Multi-target dynamic weight adaptive scheduling method for flexible verification pipeline of metrological equipment

Abstract

The application discloses a flexible verification pipeline multi-target dynamic weight adaptive scheduling method of a metrological device and belongs to the technical field of industrial intelligent scheduling. The method comprises the following steps: establishing a capacity model of N (N is an integer and belongs to [3, 20]) verification lines; collecting working condition indexes at three frequency periods of 10 kHz (electric parameters), 1 Hz (temperature) and 20 kHz (vibration) and calculating a comprehensive health degree score H i (t) according to an order capacity ratio R m Adaptively determining a four-target weight vector composed of a completion time weight w1, a device utilization rate weight w2, a switching cost weight w3 and a health risk weight w4 in three preset intervals; constructing a four-target optimization model by using allocation decision variables x ij ∈ {0, 1} and sequencing decision variables y jj′ ∈ {0, 1}, solving a Pareto front solution set by using an improved non-dominated sorting genetic algorithm under five types of constraints; automatically screening representative solutions from the Pareto front and providing the representative solutions for interactive selection of a scheduler in a visual mode. Simulation verification shows that the maximum completion time is shortened by 24.3% and the switching times are reduced by 61.1% compared with manual production scheduling.

Description

Technical Field

[0001] This invention belongs to the field of industrial intelligent scheduling technology, specifically involving a multi-objective dynamic weight adaptive scheduling method for flexible calibration production lines of metering equipment. Background Technology

[0002] With the continuous advancement of the power Internet of Things (IoT) construction, the demand for centralized verification of metering equipment such as electricity meters and data acquisition terminals is rapidly increasing. Taking a provincial-level metering center as an example, the annual verification volume has exceeded 6.6 million units, covering four major categories and more than 190 specifications, including single-phase smart meters, three-phase direct-connection meters, three-phase current transformer-connected meters, and low-voltage current transformers. This metering center is equipped with 11 verification lines (6 single-phase lines and 5 three-phase lines). The lines differ significantly in terms of equipment type compatibility, number of meter positions, and daily verification batches, forming a typical heterogeneous parallel system. The monthly verification volume varies from approximately 137,000 units in the off-season to approximately 704,000 units in the peak season, a fluctuation of more than five times.

[0003] To address the issue of flexible scheduling, existing technologies offer several solutions: CN202411993841 allocates tasks to conveyor line processes using a cache priority principle. This method is effective in single-line scenarios, but when the number of calibration lines expands to 11, the cache rule struggles to achieve global load balancing; CN202510837047 introduces dynamic metering data into multi-line scheduling, which touches upon the application scenario of this invention. However, its weighting system does not adjust with changes in production capacity load. When encountering a 5-fold fluctuation between peak and off-peak seasons, the fixed weighting scheme will continuously push tasks onto high-load lines; WO2016138737A1 quantifies the maintenance priority of the electricity meter calibration line, but the priority calculation results are only used as a post-scheduling reference and are not incorporated into the real-time scheduling objective function.

[0004] From an academic perspective, the convergence of NSGA-II in the dual-objective (completion time × utilization rate) scenario has been fully verified; reinforcement learning (DQN / PPO) responds quickly to online policy adjustments, but the training process requires a large amount of historical data, and cold-start convergence is slow, raising doubts about its practical applicability in the seasonal switching scenario of metrology verification centers.

[0005] In summary, existing technologies have the following shortcomings when applied to the scheduling of metering equipment calibration production lines: (1) Single objective or fixed objective weight. Most existing methods only focus on the single objective of minimizing the completion time. Even if multi-objective optimization is adopted, the objective weight is a fixed value set in advance, which cannot adapt to the dynamic production scenario where the monthly inspection quantity fluctuates by more than 5 times.

[0006] (2) Ignoring equipment health status. Existing scheduling methods treat the calibration line as a constant capacity resource, without considering the deterioration of the calibration equipment's operating conditions and the risk of failure. Annual standard equipment inspection (about 2 weeks of downtime), quarterly planned maintenance (about 2 days each time), and random failures all affect the effective capacity of the calibration line.

[0007] (3) Difficulty in selecting Pareto solution set. Traditional multi-objective evolutionary algorithms output dozens to hundreds of Pareto non-dominated solutions, making it difficult for schedulers to quickly select a scheduling scheme suitable for the current scenario.

[0008] (4) The crossover operator design is not suitable. The single-point or multi-point crossover operator of the standard genetic algorithm is prone to destroying the continuous arrangement relationship of the same type of task, resulting in a scheme that requires additional switching, and the hardware switching time is as long as 5 to 8 hours. Summary of the Invention

[0009] The purpose of this invention is to overcome the above-mentioned shortcomings of the prior art and provide a multi-objective dynamic weight adaptive scheduling method for flexible calibration production lines of metering equipment. This method can adapt to dynamic scenarios with large fluctuations in production capacity during peak and off-peak seasons, while also taking into account the health status of the equipment, reducing the number of product specification changes, and improving the decision-making efficiency of scheduling schemes.

[0010] A multi-objective dynamic weighted adaptive scheduling method for flexible calibration lines of metrology equipment, wherein the calibration line comprises N calibration lines with heterogeneous capabilities, N∈[3,20], includes the following steps: Step S1: Construct the verification line capability model and obtain the number of tables B for each verification line. i The set of compatible device types Γ i The average daily batch size of various types of equipment being inspected (β) k i and the switching time parameter σ between different types of devices ij Based on the above parameters, the daily production capacity C of each calibration line for each type of equipment is calculated. i =B i ×β k i The scheduling cycle time granularity is 1 month by default and 1 day by minimum. Step S2: Collect the operating condition indicators of key equipment in each calibration line at a frequency of 10kHz for voltage / current parameters, 1Hz for temperature, and 20kHz for vibration. Calculate the comprehensive health score H of each calibration line using a weighted normalization method. i (t), where the set of operating condition indicators S is non-empty and the weight coefficients w of each indicator are... s Satisfy Σ s∈S w s =1; Step S3: Calculate the total number of orders Q in the current scheduling period. m and the total monthly production capacity C of each testing linetotal,m Calculate the order capacity ratio R m =Q m / C total,m Based on the preset range in which the order capacity ratio is located, a four-objective weight vector is determined, consisting of completion time weight w1, equipment utilization weight w2, switching cost weight w3, and health risk weight w4; corresponding to three scenarios: peak season, off-season, and low season. Step S4: Assign decision variable x ij ∈{0,1} and ranking decision variable y jj′ Let ∈{0,1} be the optimization variables. With the four optimization objectives of minimizing the maximum completion time f1, maximizing equipment utilization f2, minimizing switching cost f3, and minimizing health risk f4, a four-objective optimization model is constructed under the conditions of satisfying the constraints of task integrity C1, equipment compatibility C2, capacity C3, delivery time C4, and equipment status C5. Step S5: Solve the four-objective optimization model using an improved non-dominated sorting genetic algorithm to generate a Pareto front solution set; the improvement includes a task block crossover operator, an elite retention strategy, and an adaptive mutation probability. Preferably, an improved NSGA-II multi-objective evolutionary optimization is performed, which preserves the continuous permutation relationship of the same type of tasks through the task block crossover operator, adopts an elite retention strategy to ensure that the extreme solutions of a single objective are not lost, and uses adaptive mutation probability to balance the global search and local refinement capabilities. Step S6: Automatically select representative solutions from the Pareto front solution set and display them to the scheduler for interactive selection in a visual manner.

[0011] Preferably, in step S3, the three-level determination of the weight vector based on the order capacity ratio specifically refers to: when R m A value greater than 0.8 indicates peak production season; therefore, [w1,w2,w3,w4] = [0.50,0.20,0.10,0.20] is set. When the value is greater than 0.5... <R m When R is ≤0.8, it is considered a production off-season, and the values ​​are set to [0.35, 0.25, 0.15, 0.25]. m A value ≤0.5 indicates a low season for production, with a set range of [0.20, 0.35, 0.20, 0.25].

[0012] Preferably, the formula for calculating the comprehensive health score in step S2 is:

[0013] Where S is the set of operating condition indicators, w s h represents the weighting coefficients for each indicator. s i (t) represents the actual measured value, hs ref (t) represents the standard reference value, h s max and h s min These are the upper and lower limits of the allowed range, respectively. The truncation constraint max(0,·) ensures that H is within extreme operating conditions. i (t)≥0, to prevent distortion of the objective function f4 direction; the operating condition indicators include 7 items: voltage output deviation, current output deviation, frequency output deviation, phase output deviation, voltage symmetry, power stability and waveform distortion.

[0014] Preferably, the maximum completion time is minimized:

[0015] Maximize equipment utilization (converted to minimize):

[0016] Minimize switching costs:

[0017] Minimize health risks:

[0018] The quantitative definition of the constraint is as follows: Constraint C1 is a task allocation integrity constraint, ensuring that each task is assigned to at least one calibration line; ; Constraint C2 is x ij =0,∀ i :d j ∉Γ i That is, the device type d of task j j The set of compatible equipment types that do not belong to the verification line i Γ i At that time, the task cannot be assigned to this line; Constraint C3 is that the daily workload of each calibration line shall not exceed the effective working hours T. avail,i ; Constraint C4 is C ij ≤DL j ,∀ j (where Cij is the estimated completion time of task j on calibration line i, DL) j (The promised delivery time for task j). Constraint C5 is a non-assignable task for the calibration line that is in the inspection or maintenance state (x) ij =0).

[0019] Preferably, the execution rules of the task block crossover operator in step S5 are as follows: Step S51: Identify consecutive task sequences with the same device type in the parent chromosome and define them as task blocks; Step S52: Select an intersection point at the boundary of the task block, the intersection operation with probability p c ∈[0.7,0.9] to execute, default p c =0.80; Step S53: Exchange gene segments between the two parent chromosomes in units of task blocks; Step S54: The task of repairing duplications or deletions in offspring chromosomes.

[0020] Preferably, the improved non-dominated sorting genetic algorithm further includes an elite retention strategy: in each generation of evolution, four single-objective extreme solutions are additionally retained: the optimal solution for completion time, the optimal solution for equipment utilization, the optimal solution for switching cost, and the optimal solution for health risk. The upper limit for the number of non-dominated solutions retained in the Pareto front is 100.

[0021] Preferably, the improved non-dominated sorting genetic algorithm further includes an adaptive mutation probability mechanism, wherein the mutation probability decays with the number of generations according to the following formula:

[0022] Where g is the current iteration number, G=200 is the maximum iteration number, the initial mutation probability is 0.30, and the final mutation probability is 0.10; the mutation operations include three types: task migration, intra-line swap, and task block reversal.

[0023] Preferably, the complete mathematical expression of the weighted synthesis objective function in step S6 is:

[0024] in Let be the normalized value of the k-th objective function. ; The optimal solutions for each of the four single-objective dimensions are retained from the Pareto front, and the 1 to 3 comprehensive equilibrium solutions with the smallest ftotal are selected, resulting in a total of 5 to 7 representative solutions.

[0025] Based on the above, this application also designs a multi-objective dynamic weight scheduling system for a flexible calibration line of metrology equipment, including: The order management module receives task data from the inspection and verification management system at its input end and provides a task set J={j1,...,j...} at its output end. M} and the equipment type d for each task j Quantity q j Delivery period DL j ; The equipment management module receives real-time operating data from the SCADA system at its input end and provides health scores (H) for each calibration line at its output end. i (t), Effective production capacity C i eff and availability status flags; The dynamic weight adaptive module receives the total order amount Q as input. m And total monthly production capacity C total,m The output provides a four-objective weight vector [w1, w2, w3, w4]; The four-objective optimization module receives weight vectors, task sets, and calibration parameters at its input end, and outputs Pareto front solution set using an improved non-dominated sorting genetic algorithm. The interactive decision-making module receives the Pareto front solution set as input and displays a visual comparison of 5 to 7 representative solutions to the scheduler as output, and accepts the scheme confirmation instruction; each module is executed in the order of order collection, health assessment, weight determination, optimization solution, and scheme display.

[0026] Based on the above, this application also designs a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described multi-objective dynamic weight adaptive scheduling method for flexible verification lines of metrology equipment.

[0027] The symbol table used in this application

[0028]

[0029] The advantages and effects of this application are as follows: This application designs a multi-objective dynamic weighted adaptive scheduling method for flexible calibration lines of metering equipment. By incorporating equipment health as the fourth objective, the workload assigned to calibration lines with a score below 0.7 is reduced by approximately 40%, effectively lowering the risk of equipment failure. Simultaneously, the Pareto front explicitly presents the trade-off between health and completion time to the scheduler. The task block crossover operator uses consecutive task sequences of the same equipment type as crossover units, reducing the number of specification changes from 18 times / month to 7 times / month, a reduction of 61.1%. The saved hardware changeover time (5 to 8 hours each time) is directly converted into effective production capacity. Furthermore, the elite retention strategy designed in this application ensures that the four single-objective extreme solutions are not lost during population updates, and the solution set hypervolume (HV index) is superior to the standard NSGA-II. The interactive screening framework decompresses dozens to hundreds of non-dominated solutions into 5 to 7 representative solutions, reducing the scheduler's scheme selection time from "inoperable" to minutes. This application also enables three levels of dynamic weights to automatically switch scheduling objectives according to monthly load: peak season (R... m>0.8) Maximum completion time is reduced by 15%~20% compared to the fixed-weight scheme; off-season (R m ≤0.5) Equipment utilization rate increases by 10%~15%, and the improvements in both directions come from the same judgment mechanism, requiring no manual intervention.

[0030] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0031] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0033] Figure 1 System architecture diagram (three layers and nine nodes) designed for this application; Figure 2 Flowchart of the dynamic weight adaptive mechanism designed for this application (decision diamond + three-branch path); Figure 3 A schematic diagram of the improved NSGA-II chromosome double-layer coding structure and task block identifier designed for this application; Figure 4 Flowchart of the interactive Pareto solution filtering framework designed for this application; Figure 5 Radar chart of scheduling results for the peak season scenario (July) designed for this application; Figure 6 A comparison chart of the average daily load rate of each verification line during the off-season (December) and peak season (July) designed for this application. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0035] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0036] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0037] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0038] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0039] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0040] Example 1: This example mainly introduces the overall architecture of the scheduling system.

[0041] The scheduling optimization system is deployed on the scheduling management platform of the verification center, and includes three layers: data acquisition layer, model calculation layer, and interactive decision-making layer. Figure 1 As shown, Figure 1 The overall system architecture diagram (three layers and nine nodes) designed for this application depicts the system's three-layer, nine-node architecture. The three input terminals of the data acquisition layer (ERP / MES verification management system, SCADA equipment monitoring system, and production execution system) provide order data, working condition time series, and production statistics to the model calculation layer, respectively. Inside the model calculation layer, scene recognition, weight adaptation, NSGA-II optimization, and Pareto solution filtering are executed sequentially. Finally, the interactive decision layer completes the visualization comparison between the three-dimensional scatter plot and the radar chart, and inputs the scheduler's weight fine-tuning instructions back into the weight adaptation module (closed-loop feedback).

[0042] The data acquisition layer consists of three sub-modules, responsible for obtaining the basic data required for scheduling from external systems: Order Management Module: Input data is a JSON stream of the task queue from the Inspection and Verification Management System (ERP / MES); output data is a task set J={j1,...,j...} M} and the equipment type d for each task j Quantity q j Delivery period DL j The output timing is triggered once at the beginning of the scheduling cycle; Equipment Management Module: Input data consists of multi-channel operating condition timing data from the SCADA system (electrical parameters 10kHz, temperature 1Hz, vibration 20kHz); output data consists of the health score H for each calibration line. i (t), Effective production capacity C i eff The output includes an availability status flag, and is updated hourly.

[0043] Capacity Data Module: Input data is the daily output statistics of the production execution system; output data is the total monthly capacity C. total,m The output timing is triggered once at the beginning of each scheduling cycle.

[0044] The model computation layer comprises four sub-modules: a scene recognition module, a weight adaptation module, an NSGA-II optimization engine, and a Pareto solution filter. This is the core computational unit of the invention, executing the core scheduling optimization algorithm sequentially.

[0045] The weighted adaptive module takes the total number of orders Qm and the total monthly production capacity Ctotal,m as input data, and outputs a four-objective weight vector [w1,w2,w3,w4]. The output timing is executed immediately after the order collection is completed.

[0046] NSGA-II optimization engine: Input data includes weight vectors, task set J, and calibration parameters (B). i ,Γ i, β k i ,σ ij H i The output data is the Pareto front solution set. The output time series is executed after the weight vector is determined, and the running time is about 5 to 15 minutes.

[0047] Pareto solution filter: The input data is the Pareto front solution set and the current weight vector. The output data is 5 to 7 representative solutions. The output time series is executed immediately after optimization.

[0048] The interactive decision-making layer displays a visual comparison of representative solutions to the schedulers, supporting the selection of solutions or manual fine-tuning of weight parameters; after receiving the solution confirmation instruction, it sends the scheduling solution to the production execution system.

[0049] Example 2: Based on Example 1, this example mainly introduces the dynamic weight adaptive mechanism.

[0050] This embodiment uses the order-to-capacity ratio R. m As the sole decision variable, the current cycle is categorized into three scenarios—peak, flat, and sluggish—through a two-level decision process. Each scenario outputs a fixed-weight vector, which is then passed to the NSGA-II optimization engine. Figure 2 As shown, Figure 2 The flowchart for the dynamic weight adaptive mechanism (decision diamond + three-branch path) is based on the order capacity ratio R. m As the sole decision variable, the current cycle is categorized into three scenarios—peak, flat, and sluggish—through a two-level decision diamond. Each scenario outputs a fixed-weight vector, which is then merged into a single output and passed to the NSGA-II optimization engine. It should be noted that R... m It can exceed 1.0 (peak season overcapacity scenario). In this case, the system still executes the peak season weight vector [0.50, 0.20, 0.10, 0.20]. The delivery time satisfaction of constraint C4 needs to be handled separately in conjunction with the capacity expansion strategy.

[0051] It should be noted that R m It can exceed 1.0 (peak season overcapacity scenario). In this case, the system still executes the peak season weight vector [0.50, 0.20, 0.10, 0.20]. The delivery time satisfaction of constraint C4 needs to be handled separately in conjunction with the capacity expansion strategy.

[0052] The switching between the three weight levels was verified using real monthly data, as shown in Table 1:

[0053] Table 1 Monthly Data and Three-Tier Weight Vectors Example 3: Based on Examples 1-2, this example mainly introduces the capability modeling of calibration lines. Capability profiles were established for each of the 11 calibration lines, and the actual data is shown in Table 2.

[0054] Table 2. Verification line capability parameters Switching time matrix: Program switching within the same type σ soft ≈0 (seconds), cross-type hardware switching σ hard =5~8 hours.

[0055] Effective monthly production capacity is:

[0056] The operating condition index system of the core equipment (multi-functional calibration device) of each calibration line is shown in Table 3: The overall health score is calculated using a weighted normalization method, and the truncation constraint max(0,・) ensures that H is within the range of extreme operating conditions. i (t)≥0, to prevent distortion of the objective function f4 direction.

[0057]

[0058] Table 3 Operating Condition Index System (Σw) s =1) Example 4: This example mainly introduces the improved NSGA-II optimization engine designed in this application.

[0059] This invention employs a two-layer chromosome coding structure, such as Figure 3 As shown. The first layer is the task-test line allocation code, with a length equal to the number of tasks n, and each gene bit indicates which test line the task is assigned to; the second layer is the task sorting code within each test line, using a permutation coding method.

[0060] Task block crossover operator: Step T1: Identify consecutive task sequences with the same device type in the parent chromosome and define them as task blocks; Step T2: Select an intersection point at the boundary of the task block. The intersection operation is performed with probability p. c Execute within [0.7, 0.9], default pc = 0.80; Step T3 involves exchanging gene segments between two parent chromosomes, using task blocks as the unit; Step T4 involves repairing duplicate or missing tasks in the offspring chromosomes. The fact that the crossover point falls on the boundary of a task block is the core difference between the task block crossover operator and the standard single-point crossover, and it is also the direct reason for the approximately 30% reduction in the number of switches.

[0061] Elite retention strategy: In each generation of evolution, in addition to retaining the non-dominated solutions on the Pareto front (up to 100), four additional single-objective extreme solutions are retained: f1 optimal solution (minimum completion time), f2 optimal solution (highest utilization), f3 optimal solution (minimum switching cost), and f4 optimal solution (lowest health risk).

[0062] Adaptive mutation probability: The mutation probability decreases parabolically with each generation, ensuring that the mutation rate decreases slowly in the first 100 generations (global search) and converges quickly to 0.10 in the last 100 generations (local refinement).

[0063] Mutation operations include three types: (a) Task migration – Migrating a task from one line to another compatible line; (b) Intra-line swapping—Swapping the execution order of two tasks on the same line; (c) Task block reversal - Reversing the sequence of tasks within a task block on the same line.

[0064] The algorithm parameters are shown in Table 4:

[0065] Table 4 NSGA-II Algorithm Parameters The complete interactive filtering path of this invention is as follows: Figure 4 As shown: Starting from the Pareto solution set output by NSGA-II, four single-objective extreme solutions (optimal in each dimension) are first extracted. Then, objective normalization is performed on the entire solution set and the weighted comprehensive score f is calculated. total The system obtains one to three comprehensive equilibrium solutions with the lowest scores; the solutions from both categories are then merged and sent to the visualization interface. If the scheduler has doubts about the recommended solution, they can fine-tune the weights w and trigger a re-scoring without having to rerun NSGA-II.

[0066] The filtering rules are as follows: Rule R1: Retain the solution with the minimum f1 (optimal completion time); Rule R2: Retain the solution that minimizes f2 (optimal utilization); Rule R3 retains the solution with the minimum f3 (optimal switching cost); Rule R4 preserves the solution that minimizes f4 (optimal health risk); Rule R5, calculate the weighted composite objective f total Take f total The smallest 1 to 3 solutions are taken as the overall equilibrium solution.

[0067] Example 5, please refer to Figure 6 This embodiment mainly uses actual cases of peak season (July) and off-season (December) scenarios.

[0068] Please refer to Figure 5 , Figure 5 Radar chart showing the peak season (July) scheduling results designed for this application; July scenario data: 23 working days per month, single-phase verification demand 536,880 units, three-phase direct-acting demand 118,099 units, three-phase inductive demand 48,633 units. Monthly capacity of 6 single-phase lines: 6 × 4500 × 23 = 621,000 units, order-to-capacity ratio R m =703,612 / 621,000=1.13>0.8, therefore it is determined to be the peak season, and the weight is set to w=[0.50,0.20,0.10,0.20].

[0069] 536,880 single-phase meters were allocated to 6 single-phase lines, and tasks of the same specification were merged using task block sorting. Assuming the health of line l3 is H3=0.65 (below the 0.7 threshold), the system automatically transferred some tasks from line l3 to line l4 with H4=0.92. The Pareto front ultimately generated 67 non-dominated solutions, from which 6 representative solutions were selected, and a comprehensive equilibrium solution with a maximum completion time of 17.8 days was recommended.

[0070] December scenario data: 7 working days in the month, single-phase demand 92,896 units, three-phase demand 43,799 units. Order-to-capacity ratio R m =136,695 / 189,000=0.72, 0.5 <R m ≤0.8, classified as a flat season, weighted w=[0.35,0.25,0.15,0.25]. Due to the low load, the system prioritizes merging similar tasks to fewer lines, releasing 3 single-phase lines for monthly maintenance, increasing equipment utilization from 62.3% to 73.8%.

[0071] In addition, based on a complete dataset of 11 lines and 12 months, the performance of the four methods was compared, and the results are shown in Table 5:

[0072] Table 5 Performance Comparison of Four Methods Compared to manual scheduling: maximum completion time is reduced by 24.3% ((23.5-17.8) / 23.5), and changeover times are reduced by 61.1% ((18-7) / 18). Compared to standard NSGA-II (fixed weight): maximum completion time is further reduced by 11.4%, and utilization rate is increased by 9.9%.

[0073] The above description is merely a preferred embodiment of the present invention and does not limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter alterations to these embodiments within the spirit and principles of the present invention, achieved through conventional substitutions or by achieving the same function without departing from the principles and spirit of the present invention, fall within the scope of protection of the present invention.

Claims

1. A multi-objective dynamic weighted adaptive scheduling method for a flexible calibration line of metrology equipment, wherein the calibration line comprises N calibration lines with heterogeneous capabilities, N∈[3,20], characterized in that, Includes the following steps: Step S1: Construct the verification line capability model and obtain the number of tables B for each verification line. i The set of compatible device types Γ i The average daily batch size β of various types of equipment being inspected k i and the switching time parameter σ between different types of devices ij Based on the above parameters, the daily production capacity C of each calibration line for each type of equipment is calculated. i =B i ×β k i The scheduling cycle time granularity is 1 month by default and 1 day by minimum; Step S2: Collect the operating condition indicators of key equipment in each calibration line at a frequency of 10kHz for voltage / current parameters, 1Hz for temperature, and 20kHz for vibration. Calculate the comprehensive health score H of each calibration line using a weighted normalization method. i (t), where the set of operating condition indicators S is non-empty and the weight coefficients w of each indicator are... s Satisfy Σ s∈S w s =1; Step S3: Calculate the total number of orders Q in the current scheduling period. m and the total monthly production capacity C of each testing line total,m Calculate the order capacity ratio R m =Q m / C total,m Based on the preset range in which the order capacity ratio is located, a four-objective weight vector is determined, consisting of completion time weight w1, equipment utilization weight w2, switching cost weight w3, and health risk weight w4. Step S4: Assign decision variable x ij ∈{0,1} and ranking decision variable y jj′ Let ∈{0,1} be the optimization variables. With the four optimization objectives of minimizing the maximum completion time f1, maximizing equipment utilization f2, minimizing switching cost f3, and minimizing health risk f4, a four-objective optimization model is constructed under the conditions of satisfying the constraints of task integrity C1, equipment compatibility C2, capacity C3, delivery time C4, and equipment status C5. Step S5: Solve the four-objective optimization model using an improved non-dominated sorting genetic algorithm to generate a Pareto front solution set; the improvement includes a task block crossover operator, an elite retention strategy, and an adaptive mutation probability. Step S6: Automatically select representative solutions from the Pareto front solution set and display them to the scheduler for interactive selection in a visual manner.

2. The multi-objective dynamic weight adaptive scheduling method for flexible calibration lines of metering equipment according to claim 1, characterized in that, In step S3, the three-level determination of the weight vector based on the order capacity ratio specifically refers to: when R m A value greater than 0.8 indicates peak production season; therefore, [w1,w2,w3,w4] = [0.50,0.20,0.10,0.20] is set. When the value is greater than 0.5... <R m When R is ≤0.8, it is considered a production off-season, and the values ​​are set to [0.35, 0.25, 0.15, 0.25]. m A value ≤0.5 indicates a low season for production, with a set range of [0.20, 0.35, 0.20, 0.25].

3. The multi-objective dynamic weight adaptive scheduling method for flexible calibration lines of metering equipment according to claim 1, characterized in that, The formula for calculating the comprehensive health score in step S2 is as follows: ； Where S is the set of operating condition indicators, w s h represents the weighting coefficients for each indicator. s i (t) represents the actual measured value, h s ref (t) represents the standard reference value, h s max and h s min These are the upper and lower limits of the allowed range, respectively. The truncation constraint max(0,·) ensures that H is within extreme operating conditions. i (t)≥0, to prevent distortion of the objective function f4 direction; the operating condition indicators include 7 items: voltage output deviation, current output deviation, frequency output deviation, phase output deviation, voltage symmetry, power stability and waveform distortion.

4. The multi-objective dynamic weight adaptive scheduling method for flexible calibration lines of metering equipment according to claim 1, characterized in that, Minimize the maximum completion time: ； Maximize equipment utilization (converted to minimize): ； Minimize switching costs: ； Minimize health risks: ； The quantitative definition of the constraint is as follows: Constraint C1 is a task allocation integrity constraint, ensuring that each task is assigned to at least one calibration line; ； Constraint C2 is x ij =0,∀ i :d j ∉Γ i That is, the device type d of task j j The set of compatible equipment types that do not belong to the calibration line i Γ i At that time, the task cannot be assigned to that line; Constraint C3 is that the daily workload of each calibration line shall not exceed the effective working hours T. avail,i ; Constraint C4 is C ij ≤DL j ,∀ j (where Cij is the estimated completion time of task j on calibration line i, DL) j (The promised delivery time for task j). Constraint C5 is a non-assignable task for the calibration line that is in the inspection or maintenance state (x) ij =0).

5. The multi-objective dynamic weight adaptive scheduling method for flexible calibration lines of metering equipment according to claim 1, characterized in that, The execution rules for the task block crossover operator in step S5 are as follows: Step S51: Identify consecutive task sequences with the same device type in the parent chromosome and define them as task blocks; Step S52: Select an intersection point at the boundary of the task block, the intersection operation with probability p c ∈[0.7,0.9] to execute, default p c =0.80; Step S53: Exchange gene segments between the two parent chromosomes in units of task blocks; Step S54: The task of repairing duplications or deletions in offspring chromosomes.

6. The multi-objective dynamic weight adaptive scheduling method for flexible calibration lines of metering equipment according to claim 1, characterized in that, The improved non-dominated sorting genetic algorithm also includes an elite retention strategy: in each generation of evolution, four single-objective extreme solutions are additionally retained: the optimal solution for completion time, the optimal solution for equipment utilization, the optimal solution for switching cost, and the optimal solution for health risk. The upper limit for the number of non-dominated solutions retained in the Pareto front is 100.

7. The multi-objective dynamic weight adaptive scheduling method for flexible calibration lines of metering equipment according to claim 1, characterized in that, The improved non-dominated sorting genetic algorithm also includes an adaptive mutation probability mechanism, where the mutation probability decreases with each generation according to the following formula: ； Where g is the current iteration number, G=200 is the maximum iteration number, the initial mutation probability is 0.30, and the final mutation probability is 0.10; the mutation operations include three types: task migration, intra-line swap, and task block reversal.

8. The multi-objective dynamic weight adaptive scheduling method for flexible calibration lines of metering equipment according to claim 1, characterized in that, The complete mathematical expression of the weighted synthesis objective function mentioned in step S6 is: ； in Let be the normalized value of the k-th objective function. ; The optimal solutions for each of the four single-objective dimensions are retained from the Pareto front, and the 1 to 3 comprehensive equilibrium solutions with the smallest ftotal are selected, resulting in a total of 5 to 7 representative solutions.

9. A multi-objective dynamic weight adaptive scheduling system for a flexible calibration production line of measuring equipment, characterized in that, include: The order management module receives task data from the inspection and verification management system at its input end and provides a task set J={j1,...,j...} at its output end. M } and the equipment type d for each task j Quantity q j Delivery period DL j ; The equipment management module receives real-time operating data from the SCADA system at its input end and provides health scores (H) for each calibration line at its output end. i (t), Effective production capacity C i eff and availability status flags; The dynamic weight adaptive module receives the total order amount Q as input. m And total monthly production capacity C total,m The output provides a four-objective weight vector [w1, w2, w3, w4]; The four-objective optimization module receives weight vectors, task sets, and calibration parameters at its input end, and outputs Pareto front solution set using an improved non-dominated sorting genetic algorithm. The interactive decision-making module receives the Pareto front solution set as input and displays a visual comparison of 5 to 7 representative solutions to the scheduler as output, and accepts the scheme confirmation instruction; each module is executed in the order of order collection, health assessment, weight determination, optimization solution, and scheme display.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-objective dynamic weight adaptive scheduling method for the flexible verification line of metrology equipment as described in any one of claims 1 to 8.

Images