A method and system for dynamic scheduling of irradiation conveyor line pallets

Abstract

This invention relates to the field of automated control technology for irradiation transmission lines, and discloses a dynamic scheduling method and system for irradiation transmission line trays. The dynamic scheduling method includes: installing an integrated sensor module on each tray; transmitting data to a data acquisition server via wireless communication; filtering and standardizing the raw sensor data; collecting product information and identifying product characteristics based on the timing of product arrival at the transmission line entrance; establishing a mapping relationship between product characteristics and irradiation requirements; predicting equipment operating status and production load; constructing a multi-objective optimization model using an improved non-dominated sorting genetic algorithm; generating scheduling strategies using fuzzy logic decision-making and deep reinforcement learning; and implementing trajectory tracking and dynamic adjustment using a predictive control algorithm. This invention, through equipment health monitoring, system load analysis, energy consumption monitoring, and fault prediction algorithms, can comprehensively assess operating status and predict future trends.

Description

Technical Field

[0001] This invention relates to the field of automated control technology for irradiation transmission lines, and more specifically, to a method and system for dynamic scheduling of trays in irradiation transmission lines. Background Technology

[0002] With the increasing demand for irradiation treatment of high-reliability electronic devices in aerospace, nuclear power plants, and other industries, the scheduling and management of large-scale electronic component irradiation production lines face severe challenges. Existing irradiation transfer line systems generally suffer from the following technical problems:

[0003] Traditional systems lack the ability to accurately perceive the multi-dimensional status of pallets, such as position, load, temperature, and irradiation dose, making it impossible to obtain comprehensive operational status information of the pallets and resulting in a lack of accurate data support for scheduling decisions. Existing systems mainly rely on manual identification of product types and setting of irradiation parameters, lacking automated product characteristic identification and intelligent demand analysis capabilities, which affects scheduling efficiency and accuracy. Traditional scheduling systems lack the ability to comprehensively assess equipment health status, system load, and energy consumption, as well as the ability to predict faults, making it impossible to identify and prevent potential system risks in advance. Existing scheduling methods usually only consider a single optimization objective and lack comprehensive scheduling algorithms that simultaneously optimize multiple objectives such as completion time, energy consumption, and equipment utilization, making it difficult to achieve comprehensive optimization of system performance. Traditional systems lack the ability to dynamically adjust scheduling strategies based on real-time operating status and cannot effectively cope with unexpected situations and parameter changes during the production process.

[0004] In summary, there is a need for a dynamic scheduling method and system for irradiation transfer line pallets that can achieve real-time pallet status perception, intelligent product characteristic identification, comprehensive system status evaluation, and multi-objective optimized scheduling, in order to improve the automation level, operating efficiency, and system reliability of the irradiation production line. Summary of the Invention

[0005] This invention provides a dynamic scheduling method and system for irradiation transmission line trays, which solves the technical problems in related technologies such as single tray scheduling strategy, insufficient state awareness, lack of dynamic adaptability and imperfect multi-objective optimization capabilities.

[0006] This invention provides a method for dynamic scheduling of irradiation transmission line trays, comprising:

[0007] An integrated sensor module is installed on each tray, and the data is transmitted to the data acquisition server via wireless communication. The raw sensor data is filtered and standardized, and the tray status parameter matrix is ​​output.

[0008] The timing of a product's arrival at the transmission line entrance is determined based on the pallet status parameter matrix. Product information is collected and product characteristics are identified based on the timing of the product's arrival at the transmission line entrance. Intelligent identification technology is used to establish a mapping relationship between product characteristics and irradiation requirements, thereby obtaining a product requirement parameter vector.

[0009] Based on the product demand parameter vector, a long short-term memory network is used to predict the equipment operating status and production load, and output the system status assessment results.

[0010] Based on the system state assessment results, an improved non-dominated sorting genetic algorithm is used to construct a multi-objective optimization model and output the Pareto optimal solution set.

[0011] Based on the Pareto optimal solution set, a scheduling strategy is generated by fuzzy logic decision-making and deep reinforcement learning to output a tray operation scheme.

[0012] The pallet operation plan is decomposed into specific execution instructions, and a predictive control algorithm is used to achieve trajectory tracking, output real-time control status, and establish a performance monitoring and parameter optimization mechanism based on the control status to achieve dynamic adjustment.

[0013] In a preferred embodiment, the sensor module includes a triaxial accelerometer, a laser positioning sensor, an electronic scale module, a thermocouple temperature sensor, and an irradiation dose detector. The triaxial accelerometer detects the motion and vibration of the tray, with a sampling frequency set to 100 Hz, and calculates the real-time velocity and displacement of the tray through integration. The laser positioning sensor uses triangulation to determine the precise position of the tray on the transmission line, with a measurement accuracy of 2 cm. The electronic scale module uses a strain gauge load cell to monitor changes in the tray's load in real time, with a measurement range of 0–20 kg and an accuracy of 0.1 kg.

[0014] In a preferred embodiment, the product identification station includes an industrial camera and an RFID reader. The industrial camera captures product images with a resolution of 2048×1536 pixels. It uses an edge detection algorithm to identify the product outline, performs edge extraction using the Cannibal operator, and then detects straight and circular features through Hough transform to finally calculate the product's geometric parameters. The RFID reader reads the product tag information.

[0015] In a preferred embodiment, the long short-term memory network prediction includes assessing the equipment's operating status by installing vibration sensors, current sensors, and temperature sensors to analyze the equipment's vibration spectrum characteristics to identify mechanical fault signs, monitoring motor current waveforms to identify electrical problems, and using temperature trend analysis to determine the equipment's thermodynamic state.

[0016] In a preferred embodiment, the improved non-dominated sorting genetic algorithm includes an initialization phase, a selection operation, a crossover operation, a mutation operation, and an environment selection phase. The selection operation adopts the tournament selection method, the crossover operation adopts the sequential crossover method, and the mutation operation adopts a combination of insertion mutation and exchange mutation.

[0017] In a preferred embodiment, the fuzzy logic decision-making includes establishing a fuzzy decision rule base, setting corresponding priority rules for different production modes, and the fuzzy inference engine automatically matching appropriate decision rules based on the current system state and production priority.

[0018] In a preferred embodiment, the predictive control includes a distributed control system with a master-slave control architecture, which issues scheduling commands via industrial Ethernet, and the controller uses a combination of proportional-integral-derivative control and feedforward control to achieve trajectory tracking.

[0019] In a preferred embodiment, the performance monitoring and parameter optimization includes adaptive optimization of the parameters of the multi-objective optimization algorithm and the control parameters, training the parameter optimization model using an online learning algorithm, and dynamically adjusting the learning step size through an adaptive learning rate adjustment mechanism.

[0020] In a preferred embodiment, the dynamic adjustment further includes establishing a multi-level security protection system, including hardware security protection devices and software security protection functions, which automatically triggers protection actions when a security risk is detected.

[0021] In a preferred embodiment, an irradiation transmission line tray dynamic scheduling system is used to execute the above-described irradiation transmission line tray dynamic scheduling method, comprising:

[0022] The data acquisition module is used to install an integrated sensor module on each tray, transmit data to the data acquisition server via wireless communication, filter and standardize the raw sensor data, and output a tray status parameter matrix.

[0023] The product identification module is used to determine the timing of a product's arrival at the transmission line entrance based on the tray status parameter matrix, collect product information and identify product characteristics based on the timing of the product's arrival at the transmission line entrance, and establish a mapping relationship between product characteristics and irradiation requirements using intelligent identification technology to obtain a product requirement parameter vector.

[0024] The status assessment module is used to predict equipment operating status and production load based on product demand parameter vectors and using long short-term memory networks, and output system status assessment results.

[0025] The optimization calculation module is used to construct a multi-objective optimization model based on the system state evaluation results using an improved non-dominated sorting genetic algorithm, and output a Pareto optimal solution set.

[0026] The strategy generation module is used to generate scheduling strategies based on Pareto optimal solution sets, using fuzzy logic decision-making and deep reinforcement learning, and output tray operation schemes.

[0027] The execution control module is used to decompose the pallet operation plan into specific execution instructions, use predictive control algorithms to achieve trajectory tracking, output real-time control status, and establish a performance monitoring and parameter optimization mechanism based on the control status to achieve dynamic adjustment.

[0028] The beneficial effects of this invention are as follows:

[0029] By integrating multiple sensors and advanced data processing algorithms, the system can acquire comprehensive status information such as the position, speed, load, temperature and radiation dose of the pallet in real time. It adopts technologies such as triangulation principle, filtering, anomaly detection and standardization to ensure the accuracy and reliability of status data, providing a precise data foundation for intelligent scheduling.

[0030] By integrating industrial cameras, RFID readers, and laser scanners, and combining image processing algorithms and fuzzy logic reasoning, the system can automatically identify product characteristics and intelligently analyze irradiation requirements, realizing the automated mapping between product characteristics and scheduling parameters, and improving the intelligence level and processing efficiency of scheduling decisions.

[0031] Through equipment health monitoring, system load analysis, energy consumption monitoring, and fault prediction algorithms based on long short-term memory networks, the system can comprehensively assess the operating status and predict future trends, realizing the transformation from passive response to proactive prevention and effectively improving the reliability and stability of the system.

[0032] By constructing a multi-objective optimization model that comprehensively considers completion time, energy consumption, and equipment utilization, and by adopting an improved non-dominated sorting genetic algorithm and dynamic weight adjustment mechanism, the system can find the Pareto optimal solution set under complex constraints, achieving coordinated optimization of multiple performance indicators and improving the overall system efficiency.

[0033] Through real-time performance monitoring, execution deviation analysis, adaptive parameter adjustment, and reinforcement learning algorithms, the system has the ability to dynamically adjust the scheduling strategy according to the running status, and can continuously learn and optimize scheduling performance, achieving a technological breakthrough from static scheduling to dynamic intelligent scheduling.

[0034] Through precise path planning algorithms, time scheduling algorithms, and distributed instruction execution mechanisms, combined with motion control algorithms and safety protection mechanisms, the system can accurately translate scheduling decisions into tray control actions, ensuring high-precision execution of scheduling instructions and safe and reliable system operation.

[0035] Intelligent scheduling has improved production efficiency, reduced energy consumption, and optimized equipment utilization, providing advanced technical solutions for the irradiation processing industry. It has significant industrial application value and promotion prospects, and helps to promote technological progress and transformation and upgrading of related industries. Attached Figure Description

[0036] Figure 1 This is a flowchart of a dynamic scheduling method for irradiation transmission line trays in this invention;

[0037] Figure 2 This is a block diagram of a dynamic scheduling system for irradiation transmission line trays in this invention;

[0038] Figure 3 It is a bar chart comparing the improvements of the dynamic scheduling system over the traditional scheduling system in various key indicators;

[0039] Figure 4 It is a time series graph of the real-time optimization effect of the dynamic scheduling system during 24-hour continuous operation;

[0040] Figure 5 It is a scatter plot showing the distribution of Pareto optimal solutions generated by a multi-objective optimization algorithm in a two-dimensional space. Detailed Implementation

[0041] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0042] At least one embodiment of the present invention discloses a dynamic scheduling method for irradiation transmission line trays, such as... Figure 1 As shown, it includes:

[0043] Step 1: Install an integrated sensor module on each tray and transmit the data to the data acquisition server via wireless communication. Filter and standardize the raw sensor data and output the tray status parameter matrix.

[0044] This step enables comprehensive status monitoring of all trays on the transmission line. Each tray is equipped with an integrated sensor module, including a triaxial accelerometer, a laser positioning sensor, an electronic scale module, a thermocouple temperature sensor, and an irradiation dose detector. The sensor modules transmit data to the data acquisition server wirelessly.

[0045] The specific process is as follows: A triaxial accelerometer detects the motion and vibration of the tray, with a sampling frequency set to 100 Hz. The real-time velocity and displacement of the tray are calculated through integration. Specifically, the acceleration data is integrated once to obtain the velocity; the velocity data is integrated a second time to obtain the displacement. The time interval for integration calculation is 0.01 seconds. A laser positioning sensor uses the triangulation principle to determine the precise position of the tray on the transmission line, with a measurement accuracy of 2 cm. The electronic scale module uses a strain gauge load cell to monitor changes in the tray's load in real time, with a measurement range of 0–20 kg and an accuracy of 0.1 kg. A thermocouple temperature sensor monitors the tray temperature, with a measurement range of -20℃ to 100℃ and an accuracy of 0.5℃. An irradiation dose detector uses a silicon detector to monitor the cumulative irradiation dose in real time, with a measurement range of 0–1000 Gy and an accuracy of 1%.

[0046] The specific calculation method of the triangulation principle is as follows: The laser sensor emits a laser beam onto the reflector. By measuring the emission angle and reception angle of the laser beam and the baseline distance from the sensor to the reflector, the distance from the tray to the reflector is calculated.

[0047]

[0048] Where d represents the distance from the tray to the reflector, L represents the baseline distance from the sensor to the reflector, α represents the emission angle of the laser beam, β represents the reception angle of the laser beam, and sin represents the sine function.

[0049] The absolute position coordinates of the tray were calculated using the known position of the reflector and the measured distance:

[0050] x p =x r -dcos(θ);

[0051] y p =y r -dsin(θ);

[0052] Where, x p y p These represent the x and y coordinates of the pallet, respectively. r y r Let x and y represent the x and y coordinates of the reflector, respectively; d represents the distance from the tray to the reflector; θ represents the angle between the laser beam and the transmission line; sin represents the sine function; and cos represents the cosine function.

[0053] The data preprocessing process includes: filtering the raw sensor data (including raw data collected by triaxial accelerometer, laser positioning sensor, electronic scale module, thermocouple temperature sensor and radiation dose detector) to remove high-frequency noise and outliers; synchronizing and aligning the data from different sensors in time; and standardizing the data to unify parameters of different dimensions into the 0-1 range.

[0054] The specific implementation method of filtering is as follows: a Butterworth low-pass filter is used for the acceleration sensor data, with a cutoff frequency set to 20Hz and a filter order of 4, to remove high-frequency vibration noise; a Kalman filter is used for the position sensor data. By establishing the system state equation and observation equation, and combining the process noise covariance and observation noise covariance parameters, the optimal estimation and prediction of the position data can be achieved, effectively removing measurement noise and system errors.

[0055] The specific implementation method of time synchronization alignment is as follows: the Network Time Protocol (NTP) is used to ensure the clock synchronization of each sensor node, and the clock accuracy reaches the millisecond level; for sensor data with different sampling frequencies, the timestamp interpolation method is used for alignment. The interpolation algorithm adopts the cubic spline interpolation method. By constructing a piecewise cubic polynomial function, the time alignment of sensor data with different sampling frequencies is achieved under the condition that the function and its first and second derivatives are continuous.

[0056] Outlier detection uses the 3σ criterion. When data deviates from the mean by more than 3 times the standard deviation, it is identified as an outlier and is removed or interpolated.

[0057] For single outliers, a linear interpolation method is used:

[0058]

[0059] Where, x interp Indicates the interpolation result, x i-1 and x i+1 These represent normal data points before and after the outlier, respectively, where t represents the current time point. i-1 and t i+1 These represent the timestamps of the preceding and following normal data points, respectively.

[0060] For consecutive outliers (more than 3 consecutive points), an ARIMA model based on historical data is used for prediction interpolation, and the model parameters are determined by maximum likelihood estimation. After interpolation, the interpolation points are marked, and their weight coefficients are reduced to 0.5 in subsequent data analysis to reduce the impact of interpolation error on the overall algorithm performance.

[0061] The specific method for standardization is as follows: Minimum and maximum standardization methods are used, with the following formula:

[0062]

[0063] Where, x norm This represents the standardized value, where x represents the original value. min and x max These represent the historical minimum and maximum values ​​of the parameter, respectively.

[0064] The standardized processing ranges for different parameters are as follows: the minimum value for position coordinates is 0 meters, and the maximum value is the transmission line length of 200 meters; the minimum value for speed is 0 meters per second, and the maximum value is the maximum permissible speed of 5 meters per second; the minimum value for load is 0 kilograms, and the maximum value is the maximum load of 20 kilograms; the minimum value for temperature is -20 degrees Celsius, and the maximum value is 100 degrees Celsius; the minimum value for irradiation dose is 0 grays, and the maximum value is the maximum dose of 1000 grays.

[0065] Standardized data ensures that all parameters are within the same order of magnitude, avoiding excessive numerical differences that could affect algorithm performance.

[0066] The output is a tray status parameter matrix, which includes real-time data for five dimensions: position coordinates, moving speed, load, temperature, and cumulative irradiation dose.

[0067] Furthermore, machine learning-based sensor fusion algorithms can be used to replace traditional data preprocessing methods. This algorithm employs a Kalman filter to fuse multi-sensor data, improving the accuracy of state parameter estimation through state estimation and covariance matrix updates. Simultaneously, an anomaly detection algorithm is introduced to establish normal operating modes based on historical data, automatically identifying and removing outlier data points.

[0068] Step 2: Determine the timing of the product's arrival at the transmission line entrance based on the pallet status parameter matrix. Collect product information and identify product characteristics based on the timing of the product's arrival at the transmission line entrance. Use intelligent identification technology to establish a mapping relationship between product characteristics and irradiation requirements to obtain the product requirement parameter vector.

[0069] This step uses automatic identification technology to acquire product information and analyze its irradiation treatment requirements. A product identification station, equipped with a high-resolution industrial camera, RFID reader, and laser scanner, is set up at the transmission line entrance.

[0070] Product Feature Identification Process: An industrial camera captures product images at a resolution of 2048×1536 pixels. Image processing algorithms extract the product's geometric features, including length, width, height, and volume. An edge detection algorithm identifies the product's outline, using the Canny operator for edge extraction, followed by Hough transform to detect straight lines and circles, ultimately calculating the product's geometric parameters. An RFID reader reads the product tag information, obtaining basic information such as product code, batch number, and production date. A laser scanner performs a 3D contour scan of the product, acquiring precise volume and shape data.

[0071] Requirements analysis process: A mapping database between product characteristics and irradiation requirements is established, containing standard irradiation parameters for different product types. The database is queried based on the product code to obtain process parameters such as standard irradiation dose, processing time, and maximum allowable temperature for that product. For special products, the system supports manual input of customized irradiation parameters. A fuzzy logic reasoning algorithm is used to calculate the optimal transmission speed and irradiation time allocation based on the product's geometric features and material properties.

[0072] Product geometric feature standardization: A multi-level standardization system for product characteristic parameters is established. A hierarchical standardization mechanism is established based on product category, with different product types using independent sets of standardized parameters to avoid data interference between different product types. For product length, width, and height parameters, a composite processing method combining logarithmic transformation and maximum / minimum standardization is used. Logarithmic transformation first compresses the dynamic range of the data before standardization, effectively handling large variations in product dimensions. For product volume parameters, cube root transformation standardization is used to convert three-dimensional volume data into one-dimensional linear features, facilitating subsequent algorithm processing. For product density parameters, material type-based grouping standardization is used, dividing density data into different groups such as metals, ceramics, and polymers based on product material properties, with each group using independent standardization parameters. A dynamic update mechanism for product characteristic parameters is established, periodically updating standardized parameters based on the addition of new product types and the accumulation of historical data to ensure the accuracy and adaptability of standardization processing. A product characteristic data integrity verification mechanism is established to identify and complete missing or abnormal product characteristic data, using interpolation methods based on similar products to fill in missing data.

[0073] The specific implementation of the fuzzy logic inference algorithm is as follows: First, a fuzzy rule base is established, containing the mapping relationship between product geometric features and transmission parameters; then, the input geometric features are fuzzified, and the precise values ​​are converted into fuzzy sets using the triangle membership function; next, the recommended transmission speed value is calculated by the fuzzy inference engine according to preset rules and weight coefficients, with the weight coefficients determined based on product type and historical experience; finally, defuzzification is performed using the centroid method, and the fuzzy output is converted into a precise recommended transmission speed value using a discretization calculation method.

[0074] Key algorithms in the processing include: product image preprocessing, which uses median filtering to remove noise and histogram equalization to enhance image contrast; geometric feature extraction, which uses morphological operations to remove small noise points in the image and uses connected component analysis to separate independent product regions; and RFID data parsing, which parses tag information according to a predefined data format and verifies data integrity and validity.

[0075] Non-numerical data coding conversion: Establish a complete classification data coding preprocessing system. For product type coding, a hierarchical coding strategy is adopted, combining the advantages of one-hot coding and sequential coding. Sequential coding is used for product types with clear hierarchical relationships, while one-hot coding is used for those without, ensuring that the coding results maintain the distinguishability between categories while avoiding false sequential relationships. For equipment status coding, a tree-like coding structure is used, encoding equipment status according to levels such as normal operation, minor anomalies, and serious faults. Each level is further subdivided into sub-codes, forming a multi-level tree-like coding system. For fault type coding, a weighted coding method based on fault severity and impact range is used, with the severity and impact range of the fault type as weighting factors to generate numerical codes reflecting the importance of the fault. A dynamic maintenance mechanism for the coding dictionary is established, automatically expanding the coding dictionary and maintaining coding consistency when new product types, equipment statuses, or fault types appear. A coding validity verification mechanism is established, periodically checking the completeness and accuracy of the coding dictionary, identifying and correcting coding errors or conflicts. For the coded classification data, a data distribution balance check mechanism is established to identify cases of unbalanced category distribution and to process them using resampling or weight adjustment methods.

[0076] The output is a vector of product requirement parameters, which includes four key parameters: required irradiation dose, processing time, recommended transmission speed, and maximum allowable temperature.

[0077] Furthermore, a product identification method based on deep learning can be employed. Training a convolutional neural network model can directly identify product types and predict irradiation requirements from product images, without relying on RFID tags. This method improves the system's adaptability and identification accuracy, and is particularly suitable for the rapid identification of new product types.

[0078] Step 3: Based on the product demand parameter vector, use a long short-term memory network to predict the equipment operating status and production load, and output the system status assessment results;

[0079] This step provides a comprehensive assessment of the overall operating status of the transmission line system and predicts future operating trends. Monitoring includes equipment health status, system load levels, energy consumption, and potential failure risks.

[0080] Equipment health status assessment: Vibration sensors, current sensors, and temperature sensors are installed on critical equipment to monitor operating parameters. By analyzing the equipment's vibration spectrum characteristics, mechanical fault signs such as bearing wear and imbalance are identified. Motor current waveforms are monitored, and electrical problems such as winding short circuits and bearing failures are identified through spectrum analysis. Temperature trend analysis is used to determine the equipment's thermodynamic state and provide early warnings of overheating and other abnormal conditions.

[0081] System load analysis: Real-time statistics on the number of pallets and load distribution on each transmission line are collected to calculate the total system load and load balance. Queuing theory models are used to analyze the system's processing capacity and bottlenecks, predicting system performance under different load conditions. A load forecasting model is established based on historical production data and order information to predict future production load trends.

[0082] Energy consumption monitoring and analysis: Install smart meters on major electrical equipment to monitor power consumption in real time. Establish an energy consumption baseline model, compare the difference between actual and theoretical energy consumption, and identify equipment with abnormal energy consumption. Calculate energy efficiency under different operating modes using energy efficiency analysis algorithms to provide data support for energy-saving optimization.

[0083] Fault prediction algorithm: Employing a time series prediction model based on a long short-term memory network, this algorithm learns the changing patterns of equipment operating parameters to predict the timing of equipment failures. A fault pattern library is established, recording historical fault cases and characteristics; potential fault risks are identified through pattern matching. A multi-level early warning mechanism is implemented, triggering different levels of warning signals based on the fault probability and severity.

[0084] The data processing includes: cleaning and denoising of raw monitoring data; time synchronization and spatial alignment of multi-source data; feature extraction and dimensionality reduction to extract key state feature parameters; and training and parameter optimization of the state assessment model.

[0085] The output results include equipment health scores, system load levels, energy efficiency indicators, and fault prediction results, as well as predictive analysis of the system status in the future.

[0086] Furthermore, digital twin technology can be used to create a virtual model of the transmission line system. Through real-time synchronization between the physical system and the virtual model, accurate modeling and prediction of the system state can be achieved. The digital twin model can simulate system performance under different scheduling strategies, providing a virtual verification environment for scheduling decisions.

[0087] Step 4: Based on the system state evaluation results, construct a multi-objective optimization model using an improved non-dominated sorting genetic algorithm and output the Pareto optimal solution set;

[0088] This step involves constructing a mathematical model that comprehensively considers multiple optimization objectives, thereby achieving scientific and optimized scheduling decisions. The optimization objectives include minimizing total completion time, minimizing energy consumption, and maximizing equipment utilization.

[0089] Objective function design: The first objective function is to minimize the total completion time, which is the total time from when all products enter the system to when they complete irradiation treatment. This is calculated by summing the differences between the completion time and start time of each product. The second objective function is to minimize the total energy consumption, which is the total power consumption of all devices throughout the entire scheduling cycle. This is calculated by summing the products of the power consumption and operating time of each device. The third objective function is to maximize the average equipment utilization rate, which is the average utilization rate of all devices. This is calculated by averaging the ratio of the actual operating time of each device to the total available time.

[0090] Multi-objective function numerical scaling unification: Establishing a standardized preprocessing mechanism for system performance indicators. For completion time indicators, a maximum-minimum standardization method based on historical data is adopted. The normal range of completion time is derived from historical production data statistics, and the actual completion time is standardized to a range of 0 to 1, where 0 represents the best historical completion time and 1 represents the worst historical completion time. For energy consumption indicators, a relative standardization method based on theoretical optimal values ​​is adopted. The theoretical minimum energy consumption is used as the benchmark value, and the actual energy consumption is expressed as a multiple of the theoretical optimal value. Then, logarithmic transformation and standardization are performed to effectively handle the nonlinear characteristics of energy consumption data. For equipment utilization rate indicators, a deviation standardization method based on target values ​​is adopted. The ideal utilization rate of the equipment is used as the target value, and the deviation between the actual utilization rate and the target value is standardized. Positive deviation indicates overutilization, and negative deviation indicates underutilization. A dynamic update mechanism for standardized parameters is established. Based on changes in system operating status and production demands, the value range and benchmark values ​​of standardized parameters are updated periodically to ensure the timeliness and accuracy of the standardization process. An adaptive adjustment mechanism for the weights of objective functions is established. Based on the current production priority and system status, the weight coefficients of each objective function in the comprehensive evaluation are dynamically adjusted to achieve flexibility and adaptability in multi-objective optimization.

[0091] Constraints are set as follows: Irradiation quality constraints ensure that the irradiation dose received by each product is within the specified range, avoiding excessive or insufficient irradiation; equipment capacity constraints limit the maximum load and processing capacity of each piece of equipment; safety constraints include minimum safe distance between pallets, maximum moving speed limits, etc.; time window constraints ensure that urgent orders are processed within a specified time; resource constraints limit the number of products processed simultaneously and the allocation of pallets.

[0092] Algorithm Selection and Implementation: An improved non-dominated sorting genetic algorithm is used to solve the multi-objective optimization problem. The algorithm includes an initialization phase, selection operation, crossover operation, mutation operation, and environment selection phase. In the initialization phase, an initial population is randomly generated, with each individual representing a scheduling scheme. The selection operation uses a tournament selection method, choosing superior individuals based on their dominance and crowding distance. The crossover operation uses a sequential crossover method to maintain the validity of the scheduling sequence. The mutation operation combines insertion mutation and exchange mutation to increase population diversity. The environment selection uses fast non-dominated sorting and crowding distance calculation to maintain a constant population size.

[0093] Dynamic weight adjustment mechanism: Based on the current production priority and system status, the weight coefficients of each objective function are dynamically adjusted. When the system load is high, the weight of the equipment utilization target is increased; when there are many urgent orders, the weight of the completion time target is increased; when energy consumption exceeds the limit, the weight of the energy consumption control target is increased. The weight adjustment adopts a fuzzy control method, automatically determining the weight allocation based on a fuzzy rule base.

[0094] Algorithm parameter settings: population size set to 100, number of generations set to 200, crossover probability set to 0.9, mutation probability set to 0.1. The algorithm terminates when the maximum number of generations is reached or no significant improvement is observed for 20 consecutive generations.

[0095] The output is a Pareto optimal solution set, which includes multiple optimal scheduling schemes under different weight combinations, as well as the recommended best scheduling strategy.

[0096] Based on multi-objective optimization theory, a mathematical model for comprehensive scheduling optimization is established, and the optimal scheduling strategy is derived, specifically expressed as:

[0097] minF(x1)=[f1(x1),f2(x1),f3(x1)];

[0098] Where, min represents minimizing the objective function vector, F(x1) represents the objective function vector consisting of three objective functions, and x1 represents the scheduling scheme, including all scheduling parameters of the trays (such as path, speed, time, etc.). Each objective function is defined as follows:

[0099]

[0100] Where f1(x1) is the total completion time, f2(x1) is the total energy consumption, f3(x1) is the negative value of the average equipment utilization rate, n1 represents the total number of products, and m represents the total number of equipment. and These represent the completion time and start time of the i1th product under scheduling scheme x1, respectively. This represents the power consumption of the i1th product under scheduling scheme x1. This represents the processing time of the i1th product under scheduling scheme x1. This represents the utilization rate of the i2th device under scheduling scheme x1.

[0101] Furthermore, a multi-objective algorithm based on particle swarm optimization can be used to replace the genetic algorithm. Particle swarm optimization has the advantages of fast convergence speed and simple parameter settings, making it particularly suitable for continuous optimization problems. By introducing a multi-objective particle swarm optimization algorithm, a high-quality scheduling solution can be found in a shorter time.

[0102] Step 5: Based on the Pareto optimal solution set, a scheduling strategy is generated using fuzzy logic decision-making and deep reinforcement learning to output the tray operation scheme;

[0103] This step generates a specific scheduling strategy based on the multi-objective optimization results and performs real-time optimization. The scheduling strategy includes three core elements: pallet path planning, speed adjustment, and time scheduling.

[0104] Pareto Solution Set Analysis: This section analyzes the Pareto optimal solution set generated by multi-objective optimization algorithms to evaluate the performance characteristics and applicable scenarios of each solution. The Pareto solutions are sorted and filtered using a combination of the ideal point distance method and the entropy weight method. The ideal point distance method calculates the Euclidean distance from each solution to the ideal point; a smaller distance indicates better overall performance. The entropy weight method automatically determines the weights based on the distribution characteristics of each objective function value, avoiding biases from subjective weight settings.

[0105] Decision-making scheme selection: A fuzzy logic-based decision-making method is used to select the most suitable scheduling scheme from the Pareto solution set to meet current production needs. A fuzzy decision rule base is established, containing decision rules for different production scenarios. For example, in normal production mode, the scheme with lower energy consumption and higher equipment utilization is prioritized; in emergency production mode, the scheme with the shortest completion time is prioritized; and in energy-saving production mode, the scheme with the lowest energy consumption is prioritized. The fuzzy inference engine automatically matches appropriate decision rules based on the current system state and production priority.

[0106] Reinforcement Learning Optimization: A reinforcement learning model based on a deep Q-network is established to optimize the scheduling strategy through continuous interaction with the environment. The state space includes information such as the current pallet distribution, system load, and equipment status; the action space includes pallet scheduling instructions, speed adjustment instructions, etc.; the reward function comprehensively considers factors such as completion time, energy consumption, and equipment utilization. The agent continuously improves the scheduling strategy through trial and error learning to adapt to the dynamically changing production environment.

[0107] Adaptive Parameter Adjustment: An adaptive adjustment mechanism is designed to dynamically optimize scheduling parameters based on system feedback. The execution effect of the scheduling strategy is monitored, including the deviation between actual and expected completion times, energy consumption levels, and equipment utilization. When the execution effect deviates from the expected target, the parameter adjustment mechanism is triggered, optimizing the scheduling parameters through gradient descent or a genetic algorithm. A feedback loop for parameter adjustment is established to ensure continuous improvement of the scheduling system.

[0108] Path planning algorithm: An improved A* algorithm is used for tray path planning, considering the physical constraints of the transmission line and dynamic obstacles. The algorithm finds the optimal path from the starting point to the destination through heuristic search, with the heuristic function comprehensively considering distance and congestion levels. For dynamic obstacles, a dynamic window method is used to achieve real-time obstacle avoidance, ensuring the safety of tray operation.

[0109] The specific calculation method for the heuristic function is as follows:

[0110] h(n) = w1·d euclidean (n,goal)+w2·ρ congestion (n)+w3·c energy (n);

[0111] Where h(n) represents the heuristic function value of node n, d euclidean (n,goal) represents the Euclidean distance from node n to the target point goal, ρ congestion (n) represents the congestion density at node n, c energy (n) represents the estimated energy cost to reach node n, and w1, w2, and w3 represent the distance weight, congestion weight, and energy consumption weight, respectively. The weight coefficients are set as follows: distance weight 0.6, congestion weight 0.3, and energy consumption weight 0.1.

[0112] The formula for calculating congestion density is:

[0113]

[0114] Where, N tray (R) represents the number of pallets within radius R, where R represents the statistical radius, set to 5 meters; π represents pi; ρ congestion (n) represents the congestion density at node n.

[0115] Time scheduling algorithm: Based on the critical path method, time scheduling identifies critical tasks and bottlenecks affecting the total completion time. Through task dependency analysis and resource constraint modeling, the earliest start time and latest finish time for each task are calculated. Resource balancing techniques are employed to balance resource demands across different time periods while meeting time constraints.

[0116] The output includes the optimal scheduling strategy, specifically the running path, speed curve, and time schedule for each tray, as well as real-time adjustment suggestions for scheduling parameters.

[0117] Furthermore, a scheduling strategy generation method based on graph neural networks can be employed. The transmission line system is modeled as a graph structure, with nodes representing trays and devices, and edges representing connections. Graph neural networks can learn the complex relationships within the graph structure, generating more intelligent scheduling strategies.

[0118] Step 6: Decompose the pallet operation plan into specific execution instructions, use predictive control algorithms to achieve trajectory tracking, output real-time control status, and establish a performance monitoring and parameter optimization mechanism based on the control status to achieve dynamic adjustment;

[0119] Distributed scheduling instruction execution:

[0120] This step decomposes the global scheduling strategy into local execution instructions for each pallet, and achieves precise execution through a distributed control system. The execution system adopts a master-slave control architecture, with the master controller responsible for global coordination and the slave controllers responsible for the motion control of individual pallets.

[0121] Command Decomposition and Issuance: The global scheduling strategy is decomposed into pallets, generating detailed execution instructions for each pallet. The instructions include target position coordinates, motion trajectory point sequence, target velocity for each trajectory segment, acceleration limits, and arrival time requirements. A timestamp synchronization mechanism ensures all pallets execute scheduling instructions according to a unified time base. Scheduling instructions are issued to each pallet controller via industrial Ethernet, using Modbus TCP as the communication protocol to ensure reliable data transmission.

[0122] Motion Control Algorithm: Each tray is equipped with an independent motion controller, employing a trajectory tracking algorithm based on predictive control. The predictive control algorithm uses model predictive control to establish a tray dynamics model and predict the future motion state of the tray in each control cycle. By optimizing quadratic performance indicators and comprehensively considering trajectory tracking error and control energy consumption, the algorithm uses quadratic programming to solve for the optimal control sequence under the conditions of satisfying control amplitude constraints and control increment constraints, ensuring that the tray runs accurately along the planned trajectory. The controller uses a combination of PID control and feedforward control. The PID controller eliminates trajectory tracking error, while the feedforward controller compensates for known disturbances and model uncertainties.

[0123] Control system parameter standardization: A dimensionless preprocessing system for motion control parameters is established. Position parameters are standardized based on the physical scale of the transmission line, converting absolute position coordinates into dimensionless positions relative to the transmission line length, eliminating the influence of physical scale on the control algorithm. Velocity parameters are normalized based on the maximum permissible speed, standardizing actual speed values ​​to the range of 0 to 1, where 1 represents the maximum safe speed allowed by the system. Acceleration parameters are standardized based on equipment performance limitations, determining the maximum acceleration limit according to the physical characteristics of the pallet and transmission equipment, and standardizing actual acceleration values ​​to the permissible range. Control force parameters are standardized based on actuator capability, determining the standardized range of control force according to the maximum output capability of the drive motor. An adaptive adjustment mechanism for control parameter constraints is established, dynamically adjusting the constraint range and standardized parameters based on changes in factors such as pallet load, transmission line slope, and environmental conditions. A smooth transition mechanism for control parameters is established, employing a gradual adjustment method to avoid abrupt changes and oscillations in the control system when control parameters undergo significant changes. For feedback signals of the control system, establish a signal quality assessment and preprocessing mechanism, and improve the quality and reliability of feedback signals through filtering, denoising and signal reconstruction techniques.

[0124] Positioning and Correction: A combination of laser and magnetic navigation is used to ensure high-precision control of the pallet's position. Laser reflectors are deployed along the transmission line, and laser sensors on the pallet determine their position using triangulation. Simultaneously, magnetic conductors are buried in the ground, and magnetic sensors on the bottom of the pallet detect changes in magnetic field strength to achieve path tracking. These two positioning methods serve as backups for each other, improving system reliability. When a positional deviation is detected, the controller automatically corrects the trajectory to ensure the pallet follows the predetermined path.

[0125] Speed ​​Control and Adjustment: The pallet operating speed is dynamically adjusted based on scheduling strategies and real-time traffic conditions. A speed control model is established, considering the impact of factors such as load variations, slope effects, and friction coefficients on speed. An adaptive control algorithm is employed to adjust control parameters based on actual operational performance. Speed ​​is automatically reduced in densely populated pallet areas to avoid collision risks; speed is appropriately increased in open areas to improve operational efficiency.

[0126] Safety Protection Mechanism: A multi-level safety protection system is established, encompassing both hardware and software security layers. Hardware security includes physical protection devices such as emergency stop buttons, anti-collision sensors, and limit switches. Software security includes logical protection functions such as speed limits, acceleration limits, and safe distance monitoring. When a safety risk is detected, the system automatically triggers protective actions to ensure the safety of personnel and equipment.

[0127] Execution Performance Monitoring: Real-time monitoring of the execution of scheduling commands, including indicators such as trajectory tracking accuracy, speed control effectiveness, and arrival time deviation. An execution performance evaluation model is established to quantitatively analyze the actual execution quality of the scheduling strategy. When the execution performance deviates from expectations, timely feedback is provided to the higher-level control system, triggering dynamic adjustments to the scheduling strategy.

[0128] The output is real-time status information of the pallet motion control, including current position, speed, acceleration, execution progress, and an evaluation report of the execution effect.

[0129] Furthermore, a distributed control architecture based on edge computing can be adopted. Edge computing devices are deployed at each control node to achieve local data processing and decision-making. This architecture reduces data transmission latency and improves the system's real-time response capability, making it particularly suitable for scheduling scenarios with extremely high real-time requirements.

[0130] Dynamic feedback and adaptive adjustment:

[0131] This step establishes a closed-loop feedback mechanism, enabling dynamic optimization and adaptive adjustment of scheduling strategies through continuous monitoring and analysis of system operation status. The feedback system comprises four core modules: performance monitoring, deviation analysis, strategy adjustment, and parameter optimization.

[0132] Performance Monitoring and Data Acquisition: Establish a comprehensive performance monitoring system to collect key performance indicators (KPIs) of the system in real time. Monitoring indicators include average processing time, system throughput, equipment utilization, energy consumption, scheduling response time, and task completion rate. Real-time operational data is acquired from various subsystems through a distributed data acquisition network. The data acquisition frequency is set according to the importance and frequency of change of the indicators; key indicators are collected at the millisecond level, while general indicators are collected at the second level. A data quality inspection mechanism is established to automatically identify and handle abnormal data points.

[0133] Deviation analysis involves comparing the expected results of the scheduling strategy with the actual execution results to analyze the causes and extent of the deviations. A deviation analysis model is established, categorizing deviations into systematic and random deviations. Systematic deviations are typically caused by inaccurate model parameters or unreasonable algorithm assumptions, requiring model optimization and parameter correction. Random deviations are mainly caused by environmental disturbances and measurement noise, and are handled using statistical methods. Analysis of variance (ANOVA) is employed to identify the main factors affecting system performance, providing a basis for subsequent optimization.

[0134] Strategy Adjustment Decision: Based on the deviation analysis results, corresponding scheduling strategy adjustment schemes are formulated. A strategy adjustment decision tree is established, triggering corresponding adjustment actions according to different types of deviations. When the completion time deviation is large, priority is given to adjusting task priority and resource allocation; when the energy consumption deviation is large, the focus is on optimizing equipment operating modes and idle time; when the equipment utilization deviation is large, load allocation and task scheduling are rebalanced. A multi-criteria decision-making method is adopted to comprehensively consider factors such as adjustment effect, implementation cost, and risk level to select the optimal adjustment scheme.

[0135] Adaptive Parameter Optimization: An adaptive parameter optimization mechanism is established to automatically adjust algorithm parameters based on system operating status and environmental changes. The optimization scope includes genetic algorithm parameters such as population size, generation number, and crossover / mutation probability for multi-objective optimization algorithms, as well as control parameters such as proportional, integral, and derivative coefficients of PID controllers. An online learning algorithm is employed to train the parameter optimization model using historical data. An adaptive learning rate adjustment mechanism is introduced to dynamically adjust the learning step size based on optimization performance. Constraints on parameter changes are established to prevent excessively drastic parameter adjustments from affecting system stability.

[0136] Model Updates and Learning: Regularly update the system model and knowledge base to incorporate new operational experience and optimization results. Establish a model version management mechanism to record the content and effects of each model update. Employ incremental learning methods to learn new patterns and rules without disrupting existing knowledge. Combine the prediction results of multiple models using ensemble learning techniques to improve the model's generalization ability and robustness. Establish a model performance evaluation mechanism to regularly assess the model's prediction accuracy and applicability.

[0137] Feedback control implementation: A multi-layered feedback control structure is designed, including three levels: real-time feedback, short-term feedback, and long-term feedback. Real-time feedback primarily addresses deviations at the execution level, with response times in the order of seconds; short-term feedback addresses adjustments at the strategy level, with response times in the order of minutes; and long-term feedback addresses model and algorithm optimization, with response times in the order of hours or days. Different control algorithms and parameter settings are used for different levels of feedback control to ensure system stability and responsiveness.

[0138] The output includes a system performance evaluation report, deviation analysis results, strategy adjustment suggestions and parameter optimization schemes, as well as an updated system model and knowledge base.

[0139] Furthermore, an adaptive adjustment method based on deep reinforcement learning can be employed. This method trains an intelligent adjustment agent capable of autonomously determining adjustment strategies and parameters based on the system state. Through continuous experimentation and learning, this approach gradually improves the intelligence level of adjustment decisions, reducing the need for manual intervention.

[0140] A dynamic scheduling system for irradiation transmission line trays is provided for executing the aforementioned dynamic scheduling method for irradiation transmission line trays, such as... Figure 2 As shown, it includes:

[0141] The data acquisition module is used to install an integrated sensor module on each tray, transmit data to the data acquisition server via wireless communication, filter and standardize the raw sensor data, and output a tray status parameter matrix.

[0142] The product identification module is used to determine the timing of a product's arrival at the transmission line entrance based on the tray status parameter matrix, collect product information and identify product characteristics based on the timing of the product's arrival at the transmission line entrance, and establish a mapping relationship between product characteristics and irradiation requirements using intelligent identification technology to obtain a product requirement parameter vector.

[0143] The status assessment module is used to predict equipment operating status and production load based on product demand parameter vectors and using long short-term memory networks, and output system status assessment results.

[0144] The optimization calculation module is used to construct a multi-objective optimization model based on the system state evaluation results using an improved non-dominated sorting genetic algorithm, and output a Pareto optimal solution set.

[0145] The strategy generation module is used to generate scheduling strategies based on Pareto optimal solution sets, using fuzzy logic decision-making and deep reinforcement learning, and output tray operation schemes.

[0146] The execution control module is used to decompose the pallet operation plan into specific execution instructions, use predictive control algorithms to achieve trajectory tracking, output real-time control status, and establish a performance monitoring and parameter optimization mechanism based on the control status to achieve dynamic adjustment.

[0147] In one embodiment of the present invention, a specific example is provided:

[0148] Based on the actual application in a certain electronic component irradiation production line, the data collection of tray status and product characteristic identification are shown in Table 1 and Table 2, respectively:

[0149] Table 1: Example of pallet status data collection;

[0150]

[0151] Table 1 illustrates that the pallet status data acquisition system can acquire five dimensions of pallet status information in real time. Position coordinates are obtained using a laser positioning sensor based on triangulation principles, with an accuracy of 2 cm; movement speed is obtained through integration calculation using a triaxial accelerometer, with a sampling frequency of 100 Hz; load is monitored in real time using a strain gauge weighing sensor, with an accuracy of 0.1 kg; temperature is monitored using a thermocouple temperature sensor, with an accuracy of 0.5 °C; and cumulative radiation dose is monitored in real time using a silicon detector, with an accuracy of 1%. All data undergoes filtering, time synchronization alignment, and standardization to ensure data quality and the accuracy of the algorithm processing.

[0152] Table 2: Example of product characteristic identification data collection;

[0153]

[0154] Table 2 shows the data acquisition results of the product characteristic identification system. An industrial camera captures product images at a resolution of 2048×1536 pixels, and geometric features are extracted using Canny edge detection and Hough transform. An RFID reader reads product tag information to obtain basic information such as product code and batch number. A laser scanner performs 3D contour scanning to obtain precise volume and shape data. The system establishes a mapping database between product characteristics and irradiation requirements, and uses a fuzzy logic reasoning algorithm to calculate the optimal transmission parameters based on product geometric features and material properties. All product characteristic data undergoes hierarchical standardization processing to ensure the accuracy and comparability of data for different types of products.

[0155] like Figure 3 As shown, the dynamic scheduling system demonstrates significant improvements over the traditional scheduling system in key performance indicators. Green bars represent the performance of the dynamic scheduling system, while blue bars represent the performance baseline of the traditional scheduling system. In terms of processing time, the dynamic scheduling system reduces the average processing time from 4.2 hours to 3.1 hours, resulting in a significant efficiency improvement. Regarding equipment utilization, it increases from 72% to 89%, achieving more efficient resource utilization. Pallet idle rate decreases from 18% to 8%, reducing resource waste. Emergency response time is reduced from 35 minutes to 12 minutes, improving system response speed. System throughput increases from 28 batches / hour to 41 batches / hour, significantly enhancing production capacity. Unit energy consumption decreases from 12.5 kWh / batch to 9.2 kWh / batch, demonstrating significant energy savings.

[0156] like Figure 4The figure illustrates the real-time optimization effect of the dynamic scheduling system during 24-hour continuous operation. The blue line represents the trend of equipment utilization, the orange line represents the change in system throughput, and the green line represents the change in energy efficiency. The figure shows that equipment utilization remains at a high level of 85%–94% throughout the entire operating cycle, reflecting the system's full utilization of equipment resources. System throughput exhibits a variation pattern matching the production load, reaching a peak level of 92% during peak periods. Energy efficiency maintains a stable upward trend throughout the day, gradually increasing from 88% to 95%, reflecting the system's continuous optimization and learning capabilities. The coordinated changes of the three indicators demonstrate the effectiveness of the multi-objective optimization algorithm.

[0157] like Figure 5 The figure shows the distribution of Pareto optimal solutions generated by the multi-objective optimization algorithm in a two-dimensional space. The horizontal axis represents completion time, the vertical axis represents energy consumption level, and the size of the points indicates the level of equipment utilization. The blue scatter points represent the set of optimal solutions on the Pareto front, forming a clear front curve, reflecting the trade-off between completion time and energy consumption. The large red dots indicate the scheduling scheme currently selected by the system, located near the Pareto front, indicating that the scheme achieves a good balance among the three objectives. The figure shows that when the completion time requirement is tight, the system energy consumption is relatively high; when a longer completion time is allowed, the system can achieve a lower energy consumption level. This distribution characteristic provides a scientific basis for decision-makers to select appropriate scheduling strategies based on different production needs.

[0158] Through the above practical application verification, the Netgear dynamic scheduling method has achieved significant results in improving production efficiency, optimizing resource allocation, and reducing energy consumption, providing an effective technical solution for the intelligent upgrading of irradiation transmission lines.

[0159] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for dynamic scheduling of irradiation transmission line trays, characterized in that, Includes the following steps: An integrated sensor module is installed on each tray, and the data is transmitted to the data acquisition server via wireless communication. The raw sensor data is filtered and standardized, and the tray status parameter matrix is ​​output. The timing of a product's arrival at the transmission line entrance is determined based on the pallet status parameter matrix. Product information is collected and product characteristics are identified based on the timing of the product's arrival at the transmission line entrance. Intelligent identification technology is used to establish a mapping relationship between product characteristics and irradiation requirements, thereby obtaining a product requirement parameter vector. Based on the product demand parameter vector, a long short-term memory network is used to predict the equipment operating status and production load, and output the system status assessment results. Based on the system state assessment results, an improved non-dominated sorting genetic algorithm is used to construct a multi-objective optimization model, outputting a Pareto optimal solution set. The objective functions are designed as follows: the first objective function is to minimize the total completion time, which is the total time from the entry of all products into the system to the completion of irradiation treatment. The calculation method is to accumulate the difference between the completion time and the start time of each product. The second objective function is to minimize the total energy consumption, which is the total power consumption of all devices in the entire scheduling cycle. The calculation method is to accumulate the product of the power of each device and its running time. The third objective function is to maximize the average utilization rate of the equipment, which is expressed as the average utilization rate of all equipment; The calculation method is to average the ratio of the actual working time of each device to the total available time; Constraints are set as follows: Irradiation quality constraints ensure that the irradiation dose received by each product is within the specified range, avoiding excessive or insufficient irradiation; equipment capacity constraints limit the maximum load and processing capacity of each piece of equipment; safety constraints include minimum safe distance between pallets, maximum moving speed limits, etc.; time window constraints ensure that urgent orders are processed within a specified time; resource constraints limit the number of products processed simultaneously and pallet allocation. Based on the Pareto optimal solution set, a scheduling strategy is generated using fuzzy logic decision-making and deep reinforcement learning to output a pallet operation plan; a fuzzy logic-based decision-making method is used to select the most suitable scheduling plan for the current production needs from the Pareto solution set; a reinforcement learning model based on a deep Q-network is established to optimize the scheduling strategy through continuous interaction with the environment. The pallet operation plan is decomposed into specific execution instructions, and a predictive control algorithm is used to achieve trajectory tracking, output real-time control status, and establish a performance monitoring and parameter optimization mechanism based on the control status to achieve dynamic adjustment.

2. The dynamic scheduling method for irradiation transmission line trays according to claim 1, characterized in that, The sensor module includes: Triaxial accelerometer, laser positioning sensor, electronic scale module, thermocouple temperature sensor and radiation dose detector; A three-axis accelerometer detects the motion and vibration of the tray, with a sampling frequency set to 100 Hz. The real-time velocity and displacement of the tray are calculated through integration. The laser positioning sensor uses the triangulation principle to determine the precise position of the tray on the transmission line with a measurement accuracy of 2 cm; the electronic scale module uses a strain gauge load cell to monitor changes in the tray load in real time, with a measurement range of 0-20 kg and an accuracy of 0.1 kg.

3. The dynamic scheduling method for irradiation transmission line trays according to claim 1, characterized in that, The process of collecting product information and identifying product characteristics includes: Industrial camera and RFID reader, wherein the industrial camera acquires product images with an image resolution of 2048×1536 pixels, uses an edge detection algorithm to identify the product outline, and uses the Kenny operator for edge extraction; The product's geometric parameters are calculated by detecting straight and circular features using Hough transform; an RFID reader then reads the product tag information.

4. The dynamic scheduling method for irradiation transmission line trays according to claim 1, characterized in that, The Long Short-Term Memory (LSTM) network prediction includes: The assessment of equipment operating status involves installing vibration sensors, current sensors, and temperature sensors to analyze the vibration spectrum characteristics of the equipment to identify mechanical fault signs, monitoring motor current waveforms to identify electrical problems, and using temperature trend analysis to determine the thermodynamic state of the equipment.

5. The dynamic scheduling method for irradiation transmission line trays according to claim 1, characterized in that, The improved non-dominated sorting genetic algorithm includes: The process includes an initialization phase, a selection operation, a crossover operation, a mutation operation, and an environment selection phase. The selection operation uses a tournament selection method, the crossover operation uses a sequential crossover method, and the mutation operation uses a combination of insertion mutation and exchange mutation.

6. The dynamic scheduling method for irradiation transmission line trays according to claim 1, characterized in that, The fuzzy logic decision-making includes: Establish a fuzzy decision rule base, set corresponding priority rules for different production modes, and the fuzzy inference engine automatically matches appropriate decision rules based on the current system status and production priority.

7. The dynamic scheduling method for irradiation transmission line trays according to claim 1, characterized in that, The predictive control includes: The distributed control system adopts a master-slave control architecture. It issues scheduling commands via industrial Ethernet, and the controller uses a combination of proportional-integral-derivative control and feedforward control to achieve trajectory tracking.

8. The dynamic scheduling method for irradiation transmission line trays according to claim 1, characterized in that, The performance monitoring and parameter optimization include: The multi-objective optimization algorithm parameters and control parameters are adaptively optimized by using an online learning algorithm to train the parameter optimization model and dynamically adjusting the learning step size through an adaptive learning rate adjustment mechanism.

9. The dynamic scheduling method for irradiation transmission line trays according to claim 1, characterized in that, The dynamic adjustment also includes: Establish a multi-level security protection system, including hardware security protection devices and software security protection functions, which automatically trigger protection actions when a security risk is detected.

10. A dynamic scheduling system for irradiation transmission line trays, characterized in that, A method for executing a dynamic scheduling method for irradiation transmission line trays according to any one of claims 1-9 includes: The data acquisition module is used to install an integrated sensor module on each tray, transmit data to the data acquisition server via wireless communication, filter and standardize the raw sensor data, and output a tray status parameter matrix. The product identification module is used to determine the timing of a product's arrival at the transmission line entrance based on the tray status parameter matrix, collect product information and identify product characteristics based on the timing of the product's arrival at the transmission line entrance, and establish a mapping relationship between product characteristics and irradiation requirements using intelligent identification technology to obtain a product requirement parameter vector. The status assessment module is used to predict equipment operating status and production load based on product demand parameter vectors and using long short-term memory networks, and output system status assessment results. The optimization calculation module is used to construct a multi-objective optimization model based on the system state evaluation results using an improved non-dominated sorting genetic algorithm, and output a Pareto optimal solution set. The strategy generation module is used to generate scheduling strategies based on Pareto optimal solution sets, using fuzzy logic decision-making and deep reinforcement learning, and output tray operation schemes. The execution control module is used to decompose the pallet operation plan into specific execution instructions, use predictive control algorithms to achieve trajectory tracking, output real-time control status, and establish a performance monitoring and parameter optimization mechanism based on the control status to achieve dynamic adjustment.

Images