SLT equipment optimization design method and system based on discrete event simulation

Abstract

The invention relates to the technical field of SLT equipment, in particular to an SLT equipment optimization design method and system based on discrete event simulation, which are applied to SLT equipment based on discrete event simulation, and the detection efficiency problem of the SLT equipment is jointly determined by resource configuration and action rhythm. The optimization design method is developed mainly around the following core links of problem definition and parameter extraction, simulation model construction, algorithm optimization, optimal parameter selection and verification and the like, so that reasonable resource allocation and action rhythm are solved under site constraint and mechanical structure movement speed constraint to obtain the SLT equipment detection efficiency as high as possible. According to the method, the defects of a traditional SLT equipment design mode in the aspects of resource configuration, design scheme evaluation, verification and the like are overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of SLT equipment, and in particular to a SLT equipment optimization design method and system based on discrete event simulation. Background Art

[0002] SLT, the full name of System Level Test, is a test method that verifies the functionality of a chip under conditions that simulate the actual terminal usage environment. SLT equipment does not need to create test vectors like traditional automatic test equipment (ATE), but instead completes the test by running and using the chip under test.

[0003] At present, in response to new design requirements for SLT equipment, the traditional design process mainly relies on drawing on historical plans to form a preliminary design plan. On this basis, designers will repeatedly modify and adjust the new plan, and rely on experience and a series of performance calculation formulas to verify whether it meets the design requirements. However, this traditional method has exposed several shortcomings in practice.

[0004] First, resource allocation and dynamic beat are two key factors that are usually considered separately during the design process. Designers often first determine a resource allocation scheme based on experience, and then debug the dynamic beat. This step-by-step approach may lead to a mismatch between resource allocation and dynamic beat, which in turn affects overall performance.

[0005] Secondly, the design scheme is mainly determined based on whether it meets the basic design requirements, but lacks a systematic and comprehensive evaluation standard. This leads to a large degree of blindness in the adjustment process of the design scheme, making it difficult to ensure the superiority and feasibility of the final scheme.

[0006] Finally, the verification method also has limitations. Traditional verification methods often cannot accurately simulate the complex situations of equipment in the real operating environment, which may lead to design defects being discovered only in actual applications, increasing the cost of subsequent modification and optimization.

[0007] In summary, the traditional SLT equipment design method has shortcomings in resource allocation, design scheme evaluation and verification, and it is necessary to explore more scientific and efficient design optimization methods to cope with increasingly complex and diverse design needs. Summary of the invention

[0008] One purpose of the present invention is to propose an SLT equipment optimization design method based on discrete event simulation to solve the deficiencies of traditional SLT equipment design methods in resource allocation, design scheme evaluation and verification, etc.

[0009] One purpose of the present invention is to propose an SLT equipment optimization design system based on discrete event simulation, which can further improve the design efficiency and performance level of SLT equipment by introducing advanced simulation technology, optimization algorithms and intelligent tools.

[0010] To achieve this object, the present invention adopts the following technical solutions:

[0011] An SLT equipment optimization design method based on discrete event simulation is applied to an SLT equipment based on discrete event simulation, wherein the detection efficiency of the SLT equipment is jointly determined by resource allocation and action rhythm;

[0012] The optimization design method comprises the following steps:

[0013] S1. According to the optimization objectives and constraints, determine the optimization problem definition corresponding to the detection efficiency problem of SLT equipment;

[0014] S2. According to the mechanical structure and optimization problem of SLT equipment, the resource allocation and action rhythm are quantitatively characterized through parameter setting, the core design parameters required for parameter selection are extracted, and the optimization problem model is established based on the optimization problem and the core design parameters;

[0015] S3. According to the mechanical structure of the SLT equipment, a discrete event simulation model is established, and parameter settings are performed on the discrete event simulation model, including taking the core design parameters as input and taking the optimization objectives corresponding to the core design parameters as output;

[0016] S4. Use GA-PSO hybrid algorithm to particle encode the core design parameters, generate an initialized particle swarm within the value range of the optimization problem model, input the discrete event simulation model for simulation analysis, and calculate the fitness of the initialized particle swarm and each particle in combination with the fitness function;

[0017] S5, performing selection, crossover and mutation operations on the initialized particle swarm to obtain a new particle swarm, inputting the discrete event simulation model for simulation analysis, and calculating the fitness of each new particle swarm and each particle in combination with the fitness function;

[0018] S6. Define the historical optimal fitness of a single particle as the local optimal value, and define the historical optimal fitness of the particle swarm as the global optimal value. After each generation of particle swarm completes the fitness calculation, the local optimal value and the global optimal value are updated.

[0019] S7, judging whether the local optimal value outputted by the iteration satisfies the termination condition, if not, returning to S5 to continue the iteration, if yes, jumping out of the loop, and selecting the particle with the largest historical optimal fitness as the optimal core design parameter;

[0020] S8. Use virtual debugging technology to apply the optimal core design parameters to the virtual SLT equipment for verification and debugging.

[0021] Preferably, in S1, the optimization target is the maximum number of chips detected when the SLT equipment works continuously for 8 hours;

[0022] The constraints are site constraints and mechanical structure movement speed constraints;

[0023] The optimization problem is defined as solving reasonable resource allocation and action rhythm to obtain the highest possible SLT equipment detection efficiency under site constraints and mechanical structure movement speed constraints.

[0024] Preferably, in S2, the mechanical structure of the SLT equipment includes k detection units, one detection unit includes a tray management unit and n chip detection units, one chip detection unit includes m chip detection mechanisms, wherein k>=1, n>=1, m>=2;

[0025] The tray management unit includes a tray management mechanism for testing, an empty tray management mechanism, four BIN tray management mechanisms, a transfer trolley and a lifting mechanism;

[0026] The chip detection unit includes five tray buffer mechanisms, a BIN division robot, two shuttles, a material change robot and multiple chip detection mechanisms;

[0027] A total of 8 core design parameters are extracted from the resource configuration and the action rhythm, including the number of detection units C1, the number of chip detection units C2, the number of chip detection mechanisms C3, the transfer trolley speed v1, the lifting mechanism speed v2, the BIN separation robot speed v3, the shuttle speed v4 and the material change robot speed v5;

[0028] Among them, C1=k, C2=k*m, C3=k*m*n, C1, C2 and C3 are all integer data, and v1, v2, v3, v4 and v5 are all floating-point data.

[0029] Preferably, in S2, the optimization problem model is as follows:

[0030] Max Z=f(C1,C2,C3,V1,V2,V3,V4,V5)

[0031]

[0032] Where Z represents the maximum number of chips detected by the SLT equipment when it works continuously for 8 hours, and W s , H s and L sRespectively represent the length, width and height of the SLT equipment site, W c Indicates the width of the detection unit, W r Indicates the width between adjacent detection units, H c Indicates the height spacing of the chip detection unit, H b Indicates the base height of the SLT equipment, L c Indicates the front and rear distance of the chip detection mechanism, L b Indicates the basic length of the chip detection unit.

[0033] Preferably, in S3, the following steps are specifically included:

[0034] S31. Use Desmo-J open source discrete event simulation engine to build a discrete event simulation model, including:

[0035] S311, extracting entities according to the mechanical structure of SLT equipment;

[0036] S312, extracting the events of each entity to form three event streams, namely, a chip detection event stream, a tray entry event stream, and a tray exit event stream;

[0037] S313, creating the life cycle of each entity to manage events, that is, setting the execution timing of events;

[0038] S32. Add counters and timers to the discrete event simulation model to set the output parameters, with the core design parameters as input and the maximum number of tests for the SLT equipment to work continuously for 8 hours, as well as the total of the waiting time for the disk to be tested, the waiting time for the BIN disk, the waiting time for the BIN robot and the waiting time for the material changing robot as output.

[0039] Preferably, in S4, the GA-PSO hybrid algorithm is used to perform particle encoding on the core design parameters and generate an initialization particle swarm, specifically comprising the following steps:

[0040] S41. Real number coding is used for particles, and the particle gene length is defined as 8 bits. The first to eighth bits correspond to eight core design parameters, namely the number of detection units C1, the number of chip detection units C2, the number of chip detection mechanisms C3, the transfer trolley speed v1, the lifting mechanism speed v2, the BIN division robot speed v3, the shuttle speed v4 and the material change robot speed v5; among which C1, C2 and C3 are all integer data, and v1, v2, v3, v4 and v5 are all floating point data;

[0041] S42, set the size of the particle group to 20, and arrange and combine the first three core design parameters of the particle gene length within their respective value ranges in the optimization problem model to generate a set of integer gene segments; arrange and combine the last five core design parameters of the particle gene length within their respective value ranges in the optimization problem model to take 5 equally divided points and arrange and combine to generate a size of 5 5 Floating point gene segment;

[0042] S43, randomly selecting 20 times from the group of integer gene segments and floating-point gene segments respectively, to form 20 complete codes, and generate an initialized particle swarm.

[0043] Preferably, in S4 and S5, the fitness function is as follows:

[0044]

[0045] Among them, M is the number of detections of the current individual, M max is the maximum number of detections in the current population, M min is the minimum number of tests that appear in the current population, T is the sum of the time for the current individual to wait for the test disk, the time for the BIN disk, the time for the BIN robot, and the time for the material change robot, T max is the maximum total waiting time in the current population, T min is the maximum sum of waiting times in the current population;

[0046] Normalize M to Reverse normalization of T is The two items have their respective weight coefficients ω1 and ω2, ω1+ω2=1, ω1=0.9, and ω2=0.1.

[0047] Preferably, in S5, performing selection, crossover and mutation operations on the initialized particle swarm to obtain a new particle swarm specifically includes the following steps:

[0048] S51, set the crossover probability to 0.6, that is, use the roulette method to select 60% of the particles in the current population for gene crossover to generate new individuals;

[0049] Among them, the crossover method adopts a discrete recombination method suitable for real number coding, that is, the same bit of the two parent individuals is selected with the same probability as the gene value corresponding to the child individual;

[0050] S52, set the mutation probability to a variable value, obey the function n is the number of particle swarm evolutions;

[0051] Among them, the particle decides whether to mutate based on the mutation probability. If it mutates, a gene is randomly selected for mutation. The mutation rule is to add or subtract one to the integer data within the respective value range in the optimization problem model, and multiply or divide the floating-point data by 1.1.

[0052] Preferably, in S6, the termination condition is that the fluctuation of the historical optimal fitness of the particle swarm of five consecutive generations does not exceed 5%.

[0053] A SLT equipment system based on discrete event simulation is designed using the above-mentioned optimization design method for SLT equipment based on discrete event simulation and is applied to a SLT equipment based on discrete event simulation, comprising:

[0054] An optimization problem definition module is used to determine the optimization problem definition corresponding to the detection efficiency problem of SLT equipment according to the optimization objectives and constraints;

[0055] Extract core design parameter module, which is used to quantitatively characterize resource allocation and action rhythm through parameterized settings according to the mechanical structure and optimization problem of SLT equipment, extract the core design parameters required for parameter optimization, and establish an optimization problem model based on the optimization problem and core design parameters;

[0056] Construct a simulation model module, which is used to establish a discrete event simulation model according to the mechanical structure of the SLT equipment, and perform parameter settings on the discrete event simulation model, including taking the core design parameters as input and taking the optimization objectives corresponding to the core design parameters as output;

[0057] The initialization particle swarm calculation module is used to use the GA-PSO hybrid algorithm to perform particle encoding on the core design parameters, generate the initialization particle swarm within the value range of the optimization problem model, input the discrete event simulation model for simulation analysis, and calculate the fitness of the initialization particle swarm and each particle in combination with the fitness function;

[0058] Iterative particle swarm calculation module is used to perform selection, crossover and mutation operations on the initialized particle swarm to obtain a new particle swarm, input the discrete event simulation model for simulation analysis, and calculate the fitness of each new particle swarm and each particle in combination with the fitness function;

[0059] Iterative optimization module, which is used to define the historical optimal fitness of a single particle as a local optimal value, and the historical optimal fitness of a particle swarm as a global optimal value. After each generation of particle swarm completes the fitness calculation, the local optimal value and the global optimal value are updated.

[0060] The judgment module is used to judge whether the local optimal value output by the iteration meets the termination condition. If not, it returns to S5 to continue the iteration. If yes, it jumps out of the loop and selects the particle with the largest historical optimal fitness as the optimal core design parameter.

[0061] The verification and debugging module is used to use virtual debugging technology to apply the optimal core design parameters to the virtual SLT equipment for verification and debugging.

[0062] One of the above technical solutions has the following beneficial effects:

[0063] 1. Improve detection efficiency: By optimizing resource allocation and action rhythm, the detection efficiency of SLT equipment can be significantly improved, thereby improving overall production efficiency and product quality.

[0064] 2. Reduce design costs: Use discrete event simulation and GA-PSO hybrid algorithm for optimized design to avoid the high design and time costs brought by traditional trial and error methods.

[0065] 3. Enhanced design flexibility: Ability to handle multiple optimization objectives and constraints, providing greater flexibility and customizability for the design of SLT equipment.

[0066] 4. Improve design reliability: Use virtual debugging technology to verify and debug the optimal core design parameters to ensure that the actual effect of the optimized design is consistent with expectations and improve the reliability and stability of the design. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a flow chart of a SLT equipment optimization design method based on discrete event simulation of the present invention;

[0068] Figure 2 It is a schematic diagram of the operations of particle selection, crossover and mutation in a SLT equipment optimization design method based on discrete event simulation of the present invention;

[0069] Figure 3 It is a structural schematic diagram of a SLT device based on discrete event simulation of the present invention;

[0070] Figure 4 It is a flow chart of a chip detection event flow in SLT equipment based on discrete event simulation of the present invention;

[0071] Figure 5 It is a flow chart of a tray entry event flow in SLT equipment based on discrete event simulation of the present invention;

[0072] Figure 6 The present invention is a flowchart of a tray out event flow in an SLT device based on discrete event simulation. DETAILED DESCRIPTION

[0073] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.

[0074] An SLT equipment optimization design method based on discrete event simulation is applied to an SLT equipment based on discrete event simulation, wherein the detection efficiency of the SLT equipment is jointly determined by resource allocation and action rhythm;

[0075] The optimization design method comprises the following steps:

[0076] S1. According to the optimization objectives and constraints, determine the optimization problem definition corresponding to the detection efficiency problem of SLT equipment;

[0077] S2. According to the mechanical structure and optimization problem of SLT equipment, the resource allocation and action rhythm are quantitatively characterized through parameter setting, the core design parameters required for parameter selection are extracted, and the optimization problem model is established based on the optimization problem and the core design parameters;

[0078] S3. According to the mechanical structure of the SLT equipment, a discrete event simulation model is established, and parameter settings are performed on the discrete event simulation model, including taking the core design parameters as input and taking the optimization objectives corresponding to the core design parameters as output;

[0079] S4. Use GA-PSO hybrid algorithm to particle encode the core design parameters, generate an initialized particle swarm within the value range of the optimization problem model, input the discrete event simulation model for simulation analysis, and calculate the fitness of the initialized particle swarm and each particle in combination with the fitness function;

[0080] S5, performing selection, crossover and mutation operations on the initialized particle swarm to obtain a new particle swarm, inputting the discrete event simulation model for simulation analysis, and calculating the fitness of each new particle swarm and each particle in combination with the fitness function;

[0081] S6. Define the historical optimal fitness of a single particle as the local optimal value, and define the historical optimal fitness of the particle swarm as the global optimal value. After each generation of particle swarm completes the fitness calculation, the local optimal value and the global optimal value are updated.

[0082] S7, judging whether the local optimal value outputted by the iteration satisfies the termination condition, if not, returning to S5 to continue the iteration, if yes, jumping out of the loop, and selecting the particle with the largest historical optimal fitness as the optimal core design parameter;

[0083] S8. Use virtual debugging technology to apply the optimal core design parameters to the virtual SLT equipment for verification and debugging.

[0084] like Figure 1 As shown in Figure 1, this method focuses on solving the detection efficiency problem determined by resource allocation and action rhythm. The specific steps of this method are as follows:

[0085] S1. Define the optimization problem: First, based on the optimization goal and actual constraints, accurately define the optimization requirements of the SLT equipment detection efficiency problem. This step provides a clear direction and framework for subsequent work.

[0086] S2. Extract core design parameters: In-depth analysis of the mechanical structure of the SLT equipment, combined with the characteristics of the optimization problem, quantitative description of resource allocation and action rhythm through parameterized settings. On this basis, the core design parameters that have a significant impact on the detection efficiency are extracted, and the optimization problem model is constructed based on these parameters.

[0087] S3. Construct simulation model: According to the mechanical structure characteristics of SLT equipment, a discrete event simulation model is established. The discrete event simulation model needs to accurately reflect the actual working conditions of SLT equipment, including the dynamic changes of resource allocation and action rhythm. At the same time, the discrete event simulation model is parameterized to ensure that the core design parameters can be used as input and the corresponding optimization target values ​​can be output.

[0088] S4, Initialize particle swarm and perform selection, crossover and mutation operations: Use GA-PSO hybrid algorithm to encode core design parameters and generate initialization particle swarm within the value range of the optimization problem model to ensure that the above constraints are met. These initialization particle swarms represent different combinations of design parameters, providing a basis for subsequent simulation analysis and optimization.

[0089] S5. Simulation analysis and fitness calculation: The initialized particle swarm and the new particle swarm are input into the discrete event simulation model for simulation analysis, and the fitness of each particle and its particle swarm is calculated in combination with the fitness function. The fitness value reflects the optimization degree of the SLT equipment detection efficiency under the current design parameter combination.

[0090] S6. Update the optimal value: After each generation of particle swarm completes the fitness calculation, update the local optimal value and the global optimal value. The local optimal value refers to the historical optimal fitness of a single particle, while the global optimal value refers to the historical optimal fitness of the entire particle swarm.

[0091] S7, iterative optimization: perform selection, crossover and mutation operations on the initialized particle swarm to generate new particle swarms. These new particle swarms inherit the excellent genes of the previous generation of particles and introduce new mutations to increase the diversity of the search space. The new particle swarm is input into the simulation model again for simulation analysis and the fitness value is calculated. Repeat this process until the fitness value of the iterative output meets the preset termination condition.

[0092] S8, Verification and debugging: Finally, using virtual debugging technology, the optimal core design parameters are applied to the virtual SLT equipment for verification and debugging, that is, the real control program drives the virtual SLT equipment, and the machinery and control are constantly adjusted and verified in this process. Since the virtual SLT equipment is highly consistent with the real SLT equipment in layout, action and control points, the debugging results are effective. Compared with traditional real machine debugging, the development cycle is shorter, the cost is lower, and the effect is better. This step ensures that the actual effect of the optimized design is consistent with expectations, providing reliable guarantees for subsequent practical applications.

[0093] In summary, this optimization design method mainly revolves around the following core links such as problem definition and parameter extraction, simulation model construction, algorithm optimization, and optimal parameter selection and verification, and has the following beneficial effects:

[0094] 1. Improve detection efficiency: By optimizing resource allocation and action rhythm, the detection efficiency of SLT equipment can be significantly improved, thereby improving overall production efficiency and product quality.

[0095] 2. Reduce design costs: Use discrete event simulation and GA-PSO hybrid algorithm for optimized design to avoid the high design and time costs brought by traditional trial and error methods.

[0096] 3. Enhanced design flexibility: Ability to handle multiple optimization objectives and constraints, providing greater flexibility and customizability for the design of SLT equipment.

[0097] 4. Improve design reliability: Use virtual debugging technology to verify and debug the optimal core design parameters to ensure that the actual effect of the optimized design is consistent with expectations and improve the reliability and stability of the design.

[0098] In summary, this optimization design method can systematically solve the detection efficiency problem of SLT equipment, improve the overall performance of the equipment, reduce design costs, and enhance design flexibility and reliability.

[0099] To further illustrate, in S1, the optimization target is the maximum number of chips detected when the SLT equipment works continuously for 8 hours;

[0100] The constraints are site constraints and mechanical structure movement speed constraints;

[0101] The optimization problem is defined as solving reasonable resource allocation and action rhythm to obtain the highest possible SLT equipment detection efficiency under site constraints and mechanical structure movement speed constraints.

[0102] First, in the S1 stage, the optimization goal is to set the maximum number of chips that can be detected by the SLT equipment within 8 hours of continuous operation. The setting of this optimization goal is directly related to the production efficiency of the SLT equipment and the usage needs of customers.

[0103] Next, consider two main constraints: site constraints and mechanical structure movement speed constraints. Site constraints are due to the customer's strict requirements on the size of the SLT equipment, which limits the overall layout of the equipment and the possibility of resource allocation. Mechanical structure movement speed constraints are due to the physical characteristics and movement mechanism of the SLT equipment, which determines the upper limit of the speed of the equipment when performing actions such as pallet transfer and chip transfer.

[0104] After clarifying the optimization goals and constraints, the optimization problem is defined as: Under the constraints of the site and the movement speed of the mechanical structure, solve the reasonable resource allocation and action rhythm to obtain the highest possible SLT equipment detection efficiency. The resource allocation here involves the allocation and layout of resources such as trays, chips, and test equipment, while the action rhythm refers to the timing and rhythm of the SLT equipment when performing various actions.

[0105] In order to solve this optimization problem, the subsequent steps (S2 to S8) iteratively optimize the core design parameters through parameterization settings, the establishment of discrete event simulation models, the application of GA-PSO hybrid algorithms, and virtual debugging technology, until the resource configuration and action beat combination that meets the constraints and has the highest detection efficiency is found.

[0106] Further explanation, in S2, the mechanical structure of the SLT equipment includes k detection units, one detection unit includes a tray management unit and n chip detection units, one chip detection unit includes m chip detection mechanisms, wherein k>=1, n>=1, m>=2;

[0107] The tray management unit includes a tray management mechanism for testing, an empty tray management mechanism, four BIN tray management mechanisms, a transfer trolley and a lifting mechanism;

[0108] The chip detection unit includes five tray buffer mechanisms, a BIN division robot, two shuttles, a material change robot and multiple chip detection mechanisms;

[0109] A total of 8 core design parameters are extracted from the resource configuration and the action rhythm, including the number of detection units C1, the number of chip detection units C2, the number of chip detection mechanisms C3, the transfer trolley speed v1, the lifting mechanism speed v2, the BIN separation robot speed v3, the shuttle speed v4 and the material change robot speed v5;

[0110] Among them, C1=k, C2=k*m, C3=k*m*n, C1, C2 and C3 are all integer data, and v1, v2, v3, v4 and v5 are all floating-point data.

[0111] Specifically, Figure 3 As shown in the figure, the tray management unit has two tasks: one is to transport the trays to be tested or empty trays to the chip detection unit, and the other is to recover BIN (the chips will be divided into 4 levels after the test, BIN refers to the level) trays or empty trays from the chip detection unit. The chip detection unit is responsible for testing the chips to be tested and sorting the chips into BIN after the test is completed.

[0112] Therefore, the optimization problem of SLT equipment is to improve the transfer efficiency of transfer pallets and chips under the constraints of site constraints and mechanical structure movement speed. Among them, site constraints also mean constraints on the composition of SLT equipment. Therefore, the parameters related to resource configuration are extracted as the setting quantity of each mechanical structure. It is important that the transfer trolley and the lifting mechanism are jointly responsible for the transfer of pallets, and the BIN manipulator, shuttle and material change manipulator are jointly responsible for the transfer of chips. The reasonable coordination of the speed of each transfer component can make the equipment beat more reasonable and thus improve the detection efficiency. The selection of these core design parameters is based on their direct impact on the detection efficiency, and they can be adjusted and optimized in the subsequent simulation and optimization process.

[0113] To further illustrate, in S2, the optimization problem model is as follows:

[0114] Max Z=f(C1,C2,C3,V1,V2,V3,V4,V5)

[0115]

[0116] Where Z represents the maximum number of chips detected by the SLT equipment when it works continuously for 8 hours, and W s , H s and L s Respectively represent the length, width and height of the SLT equipment site, W c Indicates the width of the detection unit, W r Indicates the width between adjacent detection units, H c Indicates the height spacing of the chip detection unit, H b Indicates the base height of the SLT equipment, L c Indicates the front and rear distance of the chip detection mechanism, L b Indicates the basic length of the chip detection unit.

[0117] Specifically, an optimization problem model was established using the core design parameters obtained in the above steps. The goal of this optimization problem model is to maximize the detection efficiency of the SLT equipment by adjusting these core design parameters under the constraints of site constraints and mechanical structure movement speed. Site constraints limit the overall layout and size of the SLT equipment, while mechanical structure movement speed constraints limit the upper speed limit of each transfer component.

[0118] To further explain, in S3, the following steps are specifically included:

[0119] It is known that Desmo-j is an open source discrete event simulation engine based on Java language. Discrete event simulation can highly restore the actual execution process of SLT equipment. The state of SLT equipment changes at discrete time points, and the advancement of simulation is based on the occurrence of events.

[0120] S31. Use Desmo-J open source discrete event simulation engine to build a discrete event simulation model, including:

[0121] S311. Extract entities according to the mechanical structure of the SLT equipment; known entities are elements that can exist independently in the discrete event simulation model and represent individual objects in the real world. Specifically, from the mechanical structure of the SLT equipment, it can be seen that the entities extracted from the pallet management unit include 1 test tray management mechanism, 1 empty tray management mechanism, 4 BIN tray management mechanisms, 1 transfer and 1 lifting mechanism; the entities extracted from the chip detection unit include 5 pallet cache mechanisms, 1 BIN separation robot, 2 shuttles, 1 material change robot and multiple chip detection mechanisms.

[0122] S312. Extract the events of each entity to form three event flows, namely chip detection event flow, pallet entry event flow and pallet exit event flow; known events are the driving force of discrete event simulation. The essence of discrete event simulation execution is the end of old events and the occurrence of new events, which are called event flows. There are pallet disassembly events in the test pallet management mechanism, pallet disassembly and pallet encoding events in the empty pallet management mechanism, pallet encoding events in the BIN pallet management mechanism, and movement events in the transfer trolley and lifting mechanism; there are pallet entry and pallet exit events in the pallet cache mechanism, and there are chip retrieval events, chip loading events and BIN separation events in the BIN sorting robot, movement events in the shuttle, and events of going to the target detection mechanism and chip replacement in the material change robot. There are chip detection events in the chip detection mechanism. Figure 4-6 As shown in the figure, the right-angled rectangle represents the event. The events of each entity of the SLT equipment form three event flows, namely the chip detection event flow, the tray entry event flow and the tray exit event flow.

[0123] S313, create the life cycle of each entity to manage events, that is, set the execution time of the event; Figure 3 As shown, taking the shuttle as an example, when the chip detection mechanism is in the state of waiting for material replacement, it waits for itself to be idle. After the BIN robot loads the chip to be tested, the shuttle starts to move. After it is in place, it waits for the material replacement robot to complete the chip replacement event. After the shuttle moves back to its original position, it waits for the BIN event to end and ends a life cycle. The creation of the life cycle of other entities is similar. In the Demo-j simulation software, it can be achieved by inheriting the SimProcess class and writing the life cycle in the lifeCycle() function.

[0124] S32. Add counters and timers to the discrete event simulation model to set the output parameters, with the core design parameters as input and the maximum number of tests for the SLT equipment to work continuously for 8 hours, as well as the sum of the waiting time for the disk to be tested, the waiting time for the BIN disk, the waiting time for the BIN manipulator, and the waiting time for the material change manipulator as output. The discrete event simulation model is parameterized to obtain better versatility. Due to the significant modular characteristics of the SLT equipment structure, this parameterization can be achieved through simple programming.

[0125] To further illustrate, in S4, the GA-PSO hybrid algorithm is used to perform particle encoding on the core design parameters and generate an initialization particle swarm, which specifically includes the following steps:

[0126] S41. Real number coding is used for particles, and the particle gene length is defined as 8 bits. The first to eighth bits correspond to eight core design parameters, namely the number of detection units C1, the number of chip detection units C2, the number of chip detection mechanisms C3, the transfer trolley speed v1, the lifting mechanism speed v2, the BIN division robot speed v3, the shuttle speed v4 and the material change robot speed v5; among which C1, C2 and C3 are all integer data, and v1, v2, v3, v4 and v5 are all floating point data;

[0127] S42, set the size of the particle group to 20, and arrange and combine the first three core design parameters of the particle gene length within their respective value ranges in the optimization problem model to generate a set of integer gene segments; arrange and combine the last five core design parameters of the particle gene length within their respective value ranges in the optimization problem model to take 5 equally divided points and arrange and combine to generate a size of 5 5 Floating point gene segment;

[0128] S43, randomly selecting 20 times from the group of integer gene segments and floating-point gene segments respectively, to form 20 complete codes, and generate an initialized particle swarm.

[0129] It is known that the basic idea of ​​the GA-PSO hybrid algorithm is to significantly improve the efficiency and accuracy of the solution by combining the global search capability of GA and the local search capability of PSO. Specifically, GA helps local search and increases the diversity of the population through mutation operations in the early stage of evolution, while in the later stage of evolution, the mutation operator may destroy the already stable population. PSO can maintain stability during the evolution process by maintaining the historical state and the current state through particles. By introducing the crossover operation of GA in PSO and reconstructing the mutation operator, a behavior pattern that manifests itself as PSO in the macro sense and GA in the micro sense is formed.

[0130] To further illustrate, in S4 and S5, the fitness function is as follows:

[0131]

[0132] Among them, M is the number of detections of the current individual, M max is the maximum number of detections in the current population, M min is the minimum number of tests that appear in the current population, T is the sum of the time for the current individual to wait for the test disk, the time for the BIN disk, the time for the BIN robot, and the time for the material change robot, T max is the maximum total waiting time in the current population, T min is the maximum sum of waiting times in the current population; since the number of tests and the waiting time have different dimensions, and the number of tests is maximized while the waiting time is minimized, M is normalized to Reverse normalization of T is The two items have their respective weight coefficients ω1 and ω2, ω1+ω2=1, ω1=0.9, and ω2=0.1.

[0133] To further illustrate, in S5, performing selection, crossover and mutation operations on the initialized particle swarm to obtain a new particle swarm specifically includes the following steps:

[0134] S51, set the crossover probability to 0.6, that is, use the roulette method to select 60% of the particles in the current population for gene crossover to generate new individuals;

[0135] Among them, the crossover method adopts a discrete recombination method suitable for real number coding, that is, the same bit of the two parent individuals is selected with the same probability as the gene value corresponding to the child individual.

[0136] S52, set the mutation probability to a variable value, obey the function n is the number of particle swarm evolutions;

[0137] Among them, the particle decides whether to mutate based on the mutation probability. If it mutates, a gene is randomly selected for mutation. The mutation rule is to add or subtract one to the integer data within the respective value range in the optimization problem model, and multiply or divide the floating-point data by 1.1.

[0138] In order to enhance the diversity of individuals, such as Figure 2 As shown, the particle codes are crossed and mutated to complete the update of the particle positions (particle codes). The particles participating in the crossover and mutation need to be selected from the current particle group. The selection method can be a common roulette method, which is a random selection method. The higher the fitness value of the particle, the higher the probability of selection.

[0139] Further explanation, in S6, the termination condition is that the fluctuation of the historical optimal fitness of the continuous five adjacent generations of particle swarms does not exceed 5%. Meeting the termination condition means that even if the selection, crossover and mutation operations are repeated, the probability of a better fitness is already very small, so it can be considered that the best particle is found in the historical optimal fitness of the continuous five adjacent generations of particle swarms, that is, the optimal core design parameters.

[0140] A SLT equipment system based on discrete event simulation is designed using the above-mentioned optimization design method for SLT equipment based on discrete event simulation and is applied to a SLT equipment based on discrete event simulation, comprising:

[0141] An optimization problem definition module is used to determine the optimization problem definition corresponding to the detection efficiency problem of SLT equipment according to the optimization objectives and constraints;

[0142] Extract core design parameter module, which is used to quantitatively characterize resource allocation and action rhythm through parameterized settings according to the mechanical structure and optimization problem of SLT equipment, extract the core design parameters required for parameter optimization, and establish an optimization problem model based on the optimization problem and core design parameters;

[0143] Construct a simulation model module, which is used to establish a discrete event simulation model according to the mechanical structure of the SLT equipment, and perform parameter settings on the discrete event simulation model, including taking the core design parameters as input and taking the optimization objectives corresponding to the core design parameters as output;

[0144] The initialization particle swarm calculation module is used to use the GA-PSO hybrid algorithm to perform particle encoding on the core design parameters, generate the initialization particle swarm within the value range of the optimization problem model, input the discrete event simulation model for simulation analysis, and calculate the fitness of the initialization particle swarm and each particle in combination with the fitness function;

[0145] Iterative particle swarm calculation module is used to perform selection, crossover and mutation operations on the initialized particle swarm to obtain a new particle swarm, input the discrete event simulation model for simulation analysis, and calculate the fitness of each new particle swarm and each particle in combination with the fitness function;

[0146] Iterative optimization module, which is used to define the historical optimal fitness of a single particle as a local optimal value, and the historical optimal fitness of a particle swarm as a global optimal value. After each generation of particle swarm completes the fitness calculation, the local optimal value and the global optimal value are updated.

[0147] The judgment module is used to judge whether the local optimal value output by the iteration meets the termination condition. If not, it returns to S5 to continue the iteration. If yes, it jumps out of the loop and selects the particle with the largest historical optimal fitness as the optimal core design parameter.

[0148] The verification and debugging module is used to use virtual debugging technology to apply the optimal core design parameters to the virtual SLT equipment for verification and debugging.

[0149] The technical principle of the present invention is described above in conjunction with specific embodiments. These descriptions are only for explaining the principle of the present invention and cannot be interpreted as limiting the scope of protection of the present invention in any way. Based on the explanations herein, those skilled in the art can associate other specific embodiments of the present invention without creative work, and these equivalent variations or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A SLT equipment optimization design method based on discrete event simulation, characterized in that: Applied to a SLT equipment based on discrete event simulation, the detection efficiency of the SLT equipment is jointly determined by resource allocation and action rhythm; The optimization design method comprises the following steps: S1. According to the optimization objectives and constraints, determine the optimization problem definition corresponding to the detection efficiency problem of SLT equipment; S2. According to the mechanical structure and optimization problem of SLT equipment, the resource allocation and action rhythm are quantitatively characterized through parameter setting, the core design parameters required for parameter selection are extracted, and the optimization problem model is established based on the optimization problem and the core design parameters; S3. According to the mechanical structure of the SLT equipment, a discrete event simulation model is established, and parameter settings are performed on the discrete event simulation model, including taking the core design parameters as input and taking the optimization objectives corresponding to the core design parameters as output; S4. Use GA-PSO hybrid algorithm to particle encode the core design parameters, generate an initialized particle swarm within the value range of the optimization problem model, input the discrete event simulation model for simulation analysis, and calculate the fitness of the initialized particle swarm and each particle in combination with the fitness function; S5, performing selection, crossover and mutation operations on the initialized particle swarm to obtain a new particle swarm, inputting the discrete event simulation model for simulation analysis, and calculating the fitness of each new particle swarm and each particle in combination with the fitness function; S6. Define the historical optimal fitness of a single particle as the local optimal value, and define the historical optimal fitness of the particle swarm as the global optimal value. After each generation of particle swarm completes the fitness calculation, the local optimal value and the global optimal value are updated. S7, judging whether the local optimal value outputted by the iteration satisfies the termination condition, if not, returning to S5 to continue the iteration, if yes, jumping out of the loop, and selecting the particle with the largest historical optimal fitness as the optimal core design parameter; S8. Use virtual debugging technology to apply the optimal core design parameters to the virtual SLT equipment for verification and debugging.

2. The SLT equipment optimization design method based on discrete event simulation according to claim 1 is characterized in that: In S1, the optimization target is the maximum number of chips detected when the SLT equipment works continuously for 8 hours; The constraints are site constraints and mechanical structure movement speed constraints; The optimization problem is defined as solving reasonable resource allocation and action rhythm to obtain the highest possible SLT equipment detection efficiency under site constraints and mechanical structure movement speed constraints.

3. The SLT equipment optimization design method based on discrete event simulation according to claim 2 is characterized in that: In S2, the mechanical structure of the SLT equipment includes k detection units, one detection unit includes a tray management unit and n chip detection units, one chip detection unit includes m chip detection mechanisms, wherein k>=1, n>=1, m>=2; The tray management unit includes a tray management mechanism for testing, an empty tray management mechanism, four BIN tray management mechanisms, a transfer trolley and a lifting mechanism; The chip detection unit includes five tray buffer mechanisms, a BIN division robot, two shuttles, a material change robot and multiple chip detection mechanisms; A total of 8 core design parameters are extracted from the resource configuration and the action rhythm, including the number of detection units C1, the number of chip detection units C2, the number of chip detection mechanisms C3, the transfer trolley speed v1, the lifting mechanism speed v2, the BIN separation robot speed v3, the shuttle speed v4 and the material change robot speed v5; Among them, C1=k, C2=k*m, C3=k*m*n, C1, C2 and C3 are all integer data, and v1, v2, v3, v4 and v5 are all floating-point data.

4. The SLT equipment optimization design method based on discrete event simulation according to claim 3 is characterized in that: In S2, the optimization problem model is as follows: Max Z=f(C1,C2,C3,V1,V2,V3,V4,V5) Where Z represents the maximum number of chips detected by the SLT equipment when it works continuously for 8 hours, and W s , H s and L s Respectively represent the length, width and height of the SLT equipment site, W c Indicates the width of the detection unit, W r Indicates the width between adjacent detection units, H c Indicates the height spacing of the chip detection unit, H b Indicates the base height of the SLT equipment, L c Indicates the front and rear distance of the chip detection mechanism, L b Indicates the basic length of the chip detection unit.

5. The SLT equipment optimization design method based on discrete event simulation according to claim 4 is characterized in that: In S3, the following steps are specifically included: S31. Use Desmo-J open source discrete event simulation engine to build a discrete event simulation model, including: S311, extracting entities according to the mechanical structure of SLT equipment; S312, extracting the events of each entity to form three event streams, namely, a chip detection event stream, a tray entry event stream, and a tray exit event stream; S313, creating the life cycle of each entity to manage events, that is, setting the execution timing of events; S32. Add counters and timers to the discrete event simulation model to set the output parameters, with the core design parameters as input and the maximum number of tests for the SLT equipment to work continuously for 8 hours, as well as the total of the waiting time for the disk to be tested, the waiting time for the BIN disk, the waiting time for the BIN robot and the waiting time for the material changing robot as output.

6. The SLT equipment optimization design method based on discrete event simulation according to claim 5, characterized in that: In S4, the GA-PSO hybrid algorithm is used to perform particle encoding on the core design parameters and generate an initialization particle swarm, which specifically includes the following steps: S41. Real number coding is used for particles, and the particle gene length is defined as 8 bits. The first to eighth bits correspond to eight core design parameters, namely the number of detection units C1, the number of chip detection units C2, the number of chip detection mechanisms C3, the transfer trolley speed v1, the lifting mechanism speed v2, the BIN division robot speed v3, the shuttle speed v4 and the material change robot speed v5; among which C1, C2 and C3 are all integer data, and v1, v2, v3, v4 and v5 are all floating point data; S42, set the size of the particle group to 20, and arrange and combine the first three core design parameters of the particle gene length within their respective value ranges in the optimization problem model to generate a set of integer gene segments; arrange and combine the last five core design parameters of the particle gene length within their respective value ranges in the optimization problem model to obtain 5 equally divided points and arrange and combine to generate a size of 5 5 Floating point gene segment; S43, randomly selecting 20 times from the group of integer gene segments and floating-point gene segments respectively, to form 20 complete codes, and generate an initialized particle swarm.

7. The SLT equipment optimization design method based on discrete event simulation according to claim 6 is characterized in that: In S4 and S5, the fitness function is as follows: Among them, M is the number of detections of the current individual, M max is the maximum number of detections in the current population, M min is the minimum number of tests that appear in the current population, T is the sum of the time for the current individual to wait for the test disk, the time for the BIN disk, the time for the BIN robot, and the time for the material change robot, T max is the maximum total waiting time in the current population, T min is the maximum sum of waiting times in the current population; Normalize M to Reverse normalization of T is The two items have their respective weight coefficients ω1 and ω2, ω1+ω2=1, ω1=0.9, and ω2=0.

1.

8. The SLT equipment optimization design method based on discrete event simulation according to claim 7 is characterized in that: In S5, the selection, crossover and mutation operations are performed on the initialized particle swarm to obtain a new particle swarm, which specifically includes the following steps: S51, set the crossover probability to 0.6, that is, use the roulette method to select 60% of the particles in the current population for gene crossover to generate new individuals; Among them, the crossover method adopts a discrete recombination method suitable for real number coding, that is, the same bit of the two parent individuals is selected with the same probability as the gene value corresponding to the child individual; S52, set the mutation probability to a variable value, obey the function n is the number of particle swarm evolutions; Among them, the particle decides whether to mutate based on the mutation probability. If it mutates, a gene is randomly selected for mutation. The mutation rule is to add or subtract one to the integer data within the respective value range in the optimization problem model, and multiply or divide the floating-point data by 1.

1.

9. The SLT equipment optimization design method based on discrete event simulation according to claim 8, characterized in that: In S6, the termination condition is that the fluctuation of the historical optimal fitness of the particle swarm in five consecutive generations does not exceed 5%.

10. An SLT equipment system based on discrete event simulation, characterized in that: The optimization design method of SLT equipment based on discrete event simulation according to any one of claims 1 to 9 is applied to SLT equipment based on discrete event simulation, comprising: An optimization problem definition module is used to determine the optimization problem definition corresponding to the detection efficiency problem of SLT equipment according to the optimization objectives and constraints; Extract core design parameter module, which is used to quantitatively characterize resource allocation and action rhythm through parameterized settings according to the mechanical structure and optimization problem of SLT equipment, extract the core design parameters required for parameter optimization, and establish an optimization problem model based on the optimization problem and core design parameters; Construct a simulation model module, which is used to establish a discrete event simulation model according to the mechanical structure of the SLT equipment, and perform parameter settings on the discrete event simulation model, including taking the core design parameters as input and taking the optimization objectives corresponding to the core design parameters as output; The initialization particle swarm calculation module is used to use the GA-PSO hybrid algorithm to perform particle encoding on the core design parameters, generate the initialization particle swarm within the value range of the optimization problem model, input the discrete event simulation model for simulation analysis, and calculate the fitness of the initialization particle swarm and each particle in combination with the fitness function; Iterative particle swarm calculation module is used to perform selection, crossover and mutation operations on the initialized particle swarm to obtain a new particle swarm, input the discrete event simulation model for simulation analysis, and calculate the fitness of each new particle swarm and each particle in combination with the fitness function; Iterative optimization module, which is used to define the historical optimal fitness of a single particle as a local optimal value, and the historical optimal fitness of a particle swarm as a global optimal value. After each generation of particle swarm completes the fitness calculation, the local optimal value and the global optimal value are updated. The judgment module is used to judge whether the local optimal value output by the iteration meets the termination condition. If not, it returns to S5 to continue the iteration. If yes, it jumps out of the loop and selects the particle with the largest historical optimal fitness as the optimal core design parameter. The verification and debugging module is used to use virtual debugging technology to apply the optimal core design parameters to the virtual SLT equipment for verification and debugging.

Images