Rapid-response intelligent scheduling method for semiconductor packaging and testing workshop, and system

Abstract

The present invention relates to a rapid-response intelligent scheduling method for a semiconductor packaging and testing workshop, and a system. Workshop operation data is transmitted to a bottleneck identification module by means of a GUI module; the bottleneck identification module identifies all bottleneck processes by means of a "buffer-bottleneck index" bottleneck identification method; the GUI module and the bottleneck identification module transmit workshop scheduling data and a bottleneck process identification result to a scheduling module, respectively; an intelligent scheduling submodule establishes a bottleneck process scheduling model in a semiconductor packaging and testing workshop, and inputs a bottleneck process scheduling solution into a rule-based scheduling submodule; and the rule-based scheduling submodule generates a global scheduling solution and transmits the global scheduling solution to the GUI module for arranging production. The problems of work redundancy and low efficiency due to the scheduling work of a semiconductor packaging and testing factory overly focusing on a certain process are solved, so that the present invention can be used for the scheduling of a semiconductor packaging and testing workshop, thereby improving the production efficiency of the workshop. The phenomenon of "bottleneck drifting" occurring in a semiconductor packaging and testing workshop can be mitigated and overcome, thereby improving production benefits.

Description

A fast-response semiconductor packaging and testing workshop intelligent scheduling method and system Technical Field

[0001] The present invention relates to a workshop scheduling technology, and in particular to a fast-response semiconductor packaging and testing workshop intelligent scheduling method and system. Background Art

[0002] Semiconductor packaging and testing is an important part of my country's semiconductor industry, with great economic value and strategic significance. Semiconductor packaging and testing production is often considered a mixed flow workshop. Semiconductor packaging and testing workshops have the characteristics of diverse product types, complex processes, a large number of equipment units and long processing cycles, making it difficult to efficiently complete the scheduling of semiconductor packaging and testing workshops. Therefore, studying the scheduling problem of semiconductor packaging and testing workshops has high practical value and theoretical significance. The semiconductor packaging and testing process includes wafer inspection, backside thinning, wafer dicing, chip mounting, wire bonding, plastic encapsulation, laser printing, rib cutting and forming, electroplating, and final testing. Among them, chip mounting, wire bonding and plastic encapsulation are often considered to be the most important processes in the packaging and testing workshop, and their processing efficiency has a significant impact on the production efficiency of the semiconductor packaging and testing workshop.

[0003] Chinese patents CN109085803A, CN105320105A, and CN103246240A primarily focus on scheduling specific processes in the semiconductor packaging and testing sector. While these applications can improve production efficiency by reducing changeover time, increasing the efficiency of certain production lines, and shortening production cycles, thereby increasing production efficiency, they fail to consider the overall nature of semiconductor packaging and testing workshop scheduling and fail to fully address the issue. Because orders received by semiconductor packaging and testing companies do not match the production capacity of each workshop process, this leads to a "bottleneck drift" phenomenon in the workshop. This means that the process where the production bottleneck occurs constantly shifts, causing the processes experiencing "blockage" and "starvation" to also shift. Considering only the scheduling of a single process fails to address bottleneck drift and cannot achieve truly effective scheduling. Summary of the Invention

[0004] To address the redundancy and inefficiency of scheduling in semiconductor packaging and testing plants, which often focuses on a single process, we propose a fast-response intelligent scheduling method and system for semiconductor packaging and testing workshops. This method aims to improve production efficiency by addressing the overall scheduling of semiconductor packaging and testing plants. This approach addresses the bottleneck phenomenon that often occurs in semiconductor packaging and testing workshops, mitigating and overcoming it, thereby improving production efficiency. Therefore, a holistic study of scheduling in semiconductor packaging and testing production workshops is of great significance.

[0005] The technical solution of the present invention is:

[0006] A fast-response intelligent scheduling method for semiconductor packaging and testing workshops includes three modules: a graphical user interface (GUI) module, a bottleneck identification module, and a scheduling module. The scheduling module includes an intelligent scheduling submodule and a rule-based scheduling submodule. These modules complete the following steps:

[0007] Step 1: The user transmits shop floor operation data to the bottleneck identification module through the GUI module;

[0008] Step 2: The bottleneck identification module identifies all bottleneck processes using the “buffer-bottleneck index” bottleneck identification method;

[0009] Step 3: The GUI module and the bottleneck identification module transmit the shop floor scheduling data and the bottleneck process identification results to the scheduling module respectively;

[0010] Step 4: The intelligent scheduling submodule establishes a bottleneck process scheduling model for the semiconductor packaging and testing workshop, uses IAHA to schedule the bottleneck process, and inputs the obtained bottleneck process scheduling plan into the rule scheduling submodule;

[0011] Step 5: Based on the rule base, the user selects rules for each process in advance. Combined with the bottleneck process scheduling solution obtained by the intelligent scheduling submodule, the rule scheduling submodule generates a global scheduling solution and transmits it to the GUI module to arrange production.

[0012] Furthermore, the specific operations of step 2 are:

[0013] Step 2.1: Determine whether the buffer of process s (s∈1…S, where S is the total number of processes) occurs in a product accumulation state. If so, the process is a bottleneck process and proceeds to step 2.3; otherwise, proceeds to step 2.2;

[0014] Step 2.2: Use formula (1) and formula (2) to calculate the bottleneck index of each process and determine whether the process has the highest bottleneck index. If so, proceed to step 2.3; otherwise, proceed to step 2.4. c s =T s -F s (t) (2)

[0015] Among them I BN is the bottleneck index, w t ,w b ,w q are the weights of the impact of the number of products produced by the process, the process buffer and the product quality on the bottleneck degree, and they satisfy w t +w b +w q =1,c s and l sare production capacity and production load, T s Assuming available processing power, F s (t) is the amount by which the production capacity of process s changes due to changes in actual production conditions, and They are the maximum number of products that can be carried in the buffer zone and the number of new products added to the buffer zone. Quality assurance capability (q ac )’s impact function on the degree of bottleneck, (q ac ) is a comprehensive reflection of quality capabilities and quality requirements;

[0016] Step 2.3: Record the process as a bottleneck process and proceed to step 2.5;

[0017] Step 2.4: Record the process as a non-bottleneck process and proceed to step 2.5;

[0018] Step 2.5: Determine whether this process is the last process. If so, proceed to step 2.6; otherwise, proceed to step 2.1.

[0019] Step 2.6: End the judgment and output all bottleneck processes to the scheduling module.

[0020] Furthermore, the specific operations of step 4 are:

[0021] Step 4.1: Establish a bottleneck process scheduling model for a semiconductor packaging and testing workshop, which includes all bottleneck processes and processes between bottlenecks;

[0022] Step 4.2: Optimize the bottleneck process shop scheduling model by improving the artificial hummingbird algorithm;

[0023] Step 4.2.1: Initialize by improving the NEH heuristic rule instead of random generation;

[0024] Step 4.2.2: Initialize the food source access table;

[0025] Step 4.2.3: Set the number of iterations to Iteration and use IAHA to perform optimization search;

[0026] Step 4.2.4: Use the improved foraging method of the foraging judgment formula to replace the 50% probability of guided foraging or territorial foraging;

[0027] Step 4.2.5: Determine the number of iterations. If the number of iterations is a multiple of the migration coefficient preset value n, perform enhanced foraging in the taboo area.

[0028] Step 4.2.6: Determine the number of iterations. If the number of iterations exceeds the predetermined migration coefficient value 2n, migration foraging is performed.

[0029] Step 4.2.7: Output the individual with the optimal fitness value, that is, the optimal solution for bottleneck process scheduling;

[0030] Step 4.3: The intelligent scheduling sub-module outputs the bottleneck process scheduling plan.

[0031] Furthermore, the specific operation of Step 4.2.1 is as follows:

[0032] Step 4.2.1.1: Calculate the total processing time of all orders j ∈ 2...n, that is, the sum of the ratios of the processing time of each stage of the order to the processing speed of each stage; Arrange the orders in non-increasing order of TP j to obtain the initial arrangement π 0 ={π 0 (1), π 0 (2), …, π 0 (n)};

[0033] Step 4.2.1.2: Take out the first two orders π 0 of π, namely π 0 (1) and π 0 (2), and sort them to obtain these two possible schedules {π 0 (1), π 0 (2)} and {π 0 (2), π 0 (1)}; Evaluate these two partial schedules, and take the one with the smaller maximum completion time as the current schedule, denoted as π = {π(1), π(2)};

[0034] Step 4.2.1.3: Take out the j-th order π 0 of π, namely π 0 (j), and insert it into all possible positions of π, obtaining a total of j partial permutations; Evaluate the obtained partial permutations, and take the partial permutation with the smallest maximum completion time as the current schedule π; <0-000189>

[0035] Step 4.2.1.4: Let j = j + 1; If j ≤ n - 1, then go to Step 2.1.3; Otherwise, output the current schedule π;

[0036] Step 4.2.1.5: Based on the current schedule π = {π(1), π(2), …, π(n)}, randomly take integers l, m, satisfying the condition l < m ≤ n, and exchange the order of the l-th order and the m-th order in the current schedule π to obtain a new individual;

[0037] Step 4.2.1.6: Repeat Step 4.1.5 until a set of hummingbird individuals with a quantity of Popsize / 2 is generated;

[0038] Step 4.2.1.7: Use the result of the improved NEH heuristic algorithm to directly generate Popsize / 2 duplicate hummingbird individuals, and merge them with the individual set generated in Step 4.1.6 to obtain the initial population.

[0039] Further, the specific operation of Step 4.2.4 is as follows:

[0040] Step 4.2.4.1: Select the foraging method through the foraging judgment formula;

[0041] Use the foraging judgment formula to replace the original 50% probability in AHA to select guided foraging and territorial foraging; it is beneficial to increase the selection ratio of guided foraging in the early stage of iteration to enhance the ability to explore other spaces, and increase the selection of territorial foraging in the later stage of iteration to enhance the ability to find local optimal solutions;

[0042] Where GT is the foraging judgment coefficient; GT max and GT min are the maximum foraging judgment coefficient and the minimum foraging judgment coefficient respectively, and the values are Popsize is the population size, Popsize ≥mean is the number of individuals in the population whose fitness value is greater than or equal to the average fitness value of the population; it is the iteration number; Iteration is the maximum iteration number; v i (t + 1) is the candidate food source position of the i-th hummingbird at time t + 1; f Guided (v i,tar (t), x i (t)) is the guided foraging search, x i (t) is the food source position of the i-th hummingbird at time t, v i,tar (t) is the position of the target food source that the i-th hummingbird intends to visit; fTerritorial(x i (t)) is the territorial foraging search;

[0043] Step 4.2.4.2: The hummingbird finds a better food source through three flight methods;

[0044] When the generated random number rand ≥ GT, the hummingbird chooses to conduct guided foraging. The three flight methods of guided foraging: draw a processing position to swap two orders, draw multiple consecutive positions to swap two order blocks, and draw multiple positions to swap multiple orders;

[0045] When the generated random number rand < GT, the hummingbird chooses to conduct territorial foraging. The three flight methods of territorial foraging: draw two positions to swap two orders, draw multiple consecutive positions to swap orders within a block, and draw multiple positions to swap multiple orders;

[0046] Step 4.2.4.3: Update the population and visit tables;

[0047] The position of the i-th food source is updated as follows:

[0048] The fitness(*) in the formula is the fitness value of the hummingbird. If the fitness of the candidate food source is lower than the current food source, the hummingbird will abandon the current food source and stay at the newly generated candidate food source. i (t+1) Feeding; updating the visit table.

[0049] Furthermore, the specific operations of step 4.2.5 are:

[0050] Step 4.2.5.1: Calculate the number of times each hummingbird performs reinforced foraging, RS, by taking a random integer randint (randint∈[2,3,4,5]) as the number of hummingbirds that perform reinforced foraging. num (RS num = Popsize / randint);

[0051] Step 4.2.5:2: Field-foraging hummingbirds enhanced by roulette wheel screening;

[0052] Step 4.2.5.3: Perform RS on each selected hummingbird in turn. num Enhanced Taboo Area Foraging: This method adds a taboo table to the area foraging method to ensure that the search results for each enhanced taboo area foraging are different.

[0053] Step 4.2.5.4: Update the population and visit tables.

[0054] Furthermore, the specific operations of step 4.2.6 are:

[0055] Step 4.2.6.1: Select the hummingbird with the worst fitness and have it migrate to the location of the hummingbird with the best fitness.

[0056] Step 4.2.6.2: Update the access table.

[0057] Furthermore, the specific operations of step 5 are:

[0058] The heuristic rule base of the rule scheduling submodule includes the heuristic processing sequence rule base and the heuristic equipment unit selection rule base;

[0059] The processing sequence rule base includes rules such as the shorter the processing time, the higher the priority, the longer the processing time, the first-come-first-served rule, and the earlier the process delivery date, the higher the priority.

[0060] Equipment unit selection rule base, including rules for selecting the equipment unit with the shortest processing time, the equipment unit with the shortest changeover time, the equipment unit with the highest precision, the equipment unit with the lowest precision, the equipment unit with the most processable order types, and the equipment unit with the least processable order types.

[0061] Step 5.1: The user selects the corresponding heuristic rules for each process in advance;

[0062] Step 5.2: Generate an overall scheduling plan using the processing sequence rules and equipment unit selection rules selected for each process;

[0063] Step 5.3: Transfer the generated global scheduling plan to the GUI module to arrange production.

[0064] A fast-response semiconductor packaging and testing workshop intelligent scheduling system includes a graphical user interface (GUI) module, a bottleneck identification module, and a scheduling module; and is used to implement the fast-response semiconductor packaging and testing workshop intelligent scheduling method;

[0065] The GUI module is used to transmit workshop operation data, workshop scheduling data and receive global scheduling plans;

[0066] The bottleneck identification module identifies all bottleneck processes through the "buffer-bottleneck index" bottleneck identification method;

[0067] The scheduling module includes an intelligent scheduling submodule and a rule scheduling submodule. The intelligent scheduling submodule solves the bottleneck process scheduling problem by improving the artificial hummingbird algorithm and generates a bottleneck scheduling plan. The rule scheduling submodule allows users to independently select rules from the rule library and combines them with the bottleneck scheduling plan to generate a global scheduling plan.

[0068] Furthermore, it also includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, it executes the "buffer-bottleneck index" bottleneck identification method of the bottleneck identification module, the IAHA algorithm of the intelligent scheduling submodule in the scheduling module, and the rule scheduling submodule in the scheduling module.

[0069] The beneficial effects of the present invention are:

[0070] The present invention discloses a fast-response intelligent scheduling method and system for semiconductor packaging and testing workshops, relating to the field of workshop scheduling. The present invention considers the overall scheduling of semiconductor packaging and testing factories, establishes a conductor packaging and testing workshop scheduling model, screens bottleneck processes using a "buffer-bottleneck index" bottleneck identification method, constructs a bottleneck process workshop scheduling model, uses an improved artificial hummingbird algorithm (IAHA) to solve bottleneck processes, and utilizes a heuristic rule library to complete the scheduling of missing processes, thereby completing the scheduling of the entire workshop. This solves the scheduling optimization problem of semiconductor packaging and testing production workshops, shortens the production cycle of bottleneck processes, improves the processing efficiency of the production line, and thus improves the production efficiency of semiconductor packaging and testing production workshops. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] FIG1 is an overall design diagram of a fast-response semiconductor packaging and testing workshop intelligent scheduling method and system according to the present invention;

[0072] FIG2 is a flow chart of the bottleneck identification method of the present invention using the "buffer-bottleneck index" method;

[0073] FIG3 is a flow chart of the IAHA of the present invention;

[0074] FIG4 is a diagram illustrating three flight modes for improved guided foraging according to the present invention;

[0075] FIG5 is a diagram showing the three flight modes for improving the field foraging of the present invention;

[0076] FIG6 is a Gantt chart of four scheduling methods for bottleneck processes according to the present invention. DETAILED DESCRIPTION

[0077] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0078] The present invention provides a fast-response semiconductor packaging and testing workshop intelligent scheduling method and system, including a graphical user interface (GUI) module, a bottleneck identification module, and a scheduling module (including an intelligent scheduling submodule and a rule scheduling submodule), as shown in Figure 1.

[0079] The GUI module is used to transmit shop floor operation data, shop floor scheduling data, and receive global scheduling plans. The bottleneck identification module uses the "buffer-bottleneck index" bottleneck identification method. The scheduling module includes intelligent scheduling and rule-based scheduling submodules. The intelligent scheduling submodule uses the IAHA algorithm to solve bottleneck process scheduling problems and generate bottleneck scheduling plans. The rule-based scheduling submodule generates a global scheduling plan by allowing users to select rules from a rule library and combining them with bottleneck scheduling plans.

[0080] The scheduling process of the system includes the following steps:

[0081] Step 1: The user transmits shop floor operation data to the bottleneck identification module through the GUI module;

[0082] Step 2: The bottleneck identification module identifies all bottleneck processes using the "buffer-bottleneck index" bottleneck identification method. The bottleneck identification process is shown in FIG2 . The bottleneck index-related data of each process in the semiconductor packaging and testing workshop of the present invention is shown in Table 1. The specific steps are as follows;

[0083] Table 1 Process bottleneck index related data

[0084] Step 2.1: Determine whether the buffer of process s (s∈1…S, where S is the total number of processes) occurs in a product accumulation (buffer addition). If so, the process is a bottleneck process and proceed to step 2.3; otherwise, proceed to step 2.4.

[0085] Step 2.2: Use formula (1) and formula (2) to calculate the bottleneck index of each process and determine whether the process has the highest bottleneck index. If so, proceed to step 2.3; otherwise, proceed to step 2.4. c s =T s -F s (t) (2)

[0086] Among them I BN is the bottleneck index, w t ,w b ,w q are the weights of the impact of the number of products produced by the process, the process buffer and the product quality on the bottleneck degree, and they satisfy w t +w b +w q =1,c s and l s They are production capacity and production load (actual processing product volume), T s Assuming available processing power, F s (t) is the amount by which the production capacity of process s changes due to changes in actual production conditions, and They are the maximum number of products that can be carried in the buffer zone and the number of new products added to the buffer zone. Quality assurance capability (q ac )’s impact function on the degree of bottleneck, (q ac ) is a comprehensive reflection of quality capabilities and quality requirements.

[0087] Step 2.3: Record the process as a bottleneck process and proceed to step 2.5;

[0088] Step 2.4: Record the process as a non-bottleneck process and proceed to step 2.5;

[0089] Step 2.5: Determine whether this process is the last process. If so, proceed to step 2.6; otherwise, proceed to step 2.1.

[0090] Step 2.6: End the judgment and output all bottleneck processes to the scheduling module;

[0091] Through the bottleneck identification method of "buffer-bottleneck index", the three connected processes of chip placement, wire bonding and plastic packaging are identified as bottlenecks due to the accumulation of buffers; the weight value w is taken t =0.5,w b =0.5,w q =0, bottleneck index I BN The maximum value is 0.443 in the chip mounting process, so the bottleneck processes in this case are chip mounting, wire bonding, and plastic packaging.

[0092] Step 3: The GUI module and the bottleneck identification module transmit the shop floor scheduling data and the bottleneck process identification results to the scheduling module respectively;

[0093] Step 4: The intelligent scheduling submodule establishes a bottleneck process scheduling model for the semiconductor packaging and testing workshop, uses IAHA to schedule the bottleneck process, and inputs the obtained bottleneck process scheduling plan into the rule scheduling submodule. The specific steps are as follows;

[0094] Step 4.1: Establish a bottleneck process scheduling model for a semiconductor packaging and testing workshop, which includes all bottleneck processes and processes between bottlenecks;

[0095] The parameter symbols of the mathematical model are defined as follows:

[0096] N--total number of orders, i order index;

[0097] K - total number of processing stages, k-stage index;

[0098] m k --Total number of equipment units in the stage;

[0099] W i,k--Processing time of order i at stage k;

[0100] P i,k,j --Processing time of order i at equipment unit j in stage k;

[0101] v k,j --speed of the stage equipment unit;

[0102] B i,k,j --The start time of the order in stage k equipment unit j;

[0103] E i,k,j --Completion time of equipment unit j of order at stage k;

[0104] C i --Completion time of order i; C max Maximum completion time;

[0105] X i,k,j -- is a 0-1 variable. If order i is scheduled to be processed on the j-th equipment unit in the stage, the value is 1, otherwise it is 0.

[0106] Furthermore, the objective function of the shop floor scheduling model is formula (3): fitness = w1*C max +w2*Cost+w3*Tardiness (3)

[0107] Among them, fitness is the weighted scheduling target, w1, w2, w3 are the weight values ​​of maximum processing time, processing cost, and average delay time respectively, and w1+w2+w3=1. The specific value of weight time is determined according to the actual situation of different semiconductor packaging and testing production workshops. In this case, the weight values ​​are w1=1, w2=0, w3=0, C max ,Cost, and Tardiness are the maximum processing time, processing cost, and average tardiness time respectively.

[0108] The constraints of the mathematical model are as follows: E i,k,j ≤B i,k+1,j (4) E i2,k,j =B i2,k,j +P i2,k,j (7) C i =E i,K,j (8)

[0109] Formula (3) is the objective function; Formula (4) indicates that the start time of the current process of each order is after the previous process; Formula (5) indicates that each process of each order is processed on only one equipment unit; Formula (6) indicates the completion time of the order in each process; Formula (7) indicates that the completion time of an order is the sum of the start processing time of the order and the processing time of the stage; Formula (8) indicates that the final completion time of the order is the end time of the K-stage processing.

[0110] The relevant data on the scheduling of bottleneck process workshops are shown in Tables 2 and 3. Table 2 shows the order processing time data for the bottleneck process workshop scheduling, which records the production volume of 20 orders and the basic processing time of the corresponding chip models in the three bottleneck processes of chip mounting, wire bonding, and plastic packaging, that is, the time spent on processing 10,000 chips; Table 3 shows the equipment unit processing rate data for the bottleneck process workshop, which records the number of scheduled equipment units and the number of equipment units with different processing rates.

[0111] Table 2. Bottleneck process workshop scheduling order processing time data

[0112] Table 3 Bottleneck process workshop equipment unit processing rate data

[0113] Step 4.2: Optimize the bottleneck process shop scheduling model through IAHA. The algorithm flow is shown in Figure 3.

[0114] Step 4.2.1: Initialize the algorithm by improving the NEH (Nawaz, Enscore, and Ham) heuristic instead of random generation.

[0115] Step 4.2.1.1: Calculate the total processing time for all orders j∈2...n, that is, the sum of the ratios of the processing time of each stage of the order to the sum of the processing speed of each stage. j Arrange the orders in non-increasing order to get the initial arrangement π 0 ={π 0 (1),π 0 (2),…,π 0 (n)}.

[0116] Step 4.2.1.2: Take out π 0 The first two orders of π 0 (1) and π 0 (2), sorting them can get these two possible scheduling {π 0 (1),π 0 (2)} and {π 0 (2),π 0(1). Evaluate these two partial schedules, and take the one with a smaller makespan as the current schedule, denoted as π = {π(1), π(2)}. Let j = 3.

[0117] Step 4.2.1.3: Take out the j-th order π 0 (j) of π, and insert it into all possible positions of π, obtaining a total of j partial permutations. Evaluate the obtained partial permutations, and take the partial permutation with the smallest makespan as the current schedule π. 0 (j) of π, and insert it into all possible positions of π, obtaining a total of j partial permutations. Evaluate the obtained partial permutations, and take the partial permutation with the smallest makespan as the current schedule π.

[0118] Step 4.2.1.4: Let j = j + 1. If j ≤ n - 1, go to Step 2.1.3; otherwise, output the current schedule π.

[0119] Step 4.2.1.5: Based on the current schedule π = {π(1), π(2), …, π(n)}, randomly select integers l and m that satisfy the condition l < m ≤ n, and swap the order of the l-th order and the m-th order in the current schedule π to obtain a new individual.

[0120] Step 4.2.1.6: Repeat Step 4.1.5 until a set of Hummingbird individuals with a quantity of Popsize / 2 is generated;

[0121] Step 4.2.1.7: Use the result of the improved NEH heuristic algorithm to directly generate Popsize / 2 duplicate Hummingbird individuals, and merge them with the set of individuals generated in Step 4.1.6 to obtain the initial population.

[0122] Step 4.2.2: Initialize the food source access table;

[0123] where i = j, VT i,j = null indicates that the hummingbird is feeding at a specific food source; for i ≠ j, VT i,j = 0 indicates that the i-th hummingbird has just visited the j-th food source in the current iteration.

[0124] Step 4.2.3: Set the number of iterations as Iteration, and use IAHA for optimization search;

[0125] Step 4.2.4: Improve the foraging method through the foraging judgment formula, replacing the 50% probability for each of the guided foraging method or the neighborhood foraging method;

[0126] Step 4.2.4.1: Select the foraging method through the foraging judgment formula;

[0127] By replacing the original 50% probability in AHA with the foraging judgment formula, the choices of guided foraging and territorial foraging are made. This is beneficial for increasing the proportion of guided foraging choices in the early stage of iteration, enhancing the ability to explore other spaces, and increasing the proportion of territorial foraging choices in the later stage of iteration, enhancing the ability to find local optimal solutions;

[0128] where GT is the foraging judgment coefficient; GT max and GT min are the maximum foraging judgment coefficient and the minimum foraging judgment coefficient respectively, with values Popsize is the population size, Popsize ≥mean is the number of individuals in the population whose fitness value is greater than or equal to the average fitness value of the population; it is the iteration number; Iteration is the maximum iteration number; v i (t + 1) is the position of the candidate food source of the i-th hummingbird at time t + 1; f Guided (v i,tar (t), x i (t)) is the guided foraging search, x i (t) is the position of the food source of the i-th hummingbird at time t, v i,tar (t) is the position of the target food source that the i-th hummingbird intends to visit; fTerritorial(x i (t)) is the territorial foraging search.

[0129] Step 4.2.4.2: The hummingbird searches for a better food source through three flight modes;

[0130] When the generated random number rand ≥ GT, the hummingbird chooses to conduct guided foraging, as shown in Figure 4. The three flight modes of guided foraging: extracting a processing position to swap two orders, extracting multiple consecutive positions to swap two order blocks, and extracting multiple positions to swap multiple orders.

[0131] When the generated random number rand < GT, the hummingbird chooses to conduct territorial foraging, as shown in Figure 5. The three flight modes of territorial foraging: extracting two positions to swap two orders, extracting multiple consecutive positions to swap orders within a block, and extracting multiple positions to swap multiple orders.

[0132] Step 4.2.4.3: Update the population and the access table.

[0133] The position of the i-th food source is updated as follows:

[0134] In the formula fitness(*), it is the fitness value represented by the hummingbird. If the fitness of the candidate food source is lower than that of the current food source, the hummingbird abandons the current food source and stays at the newly generated candidate food source vi (t+1) feeding; updating the visit table;

[0135] Step 4.2.5: Determine the number of iterations. If the number of iterations is a multiple of the migration coefficient preset value n, perform enhanced foraging in the taboo area.

[0136] Step 4.2.5.1: Calculate the number of times each hummingbird performs reinforced foraging, RS, by taking a random integer randint (randint∈[2,3,4,5]) as the number of hummingbirds that perform reinforced foraging. num (RS num = Popsize / randint);

[0137] Step 4.2.5.2: Field-foraging hummingbirds enhanced by roulette wheel screening;

[0138] Step 4.2.5.3: Perform RS on each selected hummingbird in turn. num Enhanced Taboo Area Foraging: This method adds a taboo table to the area foraging method to ensure that the search results for each enhanced taboo area foraging are different.

[0139] Step 4.2.5.4: Update the population and visit tables;

[0140] Step 4.2.6: Determine the number of iterations. If the number of iterations exceeds the predetermined migration coefficient value 2n, migration foraging is performed.

[0141] Step 4.2.6.1: Select the hummingbird with the worst fitness and have it migrate to the location of the hummingbird with the best fitness.

[0142] Step 4.2.6.2: Update the access table;

[0143] Step 4.2.7: Output the individual with the best fitness value, that is, the optimal solution for bottleneck process scheduling;

[0144] Table 4 compares the performance of the algorithms described above, using the improved artificial hummingbird algorithm (IAHA) and the genetic algorithm (GA) for 20 runs. The results show that the IAHA significantly outperforms the traditional GA in both time and algorithm performance, and significantly outperforms the FIFO (First in, First out) and NEH (Nawaz, Enscore, and Ham) heuristics. Figure 6 shows a Gantt chart for scheduling bottleneck processes using the IAHA algorithm, the genetic algorithm, the FIFO, and the NEH heuristics.

[0145] Table 4 Algorithm performance comparison

[0146] Step 4.3: The intelligent scheduling submodule outputs the bottleneck process scheduling plan;

[0147] Step 5: Based on the rule base, the user selects rules for each process in advance. Combined with the bottleneck process scheduling solution obtained by the intelligent scheduling submodule, the rule scheduling submodule generates a global scheduling solution and transmits it to the GUI module to arrange production. The specific steps are as follows;

[0148] The heuristic rule base of the rule scheduling submodule includes a heuristic processing sequence rule base and a heuristic equipment unit selection rule base. The order processing sequence rule base includes rules such as priority for shorter processing times, priority for longer processing times, first-come-first-served processing, and priority for earlier process delivery dates. The equipment unit selection rule base includes rules for selecting the equipment unit with the shortest processing time, the shortest changeover time, the highest precision, the lowest precision, the equipment unit with the most processable order types, and the equipment unit with the fewest processable order types.

[0149] Step 5.1: The user selects the corresponding heuristic rules for each process in advance;

[0150] Step 5.2: Generate an overall scheduling plan using the processing order rules and equipment unit selection rules selected for each process. In this case, the non-bottleneck process has sufficient capacity, so the FIFO rule is used to select the processing orders, and the FAM rule (limited idle equipment units) is used to select the equipment units.

[0151] Step 5.3: Transfer the generated global scheduling plan to the GUI module to arrange production.

[0152] It should be further noted that the above description is merely a detailed description of specific embodiments of the present invention, and is not intended to limit the scope of protection of the present invention. Equivalent modifications and substitutions made by those skilled in the art to the present invention are all within the scope of the present invention. Therefore, without departing from the spirit and scope of the present invention, equivalent changes and modifications made to the present invention are all within the scope of the present invention.

Claims

1. An intelligent scheduling method for a fast-response semiconductor packaging test workshop, characterized in that, It includes three parts: the Graphical User Interface (GUI) module, the bottleneck identification module, and the scheduling module. The scheduling module includes an intelligent scheduling sub-module and a rule-based scheduling sub-module. The following steps are completed through these modules: Step 1: The user transmits the workshop operation data to the bottleneck identification module through the GUI module. Step 2: The bottleneck identification module identifies all bottleneck processes through the "buffer-bottleneck index" bottleneck identification method. Step 3: The GUI module and the bottleneck identification module respectively transmit the workshop scheduling data and the bottleneck process identification results to the scheduling module. Step 4: The intelligent scheduling sub-module establishes a bottleneck process scheduling model for the semiconductor packaging and testing workshop, uses the Improved Artificial Hummingbird Algorithm (IAHA) for bottleneck process scheduling, and inputs the obtained bottleneck process scheduling plan to the rule-based scheduling sub-module. Step 5: According to the rule library, the user selects rules for each process in advance. Combining with the bottleneck process scheduling plan obtained by the intelligent scheduling sub-module, the rule-based scheduling sub-module generates a global scheduling plan and transmits it to the GUI module to arrange production.

2. The intelligent scheduling method for a fast-response semiconductor packaging test workshop according to claim 1, wherein The specific operation of Step 2 is as follows: Step 2.1: Determine whether there is a backlog of work-in-progress in the buffer of process s (s ∈ 1…S, where S is the total number of processes). If so, this process is a bottleneck process and proceed to Step 2.3; otherwise, proceed to Step 2.

2. Step 2.2: Calculate the bottleneck index of each process using Formula (1) and Formula (2), and determine whether this process is the one with the highest bottleneck index. If so, proceed to Step 2.3; otherwise, proceed to Step 2.4; c s = T s - F s (t) (2) Among which I BN is the bottleneck index, w t , w b , w q are respectively the influence weights of the number of products produced by the process, the process buffer, and the product quality on the bottleneck degree and satisfy w t + w b + w q = 1, c s and l s are respectively the production capacity and the production load, T s is the assumed available processing capacity, F s (t) is the quantity by which the production capacity of process s changes due to the change of actual production conditions, And They are the maximum number that the buffer can hold and the number of new products added to the buffer respectively, is the influence function of the quality assurance ability (q ac ) on the bottleneck degree, and (q ac ) is a comprehensive reflection of the quality ability and quality requirements; Step 2.3: Record this process as a bottleneck process and proceed to Step 2.

5. Step 2.4: Record this process as a non-bottleneck process and proceed to Step 2.

5. Step 2.5: Determine whether this process is the last process. If so, proceed to Step 2.6; otherwise, proceed to Step 2.

1. Step 2.6: End the judgment and output all bottleneck processes to the scheduling module.

3. The intelligent scheduling method for a fast-response semiconductor packaging test workshop according to claim 2, wherein The specific operation of Step 4 is as follows: Step 4.1: Establish a bottleneck process scheduling model for the semiconductor packaging and testing workshop, which includes all bottleneck processes and the processes between bottlenecks. Step 4.2: Optimize the bottleneck process workshop scheduling model through the Improved Artificial Hummingbird Algorithm (IAHA). Step 4.2.1: Initialize the population by improving the NEH heuristic rule instead of randomly generating the initial population. Step 4.2.2: Initialize the food source access table. Step 4.2.3: Set the number of iterations as Iteration and use IAHA for optimization search. Step 4.2.4: Improve the foraging method through the foraging judgment formula to replace the foraging method of guiding foraging or neighborhood foraging with a 50% probability each. Step 4.2.5: Judge the number of iterations. If the number of iterations is a multiple of the preset value n of the migration coefficient, perform enhanced tabu neighborhood foraging. Step 4.2.6: Judge the number of iterations. If the number of iterations exceeds the preset value 2n of the migration coefficient, perform migratory foraging. Step 4.2.7: Output the individual with the optimal fitness value, that is, the optimal plan for bottleneck process scheduling. Step 4.3: The intelligent scheduling sub-module outputs the bottleneck process scheduling plan.

4. The intelligent scheduling method for a fast-response semiconductor packaging test workshop according to claim 3, characterized in that The specific operation of Step 4.2.1 is as follows: Step 4.2.1.1: Calculate the total processing time of all orders That is, the sum of the ratios of the processing times at each stage of the order to the sum of the processing speeds at each stage; according to TP j Arrange the orders in non-increasing order to obtain the initial arrangement π 0 ={π 0 (1), π 0 (2), …, π 0 (n)}; Step 4.2.1.2: Take out π 0 The first two orders of π 0 (1) and π 0 (2). Sorting them can obtain these two possible schedules {π 0 (1), π 0 (2)} and {π 0 (2), π 0 (1)}; Evaluate these two partial schedules, and take the one with the smaller maximum completion time as the current schedule, denoted as π = {π(1), π(2)}; Step 4.2.1.3: Take out π 0 The j-th order π 0 (j), insert it into all possible positions of π, obtaining a total of j partial permutations; evaluate the obtained partial permutations, and take the partial permutation with the minimum maximum completion time as the current schedule π; Step 4.2.1.4: Let j = j + 1; if j ≤ n - 1, then go to Step 2.1.3; otherwise, output the current schedule π. Step 4.2.1.5: Based on the current schedule π = {π(1), π(2), …, π(n)}, randomly select integers l and m that satisfy the condition l < m ≤ n, and swap the order of the l-th order and the m-th order in the current schedule π to obtain a new individual; Step 4.2.1.6: Repeat Step 4.1.5 until a set of Hummingbird individuals with a quantity of Popsize / 2 is generated; Step 4.2.1.7: Use the result of the improved NEH heuristic algorithm to directly generate Popsize / 2 duplicate Hummingbird individuals, and merge them with the individual set generated in Step 4.1.6 to obtain the initial population.

5. The intelligent scheduling method for a fast-response semiconductor packaging test workshop according to claim 4, wherein The specific operation of Step 4.2.4 is as follows: Step 4.2.4.1: Select the foraging method through the foraging judgment formula; By replacing the original 50% probability in AHA with a foraging judgment formula, the selection of guided foraging and territorial foraging is carried out; it is beneficial to increase the selection ratio of guided foraging in the early stage of iteration to enhance the ability to explore other spaces, and increase the selection of territorial foraging in the later stage of iteration to enhance the ability to find local optimal solutions; where GT is the foraging judgment coefficient; GT max and GT min are the maximum foraging judgment coefficient and the minimum foraging judgment coefficient respectively, and the value is Popsize is the population size, Popsize ≥mean is the number of individuals in the population whose fitness value is greater than or equal to the mean fitness value of the population; it is the iteration number; Iteration is the maximum number of iterations; v i (t + 1) is the position of the candidate food source of the i-th hummingbird at time t + 1; f Guided (v i,tar (t), x i (t)) is for guiding the foraging search, and x i (t) is the food source location of the i-th hummingbird at time t, and v i,tar (t) is the location of the target food source that the i-th hummingbird intends to visit; fTerritorial(x i (t)) is the territorial foraging search; Step 4.2.4.2: The hummingbird searches for a better food source through three flight methods; When the generated random number rand ≥ GT, the hummingbird chooses to conduct guided foraging. The three flight methods of guided foraging are: extracting a processing position to swap two orders, extracting multiple consecutive positions to swap two order blocks, and extracting multiple positions to swap multiple orders; When the generated random number rand < GT, the hummingbird chooses to conduct local foraging. The three flight methods of local foraging are: extracting two positions to swap two orders, extracting multiple consecutive positions to swap orders within a block, and extracting multiple positions to swap multiple orders; Step 4.2.4.3: Update the population and the access table; The position of the i-th food source is updated as follows: The fitness value represented by the hummingbird in the formula fitness(*). If the fitness of the candidate food source is lower than that of the current food source, the hummingbird abandons the current food source and stays at the newly generated candidate food source v i (t + 1) Feeding; update the access table.

6. The intelligent scheduling method for the fast-response semiconductor packaging test workshop according to claim 5, wherein The specific operation of Step 4.2.5 is as follows: Step 4.2.5.1: Calculate the number of times RS for each hummingbird to perform enhanced foraging by taking a random integer randint (randint ∈ [2, 3, 4, 5]) as the number of hummingbirds for enhanced foraging num (RS num = Popsize / randint); Step 4.2.5:2: Screen the local foraging hummingbirds for enhancement through roulette selection; Step 4.2.5.3: Perform RS on each selected hummingbird in sequence num times of enhanced taboo area foraging. The enhanced taboo area foraging method adds a taboo list on the basis of area foraging to ensure that the search results of each enhanced taboo area foraging are different; Step 4.2.5.4: Update the population and the access table.

7. The intelligent scheduling method for a fast-response semiconductor packaging test workshop according to claim 6, characterized in that, The specific operation of Step 4.2.6 is as follows: Step 4.2.6.1: Select the hummingbird with the worst fitness value and let this hummingbird migrate to the position of the hummingbird with the best fitness value that has been discovered; Step 4.2.6.2: Update the access table.

8. The intelligent scheduling method for a fast-response semiconductor packaging test workshop according to claim 4, wherein The specific operation of Step 5 is as follows: The heuristic rule library of the rule scheduling sub-module includes a heuristic processing order rule library and a heuristic equipment unit selection rule library; The processing order rule library includes rules such as the rule of giving priority to the shorter processing time, the rule of giving priority to the longer processing time, the rule of first come first served, and the rule of giving priority to the earlier due date of the process; The equipment unit selection rule library includes rules such as the rule of selecting the equipment unit with the shortest processing time, the rule of selecting the equipment unit with the shortest changeover time, the rule of selecting the equipment unit with the highest precision, the rule of selecting the equipment unit with the lowest precision, the rule of selecting the equipment unit with the most processable order types, and the rule of selecting the equipment unit with the fewest processable order types; Step 5.1: The user selects the corresponding heuristic rules for each process in advance; Step 5.2: Use the selected processing order rules and equipment unit selection rules for each process to generate an overall scheduling plan; Step 5.3: Transmit the generated global scheduling plan to the GUI module to arrange production.

9. An intelligent scheduling system for a fast-response semiconductor packaging and testing workshop, characterized in that, It includes three parts: a graphical user interface GUI module, a bottleneck identification module, and a scheduling module; It is used to complete the intelligent scheduling method for a fast-response semiconductor packaging and testing workshop described in any one of claims 1-8; The GUI module is used to transmit workshop operation data, workshop scheduling data and receive the global scheduling plan; The bottleneck identification module identifies all bottleneck processes through the "buffer-bottleneck index" bottleneck identification method; The scheduling module includes an intelligent scheduling sub-module and a rule-based scheduling sub-module. The intelligent scheduling sub-module solves the bottleneck process scheduling problem through an improved artificial hummingbird algorithm to generate a bottleneck scheduling plan. The rule-based scheduling sub-module generates a global scheduling plan by the user independently selecting rules from the rule library and combining the bottleneck scheduling plan.

10. The intelligent scheduling system for a fast-response semiconductor packaging test workshop according to claim 9, characterized in that It also includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it executes the "buffer-bottleneck index" bottleneck identification method of the bottleneck identification module, the IAHA algorithm of the intelligent scheduling sub-module in the scheduling module, and the rule-based scheduling sub-module in the scheduling module.

Images