A flexible task scheduling method for multiple automatic lines based on dynamic measurement data

Abstract

The present invention relates to a method for flexible task scheduling of multiple automatic lines based on dynamic metering data, and belongs to the field of workshop scheduling. The method comprises: setting different acquisition frequencies based on different parameter characteristics to perform parameter acquisition, and performing different data cleaning methods on the different types of acquired parameters to perform data cleaning; constructing an equipment operation, product quality, and material supply analysis model based on the cleaned data; constructing a scheduling model including multi-objective function optimization and multi-constraint conditions based on the equipment operation, product quality, and material supply analysis model, and solving the model to obtain the optimal solution in a complex task allocation combination. The present invention can realize flexible task scheduling of multiple automatic lines, realize task scheduling planning that is more in line with complex production scenarios, provide effective task allocation for actual production, and improve task scheduling efficiency.

Description

Technical Field

[0001] The invention belongs to the field of workshop scheduling, and in particular relates to a flexible task scheduling method for multiple automatic lines based on dynamic metering data. Background Art

[0002] Flexible workshop scheduling methods address various uncertainties and changes during the workshop production process by flexibly adjusting production tasks and resource allocation to achieve efficient production. Common methods include rule-based scheduling, which uses rules based on production characteristics and objectives to quickly make decisions, but lack flexibility. Heuristic algorithms, such as genetic algorithms, use heuristic rules and search strategies to find near-optimal solutions in the solution space. They can handle complex problems but cannot guarantee global optimality. Simulated annealing algorithms start from an initial solution and search for the optimal solution through random perturbations and acceptance criteria. They have strong global search capabilities but are computationally intensive and converge slowly. Constraint satisfaction-based scheduling methods transform the problem into a constraint satisfaction problem, accurately handling complex constraints, but are difficult to solve and inefficient. Agent-based technology abstracts workshop entities into agents and implements scheduling through interactive collaboration between agents. It offers flexibility and adaptability, but its design and coordination mechanisms are complex.

[0003] Existing technologies for shop floor scheduling have numerous shortcomings that need to be addressed. For example, CN112183817A employs particle coding combined with a genetic algorithm and simulated annealing for scheduling. While this approach offers some optimization, it still has limitations. The main limitations are as follows:

[0004] Limitations of Data Collection and Processing: Existing technologies are not comprehensive and accurate in data collection. They lack systematic collection of dynamic measurement data on various aspects, such as equipment operation, product quality, and material supply, making it difficult to fully reflect the actual production process. For example, when monitoring equipment status, they may only focus on a few key parameters and ignore other factors that affect production. The fixed collection frequency cannot adapt to the dynamic changes in parameters, resulting in delayed acquisition of key information. In terms of data processing, the methods for handling outliers are relatively simple, making it difficult to effectively deal with data noise in complex production environments, affecting the accuracy and reliability of subsequent analysis.

[0005] Imperfect analytical models: Existing analytical models often focus on a single production factor, failing to build comprehensive analytical models encompassing equipment failure prediction, the relationship between product quality and production parameters, and material supply trends. For example, when predicting equipment failure, there is a lack of in-depth exploration and effective integration of equipment operating parameters, resulting in insufficient prediction accuracy. The correlation analysis between product quality and production parameters is not thorough enough, making it impossible to provide precise guidance for adjusting production parameters. Regarding material supply, it is difficult to accurately predict inventory consumption trends and replenishment timing, which can easily lead to material shortages or backlogs, impacting production continuity.

[0006] Suboptimal scheduling models: Existing scheduling models lack comprehensiveness and flexibility in their objective functions and constraint settings. Objective functions often focus solely on a single metric, such as production cycle time or equipment utilization, failing to comprehensively balance multiple production goals. Constraints are also insufficiently considered, lacking detailed consideration of complex constraints in actual production, such as equipment maintenance, task sequencing, and material supply. These models struggle to adapt to complex and ever-changing production scenarios, leading to low production efficiency and wasted resources.

[0007] Lack of flexibility in scheduling adjustment mechanisms: Existing technologies lack effective scheduling adjustment mechanisms for dynamic situations such as equipment failures and order changes. Once an equipment failure occurs, tasks cannot be quickly and reasonably reallocated, resulting in extended production interruptions. When orders change, it is difficult to adjust task allocation and production plans to meet new demands, impacting order delivery timelines and customer satisfaction, and reducing the adaptability and robustness of the production system.

[0008] The efficiency and quality of solution algorithms need to be improved: Existing solution algorithms, such as the combination of genetic algorithms and simulated annealing, lack efficiency and quality. Genetic algorithms are prone to getting stuck in local optimal solutions, while simulated annealing algorithms are computationally intensive and slow to converge. This makes it difficult to quickly find optimal solutions for complex task allocation combinations, failing to meet the demand for efficient scheduling in actual production, impacting production efficiency and enterprise competitiveness. Summary of the Invention

[0009] The purpose of the present invention is to provide a flexible task scheduling method for multiple automatic lines based on dynamic metering data, which can realize flexible task scheduling of multiple automatic lines, realize task scheduling planning that is more in line with complex production scenarios, provide effective task allocation for actual production, and improve task scheduling efficiency.

[0010] To achieve the above object, the technical solution of the present invention is: a method for flexible scheduling of tasks of multiple automatic lines based on dynamic metering data, comprising the following steps:

[0011] S1. Parameter acquisition: acquiring a first parameter text through an acquisition terminal according to a set acquisition frequency, the first parameter text including equipment operation parameters, product quality parameters, and material supply parameters;

[0012] S2. Data cleaning: Perform data cleaning operations on equipment operating parameters, product quality parameters, and material supply parameters to obtain parameter processing text;

[0013] S3. Data analysis: Build equipment operation analysis models, product quality analysis models, and material supply analysis models based on parameter processing text;

[0014] S4. Scheduling model construction: Based on the analysis results of the equipment operation analysis model, product quality analysis model, and material supply analysis model, the objective function optimization is constructed, and constraints are set according to the objective function to build a task scheduling model;

[0015] S5. Scheduling model solution: Genetic algorithm is used to solve the task scheduling model and obtain the optimal task allocation solution.

[0016] Furthermore, in S1, equipment operating parameters include speed, temperature, and pressure; product quality parameters include size; and material supply parameters include inventory quantity and material delivery speed.

[0017] Furthermore, in S1, the acquisition frequency is set as follows:

[0018] For the equipment operating parameters, set the acquisition frequency to 5Hz;

[0019] For product quality parameters, the acquisition frequency is determined by the production rhythm. The production rhythm is to produce one product every N seconds, so the acquisition frequency is set to 1 / N Hz.

[0020] For material supply parameters, set the acquisition frequency to 1 / 60Hz.

[0021] Furthermore, in S2, the data cleaning operation is as follows:

[0022] For equipment operating parameters, we set upper and lower thresholds to identify abnormal values, and use the sliding average method to clean abnormal values;

[0023] For product quality parameters, standard thresholds are set to identify outliers, and outliers are cleaned by comparing data with similar products.

[0024] For material supply parameters, outliers are identified by setting standard thresholds or upper and lower thresholds, and outliers are cleaned by comparing them with the data at the previous moment or by weighted averaging the data of adjacent time periods.

[0025] Furthermore, in S3, data analysis is implemented as follows:

[0026] The equipment operation analysis model uses the ARIMA (p, d, q) model to predict equipment failure time, and the time series data {y T},y t For {y T}, the data at time t, the model expression is:

[0027]

[0028] Among them, the autoregressive order p=2, the difference order d=1, and the moving average order q=1. 、 is the model parameter, y t-i For {y T}, B is the backshift operator, μ is the constant term, is the white noise at time t, is the white noise at time tj; the model output is the equipment failure time t fault The predicted value of

[0029] The product quality analysis model refers to the construction of a linear regression model that associates product quality with production parameters. The model expression is as follows:

[0030] Y=β0+β1X1+β2X2+β3X3+

[0031] Among them, the product size qualification rate in the product quality parameter is Y, and the production parameters are equipment speed X1, equipment temperature X2 and material conveying speed X3. is a constant, and the least square method is used to estimate the regression coefficients β0, β1, β2, and β3 to determine the quantitative relationship between product quality and production parameters;

[0032] The material supply analysis model uses time series analysis to predict material inventory consumption trends and replenishment time points. Assume that the material inventory quantity time series data is {I T}, I t For {I T By analyzing the historical data trends and seasonal characteristics, a time series prediction model I is established. t+1 =αI t +(1-α)I t-1 , where α is the smoothing coefficient, ranging from 0 to 1, I t-1 For {I T}, predict the material inventory quantity I at the next k moments t+k , determine the replenishment time point t replenish At the same time, analyze the stability of material conveying speed and calculate the standard deviation of material conveying speed fluctuation σ speed ;

[0033] S4. Scheduling model construction: Based on the analysis results of the equipment operation analysis model, product quality analysis model, and material supply analysis model, the objective function optimization is constructed, and constraints are set according to the objective function to build a task scheduling model;

[0034] S5. Scheduling model solution: Genetic algorithm is used to solve the task scheduling model and obtain the optimal task allocation solution.

[0035] Furthermore, in S4, the scheduling model is specifically implemented as follows:

[0036] The objective functions include minimizing the production cycle function, maximizing the equipment utilization function, and meeting the order delivery time function.

[0037] The process of optimizing the construction of the minimization production cycle function is as follows:

[0038] Considering the equipment failure time t fault Based on the impact of changes in product quality requirements on task processing time, the production cycle objective function is further optimized. Since equipment failure may lead to task reallocation, the set of tasks that need to be reallocated due to equipment failure is set to R. For task i∈R, its new processing time t ij,new The processing capacity of the new equipment needs to be comprehensively considered j,new As well as the task priority factor, assuming the task priority coefficient is π i , then the adjusted production cycle objective function is expressed as , n represents the number of tasks, m represents the number of devices; for normal tasks, the processing time , w i is the workload of task i, c j,normal is the normal processing capacity of the equipment; for task i∈R, ;

[0039] The optimization construction process of the function to maximize equipment utilization is as follows:

[0040] Combined with the changes in processing time caused by equipment failure and task reallocation, the equipment utilization objective function is improved. When the equipment fails, the equipment can work for a certain time, T j Need to be redefined, assuming that the working time of equipment j before failure is T j,used , the fault repair time is T j,repair , then the new working time T j,new =T j -T j,used -T j,repair , equipment utilization rate , where t ij Calculated based on actual conditions after equipment failure and task reallocation;

[0041] The process of optimizing the function to meet order delivery time is as follows:

[0042] Considering the material supply constraint, the task i starts at time s i Changes and processing time t of task i iWhen the material delivery speed is unstable, the order delivery time objective function is adjusted. Let the time for task i to be delayed due to material shortage be Δt delay , the part of the task processing time that increases due to the unstable material conveying speed is Δt increase , then the order delivery time is adjusted to (s i +Δt delay )+(t i +Δt increase )≤D k , D k For order delivery time, this adjustment ensures that in actual production, we can accurately judge whether the order can be delivered on time;

[0043] Constraints include equipment processing capacity constraints, task sequence constraints, and material supply constraints;

[0044] Equipment processing capacity constraints:

[0045] Assume that equipment j has a maintenance plan in the time period [t1, t2], and the equipment processing capacity c during this period j,t =0, for task i, its processing time t ij It must be within the normal processing time period of the equipment, that is, t start,i ≥t2 and t start,i +t ij ≤t end,j , where t start,i is the start time of task i, t end,j is the end working time of equipment j in the current production cycle, and w i / t ij ≤c j,t ;

[0046] Task sequence constraints:

[0047] Suppose task groups G1 and G2. If G1 needs to be completed before G2, for task i∈G1 and task l∈G2, it must satisfy At the same time, for tasks within the task group, follow the original task sequence constraint s l ≥s i +t i ;

[0048] Material supply constraints:

[0049] If a batch of materials has a quality problem, let the affected material be r′ and the set of tasks involved be M. For task i∈M, it is necessary to suspend the task or adjust the task allocation until the material quality problem is solved. Let the time to solve the material quality problem be t resolve , then for task i∈M, if t <t resolve, task i cannot start or needs to be replanned. At the same time, considering the unstable material conveying speed, the safe range of the material conveying speed is set as [v min , v max . If the actual conveying speed , it is necessary to adjust the material waiting time in the task processing time according to the speed fluctuation situation.

[0050] Furthermore, in S5, the genetic algorithm is used to solve the task scheduling model. The iterative termination condition is set that the fitness value does not increase significantly for N = 10 consecutive generations. When the termination condition is met, the chromosome with the highest fitness value is output, that is, the optimal task assignment scheme is obtained.

[0051] Furthermore, the specific implementation of the genetic algorithm is as follows:

[0052] Set the initial population size P = 50, that is, there are initially 50 chromosomes. Each chromosome represents a task assignment scheme. The chromosome coding adopts the integer coding method. For n tasks and m devices, the chromosome is represented as [j1, j2, ……, j n , where j i represents the equipment number to which task i is assigned;

[0053] The roulette wheel selection method is adopted to calculate the probability of each chromosome being selected according to the fitness value of the chromosome; Let the fitness value of chromosome k be F k , then the probability of its being selected , is the fitness value of chromosome ;

[0054] Set the crossover probability Pc = 0.6. For the selected chromosome pairs for crossover, the single-point crossover method is adopted. For chromosome A = [a1, a2, ……, a n and chromosome B = [b1, b2, ……, b n , randomly select a crossover point p’, 1 < p’ < n. After crossover, new chromosomes A′ = [a1, ……, a p’ , b p’+1 , ……, b n and B′ = [b1, ……, b p’ , a p’+1 , ……, a n are generated;

[0055] Set the mutation probability Pm = 0.01. For the mutated chromosome, randomly select a gene position for mutation. For chromosome C = [c1, c2, ……, c n , if the gene c q’ is selected for mutation, then randomly generate a new automatic line number j new, 1≤j new ≤m, the mutated chromosome C′=[c1,……,c q’-1 ,j new ,c q’+1 ,……,c n ];

[0056] After selection, crossover and mutation operations, a new population is generated;

[0057] Repeat the above process for multiple generations of iterative solution; set the iteration termination condition as the fitness value no longer significantly improves for N = 10 consecutive generations. When the termination condition is met, output the chromosome with the highest fitness value, and obtain the optimal task allocation solution.

[0058] The present invention also provides a computer-readable storage medium on which computer program instructions that can be executed by a processor are stored. When the processor executes the computer program instructions, any of the method steps described above can be implemented.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] 1. More comprehensive and accurate data collection: This system collects dynamic measurement data on various aspects, including equipment operation, product quality, and material supply. By setting different collection frequencies based on parameter variation characteristics, it can more comprehensively and accurately reflect the actual production process. For example, parameters with rapidly changing equipment operating status are collected at a high frequency, while parameters with slower changes are collected at a low frequency. This ensures timely acquisition of key information while avoiding data redundancy.

[0061] 2. More efficient data processing: The present invention adopts specific data cleaning methods for different types of parameters, such as threshold judgment and sliding average method for equipment operation parameters, data comparison method for similar products of product quality parameters, and verification and correction method for material supply parameters. Compared with existing technologies, it can more effectively handle outliers, improve data quality, and provide a reliable basis for subsequent analysis and scheduling decisions.

[0062] 3. Improved analysis model: A multi-dimensional analysis model has been constructed, including an equipment operation analysis model that predicts equipment failure time to affect equipment processing capacity in task scheduling, a product quality analysis model that establishes a quantitative relationship between production parameters and product quality to affect task processing time, and a material supply analysis model that predicts material inventory consumption and replenishment time points and considers the impact of material delivery speed on task scheduling. The existing technology uses single data when evaluating mail classification models, while the present invention can analyze the production process more deeply and comprehensively, providing a more scientific basis for task scheduling.

[0063] 4. More optimized scheduling model: A scheduling model that includes multi-objective function optimization and multiple constraints has been established, comprehensively considering actual factors such as production cycle, equipment utilization, order delivery time, equipment processing capacity, task sequence, material supply, etc. Compared with the existing technology that can only perform simple mail classification and single-dimensional evaluation, the present invention can achieve more complex task scheduling planning that is more in line with actual production needs, thereby improving production efficiency and resource utilization.

[0064] 5. More flexible scheduling and adjustment: It has a scheduling and adjustment mechanism for equipment failures and order changes. When there are dynamic changes in the production process, task allocation can be adjusted in a timely manner to ensure the continuity and stability of production. The existing technology lacks a similar effective adjustment mechanism for complex changes. The present invention makes the production system more adaptable and robust.

[0065] 6. More efficient solution: Genetic algorithm is used to solve the scheduling model. By reasonably setting the genetic operation parameters and iteration termination conditions, it can quickly search for the best solution in complex task allocation combinations. Compared with some traditional scheduling methods, it can more efficiently provide high-quality task allocation solutions for actual production, thereby improving scheduling efficiency and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0067] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0068] The present invention provides a method for flexible scheduling of tasks of multiple automatic lines based on dynamic metering data, comprising:

[0069] S1. Parameter acquisition: acquiring a first parameter text through an acquisition terminal according to a set acquisition frequency, the first parameter text including equipment operation parameters, product quality parameters, and material supply parameters;

[0070] S2. Data cleaning: Perform data cleaning operations on equipment operating parameters, product quality parameters, and material supply parameters to obtain parameter processing text;

[0071] S3. Data analysis: Build equipment operation analysis models, product quality analysis models, and material supply analysis models based on parameter processing text;

[0072] S4. Scheduling model construction: Based on the analysis results of the equipment operation analysis model, product quality analysis model, and material supply analysis model, the objective function optimization is constructed, and constraints are set according to the objective function to build a task scheduling model;

[0073] S5. Scheduling model solution: Genetic algorithm is used to solve the task scheduling model and obtain the optimal task allocation solution.

[0074] The following is a specific implementation process of the present invention.

[0075] like Figure 1 As shown, the present invention provides a flexible task scheduling method for multiple automatic lines based on dynamic metering data, and the specific steps include:

[0076] S1: Parameter acquisition: obtaining the first parameter text through the acquisition terminal according to the set acquisition frequency;

[0077] It should be noted that the first parameter text includes equipment operating parameters, product quality parameters, and material supply parameters;

[0078] Equipment operating parameters include:

[0079] Rotational speed: For various rotating equipment, such as motors and lathe spindles, the rotational speed data is collected. The rotational speed directly reflects the operating speed of the equipment and has a significant impact on production efficiency and product quality. For example, unstable motor speed may lead to reduced product processing accuracy.

[0080] Temperature: Monitor the temperature of key parts of the equipment, such as the engine block and transformer windings. Excessively high temperatures may indicate equipment failure. For example, long-term high-temperature operation of a motor may burn out the windings.

[0081] Pressure: For equipment involving pressure control, such as hydraulic presses and compressors, pressure data is collected. Abnormal pressure may affect the normal operation of the equipment. For example, insufficient pressure in a hydraulic press may result in incomplete machining of the workpiece.

[0082] Product quality parameters include:

[0083] Size: Use laser measuring instruments, calipers and other equipment to measure the size of the product. Failure of the product size to meet the standards will directly affect the assembly and performance of the product. For example, the size deviation of automotive parts may cause difficulties in vehicle assembly.

[0084] Material supply parameters include:

[0085] Inventory quantity: With the help of inventory management systems and sensors, the inventory quantity of raw materials and semi-finished products can be obtained in real time. Insufficient inventory may lead to production interruptions, such as a production line being forced to stop due to the lack of a certain component.

[0086] Material conveying speed: Install speed sensors on the material conveying line to monitor the material conveying speed. Unstable conveying speed may affect the production rhythm. For example, too fast or too slow material conveying speed will cause material accumulation or shortage on the production line.

[0087] The method for setting the acquisition frequency is as follows:

[0088] For parameters with fast changing equipment operating conditions such as speed, temperature, and pressure, the acquisition frequency is set to 5Hz, that is, 5 times per second. According to the formula N fast =f fast ×Δt (where the acquisition frequency f fast =5Hz, Δt=1s), we can calculate the number of acquisitions N in a 1 second time interval. fast =5 times, high-frequency acquisition can timely capture the instantaneous changes in equipment status and provide timely data support for equipment failure prediction and production process adjustment.

[0089] For product dimensional parameters, since each product is tested individually during production, the collection frequency is determined by the production cycle. Assuming a production cycle of one product every 10 seconds, the collection frequency for product quality parameters is 0.1Hz. Within a 10-second interval, the number of collections (Nquality) is 1. This ensures that quality data for each product is accurately collected.

[0090] For parameters such as material inventory quantity that change relatively slowly, the acquisition frequency is set to 1 / 60Hz, that is, once per minute. Within a 60-second time interval, the formula N slow =f slow ×Δt(acquisition frequency f slow =1 / 60Hz, Δt=60s), the acquisition times N is obtained slow =1 time, and if the fluctuation of material conveying speed has a greater impact on production, the collection frequency can be set to 1Hz, collecting once per second, so as to grasp the material conveying situation in time.

[0091] The acquisition terminal includes equipment operation parameter sensors, product quality detection sensors, and material supply sensors;

[0092] Equipment operating parameter sensors include temperature, pressure, and speed sensors, which are used to monitor the equipment's operating status, determine whether the equipment is working properly, and predict equipment failures;

[0093] Product quality detection sensors refer to dimensional measurement sensors used to obtain product quality data in real time so as to adjust production parameters or deal with unqualified products in a timely manner;

[0094] Material supply sensors include material inventory sensors and material conveying speed sensors to monitor material supply conditions and avoid production interruptions due to material shortages.

[0095] S2: Data cleaning: Perform data cleaning operations on equipment operating parameters, product quality parameters, and material supply parameters to obtain parameter processing text;

[0096] Specifically, the data cleaning operation of the equipment operation parameters is as follows:

[0097] Taking the temperature data as an example, for the equipment temperature data sequence {Tn} transmitted from the acquisition module, due to factors such as environmental interference, there may be outliers. Based on the normal working characteristics of the equipment, the normal temperature lower limit Tmin = 50°C and the upper limit Tmax = 80°C are determined. Using the outlier judgment expression, if the collected temperature value Ti < Tmin or Ti > Tmax at the i-th moment, then Ti is determined as an outlier. For other equipment operation parameters such as rotational speed and pressure, in a similar manner, thresholds are set based on the normal operation range of the equipment to identify and clean outliers. For example, the normal rotational speed range of a certain motor is 1400 - 1500 revolutions per minute, and rotational speed data outside this range can be determined as outliers;

[0098] Then, the moving average method is used to clean the outliers. The moving window size w = 10 is set. When the data point serial number k ≥ w, if the k-th data point is an outlier, its value after cleaning ; when k < w, if the k-th data point is an outlier, its value after cleaning .

[0099] The data cleaning operation of the product quality parameters is as follows:

[0100] For the product size data, if the measured value exceeds the tolerance range of the product standard size, it is determined as an outlier. For example, the standard length of a certain product is 100mm, and the tolerance range is ±0.5mm. Data with a measured value less than 99.5mm or greater than 100.5mm is an outlier.

[0101] The method of comparing data of similar products can be used to clean the outliers of product quality parameters. For size data, the average value and standard deviation of the sizes of products in the same batch can be calculated, and data deviating from the average value by more than 3 times the standard deviation is regarded as an outlier and corrected. The correction method is to replace the outlier with the median of the sizes of products in the same batch to ensure the accuracy and reliability of the data and provide effective data for subsequent product quality analysis and production parameter adjustment

[0102] The data cleaning operation of the material supply parameters is as follows:

[0103] If the material inventory quantity data appears negative, it is determined as an outlier. If the material conveying speed data appears to be zero instantaneously or a value far beyond the normal conveying speed range, it is also regarded as an outlier. For example, the normal conveying speed range of a certain material is 0.5 - 1m / s, and data with a speed value less than 0.5m / s or greater than 1m / s and a duration less than 1 second may be outliers.

[0104] Abnormal values ​​of cleaning material inventory quantity can be checked with the inventory quantity at the previous moment and the material in and out records to find out the cause of the error and make corrections. For abnormal values ​​of material conveying speed, the weighted average of the speed values ​​in adjacent time periods is used for correction. The weight is determined according to the distance of the time interval. The closer the time interval, the greater the weight, to ensure that the corrected data can truly reflect the material conveying situation.

[0105] S3: Data analysis: Build equipment operation analysis models, product quality analysis models, and material supply analysis models based on parameter processing text;

[0106] It should be noted that the equipment operation analysis model uses the ARIMA (p, d, q) model to predict equipment failure time, and the time series data {y T},y t For {y T}The data at the tth moment in the model expression is:

[0107] The equipment operation analysis model uses the ARIMA (p, d, q) model to predict equipment failure time, and the time series data {y T},y t For {y T}, the data at time t, the model expression is:

[0108]

[0109] Among them, the autoregressive order p=2, the difference order d=1, and the moving average order q=1. 、 is the model parameter, y t-i For {y T}, B is the backshift operator, μ is the constant term, is the white noise at time t, is white noise at time tj; by processing historical data, the model outputs the equipment failure time t fault The predicted value of .

[0110] It should be further explained that in the task scheduling model, the equipment processing capacity constraint is affected by the equipment failure time. Let the actual processing capacity of equipment j at time t be c j,t , if t≥t fault , then c j,t =0; if t <t fault , c j,t Determined according to the normal operating parameters of the equipment, the objective function of minimizing the production cycle In the example, the processing time t of task i on equipment j isij It will be adjusted due to changes in equipment processing capacity, n is the total number of tasks, m is the total number of equipment; when the equipment is normal , where w i is the workload of task i, c j,normal is the normal processing capacity of the equipment. When the equipment failure is predicted, t ij Need to be reallocated to other equipment to calculate processing time; maximize equipment utilization middle, , T j is the working time of equipment j, and equipment failure causes t ij Changes will affect the equipment utilization calculation.

[0111] The product quality analysis model refers to the construction of a linear regression model that relates product quality to production parameters. Suppose the product size qualification rate is Y, and the production parameters are X1 (equipment speed), X2 (equipment temperature), and X3 (material conveying speed and other parameters related to product quality). The model expression is Y=β0+β1X1+β2X2+β3X3+ The least square method is used to estimate the regression coefficients β0, β1, β2, and β3, and the quantitative relationship between product quality and production parameters is determined through a large amount of data;

[0112] It should be further explained that if the product quality requirements are improved and the product size tolerance range is reduced, the new quality requirement is set to Y new , according to the linear regression model, to satisfy Y new , production parameters need to be adjusted, which in turn affects the task processing time. Assuming that the product size qualification rate is related to the equipment speed X1 and the equipment temperature X2, in order to achieve Y new , X1 needs to be adjusted to X 1,new , X2 needs to be adjusted to X 2,new , the equipment speed and equipment temperature changes will change the equipment processing capacity, assuming the equipment processing capacity becomes c j,new , then the processing time t of task i on equipment j is ij becomes t ij,new =w i / c j,new , which will affect the calculation of the objective function T for minimizing the production cycle and the calculation of maximizing the equipment utilization U, because t ij,new Change will change The value of

[0113] The material supply analysis model uses time series analysis to predict material inventory consumption trends and replenishment time points. Assume that the material inventory quantity time series data is {I T}, I t For {I TBy analyzing the historical data trends and seasonal characteristics, a time series prediction model I is established. t+1 =αI t +(1-α)I t-1 , where α is the smoothing coefficient, ranging from 0 to 1, I t-1 For {I T}, predict the material inventory quantity I at the next k moments t+k , determine the replenishment time point t replenish At the same time, analyze the stability of material conveying speed and calculate the standard deviation of material conveying speed fluctuation σ speed ;

[0114] It should be further explained that in the material supply constraint of the task scheduling model, let the demand for material r by task i be m ir , at time t, the inventory quantity of material r is I r,t , if I r,t <m ir And t <t replenish , task scheduling needs to be adjusted, giving priority to tasks with smaller material requirements. Suppose task i was originally scheduled to start at time t, but due to material shortage, it needs to be postponed to time t+Δt. This will change the task start time s i , which in turn affects the order delivery time objective function s i +t i ≤D k At the same time, the material conveying speed is unstable, and the original task processing time t i The waiting time of the material is t wait , the material transportation speed fluctuation causes the material waiting time to become t wait,new =t wait +Δt wait , Δt wait By σ speed OK, then t i becomes t i,new =t i -t wait +t wait,new , which affects the calculation of minimizing the production cycle T and meeting the order delivery time;

[0115] Δt wait The method for determining is as follows:

[0116] Assuming that the average speed of material transportation is v, when the standard deviation of material transportation speed fluctuation is σ speed When the material waiting time increment Δt wait Calculated by the following formula:

[0117]

[0118] Among them, t i is the estimated value of the material waiting time in the original task processing time, determined based on historical data. The value of γ is generally determined by statistical analysis and actual verification of past production data. For example, after a large amount of data statistics and actual production observation, it is found that when σ speed When the ratio increases by a certain percentage, the material waiting time will increase by a specific percentage accordingly, and the value of γ can be obtained by fitting through regression analysis and other methods.

[0119] S4: Scheduling model construction: Based on the analysis results of the equipment operation analysis model, product quality analysis model, and material supply analysis model, the objective function optimization is constructed, and constraints are set according to the objective function;

[0120] It should be specifically noted that the objective functions include the optimization of minimizing the production cycle function, maximizing the equipment utilization function, and meeting the order delivery time function;

[0121] Constraints include equipment processing capacity constraints, task sequence constraints, and material supply constraints;

[0122] The process of optimizing the construction of the minimization production cycle function is as follows:

[0123] Considering the equipment failure time t fault Based on the impact of changes in product quality requirements on task processing time, the production cycle objective function is further optimized. Since equipment failure may lead to task reallocation, the set of tasks that need to be reallocated due to equipment failure is set to R. For task i∈R, its new processing time t ij,new The processing capacity of the new equipment needs to be comprehensively considered. j,new As well as factors such as task priority, assuming that the task priority coefficient is π i , then the adjusted production cycle objective function can be expressed as , where for normal task t ij Calculate according to the original rules, for task i∈R, ,In this way, high-priority tasks are prioritized while minimizing the production cycle, ,so as to reduce the impact of equipment failure on the overall production ,progress.

[0124] The optimization construction process of the function to maximize equipment utilization is as follows:

[0125] Combined with the changes in processing time caused by equipment failure and task reallocation, the equipment utilization objective function is improved. When the equipment fails, the equipment can work for a certain time, T j Need to be redefined, assuming that the working time of equipment j before failure is T j,used , the fault repair time is T j,repair , then the new working time Tj,new =T j -T j,used -T j,repair , equipment utilization rate , where t ij It needs to be calculated based on the actual situation after equipment failure and task reallocation. Through this optimization, the actual utilization efficiency of the equipment in a complex production environment can be more accurately reflected, providing more precise guidance for improving equipment utilization.

[0126] The process of optimizing the function to meet order delivery time is as follows:

[0127] Considering the material supply constraints, the task start time s i Changes and task processing time t i When factors such as unstable material delivery speed change, the order delivery time objective function is adjusted, and the time for task i to be delayed due to material shortage is Δt delay , the part of the task processing time that increases due to the unstable material conveying speed is Δt increase , then the order delivery time is adjusted to (s i +Δt delay )+(t i +Δt increase )≤D k Through this adjustment, we can ensure that in actual production, even when faced with problems such as material supply, we can accurately judge whether the order can be delivered on time, providing task scheduling with time constraints that are more in line with actual conditions.

[0128] Equipment processing capacity constraints refer to:

[0129] Assume that equipment j has a maintenance plan in the time period [t1, t2], and the equipment processing capacity c during this period j,t =0, for task i, its processing time t ij It must be within the normal processing time period of the equipment, that is, when t start,i ≥t2 and t start,i +t ij ≤t end,j , where t start,i is the start time of task i, t end,j is the end working time of equipment j in the current production cycle, and w i / t ij ≤c j,t In this way, the actual situation of equipment maintenance is fully considered to ensure that the task allocation is within the scope allowed by the equipment processing capacity.

[0130] Task sequence constraints refer to:

[0131] Suppose task groups G1 and G2. If G1 needs to be completed before G2, for task i∈G1 and task l∈G2, it must satisfy At the same time, for tasks within the task group, follow the original task sequence constraint s l ≥s i +t i ,Through this hierarchical task sequence constraint setting, it is more in line with the logical relationship between tasks in ,complex production processes, ensuring the orderly progress of production.

[0132] Material supply constraints refer to:

[0133] If a batch of materials has a quality problem, let the affected material be r′ and the set of tasks involved be M. For task i∈M, it is necessary to suspend the task or adjust the task allocation until the material quality problem is solved. Let the time to solve the material quality problem be t resolve , then for task i∈M, if t <t resolve , task i cannot start or needs to be re-planned. At the same time, considering the unstable material transportation speed, the safety range of the material transportation speed is set to [v min ,v max ], if the actual conveying speed , it is necessary to adjust the material waiting time in the task processing time according to the speed fluctuation to ensure that the material supply link is stable and does not affect the normal progress of the task.

[0134] S5: Scheduling model solution: Genetic algorithm is used to solve the task scheduling model. The iterative termination condition is set as the fitness value does not increase significantly for N=10 consecutive generations. When the termination condition is met, the chromosome with the highest fitness value is output.

[0135] It should be specifically stated that the genetic algorithm is as follows:

[0136] Set the initial population size P=50, that is, there are 50 chromosomes initially, each chromosome represents a task allocation scheme, and the chromosome encoding adopts integer encoding. For example, for n tasks and m automatic lines, the chromosome can be represented as [j1, j2, ..., j n ], where ji represents the automatic line number assigned to task i;

[0137] The roulette wheel selection method is used to calculate the probability of each chromosome being selected according to the fitness value of the chromosome; let the fitness value of chromosome k be F k , then the probability of being selected , Chromosome The fitness value of the chromosomes is selected by roulette wheel, and chromosomes with high fitness have a greater probability of being selected into the next generation, thus achieving the retention and inheritance of excellent task allocation schemes.

[0138] Set the crossover probability Pc = 0.6. For the selected chromosome pairs to perform crossover, the single-point crossover method is adopted. For example, for chromosome A = [a1, a2, ……, a n and chromosome B = [b1, b2, ……, b n , randomly select a crossover point p’ (1 < p’ < n). After crossover, new chromosomes A′ = [a1, ……, a p’ , b p’+1 , ……, b n and B′ = [b1, ……, b p’ , a p’+1 , ……, a n [[ID=十六]]] are generated. Through the crossover operation, gene exchange between different task allocation schemes is achieved to explore better task allocation combinations;

[0139] Set the mutation probability Pm = 0.01. For the mutated chromosomes, randomly select a gene position for mutation. For example, for chromosome C = [c1, c2, ……, c n , if the gene c q ’ is selected for mutation, randomly generate a new automatic line number j new (1 ≤ j new ≤ m). After mutation, the chromosome C′ = [c1, ……, c q’-1 , j new , c q’+1 , ……, c n is generated. Through the mutation operation, new genes are introduced into the population to prevent the algorithm from falling into a local optimal solution;

[0140] After the selection, crossover, and mutation operations, a new population is generated. Repeat the above steps of fitness function calculation, genetic operations, etc. for multiple generations of iterative solution. Set the iteration termination condition as the fitness value not significantly improving for N = 10 consecutive generations. When the termination condition is met, output the chromosome with the highest fitness value, that is, obtain the optimal task allocation scheme to guide the actual task scheduling of multiple automatic lines.

[0141] The fitness calculation method is as follows:

[0142] For each chromosome, that is, the task allocation scheme, calculate its corresponding production cycle T, equipment utilization rate U, and whether it meets the order delivery time constraint. Let the weight coefficients be w T , w U , w D and w T + w U + w D = 1. The fitness function F can be expressed as F = w T × T1 + w U × U + w D×δ, where δ is an indicator variable for order delivery time satisfaction. If all orders meet the delivery time, δ=1, otherwise δ=0. Through this fitness function, the three objectives of production cycle, equipment utilization, and order delivery time are balanced in the solution process;

[0143] S6: Scheduling adjustment: When a device failure triggers a scheduling adjustment, the faulty device is removed from the task scheduling model, and the processing time of each task on other available devices is recalculated. The tasks originally assigned to the faulty device are reallocated according to the remaining processing capacity of other devices and the task priority. When the order changes, the order delivery time in the objective function of the task scheduling model is updated. If the order increases production, and the material supply is sufficient, the new tasks are preferentially assigned to the automatic line with lower current load and matching processing capacity, and the scheduling plan is recalculated.

[0144] S7: Task execution: After receiving the scheduling instruction, the automatic line control terminal parses the instruction content, converts the task assignment information into equipment control signals, controls the equipment to start, stop, adjust processing parameters, and carry out production in an orderly manner according to the new task assignment plan.

[0145] The automatic line control terminal includes a central processing unit, a storage module and a communication interface;

[0146] The central processing unit (CPU) possesses powerful data processing and computing capabilities, enabling it to quickly parse received scheduling instructions and efficiently process complex task allocation information, laying the foundation for subsequent conversion into equipment control signals. For example, in a high-variety, small-batch production model, faced with frequently changing task scheduling instructions, the CPU can parse and compute large amounts of data in a short period of time, ensuring the responsiveness of the control terminal and avoiding production delays.

[0147] The storage module stores the various programs and data required for the control terminal's operation. This includes basic system software and equipment control programs, as well as historical production data and equipment parameter configuration information. Upon receiving a dispatch instruction, the control terminal quickly retrieves the relevant equipment parameters and control logic from the storage module to accurately convert task assignment information into equipment control signals. For example, when adjusting equipment processing parameters, the control terminal quickly retrieves the optimal parameter configuration for the corresponding equipment under different tasks from the storage module, enabling precise adjustment of equipment parameters.

[0148] The communication interface is used for high-speed data transmission with the server where the task scheduling model is located, the enterprise management information system, etc., to receive the latest scheduling instructions and feedback production status information. At the same time, it is also equipped with serial ports (such as RS-232, RS-485), field bus interfaces (such as Profibus, CANopen), etc., for connecting various sensors, actuators and lower-level equipment controllers. Through these communication interfaces, the control terminal can establish stable and reliable communication links with various devices in the automatic line to ensure accurate data transmission.

[0149] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions and effects do not exceed the scope of the technical solution of the present invention, shall fall within the scope of protection of the present invention.

Claims

1. A flexible task scheduling method for multiple automatic lines based on dynamic metering data, characterized in that: The steps include: S1. Parameter acquisition: acquiring a first parameter text through an acquisition terminal according to a set acquisition frequency, the first parameter text including equipment operation parameters, product quality parameters, and material supply parameters; S2. Data cleaning: Perform data cleaning operations on equipment operating parameters, product quality parameters, and material supply parameters to obtain parameter processing text; S3. Data Analysis: Build equipment operation analysis models, product quality analysis models, and material supply analysis models based on parameter processing text. The specific implementation is as follows: The equipment operation analysis model uses the ARIMA (p, d, q) model to predict equipment failure time, and the time series data {y T },y t For {y T }, the data at time t, the model expression is: Among them, the autoregressive order p=2, the difference order d=1, and the moving average order q=1. 、 is the model parameter, y t-i For {y T }, B is the backshift operator, μ is the constant term, is the white noise at time t, is the white noise at time tj; the model output is the equipment failure time t fault The predicted value of Product quality analysis model refers to the construction of a linear regression model that relates product quality to production parameters; The material supply analysis model uses time series analysis to predict material inventory consumption trends and replenishment time points. Assume that the material inventory quantity time series data is {I T }, I t For {I T By analyzing the historical data trends and seasonal characteristics, a time series prediction model I is established. t+1 =αI t +(1-α)I t-1 , where α is the smoothing coefficient, ranging from 0 to 1, I t-1 For {I T }, predict the material inventory quantity I at the next k moments t+k , determine the replenishment time point t replenish At the same time, analyze the stability of material conveying speed and calculate the standard deviation of material conveying speed fluctuation σ speed ; S4. Scheduling model construction: Based on the analysis results of the equipment operation analysis model, product quality analysis model, and material supply analysis model, the objective function optimization is constructed, and constraints are set according to the objective function to build a task scheduling model; S5. Scheduling model solution: Genetic algorithm is used to solve the task scheduling model and obtain the optimal task allocation solution.

2. The method for flexible task scheduling of multiple automatic lines based on dynamic metering data according to claim 1 is characterized in that: In S1, equipment operating parameters include speed, temperature, and pressure; product quality parameters include size; and material supply parameters include inventory quantity and material delivery speed.

3. The method for flexible task scheduling of multiple automatic lines based on dynamic metering data according to claim 1 is characterized in that: In S1, the acquisition frequency is set as follows: For the equipment operating parameters, set the acquisition frequency to 5Hz; For product quality parameters, the acquisition frequency is determined by the production rhythm. The production rhythm is to produce one product every N seconds, so the acquisition frequency is set to 1 / N Hz. For material supply parameters, set the acquisition frequency to 1 / 60Hz.

4. The method for flexible task scheduling of multiple automatic lines based on dynamic metering data according to claim 1 is characterized in that: In S2, the data cleaning operations are as follows: For equipment operating parameters, we set upper and lower thresholds to identify abnormal values, and use the sliding average method to clean abnormal values; For product quality parameters, standard thresholds are set to identify outliers, and outliers are cleaned by comparing data with similar products. For material supply parameters, outliers are identified by setting standard thresholds or upper and lower thresholds, and outliers are cleaned by comparing them with the data at the previous moment or by weighted averaging the data of adjacent time periods.

5. The method for flexible scheduling of tasks of multiple automatic lines based on dynamic metering data according to claim 1 is characterized in that: The expression of the linear regression model associated with product quality and production parameters is as follows: Y=β0+β1X1+β2X2+β3X3+ Among them, the product size qualification rate in the product quality parameter is Y, and the production parameters are equipment speed X1, equipment temperature X2 and material conveying speed X3. is a constant, and the least squares method is used to estimate the regression coefficients β0, β1, β2, and β3 to determine the quantitative relationship between product quality and production parameters.

6. The method for flexible task scheduling of multiple automatic lines based on dynamic metering data according to claim 1 is characterized in that: In S4, the scheduling model is constructed as follows: The objective functions include minimizing the production cycle function, maximizing the equipment utilization function, and meeting the order delivery time function. The process of optimizing the construction of the minimization production cycle function is as follows: Considering the equipment failure time t fault Based on the impact of changes in product quality requirements on task processing time, the production cycle objective function is optimized. Let the set of tasks that need to be reallocated due to equipment failure be R. For task i∈R, its new processing time t ij,new The processing capacity of the new equipment needs to be comprehensively considered. j,new As well as the task priority factor, assuming the task priority coefficient is π i , then the adjusted production cycle objective function is expressed as , n represents the number of tasks, m represents the number of devices; for normal tasks, the processing time , w i is the workload of task i, c j,normal is the normal processing capacity of the equipment; for task i∈R, ; The optimization construction process of the function to maximize equipment utilization is as follows: Combined with the changes in processing time caused by equipment failure and task reallocation, the equipment utilization objective function is improved. When the equipment fails, the equipment can work for a certain time, T j Need to be redefined, assuming that the working time of equipment j before failure is T j,used , the fault repair time is T j,repair , then the new working time T j,new =T j -T j,used -T j,repair , equipment utilization rate , where t ij Calculated based on actual conditions after equipment failure and task reallocation; The process of optimizing the function to meet order delivery time is as follows: Considering the material supply constraint, the task i starts at time s i Changes and processing time t of task i i When the material delivery speed is unstable, the order delivery time objective function is adjusted. Let the time for task i to be delayed due to material shortage be Δt delay , the part of the task processing time that increases due to the unstable material conveying speed is Δt increase , then the order delivery time is adjusted to (s i +Δt delay )+(t i +Δt increase )≤D k , D k For order delivery time, this adjustment ensures that in actual production, we can accurately judge whether the order can be delivered on time; Constraints include equipment processing capacity constraints, task sequence constraints, and material supply constraints; Equipment processing capacity constraints: Assume that equipment j has a maintenance plan in the time period [t1, t2], and the equipment processing capacity c during this period j,t =0, for task i, its processing time t ij It must be within the normal processing time period of the equipment, that is, t start,i ≥t2 and t start,i +t ij ≤t end,j , where t start,i is the start time of task i, t end,j is the end working time of equipment j in the current production cycle, and w i / t ij ≤c j,t ; Task sequence constraints: Suppose task groups G1 and G2. If G1 needs to be completed before G2, for task i∈G1 and task l∈G2, it must satisfy At the same time, for tasks within the task group, follow the original task sequence constraint s l ≥s i +t i ; Material supply constraints: If a batch of materials has a quality problem, let the affected material be r′ and the set of tasks involved be M. For task i∈M, it is necessary to suspend the task or adjust the task allocation until the material quality problem is solved. Let the time to solve the material quality problem be t resolve , then for task i∈M, if t <t resolve , task i cannot start or needs to be re-planned. At the same time, considering the unstable material transportation speed, the safety range of the material transportation speed is set to [v min ,v max ], if the actual conveying speed , the material waiting time in the task processing time needs to be adjusted according to the speed fluctuation.

7. The method for flexible task scheduling of multiple automatic lines based on dynamic metering data according to claim 1 is characterized in that: In S5, a genetic algorithm is used to solve the task scheduling model. The iterative termination condition is set as the fitness value no longer significantly improves for N = 10 consecutive generations. When the termination condition is met, the chromosome with the highest fitness value is output, and the optimal task allocation solution is obtained.

8. The method for flexible task scheduling of multiple automatic lines based on dynamic metering data according to claim 1 or 7, characterized in that: The specific implementation of the genetic algorithm is as follows: Set the initial population size P = 50, that is, there are 50 chromosomes initially, each chromosome represents a task allocation scheme, and the chromosome encoding adopts integer encoding. For n tasks and m devices, the chromosome is represented as [j1, j2, ..., j n ], where j i Indicates the device number assigned to task i; The roulette wheel selection method is used to calculate the probability of each chromosome being selected according to the fitness value of the chromosome; let the fitness value of chromosome k be F k , then the probability of being selected , Chromosome The fitness value of Set the crossover probability Pc = 0.

6. For the selected chromosome pairs to perform crossover, the single-point crossover method is adopted. For chromosome A = [a1, a2, ……, a n and chromosome B = [b1, b2, ……, b n , randomly select a crossover point p’, 1 < p’ < n. After crossover, generate new chromosomes A′ = [a1, ……, a p’ , b p’+1 , ……, b n and B′ = [b1, ……, b p’ , a p’+1 , ……, a n ; Set the mutation probability Pm=0.01, and for the mutated chromosome, randomly select a gene position for mutation. For chromosome C=[c1,c2,……,c n ], if gene c q’ If it is selected for mutation, a new automatic line number j is randomly generated new , 1≤j new ≤m, the mutated chromosome C′=[c1,……,c q’-1 ,j new ,c q’+1 ,……,c n ]; After selection, crossover and mutation operations, a new population is generated; Repeat the above process for multiple generations of iterative solution; set the iteration termination condition as the fitness value no longer significantly improves for N = 10 consecutive generations. When the termination condition is met, output the chromosome with the highest fitness value, and obtain the optimal task allocation solution.

9. A computer-readable storage medium storing computer program instructions that can be executed by a processor. When the processor executes the computer program instructions, the method for flexible scheduling of tasks of multiple automatic lines based on dynamic metering data as described in any one of claims 1 to 8 can be implemented.

Images