstream machine also enters the working state, that is, (B-1,1,1), and the transfer rate is
[0092] ② The upstream machine enters the setup state, and the downstream machine enters the working state, that is, (B-1,2,1), and the transfer rate is
[0093]
[0094] ③The upstream machine remains in the blocking state, and the downstream machine enters the setting state, that is, enters the state (B,3,2), and the transfer rate is
[0095] 3) When n = 0, that is, the current state (0, α, 1):
[0096] ① The downstream machine enters the starvation state, that is, enters the state (0, α, 3), and the transfer rate is μ2;
[0097] Based on the above system state transition probability analysis, 9B+3 state transition equations are established respectively; using P ij Indicates that the system changes from the jth state (n j ,α j ,β j ) transfers to the i-th state (n i ,α i ,β i ), P(i) indicates the transition probability of the system in state (n i ,α i ,β i ) probability.
[0098] Furthermore, in step 3.4, the performance index calculation expression of the two-stage production system is:
[0099] Based on the system steady-state distribution vector p(n,α,β), the analytical expression of the two-stage system performance index is given as follows:
[0100] (1) Throughput TP:
[0101]
[0102] (2) Work in process inventory level WIP:
[0103]
[0104] (3) Blocking rate BL:
[0105]
[0106] (4) Hunger rate ST:
[0107]
[0108] Furthermore, in step 4, the specific process of multi-stage production line performance evaluation considering setup time is as follows:
[0109] Step 4.1: Backward polymerization: Starting from the last two stations, perform polymerization from back to front, and establish the backward polymerization iteration equation of the multi-station production line;
[0110] Step 4.2: Forward polymerization: Starting from the initial two stations, polymerization is performed from front to back, and the forward polymerization iteration equation of the multi-station production line is established;
[0111] Step 4.3: Aggregation iteration: Repeat steps 4.1 and 4.2 until the iterative process results converge;
[0112] Step 4.4: Calculation of performance indicators of multi-stage production system: Call the results of the two-stage series-parallel production line model, solve the iterative equations, and calculate the performance indicators of the multi-stage production line.
[0113] Beneficial Effects
[0114] The performance evaluation method of unreliable series-parallel production lines considering the setup time proposed in this invention is analyzed through numerical experiments to prove the effectiveness of the method proposed in this invention in practical applications. Specifically, when the number of workstations is 4, the buffer capacity is 3 to 5 pieces, the setup time is 0.5min, the product types are 3, the processing rate is [0.8, 1], the parallel machine failure probability is [0.1, 0.3], and the parallel machine failure repair probability is [0.8, 0.95], 5 series-parallel production lines with different parameter structure types are set, including horizontal type, rising type, falling type, positive V type and inverted V type. The applicability of the proposed method to different types of series-parallel production lines is verified through numerical experimental analysis. Considering the randomness of the experimental results, 10 independent repeated simulation experiments were carried out under different experimental parameters, and the experimental results were averaged. The experimental results of the multi-stage performance evaluation model are shown in Table 1.
[0115] Table 1 Experimental results of multi-stage performance evaluation model
[0116]
[0117]
[0118] The experimental results show that the theoretical results of the multi-stage approximate evaluation model proposed in this invention have a small deviation from the simulation results and a high degree of agreement, which verifies the effectiveness of the multi-stage performance evaluation model based on the aggregation method. Specifically, for the throughput index, the deviation between the theoretical results of the five structural production lines and the simulation results obtained by numerical experiments does not exceed 1.814%; for the work-in-process inventory level index, the deviation does not exceed 3.962%; for the blocking rate index, the deviation does not exceed 6.452%; for the starvation rate index, the deviation does not exceed 6.25%.
[0119] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0120] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:
[0121] Figure 1 Consider unreliable series-parallel production lines with setup times.
[0122] Figure 2 Schematic diagram of parallel machine merging.
[0123] Figure 3 Schematic diagram of a two-stage series-parallel production line.
[0124] Figure 4 Schematic diagram of the two-stage production line after the machines are merged.
[0125] Figure 5 Schematic diagram of a multi-stage production line after the machines are merged.
[0126] Figure 6 Schematic diagram of the state transition diagram of a two-stage system.
[0127] Figure 7 Schematic diagram of backward polymerization.
[0128] Figure 8 Schematic diagram of the forward polymerization.
[0129] Fig. 9 Schematic diagram of the impact of the number of parallel machines on system performance indicators.
[0130] Fig.10 Schematic diagram of the impact of buffer capacity on system performance indicators.
[0131] Fig.11 Schematic diagram of the effect of setup time on system performance indicators.
[0132] Fig.12 Schematic diagram of the impact of product types on system performance indicators. DETAILED DESCRIPTION
[0133] Embodiments of the present invention are described in detail below. The embodiments are exemplary and intended to be used to explain the present invention, but should not be construed as limiting the present invention.
[0134] The present invention is aimed at unreliable series-parallel production lines, such as Figure 1 As shown in the figure, a performance evaluation method for unreliable series-parallel production lines considering setup time is provided, which mainly solves the following problems:
[0135] (1) Merging and parameter calculation of parallel machines in series-parallel production lines
[0136] In actual production, a production line has a process that consists of multiple parallel machines. That is, when the processing capacity of a single machine is insufficient, a multiple parallel production method is usually adopted, which makes it difficult to perform performance analysis on production lines with parallel machine structures. In order to effectively simplify the difficulty of solving the performance indicators of the series-parallel production system, one research idea is to merge the parallel machines in the series-parallel production line to make them an equivalent machine, and calculate the parameters of the merged new machine so that the system performance remains unchanged before and after the merger. By simplifying the production system with a complex structure into a standard serial production line, it is easy to analyze and improve it. Therefore, the merging of parallel machines and parameter calculation is the first technical problem to be solved by the present invention.
[0137] (2) Constructing an analytical model for performance evaluation of a two-stage series-parallel production line
[0138] For the two-stage series-parallel production line, an analytical method is used to model it. On the basis of the merging of parallel machines and parameter calculation, the system state transition process is characterized based on the Markov chain, the system state transition equation is established, the system steady-state probability distribution is solved, and then the various performance indicators of the system are calculated, laying the foundation for the performance evaluation research of multi-stage series-parallel production lines. Therefore, establishing an analytical model for the two-stage series-parallel production line considering the setup time is the second technical problem to be solved by the present invention.
[0139] (3) Constructing an approximate model for performance evaluation of multi-stage series-parallel production lines
[0140] For multi-stage series-parallel production lines, based on two-stage building blocks, a multi-stage aggregation iterative equation is constructed based on the aggregation method. Through repeated iterations of backward aggregation and forward aggregation processes, an approximate solution of the system performance index is obtained. By designing simulation experiments, the approximate solution obtained by the aggregation method is compared with the simulation solution obtained by the simulation experiment to verify the effectiveness of the approximate model for performance evaluation of multi-stage series-parallel production lines. On this basis, the influence of system parameters such as the number of parallel machines, buffer capacity, setup time, and product types on system performance indicators is explored. Therefore, establishing an aggregation model for multi-stage series-parallel production lines is the third technical problem to be solved by the present invention.
[0141] In summary, the present invention takes unreliable series-parallel production lines as the research object, and proposes a performance evaluation method for unreliable series-parallel production lines that takes setup time into consideration by considering factors such as setup time, random machine failures, limited cache capacity, and diverse product types. That is, by analyzing the relationship between system transfer states, a two-stage Markov model is established to solve performance indicators such as throughput, work-in-process inventory level, hunger rate, and blocking rate. On this basis, multi-stage series-parallel production lines are iteratively merged through an aggregation method to obtain an approximate solution to the system performance indicators, thereby achieving the purpose of performance evaluation of unreliable series-parallel production lines. The specific implementation is as follows:
[0142] 1. Description and model assumptions of unreliable series-parallel production line system
[0143] System Description
[0144] Consider an unreliable series-parallel production line with parallel machining stations that can produce multiple products, such as Figure 1 As shown in the figure, the blank circle represents the processing machine. Multiple parallel processing machines in the same area constitute a workstation. The blank rectangle represents the buffer with limited capacity. The blank triangle represents the product to be processed. The machine needs to consider changing the mold when producing different types of products. Specifically, the unreliable series-parallel production line consists of m workstations and (m-1) buffers. The workpieces to be processed pass through the workstation WS in sequence. i Processing and buffer B i Storage and transfer from the last workstation WS m Outflow system.
[0145] Model Assumptions
[0146] (1) The production line has a total of m workstations {WS1, WS2, …, WS m}, each workstation contains S i Parallel unreliable machines Where i = 1, 2, ..., m;
[0147] (2) The series-parallel production line produces a total of K products, of which the types are 1, 2, 3, …, K. All products arrive randomly, and the probability of each type of product entering the system is equal;
[0148] (3) When the product currently processed by the machine is of a different type from the previous product, a setup operation is required before processing. Assume that the setup time for different products on different machines is the same, and the setup time is TS;
[0149] (4) The buffer between the two stations has a limited capacity, which is B i , i=1,2,…,m-1, the buffer work-in-process inventory level is represented by n.
[0150] (5) In the same processing stage, all types of products share a buffer, assuming that the buffer will not fail;
[0151] (6) The time it takes to move the workpiece between the workstation and the buffer zone is not considered;
[0152] (7) All machine reliability models follow an exponential distribution. For all types of products, the workstation WS i The service rate, failure rate, and repair rate of the machine are μ i,j ,p i,j ,r i,j , i=1,2,…,m,j=1,2,…,S i ;
[0153] (8) At a certain moment, if a machine M i,j The workpiece has been processed and the upstream buffer B i-1 At the previous moment, the state is empty, so machine M i,j Being hungry at the present moment;
[0154] (9) At a certain moment, if a machine M i In working state, downstream buffer B i It was already full at the previous moment, and the machine M i+1 No workpiece will be extracted from the buffer, then machine M i is blocked at the current moment;
[0155] (10) The first station will never be hungry, and the last station will never be blocked;
[0156] (11) Assume that the failure mode of the machine is operation-dependent failure (Operation Dependent Failure, OFD), that is, the machine will only fail when it is in working state, and will not fail when it is hungry or blocked;
[0157] (12) Assume that the blocking mechanism obeys the blocking before service (BBS), that is, within a period of time, machine M i,j If the machine M is in working state, and its downstream buffer is full at the end of the previous period, and the downstream workstation cannot extract the workpiece from the box for processing at the beginning of the period, then the machine M is considered to be i,j During this period, the machine is blocked and cannot process the workpiece.
[0158] (II) Definition of performance indicators of unreliable series-parallel production lines
[0159] When evaluating the performance of a production line, performance indicators such as the throughput rate, work-in-process inventory level, blocking rate, and starvation rate of the production system are usually observed. The specific definitions of each performance indicator are as follows:
[0160] (1) Throughput (TP): The average number of workpieces produced per unit time by the last machine or workstation when the system is running in steady state.
[0161] (2) Work-in-process (WIP) inventory level: the average total occupancy of all buffers when the system is running in steady state.
[0162] (3) Blockage rate (BL i ): The probability that machine i is in working state, downstream buffer i is full, and downstream machine i+1 cannot extract the workpiece under steady-state operation of the system.
[0163] (4) Starvation, ST i ): The probability that the upstream buffer i-1 is empty when machine i takes a workpiece under steady-state operation of the system.
[0164] (III) Parallel machine combination and parameter calculation of series-parallel production lines
[0165] Parallel machine merging
[0166] Since the system contains multiple machines, a large number of machines will lead to an exponential growth of the system state space, which is not conducive to analyzing the evolution law of the system state. In order to simplify the system, the present invention merges the parallel machines in the series-parallel production line to make them into an equivalent virtual machine. The new virtual machine can approximately reflect the impact of the real parallel machine system on the entire production process. By calculating the parameters of the merged new machine, the system performance before and after the merger remains unchanged. The parallel machine merger schematic diagram is shown in the figure. Figure 2 shown.
[0167] Specifically, merge S parallel machines into one equivalent virtual machine Its parameters are in is the average working time of the combined machines, is the combined machine failure working time, For the combined machine processing cycle.
[0168] Taking the two-stage series-parallel production line as an example, the parallel machine merging is introduced. Figure 3 The two-stage production line after merging the parallel machines is shown in Figure 4 As shown in the figure, the two merged virtual machines are recorded as M1 and M2 respectively. In order to obtain the analytical solution of the performance indicators of the two-stage production line, the following assumptions need to be made for the merged production system:
[0169] (1) If machine M i , at the moment of finishing the processing, buffer B f If it is empty, it will enter the hungry state;
[0170] (2) Assume that the upstream raw materials of M1 are in unlimited supply, that is, M1 will not starve; m The downstream storage space is unlimited, that is, M m No blocking;
[0171] (3) All machines except M1 adopt a first-come, first-served approach for workpieces in the buffer.
[0172] Parameter calculation
[0173] For all types of products, the average service rate, failure rate, and repair rate of machine M1 are μ M1 ,p M1 ,r M1 ; The service rate, failure rate, and repair rate of machine M2 are μ M2 ,p M2 ,r M2 In the following derivation, the combined machine is called The service rate of the combined machine is The average working time of the combined machines is The throughput rate is The average failure time of the combined machines is
[0174] machine Service Rate for:
[0175]
[0176] Merged Machine The calculation of the average working time and the average failure time needs to consider the following factors: When all S parallel machines are in working state, the service rate of the parallel system is equal to When some of the machines fail, the service rate of the parallel system will be lower than Based on this situation, it is assumed that the merged machine still has a service rate But the working hours are appropriately shortened.
[0177] For example, when S=2, the system may have the following four states: two machines working at the same time; two machines failing at the same time, the first machine working and the second machine failing; the first machine failing and the second machine working. In this case, the average working time of the combined machines is The definition is as follows:
[0178]
[0179] In the above formula, the first term in the numerator represents the situation where the two machines are working at the same time, and the next two terms represent the situation where one machine is working and the other is faulty; the denominator represents the sum of the average working time and fault time of the two machines.
[0180] Let the throughput of the merged machine be the same as that before the merger to calculate the failure working time of the merged machine
[0181]
[0182] From this we can get:
[0183]
[0184] Extending the above reasoning to the case of S>2, it is not difficult to obtain the following formula:
[0185]
[0186] Before merging, all parallel machines are exponential reliability machines with parameter μ i ,p i ,r i , then the above formula can be rewritten as:
[0187]
[0188] In addition, the process of merging parallel machines in a multi-stage serial-parallel production line is similar to that in a two-stage production line, and will not be described in detail here. Figure 5 shown.
[0189] 4. Two-stage production line performance evaluation considering setup time
[0190] Considering that the two-stage production line is the simplest production line and the basis of the multi-stage production line aggregation and iteration model, the present invention first establishes a two-stage production line analytical model based on the Markov process, and calculates the system performance evaluation index by solving its steady-state probability distribution, providing a basis for the performance evaluation of multi-stage production lines.
[0191] System state modeling
[0192] Define α to represent the state of the upstream machine M1:
[0193]
[0194] Define β to represent the state of the downstream machine M2:
[0195]
[0196] The system state consists of three parts: the system WIP inventory level n, the machine state α and the machine state β. It can be represented by a three-element vector (n, α, β).
[0197] Define a judgment function for α and β respectively:
[0198]
[0199]
[0200] Flag is the state identification variable. In the Val(·) function, if the value of the machine state variable (α, β) is the same as the Flag value, it is true and the function return value is 1. Otherwise, the function return value is 0. Under steady-state conditions, any state of the system corresponds to a non-negative probability number p(n, α, β), which is between 0 and 1, representing the probability that the system is in its corresponding state.
[0201] The entire system state space is represented as Φ = {(n, α, β)}. Specifically, Φ can be divided into three sub-state sets, namely the upper boundary state, the internal state and the lower boundary state, denoted as Φ = Φ L ∪Φ I ∪Φ U :
[0202] (1) Lower boundary state set Φ L :When n=0, the possible states of the upstream machine are working, faulty, and set, that is, α=0,1,2; the possible states of the downstream machine are working, faulty, and starving, that is, β=0,1,3. Therefore, there are T L =3·3·1=9 states.
[0203] (2) Internal state set Φ I:When 0<n<B, the possible states of the upstream machine are working, faulty, and set, that is, α=0,1,2; the possible states of the downstream machine are working, faulty, and set, that is, β=0,1,2. Therefore, T I =3·3·(B-1)=9(B-1) states.
[0204] (3) Upper boundary state set Φ U :When n=B, the state of the upstream machine can only be blocked, that is, α=3; the state of the downstream machine can be working, faulty, or set, that is, β=0,1,2. Therefore, T L =1·3·1 states.
[0205] In summary, the total number of states in the system state space Φ is T = T L +T I +T U =9B+3.
[0206] System state transition analysis
[0207] Represent the system state as follows:
[0208]
[0209] The system state vector matrix is expanded into one-dimensional vectors from large to small according to the buffer level, and numbered in sequence, as shown in Table 2:
[0210] Table 2 System status sorting
[0211]
[0212] The system state sorting algorithm 3SA is used to sort the system state, laying the foundation for calculating the state transition rate matrix.
[0213]
[0214]
[0215] The final generated system state space list contains states that do not conform to physical meaning, a total of 16 (B+1) states. If the unreasonable state is included in the Markov process, the rank of the state transition matrix is not equal to T-1, the solution of the system's equilibrium equation is not unique, and the Markov process cannot reach a steady state. This problem can be solved by making the transition rate of other states to unreasonable states 0. Since all machine services, failures and repairs follow independent exponential distributions, in a relatively short time δ t The probability of these events occurring simultaneously on two or more workstations is δ t It is an infinitesimal quantity, so it can be ignored when establishing the system equilibrium equation.
[0216] The two-stage system state transition diagram is as follows Figure 6 The system state transition diagram clearly shows the relationship between the machine state and the WIP inventory level n. The repair, failure, and setting of the machine will not change the WIP quantity of the system, but the completion of the machine processing will change the WIP quantity of the system. Figure 6 The solid line in the middle represents the active transition caused by the machine's own events, and the dotted line represents the passive transition caused by events occurring on other machines. Specifically, for the general state (n, α, β), the transition rate of leaving the state (n, α, β) includes the following 8 cases:
[0217] (1) Leaving the current state due to a failure of the upstream machine, that is, changing from state (n, 1, β) to state (n, 0, β), with a transition rate of p1;
[0218] (2) Leaving the current state due to a failure of the downstream machine, that is, changing from state (n, α, 1) to state (n, α, 0), with a transition rate of p2;
[0219] (3) Leaving the current state due to the completion of repair by the upstream machine, that is, changing from state (n, 0, β) to state (n, 1, β), with a transfer rate of r1;
[0220] (4) The downstream machine completes the repair and leaves the current state, that is, changes from state (n, α, 0) to state (n, α, 1), with a transfer rate of r2;
[0221] (5) Leaving the current state due to the completion of the setup by the upstream machine, that is, the transition rate from state (n, 2, β) to state (n, 1, β) is
[0222] (6) Leaving the current state due to the completion of the setup by the downstream machine, that is, the state (n, α, 2) becomes (n-1, α, 1):
[0223] 1) When n≠B: the transfer rate is
[0224] 2) When n = B:
[0225] ①The upstream machine enters the working state, and the transfer rate is
[0226] ②The upstream machine enters the setup state, and the transfer rate is
[0227] (7) When the upstream machine completes the processing and leaves the current state:
[0228] 1) When 0<n<B-1, the current state is (n,1,β):
[0229] ① The upstream machine enters the working state, that is, enters the state (n+1,1,β), and the transfer rate is
[0230] ② The upstream machine enters the setup state, that is, enters the state (n+1,2,β), and the transfer rate is
[0231] 2) When n = 0, β = 0, 1, 3, that is, the current state is (0, 1, β):
[0232] When the current state is (0,1,3),
[0233] ① The upstream machine enters the working state, and the downstream machine also enters the working state, that is, enters the state (0,1,1), and the transfer rate is
[0234] ② The upstream machine enters the setup state, and the downstream machine enters the working state, that is, enters the state (0,2,1), and the transfer rate is
[0235] ③ The upstream machine enters the working state, and the downstream machine enters the setting state, that is, enters the state (1,1,2), and the transfer rate is
[0236] ④ The upstream machine enters the setup state, and the downstream machine also enters the setup state, that is, enters the state (1,2,2), and the transfer rate is
[0237] The current state is (0,1,β), β=0,1,
[0238] ① The upstream machine starts working, and the downstream machine state remains unchanged, that is, the system enters the state (1,1,β), and the transfer rate is
[0239] ② The upstream machine enters the setting, and the downstream machine state remains unchanged, that is, the system enters the state (1,2,β), and the transfer rate is
[0240] 3) When n = B-1, the current state is (B-1, 1, β):
[0241] ① The upstream machine enters the blocking state, that is, enters the state (B, 3, β), and the transfer rate is μ1;
[0242] (8) The situation where the downstream machine completes the processing and causes the departure:
[0243] 1) When 0<n<B, that is, the current state (n, α, 1):
[0244] ① The downstream machine enters the working state, that is, enters the state (n-1,α,1), and the transfer rate is
[0245] ② The downstream machine enters the setup state, that is, enters the state (n,α,2), and the transfer rate is
[0246] 2) When n = B, and the current state is (B, 3, 1):
[0247] ① The upstream machine enters the working state, and the downstream machine also enters the working state, that is, (B-1,1,1), and the transfer rate is
[0248] ② The upstream machine enters the setup state, and the downstream machine enters the working state, that is, (B-1,2,1), and the transfer rate is
[0249]
[0250] ③The upstream machine remains in the blocking state, and the downstream machine enters the setting state, that is, enters the state (B,3,2), and the transfer rate is
[0251] 3) When n = 0, that is, the current state (0, α, 1):
[0252] ① The downstream machine enters the starvation state, that is, enters the state (0, α, 3), and the transfer rate is μ2.
[0253] Based on the above system state transition probability analysis, 9B+3 state transition equations are established respectively. ij Indicates that the system changes from the jth state (n j ,α j ,β j ) transfers to the i-th state (n i ,α i ,β i ), P(i) indicates the transition probability of the system in state (n i ,α i ,β i ), and so on.
[0254] Steady-state probability solution
[0255] Steady-state distribution refers to the state distribution when the probability distribution of each state reaches stability after the Markov process runs for a long time. The matrix A is defined as the state transition probability matrix of the system. ij represents the element in the i-th row and j-th column of matrix A, and its corresponding value is the transition probability P from system state i to system state j ij When generating the transition probability matrix, traverse the 1st to Tth states of the system. Given the i-th state of the system, the transition probability of the corresponding state can be obtained according to the transition relationship given above.
[0256] The present invention constructs 9B+3 state transfer equations and 1 normalized formula to form a 9B+4-dimensional equation group, which contains 9B+3 unknown quantities. (9B+3)×1 represents the system state probability matrix, A (9B+4)×(9B+3) represents the transfer equation coefficient matrix, b (9B+4)×1 Represents the constant term, and the system state transfer equations can be transformed into the following form:
[0257] Ap=b (9)
[0258] Specifically,
[0259]
[0260] By solving the equilibrium equations (9) using Cramer's rule, we can obtain the state probability matrix p of the system when it is running stably.
[0261] Calculation of performance indicators for a two-stage production system
[0262] Based on the system steady-state distribution vector p(n,α,β), the analytical expression of the two-stage system performance index is given as follows:
[0263] (1) Throughput (TP) refers to the average number of workpieces produced per unit time by the last machine when the system is running in steady state:
[0264]
[0265] (2) Work-in-process inventory level (WIP) refers to the amount of inventory in buffer B when the system is running in a steady state. f Number of workpieces in:
[0266]
[0267] (3) The blocking rate (BL) refers to the probability that a machine is in a blocked state when the system is running in a steady state. Since the downstream machine will not be blocked, the blocking rate here refers to the blocking rate of the upstream machine M1:
[0268]
[0269] (4) The starvation rate (ST) refers to the probability that a machine is in a starvation state when the system is running in a steady state. Since the upstream machine will not be hungry, the starvation rate here refers to the starvation rate of the downstream machine M2:
[0270]
[0271] 5. Multi-stage production line performance evaluation considering setup time
[0272] For multi-stage production lines with setup times, e.g. Figure 1 As shown in the figure, an aggregation method is proposed to solve the system performance indicators. The aggregation method uses the two-stage production line analytical model to aggregate the two stations into a single equivalent station, and recursively iterates through the aggregation process until convergence to obtain an approximate solution for the system's productivity and other performance indicators. The aggregation process consists of two parts: backward aggregation and forward aggregation. The details are as follows:
[0273] Backward Aggregation
[0274] For a serial production line consisting of m machines, the last two machines M m-1 and M m Aggregate into an index machine The superscript b indicates backward aggregation. The service rate of the virtual machine after aggregation is equal to the throughput rate of the aggregated two-stage system and is obtained by the first equation in formula (10); the repair rate It is necessary to consider the impact of the upstream machine blocking rate in the two-stage system, that is, to appropriately extend the "virtual mean failure time" so that it can include the real failure time and the upstream machine blocking time. Its value can be expressed by multiplying the upstream machine repair rate by a correction determined by the blocking rate; the failure rate It can be directly expressed as the failure rate of the upstream machine, because the impact of the blocking rate has been corrected. m-2 With virtual machines Aggregate to get a new virtual machine Repeat the above aggregation process until all machines in the original system are aggregated into At this point, the backward polymerization process is completed. Figure 7 shown.
[0275]
[0276] in, are the steady-state probability values of the throughput rate and blocking rate of the two-stage production line system, which are obtained from the analytical model for performance evaluation of the two-stage series-parallel production line in Parts 3 and 4. is the machine efficiency, and its calculation formula is shown in (19):
[0277]
[0278] Forward Aggregation
[0279] Combine the first machine M1 in the system with the virtual machine obtained in the backward aggregation process Aggregate to get The superscript f indicates forward aggregation. The service rate of this machine It can be considered to be equal to the throughput rate of the aggregated two-stage system and is obtained by the second formula in formula (10); the repair rate It is necessary to consider the impact of the starvation rate of the downstream machines in the two-stage system, that is, to appropriately extend the "virtual mean failure time" so that it can include the real failure time and the starvation time of the downstream machines. Its value can be expressed by multiplying the downstream machine repair rate by a correction amount determined by the starvation rate; the failure rate can be directly expressed as the failure rate of the downstream machine, since the effect of the starvation rate has been corrected. and Aggregation Repeat the above polymerization process until At this point, the forward polymerization process is completed. Figure 8 shown.
[0280]
[0281] in, are the steady-state probability values of the throughput rate and blocking rate of the two-stage production line system, which are obtained from the analytical model for performance evaluation of the two-stage series-parallel production line in Parts 3 and 4. is the machine efficiency, and its calculation formula is shown in (25):
[0282]
[0283] Aggregate Iteration
[0284] The two processes of backward aggregation and forward aggregation are iterated alternately until convergence to obtain an approximate estimate of the performance indicators of the multi-stage production line. The initial conditions and boundary conditions of the iterative process are as follows:
[0285] (1) Initial conditions:
[0286]
[0287]
[0288] (2) Boundary conditions:
[0289]
[0290] Calculation of performance indicators for multi-stage production systems
[0291] When the aggregation iteration process ends, the system throughput of the multi-stage series-parallel production line is the virtual workstation obtained after aggregation. The throughput rate, the work-in-process inventory level is the average level of the number of workpieces in each buffer after iteration, the blocking rate is the probability that all stations are in working state, the downstream buffer is full, and the downstream machine cannot retrieve the workpiece, and the starvation rate is the probability that the upstream buffer is empty when all stations take the workpiece.
[0292] (1) System throughput
[0293]
[0294] (2) Work-in-process inventory level
[0295]
[0296] The total work-in-process inventory level of the system is:
[0297]
[0298] (3) Blockage rate
[0299]
[0300] (4) Hunger rate
[0301]
[0302] (VI) Numerical Experimental Analysis
[0303] For unreliable series-parallel production lines, the following compares the analytical results with the simulation results to verify the effectiveness and accuracy of the performance evaluation model for unreliable series-parallel production lines. The numerical experiment steps specifically include three parts: numerical experiment environment configuration, performance evaluation accuracy analysis, and parameter impact analysis on performance indicators.
[0304] Numerical experiment environment configuration
[0305] (1) Operating system: Windows 10 Professional (64-bit);
[0306] (2) Processor: Intel(R) Core(TM) i5-10875H CPU @ 2.90 GHz, 6 cores and 12 threads;
[0307] (3) System memory: 16.0 GB;
[0308] (4) Hard disk space: 1TB mechanical hard disk;
[0309] (5) Development tools: PyCharm 2020.2.1;
[0310] (6) Programming language: Python 3.11.
[0311] Performance Evaluation Accuracy Analysis
[0312] (1) Two-stage performance evaluation model
[0313] Three groups of experiments were designed here to verify the accuracy of the performance evaluation model under various production situations. The first group had the same machines at two stations; the second group had different machines at two stations but the same machines within the same station; and the third group had different machines at two stations. The production line parameters are shown in Table 3.
[0314] Table 3 Production line parameters
[0315]
[0316]
[0317] Table 4 Experimental results of the two-stage performance evaluation model
[0318]
[0319]
[0320] The experimental results of the two-stage performance evaluation model are shown in Table 4. The calculation results of the two-stage theoretical model proposed in the present invention have a small deviation from the results obtained by the numerical experiment, and the degree of agreement is high, which verifies the effectiveness and accuracy of the two-stage model. Specifically, for throughput rate, the deviation between the results obtained by the two-stage model and the numerical experiment does not exceed 1.360%; for work-in-process inventory level, the deviation does not exceed 1.431%; for blocking rate, the deviation does not exceed 3.333%; for starvation rate, the deviation does not exceed 3.226%.
[0321] (2) Multi-stage performance evaluation model
[0322] The performance indicators considered in this invention are system throughput TP, average work-in-progress inventory level WIP of each buffer zone, and i , blocking rate BL i and hunger rate ST i Specifically, based on the simulation experiment results, the error percentage of the performance index obtained by the aggregation method is calculated as an indicator to evaluate the effectiveness of the aggregation iteration model. The error percentage calculation formula of the performance index obtained by the aggregation method and the simulation experiment is as follows:
[0323]
[0324] Among them, error refers to the performance indicator error, Analysis refers to the approximate solution of the aggregation method, and Simulation refers to the approximate solution of the simulation method.
[0325] Five series-parallel production lines with different parameter structure types are set here, including horizontal type, rising type, falling type, positive V type and inverted V type, to verify the effectiveness and accuracy of the polymerization method proposed in the present invention under different conditions. To ensure accuracy, 10 independent repeated simulation experiments were carried out under different experimental parameters, each simulation run 40,000 time periods, the first 4,000 time periods were preheating periods, and the system performance indicators were statistically analyzed in the last 36,000 time periods. The experimental results took the average value of the 10 running results. The specific parameters of the production line are shown in Table 5, and the experimental results of the multi-stage performance evaluation model are shown in Table 1.
[0326] Table 5 Production line parameters
[0327]
[0328]
[0329] The steps of the simulation experiment are as follows:
[0330] 1) Initialize system parameters such as product type, setup time, machine status, initial work-in-process inventory level in the buffer, etc.
[0331] 2) At the beginning of each simulation unit step time, first determine the state of the workstation at the previous moment. If the workstation M i If the station M is in a starving or blocked state at the previous moment, the state will not change at this moment. i If the previous moment was neither hungry nor blocked, then at this moment The probability of normal operation is The probability of failure is small, the upstream raw material supply is unlimited, and it needs to be set up when the type of processed product changes;
[0332] 3) Update the machine status and buffer status information of upstream and downstream workstations in real time;
[0333] 4) When the workpiece completes all processing operations, it finally flows out of the system;
[0334] 5) Repeat steps 2)-4) until the simulation time ends, and calculate the system performance index results of this simulation.
[0335] By comparing the experimental results, it can be seen that the calculation results of the multi-stage theoretical model proposed in the present invention have small deviations from the simulation experimental results and a high degree of consistency, which verifies the effectiveness and accuracy of the multi-stage model. Therefore, using this method in actual production can effectively evaluate the steady-state performance of the production line.
[0336] Analysis of the impact of parameters on performance indicators
[0337] In order to further explore the performance of the present invention in different scenarios during the actual production process, the present invention designs numerical experiments on system parameters such as the number of parallel machines, buffer capacity, setup time, and product types for a multi-stage production system, explores the influence of the parameters on system performance indicators, and provides decision support for production managers.
[0338] (1) Number of parallel machines
[0339] The present invention sets up three five-station four-buffer production lines to explore the effect of the number of parallel machines on the performance indicators of different production lines. The number of parallel machines is set to [1, 2, 3, 4, 5, 6, 7, 8]. All machine parameters of production line 1 are the same, the service rate of production line 2 decreases, the service rate of production line 3 increases, and production line 2 and production line 3 are inverse systems. The specific parameters of the three production lines are shown in Table 6.
[0340] Table 6 Production line parameters
[0341]
[0342] 1) Impact of the number of parallel machines on throughput
[0343] The impact of the number of parallel machines on the throughput of the three production lines is as follows: Fig. 9 As shown in (a), the experimental results show that as the number of parallel machines increases, the TP of the three production lines increases linearly. The TP value and growth rate of production line 1 are higher than those of production lines 2 and 3. The TP curves of production lines 2 and 3 coincide, indicating that the arrangement of workstations in the production line has no effect on TP, that is, the inverse system has the same system performance.
[0344] 2) The impact of the number of parallel machines on the WIP inventory level in each buffer zone
[0345] The impact of the number of parallel machines on the work-in-process inventory level of the three production lines is as follows: Fig. 9 As shown in (b) to (d), the experimental results show that the order of WIP size of each buffer of the three production lines is WIP1>WIP2>WIP3>WIP4, indicating that the upstream buffer needs a larger buffer capacity. As the number of parallel machines increases, the WIP of each buffer of production line 1 changes slightly, the values of WIP1-WIP3 show an upward trend, and WIP4 remains basically unchanged; the WIP of each buffer of production line 2 shows an upward trend; the WIP of each buffer of production line 3 shows a downward trend. It is worth noting that the WIP curves of each station of production lines 2 and 3 are "symmetrical", indicating that the position of different stations in the system will affect the level of work-in-process inventory.
[0346] (2) Buffer capacity
[0347] The present invention sets up three five-station four-buffer production lines to explore the influence of buffer capacity on the performance indicators of different production lines. The buffer capacities are set to [1, 2, 3, 4, 5, 6, 7, 8] respectively. All machine parameters of production line 1 are the same, the service rate of production line 2 decreases, and the service rate of production line 3 increases. All buffer capacities are the same and change at the same time. Production lines 2 and 3 are inverse systems. The specific parameters of the three production lines are shown in Table 7.
[0348] Table 7 Production line parameters
[0349]
[0350] 1) Impact of buffer capacity on throughput
[0351] The impact of buffer capacity on the throughput of the three production lines is as follows: Fig.10 As shown in (a), the experimental results show that as the buffer capacity increases, the TP of the three production lines increases, and the rate of increase shows a downward trend. The TP of production line 1 is always higher than that of production lines 2 and 3, and the TP curves of production lines 2 and 3 basically overlap, indicating that the position of different workstations in the assembly line has no effect on TP, that is, the inverse system has the same system performance.
[0352] 2) The impact of buffer capacity on the WIP inventory level of each buffer
[0353] The impact of buffer capacity on the work-in-process inventory levels of the three production lines is as follows: Fig.10 As shown in (b) to (d), the experimental results show that the WIP of each buffer of production line 1 is linearly related to the buffer capacity, and the slope is WIP1>WIP2>WIP3>WIP4. The WIP of the first three buffers of production line 2 is also linearly related to the buffer capacity, and WIP4 is on an overall upward trend with a gradually increasing slope. The WIP of each buffer of production line 3 is on an overall upward trend with a gradually decreasing slope. The WIP size relationship of each buffer of the three production lines is WIP1>WIP2>WIP3>WIP4. The influence of the number of parallel machines on the work-in-process inventory level of each buffer is the same.
[0354] (3) Set the time
[0355] The present invention sets up three five-station four-buffer production lines to explore the effect of setup time on the performance indicators of different production lines. The setup time is set to [0.2, 0.4, 0.6, 0.8, 1, 1.2, 1.4, 1.6, 1.8, 2]. All machine parameters of production line 1 are the same, the service rate of production line 2 decreases, and the service rate of production line 3 increases. The setup time of all products on all machines is the same and changes at the same time, and production line 2 and production line 3 are inverse systems. The specific parameters of the three production lines are shown in Table 8.
[0356] Table 8 Production line parameters
[0357]
[0358] 1) The impact of setup time on throughput
[0359] The effect of setup time on the throughput of the three production lines is shown in Fig.11 As shown in (a), the experimental results show that the TP of the three production lines decreases with the increase of TS, and the rate of decrease gradually slows down. The TP of production line 1 is always greater than that of production lines 2 and 3, and the TP curves of production lines 2 and 3 basically overlap, indicating that the position of the workstation with different service rate parameters in the system does not affect the TP index, that is, the inverse system has the same system performance.
[0360] 2) The impact of setup time on WIP inventory levels in each buffer zone
[0361] The effect of setup time on the WIP inventory levels of the three production lines is as follows: Fig.11 As shown in (b) to (d), the experimental results show that as TS increases, the WIP of the first three buffers in production lines 1 and 2 all show a linear decreasing trend, WIP4 first increases and then decreases, and the overall change is small; for production line 3, WIP1 first decreases and then increases, and the overall change is small, and WIP2 to WIP4 all show a linear increasing trend. The WIP size relationship of each buffer of the three production lines is WIP1>WIP2>WIP3>WIP4, which is the same as the influence of the number of parallel machines and buffer capacity on the work-in-process inventory level of each buffer, that is, the upstream buffer needs a larger buffer capacity.
[0362] (4) Product categories
[0363] The present invention sets up three production lines with five workstations and four buffer zones to explore the influence of product types on the performance indicators of different production lines. The product types are set as [2, 3, 4, 5, 6, 7, 8] respectively. All machine parameters of production line 1 are the same, the service rate of production line 2 decreases, and the service rate of production line 3 increases. The probability of occurrence of all products is the same, and production lines 2 and 3 are inverse systems. The specific parameters of the three production lines are shown in Table 9.
[0364] Table 9 Production line parameters
[0365]
[0366] 1) Impact of product types on throughput
[0367] The impact of product types on the throughput rates of the three production lines is as follows: Fig.12As shown in (a), the experimental results show that the TP of the three production lines all show a downward trend with the increase of K. Because there are many types of products, frequent setup operations will occur during the production process. Therefore, the number of workpieces produced will decrease in the same simulation cycle. The TP of production line 1 is always greater than that of production lines 2 and 3, and the TP of production lines 2 and 3 basically coincide, indicating that the position of workstations with different service rates in the production line does not affect the TP index, that is, the inverse system has the same system performance.
[0368] 2) The impact of product types on the WIP inventory level in each buffer zone
[0369] The impact of product types on the work-in-process inventory levels of the three production lines is as follows: Fig.12 As shown in (b) to (d), the experimental results show that as the number of product types increases, the WIP of each buffer of the three production lines remains basically unchanged. The reason may be that although the number of product types has increased, the setup time of different products on different machines is the same, and the actual processing speed of each station on the three production lines is affected in the same way, so that the WIP of each buffer remains relatively balanced. In the case of different product types, the arrangement of stations with different service rates in the system has an impact on WIP, that is, the overall WIP of the production line arranged in a decreasing manner is higher than that of the production line with the same service rate, and higher than that of the production line arranged in an increasing manner. The reason is that the upstream processing speed is fast, the downstream machine rarely suffers from starvation, and the downstream buffer is always saturated in most cases, so the overall work-in-progress level of the system is relatively the highest.
[0370] In summary, the present invention provides a performance evaluation method for unreliable series-parallel production lines. By analyzing the operating rules of unreliable series-parallel production lines and calculating system performance indicators, a performance evaluation tool is provided for the lean design and continuous improvement of actual series-parallel production lines. The accuracy and effectiveness of the present invention are verified through numerical experimental analysis, and decision support can be provided for production managers.
[0371] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention without departing from the principles and intent of the present invention.
Claims
1. A performance evaluation method for unreliable series-parallel production lines considering setup time, Features: The following steps are involved: Step 1: Establish the unreliable series-parallel production line system model and determine the unreliable series-parallel production line performance indicators: Unreliable series-parallel production line consists of m workstations {WS 1 ,WS 2 ,…,WS m } and (m-1) buffers B i Each workstation contains S i Parallel unreliable machines Where i = 1, 2, ..., m; the workpiece to be processed passes through the workstation WS in sequence i Processing and buffer B i Storage and transfer from the last workstation WS m Outflow system; The series-parallel production line produces a total of K products, of types 1, 2, 3, ..., K. All products arrive randomly, and the probability of each type of product entering the system is equal; When the product currently processed by the machine is of a different type from the previous product, setup operations are required before processing. Assume that the setup time for different products on different machines is the same, and the setup time is TS; The performance indicators adopt the throughput rate, work-in-process inventory level, blocking rate and / or starvation rate of the unreliable series-parallel production line; Step 2: Merge parallel machines in series and parallel production lines: Merge S parallel machines into one equivalent virtual machine Its parameters are in is the average working time of the combined machines, is the combined machine failure working time, For the combined machine Processing cycle; Step 3: Two-stage line performance evaluation considering setup time: A two-stage production line analytical model is established based on the Markov process, and the system performance evaluation index is calculated by solving its steady-state probability distribution; Step 4: Multi-stage production line performance evaluation considering setup time: Based on the analytical model of the two-stage production line, the two workstations are aggregated into a single equivalent workstation, and the aggregation process is recursively iterated until convergence to obtain the performance indicators of the multi-stage production line.
2. According to claim 1, a method for evaluating the performance of an unreliable series-parallel production line taking into account the setup time, Features: In step 1, the unreliable series-parallel production line system model is assumed as follows: In the same processing stage, all types of products share a buffer, assuming that the buffer will not fail; the transfer time of the workpiece between the workstation and the buffer is not considered; All machine reliability models follow exponential distribution. For all types of products, the workstation WS i The service rate, failure rate, and repair rate of the machine are μ i,j ,p i,j ,r i,j , i=1,2,…,m,j=1,2,…,S i ; At a certain moment, if a machine M i,j The workpiece has been processed and the upstream buffer B i-1 At the previous moment, the state is empty, so machine M i,j Being hungry at the present moment; At a certain moment, if a machine M i In working state, downstream buffer B i It was already full at the previous moment, and the machine M i+1 No workpiece will be extracted from the buffer, then machine M i is blocked at the current moment; The first station will never be hungry, and the last station will never be blocked; Assume that the failure mode of the machine is operation-related failure, that is, the machine will fail only when it is in working state, and will not fail in hungry or blocked state; Assume that the blocking mechanism obeys the pre-service blocking, that is, within a period of time, machine M i,j If the machine M is in working state, and its downstream buffer is full at the end of the previous period, and the downstream workstation cannot extract the workpiece from the box for processing at the beginning of the period, then the machine M is considered to be i,j During this period, the machine is blocked and cannot process the workpiece.
3. According to claim 1, a method for evaluating the performance of an unreliable series-parallel production line taking into account the setup time, Features: In step 1, the performance indicator is defined as: Throughput TP: The average number of workpieces produced per unit time by the last machine or workstation when the system is running in steady state; Work-in-process inventory level WIP: the average total occupancy of all buffers when the system is running in a steady state; Blocking rate BL i : The probability that machine i is in working state, the downstream buffer is full and the downstream machine cannot extract the workpiece under the steady-state operation of the system; Hunger rate ST i : The probability that the upstream buffer is empty when machine i takes a workpiece under steady-state operation of the system.
4. According to claim 1, a method for evaluating the performance of an unreliable series-parallel production line taking into account the setup time, Features: In step 2 in machine The service rate, μ i is the service rate of the i-th machine among the S parallel machines, T down,i is the failure time of the i-th machine among the S parallel machines, T up,i is the working time of the i-th machine among the S parallel machines.
5. According to claim 1, a method for evaluating the performance of an unreliable series-parallel production line taking into account the setup time, Features: In step 3, the specific process of conducting two-stage production line performance evaluation considering setup time is as follows: Step 3.1: System state modeling: The system state is represented by a ternary vector (n, α, β) constructed by the WIP inventory level n, the upstream machine state α, and the downstream machine state β. According to the different WIP inventory levels in the system, the system state is divided into upper boundary state, internal state, and lower boundary state sets. Step 3.2: System state transition analysis: Based on the system state transition probability analysis, the state transition equation of the two-stage series-parallel production line is established to characterize the transformation relationship between the probabilities of the system being in each state; Step 3.3: Steady-state probability solution: The system state transfer equation and the normalized equation are combined to construct an equation system, and the equation system is solved using Cramer's rule to obtain the probability of being in each state under steady-state conditions as the steady-state probability distribution of the system; Step 3.4: Calculation of performance indicators of the two-stage production system: According to the steady-state probability distribution of the system, the performance indicators are solved: throughput rate, work-in-process inventory level, starvation rate and blocking rate.
6. A method for evaluating the performance of an unreliable series-parallel production line taking into account the setup time according to claim 5, Features: In step 3.1, the upstream machine state α is The downstream machine state β is Define a judgment function for α and β respectively: Flag is the state identification variable. In the Val(·) function, if the value of the machine state variable (α, β) is the same as the Flag value, it is true and the function return value is 1. Otherwise, the function return value is 0. Under steady-state conditions, any state of the series-parallel production line system corresponds to a non-negative probability number p(n, α, β), which is between 0 and 1, representing the probability that the system is in its corresponding state. The entire system state space is represented as Φ = {(n, α, β)}; where Φ can be divided into three sub-state sets, namely the upper boundary state, the internal state and the lower boundary state, denoted as Φ = Φ L ∪Φ I ∪Φ U : (1) Lower boundary state set Φ L :When n=0, the possible states of the upstream machine are working, faulty, and set, that is, α=0,1,2; the possible states of the downstream machine are working, faulty, and starving, that is, β=0,1,3. Therefore, there are T L =3·3·1=9 states; (2) Internal state set Φ I :When 0<n<B, the possible states of the upstream machine are working, faulty, and set, that is, α=0,1,2; the possible states of the downstream machine are working, faulty, and set, that is, β=0,1,2. Therefore, T I =3·3·(B-1)=9(B-1) states; (3) Upper boundary state set Φ U : When n = B, the state of the upstream machine can only be blocked, that is, α = 3; The states of the downstream machines may be working, faulty, or set, i.e., β = 0, 1, 2. Therefore, T L =1·3·1 states; In summary, the total number of states in the system state space Φ is T = T L +T I +T U =9B+3.
7. A method for evaluating the performance of an unreliable series-parallel production line considering setup time according to claim 6, Features: In step 3.2, the system state is expressed as follows: The system state transition probability is analyzed as follows: (1) Leaving the current state due to a failure of the upstream machine, that is, changing from state (n, 1, β) to state (n, 0, β), with a transfer rate of p 1 ; (2) Leaving the current state due to a failure of the downstream machine, that is, changing from state (n, α, 1) to state (n, α, 0), with a transfer rate of p 2 ; (3) Leaving the current state due to the completion of repair by the upstream machine, that is, changing from state (n, 0, β) to state (n, 1, β), with a transfer rate of r 1 ; (4) The downstream machine completes the repair and leaves the current state, that is, changes from state (n, α, 0) to state (n, α, 1), with a transfer rate of r 2 ; (5) Leaving the current state due to the completion of the setup by the upstream machine, that is, the transition rate from state (n, 2, β) to state (n, 1, β) is (6) Leaving the current state due to the completion of the setup by the downstream machine, that is, the state (n, α, 2) becomes (n-1, α, 1): 1) When n≠B: the transfer rate is 2) When n = B: ①The upstream machine enters the working state, and the transfer rate is ②The upstream machine enters the setup state, and the transfer rate is (7) When the upstream machine completes the processing and leaves the current state: 1) When 0<n<B-1, the current state is (n,1,β): ① The upstream machine enters the working state, that is, enters the state (n+1,1,β), and the transfer rate is ② The upstream machine enters the setup state, that is, enters the state (n+1,2,β), and the transfer rate is 2) When n = 0, β = 0, 1, 3, that is, the current state is (0, 1, β): When the current state is (0,1,3), ① The upstream machine enters the working state, and the downstream machine also enters the working state, that is, enters the state (0,1,1), and the transfer rate is ② The upstream machine enters the setup state, and the downstream machine enters the working state, that is, enters the state (0,2,1), and the transfer rate is ③ The upstream machine enters the working state, and the downstream machine enters the setting state, that is, enters the state (1,1,2), and the transfer rate is ④ The upstream machine enters the setup state, and the downstream machine also enters the setup state, that is, enters the state (1,2,2), and the transfer rate is The current state is (0,1,β), β=0,1, ① The upstream machine starts working, and the downstream machine state remains unchanged, that is, the system enters the state (1,1,β), and the transfer rate is ② The upstream machine enters the setting, and the downstream machine state remains unchanged, that is, the system enters the state (1,2,β), and the transfer rate is 3) When n = B-1, the current state is (B-1, 1, β): ① The upstream machine enters the blocking state, that is, enters the state (B, 3, β), and the transfer rate is μ 1 ; (8) The situation where the downstream machine completes the processing and causes the departure: 1) When 0<n<B, that is, the current state (n, α, 1): ① The downstream machine enters the working state, that is, enters the state (n-1,α,1), and the transfer rate is ② The downstream machine enters the setup state, that is, enters the state (n,α,2), and the transfer rate is 2) When n = B, and the current state is (B, 3, 1): ① The upstream machine enters the working state, and the downstream machine also enters the working state, that is, (B-1,1,1), and the transfer rate is ② The upstream machine enters the setup state, and the downstream machine enters the working state, that is, (B-1,2,1), and the transfer rate is ③The upstream machine remains in the blocking state, and the downstream machine enters the setting state, that is, enters the state (B,3,2), and the transfer rate is 3) When n = 0, that is, the current state (0, α, 1): ① The downstream machine enters the starvation state, that is, enters the state (0, α, 3), and the transfer rate is μ 2 ; Based on the above system state transition probability analysis, 9B+3 state transition equations are established respectively; using P ij Indicates that the system changes from the jth state (n j ,α j ,β j ) transfers to the i-th state (n i ,α i ,β i ), P(i) indicates the transition probability of the system in state (n i ,α i ,β i ) probability.
8. A method for evaluating the performance of an unreliable series-parallel production line taking into account the setup time according to claim 7, Features: In step 3.4, the calculation expression of the performance index of the two-stage production system is: Based on the system steady-state distribution vector p(n,α,β), the analytical expression of the two-stage system performance index is given as follows: (1) Throughput TP: (2) Work in process inventory level WIP: (3) Blocking rate BL: (4) Hunger rate ST:
9. According to claim 1, a method for evaluating the performance of an unreliable series-parallel production line taking into account the setup time, Features: In step 4, the specific process of multi-stage production line performance evaluation considering setup time is as follows: Step 4.1: Backward polymerization: Starting from the last two stations, perform polymerization from back to front, and establish the backward polymerization iteration equation of the multi-station production line; Step 4.2: Forward polymerization: Starting from the initial two stations, polymerization is performed from front to back, and the forward polymerization iteration equation of the multi-station production line is established; Step 4.3: Aggregation iteration: Repeat steps 4.1 and 4.2 until the iterative process results converge; Step 4.4: Calculation of performance indicators of multi-stage production system: Call the results of the two-stage series-parallel production line model, solve the iterative equations, and calculate the performance indicators of the multi-stage production line.
