A production sequence recovery method for linear buffer zones with return lanes

Abstract

The present application relates to the technical field of mixed-model assembly line production planning and scheduling, in particular to a production sequence recovery method suitable for linear buffer zone with return lane, which comprises the following steps: 1, inputting upstream sequence of products, target downstream sequence, configuration parameters of linear buffer zone with return lane; 2, generating arrangement scheme of products in the buffer zone according to heuristic filling rule; 3, calculating and solving multiple product release sequences corresponding to the arrangement scheme by using ant colony optimization algorithm; 4, if the termination condition of the ant colony optimization algorithm is met, the optimal downstream release sequence is obtained; otherwise, the next iteration is performed in step 3. The present application has the advantages that the disrupted upstream production sequence can be recovered into a downstream sequence with maximum or approximate maximum sequencing rate compared with the target downstream sequence, and the downstream sequence has a smaller sequence offset degree.

Description

Technical Field

[0001] This invention relates to the field of production planning and scheduling technology for hybrid assembly lines, and in particular to a method for restoring production sequences applicable to linear buffers with return lanes. Background Technology

[0002] Hybrid assembly lines are low-cost, high-efficiency, flexible production systems that can continuously process and assemble different types of products with similar structures and processes on a single assembly line. Hybrid assembly lines have been widely used in industries such as automotive, electronics, home appliances, and furniture.

[0003] Taking a modern automotive OEM using a hybrid assembly production method as an example, products (vehicles) sequentially pass through the body shop, paint shop, and final assembly shop, each with different preferences and requirements regarding the production sequence. Since most assembly work is completed in the final assembly shop, the OEM typically determines the initial planned production sequence according to the final assembly shop's requirements, while simultaneously issuing parts requirements to suppliers in advance. However, in actual production, the OEM often proactively adjusts the production sequence from upstream shops based on the production needs of the shops along the route. Furthermore, various unexpected events and uncertainties during production (such as material shortages or machine malfunctions) can also cause changes in the position of products in the planned production sequence. In reality, by the time the production sequence arrives at the final assembly shop, the actual position of most products has already deviated. Therefore, to ensure assembly proceeds as planned and to reduce the negative impact of temporary sequence changes on manufacturers and suppliers, it is necessary to adjust the disrupted sequence before reaching the final assembly shop, restoring it as close as possible to the planned production sequence.

[0004] Sequence reordering on hybrid assembly lines typically requires the use of different types of buffers. Linear buffers are the most widely used reordering devices, offering advantages such as lower cost, smaller footprint, and more effective reordering compared to other reordering devices like Automated Access Systems (ASRS). Currently, most companies use traditional linear buffers with a simple structure, requiring products assigned to the same lane to strictly adhere to the "first-in, first-out" principle, thus resulting in insufficient sequence reordering flexibility and limited reordering capabilities. This limitation can be effectively overcome by improving the buffer's structure and control methods (i.e., sequence reordering methods).

[0005] Existing technologies already include descriptions and introductions of novel linear buffers with return lanes. These buffers utilize a return lane to return products temporarily unsuitable for release to the buffer entrance, allowing for the release of products following in the same lane and thus increasing the flexibility of sequence adjustment. However, existing technologies lack methods for sequence recovery applicable to this novel buffer. Summary of the Invention

[0006] The purpose of this invention is to address the problems existing in the background art by proposing a production sequence recovery method applicable to linear buffers with return lanes, which can adjust the scrambled upstream production sequence to obtain a downstream release sequence with the maximum ordering rate compared to the target downstream sequence.

[0007] The technical solution of the present invention, a method for restoring production sequences applicable to linear buffers with return lanes, includes the following specific steps:

[0008] S1. Obtain the total number T of products in the upstream sequence, and the position f of the product at position i in the target downstream sequence. i Where T≤L×W, i=1,2,…,T, L is the number of storage lanes in the linear buffer, and W is the number of vehicles in each storage lane;

[0009] S2. Based on the heuristic filling rules, assign a storage lane to each product i in the order of arrival in the buffer, and construct the product layout scheme A in the buffer.

[0010] S3. Use the ant colony optimization algorithm to obtain the release sequence of multiple downstream products corresponding to the arrangement scheme A;

[0011] S4. Determine if the termination condition of the ant colony algorithm is met; if not, jump to step S3 for the next iteration search; if met, the current optimal downstream sequence S... opt This is the desired optimal downstream recovery sequence.

[0012] Preferably, in S2, upstream product i is assigned to the storage lane by sequentially executing the following heuristic filling rules:

[0013] A.1 If i = 1, choose the first lane;

[0014] A.2 If there is an unfilled lane in the buffer, and the positions of product i and the last product j in this lane are f in the downstream sequence of the target. i and f j Satisfy f i >f j If the number of lanes is not full (1), select that lane; otherwise, select the lane with the most available parking spaces. i -f j The lane with the lowest value;

[0015] A.3 If there is an empty lane in the buffer zone, if the number of empty lanes is 1, select that lane; otherwise, select any empty lane.

[0016] A.4 If there is an unfilled lane in the buffer, and the positions of product i and the last product j in this lane are f in the downstream sequence of the target.i and f j Satisfy f i <f j If the number of lanes is not full (1), select that lane; otherwise, select the lane with the most available parking spaces. j -f i The lane with the lowest value.

[0017] Preferably, the specific steps of the ant colony optimization algorithm in S3 include:

[0018] S31. The activity area of ​​the artificial ant is defined as a rectangular region constrained by x, y ∈ [1, T]. The ant completes one iteration by crawling from its initial position to the node x = T. The path nodes along the way are P(1, y1), P(2, y2), ..., P(t, y1). t ),…,P(T,y T ), where y t ∈{1,2,…T} corresponds to the position number of the t-th product released from the buffer in the upstream sequence, where t=1,2,…,T;

[0019] S32. Set and initialize algorithm parameters, including:

[0020] Set the number of ants N, and the maximum number of iterations r of the algorithm. max The parameters are: ρ1 (local pheromone evaporation ratio), ρ2 (global pheromone evaporation ratio), q0 (values ​​between [0,1]), α (the pheromone importance), and β and γ (the other two heuristic pheromone importances); where 0 < ρ1 < 1 and 0 < ρ2 < 1.

[0021] Initialize the pheromone τ between any pair of products i and j in the upstream sequence. ij Let τ at the initial time ij =τ0, where τ0 is a preset constant, i,j = 1,2,…,T; the iteration number r is initialized to 0; and the product candidate set Set for each ant n is... n The buffer is initialized as a set consisting of the two products at the very front of each storage lane when arranging products according to scheme A in step S2, n = 1, 2, ..., N; the initial position of ant n is P(1, y1), where y1 is the product candidate set Set. n The product i is randomly selected from the upstream sequence; the selected product i is released from the buffer. If the release of product i is blocked, the product j in front of it is sent to the buffer entrance via the return lane, and the storage lane for the returned product j is reselected according to the heuristic filling rule in step S2, and then product i is released; the product candidate set Set is updated according to the changes in the products at the front end of the lane. n The current crawling steps are 1; initialize the globally optimal downstream sequence So.pt The corresponding sequentialization rate is 0;

[0022] S33. In the t-th step of the r-th iteration, each ant n simultaneously crawls from the node x = t to the node x = t+1; calculate the heuristic information η related to each possible position of x = t+1. j :

[0023] η j =[1 / (InSeq) j +1)] p (T / 10j) v

[0024] Wherein, product j belongs to the product candidate set Set n ;

[0025] T represents the total number of products, and InSeq j Indicate whether product j is a sequential product; if product j and any product a that has been released downstream are in the target downstream sequence, then f represents the position of product j in the target downstream sequence. j and f a Satisfy f j >f a If j is a sequential product, then InSeq j The value is 0; otherwise, j is a non-sequential product. (InSeq) j The value is 1; β represents InSeq. j The importance parameter, γ, represents the importance of product serial number j;

[0026] Calculate the ant n from node P n (t,y t Crawling towards node P n (t+1,y t+1 The transition probability p) n (y t ,y t+1 ):

[0027]

[0028] in Indicates product y t and y t+1 The pheromone concentration between; y t+1 ∈Set n ;

[0029] Ant n crawls to its destination node P in step t. n (t+1,y t+1 This indicates that ant n chooses to release product y from the upstream sequence. t+1 Please select using the following method:

[0030] When q≤q0, choose the option that satisfies The product with the highest value; otherwise, according to the transition probability p n (y t ,y t+1 Determine the candidate products, where q is a random number uniformly distributed between [0,1].

[0031] Update product y t and y t+1 Apart from i, the pheromone concentrations between all other products i and j are as follows:

[0032] τ' ij =(1-ρ1)τ ij , where τ ij and τ' ij These represent the pheromone concentrations before and after the update, respectively.

[0033] Each time ant n crawls one step, it releases the product candidate set Set. n One of the products, the product candidate set Set is updated based on changes in the front-end product status of the lane. n ;

[0034] Let t = t + 1, repeat step S33 until t = T, complete one iteration, and record the crawling path of each ant in this iteration;

[0035] S34. Evaluate the downstream sequences of this iteration, including:

[0036] In the current iteration, the ordinates of the path nodes traversed by each ant constitute a downstream sequence. The sequentiality rate of all N downstream sequences is calculated, and the sequence with the highest sequentiality rate is selected as the optimal downstream sequence for this iteration.

[0037] if The corresponding sequentialization rate is greater than the globally optimal downstream sequence So. pt The corresponding sequentialization rate is updated in So. pt for

[0038] if The corresponding sequentialization rate is equal to So. pt The corresponding sequentialization rate, and The corresponding sequence offset is less than So pt Update So with the corresponding sequence offset. pt for

[0039] S35, Update the globally optimal downstream sequence So pt The pheromone concentration between each pair of adjacent products:

[0040] τ′ij =(1-ρ2)τ ij +Δτ ij

[0041] τ ij and τ' ij Δτ represents the pheromone concentrations before and after the update between products i and j, respectively. ij Calculate using the following formula:

[0042]

[0043] Where z is the globally optimal downstream sequence S opt The number of China-Africa sequential vehicles;

[0044] Let r = r + 1.

[0045] Preferably, in S4, the termination condition of the ant colony optimization algorithm is: r = r max Or the current globally optimal downstream sequence S opt The corresponding sequentialization rate is 100%.

[0046] Preferably, the formula for calculating the product sequencing rate R corresponding to any downstream sequence Seq is:

[0047] Preferably, the formula for calculating the sequence offset D of any downstream sequence Seq is: Where d i It is the position number of product i within Seq. seq With position number f in the downstream sequence of the target i The difference, i.e. d i =i seq -f i .

[0048] Compared with the prior art, the present invention has the following beneficial technical effects:

[0049] 1. This invention uses filling rules to determine the arrangement of products in the buffer, and then applies the ant colony optimization algorithm to obtain the release order of products leaving the buffer. It can restore the upstream sequence to a downstream sequence with the maximum or near-maximum ordering rate compared with the target downstream sequence, and the downstream sequence has a small sequence offset. Attached Figure Description

[0050] Figure 1 A schematic diagram of a design scheme for a linear buffer zone with a return lane;

[0051] Figure 2 This is a flowchart of the production sequence recovery method disclosed in this invention. Detailed Implementation

[0052] Example 1

[0053] In this embodiment, a set of test cases are used to test the effectiveness of the sequence recovery method proposed in this invention. In the test cases, the total number of vehicles T in the sequence is set to 30, 56, and 100. The ordering rate R′ of the upstream sequence relative to the target downstream sequence is set to 0.2, 0.4, and 0.6. The larger the R′ value, the greater the degree of scrambling of the target downstream sequence. For each combination of T and R′ parameters, 5 test cases are randomly generated, resulting in a total of 45 test case problems. For different test case sizes (T values), linear buffers with return lanes are used for sequence recovery with lane and parking space configurations of 5×6, 7×8, and 10×10. The sequence recovery method based on heuristic filling rules and ant colony optimization algorithm disclosed in this invention is written in MATLAB 2021 and run on a personal computer with a CPU of 2.30GHz and 16.00GB of memory. In the ant colony optimization release algorithm: the number of iterations r max =300, number of ants N=30, τ0=1, ρ1=ρ2=0.2, q0=0.5, α=0.5, β=2, γ=30.

[0054] The upstream sequence in the computational example is recovered using the buffer manipulation method based on filling rules and ant colony release algorithm (Method 1) disclosed in this invention. Additionally, for comparison, a traditional linear buffer without return lanes of the same capacity (i.e., the same number of storage lanes and parking spaces) is used in conjunction with a heuristic filling and release rule-based method (Method 2) for sequence recovery. Method 2 uses the same filling rules as Method 1, the difference being in the release rules. The release process includes T steps, each releasing one product. In step t (t = 1, 2, ..., T), a product is selected from the candidate set of releasable products (consisting of the product at the very front of each storage lane) according to the release rules: sequential products are preferred; when there is not a unique candidate product, |tf| is selected. i The smallest product. Table 1 lists the average results of five examples under different combinations of T and R′ parameters, i.e., the average ordering rate corresponding to the downstream sequence. and average sequence skewness Column 3 of Table 1 gives the average sequence offset corresponding to the upstream sequence. The calculation results show that both methods can significantly improve the sequencing efficiency in the sequence and have good sequence recovery capabilities. Furthermore, the result obtained by method one... The value is always higher than that of Method 2, with an improvement rate between [5.1% and 23.2%]; and the value obtained by Method 1 is higher than that of Method 2. The values ​​are always lower than those of Method 2, with a reduction rate between [48.3% and 81.8%]. This indicates that the sequence recovery method (Method 1) disclosed in this invention has superior sequence recovery capabilities.

[0055] Table 1. Sequence recovery results of the two reordering methods

[0056]

[0057] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A production sequence recovery method for a linear buffer zone with return lanes, characterized in that, The specific steps include the following: S1, obtaining total number T of products in an upstream sequence, a product at i-th position in the sequence has a position f in a target downstream sequence i wherein, T≤L×W, i=1, 2, …, T, L is a storage lane number of a linear buffer, and W is a parking space number contained in each storage lane. S2, assigning each product i a storage lane to enter according to the order of arrival in the buffer according to a heuristic filling rule, and constructing a product arrangement scheme A in the buffer; The following heuristic filling rule is executed to assign the upstream product i to enter the storage lane by order: A.1 If i = 1, select the first lane; A.2 If there is an unfilled lane in the buffer, and the product i is in the same lane as the last product j in the target downstream sequence, and the position f of product i in the target downstream sequence is i and f j satisfies f i > f j , if the number of unfilled lanes is 1, select this lane, otherwise select the lane with the smallest f i -f j value among the lanes with the most free spaces. A.3 If there is a free lane in the buffer, if the number of free lanes is 1, select the lane, otherwise, select a free lane at will; A.4 If there is an unfilled lane in the buffer and the product i is in the same lane as the last product j in the target downstream sequence, and the position f of product i in the target downstream sequence is i and f j satisfies f i < f j If the number of unfilled lanes is 1, select this lane, otherwise select the lane with the smallest value of f j - f i where the number of empty positions is the largest. S3, using an ant colony optimization algorithm to obtain a plurality of downstream product release sequences corresponding to the arrangement scheme A, and the specific steps are as follows: S31, determining that the activity area of the artificial ant is a rectangular area with x, y ∈ [1, T] constraints; S32, setting and initializing algorithm parameters; S33, in the tth step of the rth iteration, each ant n simultaneously climbs from the node of x=t to the node of x=t+1; the heuristic information η related to each possible position of x=t+1 is calculated j : let t=t+1, repeat step S33 until t=T, complete one iteration, and record the climbing path of each ant in this iteration; S34, evaluating the downstream sequence of this iteration; S35、updating the global optimal downstream sequence S opt The pheromone concentration between each pair of adjacent products: τ' ij = (1 - p2) τ ij + Δτ ij τ ij and τ' ij respectively represent the pheromone concentration between products i and j before and after updating, Δτ ij is calculated by the following equation: where z is the globally optimal downstream sequence S opt the number of non-sequential cars Let r = r + 1; S4, judging whether the termination condition of the ant colony algorithm is met; if not, jumping to step S3 for next iteration search; if yes, the current optimal downstream sequence S opt is the optimal downstream recovery sequence.

2. The production sequence recovery method for a linear buffer zone with return lanes according to claim 1, characterized in that, The specific steps of the ant colony optimization algorithm in S3 include: An ant completes one iteration when it crawls from the initial position to the node at x = T, and the nodes it passes along the way are P(1, y1), P(2, y2), …, P(t, yt), …, P(T, yT) in turn, where yt∈{1, 2, … T} corresponds to the position number of the tth product released from the buffer in the upstream sequence, t = 1, 2, …, T. t ),…,P(T,y T ), where yt∈{1, 2, … T} corresponds to the position number of the tth product released from the buffer in the upstream sequence, t = 1, 2, …, T. t ​ In step S32, the number N of ants is set, and the maximum iteration number r of the algorithm max , the local pheromone evaporation ratio parameter p1, the global pheromone evaporation ratio parameter p2, the parameter q0 with the value in [0, 1], the parameter a representing the importance degree of pheromone, and the parameters b and g representing the importance degrees of the other two kinds of heuristic information; wherein 0 < p1 < 1 and 0 < p2 < 1. Initialize pheromone τ between any pair of products i and j in the upstream sequence ij , τ ij = τ0, where τ0 is a preset constant, i, j = 1, 2, …, T; the iteration number r is initialized as 0; the product candidate set Set n of each ant n is initialized as a set composed of the two products at the front end of each storage lane in the buffer zone when the products are arranged according to the scheme A in step S2, n = 1, 2, …, N; the initial position of the ant n is P(1, y1), y1 is the position of the product randomly selected from the product candidate set Set n in the upstream sequence; the selected product i is released from the buffer zone, if the release of the product i is blocked, the product j in front of the product i is sent to the entrance of the buffer zone through the return lane, and the storage lane for the returned product j is selected again according to the heuristic filling rule in step S2, and then the product i is released; the product candidate set Set n is updated according to the change of the product at the front end of the lane; the current crawling step number is 1; the global optimal downstream sequence S opt corresponding to the sequencing rate is initialized as 0; In step S33, η j is calculated as follows: η j = [1 / (lnseq j +1)] β (T / 10j) γ wherein product j belongs to a product candidate set Set n ; T is the total number of products, InSeq j denotes whether product j is a sequential product; if product j and any product a released to the downstream are in the same position f in the target downstream sequence j and f a satisfies f j > f a , then j is a sequential product, InSeq j is 0, otherwise j is a non-sequential product, InSeq j is 1; β is a parameter representing the importance of InSeq j , and γ is a parameter representing the importance of product sequence number j; Compute the ant n to move from node P n (t,y t ) to node P n (t+1,y t+1 ) with transition probability p n (y t ,y t+1 ) wherein represents the pheromone concentration between products y t and y t+1 ; y t+1 ∈ Set n ; Ant n climbs to destination node P in step t n (t+1,y t+1 ) denotes that ant n chooses to release product y in the upstream sequence t+1 The choice is made according to the following method: When q ≤ q0, select the product that satisfies the maximum value; otherwise, according to the transition probability p n (y t ,y t+1 ) to determine the candidate product, where q is a random number uniformly distributed between [0, 1]. Update pheromone concentrations between all products i and j except for product y t and y t+1 between all products i and j except for product y τ' ij = (1 - p1) τ ij where τ ij and τ' ij represent the pheromone concentration before and after updating, respectively. The ant n releases one product from the product candidate set Set at each crawling step n The product candidate set Set is updated according to the product change at the front end of the lane n ; Step S34 further comprises: the longitudinal coordinates of the path nodes that each ant crawls in the current iteration constitute a downstream sequence, calculating the ordering rate of all N downstream sequences, and selecting the sequence with the largest ordering rate as the optimal downstream sequence of the current iteration If The corresponding sequencing rate is greater than the global optimal downstream sequence S opt The corresponding sequencing rate, update S opt For If the corresponding sequencing rate is equal to S opt the corresponding sequencing rate, and the corresponding sequence offset is less than S opt the corresponding sequence offset, update S opt is 3. The method of claim 1, wherein, In S4, the termination condition of the ant colony optimization algorithm is: r = r max , or the current global optimal downstream sequence S opt The corresponding sequencing rate is 100%.

4. The method of claim 2, wherein, The calculation formula of the product order rate R corresponding to any downstream sequence Seq is:

5. The method of claim 2, wherein, The calculation formula of the sequence offset degree D of any downstream sequence Seq is: Where d i is the difference between the position sequence number i seq of product i in Seq and the position sequence number f i in the target downstream sequence, that is, d i = i seq -f i .

Images